University of Southern California Dissertations and Theses

Mathematical modeling in bacterial communication and optogenetic systems

(USC Thesis Other)

Mathematical modeling in bacterial communication and optogenetic systems

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Mathematical Modeling in Bacterial Communication and Optogenetic
Systems
By
Ghazaleh Ostovar
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PHYSICS)

December 2024

ii
Acknowledgements
I am grateful to my Ph.D. advisor, Dr. James Boedicker, for his guidance and enduring support
throughout my research journey. I am also grateful to my Ph.D. committee members Moh ElNaggar, Rosa Di Felice, Aiichiro Nakano, and Stephan Wolfgang Haas for their constructive
feedback and encouragement. A special thanks goes to my lab mates for making the lab a
welcoming and collaborative space.
I am also deeply grateful to my family and friends, especially my mother Ezzat, for their love
and support during this journey.

iii
Table of Contents
Acknowledgements......................................................................................................................... ii
List of Tables................................................................................................................................. vii
List of Figures.............................................................................................................................. viii
Abstract........................................................................................................................................... x
Chapter 1: Introduction ................................................................................................................... 1
1.1.1: Decision making in fluctuating environments.............................................................. 1
1.1.2: Optogenetic Control of Gene Expression and Electron Transfer in Shewanella
oneidensis................................................................................................................................ 3
1.3: Brief Overview of the Thesis............................................................................................... 5
1.4: Summary of Chapters ......................................................................................................... 8
Chapter 2: Computation in bacterial communities ......................................................................... 9
2.1: Abstract ................................................................................................................................ 9
2.2: The computational potential of microbial communities...................................................... 9
2.3: Computation of the spatial environment............................................................................11
2.3.1: Computing the efficiency of signal exchange..............................................................11
2.3.2: Pattern formation in response to spatial structure....................................................... 13
2.4: Computation of community composition .......................................................................... 15
2.4.1. Signal crosstalk and interference ................................................................................ 15
2.4.2: Microbes as a neural network ..................................................................................... 17
2.4.3: Integration of stimuli in multilayer networks............................................................. 20
2.5: Optimization of signaling .................................................................................................. 21
2.5.1: Energetic costs of communication .............................................................................. 22
2.5.2: Measuring information flow ....................................................................................... 23
2.5.3: Single-cell heterogeneity in communication .............................................................. 25

iv
2.5.4: Spatial self-organization and percolation improve the efficiency of cellular
communication...................................................................................................................... 28
2.6: Future perspectives ............................................................................................................ 30
Chapter 3: Phenotypic memory in quorum sensing ...................................................................... 32
3.1: Abstract .............................................................................................................................. 32
3.2: Author summary ................................................................................................................ 33
3.3: Introduction........................................................................................................................ 33
3.4: Methods ............................................................................................................................. 35
3.5: Results................................................................................................................................ 39
3.5.1: A model for quorum sensing activation and deactivation........................................... 39
3.5.2: Carry-over of biomolecules during the deactivation of quorum sensing.................... 41
3.5.3: Phenotypic memory effect in quorum sensing............................................................ 42
3.5.4: Cellular parameters controlling the strength of phenotypic memory in quorum
sensing................................................................................................................................... 45
3.5.5: Reactivation of quorum sensing in cells with quorum sensing memory ........................ 51
3.6: Discussion.......................................................................................................................... 55
3.7: Conclusion ......................................................................................................................... 61
3.8: Supplementary information ............................................................................................... 61
Chapter 4: Red-Light-Induced Genetic System for Control of Extracellular Electron Transfer .. 68
4.1: Abstract .............................................................................................................................. 68
4.2: Introduction........................................................................................................................ 68
4.3: Results................................................................................................................................ 70
4.3.1: Importing the iLight optogenetic system to S. oneidensis.......................................... 70
4.3.2: Adjusting the expression level of the iLight repressor ............................................... 72
4.3.3: Creating an inverted iLight optogenetic system, for light-activated gene regulation. 73

v
4.3.4: The extracellular electron transfer activity of S. oneidensis can be regulated with
red light ................................................................................................................................. 76
4.4: Discussion.......................................................................................................................... 79
4.5: Methods ............................................................................................................................. 80
4.5.1: Bacterial strains and plasmids..................................................................................... 80
4.5.2: Growth conditions....................................................................................................... 81
4.5.3: Fluorescence measurements and microscopy ............................................................. 81
4.5.4: Iron reduction measurements...................................................................................... 82
4.5.5: Methyl orange (MO) reduction measurements............................................................... 82
4.5.6: Transparent-bottom bioreactor construction............................................................... 83
4.5.7: Cell culturing and biofilm formation within bioreactor.............................................. 83
4.5.8: Electrochemical activity measurements...................................................................... 84
4.5.9: Statistical analysis....................................................................................................... 84
4.6: Supporting Information...................................................................................................... 85
4.6.1: Thermodynamic model ............................................................................................... 93
Chapter 5: Concluding Remarks................................................................................................. 103
5.1: impact of my work........................................................................................................... 103
5.1.1: Introducing and quantifying the concept of phenotypic memory in QS................... 103
5.1.2: Potential for future experimental verification of phenotypic memory in QS ........... 103
5.1.3: Enhancing therapeutic approaches through QS phenotypic memory ....................... 104
5.1.4: Influence of quorum sensing phenotypic memory on ecological dynamics............. 105
5.1.4: Role of QS Phenotypic memory in synthetic biology .............................................. 105
5.1.5: Transfer of iLight optogenetic system in S. oneidensis enhances precision in
microbial bioelectronics...................................................................................................... 106
5.1.6: Transfer of iLight optogenetic system in S. oneidensis advances synthetic biology
toolkits................................................................................................................................. 106
5.1.7: Enhancing gene regulation in synthetic Biology: the role of the iLight
thermodynamic model ....................................................................................................... 106
5.1.7: Environmental and industrial applications ............................................................... 107

vi
5.2: Closing remarks............................................................................................................... 107
References................................................................................................................................... 108

vii
List of Tables
Table 3.1: List of parameters used in equations 3.1-3.8. .............................................................. 38
Table S3.1: QS biomolecule concentrations in ON and OFF states............................................. 62
Table S4.1: Plasmids used in the construction of iLight optogenetic system............................... 86
Table S4.2: Strains used in iLight optogenetic study. .................................................................. 87
Table S4.3: Primers used in iLight optogenetic study. ................................................................. 89
Table S4.4: Recipe of S. oneidensis MR-1 minimal medium....................................................... 90

viii
List of Figures
Figure 2.1: Bacteria use quorum sensing to gather information from the environment. ...............12
Figure 2.2: Signal exchange for spatial computations. ..................................................................14
Figure 2.3: Quorum sensing to gather information about community composition......................18
Figure 2.4: Energetic aspects of gathering information.................................................................23
Figure 2.5: The efficiency of collective behaviors in bacteria.......................................................27
Figure 3.1: Dynamics of quorum sensing activation and deactivation. .........................................40
Figure 3.2: Phenotypic memory impacts QS reactivation near the critical cell density. ...............44
Figure 3.3: Overall memory term dependency on time after dilution, Fold Change, and
doubling time. ................................................................................................................................51
Figure 3.4:The effect of Fold Change and initial cell density on the emergence of the
memory zone.................................................................................................................................55
Figure S3.1:Quorum sensing activity state vs. final cell density. ..................................................62
Figure S3.2: Over-dilution diminishes the memory effect in quorum sensing..............................63
Figure S3.3: Carry-over effects of individual QS biomolecules on memory zone emergence. ....64
Figure S3.4: LuxI memory term ξ vs. time after dilution, Fold Change, and doubling time. .......65
Figure S3.5: Overall memory term Θ in the absence of rapid degradation of LuxR vs. time
after dilution, Fold Change, and doubling time. ...........................................................................65
Figure S3.6: The impact of doubling time on memory zone width. ..............................................66
Figure S3.7: The Effect of the autoinducer synthesis rate (b) constant, and the threshold
required for activation (𝐴𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 ) on memory zone width.......................................................67
Figure 4.1: Characterization of the iLight genetic circuit in S. oneidensis and E. coli. ................71
Figure 4.2: Adapting the iLight optogenetic system for S. oneidensis..........................................73
Figure 4.3: Characterization of the non-inverted iLight genetic circuit iLight-J23102 and the
inverted iLight genetic circuit iLight-J23102(PhIF repressor) in S. oneidensis............................75
Figure 4.4: Using iLight-J23102(PhIF repressor) genetic circuit to control the outer
membrane cytochrome MtrC expression for light-induced extracellular electron transfer
(EET) activity in S. oneidensis. ....................................................................................................78

ix
Figure S4.1:Characterization of the iLight optogenetic system in S. oneidensis and E. coli
without containing the heme oxygenase gene ho1 in the plasmid.................................................92
Figure S4.2: The thermodynamic model of the iLight optogenetic system...................................95
Figure S4.3: Expression of the iLight reporter in S. oneidensis strains containing different
promoters of iLight repressor under red light and dark conditions, respectively. .........................96
Figure S4.4:Characterization of the iLight genetic circuits with different promoters of iLight
repressor in E. coli. ........................................................................................................................97
Figure S4.5:Characterization of the inverted iLight genetic circuits in S. oneidensis.
Expression of the iLight reporter in S. oneidensis strains containing the inverted iLight
genetic circuit................................................................................................................................98
Figure S4 6: Introducing of two optogenetic systems iLight and pDawn into Shewanella...........99
Figure S4.7: Expression of outer membrane cytochromes MtrC through inverted iLight
genetic circuit in S. oneidensis mutant ΔmtrCΔomcA. ..............................................................101
Figure S4.8: Expression of periplasm cytochrome STC through inverted iLight genetic
circuit in S. oneidensis mutant ΔstcΔfccA..................................................................................102

x
Abstract
This thesis explores two critical dimensions of bacterial behavior: the existence of
phenotypic memory in quorum sensing (QS) and the implementation of an optogenetic control
system in Shewanella oneidensis. The study introduces a metric to quantify phenotypic memory
in QS, defined by the reduced critical cell density required for QS reactivation in cells previously
in QS active state. This reflects the influence of historical cell states on current bacterial responses.
Mathematical models and simulations not only demonstrate that bacteria previously in QS-ON
states can activate at lower densities but also identify several key cellular parameters that
significantly influence this memory effect.
Additionally, this work details the successful adaptation of an optogenetic system in S.
oneidensis to control gene expression critical for electron transfer processes, pivotal for microbial
bioelectronics. My specific contribution was developing a thermodynamic model that predicts
optimal gene expression levels, providing a deeper understanding of the light-controlled regulatory
mechanisms in this system.
These investigations not only advance our comprehension of microbial communication and
control but also faciliates more effective management of bacterial populations and the development
of therapeutic strategies in medical and environmental applications. The findings emphasize the
potential of leveraging historical behavioral patterns and precise genetic control in synthetic
biology, enhancing both theoretical knowledge and practical applications.

1
Chapter 1: Introduction
1.1.1: Decision making in fluctuating environments
Bacteria continuously exchange information with their environment, enabling them to
perform complex tasks by sensing and responding to changes around them [1]. Although these
single-cell bugs lack a traditional nervous system, they are equipped with biological molecules
that can change structures when they interact with other molecules, and thereby encode
information. This adaptability allows bacteria to process external cues and make decisions that
affect not only themselves but also the structure and dynamics of their communities [2-4]. These
sophisticated interactions act like a type of "computation" among various species or strains,
showcasing microbial intelligence that boosts their metabolic efficiency, robustness, and overall
fitness.
Quorum sensing (QS) is a well-known example of how bacteria engage in collective
decision-making. QS involves the emission and detection of chemical signals known as
autoinducers [5-7]. As bacterial populations grow, the concentration of signaling molecules also
increases. Upon reaching a critical threshold concentration, these signals induce significant
changes in expression of target genes, transitioning the bacterial behavior from individualistic to
communal.
Additionally, QS systems often include positive feedback loops [8, 9] that amplify
autoinducer production once the threshold is reached. This results in a robust switch [8, 10] from
low to high expression states of QS target genes—referred to as the QS-OFF and ON states [11,
12]. This switch facilitates processes that are most effective at high cell densities, such as biofilm
formation, virulence factor production, and horizontal gene transfer, enhancing the bacteria's
adaptability and survival in competitive environments [13].
Since QS activation occurs upon reaching a critical cell density [14], it has been known as
a proxy for cell density. Unlike controlled laboratory conditions, natural QS systems frequently
experience fluctuations in both cell density and signal concentration. This raises questions about

2
whether the QS response is solely driven by current conditions or if it is also influenced by past
exposures, denoting a "memory" effect.
Past environmental conditions have been shown to impact bacterial future responses. For
instance, the lac operon in E. coli demonstrated a faster adaptation to lactose following previous
exposure, suggesting that residual LacZ proteins from earlier exposures can reduce growth lag
phases. This historical dependency, evident in various bacterial functions, supports the notion of
"phenotypic memory," [15, 16] which results from the persistence of cellular components rather
than genetic inheritance.
Evidence of presence of memory in QS has been observed previously [17]. Bacteria
previously in a high-density (QS-ON) state have been observed to reactivate quorum sensing more
rapidly upon re-exposure to similar conditions [17]. In this study, the memory effect was linked to
the conformational state of the QS receptor protein LasR, which was influenced by prior exposure
to the AHL signal. Specifically, the persistence of properly folded LasR proteins enabled faster
reactivation of QS upon subsequent re-exposure to the signal. This phenomenon, driven by
molecular carry-over rather than genetic inheritance, exemplifies phenotypic memory in QS.
Building on these experimental findings [17], the concept of phenotypic memory in QS
can be further explored by analyzing variations in the critical cell density required for activation.
This approach provides a quantitative measure to assess how prior environmental conditions may
influence future responses. Presence of memory will determine whether bacteria previously in a
QS 'ON' state may activate at lower densities than those initially 'OFF'. This behavior could stem
from the retention of proteins in charge of synthesis and detection of signaling molecules. By
employing mathematical models and analytical techniques, our study attempts to quantify how the
carry-over of these biomolecules from QS-ON state influences future response.
The concept of memory in QS systems introduces a complex layer to our understanding of
bacterial adaptation, enabling bacteria to respond to environmental fluctuations more effectively.
This memory allows bacteria to activate QS at lower cell densities, offering crucial survival
advantages under rapid changes such as antibiotic exposure or signal disruption [11, 18, 19].

3
Moreover, it conserves energy and resources by avoiding unnecessary protein synthesis [20],
directly enhancing bacterial fitness.
However, while this adaptive memory supports swift responses to environmental shifts, it
can also present risks. If bacteria prematurely respond based on outdated cues, they may exhibit
decreased fitness, especially when environmental conditions change too quickly [21]. It has been
shown that the costs and benefits of maintaining phenotypic memory are significantly influenced
by the frequency and magnitude of environmental fluctuations [22-26]. Understanding and
effectively leveraging this memory is crucial for both advancing ecological research and
developing practical applications aimed at controlling bacterial populations.
Understanding the role of memory in quorum sensing (QS) is essential for developing
strategies to manipulate bacterial behaviors in industrial and medical contexts. This knowledge
provides insights into how bacteria adapt to challenges like antibiotics or signal degradation,
influencing cell density and signal concentration. By targeting QS pathways, we can disrupt
biofilm formation or reduce virulence in pathogenic bacteria. Moreover, phenotypic memory
influences the timing and efficacy of treatments, such as QS inhibitors, by potentially altering the
intensity and schedule of virulence factor production. Understanding QS memory is essential in
these interventions, as even subtle variations in QS responses can greatly influence outcomes in
competitive settings
1.1.2: Optogenetic Control of Gene Expression and Electron Transfer in Shewanella
oneidensis
Optogenetics is a technique that uses light to control the activity of cells in living organisms
[27]. This approach has facilitated significant advancements in managing various bacterial
activities, including chemical synthesis [28-30], biofilm development [31, 32], and infection
processes [30].
S. oneidensis MR-1 is a bacterium notable for its ability to transfer electrons outside of its
cells, crucial for constructing conductive biofilms that support long-distance electron transport [33,
34]. Researchers have explored various methods to control this electron transfer capability using

4
genetic tools [35]. In this study, this system has been adapted for use in S. oneidensis MR-1 by
engineering its genetic components. This adaptation allows the use of red light to control the
expression of genes involved in electron transfer, providing precise control over gene expression
and electron transfer activity in the S. oneidensis.
Optogenetics offers powerful spatiotemporal control of gene expression. Several
optogenetic regulatory circuits [36-41] have been developed for bacterial systems and have been
successfully ported into new host strains[31, 40]. In this study, a red light-inducible transcription
factor was developed and adapted for S. oneidensis. This regulatory circuit, based on the iLight
optogenetic system, controls gene expression using red light. Promoter engineering and a
thermodynamic model were employed to adapt this system for achieving differential gene
expression under light and dark conditions within a S. oneidensis host strain. The iLight
optogenetic system was modified by introduction of an inverted genetic circuit, which activates
gene expression under red light. This inverted iLight genetic circuit was then used to control
extracellular electron transfer (EET) within S. oneidensis. Additionally, the simultaneous use of
both red and blue light-induced optogenetic circuits was demonstrated. This work expands the
synthetic biology toolbox for S. oneidensis, which could facilitate future advancements in
applications involving electrogenic bacteria.
The thermodynamic model developed in this study serves as a powerful tool for
understanding the limits and optimal conditions of gene expression controlled by the iLight
optogenetic system. By describing the equilibrium dynamics of transcription factor binding within
the promoter region, this model elucidates the probabilities of RNA polymerase binding under
different light conditions. Such insights are crucial for determining the optimal expression levels
at which the iLight switch operates most efficiently. Moreover, the model's ability to predict
differential gene expression in response to light versus dark conditions offers a valuable framework
for optimizing and extending optogenetic control mechanisms to other microbial hosts, enhancing
the precision and applicability of synthetic biology approaches.

5
1.3: Brief Overview of the Thesis
This thesis investigates two significant areas in bacterial regulatory mechanisms:
phenotypic memory in quorum sensing (QS) systems and the adaptation of an optogenetic system
in the bacterium S. oneidensis, including the development of a thermodynamic model to
understand how expression levels of the components of this genetic circuit impact optogenetic
control. Additionally, it includes a review of the literature on bacterial communication networks,
focusing on the challenges and limitations bacteria encounter in decision-making processes.
Chapter 2 offers an in-depth review of the literature on the dynamic processes by which
bacteria exchange information to coordinate community structure and activity. This review
explores the mechanisms by which bacteria use molecular sensors to gather and interpret
environmental signals, effectively “computing” reponses that benefit both individual strains and
their communities. The discussion extends to the challenges and strategies bacteria deploy to
sustain robust and efficient communication within microbial communities, particularly in the face
of biological, chemical, and physical fluctuations. Emphasis is placed on the incorporation of
mathematical frameworks, including information theory, to better understand and quantify these
bacterial communication processes. These models enhance our predictive capabilities regarding
the outcomes and behaviors associated with these complex interactions.
Chapter 3 presents a theoretical exploration of phenotypic memory in quorum sensing
systems, demonstrating that bacterial cells can retain a 'memory' of prior QS activation.
Specifically, cells previously in a QS-ON state can reactivate the quorum sensing response at lower
cell densities compared to those initially in a QS-OFF state. This memory effect is attributed to
the persistence of synthase proteins (LuxI) and receptors (LuxR) even after the initial stimuli are
removed. The chapter identifies conditions under which this memory effect becomes more
pronounced and delineates the key cellular parameters influencing its strength, exploring the
contexts within which this memory plays a significant role.
Analytical modeling and numerical simulations using ordinary differential equations
(ODEs) were employed to explore the dynamics of QS activation. The ODEs, based on the LuxRLuxI regulatory circuit in Vibrio fischeri [42], modeled changes in gene concentration, signaling

6
molecules, and cell density over time. The simulations uncovered a switch-like transition between
two steady states: low (OFF) and high (ON) cell densities and identified the critical cell density at
which this transition occurs. The difference in critical cell density between initially ON and OFF
states served as a key metric for quantifying phenotypic memory, revealing how previous QS
activation affects current bacterial behavior. By altering the initial conditions to OFF and ON state
values and monitoring gene expression in response to abrupt changes in cell density and
extracellular signal concentrations, the findings revealed that cells initially in the ON state could
reactivate QS at a lower critical cell density compared to those initially in the OFF state,
particularly when the initial cell density following perturbation was sufficiently high. This
observation underscores a near-threshold effect.
Further investigation into the molecular mechanisms behind this memory effect involved
manipulating the concentrations of key biomolecules. The analysis highlighted the role of elevated
concentrations of LuxI and dimeric LuxR in the emergence of memory. These elevated levels
accelerate signal synthesis and receptor binding, facilitating faster reactivation of quorum sensing
upon cell growth.
To better understand the parameters influencing the strength of the memory effect, a
simplified analytical model was developed, incorporating a memory term into the critical cell
density calculation for initially ON cells. By deriving an explicit equation for the critical cell
density required for activation in both initially ON and OFF states, the study reveals that
parameters such as higher Fold Change in gene expression, higher basal autoinducer synthesis,
lower autoinducer thresholds, and slower cell growth rates all contribute to strengthening the
memory effect. The chapter concludes by discussing the broader implications of phenotypic
memory in QS for ecological systems, industry, and medicine, and suggests potential applications
in disrupting pathogenic bacterial behaviors. These findings lay the groundwork for experimental
validation, offering new insights into how bacteria adapt and survive in dynamic environments.
Chapter 4 focuses on the development and adaptation of a red-light inducible transcription
factor for Shewanella oneidensis MR-1, a bacterium known for its ability to transfer electrons
outside its cells, thereby facilitating the creation of conductive biofilms. This genetic circuit,
designed based on the iLight optogenetic system, modulates gene expression in response to red

7
light. Advanced promoter engineering and a thermodynamic model were utilized to enable
differential gene expression under varying light conditions. Additionally, the system was modified
by incorporating a repressor that inverts the genetic circuit, thus allowing gene activation under
red light. This enhancement significantly improves the control over gene expression in S.
oneidensis, facilitating precise manipulations of its bioelectronic properties.
My contribution to this work included developing the thermodynamic model and
visualizing the expression of iLight reporter gene under varying repressor concentrations. Initially,
the iLight system was developed in an E. coli host, but after transferring it to Shewanella, the
switch failed to show significant differences in iLight reporter gene expression under light and
dark conditions. This issue was resolved by increasing expression levels of the repressor genes.
However, at high expression levels, the switch once more proved ineffective to show distinct
results for dark and light conditions. The thermodynamic model was instrumental in understanding
how the switch achieved optimal functionality at intermediate repressor expression levels in the
new host.
The thermodynamic model in this study describes the equilibrium dynamics of
transcription factor binding within the promoter region, influencing the probability of RNA
polymerase (RNAP) binding under different light conditions. This probability is assumed to be
proportional to iLight reporter gene expression levels. The model assumes that contrary to the
initial belief that only tetrameric forms of repressors can bind to the specific site, dimeric forms of
repressors can also bind, albeit loosely. To account for the stronger binding affinity, the model
assigns a more negative binding energy to the tetramer. It also assumes that light exposure converts
all dimers into tetramers, predominantly resulting in dimeric forms in the dark and tetrameric forms
in light. The ratio of Fold changes in expression of iLight reported gene in dark vs light were
derived, showing the maximum Fold change ratio at intermediate repressor concentrations,
corroborating experimental observations. This model was specifically developed to understand
and visualize the limits and optimal conditions of gene expression controlled by the iLight
optogenetic system, thereby defining the bounds of the parameter space.
The findings from both projects have significant implications for synthetic biology,
industrial applications, and medical treatments. These advancements could inform strategies for

8
managing bacterial infections and developing new therapeutic approaches. The details of these
studies and their comprehensive results will be described throughout the manuscript, with a full
discussion and exploration of potential future research directions presented in Chapter 5.
1.4: Summary of Chapters
Chapter 2 reviews how bacterial communities "compute" responses to environmental cues,
examining the types of computations bacteria perform and the limits of these computations in
terms of energy efficiency and response accuracy.
Chapter 3 presents a focused mathematical study on phenotypic memory within quorum
sensing, exploring how historical signal exposure shapes current bacterial responses. This chapter
provides insights supported by simulation data and analytical modeling.
Chapter 4 introduces transfer of iLight optogenetic system in Shewanella oneidensis and
introduces a thermodynamics model, predicting the optimal gene expression levels under which
the switch performs most effectively.
Chapter 5 explores the impact of my work and suggests future research directions.

9
Chapter 2: Computation in bacterial communities
This work appears as published in Physical Biology 17, no. 6 (2020): 061002.
2.1: Abstract
Bacteria across many scales are involved in a dynamic process of information exchange to
coordinate activity and community structure within large and diverse populations. The molecular
components bacteria use to communicate have been discovered and characterized, and recent
efforts have begun to understand the potential for bacterial signal exchange to gather information
from the environment and coordinate collective behaviors. Such computations made by bacteria
to coordinate the action of a population of cells in response to information gathered by a multitude
of inputs is a form of collective intelligence. These computations must be robust to fluctuations in
both biological, chemical, and physical parameters as well as to operate with energetic efficiency.
Given these constraints, what are the limits of computation by bacterial populations and what
strategies have evolved to ensure bacterial communities efficiently work together? Here the current
understanding of information exchange and collective decision making that occur in microbial
populations will be reviewed. Looking towards the future, we consider how a deeper understanding
of bacterial computation will inform future direction in microbiology, biotechnology, and
biophysics.
2.2: The computational potential of microbial communities
Bacteria interact with their environment in astonishing ways and to perform complex,
coordinated tasks. We now know that bacterial cells have developed mechanisms to monitor and
respond to changes in physical and chemical conditions as well as communicate with each other
to work as multicellular collectives. Coordination of such behavior is an example of collective or
swarm intelligence, as groups of cells gather information to collectively compute a response.
Recently the ability to gather and respond to information is gaining attention as the
essential feature of living systems. Biological molecules change structure as they interact with
other molecules and these changes transmit information. Bacteria receive and process information

10
from their extracellular environment in order to compute an adequate response or “make a
decision” [1]. What does it mean for bacteria to compute? By “compute” we mean a process by
which groups or individuals gather and use information to change the composition, spatial
structure, or activity of a community of cells. A population of multiple species or strains of cells
and their interactions through the exchange and detection of small molecules constitutes a
“network”. The coordinated behaviors that emerge from bacterial networks can be viewed as
microbial intelligence. These behaviors confer to the population metabolic and informative
benefits that improve fitness. Here we review recent work examining how signal exchange is used
to gather information from the local environment to compute collective responses. We then
examine how such bacterial computations are efficient means of information processes by bacterial
collectives.
Information is gathered by microbes using molecular sensors. Bacterial cells express a
variety of sensors, many in the form of two-component systems, although one-component systems
are widely distributed among prokaryotes [43]. Two-component systems are composed of a sensor
embedded on the membrane that detects a variety of chemical and physical changes in the
environment, and a response regulator that modulates patterns of gene expression based on such
stimuli [44]. The ability of bacterial species to utilize a collection of two-component systems, and
several other signal transduction pathways, to sense and respond to a variety of external inputs has
even been used to quantify bacterial IQ [45]. In addition to being able to passively detect changes
in the molecular environment around each cell, bacteria actively probe their immediate
surroundings and communicate with each other through the exchange of molecular signals.
A well-known example of this active process of gathering information from the
environment to regulate cellular behavior is through a process known as quorum sensing. Quorum
sensing is the ability to emit chemical signals, called autoinducers, and respond to high
concentrations of these signals by enacting large changes in gene expression profiles [7, 46-49].
Despite decades of work uncovering and characterizing molecular components related to quorum
sensing, debate remains over what specific types of information and collective benefits bacteria
gain in the process of quorum sensing [50].
Historically quorum sensing has been described as a mechanism to monitor population
size, as a large population of signal producing cells leads to a high concentration of signals, as

11
depicted in Figure 2.1A Quorum sensing-regulated gene expression usually occurs as signal
concentration exceeds a threshold. As the density of well-mixed populations exceeds around 107
cell/mL [51], a switch-like activation [10], typically results in major changes in gene expression
[13, 52]. Genes differentially expressed when quorum sensing is activated often are related to high
cell density behaviors and cooperativity within large populations [53]. Quorum sensing-regulated
genes control biofilm formation, horizontal gene transfer [52, 54], and the expression of
exoenzymes for harvesting shared, public nutrients [55, 56]. Given the fitness advantage of
cooperation, quorum sensing mechanisms are common to many bacterial species [57]. Over the
years, the simple perspective of quorum sensing as a way to measure population density has
evolved to encompass more sophisticated views of the computational potential of bacterial signal
exchange. Here we highlight two such computations that bacteria perform as the result of quorum
sensing, efficiency sensing and monitoring bacterial community composition.
2.3: Computation of the spatial environment
2.3.1: Computing the efficiency of signal exchange
The ability of quorum sensing signals to probe the propensity for released biomolecules to
act locally has been referred to as efficiency sensing [58]. In this reframing, bacterial quorum
sensing is a measurement of the local accumulation and reabsorption of any externally released
biomolecule, see Figure 2.1B. This perspective focuses on the importance of mass transfer and
the spatial distribution of cells within populations that were not well mixed.
An example of efficiency sensing is quorum sensing in confined spaces. Confinement of
even small numbers of cells restricts loss of released autoinducers due to diffusion, as would occur
in large volumes. When Pseudomonas aeruginosa cells were confined in microfluidics droplets,
quorum sensing was achieved even in populations with only a single cell, see Figure 2.1C [59].
The ability of cells to detect confinement via quorum sensing has also been observed in
Staphylococcus aureus, both in microfluidic experimental systems and in the realistic context of
cells engulfed by phagosomes [60, 61].
Quorum sensing in the presence of flow is another context in which cells could calculate
the efficiency to retain released molecules [11, 18, 19]. Emge et al. [18], studied the quorum
sensing of wild type P. aeruginosa and an engineered Escherichia coli carrying the same quorum

12
sensing gene circuit under controlled flows. High flow rates suppressed the expression of quorum
sensing-regulated genes by sweeping away signal that would otherwise be taken up by the cells,
despite cells being located in a high cell density environment, as shown in Figure 2.1D. Cells are
keenly attuned to their extracellular environment through quorum sensing.
As demonstrated in these examples, bacteria use cell-to-cell signaling pathways to gain
information about the local environment. A mathematical model by Cornforth et al. [62], also
demonstrated how quorum sensing systems with multiple autoinducers with distinct half-lives
could enable bacteria to infer both their density and mass-transfer environment at the same time.
In this model, the production of two autoinducers by P. aeruginosa enabled distinction between
four possible combinations of high and low mass transfer and cell density.
Figure 2.1: Bacteria use quorum sensing to gather information from the environment.
a) An illustration of a canonical quorum sensing scheme whereby a cell synthesizes a signal with
a synthase, the signal diffuses into the extracellular environment, and signal re-enters the cell to

13
bind to a receptor, which then regulates gene expression. b) An illustration of quorum sensing as
a means to count cells and to compute the efficiency of signal accumulation. In large populations,
signals (blue dots) accumulate to a high concentration and cells respond by altering gene
expression (green cells). Signal also accumulates to high concentrations in contexts of low
advection, confinement, or clustered spatial distributions leading to expression of quorum sensing
regulated genes. c) An experiment to trap a single cell in a high cell density microdroplet, from
Boedicker et al. (2009) [59]. Optical (left) and fluorescent (right) microscopy images of a single
cell expressing gfp after accumulation of released quorum sensing signals in the microdroplet. d)
Data from Kim et al (2016) [11], demonstrating that flow impedes quorum sensing. Fluorescent
microscopy images showing cells with a quorum sensing-regulated fluorescent reporter grown
with and without flow.
2.3.2: Pattern formation in response to spatial structure
Pattern formation is symptomatic of bacteria engaged in coordinated, regulated behavior.
Quorum sensing plays an important role in bacterial self-sorting into specialized, physically
localized domains that in some cases resemble multi-cellular organisms [63]. Myxococcus
xanthus, for example, leverage quorum sensing to produce such elaborate, self-organized fruiting
bodies that they were initially labeled as fungi [64].
The distribution of cells in space strongly influences cellular communication. The dynamics of
diffusion set the spatial and temporal scale over which cells interact with neighbors. Space also
introduces additional non-linearity to the interactions within biological networks [65]. In wellmixed conditions, all cells see the same chemical environment, within small molecular
fluctuations. The response of cells within these populations is approximately uniform. Contrast
this behavior with that of cells in a biofilm, where cell-cell communication is spatially dependent
and potentially localized, potentially giving rise to complex and intricately structured
communication networks. Years of experimental and theoretical studies provide insight into the
ways molecular exchange enables cells to monitor and respond to the microscale spatial structure
which they inhabit [2-4].

14
A simple example of cells responding to spatial structure is colony formation on solid
media plates. When a small population of cells is spread onto a solid agar plate to grow into
individual colonies, the number of cells and interactions between the cells influence the size and
shape of the colonies that form. In this way, the larger scale patterns of growth are determined by
the spatial distribution of cells on the plate. The patterns are shaped by depletion of nutrients as
well as the exchange of signaling molecules, as in the case of Paenibacillus dendritiformis where
the production of secondary metabolites by sister colonies biased colony growth [66]. Anyone that
has spread cells onto an agar plate has observed an inverse relationship between the number and
size of the colonies that grow. More colonies deplete the shared nutrient source inside the agar gel
resulting in smaller colonies [67].
Spatial computations can also be programmed using synthetic biology. In one system, a
mixture of positive and negative feedback on the production of diffusible signals self-regulated the
relative size of the activated region of cells. For example, a cell was programmed to sense – in
space and time- the amount of space available for growth and to scale the spatial pattern of activity
accordingly [68]. The emergent bulls-eye pattern exhibited scale-free behavior between the width
of the ring and the initial colony size, as shown in Figure 2.2.
Figure 2.2: Signal exchange for spatial computations.
Data and fluorescent microscope images (inset) from Cao et al. (2016) [68], illustrating
synthetically engineered colonies use signal exchange to calculate the spatial structure of the

15
environment and form a scale invariant spatial pattern of gene expression. Distance refers to either
colony radius (green data) or ring width (red data).
2.4: Computation of community composition
Signaling in realistic contexts is more than just production and detection of a single
signaling molecule. Not only do many bacterial species contain multiple signaling pathways, but
multiple species or strains in the same environment use chemically similar signals for
communication. Recent work has examined how bacteria within diverse populations are able to
integrate information from mixtures of signals, and in the process, gain information about the
composition of the local bacterial community.
2.4.1. Signal crosstalk and interference
The majority of bacterial signaling studies have focused on individual signaling pathways
in isolation. It has been recognized for many years, however, that many signaling pathways can be
influenced by the activity of neighboring cells [69]. In some cases neighboring cells produce
enzymes that chemically modify or destroy released signaling molecules, thereby interfering with
signal transduction [70]. Other species may detect molecular signals from their neighbors without
producing any signals of their own [71]. It is also known that the receptor proteins that bind and
respond to multiple chemical variants of a signal molecule, resulting in crosstalk between bacterial
species producing distinct yet chemically similar signaling molecules
Crosstalk generally refers to the response of a cellular signaling network to a non-cognate
signal from either the same species or strain or to signals from different species or strains. In the
context of quorum sensing, signaling crosstalk refers to the binding of a receptor by a chemically
similar but distinct signaling molecules that alters the downstream response of the cell. For
example, when cocultured together, signal from P. aeruginosa induced expression of quorum
sensing-regulated genes within Burkholderia cepacia [72]. Signaling crosstalk can excite or inhibit
gene activation and does so to variable degree, dependent on the signal and receptor molecules.

16
Such is the case in Figure 2.3A, where a receiver strain expresses quorum sensing-regulated genes
sooner (excitatory response) or later (inhibitory response) depending on signal produced by a
neighboring strain. Receptor protein may exhibit promiscuous behavior, integrating a mélange of
signaling molecules from various species. Early work on the promiscuity of quorum sensing
receptors revealed that receptor proteins are capable of recognizing and responding to multiple
signal variants [73]. Signal crosstalk has been reported in a variety of bacterial signaling systems,
including acyl-homoserine lactones and autoinducing peptides [74-76], and the strength of
crosstalk has been quantified in detail for several systems [74, 77]. These studies reveal a rich
diversity of excitatory and inhibitory crosstalk of varying strength, including neutral crosstalk in
which a signal variant has no measurable influence on gene regulation. Crosstalk has not been
accurately predicted from molecular structure and receptor sequence, but general trends have
emerged such as the degree of chemical similarity tends to indicate stronger activating crosstalk
[75].
An open question remains about whether receptors have evolved specifically to take
advantage of crosstalk. It has been proposed that crosstalk will naturally evolve due to the fitness
cost of signaling and the presence of cheating within bacterial communities [78]. Crosstalk would
provide additional information about the species or strains of cells that may be present. Theoretical
work has shown that inference of local community composition is a key benefit to populations
which utilize quorum sensing [79]. Signaling pathways may have evolved to activate earlier or not
at all in the presence of specific neighboring strains. For example, it has been suggested that
crosstalk between Streptococcus pneumoniae might enable individual strains to anticipate
production of antibiotic compounds by neighboring strains [80]. While a degenerate or noncognate response to signal may appear sub-optimal, crosstalk can improve the ability of cells to
sense their environment. Carballo-Pacheco et al. demonstrated that crosstalk is a broadly optimal
strategy for signaling [81]. Crosstalk might also enable some bacteria to cooperate, by activating
specific phenotypes only in the presence of combinations of signals from neighboring cells. For
example, crosstalk between three species found on olive trees was found to increase the ability of
one of these species, Pseudomonas savastanoi, to cause an infection [82]. More work is needed to
fully elucidate the potential benefits of crosstalk within diverse bacterial communities.

17
2.4.2: Microbes as a neural network
Looking beyond signal exchange between two species or strains, recent work more
formally analyzed how signal exchange among diverse cellular communities influences collective
decision making [83]. Distributed, diverse bacterial communities adaptively integrating and
processing information constitute a neural network. Neural networks were originally inspired by
networks of interconnected neurons [84], whereby nodes are connected to each other to reflect
their mutual influence. The population of each species or strain of bacteria is a node, the
connections between nodes represent signal exchange, and the weight of each connection
represents the strength and type of signal crosstalk. Each cell within the network measures the
mixed concentration of external signals, thereby probing the activity state and density of each cell
type in the network, as depicted in Figure 2.3B. The model can be used to predict the quorum
sensing activity of multiple bacterial strains within a mixed community of cells.

18
Figure 2.3: Quorum sensing to gather information about community composition.
a) (top) An illustration of the sender-receiver experiment from Silva et al. (2017) [74]. (bottom)
Measurements revealed neighboring strains inhibit or excite quorum sensing. b) An illustration of
quorum sensing as a neural net, whereby signals (inputs) of different concentrations with varying
weights are integrated by a cell to compute a response of ”ON” or “OFF” [83]. c) A hypothetical
illustration of a multilayer signaling network within a 3-species bacterial community. Each node

19
represents a bacterial species, and each layer represents an orthogonal type of signal. Strains
utilizing multiple signals appear on multiple layers. Arrows represent activation, flat headed
arrows represent repression. Such models should aid in understanding the consequences of signal
exchange within complex, microbial communities
The utility of this neural network model to predict the community-level signaling state of
bacterial network was shown in a recent publication by Silva et al. [83]. In this study,
communication was analyzed in a naturally occurring community of B. subtilis strains, each
producing a different variant of the ComX quorum sensing signal [85]. The pairwise crosstalk
terms were measured for each combination of species, enabling full reconstruction and prediction
of activity within the 5-strain network. Quorum sensing activity of each strain was accurately
predicted for different combinations of all five signals, revealing that even small changes in one
of the signals could modulate the community-level signaling state. These results demonstrate how
the community of strains uses a combination of multiple signals to compute an activity state for
each member of the community, thus setting an activity pattern for the community. Further
theoretical developments applying the neural network model to bacterial quorum sensing revealed
how the number of attractor states, the community-level profile of quorum sensing activity,
depended on the strength of interactions and the community composition [86].
Through the same approach, response to multiple types of signals in a single species also
could be analyzed. For example, in P. aeruginosa intracellular levels of c-di-GMP are altered in
response to extracellular stimuli such as chemoattractants, which could be a signal for motility, or
mechanical contact with surfaces, which could be a signal for biofilm formation. As a response to
the levels of c-di-GMP, genes controlling flagella needed for swarming motility and extracellular
matrix genes needed for biofilm formation are either repressed or activated. Yan et al. [87],
modeled the binary response to the intracellular secondary messenger c-di-GMP in P. aeruginosa
as a bow-tie network. Such bow-tie network integrates each signal with a different weight and
responds to the net signal through a non-linear function. Through this perspective, the architecture
of c-di-GMP works as a machine learning classifier whose function is to determine, from a set of
stimuli, to which of two categories an environment belongs—biofilm- favoring or motilityfavoring. The fittest network is the one with the highest geometric mean of fitness across multiple
environments, equivalent to a logistic regression criterion. A result of such analysis, knowledge of

20
the number of sensors in the network enables prediction of the evolutionary history of P.
aeruginosa.
The theoretical approach of neural networks may be widely applicable to the study of
natural quorum sensing communities. Natural bacterial ecosystems are extremely diverse,
containing thousands of species and strains of bacteria capable of diffusive signal exchange.
Estimates made using metagenomics approaches found that 8% of cells in one soil ecosystem were
capable of participating in the exchange of acyl-homoserine lactone (AHL) signals [88]. In another
study, the potential to participate in AHL signaling was found in 40% of the 129 bacterial species
isolated from the cottonwood tree [89]. It would appear typical that individual cells receive
signaling input from several neighboring species. Examples of crosstalk between multiple strains
has been reported for both S. pneumoniae and S. aureus [90, 91]. Given that in many contexts
quorum sensing is associated with outcomes such as virulence or biofilm formation, understanding
community-level influences on activity patterns and the potential benefits of such regulatory
schemes will be an important direction for future research
2.4.3: Integration of stimuli in multilayer networks
Mapping the exchange of several chemically related signaling molecules to a network of
interactions between multiple species or strains of bacteria will facilitate the prediction of output
states of the community. Real systems, however, include interdependencies that are not easily
understood as a single layer network: many species or strains may participate in multiple signaling
networks simultaneously. To account for this complexity, a more general framework, in which
different networks evolve or interact with each other, is needed. These are known as multilayer
networks, in which each layer encodes a specific type of information about the system [92]. Adding
an additional degree of freedom to the system, namely layers, might reveal the information
encoded in the network that cannot be captured otherwise.
From this view, one might be able to map communication in a diverse bacterial community to a
multilayer network, as depicted in Figure 2.3C. In such a network, each layer represents a class
of signaling molecules. Nodes on each layer represent bacterial species or strains that directly
respond to or produce the type of signal specified for that layer linked through weighted, directed
edges. Layers of such a network might represent orthogonal types of chemical signals that cells

21
use to communicate. It is known that some bacteria, such as V. harveyi, utilize multiple types of
chemical signals and integrate those input signals into a common pathway [93]. Therefore some
bacterial species will appear as connected nodes on multiple layers of the network. Although signal
crosstalk within microbial communities has not yet been analyzed using a multilayer network
framework, work on biological communication in eukaryotic cells, including neurons, reveals the
utility of this approach.
The multilayer network formalism more clearly captures the myriad of connections present
within complex interaction networks. A multilayer networks framework has been applied to study
tissue development in C. elegans, cancer complexomes, and protein-protein interaction networks
[94-96]. Not surprisingly, complex interaction networks such as the human brain have been
modeled as a multilayer network. Domenico et al. [97, 98] analyzed the connectivity among
different regions of the human brain through multilayer network analysis. Different brain regions
are nodes connected through the exchange of signals, represented as edges. Layers correspond to
different frequency intervals corresponding to these signals. The results of their study revealed that
hubs in multilayer networks are, in general, different from the hubs identified by standard methods
based on single-layer network analysis. Such hubs enabled distinction between the brain of a
schizophrenic patient from a healthy brain in resting state. Could similar clusters of species within
a microbial ecosystem play an important role in setting the activity of large groups of bacterial
species? For bacterial signaling networks, a multilayer analytical approach should elucidate how
coupled signal transduction within individual species impacts the community-wide response to
signal exchange and perturbations. Analysis of bacterial networks may identify new examples of
emergent phenomena and criticality, thought to be ubiquitous among living networks [99].
2.5: Optimization of signaling
The computations performed by bacterial signal exchange have an energetic cost,
suggesting that evolution has likely increased the efficiency of signal exchange. Efficient signal
exchange involves reduction in the cost of signaling components, and also incorporates aspects of
optimizing information transfer. How is quorum sensing an efficient mechanism of information
transfer? Recent work has begun to analyze bacterial communication from the perspective of
information transfer and energetic efficiency.

22
2.5.1: Energetic costs of communication
Gathering meaningful information from the environment carries a significant expense.
Broadly, storing information in a biomolecule or macromolecule demands a change in free energy
in the system, which may be gathered from the environment through metabolism. Berg and Purcell
[100], famously demonstrated that the information gathered by a cell is proportional to the cell’s
uncertainty in the external concentration and limited by the stochastic flux in the occupancy of the
outer membrane receptors. Metha and Schwab [101], build on that work by examining the tradeoff
between energy spent by the cell and the information gathered by a two-component system. Their
work revealed a limit on chemosensing, namely, the metabolic budget of the cell, as shown in
Figure 2.4A. As the uncertainty in the measurement of the external signal concentration decreases
the power consumption likewise increases. The important finding overall is that learning requires
energy, which places a strong constraint on biochemical networks in bacteria.
A similar cost for bacterial signal exchange has also been measured. Ruparell et al. [102],
examined the metabolic strain of autoinducer synthesis by comparing growth rates of wild-type
E.coli and two strains expressing LasI and RhlI, quorum sensing signal synthases native to P.
aeruginosa. The strains expressing LasI and RhlI grew slower than the strain that did not, as shown
in Figure 2.4B. This study revealed that participation in the production of quorum sensing signals
had the cost of a reduced growth rate.
Given that the computations performed by bacterial signal exchange are both energetically
costly and typically involve large populations of cells working in a coordinated fashion, it is likely
that cheaters would emerge. Cheaters are cells that do not contribute to the production of public
goods but still benefit from the work of others, thus gaining a fitness advantage over the
cooperators. In the context of bacterial computations, the information gained from signal exchange
can be viewed as a public good, and cheater cells potentially monitor signal concentrations without
the energetic investment into the full information gathering process. In a theoretical study Schuster
et al. [103], showed cooperation is evolutionarily selected for, when the metabolic burden of a
public good is low. At intermediate metabolic costs, cooperators and cheaters coexist. At high cost,
all cells abandon cooperation. Given that information gathering via quorum sensing is viewed as
a relatively low cost, that would suggest quorum sensing cheaters are tolerated in most contexts.
It is interesting to note that the majority of quorum sensing components identified in genomes

23
contain only a receptor protein, an orphan receptor, without an associated synthase protein [104].
Cheating in the context of public goods is a well-established field [105-107], and future work
should consider quorum sensing signals in the environment, from which cells can learn about local
physical and biological conditions, as a form of public good.
Figure 2.4: Energetic aspects of gathering information.
a) Theoretical results from Mehta et al (2012) [101] representing the idea that cells use energy to
reduce the uncertainty in information collected from their environment, in this case using a twocomponent system (inset). b) Experimental data from Ruparell et al. (2016) [102], showing that
cells that express the RhlI and LasI enzymes, which produce autoinducer signal, have reduced
growth rates.
2.5.2: Measuring information flow
Information theory framework enables us to better address questions related to the
efficiency of signal exchange. From this perspective, signaling process is assumed to be a black
box with external concentrations as input and phenotypic response as the output of the system.
Mutual Information is commonly used to quantify how much information cells can extract from
an external stimulus [108-110]. Calculating mutual information often requires a precise knowledge
of the input signal distribution which, for individual bacteria, is typically not known. To resolve
this issue with mutual information, channel capacity is used instead. Channel capacity is defined

24
as the maximum possible information that a communication channel can carry, the supremum of
mutual information over all possible choices of the input probability distributions [111]. Both
mutual information and channel capacity are generally calculated in bits, which gives a sense for
how many yes or no decisions can be made by the cell.
Theoretical work has examined information exchange within biological systems.
Suderman et al. (105), modeled the response in a generic signaling system as a sigmoid function
of signal concentrations, with a Gaussian noise term centered at zero to account for intrinsic noise.
Such simplified dose-response analysis is ubiquitous in many real signaling motifs. Having
intrinsic noise dependent on protein copy numbers in such system, the authors estimated the
channel capacity for a variety of transmembrane proteins, with copy numbers per cell ranging from
102
-105
. Their results showed signaling motifs prone to intrinsic noise can transmit of 4-6 bits of
information. Experimental measurements of the signaling systems studied to date encode less than
2.5 bits of information, with the majority transmitting significantly less than 1 bit (capacity of a
binary switch). As a result, extrinsic noise plays the major role in information integration.
Although noise presumably diminishes the information transmission in signaling systems,
it has been shown that noise affects information transmission in a more convoluted way. Rodrigo
et al. [112], modeled the output of a simple signaling motif as ordinary differential equations with
addition of Gaussian noises terms accounting for intrinsic and extrinsic noises. The authors showed
mutual information between the distribution of such noisy output and a given uniform distribution
of inputs does not always decrease with noise, however mutual information always decreases with
respect to the amplitude of extrinsic noise. Certain amplitudes of intrinsic noise can even amplify
the information transmission.
A few experimental efforts to date have examined the information capacity of bacterial
communication[77, 113, 114]. Mehta et al. [113], applied an information theoretic approach to
resolve why V. harveyi possess two similar quorum sensing channels responsible the same
downstream regulator. The channels differ in their use of input, AI-1, specific to V. harveyi, and
AI-2, shared among many bacterial species. Because the cells respond almost equivalently to both
signals, the channel, at first glance, encoded 0.8 bits, which is not sufficient to detect two
environmental states. Upon further inspection, however, V. harveyi increases the channel capacity
to ~1.5 bits by producing AI-1 and AI-2 at the same rate, which ensures that the extracellular
concentration of AI-2 is greater than or equal to the concentration of AI-1, which only V. harveyi

25
produces. Moreover, V. harveyi could increase the channel capacity to ~1.5 bits by positively
regulating the number of AI-2 receptors. This strategy allows V. harveyi to preferentially learn
about AI-2 at low cell density and about AI-1 at high cell density.
Perez et al. [77] studied noise and crosstalk Vibrio fischeri, which produces two quorum
sensing signals. Their experimental measurements along with mutual information analysis showed
the mutual information between the signal inputs and the lux output is less than one bit.
Furthermore, they showed the lux genes in V. fischeri do not appear to distinguish between the two
HSL inputs, and even with two signal inputs the regulation of lux is extremely noisy. Hence the
role of crosstalk from the C8-HSL input may not improve sensing precision, but rather suppresses
the sensitivity of the switch for as long as possible during colony growth.
A recent paper examined how reducing noise in the output signal might improve
information transmission (102). The study focused on the transcription factor LacI, which responds
to changes in lactose availability. The authors implemented a minimal model for the information
processed by a repressor gene circuit. They related channel capacity, repressor copy number, and
repressor-DNA binding affinity, as shown in Figure 2.5A. A circuit with zero repressors is a circuit
with zero channel capacity, since the gene is constitutively expressed. As the number of repressors
increases the channel capacity increases. At a very high repressor number, however, the channel
capacity decreases since the gene would be indefinitely repressed. A complementary interpretation
is that zero-channel capacity is a consequence of the circuit having an overlapping input-output
function whereas more separated input and output distributions imbue higher channel capacity to
the circuit. The change in channel capacity and number of bits encoded strongly depends on the
binding affinity of the repressor to the DNA. Notably they find that as the repressor concentration
increases the cell to cell variation in the number of repressors, the intrinsic noise, decreases. There
remains more to discover regarding the ability of biological systems to exchange information, and
continued dialogue between information theory and quantitative single-cell measurements should
prove fruitful in the coming years.
2.5.3: Single-cell heterogeneity in communication
Heterogeneity in the activity of individual cells impacts the efficiency of communication
within cellular populations. Single-cell variability can be the result of both genotypic variation and

26
phenotypic variation of genetically identical cells [115]. Heterogeneity has been reported in
bacterial communication systems. In the studies of quorum sensing of confined cells, significant
variability in the expression of quorum sensing-regulated genes was observed in P. aeruginosa
cells [59]. Only 20% of cells upregulated quorum sensing-controlled genes within small
populations, although the molecular mechanism causing such variability was not reported.
Measurements of heterogeneity in the expression of quorum sensing related genes have been
shown for V. harveyi [116]. V. harveyi strains lacking the genes for signal production were
engineered to contain a transcriptional fusion of gfp fused to a quorum sensing responsive reporter.
Upon addition of exogenously added signal, the response of individuals within the population
varied, as shown in Figure 2.5B. Variability in quorum sensing activation has been shown to
benefit group behaviors, such as biofilm formation [117]. Similar variability was observed in V.
fischeri and Listeria monocytogenes [118, 119].Heterogeneity in expression of quorum sensingregulated genes was also examined in genetically identical population of Sinorhizobium fredii.
Here the amount of heterogeneity in expression levels depended on the gene analyzed, and the
extent of variability in single-cell expression levels were modulated by environmental factors
[120]. In wild quorum sensing populations, genetic variation has been reported, with about 20%
of P. aeruginosa isolates containing variability in the genomic sequence of the lasR receptor
protein [121]. It remains unclear exactly how information transmission within bacterial
populations is affected by heterogeneity at the single-cell level.
A relatively unexplored aspect of single-cell variability is the coupling of phenotypic
variability within multi-species populations. Are there important consequences for populations
when two rare phenotypes of different species interact? One recent experiment to report on this
concept [122], examined the coupling of growth rates within small, mixed populations of E. coli
and Enterobacter cloacae. Using a microwell device to create replicate groups containing only a
few cells of each species, it was found that the mean and variability of growth rates was higher in
co-cultures than single strain cultures. More work is needed to understand how single-cell
variability, specifically heterogeneity in communication pathways, changes in diverse
environments composed of multiple species. These effects would be enhanced in fragmented
populations in which individual cells are not sampling population averages, but instead interact
locally with neighbors sampling their own phenotypic distributions [123].

27
Figure 2.5: The efficiency of collective behaviors in bacteria.
a) Information content of a gene circuit in E. coli, as reported by Razo-Mejia et al. (2019) [124].
Theoretical predictions closely matched experimental measurements of the channel capacity as a
function of the number and binding strength of the repressor protein. Given the energetic costs of
gathering information, cells may evolve to maximize information theoretic metrics such as channel
capacity. b) overlaid phase and fluorescence microscopy images of V. harvyei from Long et al.
PLOS Biol (2009) [116], The cells are engineered to express gfp in response to exogenously added
quorum sensing signal AI-1, which represses gene expression. The concentration is stated above
each image. (bottom) normalized histograms depicting the fluorescence per cell highlighting
heterogeneity of gfp expression among genetically identical cells. c) Overlaid fluorescent
microscope images (inset) from Larkin et al. [125], revealed heterogeneity in the opening of
potassium channels within Bacillus subtilis biofilms. Cells achieved long-range communication
even when only a fraction of the cells participated in potassium ion exchange. Wild-type cells

28
operated near a percolation transition, which maximized the cost to benefit ratio of communication.
d) Data and fluorescent microscope image (inset) from Silva et al (2019) [126], demonstrating a
percolation transition in the expression of quorum sensing-regulated genes. Beyond a critical
number of cells that degraded the signal, connected regions of activated cells no longer spanned
the network.
2.5.4: Spatial self-organization and percolation improve the efficiency of cellular
communication
Given the strong influence of spatial structure on the activity of cellular populations, it is
not surprising that many populations have been shown to self-regulate spatial structure to optimize
molecular exchange. Such modulation of spatial structure impacts the efficiency of information
flow within bacterial populations.
Most examples of emergent spatial structures are in the context of nutrient acquisition. A
population of B. subtilis utilizing an extracellular public good relied on mobility of the cells to
self-organize into a spatial pattern that optimized growth [127]. The impact of the public good, an
enzyme that processed a complex food source, was highly non-linear, requiring a high density of
cells before enzyme production influenced the growth rate. When seeded at low density, cells were
driven to motile states and self-organized into high density colonies in order to survive. The same
study examined the relationship of spatial patterning and cooperation by growing cells in an
environment with glucose, a resource that could be used without community wide action. In these
conditions, no distinct colonies form. Decreasing the concentration of glucose, however, drove
the cells towards distinct high-density colonies.
In other examples, spatial structures optimized the diffusive exchange of metabolites
within mixed microbial populations [128, 129]. A coculture of Pseudomonas putida and
Pseudomonas veronii was found to spatially segregate and organize as a result of food and oxygen
gradients [130]. The spatial organization of marine bacteria was essential for degradation of
external food particles [131]. Cocultures of engineered yeast strains also exhibited self-generated
spatial structure related to metabolic interactions between strains [132]. A pair of bacterial species
with linked metabolic pathways have even been found to have coevolved mechanisms to adopt
specific spatial structures when cocultured on surfaces [133].

29
These are a few of the many reported examples of the intimate connection between
metabolite exchange and spatial structure. Signaling efficiency is also strongly dependent on
spatial structure. Darch et al. studied quorum in aggregates of P. aeruginosa cells which were
confined and spatially positioned using a microscale 3D–printing platform. Aggregates containing
2,000 signal-producing cells were unable to signal neighboring aggregates, while those containing
≥5,000 cells communicated with neighbors as far away as 176 μm. These findings highlighted the
dependence of efficient communication on spatial structure [3]. Work on Vibrio harveyi cells
loaded into hydrogel microcapsules of various sizes reached a similar conclusion. Large aggregates
of cells, with a size of approximately 25 μm, accumulated many more autoinducers than did small
aggregates with a size of approximately 10 μm, thus demonstrating that the process of quorum
sensing relies on the spatial structure of the population [2]. These studies manipulated spatial
structure using laboratory methods, but there are likely similar examples of spatial selforganization to optimize signaling and group coordination within bacterial ecosystems.
Recently the concept of percolating networks has been discussed in the context of cellular
communication. Cells operating near the threshold of a percolation transition can improve the
efficiency of long-distance communication within a population of cells. In a percolating network,
nodes or cells are distributed on a spatial grid. Each cell has an activity level, and the percentage
of activated cells on the grid strongly influences the spatial range and overall activity level of the
entire network. When only a few cells are activated, activated cells are isolated and do not form
large patches. Above a critical percentage of activated cells, the activated cells form an
interconnected network that spans a very large range, often the entire length of the network. This
transition in the connectivity of activated regions has been reported in several biological contexts,
including embryo development [134]. Percolation networks often operate near a critical threshold
or phase transition, below which only short-range clusters interact and above which long-range or
system-spanning conduits emerge. Here we will focus on two examples of percolating networks
involving bacterial signal exchange.
Larkin et al. explored such behavior in biofilms of B. subtilis [125], where individual cells
open and close ion channels to communicate. Starved cells in the interior of the biofilm send K+
to neighbors, stalling their metabolism, until the signal wave meets the edge of the biofilm where
resources are available. The reduced consumption along the ion wave provides more nutrients for
the stressed, interior cells, increasing the fitness of the entire population. A minimal threshold

30
percentage of cells participating in K+
signaling led to the formation of system spanning channels.
A wild type B. subtilis biofilm operated near criticality, balancing the linear cost of signaling with
the highly non-linear benefit received, see Figure 2.5C.
Quorum sensing similarly can exhibit a percolation transition. Spatial patterns of the
expression of quorum sensing-regulated gene expression were studied in a synthetic two-strain
community composed of a signal producing strain and a signal degrading strain [126]. The signal
producing strain released C4-AHL and responded to a high concentration of signal by producing
a fluorescent reporter protein. The signal degrading strain produced an enzyme that degraded the
signal, thus acting as a sink for the signal. When mixed together, the size and connectivity of the
activated regions producing GFP depended on the ratio of two strains. Above a critical ratio, the
activated cells formed a connected region that spanned the entire system, as shown in Figure 2.5D.
The activated regions also followed scaling laws expected for such percolating networks [135],
demonstrating that the size and distribution of active regions within a spatially dispersed
population could be predicted from fundamental physical concepts. Long-range coordination of
signaling states was also demonstrated, even in the presence of interference from a neighboring
strain.
2.6: Future perspectives
Recent insights into the ability of bacterial communities to gather information from the
environment and coordinate large-scale behavior should enable the development of strategies to
both control diverse populations of microbes and design multispecies communities for new
applications in biotechnology. In recent years the advantages of division of labor within both
synthetic and natural microbial consortia have been reported [136-140]. The viability of these
approaches requires that communities are able to maintain relatively stable community
composition over time. Several strategies have been proposed to balance interactions within
multispecies communities, including designed cross feeding of metabolites and the use of spatial
niches to maintain diversity [141, 142]. What is often missing in this focus on metabolic balance,
is whether or not cells will maintain cellular activities of interest, other than growth of course, due
to often poorly defined regulatory interactions between community members. As highlighted here,

31
signaling interactions between species are likely commonplace in natural biological contexts.
Further work is needed to reveal how populations of cells learn about their surroundings to modify
their behaviors, and how such regulation within a community context is beneficial to both
individual species or strains and the community as a whole. Theoretical approaches described
above should help advance our understanding and prediction of how communities of cells
communicate and regulate activity within biologically diverse contexts Analysis of how
populations of cells gather information and regulate activity, bacterial computations, will help
identify strategies that maintain functional characteristics as well as species composition within
communities. A deeper understanding of how regulatory interactions are influenced by both spatial
structure and single-cell heterogeneity will be essential in many contexts and may lead to new
strategies for control and stability. Next steps should include engineered communities capable of
self-regulating and adapting to changes in biological, chemical, and physical conditions. These
approaches should incorporate cells using signal exchange to gather information about local
physical and biological conditions to calculate an appropriate response, a process that has already
evolved in many natural communities to maintain both diversity and function despite significant
uncertainty in conditions.
A lofty goal, which hopefully will become more realistic from advances in the design of
engineered communities, is the prediction and control of community function in the wild. Given
the tremendous diversity of real microbial ecosystems, it would appear that an exhaustive mapping
of species interactions, whether metabolic or regulatory, is impractical. An ideal solution to this
problem would be the identification of the general rules for how cells gather information to set
activity levels for each species and strain [114, 143], regardless of the specific community
composition or the activity of interest. Future work focused on universal strategies and limitations
of bacterial community computation should help elucidate such rules.

32
Chapter 3: Phenotypic memory in quorum sensing
This work appears as published in PLOS Computational Biology 20, no. 7 (2024): e1011696.
3.1: Abstract
Quorum sensing (QS) is a regulatory mechanism used by bacteria to coordinate group
behavior in response to high cell densities. During QS, cells monitor the concentration of external
signals, known as autoinducers, as a proxy for cell density. QS often involves positive feedback
loops, leading to the upregulation of genes associated with QS signal production and detection.
This results in distinct steady-state concentrations of QS-related molecules in QS-ON and QSOFF states. Due to the slow decay rates of biomolecules such as proteins, even after removal of
the initial stimuli, cells can retain elevated levels of QS-associated biomolecules for extended
periods of time. This persistence of biomolecules after the removal of the initial stimuli has the
potential to impact the response to future stimuli, indicating a memory of past exposure. This
phenomenon, which is a consequence of the carry-over of biomolecules rather than genetic
inheritance, is known as "phenotypic" memory.
This theoretical study aims to investigate the presence of phenotypic memory in QS and
the conditions that influence this memory. Numerical simulations based on ordinary differential
equations and analytical modeling were used to study gene expression in response to sudden
changes in cell density and extracellular signal concentrations. The model examined the effect of
various cellular parameters on the strength of QS memory and the impact on gene regulatory
dynamics. The findings revealed that QS memory has a transient effect on the expression of QSresponsive genes. These consequences of QS memory depend strongly on how cell density was
perturbed, as well as various cellular parameters, including the Fold Change in the expression of
QS-regulated genes, the autoinducer synthesis rate, the autoinducer threshold required for
activation, and the cell growth rate.

33
3.2: Author summary
Bacteria use a mechanism known as quorum sensing (QS) to collaborate when their numbers
are high. In various QS systems, cells detect specific signals that trigger certain genes, resulting in
increased production of certain molecules in response to these signals. Interestingly, these
molecules can linger even after the initial signal is gone, which can resemble a form of "memory."
Our theoretical study focuses on exploring this memory and the factors that influence it. To do
this, we used simulations and models to examine how history of exposure to signals can affect the
future response, when signals are removed, and cell density is reduced. We found that the prior
exposure to signals can influence how bacteria respond in the future, but this effect occurs under
specific conditions. This research contributes to our understanding of quorum sensing and how
bacteria adapt to environmental changes.
3.3: Introduction
Quorum sensing in bacteria is known as a mechanism to monitor cell density and
coordinate gene expression within populations of cells [5-7]. QS relies on the synthesis and
detection of diffusible signaling molecules, also known as autoinducers, by individual cells. As
cell density increases, the concentration of released signals in the environment also increases. As
the amount of external signal grows to a high concentration, the expression of target genes under
QS regulation is initiated, exhibiting a switch-like behavior [8, 10]. Since the concentration of
signals in the environment is proportional to the population size, QS is considered a proxy for cell
density. Several genes involved in collective behavior are subject to regulation under QS, including
those responsible for virulence, bioluminescence, and biofilm formation [13]. Activation of QS in
response to high signal concentrations leads to a transition of the intracellular concentrations of
expressed target genes from a low to high state, referred to as the QS OFF and ON steady states,
respectively [11, 12]. Such switch-like behavior is often attributed to the presence of positive
feedback loops within many QS circuits [8, 9]. Positive feedback in QS is due to increased
synthesis of proteins involved in signal transduction, the synthase and receptor proteins, upon
activation. As a result, low and high signal concentrations as environmental inputs can induce two
distinct phenotypes in a QS system, referred to as OFF and ON phenotypes, respectively. However,

34
it remains unclear whether the QS response is solely determined by the current cell density and
signal concentrations or is influenced by past exposure to signal.
Prior studies have shown that bacterial responses to current conditions may be influenced
by past environmental conditions. Examples of such history-dependent behavior have been
previously studied in metabolic systems, stress responses, and biofilm formation [144-147]. The
lac operon in E. coli exemplifies history-dependent behavior in bacteria [145]. When transitioning
from glucose to lactose as the carbon source, cells previously exposed to lactose exhibited shorter
growth lag phases, indicating a memory of past lactose exposure. Overexpression of LacZ proteins
further reduced the lag phase, suggesting that the memory effect is associated with carry-over of
intracellular LacZ proteins, in which the duration of the lag phase reflected the strength of historydependent behavior. Such a memory effect, resulting from the persistence of cellular components
rather than genetic inheritance, is known as "phenotypic" memory [15, 16]. Another instance of
phenotypic memory in bacteria can be observed in the stress response of Bacillus subtilis. Mutlu
et al. demonstrated that the ability of spores to recover is regulated by molecules transmitted from
the progenitor cells to the spores during sporulation timing [147]. Moreover, previous exposure of
Pseudomonas aeruginosa [146] cells to a surface imprints a memory of the surface through cyclic
adenosine monophosphate (cAMP), enabling cells to gradually become better adapted for sensing
and attachment upon returning to the surface. This phenotypic memory has been shown to
potentially provide adaptive advantages in fluctuating environments [146].
Previous studies have also analyzed memory within QS. It has been observed that
activation of competence in Streptococcus pneumoniae cells depended on the pH of media in past
cultures [54]. Sappington et al. [17] demonstrated that LasR, a transcriptional activator involved
in quorum sensing, can maintain its functional state even in the absence of the signaling molecule
for a certain period. The study revealed that the presence of a reservoir of signal-free LasR within
the cells enables the activation of transcription upon re-exposure to the signaling molecule.
Another study demonstrated that overexpression of LuxR in E. coli cells carrying the LuxI-LuxR
quorum sensing circuit significantly shortened the onset timing of signaling supporting the effect
of elevated receptor concentration is QS activation [17]. Moreover it has been observed that cells
with previous QS activity can expedite activation in neighboring cells, suggesting the presence of

35
a potential memory of prior stimulation in form of higher than basal rate of autoinducer synthesis
[148].
In this study we aim to further investigate the phenomenon of phenotypic memory in QS.
Specifically, we hypothesize that when bacterial cells transition from high to low cell densities,
the presence of excessive receptor and synthase proteins carried over from the QS ON state may
influence QS reactivation dynamics, thereby encoding a memory of past exposure. We utilize a
mathematical model of QS to examine the consequences of biomolecule carry-over when cells
move from a region of high cell density in QS ON state, to a region of low cell density. During
this transition, molecular concentrations are expected to gradually transition back to levels
associated with the OFF state. Our study aims to assess the potential carry-over effects and their
impact on the expression of QS-regulated genes. To investigate the potential sources of
phenotypic memory in QS, we evaluate the contribution of multiple components [149, 150].
Considering that each biomolecule has a different abundance and decays at a different rate, it is
plausible that multiple sources of memory may exist [144, 151, 152]. Furthermore, we develop a
simplified analytical model to determine the parameters that control the strength and duration of
phenotypic memory in quorum sensing.
3.4: Methods
LuxR-LuxI regulatory circuit can be modeled through a set of ordinary differential
equations (ODEs) [10] based on the mass-action kinetics formalism [153]. Such simplified
models assume the concentration of biomolecules transformed by the reactions depend solely on
the current amount of species, the rates at which these reactions proceed, and the stoichiometry
of the reactions [10]. These equations were used previously to model QS dynamics [10, 154].
The set of differential equations and the description of associated model parameters can be found
in Equations 3.1-3.8 and Table 3.1. These equations describe the dynamics of the quorum sensing
machinery including receptor protein LuxR in unbound (R), monomeric (RA), and
dimeric/complex forms (C), the synthase protein LuxI (I), AHL signaling molecule (A), a quorum
sensing target gene (G), and cell number (N). Although the model should apply to many quorum
sensing systems, it is inspired by the LuxR_LuxI circuit of Vibrio fischeri, a well-studied marine

36
bacterium known for its bioluminescent interaction with the Hawaiian bobtail squid. This choice
is based on the extensive availability of quantitative data regarding the QS system of Vibrio
fischeri.
Briefly, quorum sensing regulated proteins (LuxR, LuxI, and the target QS-regulated
protein) are produced at either a basal rate, when AHL concentration is low, or at an elevated rate,
when high AHL concentrations results in receptor dimer formation. The protein concentration per
cell is reduced due to both cell division and protein decay. The rate of protein degradation was set
to reflect that cell division sets the rate that proteins are diluted during cell growth. To simplify the
model, this rate of protein degradation does not change as cell growth slows near the saturating
cell density. Separate degradation and production rates could have been implemented, dependent
on the growth phase of the culture, however QS activation is expected to occur well before
reaching density-dependent growth limitations. This simplification prevents a ramp up in protein
levels during stationary phase. While dilution resulting from cell division is often the primary
mechanism of protein degradation, it should be noted that LuxR is inherently unstable in its nondimeric state, particularly when AHL concentrations are low [155], and thus is prone to rapid
degradation. In this model, cells produce AHLs in proportion to the concentration of synthase
protein present within each cell. The concentration of AHLs within the cells (A) and outside the
cells (𝐴𝑒𝑥) is linked through diffusion across the cell membrane. Cell division in this model follows
the logistic growth equation. Transcription and translation of mRNAs are not explicitly accounted
for in the model, as these molecules have much shorter half-lives than the timescale of quorum
sensing dynamics [156]. As a result, mRNAs are assumed to be in quasi-steady state and are not
explicitly tracked within the model.

37
[𝑑𝐼]
𝑑𝑡
= 𝑘5 +
𝑘3
[𝐶]
𝑘𝑑𝐼+[𝐶]
− 𝑘7
[𝐼] − 𝑘11[𝐼] (3.1)
𝑑[𝑅]
𝑑𝑡
= 𝑘6 +
𝑘4
[𝐶]
𝑘𝑑𝑅+[𝐶]
+ 𝑘−1[𝑅𝐴] − 𝑘1
[𝑅][𝐴] − 𝑘8
[𝑅] − 𝑘11[𝑅] (3.2)
𝑑[𝑅𝐴]
𝑑𝑡
= 2𝑘−2
[𝐶] − 2𝑘2
[𝑅𝐴][𝑅𝐴] − 𝑘−1
[𝑅𝐴] + 𝑘1
[𝑅][𝐴] − 𝑘8
[𝑅𝐴] − 𝑘11[𝑅𝐴] (3.3)
𝑑[𝐶]
𝑑𝑡
= 𝑘2
[𝑅𝐴][𝑅𝐴] − 𝑘−2
[𝐶] − 𝑘11[𝐶] (3.4)
𝑑[𝐴]
𝑑𝑡
= 𝑏[𝐼] − 𝑘1
[𝑅][𝐴] + 𝑘−1[𝑅𝐴] − 𝐷([𝐴] − [𝐴𝑒𝑥]) − 𝑘9
[𝐴] − 𝑘11[𝐴] (3.5)
𝑑[𝐴𝑒𝑥]
𝑑𝑡
= 𝑟𝐷[𝑁]([𝐴] − [𝐴𝑒𝑥]) − 𝑘9
[𝐴𝑒𝑥] (3.6)
[𝑑𝐺]
𝑑𝑡
= 𝑘10 +
𝑘4
[𝐶]
𝑘𝐷𝐼+[𝐶]
− 𝑘7
[𝐺] − 𝑘11[𝐺] (3.7)
𝑑[𝑁]
𝑑𝑡
= 𝑘11[𝑁] (1 −
[𝑁]
[𝑁𝑚𝑎𝑥]
) (3.8)

38
Parameter Value Description Reference
𝑘−1 10[𝑚𝑖𝑛
−1] Dissociation rate of monomer (RA) [10]
𝑘−2 1[𝑚𝑖𝑛
−1] Dissociation rate of dimer (C) [10]
kd1 100[𝑛𝑀] Dissociation constant of monomer
(RA)
[10, 157]
kd2 20[𝑛𝑀] Dissociation constant of dimer (C) [10]
k1 = 𝑘−1/𝑘𝑑1 0.1[𝑛𝑀−1𝑚𝑖𝑛
−1] Monomer (RA) formation rate [10]
k2 = 𝑘−2/𝑘𝑑2 0.05[𝑛𝑀−1𝑚𝑖𝑛
−1] Dimer (C) formation rate [10]
b 0.04[𝑚𝑖𝑛
−1] Synthesis rate of AHL by LuxI [10]
D 10[𝑚𝑖𝑛
−1] Diffusion rate of AHL through the
cell membrane
[10, 158]
k3 80 [𝑛𝑀−1𝑚𝑖𝑛
−1] Activated expression rate of LuxI estimated**
k4 80 [𝑛𝑀−1𝑚𝑖𝑛
−1] Activated expression rate of LuxR estimated**
k5 8 [𝑛𝑀−1𝑚𝑖𝑛
−1] Basal expression rate of LuxI estimated*
k6 8 [𝑛𝑀−1𝑚𝑖𝑛
−1] Basal expression rate of LuxR estimated*
kdR = kdI 100 [nM] Dimer (C) concentration at half
maximum activity
estimated ***
k7 0.006[𝑚𝑖𝑛
−1] Degradation rate of LuxI [159]
k8 0.15[𝑚𝑖𝑛
−1] Degradation rate of LuxR and
monomer (RA)
[159]
k9 0.005[𝑚𝑖𝑛
−1] Degradation rate of AHL [160]
k11 0.017[𝑚𝑖𝑛
−1] Growth rate doubling time
≈ 40𝑚𝑖𝑛
k10 0.5[𝑛𝑀−1𝑚𝑖𝑛
−1] Basal expression rate of QS target
gene (G)
Estimated
r 10−12 ml 𝑣𝐶𝑒𝑙𝑙 [161]
Table 3.1: List of parameters used in equations 3.1-3.8.
*Estimated based on [162], LuxI and LuxR are assumed to have approximately the same basal
expression rates and QS activation at a cell density near 109
cell/mL. ** Rate constants are
estimated to achieve a Fold Change of 10 for LuxI ***Estimated to obtain half maximum
activity at around 25-50 nM AHL concentration [163].
In this study, we numerically solved the system of ODEs using the scipy.integrate.odeint
function provided by the SciPy library in Python, which employs LSODA (Livermore Solver for
Ordinary Differential Equations). To characterize the “ON” and “OFF” states, simulations were
run using Equation 3.1-3.8 with an initial concentration of 1 cell/ml and the variables

39
𝐼, 𝑅,𝑅𝐴, 𝐴,𝐴𝑒𝑥, 𝐶, and G initialized to zero. Within a range spanning low to high maximum
achievable cell densities denoted by 𝑁𝑚𝑎𝑥, the system exhibits two distinct steady-state solutions
contingent upon value of 𝑁𝑚𝑎𝑥. These solutions are denoted as the "ON" and "OFF" steady states.
Moreover the 𝑁𝑚𝑎𝑥 value at which the QS regulated gene, denoted with G, reaches half of its
maximum value is defined as the critical cell density required for activation. Throughout the paper,
for cells initially in the "OFF" state, the initial conditions in the set of ODEs are set to the values
associated with the "OFF" state, while cells initially in the "ON" state are initialized with the values
corresponding to the "ON" state. Figure S3.1 shows expression of the QS target for a range of
final cell densities. Molecular concentrations are normalized by dividing by the concentration in
the ON state, listed in Table S3.1.
The process of dilution resets the external signal concentrations to zero and reduces the
initial cell density N to its post-dilution value. When comparing diluted "ON" cells to "OFF" cells,
both the external signal concentrations of initially "ON" and "OFF" cells are adjusted to zero.
3.5: Results
3.5.1: A model for quorum sensing activation and deactivation
Quorum sensing in bacteria is based on synthesis and detection of diffusible signaling
molecules known as autoinducers. LuxR-LuxI regulatory circuit in the marine bacterium Aliivibrio
fischeri is a well-known example of a QS regulatory circuits in Gram-negative bacteria [42]. A.
fisheri synthesizes the signaling molecule N-3-oxohexanoyl-homoserine lactone (3-oxo-C6-HSL
or AHL more generally) using LuxI synthase proteins. Binding to the signal leads to dimerization
of the receptor protein, and this dimer acts as a transcription regulator. As the cell density increases,
the concentration of AHLs in the environment rises, resulting in an increased number of LuxR
dimers and the upregulation of genes regulated by quorum sensing. Notably, the LuxI and LuxR
proteins are both upregulated by the receptor dimer, resulting in a positive autoregulatory circuit
[89, 164-166]. As a consequence of the positive feedback, activation of quorum sensing resembles
a switch-like behavior. This transition to an active state is accompanied by alterations in the levels
of several cellular components, including the LuxI and LuxR proteins, as well as the concentration
of signals [154]. A schematic representation of LuxR-LuxI circuit is shown in Figure 3.1A.

40
Next, we implemented the model to examine the process of quorum sensing activation and
deactivation, as shown in Figure 3.1B. Activation occurs as cells grow and AHLs accumulate
over time. Cell density starts at 106
cell/ml and saturates at 1011 cell/ml. Cells were initialized
in the OFF state, using molecular concentrations listed in Table S3.1. As shown in Figure 3.1C,
the levels of quorum sensing-associated biomolecules undergo a transition from low steady-state
concentrations (OFF state) to high steady-state concentrations (ON state).
Figure 3.1: Dynamics of quorum sensing activation and deactivation.
A) Schematic representation of the LuxR-LuxI QS regulatory circuit. B) Cartoon illustration of
QS activation and deactivation. Upon dilution of cells to low cell density and removal of
extracellular autoinducer, levels of QS-dependent genes begin to decrease. As cells grow and
produce signal, QS reactivates. Reactivation is potentially influenced by molecular memory of
being in the QS active state. C) Simulations examined the concentration of QS-regulated

41
molecules as a population of cells activated and then deactivated QS. As cell density increased,
QS-regulated molecules transitioned from low to high concentrations. Eventually cell growth and
signal production resulted in the reactivation of QS. D) Simulation results were used to calculate
the maximum decay rates for each QS-regulated biomolecule during deactivation.
Although a population of cells held at a high density will remain in the ON state indefinitely,
it is possible for cells to transition back to the OFF state under certain conditions, such as if the
population is diluted to a very low cell density. To investigate this deactivation process, the model
examines a scenario where the cell density is abruptly reduced from 1011 to 106
cell/ml, and the
external AHL concentration is reset to zero. Following the dilution step, the concentrations of
quorum sensing-associated biomolecules undergo changes due to several processes: dilution by
cell division, intrinsic degradation, transport across the membrane, and dissociation and
formation of molecular complexes. Consequently, the concentrations of the QS-associated
biomolecules transition from the ON state to the OFF state levels. Moreover, the low
concentration of signal in the environment favors loss of signal from inside the cell and therefore
a decrease in the number of dimer complexes, resulting in a reduction in the positive feedback
loop that promotes receptor and synthase production in the ON state of quorum sensing. Further
cell growth and signal synthesis eventually will lead to QS reactivation. This paper examines the
influence of molecular carry-over in initially ON cells on the reactivation dynamics and
investigates the conditions in which the phenotypic memory of being in the ON state affects
future activity states.
3.5.2: Carry-over of biomolecules during the deactivation of quorum sensing
In Figure 3.1B, a cartoon illustration depicts the process of removing a cell from a high cell
density/high AHL concentration environment. At low cell density, a reduced rate of gene
expression along with dilution due to cell division results in a decrease in the concentration of QSregulated molecules during deactivation. Simultaneously, the reactivation of quorum sensing
occurs upon further growth and signal production, which results in increase in concentration of
genes under QS regulation.
As seen in Figure 3.1C, upon abrupt changes in cell density and external AHL concentration
at t = 1500 min, the concentration of QS-regulated genes does not reach OFF-state concentrations

42
instantly, there is a slow decrease in concentrations over a period of time. This transition from ON
to OFF state, which occurs mainly due to cell division-mediated dilution, can take multiple
generations. Similarly, the normalized concentrations of LuxR, LuxI, monomer, dimers, and the
internal AHLs remain elevated for a period of time after the initial dilution. To better understand
how multiple QS-related molecules contribute to QS phenotypic memory, the normalized
concentration of each biomolecule during deactivation was fitted to an exponential decay equation
𝐶(𝑡) = 𝐶𝑂𝑁𝑒
−𝛾𝑡 + 𝐶𝑂𝐹𝐹. Here C refers to the concentration of each biomolecule, and with 𝐶𝑂𝑁
and 𝐶𝑂𝐹𝐹 denoting the characteristic concentrations of these molecules in the ON and OFF states
respectively, as listed in Table S3.1. The degradation rate was calculated for LuxR dimers, LuxR
monomers, and the total LuxR (the combination of unbound LuxR, bound LuxR, and dimer
complex). As shown in Figure 3.1D, The decay rate of internal AHLs is notably high, primarily
due to their rapid diffusion across the cell membrane. When external AHLs are removed and cells
are diluted to low cell density, the internal AHL concentration decreases to low levels. As a result,
the levels of dimers, which are the stable form of LuxR in the ON state [167], decline due to the
dissociation into monomers and the degradation of inherently unstable unbound LuxR proteins
[155]. This phenomenon leads to a concentration peak of LuxR that persists for approximately two
generations. Synthase proteins, LuxI on the other hand, have the slowest decay rates mainly due
to dilution by cell division. As seen in Figure 3.1C, the transition from ON to OFF state for LuxI
takes up to 4-5 generations. As a result, there are two important biomolecular sources encoding
phenotypic memory in quorum sensing, LuxI and LuxR complexes. Due to the carry-over effect
of LuxI, cells continue to produce AHLs at a higher rate than the basal level even after being
removed from a high cell density environment. This results in a more rapid accumulation of AHLs
as compared to signal production from a population of cells in the OFF state. Elevated levels of
LuxR presumably make cells more responsive to sense the current AHL concentrations. Next, we
explore if this molecular carry- over, QS memory, affects the dynamics of reactivation in a
population of cells following dilution to low cell density.
3.5.3: Phenotypic memory effect in quorum sensing
QS circuits can be approximated as a switch-like activation, requiring a critical signal
concentration threshold associated with a critical cell density [14]. This critical density, denoted

43
as 𝑁𝐶, is depicted in Figure S3.1, in which the expression of QS-regulated genes is shown for
populations grown to different maximum densities. As seen in this figure there is a significant
increase in the concentration of QS-regulated genes at approximately 109.8
cell/ml. This critical
density cell density (𝑁𝐶), is defined as the cell density at which the steady-state concentration of
the QS target gene reaches half the maximum value reached in the ON state. We postulated that
cells previously in the ON state might reactivate QS at a lower cell density than 𝑁𝐶, post dilution
to a lower cell density, thereby retaining a memory of past exposure to high signal concentrations.
To test this hypothesis, we compared the activation of QS for cells starting in the ON and
OFF states. Cells in initially ON state had the molecular concentration of cells grown to a density
of 1011 cell/ml. Cell in the initially OFF state had the molecular concentration of cells grown to a
density of 107
cell/ml. These two populations were simulated with a starting density of 𝑁𝑖 = 108.5
and grew to three different final cell densities: 𝐼: 𝑁𝑚𝑎𝑥 = 108.7
(< 𝑁𝑐
), 𝐼𝐼: 𝑁𝑚𝑎𝑥 = 109.3
(≈ 𝑁𝑐
)
and 𝐼𝐼𝐼: : 𝑁𝑚𝑎𝑥 = 1010(> 𝑁𝐶) cell/ml.
As shown in Figure 3.2A, the initial activity state of the cells can determine the final activity
state. For densities well below or above 𝑁𝐶, the previous activity state of QS had no impact on the
final activity state. However, for final cell densities near 𝑁𝐶, the initial activity state of QS
influenced the final activity state. Such history dependent behavior can be seen in Figure 3.2A,
when initially ON cells grown to 109.3
𝑐𝑒𝑙𝑙𝑠/𝑚𝐿 ended up in ON state, while those initially in the
QS OFF state remained in the QS OFF state despite reaching the same final cell density. Further
simulations of cell growth to a range of final cell densities revealed there exists a range of final
cell densities for which the final activity state depended on the initial activity state. As shown in
Figure 3.2B, we refer to this range of cell densities as the “memory zone”. The range of final cell
densities included in this memory zone was dependent on the initial density of the cells following
dilution. For example, as shown in Figure S3.2 over dilution of cells to 107
𝑐𝑒𝑙𝑙/𝑚𝑙 resulted in no
memory zone, as in this case the time scale over which the memory of the QS ON state was retained
was shorter than the time scale needed for reactivation of QS. Intuitively higher initial cell density
leads to accelerated reactivation and the emergence of the memory effect. Conversely, excessive
dilution prolongs the time needed for activation and diminishes the effects of the molecular carryover during the activation process, thereby reducing the extent of the memory zone.

44
Figure 3.2: Phenotypic memory impacts QS reactivation near the critical cell density.
A, top) Populations of cells were grown from a density of 108.5
cell/ml to three final cell
densities: 𝐼: 𝑁𝑚𝑎𝑥 = 108.7 < 𝑁𝑐
, 𝐼𝐼: 𝑁𝑚𝑎𝑥 = 109.3 ≈ 𝑁𝑐
, and 𝐼𝐼𝐼: 𝑁𝑚𝑎𝑥 = 1010 > 𝑁𝐶 cell/ml. A,
bottom) Normalized QS target gene expression levels for cells initially in the QS ON and QS
OFF states, corresponding to each final cell density. B) Normalized expression levels of the QSregulated gene for cells initially in the QS ON and QS OFF states, at the final cell densities in the
range of 108.5 − 1010.5
cell/ml.
To further explore the effect of each biomolecule in emergence of the memory zone,
starting from a high initial cell density, 108.5
cell/ml, we switched the value of each component
in the initial conditions of the initially ON cells individually to OFF values and plotted the
corresponding activation curve as shown in Figure S3.3A. This figure analyzed the isolated role
of different molecular components on QS memory. Switching either dimers or LuxI to OFF

45
values, leads to significant reduction in the memory zone, resulting in an activation curve that
overlaps the initially OFF cells. Conversely, the impact of other components, namely unbound
LuxR, monomer and the internal AHL concentration on width of memory zone was found to be
minimal. This is expected since the dimers are the most abundant form of LuxR in initially ON
cells due to stability in presence of AHL. Moreover, LuxI, is expected to encode memory of ON
state by maintaining an elevated rate of AHL synthesis hence formation of new dimers and
reactivation. These results demonstrate that the most influential phenotypic memory is stored
within elevated concentrations of LuxR dimers and LuxI synthase proteins.
Moreover, we explored the reverse scenario by switching the value of each individual
component in the initial conditions of the initially OFF cells to ON values and plotted the
corresponding activation curve as shown in Figure S3.3B. As seen in this figure, only when both
LuxI and dimer concentrations were switched to ON values the initially OFF cells activated at a
lower critical cell density. This result is in agreements with the findings of Figure S3.3A and
confirms that both the synthase and receptor encode the phenotypic memory in QS.
3.5.4: Cellular parameters controlling the strength of phenotypic memory in quorum
sensing
To further explore the other potential parameters influencing phenotypic memory in QS
beyond the initial cell density, we developed a simplified analytical approach. This approach
predicts the reduction in the critical cell density as a consequence of phenotypic memory.
In this model, we assume that the cells from the QS ON state are diluted to a new cell
density, denoted as N0
, where N0 < N𝐶. Upon dilution, external signaling molecules are removed
and initially set to a concentration of 0. As seen previously following the dilution, the internal
concentration of signals decreases, causing the dissociation of bound receptors, particularly stable
dimers, into unbound receptors and free signals. Although the dynamics of signal- receptor
dissociation and rebinding depicted in the system of ordinary differential equations are intricate,
we make several simplifications. Specifically, we consider that following the removal of signals
and dilution, both the dimeric and monomeric configurations of LuxR molecules rapidly dissociate
into individual AHL and LuxR units. This reservoir of receptors, referred to as total LuxR, remains
transient before ultimately transitioning to the OFF state. This transient pool of additional LuxR

46
within cells is hypothesized to heighten signal detection sensitivity relative to cells in the OFF
state with lower LuxR concentrations. To examine this hypothesis via numerical simulations, we
consider the instantaneous disassembly of dimers and monomers upon dilution. As a result, we
substituted the LuxR concentration in both dimeric and monomeric forms within initially ON cells
with values corresponding to the OFF state, while incorporating the respective values from the
instant dissociation into the pool of unbound LuxR and internal AHL concentrations. In Figure
S3.3, the activation curve for initially ON cells under instant dissociation conditions is shown with
a red dashed line. As seen in this figure, the small changes in the activation curves support the
assumption of instantaneous dissociation of dimers and signal-bound monomers following the
substitution.
Furthermore, presence of this free pool of LuxR, combined with the abundance of synthase
proteins (LuxI) in initially ON cells, leads to a potentially accelerated rate of signal- receptor
binding events [16], as the formation of complexes between LuxR and signal is a fundamental step
for the transition back to the QS ON state. In this section we define activation as the formation of
more than a threshold amount of signal bound LuxR.
To begin, we examine the effects resulting from the carry-over of LuxI proteins. Since
external signals are removed during cell dilution, we assume that the LuxR dimers dissociate at
the time of dilution, decreasing the concentration below the threshold required to sustain elevated
production rates of QS-regulated genes. As a result, the rate of gene expression reverts to the basal
rate. We assume LuxI is produced at a rate 𝛽I
, equivalent to parameter 𝑘5
in Equation (3.1) and
degrades at a rate of 𝛾I
. Moreover we assume the decrease in concentration of LuxI after dilution
is mainly due to dilution by cell division [168], setting 𝛾𝐼 =γ to be the cell growth rate which is
equivalent to parameter 𝑘11 in Equation (3.1). Solving Equation (3.9), the changes in LuxI
concentration, denoted with I, during the deactivation phase at time t after dilution can be
determined as follows:
𝑑𝐼
𝑑𝑡
= −𝛾𝐼 + 𝛽𝐼
, (3.9)
𝐼(𝑡) =
𝛽𝐼
𝛾
+ 𝐶𝐼𝑒
−𝛾𝑡
, (3.10)

47
in which CI
is constant. In the case of complete deactivation, as 𝑡 → ∞, 𝐼 → 𝐼OFF =
𝛽𝐼
𝛾
,
where 𝐼OFF represents the LuxI concentration in the OFF state. On the other hand, at 𝑡 = 0,
𝐼𝑂𝑁𝑡=0 =
𝛽𝐼
𝛾
+ 𝐶𝐼 = 𝐼𝑂𝐹𝐹 + 𝐶𝐼
. Therefore, since cells are initially in the ON state at 𝑡 = 0, 𝐶𝐼 can
be calculated as 𝐶𝐼 = 𝐼ON − 𝐼OFF, where 𝐼𝑂𝑁𝑡=0
denotes the LuxI concentration in initially ON cells
and 𝐼ON denotes the LuxI concentration in the ON steady state. By substituting the value of 𝐶𝐼
in
Equation (3.10), we can obtain the value of 𝐼(𝑡) in initially ON cells as follows:
𝐼𝑂𝑁𝑡=0
(𝑡) = 𝐼𝑂𝐹𝐹 + (𝐼𝑂𝑁 − 𝐼𝑂𝐹𝐹)𝑒
−𝛾𝑡
, (3.11)
where 𝐼𝑂𝑁𝑡=0
(𝑡) denotes the LuxI concentration at time t past dilution initially ON cells.
The ratio of the LuxI concentration in the ON state, denoted as 𝐼𝑂𝑁, to the LuxI concentration in
the OFF state, denoted as 𝐼𝑂𝐹𝐹, is defined as the Fold Change (FC), and can be expressed as 𝐼𝑂𝑁
𝐼𝑂𝐹𝐹
.
Fold change in expression of a target gene expressed under QS regulation is a measure of the
promoter activity in active to inactive states. Using this definition, Equation (3.11) can be rewritten
as:
𝐼𝑂𝑁𝑡=0
(𝑡) = 𝐼𝑂𝐹𝐹(1 + (𝐹𝐶 − 1)𝑒
−𝛾𝑡). (3.12)
If we assume that each cell synthesizes AHLs at a rate proportional to the concentration of
LuxI proteins within the cell, multiplied by a constant rate b, and that the AHLs are immediately
released into the extracellular environment, we can calculate the rate of changes in accumulated
concentration of autoinducer in the external environment, A, using Equation (3.13).
𝑑𝐴
𝑑𝑡
= 𝑣𝐶𝑒𝑙𝑙𝑁(𝑡)𝑏𝐼(𝑡). (3.13)
This rate can be expressed as the multiplication of the cell volume 𝑣𝐶𝑒𝑙𝑙, the cell density
N(𝑡), and the AHL synthesis rate b, and the concentration of synthase proteins in each cell, denoted
as I(t). The cell density 𝑁(𝑡) can be described by the exponential growth equation:
𝑁(𝑡) = 𝑁0𝑒
𝛾𝑡
. (3.14)
where 𝑁0
is the initial cell density after dilution, and 𝛾, represents the growth rate constant.
Moreover, we assume that the degradation of AHLs is negligible. By replacing I with the

48
expression in Equation (3.12), we can calculate the accumulated AHL concentration, denoted as
A, at time t after dilution of initially ON cells:
𝐴𝑂𝑁𝑡=0 = 𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑂𝐹𝐹 ∫ 𝑁
𝑡
0
(𝑡)(1 + (𝐹𝐶 − 1)𝑒
−𝛾𝑡)𝑑𝑡. (3.15)
The integral in Equation (3.15) represents the cumulative AHL production over time, and
the solution in Equation (3.16) provides an explicit expression for the accumulated AHL
concentration for initially ON cells as a function of time.
𝐴𝑂𝑁𝑡=0 = 𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑂𝐹𝐹𝑁0 (
𝑒
𝛾𝑡−1
𝛾
+ (𝐹𝐶 − 1)𝑡). (3.16)
Equation (3.16) includes two distinct terms. The first term, 𝑒
𝛾𝑡−1
𝛾
, characterizes the
production of AHL molecules under OFF-level synthase concentration, denoted as 𝐼𝑂𝐹𝐹, which is
applicable to OFF cells. The subsequent term, (FC-1)t, is due to the elevated production of AHL
molecules as a result of the heightened LuxI presence in initially ON cells. This term is
proportional to the initial cell density and quantifies the carry-over effect of heightened levels of
LuxI proteins on signal production. Equation (3.16) can be rewritten to highlight the contribution
of the additional LuxI proteins present at t = 0 due to this carry-over effect.
𝐴𝑂𝑁𝑡=0
(𝑡) = 𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑂𝐹𝐹𝑁0
𝑒
𝛾𝑡−1
𝛾
(1 + 𝜉(𝑡)). (3.17)
Here the term ξ(t), here referred to as the “LuxI memory” term, is given by ξ(t)= 𝛾(𝐹𝐶−1)𝑡
𝑒
𝛾𝑡−1
.
ξ(t) represents the ratio of the accumulated AHLs due to elevated level of LuxI in cells in the ON
state over the accumulated AHLs due to basal production, equivalent to the AHL produced by
initially OFF cells. For OFF cells, LuxI memory term is always zero, ξ(t)=0.
In Supplementary Figure S3.4A, S3.4B, the LuxI memory term ξ(t) is depicted with
respect to two parameters: the Fold Change (FC), and the doubling time= ln2/γ. Both plots
demonstrate that the LuxI memory term is at a maximum at the time of dilution. Subsequently,
this value gradually diminishes to zero as time progresses.
Next, we extend this analysis to include carry-over effects resulting from increased
concentrations of the pool of LuxR receptor in QS ON cells. By combining the carry-over of

49
LuxR with the increased signal production due to carry-over of LuxI, we can predict the overall
contribution to the reformation of LuxR bound with signal.
As shown in Equation (3.18), diluting QS ON cells to a low cell density initiates a slow
transition of the concentration of LuxR from 𝑅𝑂𝑁 to 𝑅𝑂𝐹𝐹, following:
𝑑𝑅
𝑑𝑡
= −𝛾𝑅𝑅 + 𝛽𝑅, (3.18)
𝑅𝑂𝑁𝑡=0
(𝑡) =
𝛽𝑅
𝛾𝑅
+ 𝐶𝑒
−𝛾𝑅𝑡
, (3.19)
Here, the degradation of the total LuxR occurs with rate constant 𝛾𝑅 = γ+ 𝛾𝑖𝑛𝑡𝑟𝑖𝑛𝑠𝑖𝑐,
including both dilution by cell division and the rapid intrinsic degradation of LuxR in non-dimeric
form.
Solving Equation (3.19), assuming complete deactivation as 𝑡 → ∞, 𝑅 → 𝑅OFF =
𝛽R
𝛾R
, where 𝑅OFF
represents the total LuxR concentration in the OFF state. On the other hand, at 𝑡 = 0, 𝑅𝑂𝑁𝑡=0 =
𝛽R
𝛾R
+ 𝐶𝑅 = 𝑅𝑂𝐹𝐹 + 𝐶𝑅. Therefore, since cells are initially in the ON state at 𝑡 = 0, 𝐶𝑅 can be
calculated as 𝐶𝑅 = 𝑅ON − 𝑅OFF, where 𝑅𝑂𝑁𝑡=0
denotes the concentration of total LuxR in initially
ON cells at time t, and 𝑅ON denotes the concentration of total LuxR in the ON steady state. By
substituting the value of 𝐶𝑅 in Equation (3.19), we can obtain the value of 𝑅𝑂𝑁𝑡=0
(𝑡) in initially
ON cells as follows:
𝑅𝑂𝑁𝑡=0
(𝑡) = 𝑅𝑂𝐹𝐹(1 + (𝐹𝐶 − 1)𝑒
−𝛾𝑅𝑡
). (3.20)
In Equation (3.20), the concentration of receptors in the OFF state is denoted by 𝑅𝑂𝐹𝐹, and
𝑅𝑂𝑁𝑡=0
(𝑡) is the concentration of total receptors (accounting for the dissociation of dimers and
monomers) following dilution of QS ON cells at t = 0. The Fold Change (FC) is assumed to be the
same for both LuxI and LuxR and is represented by FC.
The rate of formation of bound receptor is determined by a rate constant, here assumed to
be equal to one for simplicity, multiplied by the accumulated AHL concentration from Equation
(3.17), multiplied by the receptor concentration at time t after dilution from Equation (3.20). The
formation rate of bound receptors can be written as:

50
𝑅𝑂𝑁𝑡=0
(𝑡).𝐴𝑂𝑁𝑡=0
(𝑡) = 𝑅𝑂𝐹𝐹𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑂𝐹𝐹𝑁0
𝑒
𝛾𝑡−1
𝛾
(1 + (𝐹𝐶 − 1)𝑒
−𝛾𝑅𝑡
)(1 + 𝜉(𝑡)). (3.21)
Equation (3.21) can be rewritten by introducing the overall memory term, 𝛩(𝑡), which
accounts for the combined carry-over effects of both LuxI and LuxR.
𝑅𝑂𝑁𝑡=0
(𝑡).𝐴𝑂𝑁𝑡=0
(𝑡) = 𝑣𝑐𝑒𝑙𝑙𝑏𝑅𝑂𝐹𝐹𝐼𝑂𝐹𝐹𝑁0
𝑒
𝛾𝑡−1
𝛾
(1 + 𝛩(𝑡)), (3.22)
In which,
𝛩(𝑡) = 𝜉(𝑡) + (𝐹𝐶 − 1)𝑒
−𝛾𝑅𝑡
(1 + 𝜉(𝑡)). (3.23)
𝛩(𝑡) comprises two distinct components. The first term 𝜉(𝑡) accounts for the carry-over
effects of LuxI, whereas the second term accounts for the carry-over effects associated with both
LuxI and LuxR. The second introduces a second-order dependency on FC. This results in more
substantial effects and a quicker decay, attributable to the parameter 𝛾𝑅.
Figure 3.3A and Figure 3.3B represent the overall memory term 𝛩(𝑡), as a function of
time past dilution, plotted against Fold Change (FC) and doubling time. In Figure 3.3A, doubling
time is fixed at 40 minutes, and FC values ranges from 1 to 10. It can be observed that the overall
memory term dissipates over time, with a more pronounced effect observed for higher FC values.
In Figure 3.3B, FC is set to 10, and the doubling time varies in range of 20 to 200 min.
The overall memory term also dissipates over time, with a more pronounced effect seen for longer
doubling times. Intuitively a higher doubling time results in a decrease in the ratio of AHL
produced due to the basal rate, compared to the portion synthesized due to carry-over of LuxI in
initially ON cells. Larger doubling times lead to a smaller dilution rate of receptors, resulting in
a prolonged period of increased sensitivity to AHL concentrations. As shown in Figure 3.3B, the
overall memory term is nearly insensitive to changes in the doubling time when the doubling time
is large.

51
Figure 3.3: Overall memory term dependency on time after dilution, Fold Change, and
doubling time.
A) Overall memory term, Θ vs time after dilution as a function of Fold Change. Doubling time
is set to 40 min. B) Overall memory term, Θ vs time after dilution as a function of doubling time.
Fold Change set to 10.
We suspect the rapid degradation of LuxR in absence of AHL molecules, decreases the
potential memory effect. The figures presented in Figure S3.5A and S3.5B depict the overall
memory term 𝛩(𝑡) plotted against Fold Change (FC) and doubling time, in the absence of rapid
degradation of LuxR (𝛾𝑖𝑛𝑡𝑟𝑖𝑛𝑠𝑖𝑐=0, 𝛾𝑅 = 𝛾). Both plots show that in the absence of rapid
degradation of LuxR, the overall memory term remains elevated for a longer period. Furthermore,
it can be observed from Figure S3.5B that the overall memory term is more strongly influenced
by doubling times in the absence of rapid LuxR degradation, different than the results shown in
Figure 3.3B when rapid LuxR degradation is included. This highlights the impact of fast
degradation of LuxR in its non-dimeric form in history-dependent behavior.
3.5.5: Reactivation of quorum sensing in cells with quorum sensing memory
To further the analysis of QS reactivation following cell dilution, the rate of dimer formation
can be calculated by rewriting Equation (3.1) in terms of changes in cell density over time,
resulting in:

52
𝑅𝑂𝑁𝑡=0
(𝑡).𝐴𝑂𝑁𝑡=0
(𝑡) = 𝑣𝑐𝑒𝑙𝑙𝑏𝑅𝑂𝐹𝐹𝐼𝑂𝐹𝐹
𝑁(𝑡)−𝑁0
𝛾
(1 + 𝛩(𝑡)). (3.24)
To assess the contribution of the overall memory term to reactivation, we assume that once the
binding reaction rate in Equation (3.2), reaches a threshold value, the cells transition
instantaneously from the OFF to ON state and remain in a steady ON state indefinitely. We define
the activity state as a step function, denoted with u, in response to the receptor-signal binding
reaction rate as shown in Equation (3.3). We assume when this rate surpasses a certain threshed
denoted with (𝑅. 𝐴)𝑇ℎ𝑟𝑒𝑠ℎ, the activity state jumps from OFF to ON state here, represented by 0
and 1 states respectively.
𝐴𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝑠𝑡𝑎𝑡𝑒 = u(R(t). A(t)– (𝑅. 𝐴)𝑇ℎ𝑟𝑒𝑠ℎ) (3.25)
Using Equation (3.4), the cell density at which reactivation occurs can be calculated.
(𝑅. 𝐴)𝑇ℎ𝑟𝑒𝑠ℎ = 𝑅𝑂𝐹𝐹. 𝐴𝑡ℎ𝑟𝑒𝑠ℎ , in which 𝐴𝑡ℎ𝑟𝑒𝑠ℎ corresponds to the AHL concentration required
for activation. Solving for N, the critical cell density required for activation of initially ON cells,
denote by 𝑁𝐶
′
, can be calculated:
𝑁𝐶
′ = 𝛾
𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑂𝐹𝐹(1+𝛩(𝑡))
+ 𝑁0
. (3.26)
This approach can also calculate the cell density for reactivation of OFF cells, when the value of
𝛩(𝑡) is zero. The critical cell density required for activation of initially OFF, 𝑁𝐶 is:
𝑁𝐶 = 𝛾
𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
+ 𝑁0
. (3.27)
The memory zone is therefore defined as the range of 𝑁 values that satisfy the inequality 𝑁𝐶
′ <
𝑁 < 𝑁𝐶, corresponding to the range of cell densities for which initially OFF cells do not activate,
but initially ON cells do reactivate. Where 𝑁𝐶′ and 𝑁𝐶 are determined using Equation (3.26), and
(3.27), respectively.
As seen in both Equation (3.26), and (3.27),
the critical cell density required for activation increases with initial cell density after dilution,
𝑁0
. Moreover, it has been previously observed when cells are over-diluted with 𝑁0 << 𝑁𝐶,
activation curves for different dilution ratios overlap [169], leaving the value of the critical cell
density unchanged, 𝑁𝐶𝑁0<<𝑁𝐶
≈
𝛾𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
. This affect is due to the exponential growth within the

53
time interval required for activation, and has been previously observed experimentally [54, 169].
By replacing 𝑁𝑐 𝑤𝑖𝑡ℎ 𝑁0𝑒
𝛾𝑡𝑎𝑐𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛, the time required for activation can be calculated as:
𝑡𝑎𝑐𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛 ≈
1
𝛾
ln(
𝛾𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑁0𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
). (3.28)
Hence smaller threshold 𝐴𝑡ℎ𝑟𝑒𝑠ℎ, higher initial cell density 𝑁0
, and higher AHL synthesis
rate constant, b, are expected to reduce the time required for activation, regardless of the history
of the activity state.
Moreover, higher doubling times always reduce the critical cell density required for activation with
a second order dependency on gamma since 𝐼𝑜𝑓𝑓 =
𝛽I
𝛾
, 𝑁𝐶𝑁0<<𝑁𝐶
≈
𝛾
2𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏 𝛽I
The width of the memory zone region can be expressed as 𝑁𝐶′ − 𝑁𝐶 , and can be written as
Equation (3.29).
𝑁𝐶 − 𝑁′𝐶 =
𝛾𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
𝛩(𝑡)
1+𝛩(𝑡)
|𝑡𝑎𝑐𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛𝑂𝑁𝑡=0
. (3.29)
As seen in Equation (3.29), when 𝛩(𝑡) is small, the width of the memory zone increases
with 𝛩(𝑡) at activation time of initially ON cells, which itself is augmented with higher Fold
Change and higher doubling times, and shorter time past dilution. Consequently, a reduced time
for activation is anticipated to widen the memory zone region, as the decrease in 𝛩(𝑡) over time
diminishes the width memory zone.
It's important to note that since 0≤
𝛩(𝑡)
1+𝛩(𝑡)
|𝑡𝑎𝑐𝑡𝑖𝑣𝑎𝑡𝑖𝑜𝑛𝑂𝑁𝑡=0
< 1, the width of the memory
zone always falls between zero and 𝛾𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
. width of zero occurs to the overall memory term at
the reactivation time of initially ON cells converges to zero, 𝛩𝑡𝑎𝑐𝑡𝑂𝑁𝑡=0
→ 0, resulting in no
memory effect, and 𝑁𝐶′ = 𝑁𝐶. Conversely, an infinitely large overall memory term at reactivation
time (𝛩𝑡𝑎𝑐𝑡𝑂𝑁𝑡=0
→ ∞), leads to 𝑁𝐶′ = 𝑁0
, resulting in the maximum memory zone width of
𝛾𝐴𝑡ℎ𝑟𝑒𝑠ℎ
𝑣𝑐𝑒𝑙𝑙𝑏𝐼𝑜𝑓𝑓
.
In Figure 3.4A the QS activity state calculated from Equation (3.25), is shown for initially
ON and initially OFF cells as a function of the Fold Change (FC) and cell density. The initial cell

54
density in both cases is 𝑁0 = 108
cell/ml, and cells were allowed to grow exponentially for 700
minutes. As seen in this figure, for initially ON cells, at higher FC values, reactivation occurred
at lower cell densities compared to initially OFF cells, resulting in the emergence of the memory
zone, shown as purple region. The width of the memory zone varied depending on the FC values,
narrowing for lower FC and widening for larger FC values. Such effect is a direct result of higher
overall memory term values for higher Fold Changes.
In Figure 3.4B, representing the activity state of initially ON and initially OFF cells, the
initial cell density was reduced to 106
cell/ml. Regardless of the value of FC, the final activity
states of initially ON and initially OFF cells overlap, reducing the width of memory zone to zero.
In this case, the delayed reactivation due to over-dilution results in a nearly zero overall memory
term at the time of reactivation. This observation aligns with the findings in Figure S3.2B, where
over-dilution results in zero memory effect.
Figure S3.6 shows the activity state of initially ON and OFF cells as a function of doubling
time and cell density. As expected, the width of the memory zone described as Equation (3.29)
increases with 𝛩(𝑡), which, in turn, increases higher doubling time, expands for higher doubling
times. Additionally, an increase in doubling time leads to a lower critical cell density required for
activation, regardless of the past history of activation.
Moreover, a high basal level of AHL production primes the switch and yields a stronger
history-dependent behavior. Figure S3.7A shows the activity state of initially ON and initially
OFF cells per cell density and the AHL basal synthesis rate constant, b, which varies between
0.01 -0.1𝑚𝑖𝑛
−1
. The memory zone widens for higher b and shrinks for smaller values.
Moreover, as expected from Equation (3.26) and (3.27), higher b reduces the cell density required
for activation for both initially ON and initially OFF cells individually, while expanding the
memory zone through increasing the overall memory term Θ(t).
To assess the effect of autoinducer threshold required for activation on memory zone,
𝐴𝑡ℎ𝑟𝑒𝑠ℎ values were varied between 10-100 nM. Figure S3.7B show the activity state of initially
ON and initially OFF cells with respect to cell density and 𝐴𝑡ℎ𝑟𝑒𝑠ℎ . As expected, the cell density
required for activation increases with higher 𝐴𝑡ℎ𝑟𝑒𝑠ℎ values for both initially ON and initially
OFF cells, while the memory zone contracts with increased thresholds. Lower threshold values,
like higher basal rate constant values (b), lead to earlier reactivation, enhancing the impact of
molecular carry-over during reactivation.

55
Figure 3.4:The effect of Fold Change and initial cell density on the emergence of the memory
zone.
The colored regions indicate the state of the cells as density increases. In region I (blue) cells are
in the OFF state, in region III (pink) cells are in the ON state, and in region II (purple) the activity
state depends on the initial activity state. Solid and dashed lines indicate the boundary between
regions I and III for cells in initially OFF or ON states. Cells grow for 700 min, the doubling time
is set to 40min, and the Fold Change is kept constant at 10. A) The initial cell density is set to
108
cell/ml. Higher FC values result in wider memory zone B) The initial cell density is reduced
to 106
cell/ml, resulting in a delayed reactivation and a decrease in the width of the memory zone
to zero. For these calculations 𝛾 = 017 min
−1
, 𝑏 = 0.04 min
−1
, 𝐼𝑂𝐹𝐹 = 1000 nM, 𝐴𝑡ℎ𝑟𝑒𝑠ℎ =
20 nM, 𝛾𝑅 = 0.018 min
−1
, 𝑣𝑐𝑒𝑙𝑙 = 10−12 ml
−1
.
3.6: Discussion
Quorum sensing (QS) serves as a regulatory mechanism in bacteria, facilitating
coordinated group behavior and optimizing fitness through the controlled production of valuable
exoproducts. This mechanism is activated upon reaching a critical cell density [14]. However,
the understanding of whether the QS response depends solely on current cell density or is
influenced by past activity states, indicating a memory effect, remains limited. This theoretical

56
study aims to investigate the presence of phenotypic memory in QS, specifically focusing on the
reduction in critical cell density required for activation as an indicator of memory. Our analysis
revealed history of past exposure can affect the future response denoting a memory effect. With
the critical cell density require for activation as a measure of phenotypic memory in QS, we
showed previously ON cells can activate at a lower critical cell density compared to initially OFF
cells.
Activation at lower critical cell densities is largely due to the carry-over of synthase
proteins and receptors, which are predominantly in dimeric form in initially ON cells. An excess
of synthases in these cells leads to an earlier accumulation of signals above baseline levels
compared to initially OFF cells. This increased signal concentration boosts the rate of receptor
binding reactions, potentially facilitating the formation of new dimers necessary for reactivation.
Moreover, an initial surplus of receptors in ON cells enhances their sensitivity to current signal
concentrations by accelerating receptor binding reactions, thus potentially enabling activation at
lower cell densities compared to OFF cells. Furthermore, simulation results indicate that both
LuxR, mainly in its stable dimeric form, and LuxI play critical roles in sustaining the memory
effect by priming the quorum sensing (QS) switch. Substituting either the dimers or LuxI
molecules with those from OFF cells results in a significant reduction in the memory effect.
Additionally, replacing both LuxI and dimers with ON values in initially OFF cells led to earlier
QS activation, emphasizing the role of these components in hastening the QS response.
Interestingly, the presence of excess LuxR was found to be a crucial factor for earlier activation,
irrespective of its configuration. Substituting dimeric LuxR with unbound forms did not alter the
outcomes, suggesting that the availability of LuxR, rather than its dimeric configuration, plays a
pivotal role in the activation process.
We have shown that this molecular carry-over affects the reactivation specifically near the
critical cell density required for activation, where the time required for activation is smaller. We
further identified several influential parameters affecting the strength of the effect. This includes
the initial cell density following a change in cell density, the Fold Change in gene expression, the
threshold concentration for autoinducers, the basal rate of autoinducer synthesis, and the cell
growth rate.

57
Whether activation at a lower cell density is advantageous or detrimental to a population
of bacteria likely depends on the context. For example, previous studies have shown that quorum
sensing is the optimal strategy for producing costly public goods when activation occurs at high
cell density [170, 171]. Consequently, a significant phenotypic memory could hinder the
effectiveness of quorum sensing, resulting in the production of public goods at non-optimal cell
densities. Another potentially evolved strategy for minimizing significant memory effects in the
system could involve non-linear degradation rates of receptors in the absence of autoinducers and
the optimization of other cellular parameters, such as the autoinducer threshold and Fold Change.
Although phenotypic memory has been shown to provide advantages in fluctuating
environments by optimizing long-term fitness in metabolic regulatory systems [22, 145, 172], its
implications within QS systems remain unclear. Moreover, the optimal strategy for achieving
fitness may vary depending on the temporal order and speed of environmental changes. For
example, it has been shown that slower or diversified responses is preferred in rapidly changing
environments, potentially surpassing strong memory effect [22-26]. The potential disadvantages
of QS memory might arise when environmental correlations that historically informed bacterial
responses become decoupled due to these rapid environmental changes. In such scenarios,
memory-based responses might no longer confer the appropriate adaptive advantages, leading to
reduced fitness and possibly increased extinction risks [21]. Additionally, the ability to adapt to
changing environments often comes with a fitness cost for growth in stable environments [22].
In natural quorum sensing (QS) systems, fluctuations in cell density, signal concentration,
or both, frequently occur. The nonlinear degradation of LuxR in the absence of signals prevents
strong memory effects when there is a delay in growth after perturbation. Although LuxI content
might not degrade significantly, as dilution by cell division does not occur, delays in growth still
lead to degradation processes where receptor dimers dissociate, and unbound receptors degrade.
Consequently, by the time favorable growth conditions resume, much of the receptor content may
have already degraded, particularly if delays are prolonged, thereby nullifying the memory effect.
In quorum sensing (QS) systems, cell density perturbations can occur while external signal
concentrations remain unchanged. Enzymes like lactonases and acylases degrade signaling
molecules [173], leading to a delayed activation of QS pathways and a reduction in memory
effects. These enzymes act as a natural "reset" mechanism, modulating autoinducer levels to
prevent the reactivation of QS pathways from residual high-density signals. This adjustment avoids

58
unnecessary energy or virulence factor production and aligns QS responses more dynamically with
the current environmental and cell density conditions.
Phenotypic memory in QS is presumably advantageous in periodically fluctuating
environment. The disruption of quorum sensing activity through the removal of external
autoinducers and reduction of cell density imposes a metabolic cost for re-synthesis of
autoinducers [20] until reactivation occurs. In such cases, carry-over effects can be advantageous,
enabling cells to initiate a subsequent response at a lower metabolic cost and maximize fitness in
the new environment. This memory effect is most prominent when the initial cell density is close
to the critical cell density required for activation, ensuring that phenotypic memory is noticeable
only when the reduction in cell density is relatively small. Moreover, research has demonstrated
that in environments with extreme and random fluctuations, adopting a mixed strategy is often the
most effective for survival [16]. Phenotypic memory, in this context, resembles such a mixed
strategy, enhancing survivability in unpredictably changing environments.
Ecologically, even transient phenotypic memory can influence interspecies relationships,
potentially altering the pattern of associations between species. Additionally, it suggests that the
phenotypic memory of species is likely to co-evolve [15]. In environments subject to frequent
changes, bacteria equipped with advanced QS memory capabilities can gain a distinct competitive
advantage. This allows them to activate survival strategies and dominate ecological niches. Such
advantages are especially critical during abrupt environmental changes that demand rapid
responses for survival. Consequently, QS memory not only boosts the fitness of individual bacteria
but also affects community-level interactions and stability, potentially leading to the development
of dominant bacterial strains finely adapted to their fluctuating environments.
Regardless of whether phenotypic memory in QS serves an evolutionary adaptive
advantage, it is crucial to study this phenomenon due to various implications in industry and
medicine, including QS inhibitors. An example of this is bacteria detaching from a high cell
density, QS "ON" environment and infecting small, confined spaces resembling high cell density
conditions. This phenomenon is particularly observed in biofilms, where bacteria can disperse
from a mature biofilm to colonize new surfaces. The dispersed cells can exhibit behaviors
influenced by their previous high cell density environments, potentially accelerating the
establishment of new biofilms in confined areas. Phenotypic memory in QS affects population

59
dynamics by enabling faster responses to environmental changes, enhancing competitive
advantages in resource utilization and colonization. Moreover, QS memory potentially equips
bacteria to adapt more rapidly to environmental fluctuations, which can be crucial for survival in
dynamic conditions such as nutrient availability and host immune responses. In pathogenic
bacteria, QS memory can modulate the timing and intensity of virulence factor production,
potentially affecting their ability to evade or suppress host defenses. Bacteria like Pseudomonas
aeruginosa, known for its role in chronic infections, often utilize quorum sensing to regulate
virulence and biofilm formation, adapting quickly to new environments after dispersing from their
original colony. This adaptability is crucial in environments where rapid colonization is necessary
for survival, such as in the human lungs during infection. Furthermore, the carry-over effects in
QS reactivation can potentially influence the timing and strategy of therapeutic interventions,
particularly in QS inhibition approaches. Phenotypic memory could be crucial in strategies for
managing infections or treatments, especially considering the frequency and predictability of the
environmental pressures involved.
The presence of crosstalk and varying thresholds required for the detection to different
types of signals [74, 174], raises the possibility of a stronger or weaker memory effect in response
to fluctuations in specific types of autoinducers, or cell type. This variability can potentially lead
to varying degrees of history-dependent behavior, implying potential advantages within complex
microbial communities.
It's worth noting that parameters affecting the QS memory exhibit significant variation
across bacterial species and environmental conditions [12, 163, 167, 175, 176], potentially
influencing the strength of the memory effect. For example, the range of Fold Change values in
QS circuits can vary significantly, ranging from slightly above one to hundreds. For instance, in
Vibrio harveyi, LuxR expression demonstrates a Fold Change value of approximately 8 [167]. In
contrast, the LasI/LasR system of Pseudomonas aeruginosa exhibits a high Fold Change in gene
expression, with several hundred-fold increases in LasI and LasR expression at high cell density
[175, 176]. In this case higher Fold Change facilitate increased production of virulence factors,
enhancing competition and colonization of new environments. The optimal Fold Change values in
QS circuits likely depend on additional cellular parameters and the fitness costs of differential
expression of these genes in response to cell density, rather than solely controlling the carry-over
effects. A lower threshold concentration results in activation at a lower critical cell density and a

60
stronger history-dependent behavior upon reactivation. Most bacterial species exhibit threshold
concentrations ranging from 10-50 nM, such as 25-50nM in Vibrio fisheri [163]. For example,
Vibrio cholerae and Staphylococcus aureus form biofilms at low autoinducer thresholds thereby
low cell density [12]. Low thresholds are advantageous when rapid responses to changing
environmental conditions are needed. On the other hand, higher threshold levels result in delayed
reactivation and dampening history-dependent behaviors. Similar to low threshold values,
heightened rates of autoinducer production led to faster reactivation, enhancing memory effects.
The intricate relationship between these parameters and QS responses in fluctuating environments
necessitates further investigation.
Building upon previous experiments demonstrating the presence of phenotypic memory in
bacteria [17], we propose to extend these studies to measure the effects of molecular carry-over in
cells previously activated ('ON' state) compared to naive ('OFF') cells. Prior work has shown that
when a high concentration of signal is added to cells previously in a quorum sensing active state,
the carry-over effects led to a faster reactivation of quorum sensing [17]. Future work should
further probe quorum sensing memory, testing if carry-over of synthase proteins also changes
reactivation dynamics, as predicted in this study. Experimentally probing the limits of QS memory
would help determine the specific contexts in which this memory has an impact. To specifically
investigate the effects of each biomolecule, the synthase and receptor proteins, genetic circuits
could be designed to modulate the Fold Change and degradation rates of each protein
independently. To closely mimic the environmental fluctuations that bacterial communities
naturally encounter, we recommend exposing bacterial populations to periodically changing signal
concentrations and cell densities using microfluidic chambers. Employing microfluidic platforms
will enable precise control over experimental conditions [145, 177], providing insights into how
QS systems adapt over time to dynamic signal environments. The role of cell division and protein
degradation in QS memory should also be explored, as these are key factors that set the duration
of phenotypic memory. Our findings demonstrated that systems with a high Fold Change, low
activation threshold, and high basal levels of signal synthesis exhibit the strongest memory effects,
and these results could be experimentally validated. Such studies will not only corroborate our
theoretical findings but also expand our understanding of microbial communication in fluctuating
environments, potentially leading to novel strategies for managing microbial populations.

61
It is worth noting that throughout this study the single cell heterogeneity in response was
not considered. As a consequence of single cell heterogeneity [178] the population level activation
does not necessarily follow a step function, resulting in a graded response [179]. This graded
response can, in turn, affect the impact of molecular memory on reactivation. Further study is
needed to explore these effects in the emergence of phenotypic memory in quorum sensing.
3.7: Conclusion
In conclusion, investigating the factors influencing phenotypic memory in quorum sensing
would provide valuable insights into how bacteria respond to changing environments and optimize
their fitness. This knowledge could inform the development of innovative approaches for
regulating and manipulating quorum sensing circuits, with potential applications in biotechnology
and therapeutics.
3.8: Supplementary information
The values of OFF and ON steady states were obtained from solving the set of ODEs
starting from 1 cell/ml, setting the initial conditions corresponding to variables 𝐼, 𝑅, 𝑅𝐴, 𝐴, 𝐴𝑒𝑥, 𝐶,
and G to zero. Within a range of maximum achievable cell densities denoted by 𝑁𝑚𝑎𝑥, spanning
107 − 1012 cell/ml, the system exhibited two distinct steady-state solutions contingent upon value
of 𝑁𝑚𝑎𝑥. These solutions are denoted as the "ON" and "OFF" steady states. Moreover the 𝑁𝑚𝑎𝑥
value at which the QS regulated gene, denoted with G, reaches half of its maximum value
(corresponding to Nmax = 1011 cell/mL) is defined as the critical cell density required for activation
and is shown in red in Figure S3.1. Table S3.1 outlines the concentrations of LuxI, LuxR,
monomer, dimer, internal AHL, external AHL, and the QS target gene in both the QS ON and OFF
steady states. These values correspond to final cell densities of 107
and 1011 cell/ml respectively.

62
Figure S3.1:Quorum sensing activity state vs. final cell density.
The initial cell density is set to 1 cell/ml, and cells were allowed to grow to final cell densities
within the range of 107 − 1012 cell/ml. The expression of QS-regulated genes shows a significant
increase at the critical cell density of 𝑁𝐶~109.8
cell/ml.
Table S3.1: QS biomolecule concentrations in ON and OFF states
State LuxI[nM] LuxR[nM] Monomer[nM] Dimer[nM] Internal
AHL[nM]
External
AHL[nM]
QS target
gene[nM]
Cell
density[cell/ml]
OFF 349 47 0.7 0 1.4 0 22 107
ON 3605 21 173 1479 851 846 3280 1011

63
Figure S3.2: Over-dilution diminishes the memory effect in quorum sensing.
A, top) Populations of cells were grown from a density of 107
cell/ml to three final cell densities:
𝐼: 𝑁𝑚𝑎𝑥 = 108.7 < 𝑁𝑐
, 𝐼𝐼: 𝑁𝑚𝑎𝑥 = 109.3 ≈ 𝑁𝑐
, and 𝐼𝐼𝐼: 𝑁𝑚𝑎𝑥 = 1010 > 𝑁𝐶 cell/ml. A, bottom)
Normalized QS target gene expression levels for cells initially in the QS ON and QS OFF states,
corresponding to each final cell density. B) Normalized expression levels of the QS-regulated gene
for cells initially in the QS ON and QS OFF states, at the final cell densities in the range of 108.5 −
1010.5
cell/ml. The low initial cell density leads to an approximate equality in the critical cell
density between initially ON and initially OFF cells (Nc' ≈ Nc), diminishing the width of memory
zone to zero.

64
Figure S3.3: Carry-over effects of individual QS biomolecules on memory zone emergence.
The plot illustrates the final expression level of a QS-regulated gene for cells initially in the QS
ON and QS OFF states. Initial cell density of 108.5
cell/ml, final cell densities vary within the
range of 108.5 − 1010.5
cell/ml. A) For initially ON cells, the value of each component was
switched to OFF values, denoted with a '-' sign in the legend. B) For initially OFF cells, the value
of each individual component was switched to ON values, denoted with a '+' sign in the legend.
C) The dashed line corresponds to the scenario where both monomer and dimer concentrations
were switched to zero, with their values transferred to unbound receptors and internal AHL
molecules (instant dissociation upon dilution). Remarkably, instant dissociation had minimal
impact on the activation curve of initially ON cells.

65
Figure S3.4: LuxI memory term ξ vs. time after dilution, Fold Change, and doubling time.
A) ξ vs. time after dilution, with a fixed doubling time of 40min and varying Fold Change. B) ξ
vs. time after dilution, with a fixed Fold Change of 10 and varying doubling time.
Figure S3.5: Overall memory term Θ in the absence of rapid degradation of LuxR vs. time
after dilution, Fold Change, and doubling time.
A) Θ vs. time after dilution and Fold Change. Doubling time is fixed at 40min. B) Θ vs. time after
dilution and doubling time with a fixed Fold Change of 10 and varying doubling time.

66
Figure S3.6: The impact of doubling time on memory zone width.
This figure shows the activity states of initially ON and OFF cells within a cell density range from
108
cell/ml to 1013 cell/ml, all starting from an initial cell density of 108
cell/ml. Doubling time
varies between 20 and 200min. Fold change (FC) is set to 10, and the value of 𝐼𝑂𝐹𝐹 and 𝛾 were
adjusted for each varying doubling time. . In region I (blue) cells are in the OFF state, in region III
(pink) cells are in the ON state, and in region II (purple) the activity state depends on the initial
activity state. Solid and dashed lines indicate the boundary between regions I and III for cells in
initially OFF or ON states. For these calculations 𝐹𝐶 = 10, 𝑏 = 0.04𝑚𝑖𝑛
−1
, 𝛽 = 17 𝑚𝑖𝑛−1
,
𝐴𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 = 20𝑛𝑀, 𝛾𝑅 = 0.018𝑚𝑖𝑛
−1
, and 𝑣𝑐𝑒𝑙𝑙 = 10−12𝑚𝑙
−1
.

67
Figure S3.7: The Effect of the autoinducer synthesis rate (b) constant, and the threshold
required for activation (𝑨𝑻𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 ) on memory zone width.
The figure illustrates the activity states of initially ON and OFF cells. The initial cell density is set
to 108
cell/ml, and with a fixed doubling time of 40 min, cells are allowed to grow for 700min.
Fold change (FC) is set to 10. In region I (blue) cells are in the OFF state, in region III (pink) cells
are in the ON state, and in region II (purple) the activity state depends on the initial activity state.
Solid and dashed lines indicate the boundary between regions I and III for cells in initially OFF or
ON states. A) 𝐴𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 is maintained at 20nM with a varying autoinducer synthesis rate constant
(b). B) Autoinducer synthesis rate constant (b) is set at 0.04min
−1 with varying 𝐴𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 . For
these calculations 𝛾 = 017𝑚𝑖𝑛
−1
,𝐹𝐶 = 10, 𝐼𝑂𝐹𝐹 = 1000𝑛𝑀, 𝛾𝑅 = 0.018𝑚𝑖𝑛
−1
, and 𝑣𝑐𝑒𝑙𝑙 =
10−12 𝑚𝑙
−1
.

68
Chapter 4: Red-Light-Induced Genetic System for Control of
Extracellular Electron Transfer
This work appears as published in ACS Synthetic Biology 13, no. 5 (2024): 1467-1476.
4.1: Abstract
Optogenetics is a powerful tool for spatiotemporal control of gene expression. Several
light-inducible gene regulators have been developed to function in bacteria, and these regulatory
circuits have been ported into new host strains. Here, we developed and adapted a red lightinducible transcription factor for Shewanella oneidensis. This regulatory circuit is based on the
iLight optogenetic system, which controls gene expression using red light. Promoter engineering
and a thermodynamic model were used to adapt this system to achieve differential gene expression
in light and dark conditions within a S. oneidensis host strain. We further improved the iLight
optogenetic system by adding a repressor to invert the genetic circuit and activate gene expression
under red light illumination. The inverted iLight genetic circuit was used to control extracellular
electron transfer (EET) within S. oneidensis. The ability to use both red and blue light-induced
optogenetic circuits simultaneously was demonstrated. Our work expands the synthetic biology
toolbox of Shewanella, which could facilitate future advances in applications with electrogenic
bacteria.
4.2: Introduction
Optogenetics combines light-sensitive proteins and genetic techniques to control cellular
processes within living organisms [27]. Synthetic optogenetic circuits can be constructed to tune
gene expression with light-responsive transcription factors [27]. In recent years, optogenetic
circuits have been developed to control gene expression in bacteria in response to illumination
with blue, green, red, or near-infrared light [36-41]. Optogenetic circuits have been utilized to
control gene expression to regulate many microbial processes, such as biochemical production
[28-30], biofilm formation [31, 32], and bacterial infection [30]. Although optogenetics has been

69
implemented in model bacteria, such as Escherichia coli [31, 40], Bacillus subtilis [180] and
Pseudomonas aeruginosa [30, 181], there is still a need to extend these systems to other hosts for
specific microbial processes.
Shewanella oneidensis MR-1 is a model electroactive organism, whose extracellular
electron transfer (EET) pathways have been well studied [33, 34]. EET within S. oneidensis MR1 utilizes a network of multiheme c-type cytochromes to route electrons from the cellular interior
to external electron acceptors [182-187]. Synthetic biology strategies have recently been
developed to regulate the EET capabilities of S. oneidensis MR-1, such as using genetic circuits to
control the genes encoding the multiheme c-type cytochromes of the EET pathway. West et al.,
developed a native inducible system to tune the expression level of one multiheme cytochromeporin complex (MtrCAB) to control EET capabilities [35]. In another study, clustered regularly
interspaced short palindromic repeats interference (CRISPRi) and small regulatory RNA (sRNA)
were used to repress the transcription and translation of mtrA to regulate EET efficiency [188].
Recently, Dundas et al., developed chemical induced transcriptional logic gates to control the EET
flux of S. oneidensis MR-1 through tuning the transcription and translation of EET related genes
[189]. A plasmid toolkit with different promoters and replication origins were characterized and
utilized to control cytochromes expression in S. oneidensis MR-1 for EET regulation [190] . The
EET pathway of S. oneidensis MR-1 had also been reconstructed successfully in E. coli
through heterologous expression of the related genes encoding the c-type cytochromes [191, 192].
Despite these advances, synthetic biology toolboxes, in particular the use of optogenetic gene
circuits to control EET, are still limited in S. oneidensis.
S. oneidensis can transfer electrons through cytochromes on the nanometer-scale and can
also form living conductive biofilms for long-distance electron transport across neighboring cells
on the micrometer-scale [193-195]. We previously developed a lithographic strategy to pattern
conductive biofilms of S. oneidensis, using the blue light-induced genetic circuit pDawn to control
cell aggregation [196]. This technique enabled tunable current generation by varying the
dimensions of electroactive biofilms. Since our previous work showed the potential of using
optogenetics to pattern electrogenic microbes on the micrometer scale, adapting additional
optogenetic systems for S. oneidensis could enable new strategies to control EET and develop
advanced living electronics.

70
A recently published paper showed a single-component red light-induced optogenetic
system, iLight, for transcriptional regulation in E. coli [197]. We developed and adapted this iLight
optogenetic system into S. oneidensis through promoter engineering and a thermodynamic model.
Then we improved the iLight optogenetic system to activate the gene expression using red light in
S. oneidensis. Finally, we used this iLight genetic circuit to control the expression of cytochromes
and regulate the EET activity of S. oneidensis. Our work demonstrated a new circuit to control
gene expression in S. oneidensis using red light and demonstrated light-induced EET activity of S.
oneidensis.
4.3: Results
4.3.1: Importing the iLight optogenetic system to S. oneidensis
The iLight optogenetic system was originally developed and optimized in E. coli [197]. It
consists of a plasmid that encodes a light-sensitive repressor. The repressor contains a LexA408
DNA-binding domain fused to a photosensory domain IsPadC-PCM. This protein has a sfGFP
fused to the C-terminus. An RFP reporter gene was regulated by the iLight repressor (Figure
4.4A). The proposed mechanism of action of the iLight optogenetic system was red light induced
tetramerization of iLight photosensory module, which enabled the LexA408 DNA-binding
domains to form active transcription factor dimers [197]. The iLight photosensory module requires
biliverdin IXα (BV) tetrapyrrole as the chromophore to enable the light-responsive function of the
iLight repressor (Figure S4.1). In our system, the ho1 gene, a heme oxygenase, is expressed from
the iLight plasmid for BV synthesis (Figure 4.1A).
We then tested this single plasmid iLight genetic circuit in E. coli and S. oneidensis. In E.
coli, RFP fluorescence measurements confirmed the iLight genetic circuit repressed expression of
the RFP reporter gene under illumination of red light (Figure 4.1B). When cultured in the dark,
the RFP expression was derepressed resulting in 40-fold increase in RFP expression (Figure 4.1B).
However, in S. oneidensis transformed with the iLight plasmid, the expression level of the RFP
reporter was similar in both cells cultured in red light and in the dark (Figure 4.1B). To determine
if this circuit failure was caused by limited expression of the iLight repressor in a new host, we
measured expression using the sfGFP fusion to iLight. The expression level of iLight repressor in

71
S. oneidensis was much lower than in E. coli (Figure 4.1C). We hypothesized that the low
expression of the iLight repressor could be the reason why the iLight genetic circuit did not result
in light-regulated gene expression in S. oneidensis.
Figure 4.1: Characterization of the iLight genetic circuit in S. oneidensis and E. coli.
A) The single plasmid iLight genetic circuit contained iLight repressor, iLight reporter and iLight
cofactor. The RFP reporter measures light-regulated gene expression and sfGFP measures the
expression level of the iLight repressor. B) RFP fluorescence intensity measurements of iLight
reporter expression in S. oneidensis and E. coli cultured under red light and dark conditions,
respectively. p = 0.7462 for S. oneidensis Light vs S. oneidensis Dark and p = 0.0006 for E. coli
Light vs E. coli Dark (two-tailed unpaired t test). C) GFP fluorescence intensity measurements of
iLight repressor expression in S. oneidensis and E. coli. P = 0.0005 for S. oneidensis vs E. coli
(two-tailed unpaired t test). The measurements (mean ± SD) were derived from triplicate
experiments. Significance is indicated as ***p < 0.001 and ns (not significant) p > 0.05.

72
4.3.2: Adjusting the expression level of the iLight repressor
The expression of iLight repressor was much lower in S. oneidensis than in E. coli,
suggesting differential gene expression in light and dark conditions might depend on the level of
repressor expression. To test this hypothesis, we first developed a thermodynamic model to predict
how the expression level of the iLight repressor impacts the fold change of the iLight reporter
when exposed to dark and light conditions (see Supplementary Methods and Figure S4.2). The
thermodynamic model [198-200] investigates RFP gene expression considering the binding
probabilities of RNA polymerase and iLight repressor molecules to the promoter region in red light
and dark conditions (Figure S4.2A). Additionally, it assumes that the dimeric form of the iLight
repressor in dark exhibits a non-zero, but lower, probability of binding to the specific site compared
to tetrameric form in red light. We found that the fold change in gene expression was largest for
intermediate expression levels of iLight repressor (Figure S4.2B and C), suggesting that
increasing expression of the iLight repressor might improve the performance of this optogenetic
circuit within S. oneidensis.
The expression of the iLight repressor was modified by site directed mutagenesis of the
promoter region (Figure 4.2A). The original promoter J23116 was weak [201]. Based on the
Anderson promoter collection (http://parts.igem.org/Promoters/Catalog/Anderson), the promoter
of the iLight repressor was mutated to achieve a broad range of expression levels. As shown in
Fig. 2B, the modified promoters varied in expression of iLight repressor over 44-fold. Expression
of the RFP iLight reporter gene in both light and dark conditions was measured for each of these
promoters (Figure S4.3). Promoters J23102 and J23108, which expressed intermediate levels of
iLight repressor, showed the largest fold change of the iLight reporter (Figure 4.2C). The fold
changes of iLight reporter (dark/red light) were around 12 for both iLight-J23102 and iLightJ23108 (Figure 4.2C). We also tested the performance of these iLight genetic circuits with
modified expression of the iLight repressor in E. coli. The fold change in expression of the iLight
reporter decreased to 1 at high levels of iLight repressor expression (Figure S4.4). Unlike in S.
oneidensis, the weakest promoters resulted in fold changes above 20, likely due to these weak
promoters having higher expression in E. coli than in S. oneidensis. In summary, adaption of the
iLight optogenetic system for use in S. oneidensis required adjustment of the expression level of
the iLight repressor.

73
Figure 4.2: Adapting the iLight optogenetic system for S. oneidensis.
A) Site directed mutagenesis of the promoter to tune expression of the iLight repressor. B)
Expression of the iLight repressor in S. oneidensis from these promoters, as measured via
expression of co-transcribed sfGFP. (C) Fold changes of the iLight reporter (dark/red light) in S.
oneidensis strains containing different promoters of iLight repressor. The measurements (mean ±
SD) were derived from triplicate experiments
4.3.3: Creating an inverted iLight optogenetic system, for light-activated gene regulation
The iLight optogenetic system could work in S. oneidensis MR-1 after adjusting the
expression level of the iLight repressor. In the original report, the iLight system was used to repress
gene expression in E. coli with red light (Figure 4.3A), but other optogenetic circuits have been
modified for both activation and repression with light inputs. Next, we improved the iLight
optogenetic system to activate the target gene expression through red light illumination in S.
oneidensis MR-1.
To accomplish this, we incorporated the iLight repressor into an inverter genetic circuit by
adding a second repressor. Expression of the second repressor regulates the target gene and is
regulated by iLight repressor. After red light illumination, the expression of this second repressor
is repressed by the iLight repressor, resulting in derepression of the target gene (Figure 4.3D). We
tested three different repressors, the λ phage repressor cI and the TetR-family repressors PhIF and
SrpR40, with their cognate promoters to invert the iLight-J23102 genetic circuit. All three
repressors resulted in increased expression of the target gene in response to red light (Figure 4.3E

74
and Figure S4.5A-C). The fold changes of iLight reporter RFP between red light and dark
conditions were 16 for cI repressor, 20 for PhIF repressor and 12 for SrpR repressor (Figure 4.3E
and Figure S4.5A-C). Microscopic images showed that the non-inverted iLight-J23102 cells had
strong red fluorescence after being cultured under dark but weak fluorescence after being cultured
with red light (Figure 4.3C), as expected for light-induced repression of gene expression. The
microscopic images of inverted iLight-J23102(PhIF repressor) strain showed the cells had strong
red fluorescence after being cultured with red light, but weak fluorescence after being cultured in
the dark (Figure 4.3F), as expected for light-induced activation of gene expression. We also found
that inverting the iLight-J23108 genetic circuit using the cI repressor resulted in just 5-fold change
in expression of the reporter gene (Figure S4.5D), which was much lower than those of inverting
iLight-J23102 (Figure S4.5A). The reason for this could be higher leaky expression of the
inverting repressor (cI) for iLight-J23108 relative to iLight-J23102 under red light condition
(Figure S4.3). We selected the iLight-J23102(PhIF repressor) genetic circuit for subsequent
experiments, which had the highest fold change.
We then checked whether S. oneidensis could utilize two different light-regulated genetic
constructs responding to different colors of light simultaneously. To achieve this, we introduced
the blue light-induced pDawn [37] genetic circuit to control GFP expression and red light-induced
iLight genetic circuit to control RFP expression within S. oneidensis. The iLight circuit used here
did not have the sfGFP fusion to the iLight repressor. As we expected, S. oneidensis cells
containing both iLight and pDawn genetic circuits had strong red fluorescence only when cultured
with red light and strong green fluorescence only when cultured with blue light (Figure S4.6).
Taken collectively, we inverted the iLight optogenetic system to activate the target gene
expression under red light illumination in S. oneidensis MR-1 and the fold change was increased
compared with the non-inverted iLight genetic circuit. The inverted iLight genetic circuit could
work together with blue light-induced pDawn genetic circuit to control the expression of two
different genes with two lights, respectively.

75
Figure 4.3: Characterization of the non-inverted iLight genetic circuit iLight-J23102 and
the inverted iLight genetic circuit iLight-J23102(PhIF repressor) in S. oneidensis.
A) Genetic circuit of non-inverted iLight-J23102. B) RFP fluorescence intensity measurements of
iLight reporter expression in the non-inverted iLight-J23102 strain cultured under red light and
dark conditions, respectively. p = 0.0006 for Light vs Dark (two-tailed unpaired t test).
C) Microscope observation of iLight reporter and iLight repressor for non-inverted iLight-J23102
cells cultured under red light and dark conditions, respectively. D) Genetic circuit of inverted
iLight-J23102(PhIF repressor). E) RFP fluorescence intensity measurements of iLight reporter
expression in the inverted iLight-J23102(PhIF repressor) strain cultured under red light and dark
conditions, respectively. p < 0.0001 for Light vs Dark (two-tailed unpaired t test). F) Microscope

76
observation of iLight reporter and iLight repressor for inverted iLight-J23102(PhIF repressor) cells
cultured under red light and dark conditions, respectively. Scale bars 10 μm. The measurements
(mean ± SD) were derived from triplicate experiments. Significance is indicated as ***p < 0.001
and ****p < 0.0001.
4.3.4: The extracellular electron transfer activity of S. oneidensis can be regulated with red
light
S. oneidensis MR-1 can transport electrons from cytosolic metabolism to external
electrodes using an EET pathway involving c-type cytochromes located in the inner membrane,
periplasm, and outer membrane [34] (Figure 4.4A). To determine if we can control the EET
activity of S. oneidensis using light, we used the iLight optogenetic system to control the
expression of multiple cytochromes within this EET pathway. We constructed S. oneidensis strain
iLight-MtrC which utilized the inverted iLight-J23102 (PhIF repressor) genetic circuit to activate
the expression of outer membrane cytochrome MtrC with red light. The host strain for this
construct was S. oneidensis ΔmtrCΔomcA [202] in which the key genes encoding outer membrane
cytochromes were deleted from the genome. MtrC expression from the promoter was determined
by measuring the fluorescence due to the RFP gene co-transcribed from the same promoter (Figure
S4.7A). As expected, the RFP fluorescence was much higher when culturing iLight-MtrC strain
under red light illumination, while the RFP fluorescence was weak when culturing the same strain
under dark condition (Figure S4.7B). These results suggest that MtrC could be expressed under
red light illumination with very low expression in the dark.
Next, EET activity was measured for the iLight-MtrC strain cultured under light and dark
conditions using colorimetric assays for extracellular redox activity. S. oneidensis can perform
anaerobic respiration by reducing a wide range of external terminal electron acceptors [203-208].
We utilized iron citrate and methyl orange (MO) as the electron acceptors to measure the Fe3+ and
azo dye reduction capabilities of the iLight-MtrC strain. As shown in Figure 4.4B and C, the
iLight-MtrC strain had much higher Fe3+ and MO reduction rates when cultured under red light
than those of iLight-MtrC grown in the dark. The reduction activities of red light illuminated
iLight-MtrC cells were comparable with those of wild type cells cultured either under red light or

77
in the dark. The reduction activities of iLight-MtrC cells grown in the dark were as low as those
of ΔmtrCΔomcA cultured either under red light or in the dark (Figure 4.4B and C). These results
indicated the reduction activities of the iLight-MtrC strain could be controlled by light.
Finally, we performed electrochemical measurements to determine if light-induced
expression of MtrC would modulate current production in S. oneidensis biofilms. Cells were grown
in bioreactors with planar indium tin oxide (ITO) coated glass coverslips as bottom with or without
red light illumination. Cyclic voltammetry (CV) was performed for the biofilms on the ITO
electrodes. As shown in the cyclic voltammetry curves in Figure 4.4D, we found higher current
for red light illuminated iLight-MtrC cells than the same strain grown in the dark. The average
maximum current from cyclic voltammetry measurements of three biological replicates were 180
nA for red light illuminated cells and 110 nA for cells cultured under dark (Figure 4.4D inset
graph).
We also constructed S. oneidensis strain iLight-STC to use iLight-J23102(PhIF repressor)
genetic circuit to control periplasm cytochrome STC expression by red light within S. oneidensis
ΔstcΔfccA. STC could be expressed under red light illumination with very low expression in the
dark (Figure S4.8A). The Fe3+ reduction activities of the iLight-STC strain could also be
controlled by red light (Figure S4.8B). These results demonstrate that we can regulate the EET
activity of S. oneidensis using optogenetic circuits that regulate cytochrome expression.

78
Figure 4.4: Using iLight-J23102(PhIF repressor) genetic circuit to control the outer
membrane cytochrome MtrC expression for light-induced extracellular electron transfer
(EET) activity in S. oneidensis.
A) The EET pathway of S. oneidensis MR-1. B) Iron reduction assay for iLight-MtrC strain,
ΔmtrCΔomcA and wild type with blank plasmid iLight-J23102(PhIF repressor) after being
cultured under red light and dark conditions, respectively. C) Methyl orange decoloration assay
for iLight-MtrC strain, ΔmtrCΔomcA and wild type with blank plasmid iLight-J23102(PhIF
repressor) after being cultured under red light and dark conditions, respectively. D) Cyclic
voltammetry curves for iLight-MtrC strain after being cultured under red light and dark conditions,
respectively. The inset graph showed the average maximum current from cyclic voltammetry
measurements of three biological replicates. Data shows mean ± SD from triplicate experiments.

79
4.4: Discussion
We developed and adapted a red light-induced genetic circuit for S. oneidensis MR-1 based
on a previously reported iLight optogenetic system [197]. The iLight genetic circuit reported for
E. coli did not initially work in S. oneidensis, with no fold-change of gene expression in response
to light. Through modeling and experimental tests, we found that the iLight genetic circuits
required a specific expression range of the light-responsive transcription factor to function
properly. Our modeling suggested that an intermediate level of the repressor would be needed for
light-regulated gene expression. Low expression of iLight repressor could result in insufficient
formation and binding of the repressor tetramer under the red light condition, which led to
insufficient repression of the iLight reporter. High expression of iLight repressor resulted in
repression of the regulated gene under both dark and light conditions, as the large concentration of
the repressor compensated for the weak binding of the repressor in the dark state. Tuning of the
expression level of the iLight repressor via promoter engineering identified constructs with
differential expression of the reporter gene in light and dark conditions. This work demonstrated
how the expression levels of transcription factor proteins can be critical to the function of a genetic
circuit, and similar promoter optimization may be needed to adapt the iLight optogenetic system
in other bacterial host strains.
To expand the regulatory capability of the iLight system in S. oneidensis, we inverted the
iLight genetic circuit by adding an additional repressor to activate the gene expression by red light
(Figure 4.3). We selected three different repressors with their cognate promoters [209]. Inverting
iLight with cI and PhIF repressors achieved higher expression of RFP than with a circuit using the
SrpR repressor. The fold changes in gene expression were greater in the inverted iLight genetic
circuit containing cI and PhIF repressors than in the non-inverted iLight. The reason could be the
cognate promoters of cI and PhIF repressors are stronger than the ColE promoter, so expression
levels of RFP are higher in inverted iLight under red light than that of non-inverted iLight under
dark. A similar trend was observed for the blue light-induced pDawn and pDusk genetic circuits3
.
Prior efforts to regulate the EET activity of S. oneidensis usually depended on using
chemically induced genetic circuits to control the expression of cytochromes [35, 189, 190].
Compared with chemical induced genetic circuits, light-induced genetic circuits make it possible

80
to create spatial patterns of microbial activity [29, 31, 32, 210]. Our work using the red lightinduced iLight genetic circuit to control cytochromes expression further expands the synthetic
biology toolboxes of S. oneidensis and demonstrates light-induced EET activities, which can be a
promising approach to spatiotemporally control the EET of S. oneidensis. This work allowed us to
tune the electron transfer using light. Developing optogenetic tools for S. oneidensis will
have implications for both studying and harnessing bioelectronics and the development of
advanced living electronics.

4.5: Methods
4.5.1: Bacterial strains and plasmids
Escherichia coli DH5α and NEB stable were used for plasmid construction. Shewanella
oneidensis MR-1 was used as the host to characterize the performances of different non-inverted
and inverted iLight constructs in S. oneidensis. S. oneidensis ΔmtrCΔomcA was used as the host
to contain the plasmid with expression of outer membrane cytochrome MtrC controlled by the
inverted iLight genetic circuit. S. oneidensis ΔstcΔfccA was used as the host to contain the plasmid
with expression of periplasm cytochrome STC controlled by the inverted iLight genetic circuit. E.
coli NEB stable was used as the host to characterize the performances of different non-inverted
iLight constructs in E. coli.
The original iLight plasmid [197] was purchased from Addgene (Catalog #170268). A ho1
gene was amplified from plasmid pNO41 [39] (Addgene, Catalog #101067) and then added to the
original iLight plasmid to create iLight-J23116 plasmid. To construct iLight plasmids with
different promoters transcribing the iLight repressor, iLight-J23116 was used as the starting
plasmid. Site directed mutagenesis was then performed for the promoter region of iLight repressor
through using NEB Q5® Site-Directed Mutagenesis Kit with non-overlapping primers or
NEBuilder® HiFi DNA Assembly (New England BioLabs, MA, USA) with overlapping primers.
cI, PhIF and SrpR repressors with their cognate promoters were amplified from plasmid pDawnmCherry [196], pRF-PhIF [209] (Addgene, Catalog #49367) and pRF-SrpR [209] (Addgene,
Catalog #49372), respectively. Then, these DNA fragments were added to the iLight-J23102 and
iLight-J23108 plasmids after the LexA408 operator to obtain different inverted iLight genetic

81
circuits. To test whether S. oneidensis could utilize iLight and pDawn genetic circuits to respond
to red light and blue light simultaneously, we constructed a plasmid to control GFP expression by
pDawn. We removed the GFP reporter for the expression of the iLight repressor and changed the
antibiotic resistance from spectinomycin to kanamycin in the iLight plasmid. For constructing
light-induced EET plasmids, mtrC and stc genes were amplified from the genome of S. oneidensis
MR-1 and added before the mCherry gene in the inverted iLight plasmids under the control of the
cognate promoter of the second repressor.
All strains, plasmids and primers used in this study are listed in Table S1, S2 and S3.
4.5.2: Growth conditions
E. coli strains were cultivated in Lysogeny broth (LB) medium (10 g/L tryptone, 5 g/L yeast
extract and 5 g/L sodium chloride) at 37 oC, 200 rpm. S. oneidensis strains were cultivated in LB
medium or minimal medium at 30 °C, 200 rpm. The minimal medium recipe can be found in Table
S4. When necessary, media were supplemented with spectinomycin (Spec, 100 μg/mL) and
kanamycin (Kan, 50 ug/mL).
For anaerobic culturing of S. oneidensis, the minimal media with different kinds of electron
acceptors were purged with nitrogen. The minimal medium without vitamin solution was purged
with nitrogen and used for the electrochemical measurements. The anaerobic culturing was
performed in the sealed serum bottles and electrochemical measurements were performed in an
anaerobic chamber.
4.5.3: Fluorescence measurements and microscopy
To characterize the performance of iLight genetic circuits in E. coli and S. oneidensis,
overnight cultures (1%, v/v) of the strains were transferred into 5 mL fresh LB broth and grown to
late log phase (OD600nm about 1-1.5). Then, the cultures (1%, v/v) were seeded into 5 mL of LB
broth and incubated at 37 oC or 30 oC, for E. coli and S. oneidensis, respectively, either under the
red light or dark condition while shaking at 200 rpm. Red light was provided by attaching LED
strip lights to the wall inside the shaker with an intensity of 150 μW/cm2
[197]. S. oneidensis

82
containing both the pDawn and iLight genetic circuits were exposed to blue light and red light,
both with intensities of 150 μW/cm2
. An optical power meter (PM100USB, Thorlabs) was used
for measuring the illumination intensity of lights. Cultures were collected after 18 h incubation to
measure the fluorescence of RFP and GFP, and the cell optical density (OD600nm). Quantitative
RFP and GFP fluorescence measurements were performed via a plate reader (Infinite 200 PRO,
Tecan) at an excitation wavelength of 590 nm and an emission wavelength of 650 nm for RFP and
an excitation wavelength of 485 nm and an emission wavelength of 515 nm for GFP. The OD600nm
was determined using a spectrophotometer (Spectronic 200, Thermo scientific). Relative
fluorescence intensity was calculated by normalization against OD600nm of whole cells.
Autofluorescence was subtracted by measuring fluorescence of wild type strains. Fluorescence of
RFP and GFP was imaged via fluorescent microscope equipped with 100× oil immersion objective
lens (Revolve, Echo).
4.5.4: Iron reduction measurements
Resting cell ferrozine assay [35] was used to measure the Fe3+ reduction abilities of S.
oneidensis strains. Cells from late log phase LB cultures were diluted into 5 mL fresh LB broth.
Then, cells were incubated for 18 h at 30 oC under either red light or dark condition while shaking
at 200 rpm. Cells were collected by centrifuging (5840R, Eppendorf) at 4,200 rpm, 4 oC for 15
min and then washed with fresh minimal medium two times. After that, cells were inoculated into
sealed serum bottles containing 25 mL anaerobic minimal medium to an OD600nm of about 0.1. 2
mM of ferric citrate was added into the anaerobic minimal medium as the electron acceptor. The
samples were incubated at 30 oC under dark conditions without shaking. Every two hours, 10 μL
of each sample was added immediately to 90 μL 1 M HCl in a 96-well plate followed by 100 μL
0.01% ferrozine. Then, after mixing the samples well and letting them sit for 10 min, the
absorbance of the samples at 562 nm was determined with a plate reader (Infinite 200 PRO, Tecan).
A standard curve of freshly made ferrous sulfate was used to determine the Fe2+ concentrations.
4.5.5: Methyl orange (MO) reduction measurements

83
The MO decoloration assay [208] was used to measure the MO reduction abilities of S.
oneidensis strains. The preculture was the same as the iron reduction measurements. Cells after
being cultured under either red light or dark condition were washed and inoculated into sealed
serum bottles containing 25 mL anaerobic minimal medium with 100 mg/L MO as the electron
acceptor to an OD600nm of about 0.1. The samples were cultured at 30 oC, 200 rpm under either red
light or dark condition. Absorbances of the samples from different culturing times were measured
at 465 nm with a plate reader (Infinite 200 PRO, Tecan). A standard curve of absorbances of freshly
made MO with different concentrations was used to determine the MO concentrations of the
samples.
4.5.6: Transparent-bottom bioreactor construction
Bioreactors construction was performed as our previous work [196]. Simply, planar
commercial indium tin oxide (ITO) coated glass coverslips (22 mm by 40 mm) were used as the
working electrodes (WEs) and the base of the bioreactors. Thin copper wires were electrically
connected to the WEs with silver paint and the wire-electrode connections were then strengthened
by covering with epoxy. Glass tubes (2.5 cm tall with a 20 mm and 22 mm inner and outer diameter,
respectively) were adhered overtop of the WEs with siliconized sealant as the body of the
bioreactors. Custom, PEEK plastic lids were used with the bioreactors along with custom Pt wire
counter electrodes (CEs) and 1 M KCl Ag/AgCl reference electrodes (REs). During the cell
culturing within bioreactors, the PEEK plastic lids along with CEs and REs were removed from
the bioreactors and the bioreactors were simply used as culturing vessels.
4.5.7: Cell culturing and biofilm formation within bioreactor
Cell culturing within the bioreactor was modified based on our previous work [196]. The
late log phase LB cultures (OD600nm about 1-1.5) were diluted into fresh minimal medium to an
OD600nm of about 0.01. 1 mL of the diluted culture was added to the bioreactor, which was made
by attaching a 20 mm diameter and 2.5 cm tall glass tube on an ITO coated glass coverslip. The
glass tube was sealed with a microporous membrane filter and taped to the ceiling of an incubator.

84
A portable smart projector (A5 Pro, Wowoto) was secured below the bioreactor in the incubator
and pointed up at the bottom surface of the bioreactor to shine red light with intensity of 150
μW/cm2
. The dark samples were covered with aluminum foil to prevent undesired photoactivation.
After 18 h of culturing at 30 oC in the incubator under either red light or dark condition, medium
was discarded, and the biofilms on the ITO electrodes were washed three times, for 2 min each
time, with minimal medium on a table shaker at 60 rpm to remove the planktonic cells. The
bioreactors with fresh minimal medium were then moved to anaerobic chamber for
electrochemical measurements.
4.5.8: Electrochemical activity measurements
All electrochemical measurements were performed in an anaerobic chamber (Bactron 300,
Sheldon Manufacturing, Inc.) with a 95:5 (N2/H2) atmosphere. Electrochemical measurements
were performed with sterile minimal medium as a blank before bioreactors were used for cell
culturing. After culturing and washing the biofilms, the reactor media were exchanged for anoxic
media inside the anaerobic chamber before all electrochemical measurements. Biofilm cyclic
voltammetry (CV) measurements were performed from -500 mV to 300 mV at 1 mV/sec using a
four channel Squidstat (Admiral Instruments). Three cycles were performed for the CV and only
data from the third cycle was presented in this manuscript. All potentials reported in this
manuscript are vs 1 M KCl Ag/AgCl.
4.5.9: Statistical analysis
All statistical analyses were performed by the Prism software (version 9.0; GraphPad) using twotailed unpaired t test. All data are presented as the mean ± SD. P values in all graphs were generated
with tests as indicated in figure legends and are represented as follows: *p < 0.05, **p < 0.01,
***p < 0.001, ****p < 0.0001 and ns (not significant) p > 0.05.

85
4.6: Supporting Information
Table S4.1-S4.4: Plasmids used in this study; Strains used in this study; Primers used in
this study; Recipe of S. oneidensis MR-1 minimal medium. Figure S4.1-S4.8: Characterization of
the iLight optogenetic system without containing the heme oxygenase gene ho1; The
thermodynamic model of the iLight optogenetic system; Expression of the iLight reporter in S.
oneidensis strains containing different promoters of iLight repressor; Characterization of the iLight
genetic circuits with different promoters of iLight repressor in E. coli; Characterization of the
inverted iLight genetic circuits in S. oneidensis; Introducing of two optogenetic systems iLight and
pDawn into Shewanella; Expression of outer membrane cytochrome MtrC through inverted iLight
genetic circuit in S. oneidensis; Expression of periplasm cytochrome STC through inverted iLight
genetic circuit in S. oneidensis. Table S4.1-S4.4: Plasmids used in this study; Strains used in this
study; Primers used in this study.

86
Supplementary tables and figures
Table S4.1: Plasmids used in the construction of iLight optogenetic system.
Plasmids Description Source
or
literatur
e
pLEVI(408)-ColE-iLightmsfGFP
(iLight (ho1-
))
Expression of LexA_DBD-iLight-msfGFP under J23116
promoter and mCherry under ColE promoter with a
LexA408 operator in bacteria
Addgene
#170268
[197]
pNO41 A plasmid with heme oxygenase gene ho1 Addgene
#101067
[39]
iLight-J23116 pLEVI(408)-ColE-iLight-msfGFP derivative with ho1 gene
from pNO41 added to the plasmid
This
work
iLight-J23105 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23105 promoter
This
work
iLight-J23107 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23107 promoter
This
work
iLight-J23108 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23108 promoter
This
work
iLight-J23102 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23102 promoter
This
work
iLight-J23104 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23104 promoter
This
work
iLight-J23111 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23111 promoter
This
work
iLight-J23119 iLight-J23116 derivative with expression of LexA_DBDiLight-msfGFP under J23119 promoter
This
work
pRF-PhIF Plasmid contains PhIF promoter and its cognate promoter Addgene
#49367
[209]
pRF-SrpR Plasmid contains SrpR promoter and its cognate promoter Addgene
#49372
[209]
iLight-J23102(cI repressor) iLight-J23102 derivative with expression of cI repressor
under ColE promoter with a LexA408 operator and with
expression of mCherry under pR promoter
This
work
iLight-J23102(PhIF
repressor)
iLight-J23102(cI repressor) derivative with expression of
PhIF repressor under ColE promoter with a LexA408
operator and with expression of mCherry under pPhIF
promoter
This
work
iLight-J23102(SrpR
repressor)
iLight-J23102(cI repressor) derivative with expression of
SrpR repressor under ColE promoter with a LexA408
operator and with expression of mCherry under pSrpR
promoter
This
work

87
Table S4.2: Strains used in iLight optogenetic study.
Strains Description Source or
literature
MR-1 Shewanella oneidensis MR-1 wild type
strain
This work
ΔmtrCΔomcA MR-1 derivative without genes encoding
outer membrane cytochromes MtrC and
OmpA
From Dr.
Jeffery
Gralnick’s lab
[202]
ΔstcΔfccA MR-1 derivative without genes encoding
periplasmic cytochromes STC and FccA
From Dr.
Jeffery
Gralnick’s lab
S. oneidensis iLight (ho1-
) MR-1 derivative carrying iLight (ho1-
)
plasmid
This work
S. oneidensis iLight-J23116 MR-1 derivative carrying iLight-J23116
plasmid
This work
S. oneidensis iLight-J23105 MR-1 derivative carrying iLight-J23105
plasmid
This work
S. oneidensis iLight-J23107 MR-1 derivative carrying iLight-J23107
plasmid
This work
S. oneidensis iLight-J23108 MR-1 derivative carrying iLight-J23108
plasmid
This work
S. oneidensis iLight-J23102 MR-1 derivative carrying iLight-J23102
plasmid
This work
iLight-J23108(cI repressor) iLight-J23108 derivative with expression of cI repressor
under ColE promoter with a LexA408 operator and with
expression of mCherry under pR promoter
This
work
iLight-J23102(PhIF
repressor)
(Kan+
)
iLight-J23102(PhIF repressor) derivative with antibiotic
resistance changed to Kan+
iLight-J23102(PhIF
repressor)
(no sfGFP)(Kan+
)
iLight-J23102(PhIF repressor) derivative with sfGFP deleted
and antibiotic resistance changed to Kan+
This
work
pDawn-mCherry Expression of mCherry proteins under pR promoter of
pDawn genetic circuit
[196]
pDawn-GFP pDawn-mCherry derivative with expression of GFP under
pR promoter of pDawn genetic circuit by replacing mCherry
with gfp
This
work
iLight-MtrC iLight-J23102(PhIF repressor) derivative with expression of
MtrC and mCherry under pPhIF promoter, and antibiotic
resistance changed to Kan+
This
work
iLight-STC iLight-J23102(PhIF repressor) derivative with expression of
STC and mCherry under pPhIF promoter, and antibiotic
resistance changed to Kan+
This
work

88
S. oneidensis iLight-J23104 MR-1 derivative carrying iLight-J23104
plasmid
This work
S. oneidensis iLight-J23111 MR-1 derivative carrying iLight-J23111
plasmid
This work
S. oneidensis iLight-J23119 MR-1 derivative carrying iLight-J23119
plasmid
This work
S. oneidensis iLight-J23102(PhIF
repressor)
MR-1 derivative carrying iLightJ23102(PhIF repressor) plasmid
This work
S. oneidensis iLight-J23102(cI
repressor)
MR-1 derivative carrying iLight-J23102(cI
repressor) plasmid
This work
S. oneidensis iLight-J23102(SrpR
repressor)
MR-1 derivative carrying iLightJ23102(SrpR repressor) plasmid
This work
S. oneidensis iLight-J23108(cI
repressor)
MR-1 derivative carrying iLight-J23108(cI
repressor) plasmid
This work
iLight-MtrC ΔmtrCΔomcA derivative carrying iLightMtrC plasmid
This work
ΔmtrCΔomcA iLight-J23102(PhIF
repressor)
ΔmtrCΔomcA derivative carrying iLightJ23102(PhIF repressor)(Kan+
) plasmid
This work
iLight-STC ΔstcΔfccA derivative carrying iLight-STC
plasmid
This work
ΔstcΔfccA iLight-J23102(PhIF
repressor)
ΔstcΔfccA derivative carrying iLightJ23102(PhIF repressor)(Kan+
) plasmid
This work
S. oneidensis pDawn-GFP MR-1 derivative carrying pDawn-GFP
plasmid
This work
S. oneidensis iLight-J23102(PhIF
repressor)
(no sfGFP)(Kan+
)
MR-1 derivative carrying iLightJ23102(PhIF repressor)(no sfGFP)(Kan+
)
plasmid
This work
S. oneidensis iLight-pDawn MR-1 derivative carrying both pDawn-GFP
and iLight-J23102(PhIF repressor)(no
sfGFP)(Kan+
) plasmids
This work
E. coli iLight (ho1-
) E. coli NEB stable derivative carrying
iLight (ho1-
) plasmid
This work
E. coli iLight-J23116 E. coli NEB stable derivative carrying
iLight-J23116 plasmid
This work
E. coli iLight-J23105 E. coli NEB stable derivative carrying
iLight-J23105 plasmid
This work
E. coli iLight-J23107 E. coli NEB stable derivative carrying
iLight-J23107 plasmid
This work
E. coli iLight-J23108 E. coli NEB stable derivative carrying
iLight-J23108 plasmid
This work
E. coli iLight-J23102 E. coli NEB stable derivative carrying
iLight-J23102 plasmid
This work
E. coli iLight-J23104 E. coli NEB stable derivative carrying
iLight-J23104 plasmid
This work
E. coli iLight-J23111 E. coli NEB stable derivative carrying
iLight-J23111 plasmid
This work
E. coli iLight-J23119 E. coli NEB stable derivative carrying
iLight-J23119 plasmid
This work

89
Table S4.3: Primers used in iLight optogenetic study.
Primers Description Source or
literature
ho1-iLight-F ttaattgcgttgcgcCTAGCCTTCGGAGGTGGC This work
iLight-ho1-R acctccgaaggctagGCGCAACGCAATTAATGTAAGTTAGC This work
iLight-ho1-F tgagctagccgtaaaCGGGATCTCGACGCTCGG This work
ho1-iLight-R agcgtcgagatcccgTTTACGGCTAGCTCAGCC This work
J23105-F cagtcctaggtACTATGCTAGCAAACACTATAAC This work
J23105-R agctagccgtaAATACTGATTCAGGCTATCAATATTT This work
J23107-F agccctaggtatTATGCTAGCAAACACTATAAC This work
J23107-R gagctagccgtaAATACTGATTCAGGCTATC This work
J23108-F TGAATCAGTAcTGACAGCTAG This work
J23108-R GGCTATCAATATTTGTCG This work
J23102-F CAGTCCTAGGtactgTGCTAGCAAAC This work
J23102-R AGCTAGCTGTCAATACTG This work
J23104-F CAGTCCTAGGtattgTGCTAGCAAAC This work
J23104-R AGCTAGCTGTCAATACTG This work
J23111-F cctaggtatagTGCTAGCAAACACTATAAC This work
J23111-R actgagctagccGTCAATACTGATTCAGGC This work
J23119-F CAGTCCTAGGTATAATGCTAGCAAACAC This work
J23119-R AGCTAGCTGTCAATACTG This work
cI-iLight-F aaagaggattttataATGAGCACAAAAAAGAAAC This work
iLight-cI-R cttttttgtgctcatTATAAAATCCTCTTTGACTTTTAAAACAAT
AAG
This work
iLight-cI-F gaactatatccggatGCCTATGCAGCGACAAATATTG This work
cI-iLight-R tgtcgctgcataggcATCCGGATATAGTTCCTC This work
PhiF-iLight-F aaagaggattttataATGGCACGTACCCCGAGC This work
iLight-PhiF-R cggggtacgtgccatTATAAAATCCTCTTTGACTTTTAAAAC This work
iLight-PhiF-F gaggagaaatactagATGGTGAGTAAAGGCGAG This work
PhiF-iLight-R gcctttactcaccatCTAGTATTTCTCCTCTTTCTCTAGTAACC
TTAACG
This work
SrpR-iLight-F aaagaggattttataATGGCACGTAAAACCGCAG This work
iLight-SrpR-R ggttttacgtgccatTATAAAATCCTCTTTGACTTTTAAAAC This work
iLight-SrpR-F gaggagaaatactagATGGTGAGTAAAGGCGAG This work
PhiF-SrpR-R gcctttactcaccatCTAGTATTTCTCCTCTTTCTCTAGTAGTT
TAC
This work
Kan-iLight-F acgaattgttagacaTTAGAAAAACTCATCGAGC This work
iLight-Kan-R gatgagtttttctaaTGTCTAACAATTCGTTCAAG This work
iLight-Kan-F ttgaatatggctcatACTCTTCCTTTTTCAATATTATTG This work
Kan-iLight-R tgaaaaaggaagagtATGAGCCATATTCAACGG This work
iLight-no
GFP-F
tttgatgatctgaccTGAGAATTCCCCCTGTTTTG This work
iLight-no
GFP-R
cagggggaattctcaGGTCAGATCATCAAAGCTG This work
GFP-pDawn-F gtggtggtgctcgagTCACTTGTACAGCTCGTCC This work
pDawn-GFPR
gagctgtacaagtgaCTCGAGCACCACCACCAC This work
pDawn-GFP-F gcccttgctcaccatTCTAGTAGGTTTCCTGTGTGAGG This work

90
Table S4.4: Recipe of S. oneidensis MR-1 minimal medium.
GFP-pDawnR
aggaaacctactagaATGGTGAGCAAGGGCGAG This work
MtrC-iLight-F gaggagaaatactagATGATGAACGCACAAAAATC This work
iLight-MtrC-R ttgtgcgttcatcatCTAGTATTTCTCCTCTTTCTCTAGTAACC This work
iLight-MtrC-F aggattgacgataggtcacacaacataggtTTTTATGGTGAGTAAAGG
CGAG
This work
MtrC-iLight-R acctatgttgtgtgacctatcgtcaatcctatttcTTACATTTTCACTTTAG
TGTGATC
This work
STC-iLight-F gaggagaaatactagGTGAGCAAAAAACTATTAAGTG This work
iLight-STC-R tagttttttgctcacCTAGTATTTCTCCTCTTTCTCTAGTAAC This work
iLight-STC-F aggattgacgataggtcacacaacataggtTTTTATGGTGAGTAAAGG
CGAG
This work
STC-iLight-R acctatgttgtgtgacctatcgtcaatcctatttcTTACTTCTTCAGAACA
GAC
This work
Salts Concentration
(g/L)
PIPES buffer 15.1
Sodium hydroxide 3.4
Ammonium chloride 1.5
Potassium chloride 0.1
Sodium phosphate monobasic monohydrate 0.6
60% (w/w) sodium lactate 3.36
Solutions Concentration
(mL/L)
1000× mineral solution 1
1000× vitamin solution 1
100× amino acid solution 10
1000× mineral solution Concentration
(g/L)
Nitrilotriacetic acid 20
Magnesium sulfate heptahydrate 30
Manganese sulfate monohydrate 5
Sodium chloride 10
Ferrous sulfate heptahydrate 1
Calcium chloride dihydrate 1
Cobalt(II) chloride hexahydrate 1
Zinc chloride 1.3
Copper sulfate pentahydrate 0.1
Aluminum potassium sulfate dodecahydrate 0.1
Boric acid 0.1
Sodium molybdate dihydrate 0.25
Nickel chloride hexahydrate 0.24

91
Sodium tungstate dihydrate 0.25
1000× vitamin solution Concentration
(g/L)
Biotin 0.02
Folic acid 0.02
Pyridoxine hydrochloride 0.1
Riboflavin 0.05
Thiamine hydrochloride 0.05
Nicotinic acid 0.05
D-pantothenic acid hemicalcium salt 0.05
Vitamin B12 0.001
P-aminobenzoic acid 0.05
Thioctic acid 0.05
100× amino acid solution Concentration
(g/L)
L-glutamic acid 2
L-arginine 2
DL-serine 2

92
Figure S4.1:Characterization of the iLight optogenetic system in S. oneidensis and E. coli
without containing the heme oxygenase gene ho1 in the plasmid.
p = 0.9093 for S. oneidensis Light vs S. oneidensis Dark and p = 0.1021 for E. coli Light vs E. coli
Dark (two-tailed unpaired t test). The measurements (mean ± SD) were derived from triplicate
experiments. Significance is indicated as ns (not significant) p > 0.05.

93
4.6.1: Thermodynamic model
This model is based on an equilibrium thermodynamic description of transcription factor
binding within the promoter region of a target gene to regulate transcription. The expression levels
of the regulated gene are presumed to be directly proportional to the probability of the promoter
being occupied by an RNA polymerase (RNAP) molecule [211]. In this context, the iLight
repressor modulates the probability of RNAP binding. The iLight repressor converts between a
dimeric and tetrameric form, depending on the presence of red light. Figure S4.2A illustrates all
possible states of the promoter in the presence of iLight repressor molecules under both red light
and dark conditions.
Here RD and RT represent the number of iLight repressors in dimeric and tetrameric forms1
,
respectively. In the light the iLight repressor is in the tetrameric form, and in the dark iLight
repressor is in the dimeric form. The model includes binding energies for transcription factors and
RNAP to specific DNA sequences within the promoter region.
Here ∆ 𝑇 𝑎𝑛𝑑 ∆𝐷 denote the difference in binding energy of the iLight tetramers and
dimers to their specific binding site and a nonspecific site. ∆𝑃 denotes the difference in binding
energy of RNAP molecule to specific and nonspecific binding sites. In the model, the iLight
binding domain consists of two halves, with each half capable of binding a single dimer and both
halves together binding a single tetramer.
The probability of RNAP binding (Pbound) was calculated using the Boltzmann weights over
all possible states in presence of RNAP (P) and iLight repressor (R) molecules in dark and red
light separately. Pbound in red light can be written as equation S1.
𝑃𝑏𝑜𝑢𝑛𝑑|𝐿𝑖𝑔ℎ𝑡
𝑃
𝑁𝑁𝑆
𝑒
−𝑃
1+
𝑃
𝑁𝑁𝑆
𝑒
−𝑃+
𝑅𝑇
𝑁𝑁𝑆
𝑒
−𝑇
(S4.1)
in which P represents the number of RNAP molecules, 𝑁𝑁𝑆 represents the number of
nonspecific sited on DNA, 𝑅𝑇 indicates the number of repressors in tetrameric form. Similarly,
the Pbound in dark can be written as equation S2.
𝑃𝑏𝑜𝑢𝑛𝑑|𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠
𝑃
𝑁𝑁𝑆
𝑒
−𝑃
1+
𝑃
𝑁𝑁𝑆
𝑒
−𝑃+2
𝑅𝐷
𝑁𝑁𝑆
𝑒
−𝐷 +(
𝑅𝐷
𝑁𝑁𝑆
)
2
𝑒
−𝐷
(S4.2)

94
Here 𝑅𝐷 represents the number of repressors in dimeric form. Parameter 𝑃𝑏𝑜𝑢𝑛𝑑 vs. total repress
or for the values of 𝑁𝑁𝑆 = 5 ∗ 106
, P=1500, 𝑃=-5𝐾𝐵𝑇, 𝐷=-7𝐾𝐵𝑇, 𝑇=-10𝐾𝐵𝑇 is shown in
Fig. S2B.
As seen in Figure S4.2B, the difference in probability of RNAP bound to the promoter in
red light and dark is maximized for intermediate concentrations of iLight repressor. High
expression levels of iLight repressor result in near zero binding probability, which in turn lead to
full repression of the reporter gene in both red light and dark conditions. Low concentrations of
iLight repressor leads to a high probability of RNAP occupying the promoter, which results in
similar expression of the target gene in both red light and dark conditions.
The fold change in expression of the regulated reporter gene can be written as the ra
tio 𝑃𝑏𝑜𝑢𝑛𝑑(𝑃,𝑅)
𝑃𝑏𝑜𝑢𝑛𝑑(𝑃,𝑅 = 0)
. Assuming that the value of 𝑃
𝑁𝑁𝑆
𝑒
−𝑃 ~0.0445 is negligible in the above eq
uations, the FC in expression of mCherry in red light and dark can be approximated as
equations S4.3 and S4.4, respectively.
𝐹𝐶|𝐿𝑖𝑔ℎ𝑡~
1
1+
𝑅𝑇
𝑁𝑁𝑆
𝑒
−𝑇
(S4.3)
𝐹𝐶|𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠~
1
(1+
𝑅𝐷
𝑁𝑁𝑆
𝑒
−𝐷 )
2
(S4.4)
The total number of repressor protein subunits is the sum of repressors in the monomeric, dimeric,
and tetrameric form denoted as R = Rm + 2RD + 4RT. As the tetrameric form is highly favored in
light conditions and the dimeric form is favored in dark conditions, it is assumed that in light RT
≈ R/4 and in dark RD ≈ R/2. The impact of light on the change in gene expression can be evaluated
by taking the ratio of 𝐹𝐶|𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠
𝐹𝐶|𝐿𝑖𝑔ℎ𝑡
, which can be written as:
𝐹𝐶 𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠
𝐿𝑖𝑔ℎ𝑡
=
𝐹𝐶|𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠
𝐹𝐶|𝐿𝑖𝑔ℎ𝑡
=
1+
𝑅
4𝑁𝑁𝑆
𝑒
−𝑇
(1+
𝑅
2𝑁𝑁𝑆
𝑒
−𝐷 )
2
(S4.5)
Fig. S2C represents the 𝐹𝐶|𝐷𝑎𝑟𝑘𝑛𝑒𝑠𝑠
𝐹𝐶|𝐿𝑖𝑔ℎ𝑡
vs. total repressor. As observed in the Figure S4.2C,
when the tetramer has a more favorable binding energy than the dimer (𝑇 − 𝐷 < 0), the ratio
of fold change in expression in the dark vs. red light is maximized for an intermediate expression

95
level of iLight repressor. Conversely, for both low and high expression levels of iLight repressor,
corresponding to near zero repression and complete repression of the reporter gene respectively,
this ratio is minimized. These findings align with experimental observations, in which the
differential expression of the target gene in light vs. dark conditions was largest for intermediate
concentrations of the iLight repressor protein.
Figure S4.2: The thermodynamic model of the iLight optogenetic system.
A) Illustration all possible states of the promoter in red light and darkness. B) Probability of the
RNAP molecules binding to the promoter vs. the total repressor. C) The ratio of fold change in the
expression of RFP in darkness vs. light.

96
Figure S4.3: Expression of the iLight reporter in S. oneidensis strains containing different
promoters of iLight repressor under red light and dark conditions, respectively.
p = 0.7462 for J23116 Light vs Dark, p = 0.3522 for J23105 Light vs Dark, p = 0.0063 for J23107
Light vs Dark, p < 0.0001 for J23108 Light vs Dark, p = 0.0006 for J23102 Light vs Dark, p =
0.8900 for J23104 Light vs Dark, p = 0.0789 for J23111 Light vs Dark and p = 0.3684 for J231119
Light vs Dark (two-tailed unpaired t test). The measurements (mean ± SD) were derived from
triplicate experiments. Significance is indicated as ****p < 0.0001, ***p < 0.001, **p < 0.01 and
ns (not significant) p > 0.05.

97
Figure S4.4:Characterization of the iLight genetic circuits with different promoters of
iLight repressor in E. coli.
A) Expression of the iLight repressor in E. coli from these promoters, as measured via expression
of co-transcribed sfGFP. (B) Expression of the iLight reporter in E. coli strains containing different
promoters of iLight repressor under red light and dark conditions, respectively. p = 0.0006 for
J23116 Light vs Dark, p < 0.0001 for J23105 Light vs Dark, p < 0.0001 for J23107 Light vs Dark,
p = 0.0006 for J23108 Light vs Dark, p = 0.4684 for J23102 Light vs Dark, p = 0.9503 for J23104
Light vs Dark, p = 0.0435 for J23111 Light vs Dark and p = 0.0013 for J23119 Light vs Dark (twotailed unpaired t test). (C) Foldchanges of the iLight reporter (dark/red light) in E. coli strains
containing different promoters of iLight repressor. The measurements (mean ± SD) were derived
from triplicate experiments. Significance is indicated as ****p < 0.0001, ***p < 0.001, **p < 0.01,
*p < 0.05 and ns (not significant) p > 0.05.

98
Figure S4.5:Characterization of the inverted iLight genetic circuits in S. oneidensis.
Expression of the iLight reporter in S. oneidensis strains containing the inverted iLight
genetic circuit
A) iLight-J23102(cI repressor), B) iLight-J23102(PhIF repressor), C) iLight-J23102(SrpR
repressor), and D) iLight-J23108(cI repressor) under red light and dark conditions, respectively. p
< 0.0001 for J23102(cI repressor) Light vs Dark, J23102(PhIF repressor) Light vs Dark and
J23102(SrpR repressor) Light vs Dark (two-tailed unpaired t test), p = 0.0011 for J23108(cI
repressor) Light vs Dark (two-tailed unpaired t test). The measurements (mean ± SD) were derived
from triplicate experiments. Significance is indicated as ****p < 0.0001 and **p < 0.01.

99
Figure S4 6: Introducing of two optogenetic systems iLight and pDawn into Shewanella.
These measurements utilized a version of iLight that does not have the GFP reported for the
expression of the iLight repressor (Table S1 and S2). We also changed the antibiotic resistance of
iLight genetic circuit from spectinomycin to kanamycin to make pDawn and iLight genetic circuits
compatible (Table S1 and S2). Shewanella cells containing both iLight and pDawn genetic circuits
had strong red fluorescence when cultured under red light illumination and had strong green
fluorescence when cultured under blue light illumination. Both the red and green fluorescence
under dark condition were very weak. Some interferences were observed between these two light
sensors, since the fluorescence of reporters of the cells with two light sensors was weaker than the
fluorescence of same reporters of the cells with single light sensor after being cultured under red

100
light or blue light. A) Microscope observation of RFP and GFP reporters for Shewanella cells with
both iLight and pDawn genetic circuits cultured under red light, blue light, and dark conditions,
respectively. B) RFP fluorescence intensity measurements of iLight reporter expression in
Shewanella strains with iLight genetic circuit and both iLight and pDawn genetic circuits
cultured under red light and dark conditions, respectively. p < 0.0001 for iLight Red Light vs iLight
Dark and iLight-pDawn Red Light vs iLight-pDawn Dark (two-tailed unpaired t test), p = 0.0002
for iLight Red Light vs iLight-pDawn Red Light (two-tailed unpaired t test). (C) GFP fluorescence
intensity measurements of pDawn reporter expression in Shewanella strains with pDawn genetic
circuit and both pDawn and iLight genetic circuits under blue light and dark conditions,
respectively. p < 0.0001 for pDawn Blue Light vs pDawn Dark, p = 0.0003 for iLight-pDawn Blue
Light vs iLight-pDawn Dark and p = 0.0060 for pDawn Blue light vs iLight-pDawn Blue Light
(two-tailed unpaired t test). Scale bars 10 μm. The measurements (mean ± SD) were derived from
triplicate experiments. Significance is indicated as ****p < 0.0001, ***p < 0.001, and **p < 0.01.

101
Figure S4.7: Expression of outer membrane cytochromes MtrC through inverted iLight
genetic circuit in S. oneidensis mutant ΔmtrCΔomcA.
A) Genetic circuit of iLight-MtrC. (B) RFP fluorescence intensity measurements of MtrC
expression of iLight-MtrC strain cultured under red light and dark conditions, respectively. p <
0.0001 for Light vs Dark (two-tailed unpaired t test). The measurements (mean ± SD) were derived
from triplicate experiments. Significance is indicated as ****p < 0.0001.

102
Figure S4.8: Expression of periplasm cytochrome STC through inverted iLight genetic
circuit in S. oneidensis mutant ΔstcΔfccA.
A) RFP fluorescence intensity measurements of STC expression of the iLight-STC strain cultured
under red light and dark conditions, respectively. p < 0.0001 for Light vs Dark (two-tailed unpaired
t test). B) Iron reduction assay for iLight-STC strain, ΔstcΔfccA and wild type with blank plasmid
iLight-J23102(PhIF repressor) after being cultured under red light and dark conditions,
respectively. The measurements (mean ± SD) were derived from triplicate experiments.
Significance is indicated as ****p < 0.0001.

103
Chapter 5: Concluding Remarks
In this final chapter, I will describe the impact of my work and discuss possible directions
for future research.
5.1: Impact of my work
5.1.1: Introducing and quantifying the concept of phenotypic memory in QS
Our study introduces the concept of phenotypic memory within quorum sensing (QS)
systems and suggests a novel metric for its quantification. Phenotypic memory refers to how past
exposure states influence future cellular responses as a result of molecular carry-over. The metric
we propose, which measures the reduction in critical cell density required for QS activation, serves
as a quantifiable indicator of this memory effect. This phenomenon, where historical cell states
inform future QS responses, aligns with prior experimental observations. Our theoretical
framework not only corroborates these findings but also delineates logical bounds for the memory
effect, offering deeper insights into its mechanisms and implications. To model QS systems, we
employed ordinary differential equations (ODEs) which generally involve many parameters that
are not easily measurable in new systems. We addressed this challenge by introducing an analytical
approach that, while simpler than more complex models, reduces the number of unknown
parameters, thereby maintaining strong predictive capabilities. This approach simplifies analyses
and broadens the exploration of QS memory across various bacterial species and environmental
conditions.
5.1.2: Potential for future experimental verification of phenotypic memory in QS
Building on the findings from Chapter 3, there is significant scope for further experimental
validation of phenotypic memory in quorum sensing (QS). Previous studies have demonstrated
how the carry-over of receptor proteins can accelerate QS reactivation in response to high signal
concentrations [17]. Future experiments should expand on these observations by testing if synthase
proteins similarly influence QS dynamics. By designing genetic circuits that can independently
adjust the synthesis and degradation rates of these proteins, researchers can more precisely analyze
their roles in QS memory. We recommend using microfluidic chambers to simulate natural

104
environmental fluctuations, such as periodic changes in signal concentrations and cell densities.
This approach allows for tight control over experimental conditions, providing deeper insights into
how QS systems adapt over time. Our findings demonstrated that systems with a high Fold Change,
low activation threshold, and high basal levels of signal synthesis exhibit the strongest memory
effects, and these results could be experimentally validated. Additionally, exploring the role of cell
division and protein degradation will help define the temporal limits of phenotypic memory.
Confirming these dynamics experimentally could validate our theoretical predictions and enhance
our understanding of bacterial communication under variable conditions, opening pathways to
innovative microbial management strategies.
5.1.3: Enhancing therapeutic approaches through QS phenotypic memory
Understanding the nuances of phenotypic memory in quorum sensing (QS) is crucial for
developing effective therapeutic strategies that target bacterial survival mechanisms. For example,
bacteria can disperse from biofilms in high-density environments to colonize new, confined spaces,
still behaving as if they are in dense colonies. This ability is particularly significant in pathogens
like Pseudomonas aeruginosa, which is involved in chronic infections and needs to quickly adapt
to new environments, such as the human lungs has potential. This adaptation potentially showcases
stronger phenotypic memory, allowing these bacteria to behave as if still in dense colonies, which
can expedite colonization and affect survival strategies. By understanding QS memory, we can
more precisely control the timing and intensity of bacterial responses, including the production of
virulence factors, thus impacting their ability to evade or overpower host defenses.
QS inhibitors generally aim to block receptor interactions or accelerate the degradation of
signaling molecules. The carry-over effects from previous high-density states can influence the
timing and strategy of these therapeutic interventions, potentially making QS inhibitors more or
less effective depending on the bacterial memory. This understanding could lead to the design of
genetic circuits that predict and modulate bacterial behavior more effectively, providing a more
targeted approach to treatment that considers both the memory and adaptive capabilities of
bacterial populations. These strategies are essential not only for creating more effective QS
inhibitors but also for devising genetic modifications that control bacterial behavior predictively
in dynamic environments.

105
5.1.4: Influence of quorum sensing phenotypic memory on ecological dynamics
Quorum sensing (QS) phenotypic memory potentially influences ecological dynamics,
particularly in fluctuating environments. Bacteria with the ability to remember previous
environmental conditions can adjust their behavior based on past experiences, giving them a
survival advantage. This memory allows them to anticipate and respond more effectively to
recurring stressors or resource availability, leading to more robust population dynamics
Additionally, QS phenotypic memory can significantly impact interspecies interactions. Bacteria
equipped with this capability to process and react based on historical environmental data may
outperform those lacking such memory, especially under rapidly shifting conditions. This adaptive
advantage allows them to stabilize their communities and dominate ecological niches by
optimizing resource utilization and adapting swiftly to environmental changes, ultimately boosting
their ecological fitness. Our understanding of phenotypic memory in QS will enhance our
comprehension of bacterial adaptation and survival strategies in ecological contexts characterized
by fluctuating environments.
5.1.4: Role of QS Phenotypic memory in synthetic biology
Designing genetic circuits with variable history dependence is critical for applications in
industrial and synthetic biology. One effective strategy includes implementing mechanisms that
manage the system's response dynamics. Adding LAA fast-degrading tags to proteins like LuxR is
a prime example, as it ensures rapid protein degradation to prevent the buildup of high basal levels,
which can occur due to plasmid-induced overexpression. Such mechanisms are essential to avoid
undesired strong memory effects, which can result in erratic bistability. Understanding and
adjusting cellular parameters that affect these dynamics is key for maintaining bistability and
managing history dependence. This approach allows researchers to precisely calibrate genetic
circuits, enhancing control over bacterial behavior and increasing the reliability and predictability
of synthetic biological systems. By fine-tuning degradation rates and expression levels, it's
possible to design circuits that either minimize or harness memory effects based on specific needs,
facilitating tailored bacterial responses in various applications.

106
Chapter 4 illustrates the red-light-induced genetic system as a pioneering development in
synthetic biology and microbial bioelectronics for controlling electron transfer in S. oneidensis
MR-1. This system introduces a precise, adjustable, and scalable method that holds vast potential
for applications in synthetic biology, biotechnology, and environmental science.
5.1.5: Transfer of iLight optogenetic system in S. oneidensis enhances precision in
microbial bioelectronics
This study introduces a novel red-light-responsive optogenetic system that significantly
advances gene expression and extracellular electron transfer (EET) control in S. oneidensis.
Traditional gene regulation methods, often dependent on chemical inducers or variable
environmental conditions, lack the dynamic control required for precise applications. Utilizing
red light as a trigger offers a non-invasive, reversible, and finely tunable approach to gene
regulation, enhancing the capability and efficiency of microbial bioelectronics systems. This
development is pivotal in enabling the precise adjustment of genetic circuits crucial for
optimizing bioelectronic device performance.
5.1.6: Transfer of iLight optogenetic system in S. oneidensis advances synthetic biology
toolkits
The integration of the iLight system in S. oneidensis MR-1 expands the synthetic biology
toolkit available for this electrogenic bacterium. Optogenetic systems have been widely used in
various organisms, but their application in S. oneidensis MR-1 is novel. This work demonstrates
the feasibility of using light to regulate complex biological processes in this bacterium, paving the
way for future synthetic biology applications. The ability to control EET with light can be extended
to other microbial hosts, potentially leading to the development of new biotechnological
applications such as biosensors, bioenergy production, and bioremediation.
5.1.7: Enhancing gene regulation in synthetic Biology: the role of the iLight
thermodynamic model
My contribution to the project involved developing a thermodynamic model that refines
the control of gene expression using the iLight optogenetic system. This model evaluates the

107
dynamics of transcription factor binding within promoter regions, determining RNA polymerase
activity under varied light conditions. Such detailed understanding allows for the precise
calibration of gene expression to optimal levels, ensuring the effectiveness of the genetic switch.
Additionally, the model’s utility extends to adapting the iLight system for various hosts; its
accuracy and predictive power improve as more experimental parameters are integrated. This
capability allows for precise modifications in gene expression dynamics, broadening the system's
applications in synthetic biology and microbial engineering.
5.1.7: Environmental and industrial applications
The findings from this study have significant implications for environmental and industrial
applications. S. oneidensis MR-1 is known for its ability to reduce metal ions and form conductive
biofilms, making it a valuable tool for bioremediation of contaminated environments and
bioenergy production. The optogenetic control of EET enables the development of more efficient
and responsive systems for these applications. For instance, in bioremediation, the ability to
precisely control the reduction of contaminants can enhance the effectiveness and safety of the
process. In bioenergy production, optimizing EET can lead to more efficient microbial fuel cells,
contributing to the development of sustainable energy solutions.
5.2: Closing remarks
This dissertation highlights the indispensable role of mathematical modeling and
computational methods in deepening our understanding of microbial behavior and enhancing their
practical applications. These tools not only shed light on complex biological dynamics but also
offer predictive insights and shape the design and interpretation of experimental studies. The
synergy between theoretical models and empirical validation holds immense potential for
advancing our understanding and manipulation of complex biological processes, paving the way
for innovative applications across various scientific and industrial domains.

108
References
1. Ross-Gillespie, A. and R. Kümmerli, Collective decision-making in microbes. Frontiers in
microbiology, 2014. 5: p. 54-54.
2. Gao, M., et al., A crucial role for spatial distribution in bacterial quorum sensing.
Scientific reports, 2016. 6: p. 34695-34695.
3. Darch, S.E., et al., Spatial determinants of quorum signaling in a Pseudomonas aeruginosa
infection model. Proceedings of the National Academy of Sciences of the United States of
America, 2018. 115(18): p. 4779-4784.
4. Yusufaly, T.I. and J.Q. Boedicker, Spatial dispersal of bacterial colonies induces a
dynamical transition from local to global quorum sensing. Physical Review E, 2016. 94(6):
p. 062410.
5. Waters, C.M. and B.L. Bassler, Quorum sensing: cell-to-cell communication in bacteria.
Annu. Rev. Cell Dev. Biol., 2005. 21: p. 319-346.
6. Hastings, J.W. and E.P. Greenberg, Quorum sensing: the explanation of a curious
phenomenon reveals a common characteristic of bacteria. Journal of bacteriology, 1999.
181(9): p. 2667.
7. Schuster, M., et al., Acyl-Homoserine Lactone Quorum Sensing: From Evolution to
Application. Annual Review of Microbiology, 2013. 67(1): p. 43-63.
8. Williams, J.W., et al., Robust and sensitive control of a quorum‐sensing circuit by two
interlocked feedback loops. Molecular systems biology, 2008. 4(1): p. 234.
9. Dockery, J.D. and J.P. Keener, A mathematical model for quorum sensing in Pseudomonas
aeruginosa. Bulletin of mathematical biology, 2001. 63(1): p. 95-116.

109
10. Weber, M. and J. Buceta, Dynamics of the quorum sensing switch: stochastic and nonstationary effects. BMC systems biology, 2013. 7: p. 6-6.
11. Kim, M.K., et al., Local and global consequences of flow on bacterial quorum sensing.
Nature microbiology, 2016. 1: p. 15005-15005.
12. Mukherjee, S. and B.L. Bassler, Bacterial quorum sensing in complex and dynamically
changing environments. Nature Reviews Microbiology, 2019. 17(6): p. 371-382.
13. Williams, P. and M. Camara, Quorum sensing and environmental adaptation in
Pseudomonas aeruginosa: a tale of regulatory networks and multifunctional signal
molecules. Curr Opin Microbiol, 2009. 12(2): p. 182-91.
14. Whitehead, N.A., et al., Quorum-sensing in Gram-negative bacteria. FEMS microbiology
reviews, 2001. 25(4): p. 365-404.
15. Rescan, M., et al., Phenotypic memory drives population growth and extinction risk in a
noisy environment. Nature ecology & evolution, 2020. 4(2): p. 193-201.
16. Jablonka, E., et al., The adaptive advantage of phenotypic memory in changing
environments. Philosophical Transactions of the Royal Society of London. Series B:
Biological Sciences, 1995. 350(1332): p. 133-141.
17. Sappington, K.J., et al., Reversible signal binding by the Pseudomonas aeruginosa
quorum-sensing signal receptor LasR. MBio, 2011. 2(1): p. 10.1128/mbio. 00011-11.
18. Emge, P., et al., Resilience of bacterial quorum sensing against fluid flow. Scientific
reports, 2016. 6: p. 33115-33115.
19. Kirisits, M.J., et al., Influence of the hydrodynamic environment on quorum sensing in
Pseudomonas aeruginosa biofilms. Journal of bacteriology, 2007. 189(22): p. 8357-8360.

110
20. Ruparell, A., et al., The fitness burden imposed by synthesising quorum sensing signals.
Scientific Reports, 2016. 6(1): p. 33101.
21. Bernhardt, J.R., et al., Life in fluctuating environments. Philosophical Transactions of the
Royal Society B, 2020. 375(1814): p. 20190454.
22. Vermeersch, L., et al., Do microbes have a memory? History-dependent behavior in the
adaptation to variable environments. Frontiers in Microbiology, 2022. 13: p. 1004488.
23. Kussell, E. and S. Leibler, Phenotypic diversity, population growth, and information in
fluctuating environments. Science, 2005. 309(5743): p. 2075-2078.
24. Botero, C.A., et al., Evolutionary tipping points in the capacity to adapt to environmental
change. Proceedings of the National Academy of Sciences, 2015. 112(1): p. 184-189.
25. Ogura, M., et al., Delayed bet-hedging resilience strategies under environmental
fluctuations. Physical Review E, 2017. 95(5): p. 052404.
26. Fritz, G., N. Walker, and U. Gerland, Heterogeneous timing of gene induction as a
regulation strategy. Journal of molecular biology, 2019. 431(23): p. 4760-4774.
27. Lindner, F. and A. Diepold, Optogenetics in bacteria–applications and opportunities.
FEMS Microbiology Reviews, 2022. 46(2): p. fuab055.
28. Zhao, E.M., et al., Optogenetic regulation of engineered cellular metabolism for microbial
chemical production. Nature, 2018. 555(7698): p. 683-687.
29. Tandar, S.T., et al., Optogenetic switch for controlling the central metabolic flux of
Escherichia coli. Metabolic engineering, 2019. 55: p. 68-75.

111
30. Cheng, X., et al., Engineering Gac/Rsm signaling cascade for optogenetic induction of the
pathogenicity switch in Pseudomonas aeruginosa. ACS Synthetic Biology, 2021. 10(6): p.
1520-1530.
31. Jin, X. and I.H. Riedel-Kruse, Biofilm Lithography enables high-resolution cell patterning
via optogenetic adhesin expression. Proceedings of the National Academy of Sciences,
2018. 115(14): p. 3698-3703.
32. Moser, F., et al., Light‐controlled, high‐resolution patterning of living engineered bacteria
onto textiles, ceramics, and plastic. Advanced Functional Materials, 2019. 29(30): p.
1901788.
33. Myers, C.R. and K.H. Nealson, Bacterial manganese reduction and growth with
manganese oxide as the sole electron acceptor. Science, 1988. 240(4857): p. 1319-1321.
34. Shi, L., et al., Extracellular electron transfer mechanisms between microorganisms and
minerals. Nature Reviews Microbiology, 2016. 14(10): p. 651-662.
35. West, E.A., A. Jain, and J.A. Gralnick, Engineering a native inducible expression system
in Shewanella oneidensis to control extracellular electron transfer. ACS Synthetic
Biology, 2017. 6(9): p. 1627-1634.
36. Levskaya, A., et al., Engineering Escherichia coli to see light. Nature, 2005. 438(7067): p.
441-442.
37. Ohlendorf, R., et al., From dusk till dawn: one-plasmid systems for light-regulated gene
expression. Journal of molecular biology, 2012. 416(4): p. 534-542.
38. Tabor, J.J., A. Levskaya, and C.A. Voigt, Multichromatic control of gene expression in
Escherichia coli. Journal of molecular biology, 2011. 405(2): p. 315-324.

112
39. Ong, N.T., E.J. Olson, and J.J. Tabor, Engineering an E. coli near-infrared light sensor.
ACS synthetic biology, 2018. 7(1): p. 240-248.
40. Fernandez-Rodriguez, J., et al., Engineering RGB color vision into Escherichia coli. Nature
chemical biology, 2017. 13(7): p. 706-708.
41. Multamaki, E., et al., Optogenetic control of bacterial expression by red light. ACS
Synthetic Biology, 2022. 11(10): p. 3354-3367.
42. Stevens, A.M. and E. Greenberg, Quorum sensing in Vibrio fischeri: essential elements for
activation of the luminescence genes. Journal of bacteriology, 1997. 179(2): p. 557-562.
43. Ulrich, L.E., E.V. Koonin, and I.B. Zhulin, One-component systems dominate signal
transduction in prokaryotes. Trends in microbiology, 2005. 13(2): p. 52-56.
44. Mitrophanov, A.Y. and E.A. Groisman, Signal integration in bacterial two-component
regulatory systems. Genes & development, 2008. 22(19): p. 2601-2611.
45. Galperin, M.Y., A census of membrane-bound and intracellular signal transduction
proteins in bacteria: Bacterial IQ, extroverts and introverts. BMC Microbiology, 2005.
5(1): p. 35.
46. Waters, C.M. and B.L. Bassler, QUORUM SENSING: Cell-to-Cell Communication in
Bacteria. Annual Review of Cell and Developmental Biology, 2005. 21(1): p. 319-346.
47. Boedicker, J. and K. Nealson, Microbial Communication via Quorum Sensing. IEEE
Transactions on Molecular, Biological and Multi-Scale Communications, 2015. 1(4): p.
310-320.
48. Miller, M.B. and B.L. Bassler, Quorum Sensing in Bacteria. Annual Review of
Microbiology, 2001. 55(1): p. 165-199.

113
49. Hastings, J.W. and E.P. Greenberg, Quorum sensing: the explanation of a curious
phenomenon reveals a common characteristic of bacteria. Journal of bacteriology, 1999.
181(9): p. 2667-2668.
50. West, S.A., et al., Quorum sensing and the confusion about diffusion. Trends Microbiol,
2012. 20(12): p. 586-94.
51. Nealson, K.H., T. Platt, and J.W. Hastings, Cellular Control of the Synthesis and Activity
of the Bacterial Luminescent System. Journal of Bacteriology, 1970. 104(1): p. 313.
52. Ng, W.-L. and B.L. Bassler, Bacterial quorum-sensing network architectures. Annual
review of genetics, 2009. 43: p. 197-222.
53. Popat, R., et al., Collective sensing and collective responses in quorum-sensing bacteria. J
R Soc Interface, 2015. 12(103).
54. Moreno-Gámez, S., et al., Quorum sensing integrates environmental cues, cell density and
cell history to control bacterial competence. Nature Communications, 2017. 8(1): p. 854.
55. Bai A, J. and R. Rai Vittal, Quorum Sensing Regulation and Inhibition of Exoenzyme
Production and Biofilm Formation in the Food Spoilage Bacteria Pseudomonas
psychrophila PSPF19. Food Biotechnology, 2014. 28(4): p. 293-308.
56. Barnard, A.M.L., et al., Quorum sensing, virulence and secondary metabolite production
in plant soft-rotting bacteria. Philosophical transactions of the Royal Society of London.
Series B, Biological sciences, 2007. 362(1483): p. 1165-1183.
57. Rajput, A., K. Kaur, and M. Kumar, SigMol: repertoire of quorum sensing signaling
molecules in prokaryotes. Nucleic Acids Research, 2015. 44(D1): p. D634-D639.

114
58. Hense, B.A., et al., Does efficiency sensing unify diffusion and quorum sensing? Nature
Reviews Microbiology, 2007. 5(3): p. 230-239.
59. Boedicker, J.Q., M.E. Vincent, and R.F. Ismagilov, Microfluidic Confinement of Single
Cells of Bacteria in Small Volumes Initiates High-Density Behavior of Quorum Sensing
and Growth and Reveals Its Variability. Angewandte Chemie International Edition, 2009.
48(32): p. 5908-5911.
60. Carnes, E.C., et al., Confinement-induced quorum sensing of individual Staphylococcus
aureus bacteria. Nature Chemical Biology, 2010. 6(1): p. 41-45.
61. Qazi, S.N., et al., agr expression precedes escape of internalized Staphylococcus aureus
from the host endosome. Infect Immun, 2001. 69(11): p. 7074-82.
62. Cornforth, D.M., et al., Combinatorial quorum sensing allows bacteria to resolve their
social and physical environment. Proceedings of the National Academy of Sciences, 2014.
111(11): p. 4280.
63. Shapiro, J.A., Bacteria are small but not stupid: cognition, natural genetic engineering
and socio-bacteriology. Studies in History and Philosophy of Science Part C: Studies in
History and Philosophy of Biological and Biomedical Sciences, 2007. 38(4): p. 807-819.
64. Dworkin, M., Prokaryotic Life Cycles, in The Prokaryotes: Volume 2: Ecophysiology and
Biochemistry, M. Dworkin, et al., Editors. 2006, Springer New York: New York, NY. p.
140-166.
65. Pompano, R.R., et al., Conceptual and Experimental Tools to Understand Spatial Effects
and Transport Phenomena in Nonlinear Biochemical Networks Illustrated with Patchy
Switching. Annual Review of Biochemistry, 2017. 86(1): p. 333-356.

115
66. Be'er, A., et al., Deadly competition between sibling bacterial colonies. Proceedings of the
National Academy of Sciences of the United States of America, 2009. 106(2): p. 428-433.
67. Chacón, J.M., W. Möbius, and W.R. Harcombe, The spatial and metabolic basis of colony
size variation. The ISME Journal, 2018. 12(3): p. 669-680.
68. Cao, Y., et al., Collective Space-Sensing Coordinates Pattern Scaling in Engineered
Bacteria. Cell, 2016. 165(3): p. 620-630.
69. Rampioni, G., L. Leoni, and P. Williams, The art of antibacterial warfare: Deception
through interference with quorum sensing–mediated communication. Bioorganic
Chemistry, 2014. 55: p. 60-68.
70. Dong, Y.H., et al., AiiA, an enzyme that inactivates the acylhomoserine lactone quorumsensing signal and attenuates the virulence of Erwinia carotovora. Proceedings of the
National Academy of Sciences of the United States of America, 2000. 97(7): p. 3526-3531.
71. Michael, B., et al., SdiA of Salmonella enterica is a LuxR homolog that detects mixed
microbial communities. Journal of bacteriology, 2001. 183(19): p. 5733-5742.
72. Riedel, K., et al., N-acylhomoserine-lactone-mediated communication between
Pseudomonas aeruginosa and Burkholderia cepacia in mixed biofilms. Microbiology,
2001. 147(Pt 12): p. 3249-62.
73. McClean, K.H., et al., Quorum sensing and Chromobacterium violaceum: exploitation of
violacein production and inhibition for the detection of N-acylhomoserine lactones.
Microbiology, 1997. 143 ( Pt 12): p. 3703-11.
74. Silva, K.P.T., P. Chellamuthu, and J.Q. Boedicker, Quantifying the strength of quorum
sensing crosstalk within microbial communities. PLoS computational biology, 2017.
13(10): p. e1005809-e1005809.

116
75. Wellington, S. and E.P. Greenberg, Quorum Sensing Signal Selectivity and the Potential
for Interspecies Cross Talk. mBio, 2019. 10(2): p. e00146-19.
76. Wu, F., D.J. Menn, and X. Wang, Quorum-sensing crosstalk-driven synthetic circuits:
from unimodality to trimodality. Chem Biol, 2014. 21(12): p. 1629-38.
77. Pérez, P.D., J.T. Weiss, and S.J. Hagen, Noise and crosstalk in two quorum-sensing inputs
of Vibrio fischeri. BMC Systems Biology, 2011. 5(1): p. 153.
78. Eldar, A., Social conflict drives the evolutionary divergence of quorum sensing.
Proceedings of the National Academy of Sciences, 2011. 108(33): p. 13635.
79. Schluter, J., et al., The Evolution of Quorum Sensing as a Mechanism to Infer Kinship.
PLOS Computational Biology, 2016. 12(4): p. e1004848.
80. Miller, E.L., et al., Eavesdropping and crosstalk between secreted quorum sensing peptide
signals that regulate bacteriocin production in Streptococcus pneumoniae. The ISME
journal, 2018. 12(10): p. 2363-2375.
81. Carballo-Pacheco, M., et al., Receptor crosstalk improves concentration sensing of
multiple ligands. Physical Review E, 2019. 99(2): p. 022423.
82. Hosni, T., et al., Sharing of quorum-sensing signals and role of interspecies communities
in a bacterial plant disease. The ISME Journal, 2011. 5(12): p. 1857-1870.
83. Silva, K.P.T. and J.Q. Boedicker, A neural network model predicts community-level
signaling states in a diverse microbial community. PLoS computational biology, 2019.
15(6): p. e1007166-e1007166.

117
84. Hopfield, J.J., Neural networks and physical systems with emergent collective
computational abilities. Proceedings of the National Academy of Sciences, 1982. 79(8): p.
2554.
85. Ansaldi, M., et al., Specific activation of the Bacillus quorum-sensing systems by
isoprenylated pheromone variants. Mol Microbiol, 2002. 44(6): p. 1561-73.
86. Yusufaly, T.I. and J.Q. Boedicker, Mapping quorum sensing onto neural networks to
understand collective decision making in heterogeneous microbial communities. Physical
Biology, 2017. 14(4): p. 046002.
87. Yan, J., et al., Bow-tie signaling in c-di-GMP: Machine learning in a simple biochemical
network. PLoS computational biology, 2017. 13(8): p. e1005677-e1005677.
88. Williamson, L.L., et al., Intracellular screen to identify metagenomic clones that induce or
inhibit a quorum-sensing biosensor. Applied and environmental microbiology, 2005.
71(10): p. 6335-6344.
89. Schaefer, A.L., et al., LuxR- and luxI-type quorum-sensing circuits are prevalent in
members of the Populus deltoides microbiome. Applied and environmental microbiology,
2013. 79(18): p. 5745-5752.
90. Pinchas, M.D., N.C. LaCross, and S. Dawid, An electrostatic interaction between BlpC
and BlpH dictates pheromone specificity in the control of bacteriocin production and
immunity in Streptococcus pneumoniae. Journal of bacteriology, 2015. 197(7): p. 1236-
1248.
91. Lyon, G.J., et al., Key Determinants of Receptor Activation in the agr Autoinducing
Peptides of Staphylococcus aureus. Biochemistry, 2002. 41(31): p. 10095-10104.

118
92. Aleta, A. and Y. Moreno, Multilayer Networks in a Nutshell. Annual Review of Condensed
Matter Physics, 2019. 10(1): p. 45-62.
93. Henke, J.M. and B.L. Bassler, Three parallel quorum-sensing systems regulate gene
expression in Vibrio harveyi. Journal of bacteriology, 2004. 186(20): p. 6902-6914.
94. Zitnik, M. and J. Leskovec, Predicting multicellular function through multi-layer tissue
networks. Bioinformatics, 2017. 33(14): p. i190-i198.
95. Shinde, P. and S. Jalan, A multilayer protein-protein interaction network analysis of
different life stages in Caenorhabditis elegans. EPL (Europhysics Letters), 2015. 112(5):
p. 58001.
96. Rai, A., et al., Understanding cancer complexome using networks, spectral graph theory
and multilayer framework. Scientific reports, 2017. 7: p. 41676-41676.
97. De Domenico, M., S. Sasai, and A. Arenas, Mapping Multiplex Hubs in Human Functional
Brain Networks. Frontiers in neuroscience, 2016. 10: p. 326-326.
98. De Domenico, M., Multilayer modeling and analysis of human brain networks.
GigaScience, 2017. 6(5).
99. Bak, How Nature Works: The Science of Self-Organized Criticality 1996, Copernicus, New
York.
100. Berg, H.C. and E.M. Purcell, Physics of chemoreception. Biophysical journal, 1977. 20(2):
p. 193-219.
101. Mehta, P. and D.J. Schwab, Energetic costs of cellular computation. Proceedings of the
National Academy of Sciences, 2012. 109(44): p. 17978.

119
102. Ruparell, A., et al., The fitness burden imposed by synthesising quorum sensing signals.
Scientific Reports, 2016. 6(1): p. 33101.
103. Schuster, S., et al., Cooperation and cheating in microbial exoenzyme production –
Theoretical analysis for biotechnological applications. Biotechnology Journal, 2010. 5(7):
p. 751-758.
104. Schaefer, A.L., et al., A LuxR Homolog in a Cottonwood Tree Endophyte That Activates
Gene Expression in Response to a Plant Signal or Specific Peptides. mBio, 2016. 7(4): p.
e01101-16.
105. Whiteley, M., S.P. Diggle, and E.P. Greenberg, Progress in and promise of bacterial
quorum sensing research. Nature, 2017. 551(7680): p. 313-320.
106. Drescher, K., et al., Solutions to the public goods dilemma in bacterial biofilms. Current
biology : CB, 2014. 24(1): p. 50-55.
107. Pai, A., Y. Tanouchi, and L. You, Optimality and robustness in quorum sensing (QS)-
mediated regulation of a costly public good enzyme. Proceedings of the National Academy
of Sciences, 2012. 109(48): p. 19810.
108. Waltermann, C. and E. Klipp, Information theory based approaches to cellular signaling.
Biochimica et Biophysica Acta (BBA) - General Subjects, 2011. 1810(10): p. 924-932.
109. Mousavian, Z., J. Díaz, and A. Masoudi-Nejad, Information theory in systems biology. Part
II: protein–protein interaction and signaling networks. Seminars in Cell & Developmental
Biology, 2016. 51: p. 14-23.
110. Suderman, R. and E.J. Deeds, Intrinsic limits of information transmission in biochemical
signalling motifs. Interface focus, 2018. 8(6): p. 20180039-20180039.

120
111. Chu, D., M. Prokopenko, and J.C.J. Ray, Computation by natural systems. Interface Focus,
2018. 8(6): p. 20180058.
112. Rodrigo, G. and J.F. Poyatos, Genetic Redundancies Enhance Information Transfer in
Noisy Regulatory Circuits. PLOS Computational Biology, 2016. 12(10): p. e1005156.
113. Mehta, P., et al., Information processing and signal integration in bacterial quorum
sensing. Molecular systems biology, 2009. 5: p. 325-325.
114. Friedman, J., L.M. Higgins, and J. Gore, Community structure follows simple assembly
rules in microbial microcosms. Nature Ecology & Evolution, 2017. 1(5): p. 0109.
115. Sumner ER, A.S., Phenotypic heterogeneity: differential stress resistance among
individual cells of the yeast Saccharomyces cerevisiae. 2002: Microbiology. p. 345-351.
116. Long, T., et al., Quantifying the integration of quorum-sensing signals with single-cell
resolution. PLoS biology, 2009. 7(3): p. e68-e68.
117. Anetzberger, C., T. Pirch, and K. Jung, Heterogeneity in quorum sensing-regulated
bioluminescence of Vibrio harveyi. Molecular Microbiology, 2009. 73(2): p. 267-277.
118. Pérez, P.D. and S.J. Hagen, Heterogeneous response to a quorum-sensing signal in the
luminescence of individual Vibrio fischeri. PloS one, 2010. 5(11): p. e15473-e15473.
119. Garmyn, D., et al., Evidence of autoinduction heterogeneity via expression of the Agr
system of Listeria monocytogenes at the single-cell level. Applied and environmental
microbiology, 2011. 77(17): p. 6286-6289.
120. Grote, J., et al., Evidence of autoinducer-dependent and -independent heterogeneous gene
expression in Sinorhizobium fredii NGR234. Appl Environ Microbiol, 2014. 80(18): p.
5572-82.

121
121. Feltner, J.B., et al., LasR Variant Cystic Fibrosis Isolates Reveal an Adaptable QuorumSensing Hierarchy in Pseudomonas aeruginosa. mBio, 2016. 7(5): p. e01513-16.
122. Guo, X., K.P.T. Silva, and J.Q. Boedicker, Single-cell variability of growth interactions
within a two-species bacterial community. Physical Biology, 2019. 16(3): p. 036001.
123. Guo, X. and J.Q. Boedicker, The Contribution of High-Order Metabolic Interactions to the
Global Activity of a Four-Species Microbial Community. PLoS computational biology,
2016. 12(9): p. e1005079-e1005079.
124. Razo-Mejia, M. and R. Phillips, First-principles prediction of the information processing
capacity of a simple genetic circuit. bioRxiv, 2019: p. 594325.
125. Larkin, J.W., et al., Signal Percolation within a Bacterial Community. Cell systems, 2018.
7(2): p. 137-145.e3.
126. Silva, K.P.T., et al., Disruption of microbial communication yields a two-dimensional
percolation transition. Physical Review E, 2019. 99(4): p. 042409.
127. Ratzke, C. and J. Gore, Self-organized patchiness facilitates survival in a cooperatively
growing Bacillus subtilis population. Nature Microbiology, 2016. 1(5): p. 16022.
128. Kim, H.J., et al., Defined spatial structure stabilizes a synthetic multispecies bacterial
community. Proceedings of the National Academy of Sciences, 2008. 105(47): p. 18188.
129. Nadell, C.D., K. Drescher, and K.R. Foster, Spatial structure, cooperation and competition
in biofilms. Nature Reviews Microbiology, 2016. 14(9): p. 589-600.
130. Borer, B., R. Tecon, and D. Or, Spatial organization of bacterial populations in response
to oxygen and carbon counter-gradients in pore networks. Nature communications, 2018.
9(1): p. 769-769.

122
131. Ebrahimi, A., J. Schwartzman, and O.X. Cordero, Cooperation and spatial selforganization determine rate and efficiency of particulate organic matter degradation in
marine bacteria. Proceedings of the National Academy of Sciences, 2019. 116(46): p.
23309.
132. Momeni, B., A.J. Waite, and W. Shou, Spatial self-organization favors heterotypic
cooperation over cheating. eLife, 2013. 2: p. e00960-e00960.
133. Haagensen, J.A.J., et al., Development of Spatial Distribution Patterns by Biofilm Cells.
Applied and Environmental Microbiology, 2015. 81(18): p. 6120.
134. Chandrasekaran, K. and B. Bose, Percolation in a reduced equilibrium model of planar
cell polarity. Physical Review E, 2019. 100(3): p. 032408.
135. Essam, J.W., Percolation theory. Reports on Progress in Physics, 1980. 43(7): p. 833-912.
136. Chen, A.Y., et al., Synthesis and patterning of tunable multiscale materials with engineered
cells. Nature materials, 2014. 13(5): p. 515-523.
137. Hong, S.H., et al., Synthetic quorum-sensing circuit to control consortial biofilm formation
and dispersal in a microfluidic device. Nature Communications, 2012. 3(1): p. 613.
138. Brenner, K. and F.H. Arnold, Self-organization, layered structure, and aggregation
enhance persistence of a synthetic biofilm consortium. PloS one, 2011. 6(2): p. e16791-
e16791.
139. Choudhary, S. and C. Schmidt-Dannert, Applications of quorum sensing in biotechnology.
Applied Microbiology and Biotechnology, 2010. 86(5): p. 1267-1279.
140. Jiang, Q., et al., Quorum Sensing: A Prospective Therapeutic Target for Bacterial
Diseases. BioMed Research International, 2019. 2019: p. 2015978.

123
141. Hsu, R.H., et al., Microbial Interaction Network Inference in Microfluidic Droplets. Cell
systems, 2019. 9(3): p. 229-242.e4.
142. Ziesack, M., et al., Engineered Interspecies Amino Acid Cross-Feeding Increases
Population Evenness in a Synthetic Bacterial Consortium. mSystems, 2019. 4(4): p.
e00352-19.
143. Goldford, J.E., et al., Emergent simplicity in microbial community assembly. Science (New
York, N.Y.), 2018. 361(6401): p. 469-474.
144. Wolf, D.M., et al., Memory in microbes: quantifying history-dependent behavior in a
bacterium. PLOS one, 2008. 3(2): p. e1700.
145. Lambert, G. and E. Kussell, Memory and fitness optimization of bacteria under fluctuating
environments. PLoS genetics, 2014. 10(9): p. e1004556.
146. Lee, C.K., et al., Multigenerational memory and adaptive adhesion in early bacterial
biofilm communities. Proceedings of the National Academy of Sciences, 2018. 115(17): p.
4471-4476.
147. Mutlu, A., et al., Phenotypic memory in Bacillus subtilis links dormancy entry and exit by
a spore quantity-quality tradeoff. Nature Communications, 2018. 9(1): p. 69.
148. Chu, E.K., A. Groisman, and A. Levchenko, Environmental sensing in dynamic quorum
responses. bioRxiv, 2019: p. 745091.
149. Pedraza, J.M. and J. Paulsson, Effects of molecular memory and bursting on fluctuations
in gene expression. Science, 2008. 319(5861): p. 339-343.

124
150. Mathis, R. and M. Ackermann, Response of single bacterial cells to stress gives rise to
complex history dependence at the population level. Proceedings of the National Academy
of Sciences, 2016. 113(15): p. 4224-4229.
151. Casadesús, J. and R. D'Ari, Memory in bacteria and phage. Bioessays, 2002. 24(6): p. 512-
518.
152. Gokhale, C.S., S. Giaimo, and P. Remigi, Memory shapes microbial populations. PLoS
Computational Biology, 2021. 17(10): p. e1009431.
153. Chellaboina, V., et al., Modeling and analysis of mass-action kinetics. IEEE Control
Systems Magazine, 2009. 29(4): p. 60-78.
154. Melke, P., et al., A cell-based model for quorum sensing in heterogeneous bacterial
colonies. PLoS computational biology, 2010. 6(6): p. e1000819.
155. Buchler, N.E., U. Gerland, and T. Hwa, Nonlinear protein degradation and the function of
genetic circuits. Proceedings of the National Academy of Sciences, 2005. 102(27): p. 9559-
9564.
156. Laalami, S., L. Zig, and H. Putzer, Initiation of mRNA decay in bacteria. Cellular and
molecular life sciences, 2014. 71: p. 1799-1828.
157. Urbanowski, M., C. Lostroh, and E. Greenberg, Reversible acyl-homoserine lactone
binding to purified Vibrio fischeri LuxR protein. Journal of bacteriology, 2004. 186(3): p.
631-637.
158. Kaplan, H.B. and E. Greenberg, Diffusion of autoinducer is involved in regulation of the
Vibrio fischeri luminescence system. Journal of bacteriology, 1985. 163(3): p. 1210-1214.

125
159. Goryachev, A., D. Toh, and T. Lee, Systems analysis of a quorum sensing network: design
constraints imposed by the functional requirements, network topology and kinetic
constants. Biosystems, 2006. 83(2-3): p. 178-187.
160. Weber, M. and J. Buceta, Noise regulation by quorum sensing in low mRNA copy number
systems. BMC systems biology, 2011. 5: p. 1-13.
161. Milo, R. and R. Phillips, Cell biology by the numbers. 2015: Garland Science.
162. Boada, Y., A. Vignoni, and J. Picó, Promoter and transcription factor dynamics tune
protein mean and noise strength in a quorum sensing-based feedback synthetic circuit.
bioRxiv, 2017: p. 106229.
163. Pai, A. and L. You, Optimal tuning of bacterial sensing potential. Molecular systems
biology, 2009. 5(1): p. 286.
164. Egland, K.A. and E. Greenberg, Quorum sensing in Vibrio fischeri: elements of the luxI
promoter. Molecular microbiology, 1999. 31(4): p. 1197-1204.
165. Lupp, C., et al., The Vibrio fischeri quorum‐sensing systems ain and lux sequentially induce
luminescence gene expression and are important for persistence in the squid host.
Molecular microbiology, 2003. 50(1): p. 319-331.
166. Nasser, W. and S. Reverchon, New insights into the regulatory mechanisms of the LuxR
family of quorum sensing regulators. Analytical and bioanalytical chemistry, 2007. 387: p.
381-390.
167. Teng, S.-W., et al., Measurement of the copy number of the master quorum-sensing
regulator of a bacterial cell. Biophysical journal, 2010. 98(9): p. 2024-2031.

126
168. Hintsche, M. and S. Klumpp, Dilution and the theoretical description of growth-rate
dependent gene expression. Journal of biological engineering, 2013. 7(1): p. 1-3.
169. Pai, A., et al., Generic metric to quantify quorum sensing activation dynamics. ACS
synthetic biology, 2014. 3(4): p. 220-227.
170. Pai, A., Y. Tanouchi, and L. You, Optimality and robustness in quorum sensing (QS)-
mediated regulation of a costly public good enzyme. Proceedings of the National Academy
of Sciences, 2012. 109(48): p. 19810-19815.
171. Gangan, M.S., et al., Intertemporal trade-off between population growth rate and carrying
capacity during public good production. Iscience, 2022. 25(4).
172. Sood, V. and J.H. Brickner, Genetic and epigenetic strategies potentiate Gal4 activation
to enhance fitness in recently diverged yeast species. Current Biology, 2017. 27(23): p.
3591-3602. e3.
173. Chen, F., et al., Quorum quenching enzymes and their application in degrading signal
molecules to block quorum sensing-dependent infection. International journal of molecular
sciences, 2013. 14(9): p. 17477-17500.
174. Wellington, S. and E.P. Greenberg, Quorum sensing signal selectivity and the potential for
interspecies cross talk. MBio, 2019. 10(2): p. 10.1128/mbio. 00146-19.
175. Köhler, T., et al., Quorum sensing-dependent virulence during Pseudomonas aeruginosa
colonisation and pneumonia in mechanically ventilated patients. Thorax, 2010. 65(8): p.
703-710.
176. Pesci, E.C., et al., Regulation of las and rhl quorum sensing in Pseudomonas aeruginosa.
Journal of bacteriology, 1997. 179(10): p. 3127-3132.

127
177. Wei, G. and J.Q. Yang, Microfluidic investigation of the impacts of flow fluctuations on
the development of Pseudomonas putida biofilms. npj Biofilms and Microbiomes, 2023.
9(1): p. 73.
178. Grote, J., D. Krysciak, and W.R. Streit, Phenotypic Heterogeneity, a Phenomenon That
May Explain Why Quorum Sensing Does Not Always Result in Truly Homogenous Cell
Behavior. Applied and Environmental Microbiology, 2015. 81(16): p. 5280.
179. Rattray, J.B., et al., Bacterial quorum sensing allows graded and bimodal cellular
responses to variations in population density. MBio, 2022. 13(3): p. e00745-22.
180. Castillo-Hair, S.M., et al., Optogenetic control of Bacillus subtilis gene expression. Nature
communications, 2019. 10(1): p. 3099.
181. Pu, L., et al., Optogenetics manipulation enables prevention of biofilm formation of
engineered Pseudomonas aeruginosa on surfaces. ACS synthetic biology, 2018. 7(1): p.
200-208.
182. McMillan, D.G., et al., Menaquinone-7 is specific cofactor in tetraheme quinol
dehydrogenase CymA. Journal of Biological Chemistry, 2012. 287(17): p. 14215-14225.
183. Edwards, M.J., et al., Structural modeling of an outer membrane electron conduit from a
metal-reducing bacterium suggests electron transfer via periplasmic redox partners.
Journal of Biological Chemistry, 2018. 293(21): p. 8103-8112.
184. Edwards, M.J., et al., The crystal structure of a biological insulated transmembrane
molecular wire. Cell, 2020. 181(3): p. 665-673. e10.
185. Fonseca, B.M., et al., Mind the gap: cytochrome interactions reveal electron pathways
across the periplasm of Shewanella oneidensis MR-1. Biochemical Journal, 2013. 449(1):
p. 101-108.

128
186. Hartshorne, R.S., et al., Characterization of an electron conduit between bacteria and the
extracellular environment. Proceedings of the National Academy of Sciences, 2009.
106(52): p. 22169-22174.
187. White, G.F., et al., Rapid electron exchange between surface-exposed bacterial
cytochromes and Fe (III) minerals. Proceedings of the National Academy of Sciences,
2013. 110(16): p. 6346-6351.
188. Cao, Y., et al., CRISPRi–sRNA: transcriptional–translational regulation of extracellular
electron transfer in Shewanella oneidensis. ACS synthetic biology, 2017. 6(9): p. 1679-
1690.
189. Dundas, C.M., D.J. Walker, and B.K. Keitz, Tuning extracellular electron transfer by
Shewanella oneidensis using transcriptional logic gates. ACS synthetic biology, 2020.
9(9): p. 2301-2315.
190. Cao, Y., et al., A synthetic plasmid toolkit for Shewanella oneidensis MR-1. Frontiers in
Microbiology, 2019. 10: p. 410.
191. Jensen, H.M., et al., Engineering of a synthetic electron conduit in living cells. Proceedings
of the National Academy of Sciences, 2010. 107(45): p. 19213-19218.
192. Jensen, H.M., et al., CymA and exogenous flavins improve extracellular electron transfer
and couple it to cell growth in Mtr-expressing Escherichia coli. ACS synthetic biology,
2016. 5(7): p. 679-688.
193. Zacharoff, L.A. and M.Y. El-Naggar, Redox conduction in biofilms: From respiration to
living electronics. Current Opinion in Electrochemistry, 2017. 4(1): p. 182-189.
194. Atkinson, J.T., et al., Living electronics: A catalogue of engineered living electronic
components. Microbial Biotechnology, 2023. 16(3): p. 507-533.

129
195. Xu, S., et al., Multiheme cytochrome mediated redox conduction through Shewanella
oneidensis MR-1 cells. Journal of the American Chemical Society, 2018. 140(32): p.
10085-10089.
196. Zhao, F., et al., Light-induced patterning of electroactive bacterial biofilms. ACS synthetic
biology, 2022. 11(7): p. 2327-2338.
197. Kaberniuk, A.A., et al., Single-component near-infrared optogenetic systems for gene
transcription regulation. Nature communications, 2021. 12(1): p. 3859.
198. Kreamer, N.N., et al., Predicting the impact of promoter variability on regulatory outputs.
Scientific Reports, 2015. 5(1): p. 18238.
199. Berg, O.G. and P.H. von Hippel, Selection of DNA binding sites by regulatory proteins:
Statistical-mechanical theory and application to operators and promoters. Journal of
molecular biology, 1987. 193(4): p. 723-743.
200. Guharajan, S., et al., Quantifying the regulatory role of individual transcription factors in
Escherichia coli. Cell reports, 2021. 37(6).
201. Kelly, J.R., et al., Measuring the activity of BioBrick promoters using an in vivo reference
standard. Journal of biological engineering, 2009. 3: p. 1-13.
202. Coursolle, D. and J.A. Gralnick, Modularity of the Mtr respiratory pathway of Shewanella
oneidensis strain MR‐1. Molecular microbiology, 2010. 77(4): p. 995-1008.
203. Heidelberg, J.F., et al., Genome sequence of the dissimilatory metal ion–reducing
bacterium Shewanella oneidensis. Nature biotechnology, 2002. 20(11): p. 1118-1123.

130
204. Burns, J.L., et al., Outer membrane-associated serine protease involved in adhesion of
Shewanella oneidensis to Fe (III) oxides. Environmental science & technology, 2010.
44(1): p. 68-73.
205. Gralnick, J.A., et al., Extracellular respiration of dimethyl sulfoxide by Shewanella
oneidensis strain MR-1. Proceedings of the National Academy of Sciences, 2006. 103(12):
p. 4669-4674.
206. Luan, F., et al., Bioreduction of nitrobenzene, natural organic matter, and hematite by
Shewanella putrefaciens CN32. Environmental science & technology, 2010. 44(1): p. 184-
190.
207. Burns, J.L. and T.J. DiChristina, Anaerobic respiration of elemental sulfur and thiosulfate
by Shewanella oneidensis MR-1 requires psrA, a homolog of the phsA gene of Salmonella
enterica serovar typhimurium LT2. Applied and environmental microbiology, 2009.
75(16): p. 5209-5217.
208. Cai, P.-J., et al., Anaerobic biodecolorization mechanism of methyl orange by Shewanella
oneidensis MR-1. Applied microbiology and biotechnology, 2012. 93: p. 1769-1776.
209. Stanton, B.C., et al., Genomic mining of prokaryotic repressors for orthogonal logic gates.
Nature chemical biology, 2014. 10(2): p. 99-105.
210. Lalwani, M.A., et al., Optogenetic control of the lac operon for bacterial chemical and
protein production. Nature chemical biology, 2021. 17(1): p. 71-79.
211. Bintu, L., et al., Transcriptional regulation by the numbers: models. Current opinion in
genetics & development, 2005. 15(2): p. 116-124.

Abstract (if available)

Abstract This thesis explores two critical dimensions of bacterial behavior: the existence of phenotypic memory in quorum sensing (QS) and the implementation of an optogenetic control system in Shewanella oneidensis. The study introduces a metric to quantify phenotypic memory in QS, defined by the reduced critical cell density required for QS reactivation in cells previously in QS active state. This reflects the influence of historical cell states on current bacterial responses. Mathematical models and simulations not only demonstrate that bacteria previously in QS-ON states can activate at lower densities but also identify several key cellular parameters that significantly influence this memory effect.
Additionally, this work details the successful adaptation of an optogenetic system in S. oneidensis to control gene expression critical for electron transfer processes, pivotal for microbial bioelectronics. My specific contribution was developing a thermodynamic model that predicts optimal gene expression levels, providing a deeper understanding of the light-controlled regulatory mechanisms in this system.
These investigations not only advance our comprehension of microbial communication and control but also faciliates more effective management of bacterial populations and the development of therapeutic strategies in medical and environmental applications. The findings emphasize the potential of leveraging historical behavioral patterns and precise genetic control in synthetic biology, enhancing both theoretical knowledge and practical applications.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

From cables to biofilms: electronic and electrochemical characterization of electroactive microbial systems

PDF

Quorum sensing interaction networks in bacterial communities

PDF

Bacterial nanowires of Shewanella oneidensis MR-1: electron transport mechanism, composition, and role of multiheme cytochromes

PDF

From fuel cells to single cells: electrochemical measurements of direct electron transfer at microbial-electrode interfaces

PDF

Entanglement parity effects in quantum spin chains

PDF

Electrochemical studies of outward and inward extracellular electron transfer by microorganisms from diverse environments

PDF

Designing data-effective machine learning pipeline in application to physics and material science

PDF

Electronic, electrochemical, and spintronic characterization of bacterial electron transport

PDF

Semantic structure in understanding and generation of the 3D world

PDF

Kinetic Monte Carlo simulations for electron transport in redox proteins: from single cytochromes to redox networks

PDF

Microbial electron transfer and biogenic nanomaterials: experiments, theory, and simulations at the bacterial-inorganic interface

PDF

Coulomb interactions and superconductivity in low dimensional materials

PDF

Microbial interaction networks: from single cells to collective behavior

PDF

Topological protection of quantum coherence in a dissipative, disordered environment

PDF

Disentangling the ecology of bacterial communities in cnidarian holobionts

PDF

Deterministic evolutionary game theory models for the nonlinear dynamics and control of chemotherapeutic resistance

PDF

Survival and evolution of Shewanella oneidensis MR-1: applications for microbial fuel cells

PDF

Signatures of topology in a quasicrystal: a case study of the non-interacting and superconducting Fibonacci chain

PDF

Packaging and delivery of plasmids by bacterial extracellular vesicles

PDF

Integrative mathematical and computational approaches to investigate metastatic cancers

Asset Metadata

Creator Ostovar, Ghazaleh (author)

Core Title Mathematical modeling in bacterial communication and optogenetic systems

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Physics

Degree Conferral Date 2024-12

Publication Date 12/06/2024

Defense Date 09/24/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag gene expression regulation,mathematical modeling,microbial bioelectronics,microbial communication,OAI-PMH Harvest,optogenetic control,phenotypic memory,quorum sensing (QS),Shewanella oneidensis,thermodynamic modeling

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Boedicker, James (committee chair), Di Felice, Rosa (committee member), El-Naggar, Moh (committee member), Haas, Stephan (committee member), Nakano, Aiichiro (committee member)

Creator Email ostovar@usc.edu

Unique identifier UC11399EHFR

Identifier etd-OstovarGha-13677.pdf (filename)

Legacy Identifier etd-OstovarGha-13677

Document Type Dissertation

Format theses (aat)

Rights Ostovar, Ghazaleh

Internet Media Type application/pdf

Type texts

Source 20241210-usctheses-batch-1227 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu