Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Spatial and temporal precision of inhibitory and excitatory neurons in the murine dorsal lateral geniculate nucleus
(USC Thesis Other)
Spatial and temporal precision of inhibitory and excitatory neurons in the murine dorsal lateral geniculate nucleus
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Copyright 2022 Alexis Sara Gorin
Spatial and temporal precision of inhibitory and excitatory neurons in the murine dorsal lateral
geniculate nucleus
by
Alexis Sara Gorin
A Dissertation Presented to the
FACULTY OF THE USC NEUROSCIENCE GRADUATE PROGRAM
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)
December 2022
ii
ACKNOWLEDGEMENTS
I would like to first thank my advisor, Dr. Judith Hirsch, for her guidance and support over
the years. Working in her laboratory has been a transformative experience, both at a professional
and personal level. I would also like to thank my co-advisor Dr. Fritz Sommer for his guidance,
support, and good wit. I am thankful towards my committee members Dr. Andrew Hires, Dr.
Bartlett Mell, Dr. Hong Wei Dong, and Dr. Bosco Tjan, as well as Dr. Radha Kalluri who graciously
stepped into the role after the sudden loss of Dr. Tjan.
I would like to express my deep appreciation to my colleagues throughout the years: Dr.
Vandana Suresh, Dr. Ulas Ciftioglu, Seohee Ahn, Yizhan Miao, as well as the undergraduates
who have hopefully enjoyed watching me stumble through teaching over the years: Eric Qu, Arjun
Kumar, Yinan Su, and Rose Meltzer. Thank you to all my friends who have either helped or
pushed me every time including, but not limited to, Jen Do, Mark Haber, Chris Im, Ahyun Jung,
Samson King, Charles Lee, Sadhna Rao, Mariana Uchoa, Christopher Ventura, and Lily Zou.
Lastly, this PhD would not be possible without the support of my father and mother, Tuvia Gorin
and Ronnie Wax.
iii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ................................................................................................................................. ii
TABLE OF CONTENTS .................................................................................................................................... iii
LIST OF FIGURES ............................................................................................................................................ v
ABBREVIATIONS ........................................................................................................................................... vi
ABSTRACT .................................................................................................................................................... vii
CHAPTER 1 – INTRODUCTION ....................................................................................................................... 1
VISUALLY-EVOKED INHIBITION IN THE DORSAL LATERAL GENICULATE NUCLEUS ................................... 1
FRAMEWORK FOR DESCRIBING INHIBITION IN THE THALAMUS .............................................................. 3
Inhibition evoked by stimuli of the non-preferred contrast: pull ......................................................... 4
Inhibition evoked by stimuli of the preferred contrast: same-sign inhibition ...................................... 9
Inhibition independent of stimulus contrast: stimulus-selective inhibition ....................................... 11
Inhibition evoked by vast versus small stimuli: global versus local inhibition .................................... 12
Inhibition independent of the stimulus: tonic inhibition .................................................................... 13
POTENTIAL ANATOMICAL CIRCUITS FOR VISUALLY-EVOKED INHIBITION IN RODENTS ......................... 13
Comparison of the early visual pathway in rodent to higher order mammals ................................... 13
Putative anatomical circuits for push-pull across species .................................................................. 15
CONCLUSION ........................................................................................................................................... 16
CHAPTER 2 – RECEPTIVE FIELDS .................................................................................................................. 17
INTRODUCTION ....................................................................................................................................... 17
MATERIALS AND METHODS .................................................................................................................... 19
Key Resources Table ........................................................................................................................... 19
Resource availability ........................................................................................................................... 20
Experimental models and subject details ........................................................................................... 20
Method details .................................................................................................................................... 20
Quantification and statistical analysis ................................................................................................ 23
RESULTS .................................................................................................................................................. 30
Optotagging interneurons and relay cells ........................................................................................... 31
Receptive field structures of interneurons are nearly as diverse as those of relay cells ................... 33
The distribution of receptive field size for interneurons and relay cells is similar ............................. 35
Stability of stimulus preference across a range of stimulus strength ................................................. 35
Reconstructing inhibition in the relay cells’ receptive fields .............................................................. 38
iv
DISCUSSION ............................................................................................................................................. 44
Receptive field size .............................................................................................................................. 44
Receptive field structure ..................................................................................................................... 45
Receptive field constancy ................................................................................................................... 46
Emergence of the suppressive component of the relay cell’s receptive field .................................... 47
SUPPLEMENTARY MATERIAL .................................................................................................................. 48
AUTHOR CONTRIBUTIONS ...................................................................................................................... 51
CHAPTER 3 – TEMPORAL PRECISION .......................................................................................................... 52
INTRODUCTION ....................................................................................................................................... 52
MATERIALS AND METHODS .................................................................................................................... 53
Experimental models and subject details ........................................................................................... 53
Method details .................................................................................................................................... 53
Quantification and statistical analysis ................................................................................................ 57
RESULTS .................................................................................................................................................. 60
Temporal precision of relay cells versus interneurons ....................................................................... 61
Variation in the relationship between contrast and reliability ........................................................... 64
Relationship between receptive field size and precision.................................................................... 66
DISCUSSION ............................................................................................................................................. 67
Retinal contributions to precision ....................................................................................................... 67
Inhibitory contributions to precision .................................................................................................. 68
Reliability and contrast ....................................................................................................................... 69
CHAPTER 4 – CONCLUSION ......................................................................................................................... 70
REFERENCES ................................................................................................................................................ 71
v
LIST OF FIGURES
Figure 1.1. Introducing receptive fields in the LGN ...................................................................................... 2
Figure 1.2. The early visual pathway ............................................................................................................. 3
Figure 1.3. Push-pull organization of excitation and inhibition in the receptive fields of a mouse ............. 4
Figure 1.4. Image encoding and subsequent thalamic processing. .............................................................. 5
Figure 1.5. The largest amount of information within a natural image is skewed toward low spatial
frequencies. .................................................................................................................................................. 7
Figure 1.6. Response properties and receptive field structure of four different types of relay cells:
Center-Surround, On-Off, W3-like, and suppressed-by-contrast like. ....................................................... 10
Figure 1.7. Potential wiring diagrams for push-pull excitation and inhibition in the
center-surround cell’s receptive fields ....................................................................................................... 15
Figure 2.1. Utilizing optogenetics in anesthetized mice to identify interneurons and relay cells
during in vivo recordings ............................................................................................................................. 32
Figure 2.2. Utilizing optogenetics in anesthetized mice to identify interneurons and relay cells
during in vivo recordings ............................................................................................................................. 33
Figure 2.3. Receptive fields of interneurons range from small to large ..................................................... 34
Figure 2.4. Receptive fields of thalamic neurons change as a function of stimulus strength .................... 36
Figure 2.5. LNP models of interneurons.. ................................................................................................... 39
Figure 2.6. Ensemble of interneurons can provide inhibition to center-surround relay cells .................... 42
Figure 2.7. Ensemble of interneurons can provide inhibition to SBC relay cells ........................................ 43
Figure 2.S1. Bright-Dark Polarity Scores and cells with prolonged inhibition to both bright and dark
stimuli during priming phase ...................................................................................................................... 49
Figure 2.S2. Dictionary of all interneuron LNP models (without spatial and temporal translation) .......... 50
Figure 3.1. Interneurons fire less likely than chance than relay cells. ........................................................ 61
Figure 3.2. As a population, interneurons fire less reliably than relay cells at most contrasts
although firing rates for both cell types are indistinguishable for all contrasts ......................................... 63
Figure 3.3. Relay cells can a non-monotonic contrast-reliability relationship ........................................... 65
Figure 3.4. Contrast-reliability responses organized by receptive field types of interneurons and
relay cells. ................................................................................................................................................... 66
Figure 3.5. Relay cell precision is larger for cells with smaller receptive fields .......................................... 66
vi
ABBREVIATIONS
ChR2 Channelrhodopsin
eNphR3.0 Halorhodopsin
EPSP Excitatory post-synaptic potential
EYFP Yellow fluorescent protein
IPSP Inhibitory post-synaptic potential
LGN Lateral geniculate nucleus
dLGN Dorsal lateral geniculate nucleus
LNP Linear-nonlinear Poisson model
PSTH Peristimulus time histogram
RGC Retinal ganglion cell
SBC Suppressed-by-contrast
stGtACR2 Soma-targeted anion channelrhodopsin
TRN Thalamic reticular nucleus (also referred to perigeniculate nucleus)
vii
ABSTRACT
In the early visual pathway, retina sends visual information to the dorsal lateral geniculate
nucleus’ (dLGN) relay cells that project to cortex. Historically, the dLGN was considered a relay station
(hence “relay” cells), one that passes on information encoded in retina to cortex unaltered. However,
complex inhibitory circuits in the visual thalamus suggest that retinal information is not simply relayed to
the primary visual cortex but is altered by inhibition with the dLGN. There are two types of inhibitory
neurons that synapse onto relay cells: cells in the thalamic reticular nucleus (TRN) and local interneurons.
Thus, inhibition influences all visual information passing through the dLGN. The research presented in this
dissertation will focus on the local interneurons, which dominate the intrinsic circuit in the dLGN and can
provide powerful inhibition to relay cells.
In Chapter 2, in the first study of its kind, I use intracellular, juxtacellular, or multielectrode
methods to record from murine dLGN during visual stimulation and use optogenetics to distinguish local
interneurons from relay cells and to characterize the functional properties of interneurons learn how they
might influence visual information sent to cortex. Both relay cells and interneurons have diverse receptive
field structures. We found that interneurons, like relay cells, have diverse types of receptive fields. The
distribution of receptive field structure, but not size, differed between the two populations. To learn how
interneurons might contribute to responses of relay cells, we turned to computational approaches. We
built simple LNP models of interneurons and used the output of these models to reconstruct suppressive
components of the relay cell’s receptive field that we extracted from recordings of the membrane current.
We were able to reconstruct these suppressive fields for a group of relay cells including those center-
surround and suppressed-by-contrast profiles using a biologically plausible number of inputs. Thus, we
provide proof of concept that convergent input from diverse types of interneurons can provide feature-
specific inhibition to relay cells.
viii
In Chapter 3, using data obtained in conjunction with Chapter 2, I investigated the temporal
precision of interneurons versus that of relay cells. We made recordings from optotagged interneurons
and relay cells in the murine dLGN and compared responses evoked by natural movies at different
contrasts. While some interneurons fired as precisely as relay cells, they were less reliable as a population
across stimulus contrasts. Firing rates were similar for both populations, however. Reliability increased
steadily with contrast for most relay cells but saturated with the step from the lowest contrast to the next
for interneurons. There was also a modest correlation between smaller receptive field size and greater
temporal precision for relay cells but not for interneurons.
This work is the first study of receptive field structure and temporal response properties of
identified local interneurons in murine dLGN and, thus, provides unique insight into how these cells
influence vision.
1
CHAPTER 1 – INTRODUCTION
VISUALLY-EVOKED INHIBITION IN THE DORSAL LATERAL GENICULATE NUCLEUS
Visual information encoded by the retina is sent to the lateral geniculate nucleus of the thalamus
(LGN) before reaching the primary visual cortex. Each cell along the visual pathway has a receptive field:
a small region of visual space that these cells are sensitive to. In the retina and LGN, the most common
receptive field is the center-surround, made up of two concentric subregions (Figure 1.1A) (Hubel and
Wiesel, 1961; Wiesel, 1959). Each subregion prefers either a change from darker to brighter, known as an
On subregion, or a change from brighter to darker, known as an Off subregion. These cells respond best
to contrast borders in a discrete area of visual space. Ultimately, this will represent a segment of an edge
in visual cortex (Figure 1.1B).
Historically, the lateral geniculate nucleus was considered to be a simple relay station; a structure
that transmits information unaltered. This simplistic view was mainly because receptive fields of cells in
the LGN are similar to those of retinal ganglion cells (RGCs) in terms of their visual response properties,
size, and the spatial organization of the On and Off subregions of the receptive field (Cleland and Lee,
1985; Hubel and Wiesel, 1961; Levick et al., 1972; Mastronarde et al., 1987; Usrey et al., 1999). However,
the presence of complex inhibitory circuits in the LGN suggest that retinal information is not simply
relayed to the primary visual cortex but is altered by inhibition within the LGN (Figure 1.2).
Two types of inhibitory neurons synapse onto relay cells that convey visual information to cortex
(Figure 1.2). The first are local interneurons, whose projections are confined within the LGN (Bickford et
al., 2010; Bickford et al., 2008; Sherman and Guillery, 2002); these neurons receive direct retinal input
(Seabrook et al., 2013; Van Horn et al., 2000) and inhibit relay cells and each other in a feedforward
manner (Ahlsen et al., 1985; Blitz and Regehr, 2005; Lam et al., 2005). The second are thalamic reticular
neurons (TRN) (Uhlrich et al., 1991) which receive input from axon collaterals of relay
2
Figure 1.1. Introducing receptive fields in the LGN. (A) Defining subregions. A subregion with preference for bright stimuli
(represented here in white) is an On subregion (represented here in red). A subregion with a preference for dark stimuli
(represented here in black) is an Off subregion (represented here in blue). (B) Hubel and Wiesel’s model for simple cell
connectivity in the primary visual cortex. A simple cell receives excitation from LGN relay cells whose receptive fields are aligned
in space, such that the simple cell sums the inputs from LGN to build an elongated receptive field responsive to bars and edges.
Adapted from Hubel and Wiesel, 1962.
cells and, in turn, provide feedback inhibition to the LGN in a topographical manner (Cucchiaro et al.,
1991; Kim et al., 1997). These two inhibitory circuits influence all visual information passing through the
LGN, adding complexity to visual processing in the LGN. The research presented in this dissertation seeks
to understand how inhibitory mechanisms in the LGN help serve vision.
Previous studies in vivo explored inhibition both, post-synaptically, by recording intracellularly
from relay cells (Martinez et al., 2014; Suresh et al., 2016; Wang et al., 2011b; Wang et al., 2007) and pre-
synaptically, by recording from local interneurons (Dubin and Cleland, 1977; Durand et al., 2016; Wang et
al., 2011b) and from TRN (Sanderson, 1971; So and Shapley, 1981; Soto-Sánchez et al., 2017; Uhlrich et
al., 1991; Vaingankar et al., 2012; Xue et al., 1988). However, the previous studies of local interneurons
3
were limited; sample sizes were small (Wang et al., 2011b) or cell type was inferred rather than
demonstrated directly based on antidromic stimulation (Dubin and Cleland, 1977) or spike shape (Durand
et al., 2016). Moreover, no one to our knowledge has attempted to disentangle the effect of local
interneurons on relay cells. Thus, we seek to elucidate our understanding of both the visual response
properties of the local interneurons and the extent to which these inhibitory neurons individually
influence relay cells. This will provide fundamental insight into the role of inhibitory processing in the
thalamus.
FRAMEWORK FOR DESCRIBING INHIBITION IN THE THALAMUS
Describing patterns of inhibition is an arduous task and few have described it before. Thus, within
this introduction, I will begin to develop a framework for describing inhibition in the thalamus. From now
Figure 1.2. The early visual pathway. Retina projects to relay cells and the local interneurons. Local interneurons synapse onto
relay cells and each other. Relay cells across species projects to layer IV and VI of cortex. TRN receives input from relay cells and
provides feedback onto relay cells.
4
on, patterns of inhibition within the receptive field shall be described by their response to specific
contrasts. Furthermore, patterns of inhibition based on selectivity for stimulus attributes such as
orientation, direction, spatial frequency tuning, contrast, etc, shall be described as well.
Inhibition evoked by stimuli of the non-preferred contrast: pull
Push-pull is found in relay cells that have receptive fields built from concentric subregions with
the opposite preference for luminance contrast, known as center-surround (Wang et al., 2007). These
neurons are excited by a bright (On) or dark (Off) stimulus presented in one subregion and inhibited when
a stimulus of the opposite contrast is presented to the same area. For example, a bright stimulus covering
the center of a receptive field of an On-center neuron excites (push) (Figure 1.3, Top-Left), while a dark
stimulus covering the same subregion inhibits (pull) (Figure 1.3, Bottom-Left). Push-pull was first found in
the retina (Wiesel, 1959), where it is often called cross-over inhibition (Werblin, 2010), but is generated
de novo in the feline thalamus (Martinez et al., 2014; Wang et al., 2011b; Wang et al., 2007).
Figure 1.3. Push-pull organization of excitation and inhibition in the receptive fields of a mouse. Anatomical reconstruction
of an On-center relay cell in mouse shown above membrane currents evoked by dark and bright discs flashed in the center (left)
and annuli flashed in the surround of the receptive field (right). Each panel represents two individual responses to the stimulus,
with averaged responses to multiple stimulus trials (bold traces) shown below at 2× gain. Icons at left represent stimuli. Black
traces represent responses to dark stimuli. Gray traces represent responses to bright stimuli. Pale gray lines indicate baseline.
Dark gray bars represent stimulus duration.
5
Push-pull plays many roles in sensory processing. It extends the dynamic range of response and
restores linearity of response lost by rectification across the synapse (Werblin, 2010). Push-pull also
enhances selectivity for the spatial position of contrast borders, as follows. First, it produces a mutually
antagonistic relationship between neighboring subregions, known as subfield antagonism (Kuffler, 1953;
Wiesel, 1959). A spot confined to the center of the preferred luminance contrast evokes excitation, push.
However, if that same stimulus expands to fill the surround, pull is recruited, weakening the excitatory
response (Wang et al., 2007). Second, at least in cat, the spatial extent of the pull is larger than that of
push, which further sharpens neural sensitivity to local differences in luminance (Martinez et al., 2014).
This has been shown to boost the response to contrast borders.
Furthermore, one must consider how push-pull can enhance contrast borders within the context
of the retinogeniculate circuit. In cat, there are more relay cells than RGCs (Hamos et al., 1985; Usrey et
Figure 1.4. Image encoding and subsequent thalamic processing. (a) Encoding: Retinal ganglion cells encode visual information.
The small discs illustrate the receptive field centers of three Off relay cells that encode part of the photo of the eye above. The
retinal representation of the image is noticeably pixelated. (b) Upsampling and interpolation: Relay cells outnumber ganglion
cells by a ratio of ∼2:1. Retinal axons diverge to innervate multiple thalamic relay cells. Each relay cell receives convergent input
from neighboring ganglion cells organized such that no two thalamic receptive fields are the same. Image of the eye shows that
upsampling improves resolution but that interpolation introduces blur. (c) Sharpening: The spatial extent of the pull (indicated
by reverse shading) in the relay cell's receptive field is larger than that of the push. This arrangement boosts contrast borders,
improving image quality (to generate the snapshot in panel c, an unsharp mask was applied to the image in panel B. LGN, lateral
geniculate nucleus.
6
al., 1999; Yeh et al., 2009). Thus, when RGCs encode information about visual space (Figure 1.4a),
information sent from retina to the LGN is upsampled (Figure 1.4b), increasing the resolution of the image
representation at the level of the LGN but introducing artifact. Diverging retinal afferents converge onto
different relay cells (Eglen et al., 2005; Madarasz et al., 1978; Martinez et al., 2014; Peters and Payne,
1993; Wassle et al., 1983), averaging out artifact introduced by upsampling, but this introduces blur
(Figure 1.4b). The mechanism by which pull sharpens the neural response to contrast borders can help
mitigate blur (Figure 1.4c). Thus, the geometry of the receptive field, coupled with interactions between
push and pull across subregions, fine-tunes sensitivity to stimulus size and location and improves overall
image perception without increasing the number of cells at the initial encoding stage.
The weakened responses to diffuse stimuli that cover both center and surround due to push-pull
also supports Barlow’s idea of efficient coding (Barlow, 1961). That is, natural images are redundant since
there exists statistical dependencies amongst pixel values in space and time; in order to make efficient
use of resources, early stages of sensory processing reduce redundancy in signals sent downstream by
maximizing the amount of information carried per spike. This is analogous to lossless compression of
photographs, which minimizes the number of bits while keeping the integrity of the image to save
resources. In nature, neighboring regions within a given field of view of an image often have similar
luminance values, leading to extensive spatial correlations and substantial redundancy. That is, natural
scenes have 1/f
α
statistics—the largest amount of information within a given image is skewed toward low
spatial frequencies representative of global information in an image such as general orientation and
proportions rather than the high spatial frequencies which contain information of the details within the
image (Figure 1.5A) (Field, 1987; Simoncelli and Olshausen, 2001). Thus, the efficient coding hypothesis
suggests transforming the image to equally represent both the global information (low spatial
frequencies) and the details (high spatial frequencies), a process known as whitening (Figure 1.5B). It has
long been recognized that the structure of the retinal and thalamic receptive fields performs this
7
transformation. These fields can be represented as a difference of two Gaussians, one matched to the
center and the other to the surround (Enroth-Cugell and Robson, 1966; Marr and Hildreth, 1980). This
filter operates on the image to form the second spatial derivative of the stimulus, enhancing edges
(Chandler and Field, 2007; Marr and Hildreth, 1980); thus, the emphasis on the higher frequency
components reduces redundancy (Figure 1.5B, C).
Push-pull operates in time as well as space. Generally, push-pull speeds temporal transitions
between dark and bright stimuli (Alitto and Usrey, 2005; Krumin et al., 2014). For example, a dark annulus
flashed in an Off subregion elicits push when the stimulus appears and induces pull when it exits (Figure
1.3, Top-right). Conversely, for a cell with a center-surround receptive field structure, introducing and
then removing a bright annulus to and from the same Off subregion evoke a sequence of pull then push
(Figure 1.3, bottom-right). This can occur when the transition is for individual subregions or both. In this
fashion, relay cells select for biphasic (dark to bright or bright to dark) stimuli, thereby forming the
Figure 1.5. The largest amount of information within a natural image is skewed toward low spatial frequencies. (A) Example
sinusoidal frequencies represented in space (top) and time (bottom). (B) Fourier representation of the spatial frequency
information within natural images along a single dimension (left) and Fourier representation of the spatial frequency
information within a whitened natural image along a single dimension, which can happen due to difference-of-Gaussians
filtering (right). (C) An image (left) with only low spatial frequency information (middle) and with only high spatial frequency
information (right).
8
temporal derivative of the image and reducing redundancy of signals dominated by low spatial
frequencies such as natural stimuli (Dan et al., 1996; Olshausen and Field, 1996; Simoncelli and Olshausen,
2001). Hence, the sequential activation of push-pull promotes efficient coding in time as well as in space.
A specific mechanism by which push-pull acts in the spatiotemporal domain is that relay cells’ fire
action potentials between two modes, tonic and burst. These two modes are driven by luminance
transitions from the nonpreferred to the preferred luminance contrast (Alitto and Usrey, 2005; Wang et
al., 2007). Relay cells typically fire tonic spikes when they are relatively depolarized; for example, in the
presence of a preferred luminance contrast. Tonic firing in relay cells typically track (an edited version of)
retinal input patterns (Carandini et al., 2007; Wang et al., 2010); in this mode it is thought that virtually
every spike is preceded by a retinogeniculate excitatory postsynaptic potential (EPSP) (Carandini et al.,
2007; Koepsell et al., 2009; Sincich et al., 2007; Wang et al., 2010). Bursts are rapid trains of action
potentials that provide a means of amplifying retinal input (Sherman, 2001). Bursts are triggered by the
presence of a prolonged, strongly suppressive stimulus, de-inactivating the T-type calcium channels
(Huguenard and McCormick, 1992), followed by an excitatory stimulus. For example, a dark disc covering
the center of an On relay cell inhibits the cell; if the dark stimulus is held and then removed (the stimulus
covering the center becomes brighter), the cell will fire a burst of spikes on top of a calcium wave upon
removal of the stimulus. Thus, bursts, induced by push-pull, signal transitions from periods of lasting
brightness or darkness.
Though bursts predominate during sleep (Steriade et al., 1993), when neuromodulators that
hyperpolarize relay cells are released, bursts can also play a role in vision. When the animal moves through
its environment or scans the terrain, luminance values can remain similar for long periods and then
abruptly change (Simoncelli and Olshausen, 2001). Thus, there will be times when the receptive field is
covered by a stimulus of the nonpreferred sign for long periods of time before abruptly changing to a
stimulus of the preferred sign (Denning and Reinagel, 2005; Lesica and Stanley, 2004; Wang et al., 2007),
9
such as when you are focusing on a person’s face during a conversation but then your attention (and
vision) is pulled towards what caused a loud, abrupt noise. In these cases, hyperpolarization due to pull
can de-inactivate T-type calcium channels such that when a stimulus contrast reverses and retinal input
resumes, a burst is fired (Wang et al., 2007). Spike timing also has bearing on the amount and content of
information that reaches cortex (Denning and Reinagel, 2005; Lesica et al., 2006). Thalamic bursts activate
the cortex more effectively than slower trains of spikes (Swadlow and Gusev, 2001; 2002), likely because
they evoked synaptic potentials that summate in time (Usrey and Reid, 2000), and, also, because they
occur after long silences that permit recovery from synaptic depression (Swadlow and Gusev, 2001; 2002).
Hence, pull within the receptive field not only suppresses firing, but can also heighten the cell’s response
to changes in stimulus polarity.
Inhibition evoked by stimuli of the preferred contrast: same-sign inhibition
Pull can be thought of as opposite-sign inhibition. Accordingly, same-sign inhibition occurs when
the same stimulus, bright or dark, presented to a given region of the receptive field, evokes excitation and
inhibition at once. Across mammalian species, there is anatomical evidence for same-sign inhibition: a
single retinal ganglion cell synapses onto the dendrites of both a relay cell and an interneuron, which
inhibits the relay cell in turn via a dendrodendritic synapse; this arrangement is known as a triad (Cox,
2014; Cox and Beatty, 2017; Sherman, 2004). Other scenarios for same-sign inhibition such as an On
interneuron contacting an On relay cells via an axonal synapse are also likely, but have yet to appear in
the literature.
Same-sign inhibition in the cat has been implicated in contrast gain control—the dynamic
rescaling of neuron responses such that a neuron’s response to contrast decreases as the average stimulus
contrast increases (Shapley and Victor, 1978; Sherman, 2004). For example, as contrast increases, retinal
activity can activate the metabotropic glutamate receptors at dendrodendritic terminals, increasing the
amount of same-sign inhibition on the postsynaptic relay cell (Sherman, 2004; 2014). This would prevent
10
relay cells from reaching response saturation. Same-sign inhibition can also serve another function;
computational modelling demonstrates that like sequential push-pull, same-sign inhibition can increase
the temporal precision of relay cell responses by curtailing the time window within which it fires spikes
(Babadi et al., 2010; Butts et al., 2011; Casti et al., 2008).
Same-sign inhibition can also describe, in another context, the On-Off cells in rodent and rabbit
(see section POTENTIAL ANATOMICAL CIRCUITS). Present in retina (Berson, 2010) as well as the LGN
(Dhande et al., 2015; Piscopo et al., 2013; Suresh et al., 2016), these cells are driven by stimuli of either
contrast polarity presented to the same regions of visual space (Figure 1.6B-D) (Huberman et al., 2008;
Figure 1.6. Response properties and receptive field structure of four different types of relay cells: Center-Surround, On-Off,
W3-like, and suppressed-by-contrast like. Adapted from Suresh et al, 2016. (A) Receptive field of an On-center relay cell mapped
using sparse noise shown as a contour plot (left; yellow box represents stimulus size) and as an array of trace pairs (right) in which
the averaged responses to bright (gray traces) and dark (black traces) to each stimulus are placed at corresponding positions in
the stimulus grid. Vertical dashes indicate stimulus onset. Red dashed contour indicates the On subregion. Blue dashed contour
represents the Off subregion. (B-D) Receptive field maps for an On-Off cell, W3-like cell, and suppressed-by-contrast like cell
respectively, conventions as in A. Unlike A, in which the difference map is shown (left), the separate On and Off receptive field
maps are shown for On-Off cells.
11
Huberman et al., 2009; Kim et al., 2010; Piscopo et al., 2013; Rivlin-Etzion et al., 2012; Tien et al., 2015)
rather than having spatially opponent On and Off subregions, unlike carnivore and primate in which an
overwhelming majority relay cells have center-surround structures (Figure 1.6A). The suppressed-by-
contrast (SBC) cells (Levick, 1967; Piscopo et al., 2013; Suresh et al., 2016) are a subclass of On-Off relay
cells that are inhibited by both bright and dark contrasts (Figure 1.6D), while other On-Off cells are excited
by both contrasts (W3 – Figure 1.6C; general On-Off – Figure 1.6B). SBC cells are thought to detect the
absence of contrast in a visual scene (Piscopo et al., 2013; Suresh et al., 2016). Thus, same-sign inhibition
may occur via multiple mechanisms across species.
Inhibition independent of stimulus contrast: stimulus-selective inhibition
So far, we have discussed how the spatial distribution of inhibition in the receptive field contributes
to visual processing. However, other types of inhibition can refine stimulus selectivity in the LGN for tuning
properties such as orientation and direction, temporal frequency, temporal precision, and global
(average) contrast (of the overall image versus local differences in space i.e. contrast borders). The role
of stimulus-selective inhibition in visual processing has received little attention.
In higher order mammals, stimulus-selective inhibition has only been implicated in visual processing.
With respect to temporal precision, the addition of suppression can improve the temporal precision of
relay cells responses (Babadi et al., 2010; Butts et al., 2016; Butts et al., 2011; Casti et al., 2008).
Specifically, the contribution is likely due to the local interneurons based on the detected timescale of the
inhibitory response (Butts et al., 2016). With respect to contrast, gain control significantly changes
between retina and the LGN. The LGN has reduced firing across contrasts compared to retina (though it
still has a nonlinear, increasing relationship to contrast) (Kaplan et al., 1987). Kaplan et al. proposed the
reduction in response transmission from retina to LGN may be due to the local interneurons or
perigeniculate nucleus, the carnivore and primate equivalent of the TRN.
12
In rodent, our group has begun to elucidate how stimulus-selective suppression contributes to
stimulus properties such as orientation and direction in the rodent. Subsets of On-Off cells in rodent dLGN
are specifically selective for stimulus properties such as orientation, direction, or size (Levick, 1967;
Piscopo et al., 2013; Suresh et al., 2016; Zhang et al., 2012) in the retina and LGN compared to carnivore
and primate whose cells show such responses in large number in cortex (Hubel and Wiesel, 1959; 1968)
(although there are a small number of orientation selective relay cells in the LGN of some species of
primates (Cheong et al., 2013; Smith et al., 1990; Xu et al., 2002)). Intracellular recordings implicate
untuned suppression in sharpening On-Off cells’ responses for their preferred orientation or direction
compared to their retinal inputs (Suresh et al., 2016). Furthermore, subsets of On-Off cells (W3-like or
suppressed-by-contrast neurons) can be sensitive to stimulus size and are inhibited by large, uniform
stimuli that are bright or dark (Suresh et al., 2016). The sources of these types of stimulus-selective
suppression are still unknown, but stimulus-selective inhibition seems to enhance feature selectivity in
relay cells.
Inhibition evoked by vast versus small stimuli: global versus local inhibition
In vitro slice preparations in mouse LGN suggest that inhibition can operate in two modes; local
and global (Acuna-Goycolea et al., 2008). The unique synaptic and cable properties of interneuron
dendrites make this feasible. LGN interneurons are capable of releasing neurotransmitter from their axons
and dendrites (Cox, 2014; Cox and Beatty, 2017; Sherman, 2004). Furthermore, distal dendrites receive
retinal input and are electrotonically remote. Thus, they can release neurotransmitter independent of the
soma (Crandall and Cox, 2012). In this way, LGN interneurons can provide localized inhibition. However,
the co-activation of several retinal inputs synapsing on an interneuron can cause simultaneous GABA
release from both axon and dendrites (Acuna-Goycolea et al., 2008), leading to large-scale or global
inhibitory effects on downstream neurons. Global inhibition is thus, evoked by vast stimuli and is
independent of stimulus contrast. Despite being broadly tuned, global inhibition can sharpen neuronal
13
responses in the LGN as in cortex (Lien and Scanziani, 2013; Liu et al., 2011). This sharpening could in turn
lead to the enhancement of reliability and information transmission in the LGN.
Inhibition independent of the stimulus: tonic inhibition
So far, we have discussed only stimulus-evoked inhibition. There is also evidence for inhibition
that is generated constitutively. This tonic form of inhibition has been studied most intensively in the
context of neurological and psychiatric disease, sleep and consciousness, and learning and memory
(Brickley and Mody, 2012). Current evidence suggests that tonic inhibition is mediated by extrasynaptic
GABA receptors throughout the brain (Brickley and Mody, 2012) unlike inhibition driven by sensory input,
which acts mainly at the synaptic locus. These extrasynaptic receptors have been identified as the targets
of various drugs (Brickley and Mody, 2012) and anesthetics (Jia et al., 2008) and, thus, are of therapeutic
interest. The role of tonic inhibition in sensory processing per se has received little attention, however.
Tonic inhibition may set the tone of neural excitability, either by subtly altering transmission or by
promoting the transition from tonic to burst modes (Bright et al., 2007; Cope et al., 2005). Thus, tonic
inhibition may influence the pattern of activity that a given stimulus evokes.
POTENTIAL ANATOMICAL CIRCUITS FOR VISUALLY-EVOKED INHIBITION IN RODENTS
The LGN is a conserved structure across vertebrates. Furthermore, while the density and
anatomical properties of local interneurons are not conserved in all thalamic nuclei across species, the
presence of local interneurons is consistent (Arcelli et al., 1996; Evangelio et al., 2018; Seabrook et al.,
2013). Different species offer their unique advantage to study thalamic circuitry. In this dissertation, I will
draw upon findings and ideas derived from carnivore and primate to inform my hypotheses, but I will take
advantage of the rodent animal model to investigate inhibitory mechanisms in the LGN.
Comparison of the early visual pathway in rodent to higher order mammals
Rodents such as rats and mice are nocturnal burrowing animals, unlike carnivores and primates.
Consequently, their visual system developed differently compared to more highly visual mammals. At the
14
level of the primary visual cortex, rodent cortical cells have simple cell-like receptive fields like higher
mammals such as cat (Bonin et al., 2011; Liu et al., 2010; Niell and Stryker, 2008; Smith and Häusser,
2010). However, while simple cells in cat are restricted to thalamo-recipient layers (layer 4 and 6) (Hubel
and Wiesel, 1962; Martinez et al., 2005), rodent simple cells are present across all cortical layers (Liu et
al., 2010; Liu et al., 2011; Niell and Stryker, 2008). Preference for orientation in higher mammals is
arranged in orientation columns where the transition from one orientation to another is smooth and
systematic (Hubel and Wiesel, 1962). On the other hand, in rodents, orientation preference takes on a
salt and pepper arrangement (Ohki and Reid, 2007). There is evidence pointing to differences in the
inhibitory circuits as well. In cat, there is a special class of interneurons that have simple receptive fields
and are narrowly tuned (Hirsch et al., 2003). However, in the rodent, a large fraction of interneurons are
broadly tuned (Camillo et al., 2018; Kerlin et al., 2010; Liu et al., 2010). Consequently, rodent simple-like
cells lack the classic push-pull structure of excitation and inhibition that is observed in all true simple cells.
There are also species differences in the LGN. Unlike cat, in which most or all cells are center-
surround, approximately 40% of rodent LGN cells have center-surround receptive fields with push-pull
(Suresh et al., 2016). The other approximate 60% of rodent LGN cells are selective for several stimulus
features such orientation and direction (Marshel et al., 2012; Piscopo et al., 2013; Scholl et al., 2013; Zhao
et al., 2013). There is evidence though that suggests that the koniocellular pathway in primates (Cheong
et al., 2013; Smith et al., 1990) and cells within the A-laminae in cat (Shou and Leventhal, 1989; Thompson
et al., 1994; Vidyasagar and Heide, 1984) may be analogous to the orientation- and direction- selective
neurons in rodent in that they are at least biased for orientation or direction. Rodent interneurons also
have unique structural attributes: they have expansive dendrites that span huge distances (Morgan and
Lichtman, 2020; Seabrook et al., 2013; Zhu et al., 1999), thus traversing large retinotopic distances
compared to interneurons in carnivores, which have spatially restricted dendritic extents (Sherman,
15
2004). As rodent interneurons, like carnivore and primate interneurons, can release GABA at their
dendrites, all together, this suggests that rodent interneurons may be more compartmentalized than
carnivore interneurons.
Putative anatomical circuits for push-pull across species
In cat, the prime candidate to supply the pull in the push-pull mechanism is the local interneurons,
which have receptive fields with a center-surround, push-pull structure (Wang et al., 2011b). There is also
evidence that interneurons preferentially select relay cells with the reverse center-surround sign
(Martinez et al., 2014). A potential circuit for push-pull, based on cat (Wang et al., 2011a; Wang et al.,
2007), is illustrated in Figure 1.7, Left. In this circuit, push is generated by retinal ganglion cells whose
receptive fields have the same position and stimulus preference as the relay cell; interneurons whose
center-surround receptive fields share the similar position but have the opposite preference for
luminance contrast generate pull (Suresh et al., 2016; Wang et al., 2007). This potential circuit might be
conserved in rodent. However, the presence of On-Off, W3, and suppressed-by-contrast cells in rodent
retina and LGN suggest other circuits might exist as well (Figure 1.7, Right).
Figure 1.7. Potential wiring diagrams for push-pull excitation and inhibition in the center-surround cell’s receptive fields.
(Left) Push-pull generated for an On-center relay cell via direct input from an On-center ganglion cell and disynaptic input from
an Off-center interneuron. Cells are represented as their receptive fields whose displacement is exaggerated for clarity.
Excitatory cells are outlined in black and interneurons in green with negative signs. (Right) An alternative circuit. Here, inhibition
is supplied by On-Off interneurons; other variations could include convergent input from On, Off, or On-Off interneurons.
16
CONCLUSION
Chapters 2 and 3 will investigate how inhibitory mechanisms in the LGN serve vision. This
dissertation will address (1) how the receptive field properties of local interneurons can contribute to the
patterns of inhibition recorded from in relay cells, and (2) how the spike timing of the local interneurons
contributes to the temporal structure of the relay cells’ response. Although there are differences between
rodent and highly visual animals, with careful consideration, the findings in this thesis have bearing on
visual processing across species.
17
CHAPTER 2 – RECEPTIVE FIELDS
INTRODUCTION
Local interneurons in the dorsal lateral geniculate nucleus of the thalamus (dLGN) provide
powerful inhibition that edits every signal relay cells receive and transmit to cortex (Dubin and Cleland,
1977; Hirsch et al., 2015; Lorincz et al., 2009; Seabrook et al., 2013; Sherman, 2004; Wilson et al., 1996).
To date, the visual response properties of these cells have best been studied in the cat (Hirsch et al., 2015)
and monkey (Wilson, 1989; Wilson et al., 1996), species in which the large majority of both interneurons
and relay cells have receptive fields with center-surround structures that mirror those in retina. Further,
work in cat shows that relay cells have push-pull responses to stimuli of the opposite contrast (in
subregions where bright light is excitatory, dark stimuli are inhibitory and vice versa) (Wang et al., 2011b;
Wang et al., 2007). It is easy to imagine that the center-surround pattern of inhibition in the relay cell’s
receptive field is supplied by interneurons with the opposite preference for stimulus contrast, and there
is substantial evidence for this arrangement (Martinez et al., 2014; Wang et al., 2007; Wilson, 1989;
Wilson et al., 1996). Moreover, this simple push-pull scheme is able to contribute to visual processing in
myriad ways (Hirsch et al., 2015; Werblin, 2010).
In comparison, the receptive field structures of murine retinal ganglion cells (Kerschensteiner and
Guido, 2017; Sanes and Masland, 2014) and relay cells are diverse (Grubb and Thompson, 2003; Piscopo
et al., 2013; Suresh et al., 2016). Although the largest single population of relay cells has receptive fields
with a center-surround structure (Piscopo et al., 2013; Suresh et al., 2016) and push-pull responses
(Suresh et al., 2016), the majority have different types of On-Off receptive fields (bright and dark stimuli
presented to the same regions of visual space evoke responses of the same sign). There are anatomical
differences between species as well. Interneurons in carnivores and primates have compact dendritic
arbors that subtend small regions of the dLGN and, hence, retinotopic space, whereas those in mouse are
18
so extensive that they nearly traverse the whole of the nucleus and visual field. Further, the proportion
of interneurons (approximately 6%) is a quarter of that in carnivore and primate (approximately 25%).
Last, the number of retinal ganglion cells that converge on single postsynaptic target in dLGN is far greater
in mouse (Hammer et al., 2015; Morgan et al., 2016; Rompani et al., 2017) than in carnivore and primate
(Martinez et al., 2014; Sincich et al., 2007; Usrey et al., 1999). These differences between species raise
the question of how interneurons meet the demands of processing diverse forms of retinal input.
Despite their importance, progress in understanding the role of interneurons in sensory
processing has been slow. They are small and scarce and cannot be distinguished by standard methods of
extracellular recording. Thus, we took advantage of the optogenetic approaches mouse offers to identify
interneurons and relay cells in vivo and compare their receptive fields structures. Most interneurons in
mouse primary visual cortex are not selective for stimulus polarity (Liu et al., 2009) or other properties
(Kerlin et al., 2010), and we had expected to find similar results in dLGN. However, this was not the case.
Interneurons had receptive field structures and sizes almost as diverse as those of relay cells, although
the distribution of receptive field types was different. To ask how these interneuron might provide the
types of feature specific suppression that we had previously recorded from relay cells in mouse dLGN
(Suresh et al., 2016), we built computational models fitted to our data. This analysis provided proof-of-
concept that combined input from several different type of interneurons is able to recapitulate the
suppressive components of the relay cell’s receptive field. Thus, despite substantial variations in their
morphology and receptive field structure, interneurons are able to provide precisely tailored inhibition to
relay cells across species.
19
MATERIALS AND METHODS
Key Resources Table
REAGENT or RESOURCE SOURCE IDENTIFIER
Bacterial and virus strains
AAV1.EF1a-double floxed-
hChR2(H134R)-EYFP-WPRE-
HGHpA
unpublished Addgene 20298-AAV1; RRID: Addgene_20298
AAV1.Ef1a-DIO eNpHR 3.0-
EYFP
(Gradinaru et
al., 2010)
Addgene 26966-AAV1; RRID: Addgene_26966
AAV-stGtACR2 is
AAV1/hSyn1-SIO-stGtACR2-
FusionRed
(Mahn et al.,
2018)
Addgene 105677-AAV1; RRID: Addgene_1005677
Experimental models: Organisms/strains
Gad2-IRES-cre The Jackson
Laboratory or
(Taniguchi et
al., 2011)
RRID:IMSR_JAX:010802
Ai32(RCL-ChR2(H134R)/EYFP) The Jackson
Laboratory or
(Madisen et
al., 2012)
RRID:IMSR_JAX:024109
Chemicals, peptides, and recombinant proteins
DiI Thermo-
Fisher
Cat# C7000
DiD Thermo-
Fisher
Cat# D12730
Fast Green FCF Sigma-Aldrich Cat# F7252
ProLong Glass Antifade
Mountant
Thermo-
Fisher
Cat # P36980
Software and algorithms
MATLAB Mathworks https://www.mathworks.com/products/matlab.html
RRID: SCR_001622
Spike2 Cambridge
Electronic
Design
https://ced.co.uk/products/spkovin;
RRID:SCR_000903
Fiji/Image J (Schindelin et
al., 2012)
https://fiji.sc/; RRID: SCR_002285
Kilosort2 (Pachitariu et
al., 2016)
https://github.com/MouseLand/Kilosort
RRID: SCR_016422
Phy (Rossant et al.,
2016)
https://github.com/cortex-lab/phy
allenCCF (Shamash et
al., 2018)
https://github.com/cortex-lab/allenCCF
Table 2.1. Key Resources Table
20
Resource availability
Lead contact
Further information and request for resources and reagents should be directed to Judith Hirsch
(jhirsch@usc.edu).
Materials availability
This study did not generate new unique reagents.
Data and code availability
Not yet available.
Experimental models and subject details
Animals
All experiments were performed using mice ≥ 8 weeks old to avoid the visual critical period; there
was no detectable difference between males and females, data from both sexes were pooled. Subjects
were either Gad2-IRES-cre mice (n=28, (Taniguchi et al., 2011), JAX 010802) or the progeny of these
animals crossed with Ai32 mice expressing ChR2(H134R)/EYFP ((Madisen et al., 2012), JAX 012569) (Gad2-
Ai32) (n=37 for Gad2-Ai32 mice). All procedures were approved by the Institutional Animal Care and Use
Committees of the University of Southern California following guidelines from the National Institutes of
Health.
Method details
Virus expression
Opsins were introduced into GABAergic cells by crossing Gad2-IRES-cre and Ai32 lines or by
injection of opsin via adeno-associated viruses (AAV) directly into the dLGN of Gad2-IRES-cre mice. To
excite interneurons, we introduced channelrhodopsin using: AAV1-EF1a-double floxed-hChR2(H134R)-
EYFP-WPRE-HGHpA (Addgene: 20298-AAV1, provided by Karl Deisseroth, ≥ 7 x 10
12
vg/mL).
To inhibit cells, we introduced one of two opsins: AAV1-hSyn1-SIO-stGtACR2-FusionRed
(Addgene: 105677-AAV1, provided by Ofer Yizhar (Mahn et al., 2018), ≥ 1 x 10
13
vg/mL) or AAV1-Ef1a-DIO
21
eNpHR 3.0-EYFP (Addgene: 26966-AAV1, provided by Karl Deisseroth (Gradinaru et al., 2010), ≥ 1 x 10
13
vg/mL).
Surgical preparation for viral injection
For some experiments, we injected AAV into dLGN to restrict expression of opsins to local
interneurons as follows: Gad2-IRES-cre mice were anesthetized with isoflurane prior to (2% in oxygen 2
L/min) and during (1% in oxygen 1 L/min) surgery. The anesthetized animal was positioned in a stereotaxic
device (Kopf) and a small craniotomy made above each dLGN at coordinates -2.2 mm anteroposterior and
± 2.2 mm lateral relative to bregma. Virus was injected via a glass micropipette filled with 2 μl of AAV
mixed with Fast Green (Sigma-Aldrich) lowered 2.5 mm below the brain’s surface. 100 nL of AAV was
delivered iontophoretically (Nanoject II, Drummond Scientific Company) for 5 min and the pipette
withdrawn 10 min after. The scalp was then sutured, and subjects were treated with Ketoprofen prior to
return to the home cage for recovery. Recordings were performed three to five weeks post-injection and
viral expression was confirmed postmortem by fluorescence microscopy.
Surgical preparation for recordings
Prior to electrophysiological recording animals were given an injection of chlorprothixene (5
mg/kg, i.p.) after which anesthesia was initiated and maintained with urethane (0.5-1 g/kg, 10% w/v in
saline, i.p.) (Ciftcioglu et al., 2020; Niell and Stryker, 2008; Suresh et al., 2016). After an incision was made
to expose the skull, a headpost was affixed caudal to the recording site and a small craniotomy over the
dLGN was made. All wound margins were infiltrated with bupivacaine and the brain and eyes were kept
moist with saline. Body temperature was measured using a rectal probe and maintained at 37
O
C.
Electrophysiological recording
Borosilicate glass pipettes were used to record from single neurons in whole-cell or cell-attached
mode, or a multi-conductor electrode was used to record from multiple cells simultaneously. Whole-cell
and cell-attached recordings were made with a Multiclamp 700B amplifier (Axon Instruments) in voltage-
22
clamp mode using standard techniques (Hirsch et al., 2003) and were digitized at 10 kHz with a Power
1401 data acquisition system (Cambridge Electronic Design); whole-cell recording in voltage-clamp mode
damped intrinsic conductances. Multi-conductor electrodes were single-shank, 32-channel, H6b probes
(Cambridge NeuroTech) connected with a 32-channel digital multiplexing headstage and a Digital Lynx
4SX-M acquisition system (Neuralynx); filters were set to 0.1 Hz and 9 kHz and the sampling rate was 30
kHz.
Optogenetic manipulation
The LED light source was either an LSD-1 (A-M Systems) or Optogenetics-LED-Dual (Prizmatix) that
was paired with a 200 μm fiber patch cable to a single cannula. Blue light (455 nm) was used to activate
channelrhodopsin or stGtACR2, and orange-red light (650 nm) was used to activate eNphR3.0. All light
from the fiber cannula was blocked to prevent it from reaching the either eye. For whole-cell and cell
attached recordings, we used an Optopatcher (Katz et al., 2013) (A-M Systems), in which an optical
cannula threaded through the pipette delivered light to the tip of the pipette. For multi-site recordings, a
Lambda-B optical fiber (100-core; 1.2 mm taper-tip) (Cambridge NeuroTech) that guided light to the tip
of each shank was fixed to the multi-electrode.
Stimulus presentation
All stimuli were generated using a ViSaGe stimulus generator (Cambridge Research Systems) and
displayed on a gamma corrected Dell U2211H LCD monitor at a 70 Hz refresh rate and a viewing distance
of 180 mm. The experimental stimulus set included constant luminance and sparse noise (Figure 2.1A).
The constant luminance stimulus consisted of a full-field white, black, or grey stimulus being presented
on the screen for the duration of recording (at least 20 trials; 1 s trial length; 1.5 s interval period). The
dLGN was stimulated by light pulses for varying periods and frequencies, including 10 ms-10 Hz, to 250
ms-1 Hz pulses (Figure 2.1B-I). The sparse noise stimulus consisted of bright and dark squares, two to
23
thirty degrees, shown at half or full contrast over sixteen to twenty repeats in pseudorandom order on a
16x16 grid (5
O
grid resolution) (Ciftcioglu et al., 2020; Jones and Palmer, 1987; Suresh et al., 2016).
Histological reconstruction of recording sites and confirmation of viral expression
At the end of each experiment, the animal was perfused with 3% paraformaldehyde. The brain
was removed, placed in phosphate, and then cut in 100 μm coronal section using a vibratome. 100 μm
sections were mounted using ProLong Glass Antifade Mountant (ThermoFisher) and viewed using a
fluorescent (Zeiss) or confocal microscope. Micrographs were processed with FIJI (Schindelin et al., 2012).
For mice injected with AAV, we confirmed that opsin was expressed and limited to dLGN.
For many experiments, multi-site electrodes were coated with a yellow- or red-shifted lipophilic
tracer (DiI, ThermoFisher C7000, and DiD, ThermoFisher D12730 respectively). Thus, it was possible to
view the electrode track postmortem (Figure 2.1A, right). We then use the ‘allenCCF’ software package
(Shamash et al., 2018) to register our brain sections to coordinates the Allen Brain Mouse Atlas and
created a best-fit line for each track using orthogonal distance regression. After identifying the location of
each cluster along the electrode (see ‘Event Detection and Sorting’), we localized each cluster to a
stereotaxic coordinate in the dLGN. In cases for which the multi-site electrode was not coated with dye
but, nonetheless, left a visible track in the brain, we used the same method to localize recording sites.
For whole-cell or cell-attached recordings, visible tracks were compared to the Allen Brain Mouse
Atlas either using the ‘allenCCF’ software package or via Neurolucida (MBF Science). For recordings that
left no visible track (whole-cell, cell-attached, or a subset of multisite recordings), approximate cell or
electrode position was estimated from stereotaxic coordinates and depth measurements (Suresh et al.,
2016). These estimates were corroborated by the experiments with dye-coated electrodes.
24
Quantification and statistical analysis
Optogenetic tagging of interneurons and relay cells
For animals in which interneurons expressed channelrhodopsin, we determined cell type from
responses to pulse trains of blue light of varied duration and frequency while the animal viewed a constant
full field bright, dark, or gray stimulus. Cells were classified as interneurons if they responded to LED pulses
within 10 ms (Figure 2.1C, G) or as relay cells if they remained suppressed during pulses ≤ 250 ms (Figure
2.1K). Periods of suppression during LED stimulation in multi-electrode recordings of relay cells (Figure
2.1H, I) is indicative of strong inhibition seen in whole-cell recordings of relay cells during LED stimulation
(Figure 2.1D). For animals in which interneurons expressed an inhibitory opsin, it was necessary to excite
cells with a visual stimulus and then assess optogenetic modulation of the visually evoked response. We
used full-field stimulus sequences that comprised three luminance steps, either bright-dark-bright or dark-
bright-dark. The LED was switched on for the 70% of duration of the central luminance step and then
ramped down up to 30% of the duration of the central luminance step past the end of the step. Cells were
defined as interneurons if the LED suppressed the visual response or as relay cells if the visual response
was enhanced (Figure 2.1J-M).
Event detection and sorting
For whole-cell recordings of membrane currents, EPSCs and spikes were detected via techniques
we used previously (Suresh et al., 2016; Wang et al., 2011b; Wang et al., 2007). Briefly, we applied an
adaptive threshold to the first derivative of the intracellular signal such that the smallest potential events
included both EPSCs and noise. These events were then sorted using a support vector machine (SVM)
(Chang and Lin, 2013) trained with randomly selected events that were manually labeled as EPSCs or
noise. Because events near the decision boundary were prone to misclassification, we labeled these
manually for additional training and then reclassified the dataset. Finally, spikes were sorted from the
EPSCs by repeating the algorithm exclusively on SVM-classified neural events.
25
Using Kilosort2 (Pachitariu et al., 2016) we sorted spikes into clusters that we manually assessed
with the visualization tool Phy (Rossant et al., 2016) to determine if single clusters represented activity
from just one versus multiple neurons, and also if spikes from a single cell had been divided into several
clusters that should be merged. We assigned depth for each unit based the position of the conductor that
recorded the largest amplitude waveform. In some cases, artifact from the photo-stimulus interfered with
recording of neural signals. To address this, we used a local polynomial approximation algorithm known
as SALPA to remove LED-induced artifact without distorting the individual spike waveforms (Wagenaar
and Potter, 2002).
Receptive field mapping
Receptive field maps were constructed from responses to sparse noise by calculating the spike-
triggered average (STA) of the stimulus ensemble (Schwartz et al., 2006) for time windows up to 300
milliseconds (t = -300ms) prior to the spike (t = 0). Thus, we estimated the spatiotemporal receptive field
for each cell. Cells were qualitatively identified as center-surround with On or Off centers, On, Off, or On-
Off based on the smallest stimulus size that elicited a robust response. Because visual responses are often
biphasic—the first phase corresponds to the stimulus that triggers spikes (known as the “triggering
phase”) and the second phase corresponds to the stimulus that primes the triggering phase (the “priming
phase”)—receptive field classification was based on the earliest (closest to t=0) robust response to bright
and/or dark squares.
Determining receptive field size
To measure receptive field size, we used standard techniques (Wang et al., 2007). We fit the
sparse noise generated maps with 2D Gaussian functions and used the 1σ contours to define the radius
of center sub-regions or single regions. Specifically, the sizes of the receptive fields were quantified using
the average of the semi-major and semi-minor axes of the 2D fits and by the area of these fits. The
26
receptive fields of several cells could not be fit; for these, we report the qualitative preference for stimulus
contrast but not receptive field size nor their quantitatively derived preference for stimulus contrast.
Time course of the On and Off responses of dLGN receptive field
We calculated the temporal STA (tSTA) from the mean activity over time within each 1σ contour
of the 2D Gaussian fits of the On and Off receptive field maps; thus, we obtained the On tSTA and Off tSTA
for each cell (Alonso et al., 2001; Roy et al., 2021; Weng et al., 2005; Yeh et al., 2009), which were each
mean-subtracted. On and Off tSTAs were normalized by the absolute maximum for a given cell. We
established time points in each tSTA as significant if the z-scored first-derivative of itself had values ± 0.75.
For each tSTA, a zero-crossing or distinct inflection point between two other global inflection
points was identified. The first occurred if the tSTA was biphasic with robust excitation followed by
inhibition or vice-versa (had a peak above the zero-crossing and trough below the zero-crossing); the
second could occur if the tSTA was biphasic but had two periods of excitation (two peaks) and our method
detected a significant decrease in activity in between (a trough) but had yet to return to mean-firing. We
ruled that the triggering phase occurred between t=0 and the zero-crossing or distinct inflection point,
and the priming phase occurred between the zero-crossing or distinct inflection point and earlier times.
Some tSTAs had neither a detectable zero-crossing nor distinct inflection point between two global
inflection points, and thus only their triggering phases could be identified. The first significant time point
closest to the spike (t=0) was defined as the latency of the triggering phase (triggering latency) (Figure
2.4A). The strength of the priming phase was defined as the integral of the tSTA during the defined priming
phase.
Spike-based bright-dark polarity score
To score the sign of each cell’s responses to bright and dark spots, we modified a bright-dark
response polarity score used for membrane current (Suresh et al., 2016) for use with spikes. The index
(Ω), is defined as follows:
27
Ω
𝐵𝑟𝑖𝑔 ℎ𝑡 =
𝑅 𝑚𝑎𝑥 (𝐵𝑟𝑖𝑔 ℎ𝑡 )
√(𝑅 𝑚𝑎𝑥 (𝐵𝑟𝑖𝑔 ℎ𝑡 ))
2
+ (𝑅 𝑚𝑎𝑥 (𝐷𝑎𝑟𝑘 ))
2
Ω
𝐷𝑎𝑟 𝑘 =
𝑅 𝑚𝑎𝑥 (𝐷𝑎𝑟𝑘 )
√(𝑅 𝑚𝑎𝑥 (𝐵𝑟𝑖𝑔 ℎ𝑡 ))
2
+ (𝑅 𝑚𝑎𝑥 (𝐷𝑎𝑟𝑘 ))
2
where R max is the value of the On or Off tSTA at its triggering latency. Ω Bright and Ω Dark values lie on a circle.
When bright and dark spots evoke responses of the same sign, either excitatory or suppressive, Ω Bright and
Ω Dark have the same polarity and so the index values occupy the first or third quadrants of the unit circle.
If bright and dark spots evoke responses of the opposite sign, Ω Bright and Ω Dark have opposite polarities and
the index values occupy the second or fourth quadrants of the unit circle.
To further summarize this, we converted the bright and dark indices from Cartesian coordinates
to polar coordinates and extracted the angle (Θ). Angle values between -180 < Θ ≤ -90 degrees indicated
a set of responses were suppressed by both bright and dark spots, between -90 < Θ ≤ 0 degrees indicated
a set of responses were excited by bright spot and suppressed by dark spots, between 0 < Θ < 90 degrees
indicated a set of responses were excited by both bright and dark spots, and between 90 ≤ Θ ≤ 180 degrees
indicated a set of responses were excited by dark spots and suppressed by bright spots.
Correlating interneuron spatiotemporal responses to synaptic inhibition from the spatiotemporal
receptive fields of relay cells
Characterizing interneurons’ spatiotemporal responses
To investigate if interneurons could explain the spatiotemporal inhibition we observed in relay
cells, we first implemented standard methods of linear-nonlinear-Poisson (LNP) models to characterize
the spatiotemporal response of On and Off interneurons to visual stimuli (Chichilnisky, 2001) across sparse
noise stimuli with varying parameters. We extended this model for On-Off interneurons to account for
difference in the fields of these cells’ On and Off responses. The LNP model for On-Off cells was modified
as:
28
𝑝 (𝑠𝑝𝑖𝑘𝑒 |𝑠 ⃗) = 𝑓 𝑜𝑛
(𝑤 𝑜𝑛
𝑠⃗
𝑜𝑛
) + 𝑓 𝑜𝑓𝑓 (𝑤 𝑜𝑓𝑓 𝑠⃗
𝑜𝑓𝑓 )
Where 𝑓 is the nonlinear filter, 𝑤 is the linear filter recapitulating by the spatiotemporal separable kernel,
and 𝑠 ⃗ is the stimulus input.
We computed the model from 80% of at least one sparse noise stimulus and use these are the
linear component of the model (the remaining 20% was reserved to assess performance of the model). If
multiple sparse-noise recordings with the same sequence were available, the dataset was concatenated,
and each recording was normalized by its 90
th
percentile. The time bin, or temporal resolution, was set by
the rate of stimulus update (28 ms). The static nonlinearity function was a soft-plus function, a smooth
approximation to the rectifying linear unit. Finally, the performance of the model was assessed by cross-
validation (using the reserved 20% of the data) and quantified as the explained variance or corrected
explained variance (Haefner and Cumming, 2009; Vaingankar et al., 2012) the response that the model
predicted. The explained variance can be biased by limitations in the length of the stimulus sequence and
number of repeated trials (placement and contrast of sparse noise stimulus squares). This can be
addressed by the corrected explained variance that corrects for two potential confounds: namely
uncertainty due to finite amount of data available for cross-validation as well as the number of free
parameters in the model.
Isolating net inhibition from relay cells
Next, we measured the net outward extent of membrane currents evoked by sparse noise stimuli.
As inhibition could not be resolved as individual currents, our strategy was to remove spikes and EPSCs
from the total membrane current. We first detected the times of EPSCs using event detection and sorting
methods mentioned earlier. We then removed spikes (MATLAB: medfilt2 function, The MathWorks)
(Suresh et al., 2016; Wang et al., 2007). EPSCs were less simple to remove since their size varied and their
long-time courses lead to frequency overlap. To identify distinct EPSCs coming from different retinal
29
sources, we clustered EPSCs at the previously detected times based on amplitude (area under the peak of
the first derivative of the EPSCs) and slope (the peak value of the first derivative of the EPSCs) (Wang and
Burkhalter, 2007). The average of each cluster was modeled by a linear rise-exponential-decay function
(Wang et al., 2007), and thus we generated EPSCs templates for each cluster. Given the times for each
EPSC (𝑡 𝑖𝑗
) and corresponding EPSC template (𝜎 𝑗 𝐸𝑃𝑆𝐶 ), we calculated the predicted EPSC current trace as:
𝐼 𝐸𝑃𝑆𝐶 (𝜏 ) = ∑ ∑ 𝜎 𝑗 𝐸𝑃𝑆𝐶 (𝜏 − 𝑡 𝑖𝑗
)
𝑖 𝑗
where 𝑡 𝑖𝑗
is the time of the 𝑖 th EPSC of the 𝑗 th cluster. The current that remained after subtracting the
predicted EPSC current trace from the filtered trace was mostly slow outward current. We call this the
residual trace. The time bin, or temporal resolution, of residual trace was set by the rate of stimulus
update (28 ms), and the residual trace was rectified. Standard methods of reverse correlation (Schwartz
et al., 2006) were used to compute the spatiotemporal receptive fields of the residual trace, which we call
the suppressive spatiotemporal receptive field. Additionally, we used standard linear-nonlinear (LN)
models of each residual trace with and without L1 regularization and a rectified linear nonlinearity to
increase the signal to noise ratio of the suppressive spatiotemporal receptive fields (Suresh et al., 2016;
Vaingankar et al., 2012; Wang et al., 2011b). Optimal L1 regularization for each residual trace was
manually chosen to minimize noise without altering the suppressive spatiotemporal receptive field.
Comparing ensembles of interneuron suppression to net inhibitory spatiotemporal responses in relay cells
We then found the optimal combination of spatiotemporal receptive fields of increasing numbers
of interneurons (ensembles ranged from one interneuron to twenty) that was correlated with a relay cell’s
suppressive spatiotemporal receptive fields (of the residual trace and its LN kernel) via linear regression.
For each ensemble of increasing size, a search heuristic known as a genetic algorithm was used to find the
optimal combination of interneuron spatiotemporal receptive fields. Free parameters included restricted
30
spatial and temporal translation for each interneuron field. Model performance was quantified by
explained variance (Pearson’s correlation coefficient, r).
Statistics
All statistics were performed in MATLAB (Mathworks). For datasets in which two groups are
compared, data from each group was first tested for normality by means of the Shapiro-Wilk test (null
hypothesis is that the distribution is normal) and then for equal variance between the two groups by
means of the two-sample F-test (null hypothesis is that the two distributions come from normal
distributions with the same variance; MATLAB function ‘vartest2’). If both conditions were met, the two-
sample t-test (assumes normality and equal but unknown variance; MATLAB function ‘ttest2’) was used.
If only the first condition was met, the two-sample Welch’s t-test was used. Otherwise, the non-
parametric test used was the Wilcoxon rank sum test (equivalent to two-sample Mann-Whitney U-test;
MATLAB function ‘ranksum’) in addition to the two-sample Kolmogorov-Smirnov test (MATLAB function
‘kstest2’). All statistical tests were two-tailed with sample size indicated in each figure and/or legend.
Significance levels were indicated by asterisks: p>=0.05, *; p<0.05; **: p<0.01; ***; p<0.001,****.
RESULTS
Our goal is to understand how local interneurons in the dLGN contribute to the receptive field
structure and response properties in relay cells. To address this question, we recorded visual responses
from local interneurons in mouse and used computational approaches to determine how these responses
might give rise to patterns of inhibition recorded from relay cells. Sampling interneurons in vivo is
challenging for many reasons; they are difficult to target because they are small and sparse (Arcelli et al.,
1996; Evangelio et al., 2018) and cannot be identified by spike statistics (Wang et al., 2011b). Thus, we
took an optogenetic approach and compared responses of 27 interneurons to 127 relay cells distributed
throughout the dLGN in 65 mice and combined these data from whole-cell recordings from 24 relay cells
made in the course of an earlier study (Suresh et al., 2016).
31
Optotagging interneurons and relay cells
Optogenetics has become an established method to distinguish between different types of cells
(Lima et al., 2009). Local interneurons are GAD-positive (Sabbagh et al., 2021); thus, we were able to use
the Gad2-cre mouse line to drive expression of Channelrhodopsin-2 (ChR2), halorhodopsin (eNphR3.0),
or soma-targeted anion-conducting channelrhodopsin (stGtACR2), either by breeding these mice with
Ai32 mice or by injecting AAV vectors into the dLGN (Figure 2.1).
While most of our recordings of visual and optogenetic responses (Figure 2.1) were made
extracellularly with multichannel silicon probes (Figure 2.1F-L), we began our studies with whole-cell
recordings to verify that ChR2-mediated changes in spike rate reflected changes in membrane currents
(Figure 2.1B). Pulses of blue light delivered through a cannula in the dLGN evoked depolarizations (often
capped by spikes) for a subset of cells we thus classify as interneurons (Figure 2.1C) and hyperpolarizations
in cells we thus classify as relay cells (Figure 2.1D). Each panel shows two individual trials of a pulse train
above the average (bold); recordings were made while the animal viewed a full-field gray stimulus (banner
at top). Expression of ChR2 was confined to interneurons (Figure 2.1E). The same optogenetic protocols
during extracellular recordings (Figure 2.1F) evoked volleys of spikes at short latency for interneurons
(Figure 1G) and either prolonged suppression (Figure 2.1H) or periods of suppression followed by bursts
of spikes (presumably T-type calcium channel mediated bursts) (Huguenard and McCormick, 1992; Lu et
al., 1992; Scharfman et al., 1990) for relay cells as illustrated by raster plots over peristimulus time
histograms (PSTHs) (Figure 2.1I).
We complemented the experiments with ChR2 by using inhibitory opsins (Figure 2.1J-M). For
these experiments, it was often necessary to increase neural firing rate with a visual stimulus in order to
assess optogenetic modulation of response. We identified cells whose visually evoked activity was
reduced by photostimulation as interneurons (Figure 2.1K) and classified those whose visually evoked
suppression was relieved as relay cells (Figure 2.1L); expression of hyperpolarizing opsins was also
32
Figure 2.1. Utilizing optogenetics in anesthetized mice to identify interneurons and relay cells during in vivo recordings. (A)
Constant luminance, changes in luminance, and sparse noise (low and high contrast) were shown to mice while recording from
the dLGN (left and middle). Histological reconstruction of dLGN (expressing ChR2-EYFP) and electrode track identified
postmorterm (DiI) (right). (B) Intracellular (and juxtacellular recordings) were performed using the Optopatcher (Katz et al, 2013)
in mice expressing ChR2-EYFP. (C) In intracellular recordings, interneurons could be identified by optogenetically-activated EPSCs
and spikes that closely followed the blue light (within 10 ms). (D) In intracellular recordings, relay cells could be identified by time-
locked optogenetically-activated IPSCs to the blue light. (E) Expression of ChR2-EYFP in dLGN interneurons. (F) Multi-electrode
recordings were performed in mice expressing ChR2-EYFP. (G) Blue light evokes short latency responses in interneurons. (H) Blue
light evokes sustained suppression or (I) suppression followed by burst spikes upon removal of light in relay cells. (J) Multi-
electrode recordings were performed in mice expressing eNphR3.0-EYFP. (K) Orange light inhibits interneurons and (L) disinhibits
relay cells in mice expressing eNphR3.0-EYFP; the banner at top indicates when the animal viewed a white or black screen (a
change in luminance). (M) Expression of eNphR3.0-EYFP in dLGN interneurons.
33
confined to interneurons (Figure 2.1M). Thus, complementary optogenetic techniques provide a means
to classify cell types.
Receptive field structures of interneurons are nearly as diverse as those of relay cells
Unlike cat, for which almost all relay cells have center-surround receptive fields and push-pull
responses (Hubel, 1960), there is great diversity in mouse (Marshel et al., 2012; Piscopo et al., 2013; Scholl
et al., 2013; Suresh et al., 2016; Zhao et al., 2013). Although many relay cells in murine dLGN have
receptive fields that resemble those in cat, the rest of the population has various types of On-Off (On and
Off subregions overlap) receptive fields (Suresh et al., 2016). Thus, we asked if the diversity of receptive
field structure was mirrored by local interneurons, or if they, like GABAergic cells in murine visual cortex
(Liu et al., 2009), pool local inputs to supply nonspecific inhibition to their excitatory counterparts.
Our finding is that interneurons have diverse receptive field structures (Figure 2.2). We mapped
receptive fields with sparse noise, sequences of individually flashed bright and dark squares, so we could
separate the contributions of the On and Off responses (see STAR methods) and depicted receptive fields
Figure 2.2. Receptive fields of interneurons are diverse. (A) Receptive fields of On-Off, On, and Off interneurons identified by
optogenetics; each contour plot was made from responses to dark (blue) or bright (red) stimuli. Yellow squares indicate stimulus
size; grid spacing, 5°. Note, large stimuli were often necessary to evoke responses from cells with large receptive fields. (B)
Receptive fields of On-center and Off-center interneurons identified by optogenetics; each contour plot was made from the
summed mapped responses to dark and bright stimuli. Otherwise, similar to A. (C) Pie-charts of different types of receptive fields
of interneurons and relay cells in mouse LGN (qualitatively assessed; interneurons, n=27; relay cells, n=205).
34
as contour plots (from the peak frame of the spike-triggered average of the stimulus ensemble, a STA).
The most common types of receptive fields we recorded were On-Off (cells that are excited by bright and
dark stimuli placed in the same or overlapping regions of visual space) (Figure 2.2A); maps constructed
from responses to bright (On) stimuli are shown next those made from responses to dark (Off) ones. The
receptive fields we mapped ranged from large and patchy to small and discrete (Figure 2.2A). STAs for On
and Off cells spanned a similar range in size (Figures 2.2B, C). Finally, a subset of On and Off cells had a
surround and assumed the classical On-Center or Off-Center receptive field shape (Figure 2.2B). Plotting
the distribution of receptive field types (qualitatively determined based on the smallest sparse noise
stimulus that elicited a robust response) for interneurons alongside that for relay cells revealed
differences between the two populations (Figure 2.2C). There was a relatively small proportion of
interneurons (26%) with center-surround receptive fields compared to that for relay cells (56%); this result
might be categorical or simply indicate that interneurons have surrounds too weak to detect with our
standard stimulus. There were roughly double the proportion of interneurons with conventional On-Off
receptive fields in Figure 2.2A; we have yet to find interneurons with receptive fields that were defined
by exceptionally strong inhibition, i.e., suppressed-by-contrast-like cells, for which both bright and dark
stimuli evoke strong inhibition, and W3-like cells, for which all but small stimuli evoke strong suppression.
Figure 2.3. Receptive fields of interneurons range from small to large. (A) The range in receptive-field sizes for local interneurons
and relay cells is similar. 1σ contours of 2D Gaussian fits to the receptive fields of interneurons (green) and relay cells (black)
aligned to the center of the stimulus grid (interneurons, n = 27; relay cells, n = 205). Fits were made to the full extent of each
(sub)region for On, Off, On-Off cells or to the centers for center-surround cells. (B) Box and violin plots show that the distributions
of all receptive field sizes (left) for interneurons (green) and relay cells (black) are statistically indistinguishable; Wilcoxon rank-
sum test and two-sample Kolmogorov-Smirnov test (p>0.05). Distribution of receptive field sizes for On, Off, On-Off, SBC, and
adaptive cells (see Figure 2.4) (right). For the box plots, horizontal lines are means of distributions, circles indicate medians. For
the violin plots, horizontal lines are individual values.
35
Conceivably, the balance of cell types might change slightly with sample size, but the most important
observation is that interneurons have receptive fields almost as diverse as those of relay cells.
The distribution of receptive field size for interneurons and relay cells is similar
Past results from carnivore (Martinez et al., 2014) and primate (Wilson, 1989; Wilson et al., 1996)
suggest that receptive fields of interneurons are larger than those of neighboring relay cells, a result
consistent with the relative frequency of the two cell types (Arcelli et al., 1996). Given that the proportion
of interneurons is even smaller in mouse than carnivore and primate (Arcelli et al., 1996; Evangelio et al.,
2018) and the extent of dendritic arbors with respect to the retinotopic map even greater (Morgan and
Lichtman, 2020; Seabrook et al., 2013), it might seem as if the receptive fields of murine interneurons
should be markedly larger than those of relay cells (although some relay cells in mouse dLGN have large
receptive fields themselves) (Ciftcioglu et al., 2020; Kerschensteiner and Guido, 2017).
The size of an interneuron’s receptive field determines the spatial specificity of the inhibition it
provides. Thus, we quantified the sizes of receptive fields using 2D Gaussian fits to the peak frame of the
STA. Our results suggest that the distribution of receptive fields sizes (plotted as the 1σ contour of the fit
for each cell) are similar for both populations overall (Figure 2.3A). This overall similarity in response area
remained when we compared measurements for separate receptive field types; an apparent trend
towards slightly larger fields for interneurons did not reach statistical significance (Figure 2.3B).
Stability of stimulus preference across a range of stimulus strength
It was important to compare visual responses of relay cells and interneurons with a standardized
stimulus; the receptive fields illustrated in Figures 2.2 and 2.3 were mapped using the smallest effective
stimulus size at half (50%) contrast (i.e., black and white squares on a mean gray background). It has
become clear, however, that neurons in dLGN receive input from many ganglion cells (Hammer et al.,
2015; Litvina and Chen, 2017; Morgan et al., 2016; Seabrook et al., 2013) and this led us to ask if stimuli
of different strengths would recruit responses with different properties. To address this question, we
36
adapted an index called the Bright-Dark Polarity Score to quantify the extent to which a cell preferred, or
37
was inhibited by, dark or bright stimuli of a given size or contrast. First, we calculated the temporal STA
(tSTA) from the region that fell within the 1σ contour of the 2D Gaussian fit of the initial peak of the
receptive field. An idealized tSTA is plotted as normalized stimulus contrast against time (moving
backwards from zero milliseconds, the instant at which each spike occurred) (Figure 2.4A). The curve splits
into two phases. These are a triggering phase (in response to the stimulus that immediately evokes a
spike) and an earlier priming phase (in response to the stimulus that facilitates firing). The latency and
duration of each phase was calculated using time points that deviated significantly from the mean. We
used the triggering phase to calculate the Bright-Dark Polarity Score and plotted index values along a circle
divided into four (color-coded) quadrants that indicate receptive field type (Figure 2.4B). Thus, tSTAs for
which the triggering phase yielded a positive Bright Score and/or a negative Dark Score, fell in the fourth
(lower right) quadrant and were classified as “On”; those with negative Bright and Dark Scores fell in the
third quadrant were classified as suppressed-by-contrast-like and so on. The response types can be
succinctly described by the angle (Ɵ) of each response in polar coordinates (Figure 2.4C and 2.4E; Figure
2.S1A-B). Last, note that this index only scores responses from the peak of the receptive fields and does
not take surrounds into account.
For most cells, receptive field type remained the same across stimulus strength (size or contrast),
as illustrated for Off-center, On-surround relay cell with push-pull responses (Figure 2.4C-D). Sequences
of frames from the STAs computed from responses to small stimuli displayed at 50% contrast are shown
above those made using full contrast stimuli (Figure 2.4D). Modest push-pull excitation and suppression
Figure 2.4. Receptive fields of thalamic neurons change as a function of stimulus strength. (A) Mean activity of 2D Gaussian fit
overlaying the spatiotemporal receptive field (STRF) for On and Off responses is measured and the triggering latency is identified.
(see Methods) (B) The amplitude at this latency for the On and Off responses is used to compute the Bright and Dark Polarity
Scores, and from these, an angle (Θ) of where these scores lie on a unit circle (see Methods). (C-D) Example of an Off cell; (C) Θ
values for an Off cell across stimulus size and contrast. (D) STRFs from the Off cell in response to low contrast sparse noise stimuli
at 5
O
(first and second row) and a high contrast sparse noise stimuli at 20
O
(On responses in first and third row; Off responses in
second and fourth row) (E-F) Example of an adaptive interneuron. (E) Responds as an On cell for low contrast sparse noise stimuli,
as an Off cell for high contrast and 10
O
– 30
O
degrees sparse noise stimuli, and as an On-Off cell for 5
O
high contrast sparse noise
stimulus. (F) Similar conventions as in D. (G) Pie-charts of different types of receptive fields of interneurons (n = 27) and relay
cells (n = 205) in mouse LGN (categorized by quantitative Bright-Dark Polarity Scores and qualitative assessment of center-
surround structure; cells were considered adaptive if Θ values changed across stimulus size and/or contrast).
38
are seen in the third frame (t=-57 milliseconds) of the STA computed from weaker stimuli; the emergence
of the surround takes shape more slowly. Larger and high contrast stimuli evoked stronger responses but
did not change the overall type of the receptive field. The Bright-Dark Polarity score defines the cell as an
Off cell at both stimulus contrasts and for all sizes (Figure 2.4C; Figure 2.S1A).
For a subset of cells, however, receptive field class changed as a function of stimulus parameters.
We use the term “adaptive” to describe these cells. The responses of one such interneuron to different
stimulus sizes at half and full contrast are illustrated in Figures 2.4E-F (and Figure 2.S1B). Figure 2.4F shows
that small spots at 50% contrast evoked a slowly evolving On response but no Off response (top two rows);
larger spots at 50% contrast did not evoke measurable responses (STAs not shown). At high contrast, small
stimuli evoked both On and Off responses (middle rows), whereas large stimuli elicited an Off response
only (bottom rows). These stimulus-dependent changes in response type are quantified by the Bright-
Dark Polarity score (Figure 2.4E). After analyzing responses for our entire dataset across stimulus strengths
and sizes, we found that 15% of interneurons and 8% of relay cells had adaptive receptive fields. We
revised the pie charts shown in Figure 2.2C to reflect this finding (Figure 2.4G).
Last, response dynamics in the priming phase varied so widely from cell to cells that we have not
included this response interval in our analysis. Still, we wish to make note of a small population of relay
cells and interneurons that were noticeably suppressed by both bright and dark during the priming phase
(Figure 2.S1C-I). We do not recall observing such variability in our previous recordings from cat.
Reconstructing inhibition in the relay cells’ receptive fields
We next asked if the receptive field structures of interneurons that we had mapped might explain
the suppressive component of the relay cell’s receptive field. To address this issue, it would be ideal to
optimize stimulus parameters for each interneuron as well as record its responses to the identical stimuli
used during whole-cell recordings of the membrane current from relay cells. Interneurons are rare,
39
however (Evangelio et al., 2018). Hence, for most experiments, we used a multielectrode so that we could
record from many cells at once in hopes of finding a single interneuron. Consequently, it was not possible
to tailor stimuli for every cell. Thus, we turned to a computational approach to predict how interneurons
would respond to stimuli of a given size or contrast. We built linear-nonlinear Poisson models (LNP)
(Paninski, 2004) using visual responses recorded from different types of interneurons. The receptive field
of a given On or Off cell was described with two linear kernels, one spatial and one temporal (Chichilnisky,
2001), and the nonlinearity was generated using standard techniques (Karklin and Simoncelli, 2011;
Zoltowski and Pillow, 2018). The only nonstandard approach we took was to separate the analysis of On
and Off responses. The shapes of the STAs computed from recordings of the interneurons were well
matched by the output of the corresponding models; compare the top and bottom sequences for an
Figure 2.5. LNP models of interneurons. (A-B) Example of an On interneuron. (A) Spatial and temporal kernels of an On
interneuron. (B) Spike-triggered average (top) and spatiotemporal receptive field of the LNP model (bottom) for an On
interneuron. (C-D) Example of an Off interneuron; conventions similar to A and B. (E-F) Example of an On-Off interneuron. (E1-
E2) Spatial and temporal kernels for the On and Off component, respectively. (F1-F2) Spike-triggered average (top) and
spatiotemporal receptive field of the LNP model (bottom) for the On and Off component, respectively.
40
example On cell (Figure 2.5A, B; Figure 2.S2A), Off cell (Figure 2.5C, D; Figure 2.S2B), and an On-Off cell
(Figure 2.E, F; Figure 2.S2C). The modeled STAs also had the benefit of reducing noise. Under the
assumption that output from the types of interneurons we sampled are available to relay cells distributed
in different regions of the retinotopic map (Morgan and Lichtman, 2020) and, as we have observed in the
population, have slightly different latencies, we took the liberty of translating the modeled interneurons
across space and shifting them forward or back by one frame of the STA for the analyses we describe
below.
It is possible that every murine relay cell receives input from an interneuron that shares its
receptive field structure, along the lines of the scheme proposed for cat, where receptive fields with
center-surround shapes are common to both relay cells and interneurons (Martinez et al., 2014; Wang et
al., 2011b). The differences between species suggest that there might be alternative schemes to produce
feature-specific inhibition in mouse, however. Not only do murine relay cells and interneurons have
diverse receptive fields, but the distributions of receptive field types were different—proportionally fewer
interneurons had obvious center-surround receptive fields with the balance comprising On, Off, and On-
Off profiles (Figure 2.2C, 2.4G). Also, interneurons compose a much smaller percentage of the population
in mouse than cat (≤ 6% versus 25%) (Arcelli et al., 1996; Evangelio et al., 2018), and a single murine
interneuron innervates much of the dLGN (Morgan and Lichtman, 2020) but only a small region in cat.
Thus, is possible that different varieties of interneurons converge to form feature-specific inhibition in
relay cells. Directly testing this idea is not currently feasible, but it was possible to design analyses using
our existing data to explore proof of concept.
We illustrate the overall approach for an On-Center relay cell (Figure 2.6). Its receptive field is
shown as a contour plot next to an array of trace pairs in which averaged responses to bright (gray traces)
and dark (black traces) squares are organized point by point along the stimulus grid; the center and
surround are approximated by dashed contours (Figure 2.6A). Membrane currents in the center reveal
41
strong push-pull currents and weaker ones in the surround, shown at increased gain in the insets placed
below the contour maps. To move forward with our analysis, it was necessary to isolate the underlying
net suppressive (outward) currents from the raw signal. While we had, solved a similar problem for
recordings from cat (Wang et al., 2011b; Wang et al., 2007), mouse presented a greater challenge. In cat,
retinogeniculate convergence is limited (Cleland et al., 1971; Hamos et al., 1987; Mastronarde, 1992;
Usrey et al., 1999); it was not difficult to detect EPSCs in the membrane current and form a template that
matched their average shape. We were thus able to subtract spikes and template EPSCs from the raw
trace with relative ease and isolate slow currents, which are mainly inhibitory (Wang et al., 2011b; Wang
et al., 2007) In mouse, however, retinogeniculate convergence ratios are high. As a result, EPSCs not only
vary in size and shape, but are often superimposed. Thus, we used a support vector machine (Chang and
Lin, 2013; Suresh et al., 2016) to classify as many EPSCs as possible (usually the large majority) for a given
relay cell, and then performed K-means clustering (Arthur and Vassilvitskii, 2006) to generate EPSC
templates of different sizes. We subtracted spikes and EPSCs (represented as their templates) from the
raw recording, downsampled the residual, suppressive, signal to the stimulus update rate, and rectified it
to remove uncaptured excitation (Figure 2.6B). The receptive field generated from spikes and EPSCs
(Figure 2.6C, top) had the opposite sign of that generated from the suppressive signal (Figure 2.6C, second
row). Thus, our method captured both the push and pull. We extracted the basic shape of the suppressive
field from noise using an LN model (Suresh et al., 2016; Wang et al., 2011b) (Figure 2.6C, third row) and
proceeded to determine its potential origin by using the genetic algorithm to select candidate inputs from
our dictionary of interneuron models. For this cell, an aggregate of seven inputs recapitulated the overall
structure of suppressive field receptive (Figure 2.6C, bottom row). Inputs with negative weights were
excluded as these provide an indirect means of sculpting the aggregate field. The receptive fields and
relative weight contribution of each input are provided in Figures 2.6D and the correlation between the
suppressive and aggregate fields for increasing numbers of input is shown in Figure 2.6E.
42
A parallel analysis for a suppressed-by contrast-relay cell is provided in Figure 2.7; conventions
are the same as for Figure 2.6 except that both the On and Off maps for both the relay cell and the
interneurons that simulated the suppressive field are shown separately. Similar to the On-Center relay
cell, the suppressive field of the suppressed-by-contrast cell has the opposite sign; five interneurons
Figure 2.6. Ensemble of interneurons can provide inhibition to center-surround relay cells. (A) Receptive field of an On-center
relay cell mapped using sparse noise shown as a contour plot (left-top; yellow box represents stimulus size) and as an array of
trace pairs (right) in which the averaged responses to bright (gray traces) and dark (black traces) to each stimulus are placed
at corresponding positions in the stimulus grid. Vertical dashes indicated stimulus onset. Red dashed contour indicates the On
subjection. Blue dashed contour represents the Off subregion. Insets (left-bottom) of select positions within the center and
surround portion. (B) Responses of the On-center relay cell to approximately 1.2 seconds of sparse noise stimuli (top). The
components are separated into following from the top: the spike-subtracted current, the templates fit to each EPSC, and the
residual (spike and EPSC subtracted) currents (in pink-dashed). The rectified residual overlays the residual trace (pink solid).
(C) The event-triggered average for the relay cell (first row), the reverse correlation of the residual/net-suppressive current
(second row), its linear-nonlinear model (third row), and the receptive field of the optimal ensemble of seven interneurons
that correlates with the LN kernel (fourth row). (D) The spatiotemporal receptive fields of the seven LNP models of
interneurons that make up the ensemble field (left) with their amount of contribution to the ensemble (right). (E) The
correlation between each ensemble of increasing number of interneuron LNP models and the suppressive LN kernel.
43
recapitulated its overall structure. It is likely that our results would improve with a larger sample of
Figure 2.7. Ensemble of interneurons can provide inhibition to SBC relay cells. (A) Receptive field of an SBC relay cell mapped
using sparse noise shown as a contour plot (left-top; yellow box represents stimulus size) and as an array of trace pairs (right) in
which the averaged responses to bright (gray traces) and dark (black traces) to each stimulus are placed at corresponding
positions in the stimulus grid. Vertical dashes indicated stimulus onset. Red dashed contour indicates the On subjection. Blue
dashed contour represents the Off subregion. Insets (left-bottom) of select positions within the center and surround portion. (B)
Responses of the SBC relay cell to approximately 1.2 seconds of sparse noise stimuli (top). The components are separated into
following from the top: the spike-subtracted current, the templates fit to each EPSC, and the residual (spike and EPSC subtracted)
currents (in pink-dashed). The rectified residual overlays the residual trace (pink solid). (C) The event-triggered average for the
relay cell (first row), the reverse correlation of the residual/net-suppressive current (second row), its linear-nonlinear model (third
row), and the receptive field of the optimal ensemble of seven interneurons that correlates with the LN kernel (fourth row). (D)
The spatiotemporal receptive fields of the seven LNP models of interneurons that make up the ensemble field (left) with their
amount of contribution to the ensemble (right). (E) The correlation between each ensemble of increasing number of interneuron
LNP models and the suppressive LN kernel.
44
interneurons, or a more elaborate models, but our goal was to provide a proof-of-concept that convergent
input from diverse types of interneurons can provide, in principle, feature-specific inhibition to single relay
cells in the murine dLGN.
DISCUSSION
Every relay cell in the dLGN receives powerful inhibition from local interneurons. To understand
how thalamus processes the information it receives and transmits, it is imperative to identify the features
the interneurons encode. Interneurons are difficult to sample, however, because they are small and
sparse (approximately 6% of the population) and cannot be distinguished by spike statistics. Thus, we used
optogenetics to identify local interneurons in vivo, mapped their receptive fields, and asked how the
inhibition these cells supply might account for suppression in the relay cell’s visual response. We found
that interneurons, like relay cells, have diverse types of receptive fields. The distribution of receptive field
structure, but not size, differed between the two populations. To learn how interneurons might contribute
to responses of relay cells, we turned to computational approaches. We built simple LNP models of
interneurons and used the output of these models to reconstruct suppressive components of the relay
cell’s receptive field (that we extracted from recordings of the membrane current). We were able to
reconstruct these suppressive fields for a group of relay cells including those center-surround and
suppressed-by-contrast profiles using a biologically plausible number of inputs (a minimum of
approximately five to seven). Thus, we provide proof-of-concept that convergent input from diverse types
of interneurons can provide feature-specific inhibition to relay cells.
Receptive field size
Receptive field sizes of interneurons and relay cells ranged from small and compact to large and
amorphous. We had not anticipated finding interneurons with small receptive fields. Previous work in
primate (Wang et al., 2011b; Wilson, 1989; Wilson et al., 1996) and cat (Martinez et al., 2014) showed
that interneurons had receptive fields larger than those of neighboring relay cells. This makes sense
45
because the former vastly outnumber the latter (Evangelio et al., 2018; Golding et al., 2014; Jager et al.,
2021; LeVay and Ferster, 1979; Montero, 1987) yet must represent the full extent of visual space.
Accordingly, more retinal afferents converge onto interneurons than onto relay cells (Hamos et al., 1985;
Humphrey and Weller, 1988; Morgan et al., 2016; Morgan and Lichtman, 2020). Further, unlike carnivore
(Humphrey and Weller, 1988; Van Horn et al., 2000) or primate (Wilson et al., 1996), the dendritic arbors
of murine interneurons extend over large regions of retinotopic space (Morgan et al., 2019; Seabrook et
al., 2013). Finally, strong synaptic excitation activates L-type calcium channels that propagate even
remote input to the soma (Acuna-Goycolea et al., 2008; Casale and McCormick, 2011), and see (Perreault
and Raastad, 2006).
Why then, might some interneurons have small receptive fields in the face of high retinal
convergence? In addition to the likelihood that only subset of presynaptic inputs dominate the neural
response at the soma (Litvina and Chen, 2017), it is possible that some retinal inputs are processed locally
within the dendritic arbor (Sherman, 2004), reminiscent of amacrine cells in retina. Interneurons in mouse
dLGN have long thin dendrites that branch extensively at locations far from the soma (Charalambakis et
al., 2019) and receive dense retinal input at these distal sites (Morgan and Lichtman, 2020). Further,
interneurons connect with other cells via dendrodentric synapses (Guillery, 1969; Morgan and Lichtman,
2020; Wilson, 1989), so they are able to release transmitter in the absence of spikes (Cox et al., 1998;
Errington et al., 2011). If retinal EPSPs were to attenuate before reaching the soma (Sherman, 2004) or
synaptic input introduce shunts along the dendritic length (Halnes et al., 2011; Perreault and Raastad,
2006), then single dendrites would function in isolation (Sherman, 2004). Since we record action
potentials, we would not have been able to detect the impact of compartmentalized dendrodentric
interactions. It is also possible that different stimuli activate different subsets of inputs, as we discuss in a
later section.
46
Receptive field structure
Prior to recording from murine interneurons, we had hypothesized that they might pool retinal
input provide nonselective inhibition to cortex, as is the case for many (though not all) interneurons in
murine visual cortex (Camillo et al., 2018; Kerlin et al., 2010; Liu et al., 2009; Runyan et al., 2010). Rather,
we found that the diversity of receptive field shapes for interneurons was almost as great as that for relay
cells though the distribution of receptive fields categories differed. For example, there was a much larger
representation of conventional On-Off cells in the interneuron population, perhaps a reflection of greater
retinal convergence than for relay cells. There were also proportionally more On or Off interneurons and
fewer with detectable center-surround profiles; one can imagine that this finding resulted from masking
or dilution of the surrounds by convergent input. Last, On-Off responses that were largely suppressed
rather than excited by the stimulus were observed only for relay cells, as might be expected because
interneurons with this profile would disinhibit rather than suppress their targets.
Sampling bias is always a consideration. We tried to mitigate this concern by recording responses
from interneurons and relay cells simultaneously. We also confirmed recording location post hoc; this
analysis indicated that we sampled much of the dLGN core but not the shell. Last, it is also possible that
some interneurons were suppressed by input by their neighbors before optogenetically-activated
currents reached threshold for spikes and were thus classified as relay cells. Given the small percentage
of interneurons in the dLGN, however, this would have been unlikely to change our results meaningfully.
Overall, this imbalance between the distribution of receptive field types of relay cells and interneurons,
and the small number of interneurons in mouse compared to other species, suggests that different types
of interneurons might converge onto a single postsynaptic target.
Receptive field constancy
Although variations in stimulus size or contrast alter neural response, it is rare that receptive field
structure changes categorically in carnivore and primate when On and Off pathways are intact (Archer et
47
al., 2021; Moore et al., 2011). There are, however, stimulus dependent changes in receptive field type in
murine retina (Goldin et al., 2021; Pearson and Kerschensteiner, 2015; Tikidji-Hamburyan et al., 2015;
Wienbar and Schwartz, 2018), some of which seem to propagate to dLGN (Tikidji-Hamburyan et al., 2015).
The sparse noise stimulus, along with our method to quantify a response as On, Off, On-Off, or
suppressed-by-contrast (the Bright-Dark Polarity Score) allowed us to address the question of receptive
field constancy in different stimulus regimes. We found that the receptive fields of 15% of interneurons
and 8% of relay cells changed as a function of stimulus size and/or contrast. We refer to these cells as
having adaptive receptive fields. Examples include simple shifts from On-Off responses to sole preference
for a single polarity, to more complex transformations such as from On to On-Off to Off responses. We
also observed changes that indicated a contribution from local inhibition.
The observation of adaptive receptive fields has bearing on the question of how high
retinogeniculate convergence in mouse is compatible with feature-selective receptive fields in dLGN and
suggests that the balance of retinal and intrathalamic inputs that different visual stimuli recruit can alter
receptive field type categorically. Relay cells whose feature selectivity changes as function of stimulus
strength or size might receive input from ganglion cells that have different stimulus requirements, such
that some inputs are active under stimulus conditions that fail to drive others (Liang et al., 2018).
Emergence of the suppressive component of the relay cell’s receptive field
The intuitive explanation for the stereotyped pattern of center-surround receptive fields and
push-pull responses recorded from relay cells in cat is that it is supplied by local interneurons, since these
too have center-surround receptive fields (Martinez et al., 2014; Wang et al., 2011b; Wang et al., 2007).
In mouse, however, the situation is less clear. The receptive field structures of relay cells (Durand et al.,
2016; Piscopo et al., 2013; Suresh et al., 2016) and interneurons (as we show here) are varied. Thus, we
built simple computational models of the interneuron receptive fields we mapped to ask if, in principle,
convergent input from these cells could generate suppressive receptive fields relay cells. We made simple
48
assumptions when developing the model framework. These were that all types of interneurons were
available to relay cells at all retinotopic positions, and interneurons with the same type of receptive varied
slightly in latency (as we had observed). We cycled through all interneuron models and chose the one that
contributed most to the suppressive component of a given relay cell’s receptive field. We then iterated
this process to select the additional inputs, which were summed linearly. Importantly, we excluded any
contributions with negative weights (which, in effect, are disinhibitory versus inhibitory). Thus, we were
able to identify pools of interneurons whose collective receptive field resembled the overall shape of the
suppressive component of a given relay cell’s receptive field.
Altogether, however, our approach provides proof-of-concept that convergent input from
interneurons with diverse types of receptive fields can explain the patterns of inhibition recorded from
relay cells in the murine dLGN. In the future, it will be important to include input from the thalamic
reticular nucleus, which provides feedback inhibition to relay cells, to the model and, perhaps, estimate
dendrodendritic contributions. We hope that our work will motivate future experiments that can resolve
inhibitory input to the thalamic receptive field directly, perhaps by imaging inhibitory boutons that
synapse on single dendrites cells in vivo as has been done for the retinogeniculate synapse (Liang et al.,
2018).
49
SUPPLEMENTARY MATERIAL
Figure 2.S1. Bright-Dark Polarity Scores and cells with prolonged inhibition to both bright and dark stimuli during priming
phase. (A) Bright-Dark polarity scores for Figure 2.4C-D. Markers with outlines and without outlines are data points calculated
from responses to high contrast and low-contrast stimuli respectively. Triangle, diamond, square, and circle markers are data
pointed calculated from responses to 5
O
, 10
O
, 20
O
, and 30
O
sparse noise stimuli respectively. Colors represent of quadrant markers
are located in. (B) Bright-Dark polarity scores for Figure 2.4E-F as in A. (C-E) Example of an On relay cell that has prolonged
inhibition to both bright and dark stimuli during the priming phase. (C) Bright-Dark polarity scores for example relay cell. (D) Θ
values show that this is an On cell across stimuli size and at high contrast. Cell did not respond to low contrast stimuli. (E) On (top)
and Off (bottom) spatiotemporal receptive field of relay cell to high-contrast 5
O
sparse noise stimuli. (F-H) Example of an
interneuron that is On in response to low contrast stimuli that has prolonged inhibition to both bright and dark stimuli during the
priming phase; conventions similar to C-E except H contains a spatiotemporal receptive field mapped from low contrast 20
O
sparse noise stimuli. (I) Pie-charts of different types of receptive fields of interneurons (n = 27) and relay cells (n = 205) in mouse
LGN (categorized by quantitative Bright-Dark Polarity Scores, quantitative assessment of priming phase, and qualitative
assessment of center-surround structure; cells were considered adaptive if Θ values changed across stimulus size and/or contrast;
shading indicates proportion of cells that are inhibited to both bright and dark stimuli during priming, interneurons (On-Center:
4%, Off-Center: 4%, Off: 4%, On-Off 4%), relay cells (On-Center: <1%, On: 1%, Off-Center: 3%, On-Off: 2%, Adaptive: <1%)).
50
Figure 2.S2. Dictionary of all interneuron LNP models
(without spatial and temporal translation). (A) Spatial
(left) and temporal (right) kernels for On interneurons.
(B) Spatial (left) and temporal (right) kernels for Off
interneurons. (C) On spatial (first column), On
temporal (second column), Off spatial (third column),
and Off temporal (fourth column) kernels for On-Off
cells. For A-C, bottommost LNP models can be found
in Figure 5. (D) Calculated correlation coefficient and
corrected variance explained between LNP models
and spatiotemporal STAs for 16 interneurons.
Corrected variance explained could only be calculated
for interneurons with three or more sparse noise
stimuli recordings (at least 48 repeats of black and
white squares on a 16x16 grid).
51
AUTHOR CONTRIBUTIONS
Alexis Gorin and Judith Hirsch designed the experiments. Alexis Gorin, Seohee Ahn, and Yizhan
Miao made recordings from interneurons and relay cells. Additional recordings from relay cells were made
previously by Vandana Suresh. Alexis Gorin developed quantitative tools to analyze the data. Yizhan Miao
created the proof-of-concept models with assistance from Friedrich Sommer. Yinan Su created the
pipeline for electrode track localization. Alexis Gorin and Judith Hirsch wrote the chapter.
52
CHAPTER 3 – TEMPORAL PRECISION
INTRODUCTION
Neurons encode information in both the overall rate and the precise temporal pattern of their
spike trains (Brenner et al., 2000; Butts et al., 2007; Koepsell and Sommer, 2008; Rathbun et al., 2010;
Reich et al., 1997; Reinagel and Reid, 2000; Schreiber et al., 2003; Uzzell and Chichilnisky, 2004). For
example, relay cells in the dorsal lateral geniculate nucleus of the thalamus (dLGN) are able to preserve
the temporal structure of retinal input and thus transmit more information to cortex than would be
possible using rate alone (Koepsell et al., 2009; Liu et al., 2001; Reinagel and Reid, 2000). At the same
time, relay cells edit retinal input such that they transmit visual information using only half as many action
potentials as ganglion cells do (Kaplan et al., 1987; Kara et al., 2000), and, thus, increase the efficiency of
the neural code.
Mechanisms such temporal summation (Carandini et al., 2007; Usrey et al., 1998; Wang et al.,
2010) and the refractory period (Berry and Meister, 1998; Reinagel and Reid, 2000; Uzzell and Chichilnisky,
2004) help explain how retinal input is filtered in part. Critically, inhibition plays an important role in
determining the structure of thalamic output. For example, inhibition timed to follow excitation at short
delay improves precision of a relay cell’s spike train (Butts et al., 2011). Studies of structures in which
inhibition plays a particularly dramatic role, such as the cerebellum, provide examples of how inhibition
operating over distributed timescales refines spike timing (Gauck and Jaeger, 2000).
To gain insight into pattern of inhibitory inputs that cell in dLGN receive, we compared the
temporal precision of local interneurons to that of relay cells in the murine dLGN, using optogenetic
approaches to distinguish cell types in vivo. While some interneurons fired with the same level of temporal
precision as relay cells, they were less reliable as a population, and this held true across stimulus strengths.
However, the relationship between spike rate and contrast was similar for both cell types. Theoretically,
53
our results suggest that interneurons, like relay cells, encode information about stimulus features using
both by rate and by timing. Mechanistically, temporally precise inhibitory input could provide a means to
enforce the tight temporal coupling between retina and dLGN by determining the length of the window
in which a an EPSP can trigger a spike, and conversely, a way to disinhibit the membrane at rapid
timescales as well. Input from interneurons with variable firing rates would regulate the membrane
voltage and conductance over slower timescales, and, therefore, determine the threshold required for
patterned retinal input to cross threshold and also suppress spurious firing. Taken together, our results
provide evidence that inhibition in the thalamus operates over a wide range of temporal scales and, thus,
has the potential to influence both the precision and rate of the spike trains that convey information to
cortex.
MATERIALS AND METHODS
Experimental models and subject details
Animals
All experiments were performed using mice ≥ 8 weeks old to avoid the visual critical period; there
was no detectable difference between males and females, data from both sexes were pooled. Subjects
were either Gad2-IRES-cre mice (n=28, Taniguchi et al., 2011, JAX 010802) or the progeny of these animals
crossed with Ai32 mice expressing ChR2(H134R)/EYFP (Madisen et al., 2012, JAX 012569) (Gad2-Ai32)
(n=37 for Gad2-Ai32 mice). All procedures were approved by the Institutional Animal Care and Use
Committees of the University of Southern California following guidelines from the National Institutes of
Health.
Method details
Virus expression
Opsins were introduced into GABAergic cells by crossing Gad2-IRES-cre and Ai32 lines or by
injection of opsin via adeno-associated viruses (AAV) directly into the dLGN of Gad2-IRES-cre mice. To
54
excite interneurons, we introduced channelrhodopsin using: AAV1-EF1a-double floxed-hChR2(H134R)-
EYFP-WPRE-HGHpA (Addgene: 20298-AAV1, provided by Karl Deisseroth, ≥ 7 x 10
12
vg/mL).
To inhibit cells, we introduced one of two opsins: AAV1-hSyn1-SIO-stGtACR2-FusionRed
(Addgene: 105677-AAV1, provided by Ofer Yizhar (Mahn et al., 2018), ≥ 1 x 10
13
vg/mL) or AAV1-Ef1a-DIO
eNpHR 3.0-EYFP (Addgene: 26966-AAV1, provided by Karl Deisseroth (Gradinaru et al., 2010), ≥ 1 x 10
13
vg/mL).
Surgical preparation for viral injection
For some experiments, we injected AAV into dLGN to restrict expression of opsins to local
interneurons, as follows: Gad2-IRES-cre mice were anesthetized with isoflurane prior to (2% in oxygen 2
L/min) and during (1% in oxygen 1 L/min) surgery. The anesthetized animal was positioned in a stereotaxic
device (Kopf) and a small craniotomy made above each dLGN at coordinates -2.2 mm anteroposterior and
± 2.2 mm lateral relative to bregma. Virus was injected via a glass micropipette filled with 2 μl of AAV
mixed with Fast Green (Sigma-Aldrich) lowered 2.5 mm below the brain’s surface. 100 nL of AAV was
delivered iontophoretically (Nanoject II, Drummond Scientific Company) for 5 min and the pipette
withdrawn 10 min after. The scalp was then sutured, and subjects were treated with Ketoprofen prior to
return to the home cage for recovery. Recordings were performed three to five weeks post-injection and
viral expression was confirmed postmortem by fluorescence microscopy.
Surgical preparation for recordings
Prior to electrophysiological recording animals were given an injection of chlorprothixene (5
mg/kg, i.p.) after which anesthesia was initiated and maintained with urethane (0.5-1 g/kg, 10% w/v in
saline, i.p.) (Ciftcioglu et al., 2020; Niell and Stryker, 2008; Suresh et al., 2016). After an incision was made
to expose the skull, a headpost was affixed caudal to the recording site and a small craniotomy over the
dLGN was made. All wound margins were infiltrated with bupivacaine and the brain and eyes were kept
moist with saline. Body temperature was measured using a rectal probe and maintained at 37
O
C.
55
Electrophysiological recording
Borosilicate glass pipettes were used to record from single neurons in whole-cell or cell-attached
mode, or a multi-conductor electrode was used to record from multiple cells simultaneously. Whole-cell
and cell-attached recordings were made with a Multiclamp 700B amplifier (Axon Instruments) in voltage-
clamp mode using standard techniques (Hirsch et al., 2003) and were digitized at 10 kHz with a Power
1401 data acquisition system (Cambridge Electronic Design); whole-cell recording in voltage-clamp mode
damped intrinsic conductances. Multi-conductor electrodes were single-shank, 32-channel, H6b probes
(Cambridge NeuroTech) connected with a 32-channel digital multiplexing headstage and a Digital Lynx
4SX-M acquisition system (Neuralynx); filters were set to 0.1 Hz and 9 kHz and the sampling rate was 30
kHz.
Optogenetic manipulation
The LED light source was either an LSD-1 (A-M Systems) or Optogenetics-LED-Dual (Prizmatix) that
was paired with a 200 μm fiber patch cable to a single cannula. Blue light (453 nm) was used to activate
channelrhodopsin or stGtACR2, and orange-red light (626 nm) was used to activate eNphR3.0. All light
from the fiber cannula was blocked to prevent it from reaching the either eye. For whole-cell and cell
attached recordings, we used an Optopatcher (Katz et al., 2013) (A-M Systems), in which an optical
cannula threaded through the pipette delivered light to the tip of the pipette. For multi-site recordings, a
Lambda-B optical fiber (100-core; 1.2 mm taper-tip) (Cambridge NeuroTech) that guided light to the tip
of each shank was fixed to the multi-electrode.
Stimulus presentation
All stimuli were generated using a ViSaGe stimulus generator (Cambridge Research Systems) and
displayed on a gamma corrected Dell U2211H LCD monitor at a 70 Hz refresh rate and a viewing distance
of 180 mm. Mice were shown sparse noise and naturalistic movie stimuli. The sparse noise stimulus
consisted of bright and dark squares, 5-30
o
, shown at 50% or 100% contrast sixteen to twenty times in
56
pseudorandom order on a 16x16 grid (5
o
grid resolution) (Ciftcioglu et al., 2020; Jones & Palmer, 1987;
Suresh et al., 2016). The naturalistic movies were displayed at minimum four contrasts (12.5%, 25%, 50%,
and 100%) up to six contrasts (12.5%, 25%, 37.5%, 50%, 75%, 100%), and up to three different movie
sequences were shown.
Histological reconstruction of recording sites and confirmation of viral expression
At the end of each experiment, the animal was perfused with 3% paraformaldehyde. The brain
was removed, placed in phosphate, and then cut in 100 μm coronal section using a vibratome. 100 μm
sections were mounted using ProLong Glass Antifade Mountant (ThermoFisher) and viewed using a
fluorescent (Zeiss) or confocal microscope. Micrographs were processed with FIJI (Schindelin et al., 2012).
For mice injected with AAV, we confirmed that opsin was expressed and limited to dLGN.
For many experiments, multi-site electrodes were coated with a yellow- or red-shifted lipophilic
tracer (DiI, ThermoFisher C7000, and DiD, ThermoFisher D12730 respectively). Thus, it was possible to
view the electrode track postmortem (Figure 3.1A, right). We then use the ‘allenCCF’ software package
(Shamash et al., 2018) to register our brain sections to coordinates the Allen Brain Mouse Atlas and
created a best-fit line for each track using orthogonal distance regression. After identifying the location of
each cluster along the electrode (see ‘Event Detection and Sorting’), we localized each cluster to a
stereotaxic coordinate in the dLGN. In cases for which the multi-site electrode was not coated with dye
but, nonetheless, left a visible track in the brain, we used the same method to localize recording sites.
For whole-cell or cell-attached recordings, visible tracks were compared to the Allen Brain Mouse
Atlas either using the ‘allenCCF’ software package or via Neurolucida (MBF Science). For recordings that
left no visible track (whole-cell, cell-attached, or a subset of multisite recordings), approximate cell or
electrode position was estimated from stereotaxic coordinates and depth measurements (Suresh et al.,
2016). These estimates were corroborated by the experiments with dye-coated electrodes.
57
Quantification and statistical analysis
Optogenetic tagging of interneurons and relay cells
For animals in which interneurons expressed channelrhodopsin, we determined cell type from
responses to pulse trains of blue light of varied duration and frequency while the animal viewed a constant
full field bright, dark, or gray stimulus. Cells were classified as interneurons if they responded to LED pulses
within 10 ms (Figure 2.1C, G) or as relay cells if they remained suppressed during pulses ≤ 250 ms (Figure
2.1K). Periods of suppression during LED stimulation in multi-electrode recordings of relay cells (Figure
2.1H-I) is indicative of strong inhibition seen in whole-cell recordings of relay cells during LED stimulation
(Figure 2.1D). For animals in which interneurons expressed an inhibitory opsin, it was necessary to excite
cells with a visual stimulus and then assess optogenetic modulation of the visually evoked response. We
used full-field stimulus sequences that comprised 3 luminance steps, either bright-dark-bright or dark-
bright-dark. The LED was switched on for the 70% of duration of the central luminance step and then
ramped down up to 30% of the duration of the central luminance step past the end of the step. Cells were
defined as interneurons if the LED suppressed the visual response or as relay cells if the visual response
was enhanced (Figure 2.1J-M).
Event detection and sorting
For whole-cell recordings of membrane currents, EPSCs and spikes were detected via techniques
we used previously (Suresh et al., 2016; Wang et al., 2011b; Wang et al., 2007). Briefly, we applied an
adaptive threshold to the first derivative of the intracellular signal such that the smallest potential events
included both EPSCs and noise. These events were then sorted using a support vector machine (SVM)
(Chang and Lin, 2013) trained with randomly selected events that were manually labeled as EPSCs or
noise. Because events near the decision boundary were prone to misclassification, we labeled these
manually for additional training and then reclassified the dataset. Finally, spikes were sorted from the
EPSCs by repeating the algorithm exclusively on SVM-classified neural events.
58
Using Kilosort2 (Pachitariu et al., 2016) we sorted spikes into clusters that we manually assessed
with the visualization tool Phy (Rossant et al., 2016) to determine if single clusters represented activity
from just one versus multiple neurons, and also if spikes from a single cell had been divided into several
clusters that should be merged. We assigned depth for each unit based the position of the conductor that
recorded the largest amplitude waveform. In some cases, artifact from the optical stimulus interfered
with recording of neural signals. To address this, we used a local polynomial approximation algorithm
known as SALPA to remove LED-induced artifact without distorting the individual spike waveforms
(Wagenaar and Potter, 2002).
Receptive field mapping
Determining receptive field structure
Receptive field maps were constructed from responses to sparse noise by calculating the spike-
triggered average (STA) of the stimulus ensemble (Schwartz et al., 2006) for time windows up to 300
milliseconds (t = -300ms ) prior to the spike (t = 0). Thus, we estimated the spatiotemporal receptive field
for each cell. We then utilized a method from Chapter 2 to categorize the receptive field preferences for
each cell which includes: On, Off, On-Off, Suppressed-by-Contrast-Like. We also determined if the priming
phase was suppressed by both bright and dark stimuli by determining if the integral of the priming phase
was less than zero for responses to both stimuli.
Determining receptive field size
To measure receptive field size, we used standard techniques (Wang et al., 2007). We fit the
sparse noise generated maps with 2D Gaussian functions and used the 1σ contours to define the radius
of center sub-regions or single regions. For cells that were excited or inhibited by both bright and dark
stimuli at similar time points, the map for the dominant contrast was used to determine receptive field
size. Specifically, the sizes of the receptive fields were quantified using the average of the semi-major and
59
semi-minor axes of the 2D fits and by the area of these fits. The receptive fields of several cells could not
be fit and are not included for comparisons with precision.
Temporal precision of neuronal responses
To quantify the temporal precision of neuronal responses across stimulus repetitions, we used a
reliability measure (Ciftcioglu et al., 2020; Schreiber et al., 2003) that reflects the correlation between
pairs of filtered spike trains (R corr) as follows:
𝑅 𝑐𝑜𝑟𝑟 =
2
𝑁 (𝑁 − 1)
∑ ∑
𝑆 𝑖 →∙
𝑆 𝑗 →
|𝑆 𝑖 ||𝑆 𝑗 |
𝑁 𝑗 =𝑖 +1
𝑁 𝑖 =1
Where N and s i indicate the number of stimulus repetitions ad the filtered spike trains for individual
repetitions, respectively. Note s i is computed as the convolution of the binary spike train with a Gaussian
filter set to have a standard deviation of 10 ms. This measured is referred to as Schreiber reliability, or
reliability, from now on. It takes values between zero and one; values close to 0 represent little correlation
between trials, and values close to 1 represent high correlation between trials.
To estimate the lower bounds for temporal precision and information content within a given spike
train, we created artificial spike trains based on a homogeneous Poisson process with constant firing rate.
We then ran simulations that spanned the range of physiological firing rates.
Because reliability is influenced by mean firing rate, we previously devised a model with only a
single free parameter to capture the relationship between reliability and mean firing as follows (for fuller
explanation, see Ciftcioglu et al., 2020):
𝑅 𝑐𝑜𝑟𝑟 (𝛼 , 𝑟 𝑖 ) =
𝑟 𝑥 𝛼 1 + 𝑟 𝑥 𝛼
60
In addition, we compared Schreiber Reliability against a well-known measure of neuronal variability: fano
factor, the ratio of the spike count variance to mean (Eden & Kramer, 2010; Fano, 1947; Geisler, 2008;
Shadlen & Newsome, 1998).
Statistics
All statistics were performed in MATLAB (Mathworks). For datasets in which two groups are
compared, data from each group was first tested for normality by means of the Shapiro-Wilk test (null
hypothesis is that the distribution is normal) and then for equal variance between the two groups by
means of the two-sample F-test (null hypothesis is that the two distributions come from normal
distributions with the same variance; MATLAB function ‘vartest2’). If both conditions were met, the two-
sample t-test (assumes normality and equal but unknown variance; MATLAB function ‘ttest2’) was used.
Otherwise, the non-parametric test used was the Wilcoxon rank sum test (equivalent to two-sample
Mann-Whitney U-test; MATLAB function ‘ranksum’) in addition to the two-sample Kolmogorov-Smirnov
test (MATLAB function ‘kstest2’). All statistical tests were two-tailed with sample size indicated in each
figure and/or legend. Significance levels were indicated by asterisks: p>=0.05; *; p<0.05; **: p<0.01; ***;
p<0.001; ****.
RESULTS
Our goal was to explore the temporal response properties of interneurons in the dLGN in the
context of understanding how inhibition contributes to the responses of relay cells. We used an
optogenetic approach to identify interneurons and relay cells (Chapter 2) and used multiconductor silicon
probes to record responses to various visual stimuli including naturalistic movies shown at different
contrasts. Our sample includes 19 interneurons and 38 relay cells distributed throughout the dLGN.
61
Temporal precision of relay cells versus interneurons
Retinal ganglion cells often fire with a high degree of temporal precision and, thus, can encode
information by means of spike timing as well as rate. Studies in cat (Koepsell et al., 2009; Wang et al.,
2010) and monkey (Sincich et al., 2007; Sincich et al., 2009) show relay cells lock to retinal spikes with
millisecond fidelity and that this tight temporal coupling of input and output provides a means to transmit
temporally encoded information downstream. Furthermore, computational studies show that inhibition
that arrives shortly after excitation improves precision allowed by excitatory mechanisms (Butts et al.,
2007, 2011). This finding raised the question of whether there was a source of temporally precise
inhibition in dLGN. Thus, we recorded responses of interneurons and relay cells to natural movies at
different contrasts. We found that there was a wide range in temporal precision of response for both cell
types; raster plots and PSTHs of spikes recorded from interneurons (green, Figure 3.1A, B) and relay cells
(black, Figure 3.1C, D) are shown alongside the corresponding receptive fields. The spike trains of cells
Figure 3.1. Interneurons fire less likely than chance than relay cells. (A-D) Responses of two interneurons and two relay cells to
a movie of snow monkeys at four contrasts. The receptive field of each cell is shown above rasters and PSTHs for each stimulus
condition. Examples in A and C fire more precisely than those in B and D. (E) Response reliability (Schreiber et al., 2003) plotted
against firing rate; each point was calculated from a cell’s response to one of three different movies (max three points per cell
per plot); dashed lines show best fits. Responses for interneurons are green (n = 14, 11, 14, 14 for 12.5%, 50%, 75%, 100% contrast
respectively; n = 58 total) and for relay cells, black (n = 31, 30, 33, 43 for 12.5%, 50%, 75%, 100% contrast respectively; n = 137
total); simulated Poisson spike trains are gray. Data in A-D are represented by open circles. (F) Pie-charts, the fraction of cells that
fire below vs above Poisson reliability. Relay cells, black (n = 31, 30, 33, 43 for 12.5%, 50%, 75%, 100% contrast respectively; n =
137 total) and interneurons, green (n = 14, 11, 14, 14 for 12.5%, 50%, 75%, 100% contrast respectively; n = 58 total).
62
depicted in Figures 3.1A and 3.1C were similar from one trial to the next (this is easiest to see at lower
contrasts, when firing rate is slow), whereas spike timing varied markedly across trials the cells depicted
in Figures 3.1B and 3.1D.
In order to quantify the differences in temporal precision across cell types and to disambiguate
firing rate from precision, we used a correlation-based reliability measure (Schreiber et al., 2003) (see
Methods). This index reflects cross trial differences in spike timing; values range between 0 and 1, with 1
indicating maximal precision. Because the measure is influenced by firing rate, we chose to plot index
values against this parameter and display the results for the entire dataset divided according to cell type
and stimulus contrast (Figure 3.1E). Each point in each plot represents an index value computed for an
individual cell’s response to a single movie stimulus (maximum number of points per cell is three). We
determined best fit for each distribution (dashed black and green lines) using maximum likelihood
estimation and compared these fits to that for random firing, simulated with a homogenous Poisson spike
train (gray solid lines). The points that lie above the gray line fire more reliably than would be predicted
by a random spike generation process. In general, relay cells fired more precisely than interneurons, and
fits to the data from both populations lay above the Poisson boundary (except for responses of
interneurons at the lowest contrast). To give an intuitive illustration of this result, we plotted the data as
pie charts (Figure 3.1F). These charts show that as contrast increases, the proportion of relay cells that
fire more reliably than chance continues to grow. The situation for interneurons was different. There was
a pronounced increase in reliability as contrast stepped from 12.5% to 25%, but only negligible change at
higher contrasts. Further, at every contrast, a higher percentage of relay cells than interneurons fired
reliably. This difference was so pronounced that percentage of interneurons that fired reliably at the
highest contrast was lesser than the value for relay cells at the lowest contrast.
63
We next asked whether there were a relationship between firing rate and precision, using violin
plots to visualize the relative shape of the distributions for relay cells and interneurons (Figure 3.2). This
comparison served to disambiguate the influence of the former on the latter. At most contrasts (12.5%,
50% and 100%) there was a significant difference in the Schreiber values between relay cells and
interneurons (Figure 3.2A) (with a larger sample, the trend at 25% would likely have achieved
Figure 3.2. As a population, interneurons fire less reliably than relay cells at most contrasts although firing rates for both cell
types are indistinguishable for all contrasts. (A) Box and violin plots of reliability index values for interneurons and relay cells at
different contrasts (100% contrast: p = 0.029, n = 19 for interneurons, n = 42 for relay cells; 50% contrast: p = 0.009, n = 14 for
interneurons, n = 31 for relay cells; 25% contrast: p = 0.102, n = 11 for interneurons and n = 29 for relay cells; 12.5% contrast: p
= 0.0474, n = 14 interneurons and n = 29 for relay cells; Wilcoxin rank-sum). (B) Conventions and cell number as in A for firing
rate (12.5% - 100% contrast: p>0.05). (C) Conventions and cell number as in A for Fano Factor (12.5% - 100% contrast: p>0.05).
64
significance). Conversely, there was no difference between the increase in firing rate with contrast
between the two cell types (Figure 3.2B). Last, we computed Fano factor for the dataset (Figure 3.2C).
This measure assesses cross-trial variability in rate versus the cross-trial temporal correlation that the
Schreiber measure expresses. There was no significant difference between populations.
Variation in the relationship between contrast and reliability
While a monotonic increase in reliability with contrast was clear cut at the population level, there
were a few relay cells (5/33) whose responses clearly deviated from this pattern. The spike trains of these
cells reached peak reliability at half (50%) contrast, and the Schreiber value decreased by more than
greater than 20% at full (100%), contrast. This finding is illustrated in the form of a scatter plot and a
histogram for the full sample of cells (nonmonotonic cells are indicated by the points above unity slope in
the scatter plot and entries to left of the dashed vertical line in the histogram) (Figure 3.3A). Example
rasters from a cell with a nonmonotonic relationship between contrast and reliability, and one with a
monotonic relationship, illustrate changes in firing pattern at different stimulus strengths (Figures 3.3B,
C).
It was possible that our population results obscured differences between cells with different
receptive field types. We computed receptive fields by generating spike-triggered averages of neural
responses to sparse noise and classified them as On or Off (excited by bright but not dark stimuli and vice
versa, On-Off (excited by stimuli of both contrasts presented in overlapping regions of visual space) or
suppressed-by-contrast (excited by stimuli of both contrasts presented in overlapping regions of visual
65
space). Plots of reliability against contrast for each population revealed no apparent differences across
receptive field type (Figure 3.4).
Figure 3.3. Relay cells can a have non-monotonic contrast-reliability relationship. (A) Interneuron (green) and relay cell (black)
reliability at 100% contrast against their respective reliability scores at 50% contrast. Inset includes histograms of interneurons’
and relay cells’ relative difference between their reliability at 100% and 50% contrast. (B) Responses (raster plots and PSTHs) of
a relay cell with a non-monotonic contrast-reliability relationship at 100% contrast (top), 50% contrast (middle), and 25% contrast
(bottom). Reliability scores are in the upper-right hand corner of each response. (C) Responses of a relay cell with a monotonic
contrast-reliability relationship. Conventions follow B. (D) Reliability at maximal contrast against each interneuron’s (green) and
relay cell’s (black) relative difference between its reliability at maximal and minimal contrast (for relay cells, R = 0.369, p = 0.049,
n = 31; for interneurons, R = 0.202, p = 0.489, n = 14; Pearson’s correlation). (E) Responses (raster plots and PSTHs) at maximal
(top) and minimal (bottom) contrast for a relay cell that has high reliability (R 100 = 0.63) at maximal contrast. Reliability scores
are in the upper-right hand corner of each response. (F) Responses for a relay cell that has low reliability (R 100 = 0.26) at maximal
contrast. Conventions follow E.
66
Relationship between receptive field size and precision
We compared receptive field structure and spike timing of interneurons and relay cells to ask if
spatial and temporal precision covaried. We estimated the size of receptive fields with 2D Gaussian fits
and plotted the resulting values against the reliability score at maximum contrast. There was a significant
correlation for relay cells (R = -0.421, p = 0.009, n = 37) but not for interneurons (R = -0.134, p = 0.634, n
= 14) (Figure 3.5A-C), even though some interneurons with smaller fields fired precisely (Figure 3.5B, top)
and others with larger fields fired almost randomly (Figure 3.5B, bottom).
Figure 3.4. Contrast-reliability responses organized by receptive field types of interneurons and relay cells. (A) On interneurons
(green) and relay cell (black) reliability plotted across all contrasts recorded for each cell. Empty circles represent cells that
additionally are inhibited by both bright and dark stimuli during the priming phase (see Chapter 2). (B–D) Conventions as in A,
except for Off (B), On-Off (C), and SBC (D) interneurons and relay cells.
Figure 3.5. Relay cell precision is larger for cells with smaller receptive fields. (A) Reliability at maximal contrast against
receptive field size (calculated via 2D Gaussian fits; see Methods) for interneurons (green, R = -0.134, p = 0.634, n = 14; Pearson’s
correlation) and relay cells (black, R = -0.421, p = 0.009, n = 37; Pearson’s correlation). Empty circles indicate responses seen in
B and C. (B) Responses (raster plots and PSTHs) of an interneuron to natural movies at maximum contrast with a small receptive
field (top, R 100 = 0.59) and a large field (bottom, R 100 = 0.09). Yellow squares indicate stimulus size; grid spacing, 5°. (C)
Conventions as in B except for relay cells (top, R 100 = 0.62; bottom, R 100 = 0.38).
67
DISCUSSION
By regulating the firing rate and timing of their postsynaptic partners, local interneurons in
thalamus have the potential to influence the amount of information relay cells send to cortex. To explore
this topic, we made recordings from optogenetically identified interneurons and relay cells in the murine
dLGN and compared the temporal precision of responses evoked by natural movies at different contrasts.
While some interneurons fired as precisely as relay cells, they were less reliable as a population across
stimulus strengths. Reliability increased steadily with contrast for most relay cells but saturated with the
step from the lowest contrast to the next for interneurons. Firing rates were similar for both populations,
however. Thus, in principle, these interneurons are variously able to provide asynchronous inhibition that
generates subtle shifts in membrane voltage and conductance to regulate spike threshold, as well as
precisely timed inhibition that rapidly open and closes the windows in which spikes can fire.
Retinal contributions to precision
When a retinal input crosses threshold, the postsynaptic relay cell fires within milliseconds
(Koepsell et al., 2009; Usrey et al., 1999). This tight coupling provides a means to transmit information
encoded in the fine timing of retinal spike trains (Kara et al., 2000; Liu et al., 2001; Reinagel and Reid,
2002). It is likely that the coupling mechanism depends on membrane dynamics engaged by the sharp rise
of the retinogeniculate EPSP. Many more ganglion cells converge onto single thalamic cells in mouse
(Hammer et al., 2015; Litvina and Chen, 2017; Morgan et al., 2016; Rompani et al., 2017) than cat (Cleland
et al., 1971; Usrey et al., 1999) and monkey (Sincich et al., 2007). This profound species difference suggests
convergent EPSPs might blend into a smooth compound event and, thus, lose the capacity to promote
precise firing. There is evidence, however, that at least for some cell relay cells, a subset of many retinal
inputs generate large EPSPs and dominate the neural response (Litvina and Chen, 2017). This observation
might explain our finding that at least some murine relay cells and interneurons fire precisely.
68
As a population, interneurons fired less reliably than relay cells. This difference might reflect the
fact that interneurons pool even more retinal inputs than relay cells (Seabrook et al., 2013) and fewer
dominant afferents. Further, retinal afferents terminate on stout proximal dendrites of relay cells and are
electrotonically close to the soma. By contrast retinal axons often synapse with the distal dendrites of
interneurons (Morgan and Lichtman, 2020), which are long, thin and electrotonically remote. These distal
inputs reach the soma indirectly, propgated by L-type calcium channels. This interaction decouples the
timing of the retinal input and thalamic spike (Acuna-Goycolea et al., 2008) and provides an additional
mechanism to explain the difference in precision between relay cells and interneurons.
Inhibitory contributions to precision
Temporal precision can also be modulated by extraretinal sources. Computational studies show
how inhibition following excitation with short delay increases precision in relay cells’ responses by
narrowing the window in which a cell can fire (Butts et al., 2011). Temporally precise inhibitory inputs
could help fine tune the delay as follows. Since relay cells receive input from many interneurons, the
relative timing of these inputs will influence the strength and duration of their summed impact. That is,
coincident input should create a fast and strong IPSP that stops excitation in its tracks.
Asynchronous inhibitory inputs can also enhance reliability by hyperpolarizing and shunting the
membrane such that only stronger inputs cross threshold. Further, strong or lasting hyperpolarization de-
inactivates the low threshold, T-type calcium channels that promote bursts (Huguenard and McCormick,
1992; Lu et al., 1992; Scharfman et al., 1990). Thus, the amount of inhibition that a relay cell receives
determines whether it will respond to a retinal input with single spike or a burst.
An interneuron influences spike timing in postsynaptic cells whether it fires a spike or remains
silent. In vitro studies of the deep cerebellar nuclei, whose precisely timed output is determined by
patterned inhibitory input from Purkinje cells, demonstrate the impact of withdrawal of inhibition on spike
69
timing; even slight reductions in the membrane conductance and depolarization greatly increase the
probability that excitatory input crosses threshold (Gauck and Jaeger, 2000). The degree of temporal
overlap between inhibitory inputs regulates this disinhibitory process. It is easy to imagine that a similar
scheme applies to dLGN and other brain regions.
Reliability and contrast
Most studies of contrast response functions focus on the contrast gain control of firing rate. In
essence, the ratio of firing rate to stimulus contrast grows smaller as contrast increases, allowing cells to
operate efficiently over a wide dynamic range. Here we focused on the relationship between stimulus
contrast and precision. We found that increments in contrast improves precision for both cell types, but
the reliability of interneuronal firing above chance saturates early. Our finding is particularly interesting
in the context of contrast dependent changes in “efficacy”, the probability that a single retinal spike will
trigger a postsynaptic relay cell. Efficacy values are larger at low versus high contrasts (Alitto et al., 2019).
One hypothesis to explain this behavior is that when stimulus strength is weak, synaptic drive is sparse
and membrane conductance commensurately low, thus, heightening sensitivity to even weak excitatory
inputs (Alitto et al., 2019). In this low contrast regime, inhibition might play a role by suppressing spurious
spikes and preserving the fidelity of transmission from the eye to the brain.
70
CHAPTER 4 – CONCLUSION
Visual information encoded by the retina is sent to the lateral geniculate nucleus of the thalamus
(LGN) before reaching the primary visual cortex. Within the LGN, retinal information is subject to
inhibitory influences from the local interneurons and from the TRN. Understanding how these inhibitory
circuits operate during vision is therefore critical to our understanding of the function of the LGN in the
early visual pathway. In this thesis, I utilized optogenetics to tag interneurons and relay cells in mice and
then investigated the visual properties of interneurons to determine how they could contribute to vision.
In my first project, I explored the receptive field properties of interneurons and relay cells in the
murine dLGN and see how they compared. Interneurons receptive field were diverse in both type and size
like relay cells. In addition, by presenting stimuli of different stimulus strengths, I found a group of
interneurons and relay cells whose stimulus luminance preference changed with contrast and size. This
had previously only been reported by (Tikidji-Hamburyan et al., 2015) in the context of similar cells found
in retina. We then found, using computational analyses, that ensembles of interneuron receptive fields
could correlate with patterns of spatiotemporal inhibition seen in relay cells in response to sparse noise
stimuli. Thus, my work shows that interneurons can supply feature-selective inhibition to relay cells.
In my second project, I explored the temporal precision of murine interneurons and relay cells,
which underlies that rationale that firing patterns in relay cells can be explained by inhibition. I found that
interneurons have a comparable range of temporal precision to relay cells, but, as a population, are less
precise than relay cells over contrast. Mechanisms for precision in downstream targets from basal ganglia
offer potential roles for the range of precision in interneurons.
Taken together, interneurons are a strong candidate to explain many of the patterns of inhibition
seen in relay cells. Further studies that perturb neural activity (such as optogenetic stimulation) is needed
to investigate the role of interneurons, including the role of axonal versus dendrodendritic inhibition.
71
REFERENCES
Acuna-Goycolea, C., Brenowitz, S.D., and Regehr, W.G. (2008). Active Dendritic Conductances Dynamically
Regulate GABA Release from Thalamic Interneurons. Neuron 57, 420-431. 10.1016/j.neuron.2007.12.022.
Ahlsen, G., Lindstrom, S., and Lo, F.S. (1985). Interaction between inhibitory pathways to principal cells in
the lateral geniculate nucleus of the cat. Experimental brain research 58, 134-143. 10.1007/bf00238961.
Alitto, H.J., Rathbun, D.L., Fisher, T.G., Alexander, P.C., and Usrey, W.M. (2019). Contrast gain control and
retinogeniculate communication. European Journal of Neuroscience 49, 1061-1068. 10.1111/ejn.13904.
Alitto, H.J., and Usrey, W.M. (2005). Dynamic properties of thalamic neurons for vision. Progress in brain
research 149, 83-90. 10.1016/S0079-6123(05)49007-X.
Alonso, J.M., Usrey, W.M., and Reid, R.C. (2001). Rules of connectivity between geniculate cells and simple
cells in cat primary visual cortex. Journal of Neuroscience 21, 4002-4015. 10.1523/jneurosci.21-11-
04002.2001.
Arcelli, P., Frassoni, C., Regondi, M.C., Biasi, S.D., and Spreafico, R. (1996). GABAergic neurons in
mammalian thalamus: A marker of thalamic complexity? Brain Research Bulletin 42, 27-37.
10.1016/S0361-9230(96)00107-4.
Archer, D.R., Alitto, H.J., and Usrey, W.M. (2021). Stimulus Contrast Affects Spatial Integration in the
Lateral Geniculate Nucleus of Macaque Monkeys. The Journal of Neuroscience, JN-RM-2946-2920.
10.1523/JNEUROSCI.2946-20.2021.
Arthur, D., and Vassilvitskii, S. (2006). k-means++: The advantages of careful seeding. Stanford.
Babadi, B., Casti, A., Xiao, Y., Kaplan, E., and Paninski, L. (2010). A generalized linear model of the impact
of direct and indirect inputs to the lateral geniculate nucleus. Journal of Vision 10, 1-14.
10.1167/10.10.22.Introduction.
Barlow, H.B. (1961). The coding of sensory messages. In T.a. Zangwill, ed. (Cambridge University Press),
pp. 331-360.
72
Berry, M.J., and Meister, M. (1998). Refractoriness and neural precision. The Journal of Neuroscience 18,
2200-2211.
Berson, D. (2010). Retinal Ganglion Cell Types and Their Central Projections. The Senses: A Comprehensive
Reference 1, 491-519. 10.1016/B978-012370880-9.00280-2.
Bickford, M.E., Slusarczyk, A., Dilger, E.K., Krahe, T.E., Kucuk, C., and Guido, W. (2010). Synaptic
development of the mouse dorsal lateral geniculate nucleus. Journal of Comparative Neurology 518, 622-
635. 10.1002/cne.22223.
Bickford, M.E., Wei, H., Eisenback, M.A., Chomsung, R.D., Slusarczyk, A.S., and Dankowsi, A.B. (2008).
Synaptic organization of thalamocortical axon collaterals in the perigeniculate nucleus and dorsal lateral
geniculate nucleus. The Journal of comparative neurology 508, 264-285. 10.1002/cne.21671.
Blitz, D.M., and Regehr, W.G. (2005). Timing and specificity of feed-forward inhibition within the LGN.
Neuron 45, 917-928. 10.1016/j.neuron.2005.01.033.
Bonin, V., Histed, M.H., Yurgenson, S., and Reid, R.C. (2011). Local diversity and fine-scale organization of
receptive fields in mouse visual cortex. Journal of Neuroscience 31, 18506-18521.
10.1523/JNEUROSCI.2974-11.2011.
Brenner, N., Strong, S.P., Koberle, R., Bialek, W., and Steveninck, R.R.D.R.V. (2000). Synergy in a Neural
Code. Neural Computation 1552, 1-22. 10.1162/089976600300015259.
Brickley, S.G., and Mody, I. (2012). Extrasynaptic GABA(A) receptors: their function in the CNS and
implications for disease. Neuron 73, 23-34. 10.1016/j.neuron.2011.12.012.
Bright, D.P., Aller, M.I., and Brickley, S.G. (2007). Synaptic Release Generates a Tonic GABA A Receptor-
Mediated Conductance That Modulates Burst Precision in Thalamic Relay Neurons. 27, 2560-2569.
10.1523/JNEUROSCI.5100-06.2007.
Butts, D.A., Cui, Y., and Casti, A.R.R. (2016). Nonlinear computations shaping temporal processing of
precortical vision. Journal of Neurophysiology 116, 1344-1357. 10.1152/jn.00878.2015.
73
Butts, D.A., Weng, C., Jin, J., and Alonso, J.-M. (2011). Temporal precision in the visual pathway through
the interplay of excitation and stimulus-driven suppression. Journal of Neuroscience 31, 11313-11327.
10.1523/JNEUROSCI.0434-11.2011.
Butts, D.A., Weng, C., Jin, J., Yeh, C.-I., Lesica, N.a., Alonso, J.-M., and Stanley, G.B. (2007). Temporal
precision in the neural code and the timescales of natural vision. Nature 449, 92-95.
10.1038/nature06105.
Camillo, D., Ahmadlou, M., Saiepour, M.H., Yasaminshirazi, M., Levelt, C.N., and Heimel, J.A. (2018). Visual
Processing by Calretinin Expressing Inhibitory Neurons in Mouse Primary Visual Cortex. Scientific Reports
8, 1-13. 10.1038/s41598-018-30958-w.
Carandini, M., Horton, J.C., and Sincich, L.C. (2007). Thalamic filtering of retinal spike trains by postsynaptic
summation. Journal of Vision 7, 1-11. 10.1167/7.14.20.
Casale, A.E., and McCormick, D.A. (2011). Active Action Potential Propagation But Not Initiation in
Thalamic Interneuron Dendrites. The Journal of Neuroscience 31, 18289. 10.1523/JNEUROSCI.4417-
11.2011.
Casti, A., Hayot, F., Xiao, Y., and Kaplan, E. (2008). A simple model of retina-LGN transmission. Journal of
computational neuroscience 24, 235-252. 10.1007/s10827-007-0053-7.
Chandler, D.M., and Field, D.J. (2007). Estimates of the information content and dimensionality of natural
scenes from proximity distributions. Journal of the Optical Society of America. A, Optics, image science,
and vision 24, 922-941. 10.1364/josaa.24.000922.
Chang, C.-c.C., and Lin, C.-j.J. (2013). LIBSVM : A Library for Support Vector Machines. ACM Transactions
on Intelligent Systems and Technology (TIST) 2, 2:27.21-27.27. 10.1145/1961189.1961199.
Charalambakis, N.E., Govindaiah, G., Campbell, P.W., and Guido, W. (2019). Developmental Remodeling
of Thalamic Interneurons Requires Retinal Signaling. The Journal of Neuroscience 39, 3856-3866.
10.1523/jneurosci.2224-18.2019.
Cheong, S.K., Tailby, C., Solomon, S.G., and Martin, P.R. (2013). Cortical-like receptive fields in the lateral
geniculate nucleus of marmoset monkeys. Journal of Neuroscience 33, 6874-6876.
10.1523/JNEUROSCI.5208-12.2013.
74
Chichilnisky, E.J. (2001). A simple white noise analysis of neuronal light responses. Network 12, 199-213.
10.1080/713663221.
Ciftcioglu, U.M., Suresh, V., Ding, K.R., Sommer, F.T., and Hirsch, J.A. (2020). Visual information processing
in the ventral division of the mouse lateral geniculate nucleus of the thalamus. The Journal of
Neuroscience, JN-RM-2602-2619. 10.1523/JNEUROSCI.2602-19.2020.
Cleland, B.G., Dubin, M.W., and Levick, W.R. (1971). Simultaneous Recording of Input and Output of
Lateral Geniculate Neurones. Nature New Biology 231, 191-192. 10.1038/newbio231191a0.
Cleland, B.G., and Lee, B.B. (1985). A comparison of visual responses of cat lateral geniculate nucleus
neurones with those of ganglion cells afferent to them. The Journal of Physiology 369, 249-268.
10.1113/jphysiol.1985.sp015899.
Cope, D.W., Hughes, S.W., and Crunelli, V. (2005). GABA A Receptor-Mediated Tonic Inhibition in Thalamic
Neurons. 25, 11553-11563. 10.1523/JNEUROSCI.3362-05.2005.
Cox, C.L. (2014). Complex regulation of dendritic transmitter release from thalamic interneurons. Current
Opinion in Neurobiology 29, 126-132. 10.1016/j.conb.2014.07.004.
Cox, C.L., and Beatty, J.A. (2017). The multifaceted role of inhibitory interneurons in the dorsal lateral
geniculate nucleus. Visual Neuroscience 34, E017-E017. 10.1017/S0952523817000141.
Cox, C.L., Zhou, Q., and Sherman, S.M. (1998). Glutamate locally activates dendritic outputs of thalamic
interneurons. July 394, 478-482. 10.1038/nature01372.1.
Crandall, S.R., and Cox, C.L. (2012). Local Dendrodendritic Inhibition Regulates Fast Synaptic Transmission
in Visual Thalamus. Journal of Neuroscience 32, 2513-2522. 10.1523/JNEUROSCI.4402-11.2012.
Cucchiaro, J.B., Uhlrich, D.J., and Sherman, S.M. (1991). Electron‐microscopic analysis of synaptic input
from the perigeniculate nucleus to the A‐laminae of the lateral geniculate nucleus in cats. Journal of
Comparative Neurology 310, 316-336. 10.1002/cne.903100304.
Dan, Y., Atick, J., and Reid, R.C. (1996). Efficient Coding of Natural Scenes in the Lateral Geniculate Nucleus
: Experimental Test of a Computational Theory. J. Neurosci 16, 3351-3362.
75
Denning, K.S., and Reinagel, P. (2005). Visual control of burst priming in the anesthetized lateral geniculate
nucleus. The Journal of neuroscience : the official journal of the Society for Neuroscience 25, 3531-3538.
10.1523/JNEUROSCI.4417-04.2005.
Dhande, O.S., Stafford, B.K., Lim, J.-H.A., and Huberman, A.D. (2015). Contributions of Retinal Ganglion
Cells to Subcortical Visual Processing and Behaviors. Annual Review of Vision Science 1, 291-328.
10.1146/annurev-vision-082114-035502.
Dubin, M.W., and Cleland, B.G. (1977). Organization of visual inputs to interneurons of lateral geniculate
nucleus of the cat. Journal of neurophysiology 40, 410-427.
Durand, S., Iyer, R., Mizuseki, K., de Vries, S., Mihalas, S., and Reid, R.C. (2016). A Comparison of Visual
Response Properties in the Lateral Geniculate Nucleus and Primary Visual Cortex of Awake and
Anesthetized Mice. Journal of Neuroscience 36, 12144-12156. 10.1523/JNEUROSCI.1741-16.2016.
Eglen, S.J., Diggle, P.J., and Troy, J.B. (2005). Homotypic constraints dominate positioning of on- and off-
center beta retinal ganglion cells. Visual neuroscience 22, 859-871. 10.1017/S0952523805226147.
Enroth-Cugell, C., and Robson, J.G. (1966). The contrast sensitivity of retinal ganglion cells of the cat. The
Journal of physiology 187, 517-552. 10.1113/jphysiol.1966.sp008107.
Errington, A.C., Di Giovanni, G., Crunelli, V., and Cope, D.W. (2011). mGluR Control of Interneuron Output
Regulates Feedforward Tonic GABA(A) Inhibition in the Visual Thalamus. Journal of Neuroscience 31,
8669-8680. 10.1523/jneurosci.0317-11.2011.
Evangelio, M., García-Amado, M., and Clascá, F. (2018). Thalamocortical Projection Neuron and
Interneuron Numbers in the Visual Thalamic Nuclei of the Adult C57BL/6 Mouse. Frontiers in
Neuroanatomy 12, 27-27. 10.3389/fnana.2018.00027.
Field, D.J. (1987). Relations between the statistics of natural images and the response properties of
cortical cells. Journal of the Optical Society of America. A, Optics and image science 4, 2379-2394.
10.1364/josaa.4.002379.
Gauck, V., and Jaeger, D. (2000). The Control of Rate and Timing of Spikes in the Deep Cerebellar
Nuclei by Inhibition. Journal of Neuroscience 20, 3006-3016.
76
Goldin, M.A.M.A., Lefebvre, B., Virgili, S., Ecker, A., Mora, T., Ferrari, U., and Marre, O. (2021). Context-
dependent selectivity to natural scenes in the retina. bioRxiv, 2021.2010.2001.462157-
462021.462110.462101.462157. 10.1101/2021.10.01.462157.
Golding, B., Pouchelon, G., Bellone, C., Murthy, S., Di Nardo, A.A., Govindan, S., Ogawa, M., Shimogori, T.,
Lüscher, C., Dayer, A., and Jabaudon, D. (2014). Retinal input directs the recruitment of inhibitory
interneurons into thalamic visual circuits. Neuron 81, 1057-1069. 10.1016/j.neuron.2014.01.032.
Gradinaru, V., Zhang, F., Ramakrishnan, C., Mattis, J., Prakash, R., Diester, I., Goshen, I., Thompson, K.R.,
and Deisseroth, K. (2010). Molecular and Cellular Approaches for Diversifying and Extending Optogenetics.
Cell 141, 154-165. 10.1016/j.cell.2010.02.037.
Grubb, M.S., and Thompson, I.D. (2003). Quantitative characterization of visual response properties in the
mouse dorsal lateral geniculate nucleus. Journal of Neurophysiology 90, 3594-3607.
10.1152/jn.00699.2003.
Guillery, R.W. (1969). A Quantitative Study of Synaptic Interconnections in the Dorsal Lateral Geniculate
Nucleus of the Cat. Zeitschrift für Zellforschung und Mikroskopische Anatomie 96, 39-48.
Haefner, R.M., and Cumming, B.G. (2009). An improved estimator of Variance Explained in the presence
of noise. Advances in Neural Information Processing Systems 21 - Proceedings of the 2008 Conference,
585-592.
Halnes, G., Augustinaite, S., Heggelund, P., Einevoll, G.T., and Migliore, M. (2011). A multi-compartment
model for interneurons in the dorsal lateral geniculate nucleus. PLoS Computational Biology 7.
10.1371/journal.pcbi.1002160.
Hammer, S., Monavarfeshani, A., Lemon, T., Su, J., and Fox, M.A. (2015). Multiple Retinal Axons Converge
onto Relay Cells in the Adult Mouse Thalamus. Cell Reports 12, 1575-1583. 10.1016/j.celrep.2015.08.003.
Hamos, J.E., Van Horn, S.C., Raczkowski, D., and Sherman, S.M. (1987). Synaptic circuits involving an
individual retinogeniculate axon in the cat. Journal of Comparative Neurology 259, 165-192.
https://doi.org/10.1002/cne.902590202.
77
Hamos, J.E., Van Horn, S.C., Raczkowski, D., Uhlrich, D.J., Sherman, S.M., James E. Hamos,
S.C.V.H.D.R.D.J.U., and Sherman, S.M. (1985). Synaptic connectivity of a local circuit neurone in lateral
geniculate nucleus of the cat. Nature 317, 618-621. 10.1038/317618a0.
Hirsch, J.A., Martinez, L.M., Pillai, C., Alonso, J.-m., Wang, Q., and Sommer, F.T. (2003). Functionally
distinct inhibitory neurons at the first stage of visual cortical processing. 6, 1300-1308. 10.1038/nn1152.
Hirsch, J.A., Wang, X., Sommer, F.T., and Martinez, L.M. (2015). How Inhibitory Circuits in Thalamus Serve
Vision. Annual Review of Neuroscience 38, null-null. 10.1146/annurev-neuro-071013-014229.
Hubel, D.H. (1960). Single unit activity in lateral geniculate body and optic tract of unrestrained cats. The
Journal of Physiology 150, 91-104. https://doi.org/10.1113/jphysiol.1960.sp006375.
Hubel, D.H., and Wiesel, T.N. (1959). Receptive fields of single neurones in the cat's striate cortex. Journal
of Physiology, 574-591.
Hubel, D.H., and Wiesel, T.N. (1961). Integrative action in the cat's lateral geniculate body. The Journal of
physiology 155, 385-398. 10.1113/jphysiol.1961.sp006635.
Hubel, D.H., and Wiesel, T.N. (1968). Receptive fields and functional architecture of monkey striate cortex.
The Journal of physiology 195, 215-243.
Hubel, D.H.H., and Wiesel, T.N. (1962). Receptive fields, binocular interaction and functional architecture
in the cat's visual cortex. The Journal of Physiology 160, 106-154.
Huberman, A.D., Manu, M., Koch, S.M., Susman, M.W., Lutz, A.B., Ullian, E.M., Baccus, S.A., and Barres,
B.A. (2008). Architecture and Activity-Mediated Refinement of Axonal Projections from a Mosaic of
Genetically Identified Retinal Ganglion Cells. Neuron 59, 425-438. 10.1016/j.neuron.2008.07.018.
Huberman, A.D., Wei, W., Elstrott, J., Stafford, B.K., Feller, M.B., and Barres, B.A. (2009). Genetic
Identification of an On-Off Direction- Selective Retinal Ganglion Cell Subtype Reveals a Layer-Specific
Subcortical Map of Posterior Motion. Neuron 62, 327-334. 10.1016/j.neuron.2009.04.014.
Huguenard, J.R., and McCormick, D.A. (1992). Simulation of the currents involved in rhythmic oscillations
in thalamic relay neurons. Journal of neurophysiology 68, 1373-1383.
78
Humphrey, A.L., and Weller, R.E. (1988). Structural correlates of functionally distinct X-cells in the lateral
geniculate nucleus of the cat. The Journal of Comparative Neurology 268, 448-468.
10.1002/cne.902680312.
Jager, P., Moore, G., Calpin, P., Durmishi, X., Salgarella, I., Menage, L., Kita, Y., Wang, Y., Kim, D.W.,
Blackshaw, S., et al. (2021). Dual midbrain and forebrain origins of thalamic inhibitory interneurons. eLife
10, e59272-e59272. 10.7554/eLife.59272.
Jia, F., Yue, M., Chandra, D., Homanics, G.E., Goldstein, P.A., and Harrison, N.L. (2008). Isoflurane is a
potent modulator of extrasynaptic GABA(A) receptors in the thalamus. The Journal of pharmacology and
experimental therapeutics 324, 1127-1135. 10.1124/jpet.107.134569.
Kaplan, E., Purpura, K.P., and Shapley, R.M. (1987). Contrast affects the transmission of visual information
through the mammalian lateral geniculate nucleus. Journal of Physiology 391, 267-288.
Kara, P., Reinagel, P., and Reid, R.C. (2000). Low Response Variability in Simultaneously Recorded Retinal,
Thalamic, and Cortical Neurons. Neuron 27, 635-646. https://doi.org/10.1016/S0896-6273(00)00072-6.
Karklin, Y., and Simoncelli, E. (2011). Efficient coding of natural images with a population of noisy Linear-
Nonlinear neurons.
Katz, Y., Yizhar, O., Staiger, J., and Lampl, I. (2013). Optopatcher-An electrode holder for simultaneous
intracellular patch-clamp recording and optical manipulation. Journal of Neuroscience Methods 214, 113-
117. 10.1016/j.jneumeth.2013.01.017.
Kerlin, A.M., Andermann, M.L., Berezovskii, V.K., and Reid, R.C. (2010). Broadly Tuned Response
Properties of Diverse Inhibitory Neuron Subtypes in Mouse Visual Cortex. Neuron 67, 858-871.
10.1016/j.neuron.2010.08.002.
Kerschensteiner, D., and Guido, W. (2017). Organization of the dorsal lateral geniculate nucleus in the
mouse. Visual Neuroscience 34, E008-E008. 10.1017/S0952523817000062.
Kim, I.-J., Zhang, Y., Meister, M., and Sanes, J.R. (2010). Laminar restriction of retinal ganglion cell
dendrites and axons: subtype-specific developmental patterns revealed with transgenic markers. The
Journal of neuroscience : the official journal of the Society for Neuroscience 30, 1452-1462.
10.1523/JNEUROSCI.4779-09.2010.
79
Kim, U., Sanchez-Vives, M.V., and McCormick, D.A. (1997). Functional dynamics of GABAergic inhibition in
the thalamus. Science 278, 130-134. 10.1126/science.278.5335.130.
Koepsell, K., and Sommer, F.T. (2008). Information transmission in oscillatory neural activity. Biological
cybernetics 99, 403-416. 10.1007/s00422-008-0273-6.
Koepsell, K., Wang, X., Vaingankar, V., Wei, Y., Wang, Q., Rathbun, D.L., Usrey, W.M., Hirsch, J.a., and
Sommer, F.T. (2009). Retinal oscillations carry visual information to cortex. Frontiers in systems
neuroscience 3, 1-18. 10.3389/neuro.06.004.2009.
Krumin, M., A.B, S., K.D, H., and M, C. (2014). Factors determining narrowband gamma oscillation in mouse
visual cortex. SfN 2014.
Kuffler, S.W. (1953). Discharge Patterns and Functional of Mammalian Retina. J. Neurophysiol. 16, 37-68.
Lam, Y.W., Cox, C.L., Varela, C., and Sherman, S.M. (2005). Morphological correlates of triadic circuitry in
the lateral geniculate nucleus of cats and rats. Journal of Neurophysiology 93, 748-757.
10.1152/jn.00256.2004.
Lesica, N.A., and Stanley, G.B. (2004). Encoding of natural scene movies by tonic and burst spikes in the
lateral geniculate nucleus. The Journal of neuroscience 24, 10731-10740. 10.1523/JNEUROSCI.3059-
04.2004.
Lesica, N.a., Weng, C., Jin, J., Yeh, C.-I., Alonso, J.-M., and Stanley, G.B. (2006). Dynamic encoding of natural
luminance sequences by LGN bursts. PLoS biology 4, e209-e209. 10.1371/journal.pbio.0040209.
LeVay, S., and Ferster, D. (1979). Proportion of interneurons in the cat's lateral geniculate nucleus. Brain
Research 164, 304-308. https://doi.org/10.1016/0006-8993(79)90026-X.
Levick, W.R. (1967). Receptive fields and trigger features of ganglion cells in the visual streak of the rabbit's
retina. The Journal of Physiology 188, 285-307. 10.1113/jphysiol.1967.sp008140.
Levick, W.R., Cleland, B.G., and Dubin, M.W. (1972). Lateral geniculate neurons of cat: retinal inputs and
physiology. Investigative ophthalmology 11, 302-311.
80
Liang, L., Fratzl, A., Goldey, G., Ramesh, R.N., Sugden, A.U., Morgan, J.L., Chen, C., and Andermann, M.L.
(2018). A Fine-Scale Functional Logic to Convergence from Retina to Thalamus. Cell 173, 1343-1355.e1324.
10.1016/j.cell.2018.04.041.
Lien, A.D., and Scanziani, M. (2013). Tuned thalamic excitation is amplified by visual cortical circuits.
Nature Neuroscience 16, 1315-1323. 10.1038/nn.3488.
Lima, S.Q., Hromádka, T., Znamenskiy, P., and Zador, A.M. (2009). PINP: A new method of tagging neuronal
populations for identification during in vivo electrophysiological recording. PLoS ONE 4.
10.1371/journal.pone.0006099.
Litvina, E.Y., and Chen, C. (2017). Functional Convergence at the Retinogeniculate Synapse. Neuron 96,
330-338.e335. 10.1016/j.neuron.2017.09.037.
Liu, B.-h., Li, P., Sun, Y.J., Li, Y.-t., Zhang, L.I., and Tao, H.W. (2010). Intervening inhibition underlies simple-
cell receptive field structure in visual cortex. Nature Neuroscience 13, 89-96. 10.1038/nn.2443.
Liu, B.H., Li, P., Li, Y.T., Sun, Y.J., Yanagawa, Y., Obata, K., Zhang, L.I., and Tao, H.W. (2009). Visual receptive
field structure of cortical inhibitory neurons revealed by two-photon imaging guided recording. J Neurosci
29, 10520-10532. 10.1523/JNEUROSCI.1915-09.2009.
Liu, B.h., Li, Y.t., Ma, W.p., Pan, C.j., Zhang, L.I., and Tao, H.W. (2011). Broad inhibition sharpens orientation
selectivity by expanding input dynamic range in mouse simple cells. Neuron 71, 542-554.
10.1016/j.neuron.2011.06.017.
Liu, R.C., Tzonev, S., Rebrik, S., and Miller, K.D. (2001). Variability and information in a neural code of the
cat lateral geniculate nucleus. Journal of Neurophysiology 86, 2789-2806. 10.1152/jn.2001.86.6.2789.
Lorincz, M.L., Kékesi, K.a., Juhász, G., Crunelli, V., and Hughes, S.W. (2009). Temporal framing of thalamic
relay-mode firing by phasic inhibition during the alpha rhythm. Neuron 63, 683-696.
10.1016/j.neuron.2009.08.012.
Lu, S.M., Guido, W., and Sherman, S.M. (1992). Effects of membrane voltage on receptive field properties
of lateral geniculate neurons in the cat: contributions of the low-threshold Ca2+ conductance. Journal of
Neurophysiology 68, 2185-2198. 10.1152/jn.1992.68.6.2185.
81
Madarasz, M., Gerle, J., Hajdu, F., Somogyi, G., and Tombol, T. (1978). Quantitative histological studies on
the lateral geniculate nucleus in the cat. II. Cell numbers and densities in the several layers. Journal fur
Hirnforschung 19, 159-164.
Madisen, L., Mao, T., Koch, H., Zhuo, J.-m., Berenyi, A., Fujisawa, S., Hsu, Y.-W.A., Garcia, A.J., Gu, X.,
Zanella, S., et al. (2012). A toolbox of Cre-dependent optogenetic transgenic mice for light-induced
activation and silencing. Nat Neurosci 15, 793-802.
Mahn, M., Gibor, L., Patil, P., Cohen-Kashi Malina, K., Oring, S., Printz, Y., Levy, R., Lampl, I., and Yizhar, O.
(2018). High-efficiency optogenetic silencing with soma-targeted anion-conducting channelrhodopsins.
Nature Communications 9, 4125-4125. 10.1038/s41467-018-06511-8.
Marr, D., and Hildreth, E. (1980). Theory for Edge Detection. Proceedings of the Royal Society of London
B: Biological Sciences 207, 187-217. 10.1098/rspb.1980.0020.
Marshel, J.H., Kaye, A.P., Nauhaus, I., and Callaway, E.M. (2012). Anterior-Posterior Direction Opponency
in the Superficial Mouse Lateral Geniculate Nucleus. Neuron 76, 713-720. 10.1016/j.neuron.2012.09.021.
Martinez, L.M., Molano-Mazón, M., Wang, X., Sommer, F.T., and Hirsch, J.A. (2014). Statistical wiring of
thalamic receptive fields optimizes spatial sampling of the retinal image. Neuron 81, 943-956.
10.1016/j.neuron.2013.12.014.
Martinez, L.M., Wang, Q., Reid, R.C., Pillai, C., Alonso, J.M., Sommer, F.T., and Hirsch, J.A. (2005). Receptive
field structure varies with layer in the primary visual cortex. Nature Neuroscience 8, 372-379.
10.1038/nn1404.
Mastronarde, D.N. (1992). Nonlagged relay cells and interneurons in the cat lateral geniculate nucleus:
Receptive-field properties and retinal inputs. Visual Neuroscience 8, 407-441.
10.1017/S0952523800004934.
Mastronarde, D.N.D., Mastronarde, N., Mastronarde, D.N.D., Mastronarde, N., and Mastronarde, D.N.D.
(1987). Two classes of single-input X-cells in cat lateral geniculate nucleus. II. Retinal inputs and the
generation of receptive-field properties. Journal of neurophysiology 57, 381-413.
10.1152/jn.1987.57.2.381.
82
Montero, V.M. (1987). Ultrastructural identification of synaptic terminals from the axon of type 3
interneurons in the cat lateral geniculate nucleus. Journal of Comparative Neurology 264, 268-283.
https://doi.org/10.1002/cne.902640210.
Moore, B.D.t., Kiley, C.W., Sun, C., and Usrey, W.M. (2011). Rapid plasticity of visual responses in the adult
lateral geniculate nucleus. Neuron 71, 812-819. 10.1016/j.neuron.2011.06.025.
Morgan, J.L., Berger, D.R., Wetzel, A.W., and Lichtman, J.W. (2016). The Fuzzy Logic of Network
Connectivity in Mouse Visual Thalamus. Cell 165, 192-206. 10.1016/j.cell.2016.02.033.
Morgan, J.L., and Lichtman, J.W. (2020). An individual interneuron participates in many kinds of inhibition
and spans much of the mouse visual thalamus. Neuron 106, 468-481. 10.1101/683276.
Morgan, J.L., Lichtman, J.W., Science, V., and Biology, C. (2019). An individual interneuron participates in
many kinds of inhibition and spans much of the mouse visual thalamus. Neuron 106, 683276-683276.
10.1101/683276.
Niell, C.M., and Stryker, M.P. (2008). Highly Selective Receptive Fields in Mouse Visual Cortex. The Journal
of Neuroscience 28, 7520-7536. 10.1523/JNEUROSCI.0623-08.2008.
Ohki, K., and Reid, R.C. (2007). Specificity and randomness in the visual cortex. Current opinion in
neurobiology 17, 401-407. 10.1016/j.conb.2007.07.007.
Olshausen, B.A., and Field, D.J. (1996). Natural image statistics and efficient coding. 7, 333-339.
Pachitariu, M., Steinmetz, N.A., Kadir, S.N., Carandini, M., and Harris, K.D. (2016). Fast and accurate spike
sorting of high-channel count probes with KiloSort. NIPS Proceedings, 4448-4456.
Paninski, L. (2004). Maximum likelihood estimation of cascade point-process neural encoding models.
Network: Computation in Neural Systems 15, 243-262. 10.1088/0954-898X/15/4/002.
Pearson, J.T., and Kerschensteiner, D. (2015). Ambient illumination switches contrast preference of
specific retinal processing streams. Journal of Neurophysiology 114, 540-550. 10.1152/jn.00360.2015.
83
Perreault, M.-C., and Raastad, M. (2006). Contribution of morphology and membrane resistance to
integration of fast synaptic signals in two thalamic cell types. The Journal of physiology 577, 205-220.
10.1113/jphysiol.2006.113043.
Peters, A., and Payne, B.R. (1993). Numerical relationships between geniculocortical afferents and
pyramidal cell modules in cat primary visual cortex. Cerebral cortex (New York, N.Y. : 1991) 3, 69-78.
10.1093/cercor/3.1.69.
Piscopo, D.M., El-Danaf, R.N., Huberman, A.D., and Niell, C.M. (2013). Diverse Visual Features Encoded in
Mouse Lateral Geniculate Nucleus. Journal of Neuroscience 33, 4642-4656. 10.1523/JNEUROSCI.5187-
12.2013.
Rathbun, D.L., Warland, D.K., and Usrey, W.M. (2010). Spike timing and information transmission at
retinogeniculate synapses. The Journal of neuroscience : the official journal of the Society for
Neuroscience 30, 13558-13566. 10.1523/JNEUROSCI.0909-10.2010.
Reich, D.S., Victor, J.D., Knight, B.W., Ozaki, T., and Kaplan, E. (1997). Response variability and timing
precision of neuronal spike trains in vivo. Journal of Neurophysiology 77, 2836-2841.
10.1152/jn.1997.77.5.2836.
Reinagel, P., and Reid, R.C. (2000). Temporal coding of visual information in the thalamus. The Journal of
Neuroscience 20, 5392-5400.
Reinagel, P., and Reid, R.C. (2002). Precise firing events are conserved across neurons. Journal of
Neuroscience 22, 6837-6841. 10.1523/jneurosci.22-16-06837.2002.
Rivlin-Etzion, M., Wei, W., and Feller, M.B. (2012). Visual Stimulation Reverses the Directional Preference
of Direction-Selective Retinal Ganglion Cells. Neuron 76, 518-525. 10.1016/j.neuron.2012.08.041.
Rompani, S.B., Müllner, F.E., Wanner, A., Zhang, C., Roth, C.N., Yonehara, K., and Roska, B. (2017).
Different Modes of Visual Integration in the Lateral Geniculate Nucleus Revealed by Single-Cell-Initiated
Transsynaptic Tracing. Neuron 93, 767-776. 10.1016/j.neuron.2017.01.028.
Rossant, C., Kadir, S.N., Goodman, D.F.M., Schulman, J., Hunter, M.L.D., Saleem, A.B., Grosmark, A.,
Belluscio, M., Denfield, G.H., Ecker, A.S., et al. (2016). Spike sorting for large, dense electrode arrays.
Nature Neuroscience 19, 634-641. 10.1038/nn.4268.
84
Roy, S., Jun, N.Y., Davis, E.L., Pearson, J., and Field, G.D. (2021). Inter-mosaic coordination of retinal
receptive fields. Nature 592, 409-413. 10.1038/s41586-021-03317-5.
Runyan, C.A., Schummers, J., Van Wart, A., Kuhlman, S.J., Wilson, N.R., Huang, Z.J., and Sur, M. (2010).
Response features of parvalbumin-expressing interneurons suggest precise roles for subtypes of inhibition
in visual cortex. Neuron 67, 847-857. 10.1016/j.neuron.2010.08.006.
Sabbagh, U., Govindaiah, G., Somaiya, R.D., Ha, R.V., Wei, J.C., Guido, W., and Fox, M.A. (2021). Diverse
GABAergic neurons organize into subtype-specific sublaminae in the ventral lateral geniculate nucleus. J
Neurochem 159, 479-497. 10.1111/jnc.15101.
Sanderson, K.J. (1971). The projection of the visual field to the lateral geniculate and medial interlaminar
nuclei in the cat. Journal of Comparative Neurology 143, 101-118. 10.1002/cne.901430107.
Sanes, J.R., and Masland, R.H. (2014). The Types of Retinal Ganglion Cells: Current Status and Implications
for Neuronal Classification. Annual Review of Neuroscience 38, 150421150146009-150421150146009.
10.1146/annurev-neuro-071714-034120.
Scharfman, H.E., Lu, S.M., Guido, W., Adams, P.R., and Sherman, S.M. (1990). N-methyl-D-aspartate
receptors contribute to excitatory postsynaptic potentials of cat lateral geniculate neurons recorded in
thalamic slices. Proceedings of the National Academy of Sciences 87, 4548-4552.
10.1073/pnas.87.12.4548.
Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C.,
Saalfeld, S., Schmid, B., et al. (2012). Fiji: An open-source platform for biological-image analysis. Nature
Methods 9, 676-682. 10.1038/nmeth.2019.
Scholl, B., Tan, A.Y.Y., Corey, J., and Priebe, N.J. (2013). Emergence of orientation selectivity in the
mammalian visual pathway. Journal of Neuroscience 33, 10616-10624. 10.1523/JNEUROSCI.0404-
13.2013.
Schreiber, S., Fellous, J.M., Whitmer, D., Tiesinga, P., and Sejnowski, T.J. (2003). A new correlation-based
measure of spike timing reliability. Neurocomputing 52-54, 925-931.
Schwartz, O., Pillow, J.W., Rust, N.C., and Simoncelli, E.P. (2006). Spike-triggered neural characterization.
Journal of vision 6, 484-507. 10.1167/6.4.13.
85
Seabrook, T.A., Krahe, T.E., Govindaiah, G., and Guido, W. (2013). Interneurons in the mouse visual
thalamus maintain a high degree of retinal convergence throughout postnatal development. Neural
development 8, 24-24. 10.1186/1749-8104-8-24.
Shamash, P., Carandini, M., Harris, K., and Steinmetz, N. (2018). A tool for analyzing electrode tracks from
slice histology. bioRxiv, 447995-447995. 10.1101/447995.
Shapley, R., and Victor, J.D. (1978). The effect of contrast on the transfer properties of cat retinal ganglion
cells. The Journal of Physiology 285, 275-298. 10.1113/jphysiol.1978.sp012571.
Sherman, S.M. (2001). Tonic and burst firing: dual modes of thalamocortical relay. Trends in
Neurosciences 24, 122-126. 10.1016/S0166-2236(00)01714-8.
Sherman, S.M. (2004). Interneurons and triadic circuitry of the thalamus. Trends in neurosciences 27, 670-
675. 10.1016/j.tins.2004.08.003.
Sherman, S.M. (2014). The function of metabotropic glutamate receptors in thalamus and cortex. The
Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry 20, 136-149.
10.1177/1073858413478490.
Sherman, S.M., and Guillery, R.W. (2002). The role of the thalamus in the flow of information to the cortex.
Philosophical Transactions of the Royal Society B: Biological Sciences 357, 1695-1708.
10.1098/rstb.2002.1161.
Shou, T., and Leventhal, A.G. (1989). Organized arrangement of orientation-sensitive relay cells in the cat's
dorsal lateral geniculate nucleus. Journal of Neuroscience 9, 4287-4302. 10.1523/jneurosci.09-12-
04287.1989.
Simoncelli, E.P., and Olshausen, B.A. (2001). Natural Image Statistics and Neural Representation. Annual
Review of Neuroscience 24, 1193-1216.
Sincich, L.C., Adams, D.L., Economides, J.R., and Horton, J.C. (2007). Transmission of Spike Trains at the
Retinogeniculate Synapse. 27, 2683-2692. 10.1523/JNEUROSCI.5077-06.2007.
86
Sincich, L.C., Horton, J.C., and Sharpee, T.O. (2009). Preserving information in neural transmission. The
Journal of neuroscience : the official journal of the Society for Neuroscience 29, 6207-6216.
10.1523/JNEUROSCI.3701-08.2009.
Smith, E.L., Chino, Y.M., William, H., Ridder, R., Kitagawa, K., and Langston, A. (1990). Orientation bias of
neurons in the lateral geniculate nucleus of macaque monkeys. Visual Neuroscience 5, 525-545.
10.1017/S0952523800000699.
Smith, S.L., and Häusser, M. (2010). Parallel processing of visual space by neighboring neurons in mouse
visual cortex. Nature Publishing Group 13. 10.1038/nn.2620.
So, Y.T., and Shapley, R. (1981). Spatial tuning of cells in and around lateral geniculate nucleus of the cat:
X and Y relay cells and perigeniculate interneurons. Journal of neurophysiology 45, 107-120.
Soto-Sánchez, C., Wang, X., Vaingankar, V., Sommer, F.T., and Hirsch, J.A. (2017). Spatial scale of receptive
fields in the visual sector of the cat thalamic reticular nucleus. Nature Communications 8, 1-7.
10.1038/s41467-017-00762-7.
Steriade, M., McCormick, D.A., and Sejnowski, T.J. (1993). Thalamocortical oscillations in the sleeping and
aroused brain. Science (New York, N.Y.) 262, 679-685. 10.1126/science.8235588.
Suresh, V., Ciftcioglu, U.M., Wang, X., Lala, B.M., Ding, K.R., Smith, W.A., Sommer, F.T., and Hirsch, J.A.
(2016). Synaptic Contributions to Receptive Field Structure and Response Properties in the Rodent Lateral
Geniculate Nucleus of the Thalamus. Journal of Neuroscience 36, 10949-10963.
10.1523/JNEUROSCI.1045-16.2016.
Swadlow, H.A., and Gusev, A.G. (2001). The impact of ‘ bursting ’ thalamic impulses at a neocortical
synapse. 4.
Swadlow, H.A., and Gusev, A.G. (2002). Receptive-field construction in cortical inhibitory interneurons.
Nature Neuroscience 5, 403-404. 10.1038/nn847.
Taniguchi, H., He, M., Wu, P., Kim, S., Paik, R., Sugino, K., Kvitsani, D., Fu, Y., Lu, J., Lin, Y., et al. (2011). A
Resource of Cre Driver Lines for Genetic Targeting of GABAergic Neurons in Cerebral Cortex. Neuron 71,
995-1013. 10.1016/j.neuron.2011.07.026.
87
Thompson, K.G., Leventhal, A.G., and Zhou, Y. (1994). Direction-sensitive X and Y cells within the A laminae
of the cat's LGNd. Visual Neuroscience 11, 927-938. 10.1017/S0952523800003886.
Tien, N.W., Pearson, J.T., Heller, C.R., Demas, J., and Kerschensteiner, D. (2015). Genetically Identified
Suppressed-by-Contrast Retinal Ganglion Cells Reliably Signal Self-Generated Visual Stimuli. Journal of
Neuroscience 35, 10815-10820. 10.1523/JNEUROSCI.1521-15.2015.
Tikidji-Hamburyan, A., Reinhard, K., Seitter, H., Hovhannisyan, A., Procyk, C.A., Allen, A.E., Schenk, M.,
Lucas, R.J., and Münch, T.A. (2015). Retinal output changes qualitatively with every change in ambient
illuminance. Nature Neuroscience 18, 66-74. 10.1038/nn.3891.
Uhlrich, D.J., Cucchiaro, J.B., Humphrey, A.L., and Sherman, S.M. (1991). Morphology and axonal
projection patterns of individual neurons in the cat perigeniculate nucleus. Journal of Neurophysiology
65, 1528 LP-1541.
Usrey, W.M., and Reid, R.C. (2000). Visual physiology of the lateral geniculate nucleus in two species of
new world monkey: Saimiri sciureus and Aotus trivirgatis. The Journal of physiology 523 Pt 3, 755-769.
Usrey, W.M., Reppas, J.B., and Reid, R.C. (1998). Paired-spike interactions and synaptic efficacy of retinal
inputs to the thalamus. Nature 395, 384-387. 10.1038/26487.
Usrey, W.M., Reppas, J.B., and Reid, R.C. (1999). Specificity and strength of retinogeniculate connections.
Journal of Neurophysiology 82, 3527-3540. 10.1152/jn.1999.82.6.3527.
Uzzell, V.J., and Chichilnisky, E.J. (2004). Precision of spike trains in primate retinal ganglion cells. Journal
of Neurophysiology 92, 780-789. 10.1152/jn.01171.2003.
Vaingankar, V., Sanchez Soto, C., Wang, X., Sommer, F.T., and Hirsch, J.A. (2012). Neurons in the thalamic
reticular nucleus are selective for diverse and complex visual features. Frontiers in integrative
neuroscience 6, 1-17. 10.3389/fnint.2012.00118.
Van Horn, S.C., Erisir, A., and Sherman, S.M. (2000). Relative distribution of synapses in the A-laminae of
the lateral geniculate nucleus of the cat. The Journal of comparative neurology 416, 509-520.
88
Vidyasagar, T.R., and Heide, W. (1984). Geniculate orientation biases seen with moving sine wave gratings:
implications for a model of simple cell afferent connectivity. Experimental brain research 57, 176-200.
10.1007/bf00231146.
Wagenaar, D.A., and Potter, S.M. (2002). Real-time multi-channel stimulus artifact suppression by local
curve fitting. Journal of Neuroscience Methods 120, 113-120. 10.1016/S0165-0270(02)00149-8.
Wang, Q., and Burkhalter, A. (2007). Area Map of Mouse Visual Cortex. The Journal of comparative
neurology 504, 287-297. 10.1002/cne.
Wang, X., Hirsch, J.A., and Sommer, F.T. (2010). Recoding of sensory information across the retinothalamic
synapse. The Journal of Neuroscience 30, 13567-13577. 10.1523/JNEUROSCI.0910-10.2010.
Wang, X., Sommer, F.T., and Hirsch, J.A. (2011a). Inhibitory circuits for visual processing in thalamus.
Current opinion in neurobiology 21, 726-733. 10.1016/j.conb.2011.06.004.
Wang, X., Vaingankar, V., Sanchez, C.S., Sommer, F.T., and Hirsch, J.A. (2011b). Thalamic interneurons and
relay cells use complementary synaptic mechanisms for visual processing. Nature neuroscience 14, 224-
231. 10.1038/nn.2707.
Wang, X., Wei, Y., Vaingankar, V., Wang, Q., Koepsell, K., Sommer, F.T., and Hirsch, J.A. (2007).
Feedforward Excitation and Inhibition Evoke Dual Modes of Firing in the Cat’s Visual Thalamus during
Naturalistic Viewing. Neuron 55, 465-478. 10.1016/j.neuron.2007.06.039.
Wassle, H., Peichl, L., and Boycott, B.B. (1983). Mosaics and territories of cat retinal ganglion cells.
Progress in brain research 58, 183-190. 10.1016/S0079-6123(08)60019-9.
Weng, C., Yeh, C.I., Stoelzel, C.R., and Alonso, J.M. (2005). Receptive field size and response latency are
correlated within the cat visual thalamus. Journal of Neurophysiology 93, 3537-3547.
10.1152/jn.00847.2004.
Werblin, F.S. (2010). Six different roles for crossover inhibition in the retina: correcting the nonlinearities
of synaptic transmission. Visual neuroscience 27, 1-8. 10.1017/S0952523810000076.
Wienbar, S., and Schwartz, G.W. (2018). The dynamic receptive fields of retinal ganglion cells. Progress in
Retinal and Eye Research 67, 102-117. 10.1016/j.preteyeres.2018.06.003.
89
Wiesel, T.N. (1959). Recording inhibition and excitation in the cat's retinal ganglion cells with intracellular
electrodes. Nature 183, 264-265. 10.1038/183264a0.
Wilson, J.R. (1989). Synaptic organization of individual neurons in the macaque lateral geniculate nucleus.
The Journal of Neuroscience 9, 2931-2953.
Wilson, J.R., Forestner, D.M., and Cramer, R.P. (1996). Quantitative analyses of synaptic contacts of
interneurons in the dorsal lateral geniculate nucleus of the squirrel monkey. Visual Neuroscience 13, 1129-
1142. 10.1017/S095252380000777X.
Xu, X., Ichida, J., Shostak, Y., Bonds, A.B., and Casagrande, V.A. (2002). Are primate lateral geniculate
nucleus (LGN) cells really sensitive to orientation or direction? Visual Neuroscience 19, 97-108.
10.1017/S0952523802191097.
Xue, J.T., Carney, T., Ramoa, A.S., and Freeman, R.D. (1988). Binocular interaction in the perigeniculate
nucleus of the cat. Experimental Brain Research 69, 497-508.
Yeh, C.I., Stoelzel, C.R., Weng, C., and Alonso, J.M. (2009). Functional consequences of neuronal
divergence within the retinogeniculate pathway. Journal of Neurophysiology 101, 2166-2185.
10.1152/jn.91088.2008.
Zhang, Y., Kim, I.J., Sanes, J.R., and Meister, M. (2012). The most numerous ganglion cell type of the mouse
retina is a selective feature detector. Proceedings of the National Academy of Sciences 109, E2391-E2398.
10.1073/pnas.1211547109.
Zhao, X., Chen, H., Liu, X., and Cang, J. (2013). Orientation-selective responses in the mouse lateral
geniculate nucleus. The Journal of Neuroscience 33, 12751-12763. 10.1523/JNEUROSCI.0095-13.2013.
Zhu, J.J., Uhlrich, D.J., and Lytton, W.W. (1999). Properties of a hyperpolarization-activated cation current
in interneurons in the rat lateral geniculate nucleus. Neuroscience 92, 445-457. 10.1016/S0306-
4522(98)00759-3.
Zoltowski, D., and Pillow, J. (2018). Scaling the Poisson GLM to massive neural datasets through
polynomial approximations. Advances in neural information processing systems 31, 3517-3527.
Abstract (if available)
Abstract
In the early visual pathway, retina sends visual information to the dorsal lateral geniculate nucleus’ (dLGN) relay cells that project to cortex. Historically, the dLGN was considered a relay station (hence “relay” cells), one that passes on information encoded in retina to cortex unaltered. However, complex inhibitory circuits in the visual thalamus suggest that retinal information is not simply relayed to the primary visual cortex but is altered by inhibition with the dLGN. There are two types of inhibitory neurons that synapse onto relay cells: cells in the thalamic reticular nucleus (TRN) and local interneurons. Thus, inhibition influences all visual information passing through the dLGN. The research presented in this dissertation will focus on the local interneurons, which dominate the intrinsic circuit in the dLGN and can provide powerful inhibition to relay cells.
In Chapter 2, in the first study of its kind, I use intracellular, juxtacellular, or multielectrode methods to record from murine dLGN during visual stimulation and use optogenetics to distinguish local interneurons from relay cells and to characterize the functional properties of interneurons learn how they might influence visual information sent to cortex. Both relay cells and interneurons have diverse receptive field structures. We found that interneurons, like relay cells, have diverse types of receptive fields. The distribution of receptive field structure, but not size, differed between the two populations. To learn how interneurons might contribute to responses of relay cells, we turned to computational approaches. We built simple LNP models of interneurons and used the output of these models to reconstruct suppressive components of the relay cell’s receptive field that we extracted from recordings of the membrane current. We were able to reconstruct these suppressive fields for a group of relay cells including those center-surround and suppressed-by-contrast profiles using a biologically plausible number of inputs. Thus, we provide proof of concept that convergent input from diverse types of interneurons can provide feature-specific inhibition to relay cells.
In Chapter 3, using data obtained in conjunction with Chapter 2, I investigated the temporal precision of interneurons versus that of relay cells. We made recordings from optotagged interneurons and relay cells in the murine dLGN and compared responses evoked by natural movies at different contrasts. While some interneurons fired as precisely as relay cells, they were less reliable as a population across stimulus contrasts. Firing rates were similar for both populations, however. Reliability increased steadily with contrast for most relay cells but saturated with the step from the lowest contrast to the next for interneurons. There was also a modest correlation between smaller receptive field size and greater temporal precision for relay cells but not for interneurons.
This work is the first study of receptive field structure and temporal response properties of identified local interneurons in murine dLGN and, thus, provides unique insight into how these cells influence vision.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Exploring sensory responses in the different subdivisions of the visual thalamus
PDF
Sensory information processing by retinothalamic neural circuits
PDF
Excitatory-inhibitory interactions in pyramidal neurons
PDF
A million-plus neuron model of the hippocampal dentate gyrus: role of topography, inhibitory interneurons, and excitatory associational circuitry in determining spatio-temporal dynamics of granul...
PDF
Contextual modulation of sensory processing via the pulvinar nucleus
PDF
Encoding of natural images by retinal ganglion cells
PDF
Functional synaptic circuits in primary visual cortex
PDF
Functional magnetic resonance imaging characterization of peripheral form vision
PDF
The behavioral and neural bases of tactile object localization
PDF
Synaptic integration in dendrites: theories and applications
PDF
The impact of genetic pleiotropy on heterogeneity in the developing forebrain and hindbrain
PDF
Computational investigation of cholinergic modulation of the hippocampus and directional encoding of touch in somatosensory cortex
PDF
Functional models of fMRI BOLD signal in the visual cortex
PDF
Synaptic circuits for information processing along the central auditory pathway
PDF
Synaptic mechanism underlying development and function of neural circuits in rat primary auditory cortex
PDF
Modeling and analysis of parallel and spatially-evolving wall-bounded shear flows
PDF
Investigating the modulatory mechanisms of ILC2s as a therapeutic strategy in the context of health and disease
PDF
On the electrophysiology of multielectrode recordings of the basal ganglia and thalamus to improve DBS therapy for children with secondary dystonia
Asset Metadata
Creator
Gorin, Alexis Sara
(author)
Core Title
Spatial and temporal precision of inhibitory and excitatory neurons in the murine dorsal lateral geniculate nucleus
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Neuroscience
Degree Conferral Date
2022-12
Publication Date
09/12/2023
Defense Date
08/31/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
computational models,dLGN,dorsal lateral geniculate nucleus,electrophysiology,excitatory,inhibition,inhibitory,interneuron,mouse,neuron,OAI-PMH Harvest,optogenetics,relay cell,spatial,temporal precision
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Hires, Samuel Andrew (
committee chair
), Hirsch, Judith Ann (
committee member
), Mel, Bartlett (
committee member
)
Creator Email
agorin@usc.edu,asgorin@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111992647
Unique identifier
UC111992647
Legacy Identifier
etd-GorinAlexi-11204
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Gorin, Alexis Sara
Type
texts
Source
20220917-usctheses-batch-980
(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. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
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
Tags
computational models
dLGN
dorsal lateral geniculate nucleus
electrophysiology
excitatory
inhibition
inhibitory
interneuron
mouse
neuron
optogenetics
relay cell
spatial
temporal precision