University of Southern California Dissertations and Theses

Structure and kinetics of the Orb2 functional amyloid

(USC Thesis Other)

Structure and kinetics of the Orb2 functional amyloid

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content 1

Structure and Kinetics of the Orb2 Functional Amyloid

By

Alexander Stover Falk

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA

In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Medical Biophysics)

August 2019
2

Contents
ACKNOWLEDGEMENTS ..................................................................... 5
ABSTRACT ........................................................................................... 8
CHAPTER 1. INTRODUCTION .......................................................... 10
1.1 Biophysical Concepts Explored ............................................. 10
1.1.1. Amyloids.......................................................................... 10
1.1.2. Orb2 ................................................................................ 11
1.1.3. Droplets ........................................................................... 13
1.1.4. Huntingtin Exon 1 ............................................................ 14
1.2 Methods Used........................................................................ 15
1.2.1 NMR ................................................................................ 15
1.2.2 EPR ................................................................................. 18
1.2.3 Computer Modelling ........................................................ 19
1.2.4 Thioflavin T Fluorescence ............................................... 22
1.2.5 Chromatography ............................................................. 22
CHAPTER 2. PROJECT I: STUDIES ON NON-SOLUBLE
FRACTION ORB2 ............................................................................... 24
2.1 Introduction ............................................................................ 24
2.2 Methods ................................................................................. 24
2.2.1 Protein Expression .......................................................... 24
2.2.2 Labeled Expression ......................................................... 25
2.2.3 Protein Purification .......................................................... 26
2.2.4 Fibril Growth .................................................................... 26
2.3 Results ................................................................................... 27
2.3.1 Successful Protein Expression and Purification .............. 27
2.3.2 Not All Protein Binds the Ni-Column ............................... 28
2.3.3 Orb2 Purified from the Non-Soluble Fraction Forms
Fibrils Rapidly .............................................................................. 29
2.3.4 Fibrils Grown from Non-Soluble Fraction Protein Are
Large and Highly Bundled ........................................................... 30
2.3.5 NMR Experiments that Highlight the Static Domains
of the Fibril Show Extensive Overlap and Poor Resolution .......... 31
2.3.6 Fibrils Grown from Non-Soluble Fraction Protein
Show Limited Fibrilization Condition Impact ................................ 32
2.3.7 Fibrils Grown from Non-Soluble Fraction Protein
Have Well Resolved INEPT based Spectra ................................. 33
3

2.4 Discussion ............................................................................. 34
CHAPTER 3. PROJECT II: STUDIES ON SOLUBLE ORB2 AND
ΔRRM FRAGMENTS .......................................................................... 36
3.1 Introduction ............................................................................ 36
3.2 Methods ................................................................................. 36
3.2.1 Protein Expression .......................................................... 36
3.2.2 Protein Purification .......................................................... 37
3.2.3 Droplet Formation and Analysis ...................................... 39
3.2.4 Growth of Fibrils from Droplets ........................................ 39
3.2.5 Protein EPR Labelling ..................................................... 40
3.3 Results ................................................................................... 40
3.3.1 Successful Protein Expression and Purification .............. 40
3.3.2 Formation of An Apparent Oligomer ................................ 43
3.3.3 Formation of Droplets and Fibrils .................................... 44
3.3.4 NMR on Orb2BΔRRM Fibrils ........................................... 55
3.3.5 Successful EPR Labelling of Soluble Protein .................. 59
3.4 Discussion ............................................................................. 60
CHAPTER 4. PROJECT III: Computational Modelling of
Huntingtin C-terminus ......................................................................... 62
4.1 Introduction ............................................................................ 62
4.2 Methods ................................................................................. 62
4.2.1 Modelling Attempts .......................................................... 62
4.2.2 Recovery of Physical Parameters from Models ............... 66
4.3 Results ................................................................................... 69
4.3.1 Assessment of Simulation Accuracy ............................... 69
4.3.2 Simulation 7 Analysis ...................................................... 78
4.4 Discussion ............................................................................. 83
CHAPTER 5. CONCLUSIONS ........................................................... 86
CHAPTER 6. FUTURE DIRECTIONS ................................................ 88
BIBLIOGRAPHY ................................................................................. 91
APPENDIX .......................................................................................... 97
Transfection Procedure ................................................................... 97
Plate Reader Settings ..................................................................... 97
Electron Microscopy Grid Preparation Procedure ........................... 99
4

Dynamic domains of amyloid fibrils can be site-specifically
assigned with proton detected 3D NMR spectroscopy
(Full Copy) ..................................................................................... 100

5

ACKNOWLEDGEMENTS
First and foremost, I want to thank my mentor, Dr. Ansgar
Siemer. Dr. Siemer is a brilliant, passionate, and dedicated scientist,
but perhaps more importantly, he is pretty much the best mentor a
graduate student could hope for. He takes time to teach his students
and is always available for advice and guidance. He is patient, honest,
open and considers the impact of every decision he makes on the
people in his lab. Knowing the quality of his leadership has repeatedly
been a reassurance to me when I was at the most frustrating points in
my PhD.
I want to thank the rest of my thesis committee, Dr. Ralf Langen,
Dr. Robert Farley and Dr. Ian Haworth, for taking so much time to
advice and guide me through the previous years. I also want to thank
these individuals, along with Dr. Tobias Ulmer and Dr. Hayoun Lee for
being on my qualifying exam committee and helping to ensure I was
prepared for the rigors of a PhD.
I want to thank everyone who has worked in the Siemer lab with
me: Siliva Cervantes, Maria Soria, Connor Hurd, Samridhi Garg, Shruti
Bendre, Manjima Sarkar, Jina Kim, Dr. Bethany Caulkins, Thalia
Bajakian, Ninad Agashe, Dr. Rachel Service, Dr. Maiwenn Beaugrand,
and Rajashree Venkatraman, for their help and input over the years
and for making being in the lab such a great experience with their
wonderful personalities. I have heard that great bosses attract great
employees, and I think the Siemer lab is a perfect case study in this. I
would like to extend particular thanks to my fellow PhD students Maria
Soria and Silvia Cervantes, who have been with me, as friends as well
as coworkers, for the whole duration of my time in the Siemer lab. I
would also like to give particular thanks Connor Hurd, who has shown
6

almost boundless enthusiasm and dedication helping me on this
project over the past year. Much of what is presented in this thesis was
only doable thanks to him, and I am confident that everything I have
worked on will be in good hands as he takes over.
I want to thank everyone who has been in the protein structure
center on the first floor of the Zilkha Neurogenetic Institute during my
time here. Not just for their helpful advice and generous sharing of
resources, but for their friendliness and openness which has built a
sense of community and camaraderie and made this such a great
place to work. I want to give a special thanks to Dr. Thomas Schmidt
for introducing me to molecular dynamics and giving the protein
structure center such a lively atmosphere.
I want to thank the director of the PIBBS program, Dr. Ite
Offringa and the PIBBS administrative staff: Joyce Perez, Bami
Andrada, Marisela Zungia and Ashley Flinn for running their program
so well and giving me the chance to be part of it.
I want to thank Zilkha Neurogenetic Institute administrative staff
for keeping the institute running so well, for throwing parties for
everyone, and for handling that just about every non-scientific problem
we faced. I would like to particularly thank them, and especially Muoi
Thang, for their help in putting together and submitting my F31
application. Their knowledge and dedication turned what could have
been a cryptic and tedious process into a simple a straightforward one.
I want to thank our collaborator at the Stowers Institute for
Medical Research, Dr. Kausik Si and his postdoc, Dr. Ruben Hervas
for continuously working with me and the rest of the Siemer lab,
sharing samples, insights and helping us establish the critical biological
context of our work.
I want to thank the director of the USC Stem Cell Optical
Imaging Facility, Dr. Seth Ruffins, for providing me with the equipment,
7

expertise and help needed to produce all of the microscopy data
presented in this thesis.
I want to thank my undergraduate research mentors, Dr. Martin
Pomper at Johns Hopkins and Dr. Osnat Herzberg, and Dr. Thomas
Kocher at the University of Maryland, for first giving me the chance to
do research.
I want to thank National Institutes of Health for providing my F31
to me and for all the other funding they have given the Siemer lab to
make this project possible. I also want to thank the University of
Southern California for supporting me, first with a Provost Fellowship
and later with the Manning Endowed Fellowship.
I want to thank my wife, Shuangchao, for all the support she has
given me through this PhD, even while pursuing her own. I want to
thank my mother, father, sister, step-father, and step-sisters: Mary,
Robert, Anna, Edward, Evelyn and Audrey for being such a wonderful
family and supporting me through my life. I want to thank my wife’s
parents for supporting me even though they cannot to speak to me.
Finally, I want to thank my entire extended family (with particular
thanks to my grandfather and namesake, Sandy Stover) for being so
awesome and raising me in an environment that values education and
intellectual pursuits so highly.

8

ABSTRACT
Orb2 is a functional, amyloid forming, protein that plays a critical
role in long term memory formation in Drosophila melanogaster, with
homologs that play a similar role in other species, including humans.
Orb2 has two isoforms, Orb2A and Orb2B, and their interaction to form
functional amyloids is hypothesized to be a key step in the memory
formation process.
Huntingtin is a protein that can form pathological amyloids and
cause Huntington’s disease when mutated. It also has critical
functional roles in normal neural development when not mutated.
In this thesis, I present a series of experiments on Orb2, with
goal of understanding the amyloid structure and how it relates to
Orb2’s biological role. I show that Orb2 is capable of forming protein
droplets and two different types of fibrils (one of which seems to be
derived from droplets). I also show that fibrils formed by Orb2A and
Orb2B are highly similar and that within a single monomer in a fibril,
residues range in dynamics from extremely static to almost liquid-like. I
also present a series of molecular dynamics simulations of a domain of
Huntingtin. I show that these simulations can reproduce EPR and NMR
data gathered on Huntingtin fibrils, and also help to identify what
appears to be a highly conserved protein-protein interaction site in the
domain.
Because this thesis covers a range of experiments on different
proteins from different sources, I first present an introduction providing
a general background for all topics and methods discussed (CHAPTER
1. INTRODUCTION) before delving into experiments on Orb2 derived
from non-soluble protein fractions (CHAPTER 2. PROJECT I:
STUDIES ON NON-SOLUBLE FRACTION ORB2), experiments on
9

Orb2 derived from soluble protein fractions (CHAPTER 3. PROJECT II:
STUDIES ON SOLUBLE ORB2 AND ΔRRM FRAGMENTS) and
simulations of the Huntingtin exon-1 C-terminus (CHAPTER 4.
PROJECT III: Computational Modelling of Huntingtin C-terminus).
10

CHAPTER 1. INTRODUCTION
1.1 Biophysical Concepts Explored
1.1.1. Amyloids
Amyloids are fibrils composed of numerous copies of the same or
highly similar proteins in a stacked cross-β structure. Once formed, an
amyloid fibril has the potential to grow rapidly as additional protein
monomers are recruited into the fibril, adopting its conformation. How
an amyloid fibril initially forms is not completely clear and may vary
from protein to protein. However, it is known that initial formation of the
amyloid often requires overcoming a high activation energy barrier,
which is almost always much higher than the (relatively minimal)
barrier for addition of monomers to the amyloid. Most known amyloids
display a remarkable stability and thermodynamic favorability and are
resistant to conventional protein denaturants such as heat and SDS.
The result of the combined effects of high initial activation energy,
rapid grow and high thermodynamic favorability is that amyloid
proliferation from a solution of monomers often follows an S-curve
(Kumar, Haque, & Prabhu, 2017) (Chuang, Hori, Hesketh, & Shorter,
2018).
Amyloids have been found in all domains of life and have long been
suspected or identified as the causative agents in a number of human
diseases (especially neurodegenerative disorders) such as
Huntington’s disease, Alzheimer’s and ALS. This is not surprising,
given their potential for rapid proliferation and resistance to
degradation. It is hypothesized that the rapid and irreversible growth of
amyloids overwhelms and kills cells in these diseases (Chiti & Dobson,
2017).
11

Despite their strong disease association, amyloids have also been
found in functional roles. This includes the Curli fibrils of bacteria,
certain spider silk proteins, a number of yeast proteins, and even
human proteins (Fowler, Koulov, Balch, & Kelly, 2007) (Bergman,
Roan, Römling, Bevins, & Münch, 2016) (Chiti & Dobson, 2017).
In terms of occurrence, proteins or domains with a simplistic or
highly repetitive sequence, that are rich in disorder promoting residues,
and seem to adopt a disordered and dynamic conformation on initially
being dissolved in solution, are often the most likely to form amyloids
(Pang Benny Yiu & Wai Chen, 2017). It has also been shown that
proteins with apparent amyloid forming domains are also more likely to
contain RNA recognition motif (RRM) domains (Kato et al., 2012),
enhancing speculation about a functional role for amyloids as part of
an RNA regulatory network.
1.1.2. Orb2
Orb2 is a member of the cytoplasmic polyadenylation element
binding protein (CPEB) family expressed in Drosophila melanogaster.
Like all members of the family, it has two RRM domains followed by
zinc finger near its C-terminus (Ivshina, Lasko, & Richter, 2014). The
zinc finger is of the ZZ type, which is more commonly used in protein-
protein interactions than protein-DNA interactions (Merkel et al., 2013).
Prior to the RRM motifs, Orb2 has a long (~200 residues) domain that
lacks any clear structural order or sequence homology (including to
other CPEB proteins), which I refer to as the linker domain. This
domain is itself proceeded by an ~40 residue glutamine rich
(reminiscent of the poly-glutamine tract found in huntingtin) domain
that is interspersed with a few other residues (mainly histidine). Finally,
the N-terminus of Orb2 can vary depending on which isoform is made
through differential mRNA splicing. The heavily expressed Orb2B
isoform has a long (~170 residue) N-terminus that is rich in glycine and
12

serine. The much more lightly expressed Orb2A has a short (~20
residue) N-terminus that produces an amphipathic helix when wrapped
in a standard α-helical conformation (Krüttner et al., 2012) (Soria,
Cervantes, Bajakian, & Siemer, 2017).

Figure 1. The sequence of Orb2 isoforms A and B, showing the N-terminal, glutamine
rich, linker, first RRM, second RRM and zinc finger domains.
Orb2 is expressed in Drosophila melanogaster neurons, and
mutagenesis experiments and knockouts have shown that Orb2 is
essential for long-term (>24 hours) memory formation, while having no
impact on short term memory (Keleman, Krüttner, Alenius, & Dickson,
2007). These experiments have also demonstrated that Orb2A plays
an essential role in this process, despite its very low expression levels,
and that it can fulfill this role even if the RRM and zinc finger domains
are removed (Krüttner et al., 2012) (Majumdar et al., 2012). Orb2 has
been shown to concentrate in small points, referred to as puncta, in
insect cells (Majumdar et al., 2012). The exact structure of these
puncta is currently unknown, but Orb2 has been demonstrated to form
amyloids in-vitro, and this has led to the hypothesis that these puncta
are amyloids (Majumdar et al., 2012). It has also been shown that
small mutations in Orb2 that prevent these puncta from forming in cells
also block or disable memory formation in fruit flies. Notable among
these mutations are several demonstrating the importance of the
Orb2A N-terminus, such as F5Y (Majumdar et al., 2012).
Experiments by our collaborator, using cell extract, have shown
that Orb2 binds target mRNAs with equal affinity in both the amyloid
13

and monomer form. However, monomeric Orb2 causes the
degradation of mRNA it binds, while amyloid Orb2 binds the same
mRNA and then promotes its polyadenylation and translation (Khan et
al., 2015).
Many of the features of Orb2 discussed so far are consistent
with what has been observed for other CPEB proteins, for example
CPEB1 switches between a degrader and promoter of target mRNA
based on phosphorylation state (Ivshina et al., 2014), while ApCPEB in
Aplysia californica has also been shown to form amyloids and control
long term memory formation (Miniaci et al., 2008) (Raveendra et al.,
2013)(Si, Giustetto, et al., 2003; Si, Lindquist, & Kandel, 2003).
All the features described above have led to a popular
hypothesis where Orb2 amyloids are used to mark newly formed
synapses and prevent their otherwise inevitable degradation. Since
synaptic formation and preservation is a key step in long term memory
formation, this would explain Orb2’s key role in the process (Khan et
al., 2015) (Hervás et al., 2016) (Li et al., 2016). It has also been
hypothesized that amyloid formation is seeded by Orb2A, thus
explaining its essentiality, low expression levels, and the fact that it
does not require its RRM or zinc finger domains (Krüttner et al., 2012)
(Majumdar et al., 2012).
1.1.3. Droplets
Protein droplets are formed when proteins in solution undergo a
phase separation from the main bulk water. This is a true phase
separation, rather than aggregation, similar to what oils or lipids do in
water, and the droplets formed are nearly perfect spheres that undergo
fusion and fission (again parallel to lipids and oils) (Zhang et al., 2017)
(Wegmann et al., 2018) (Ambadipudi, Biernat, Riedel, Mandelkow, &
Zweckstetter, 2017) (Molliex et al., 2015) (Kato et al., 2012) (Elbaum-
Garfinkle et al., 2015). Fluorescence recovery after photobleaching
14

(FRAP) experiments show that, at least initially after droplet formation,
the proteins remain highly dynamic within the droplets, although some
droplets slowly “harden” and become less dynamic with time (Elbaum-
Garfinkle et al., 2015).
The internal structure of droplets is currently unknown, but a
variety of proteins have been shown to form them and several trends in
droplet formation have been identified. First, droplet formation almost
always increases as protein concentration increases and ionic strength
decreases (Zhang et al., 2017) (Ambadipudi et al., 2017) (Molliex et al.,
2015) (Elbaum-Garfinkle et al., 2015). Second, RNA has also been
shown to frequently drive droplet formation, even of proteins that lack
RRM domains (Wegmann et al., 2018) (Zhang et al., 2017) (Molliex et
al., 2015) (Kato et al., 2012). Third, amyloid forming proteins and
proteins with disordered domains (already overlapping categories, as
discussed above) seem to be particularly prone to droplet formation
(Kato et al., 2012).
Two important consequences are associated with protein
droplet formation. First, protein droplets are hypothesized to be the
basis membrane free organelles such as the P-bodies and RNA
granules. Thus, they provide an organizational and regulatory system
within the cell (Shin & Brangwynne, 2017). Second, some proteins
appear to progress from droplets into fibrils. This is consistent with the
“hardening” of droplets previously mentioned, and provides a route for
overcoming the initiation stage of fibril formation (Ambadipudi et al.,
2017) (Molliex et al., 2015) (Kato et al., 2012).
1.1.4. Huntingtin Exon 1
Huntingtin is one of the largest proteins expressed in humans,
with 3144 residues. It contains a polyglutamine domain in its first exon,
with 6 to 35 glutamine residues in healthy people. Expansion of this
domain beyond 35 residues causes Huntington’s disease, and animal
15

studies have shown that mere expression the exon 1 region alone, with
more than 35 glutamines, is enough to reproduce the disease. The root
of the disease is believed to be the uncontrolled growth of amyloid
fibrils formed from Exon 1 in the brain. It is hypothesized that a larger
glutamine tract makes the fibrils more stable and thus ensures their
proliferation. (Mangiarini et al., 1996) (Walker, 2007).
In addition to the central polyglutamine domain, Huntingtin exon
1 contains an N-terminal domain that resembles an amphipathic helix
and a C-terminal domain that is rich in proline. When expressed in
bacteria and purified, Huntingtin exon 1 can form both toxic and non-
toxic amyloid fibrils based on fibrilization conditions. Interestingly, NMR
and EPR studies have shown that these toxic and non-toxic fibrils do
not differ in their polyglutamine amyloid core, but rather in their proline
rich C-terminus (Nekooki-Machida et al., 2009) (Isas, Langen, &
Siemer, 2015) (Isas, Langen, Isas, Pandey, & Siemer, 2017).

1.2 Methods Used
1.2.1 NMR
Nuclear magnetic resonance (NMR) spectroscopy is based on
the magnetic moment and intrinsic angular momentum possessed by
many atoms. In particular, the atoms used for NMR in this study (
1
H,
13
C and
15
N) all have a spin quantum number of ½, meaning that they
have an angular momentum that is quantized between two distinct
states. These atoms also have a magnetic moment that always points
with (
1
H and
13
C) or against (
15
N) the angular momentum (Levitt, n.d.).
When placed in an external magnetic field (provided by a
massive magnet in the NMR spectrometer), an atom’s magnetic
moment can be observed as either aligned with or against the field.
The magnetic moment is slightly more likely to align with field, since
16

this is more energetically favorable. If numerous atoms are placed in a
field (as in a normal sample), a slight majority will align with the field,
giving the sample a collective magnetic moment. Due to their innate
angular momentum, the magnetic moment of an atom will rotate
around the field it is aligned with at a given frequency (the Larmor
frequency) (Levitt, n.d.).
If a radio frequency (RF) pulse is applied to the sample with the
same frequency as the Larmor frequency, the magnetic moments of all
atoms can be redirected from their alignment with or against the
external field to be perpendicular to the field. Thus, the sample as a
whole will now have its magnetic moment pointed perpendicular to the
external field (Levitt, n.d.).
Once the RF pulse has ended, the magnetic moment of the
sample will begin to rotate around external field, producing a
detectable signal, the free induction decay (FID), in the form of current
generated by this rotating magnetic field. The frequency of this signal
will match the Larmor frequency of the atoms and can be determined
by taking Fourier transform of the signal. The signal will dissipate as
atoms return to their equilibrium state, aligned with or against the
magnetic field (longitudinal relaxation) (Levitt, n.d.).
This method is informative because the electron environment
surrounding a nucleus can slightly change its Larmor frequency
(chemical shielding). Thus, an RF pulse can rotate the magnetic
moments of all
13
C atoms, and then each atom position in the sample
will give a slightly different signal (chemical shift), and thus a different
peak following the Fourier transform, based on their local environment
(Levitt, n.d.).
Atoms will also experience slight variations in their Larmor
frequency during the experiment, as molecular motions cause changes
to their environment and thus chemical shielding. These variations will
drive the dissipation of signal, as all atoms eventually fall out of sync
17

with each other and thus the sample no longer has an overall magnetic
moment (transverse relaxation) (Levitt, n.d.).
This study frequently utilized multi-dimensional NMR. In multi-
dimensional NMR, magnetization is transferred from one atom to
another at some time point after the magnetization of the first atom has
been rotated perpendicular to the static field and allowed to oscillate.
This is repeated at a different time points in the detection period to
eventually produce a multi-dimensional FID where each peak
represents atoms that have transferred magnetization between them.
Applying a multiple Fourier transforms to this FID will produce a multi-
dimensional spectrum where each peak corresponds to two or more
atoms that magnetization was transferred between (Levitt, n.d.).
This study also depended heavily on solid-state NMR (ssNMR).
ssNMR is performed on samples where molecules are too large to
tumble quickly in solution. This is important because the chemical shift
of an atom can depend heavily on its orientation in the external field. If
the molecules move fast enough in solution, then this orientation
dependence will be averaged out and not be an issue. However, if the
molecules do not move fast enough, the same atom in different
molecules will have very different chemical shifts, giving broad and
largely featureless peaks. The amyloids analyzed in this study are far
too large for effective molecular tumbling, and we addressed this using
magic angle spinning (MAS). In MAS, a sample is rapidly spun on an
air current (at 25 kHz in our case) to partially recreate the effect of
molecular tumbling. This averages most (but not all) of the line
broadening associated with molecular orientation, making the lines
much more narrow and easy to analyze (Levitt, n.d.).
This study utilized a number of different ssNMR pulse
sequences. Key to the difference between these pulse sequences, was
how magnetization was transferred between atoms in the sequence.
Magnetization can be transferred either through space via dipolar spin-
18

spin couplings (as was the case with variable amplitude cross
polarization (VACP), dipolar assisted rotational resonance (DARR),
dipolar recoupling enhanced by amplitude modulation (DREAM),
phase-alternated recoupling irradiation scheme (PARIS), and NCA
experiments) or transferred through molecular bonds via J-couplings
(as was the case with insensitive nuclei enhanced by polarization
transfer (INEPT) and total through bond correlation spectroscopy
(TOBSY) experiments). When magnetization is transferred through
space, the experiment is effectively selective for static regions of the
protein, because through space magnetization transfers only work
when the atoms remain fixed relative to each other (the dipolar
couplings that they depend upon can be averaged by motion). By
contrast, highly dynamic regions can be seen using through bond
transfers, and these experiments can be made selective for highly
dynamic regions by the fast
1
H based relaxation of the static regions
(Sun et al., 2012; Weingarth, Demco, Bodenhausen, & Tekely, 2009).
1.2.2 EPR
This project made use of electron paramagnetic resonance
spectroscopy (EPR) data and performed a small number of EPR
experiments. EPR is based on the same physical principals as NMR
(described above) but it works on the spin states of electrons rather
than nuclei. For an EPR signal to be generated, a protein must have
an unpaired electron, since paired electrons have opposite spin states
and thus effectively cancel each other out. This is typically
accomplished by attaching a S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-
1H-pyrrol-3-yl)methyl methanesulfonothioate (MTSL) label, referred to
in the following as a spin label, to the protein via a disulfide bond (Sahu,
McCarrick, & Lorigan, 2013) (Klare, 2013).
EPR experiments begin the same way as NMR experiments, by
placing a labeled sample in a magnetic field. An experiment can be
19

carried out using the same logic as described above for NMR (pulsed
EPR) or it can be done with the continuous wave (CW EPR) method.
In CW EPR, the sample is placed under a fixed microwave frequency
and its absorbance is measured. The strength of the magnetic field is
then scanned. Since the energy difference between the unpaired
electrons aligning with or against the magnetic field is dependent on
the strength of the field, a point will be reached where energy
difference matches the energy of the microwave frequency being
applied. At this point, an absorbance signal will be recorded as
unpaired electrons absorbed the microwave to transition spin states
(Sahu et al., 2013) (Klare, 2013).
CW EPR experiments are ideal for studying the dynamics of
MTSL labels and the distance between labels that are close (<20 Å) to
each other. Interactions with nearby spin labels and dynamics of the
MTSL label will both alter the CW EPR spectrum. Pulsed EPR is ideal
for studying the distance between spin labels that are further apart (20-
60 Å). Pulsed EPR can actually give distance distributions for a spin
label pair across all molecules in a sample (Sahu et al., 2013) (Klare,
2013).
1.2.3 Computer Modelling
Computer modelling in this project was done using Rosetta and
different molecular dynamics programs.
Rosetta is a molecular modelling software suite dedicated to
predicting the structure of proteins from sequence alone or from
sequence with additional physical information. Rosetta is built around a
simulated annealing algorithm and collection of force fields used to find
the lowest energy conformation for a given protein (Simons,
Kooperberg, Huang, & Baker, 1997).
In basic usage, Rosetta begins by taking a given protein
sequence and using it to construct an extended conformation peptide
20

matching that sequence. Rosetta then makes random “moves” on the
peptide, by adjusting the dihedral angles of a set of randomly selected
adjacent residues to match angles taken from a library of known
protein structures. After making a move, Rosetta scores the move and
the previous structure using the currently active force field. The move
is accepted or rejected based on a simulated annealing logic: moves
that lead to a lower system energy are always accepted, moves that
lead to a higher system energy are randomly accepted or rejected
based on the current system temperature and how unfavorable the
move is. As modelling progresses, the system temperature is lowered
(making high energy moves less likely to be accepted) and force field
used becomes more detailed. Eventually the temperature hits zero and
the current structure is returned (Simons et al., 1997).
As mentioned before, the force field used by Rosetta considers
a number of factors such as hydrophobic interactions, electrostatic
interactions and structure packing. If physical parameters are known
about the structure, they can be used to guide modelling by adding
constraints to the process. These are artificial forces that impact the
energy score of the model. For example, an idealized spring can be
placed between two atoms that are known to be a certain distance
from each other (Simons et al., 1997).
Molecular dynamics simulations model the time progression of a
molecular system (such as a protein) from a given starting
conformation. The basic design of molecular dynamics is that a system
is described in terms of atom positions, velocities, types (e.g. the Cα
atom of an alanine) and bonds. This information is combined with a
selected forcefield to calculate the current force vector on all atoms
and the system is advanced in time (typically .5 to 2 fs) based on this
information. Force vectors are then recalculated, and the process is
repeated (Brooks et al., 2009) (Eastman et al., 2013) (Phillips et al.,
2005).
21

When calculating force vectors, the most forcefields use several
simplified models to capture atomic interactions. First, bonds are
represented as idealized springs, holding a pair of atoms at a given
distance. Second, harmonic functions are also used to describe
dihedral and improper angles. Third, atoms are assigned a charge
based on the structure they are in (for example, a free chloride ion is
assigned a charge of -1, while the oxygen in water is assigned a partial
negative charge) and the electrostatic interactions between atoms are
calculated. Finally, the Van der Waals interactions between atoms are
described using the Lennard Jones (or 6-12) potential. This potential
requires 2 constants to be fitted for each atom type and gives a
simplified description of both the repulsive and attractive elements of
the Van der Waals interactions (Brooks et al., 2009) (Eastman et al.,
2013) (Phillips et al., 2005).
In performing molecular dynamics, several modifications are
often applied to the system described above to make calculations more
feasible and convenient. Some or all bonds may be classed as rigid
and completely fixed at their ideal length to make the system more
stable. The Van der Waals interaction is typically only calculated and
applied if atoms are within a given distance, otherwise it would need to
be calculated for all atom pairs in the system. In some simulations,
water may not be modeled at all (implicit solvent), and its impact is
approximated by changes to the Van der Waals constants for all atoms.
If water is modeled, the edges of the system present an obvious
challenge, and this is typically addressed via a periodic boundary
condition (PBC) where the simulation occurs in a box and atoms or
force vectors that exit one end of the box reenter in the opposite end.
The size of this box can be dynamically adjusted as the simulation runs
to keep a consistent pressure in the system. Likewise, random forces
or dampening can be applied to atom movement to control
temperature. Finally, just as in Rosetta, artificial forces based on
22

known physical data can be added into the system (Brooks et al., 2009)
(Eastman et al., 2013) (Phillips et al., 2005).
1.2.4 Thioflavin T Fluorescence
Thioflavin T (ThT) is fluorescent dye that is widely used to
detect the presence of amyloid fibrils. On binding β-sheet rich
structures (such as amyloids), ThT typically undergoes an increase in
fluorescence efficiency and a red shift of its emissions spectrum. Thus,
adding ThT and checking fluorescence relative to a blank is an easy
way to test for the presence of most amyloids. However, it should be
noted that some amyloids are known to not change ThT fluorescence,
while some non-amyloid protein aggregates are (Gade Malmos et al.,
2017).
1.2.5 Chromatography
This project was dependent on two types of chromatography:
Nickle affinity and size exclusion. Like all chromatographic methods,
these techniques depend on the separation of proteins and other
molecules between mobile and immobile phases, and if two molecules
or proteins show different preferences for the mobile and immobile
phases, they can be separated (Oliveira & Domingues, 2018).
In nickel affinity chromatography, the immobile resin displays Ni
ions that proteins can bind to. Proteins of interest are typically
expressed with a series of C-terminal histidine residues (the his-tag)
that will bind this ion, while most proteins show no preference for it. A
solution containing numerous proteins and the his-tagged expressed
protein can be run through a Ni resin containing column with only the
desired protein remaining on the column. This protein can then be
eluted using a low pH or imidazole, both of which disrupt the
interaction between Ni ions and his-tags (Oliveira & Domingues, 2018).
23

Size exclusion chromatography is based on a resin made of
porous beads. As a solution containing proteins or other molecules is
run through the beads, some molecules may become temporarily stuck
in the pores of the beads (the immobile phase) if they are small
enough. The smaller a molecule is, the more likely it is to become
stuck and the slower it will pass through the column. At a certain size,
molecules cannot enter the beads at all, and everything over this size
(the exclusion limit) passes through the column at the same pace
(Barth, Boyes, & Jackson, 1994).

24

CHAPTER 2. PROJECT I: STUDIES
ON NON-SOLUBLE FRACTION
ORB2
2.1 Introduction
I began this project hoping to address two major questions
about Orb2. First, the studies discussed in the introduction suggested
that Orb2A seeded fibril formation (Krüttner et al., 2012; Majumdar et
al., 2012), and I aimed to reproduce and study this seeding effect in
vitro. Knowledge of how Orb2 fibril formation is induced (and thus
regulated) can help us understand how fibrils are formed at only the
right place and time (in contrast to pathological amyloids). Second,
since amyloid formation has been shown to change Orb2’s
biochemical activity (Khan et al., 2015), I sought to obtain the structure
of the amyloid to determine how this could occur. This could offer
major insights into the mechanisms of other functional amyloids.
To begin my studies on Orb2, I started purifying and working
with protein expressed in the non-soluble fraction of Escherichia coli.
This was a logical choice since other groups had already shown that
Orb2 from this source could be purified and grown into amyloid fibrils
(Majumdar et al., 2012).
2.2 Methods
2.2.1 Protein Expression
The Orb2 expression, purification and fibril formation methods I
used in this study were based on the methods used in (Majumdar et al.,
2012). Plasmids containing the Orb2A and Orb2B genes, as well as a
kanamycin resistance gene, were generously provided by Dr. Kausik
25

Si at the University of Kansas School of Medicine. These plasmids
were based on the Drosophila melanogaster Orb2 coding sequence
and thus contained codons that are rare in Escherichia coli. To
compensate for this, expression was done using Rosetta (DE3) cells
from Novagen. These cells carry an additional plasmid encoding
tRNAs matching rare codons and a chloramphenicol resistance gene.
To grow cells needed for protein expression, a starter culture
was first made by resuspending one colony of transfected cells (the
transfection procedure is available in the appendix) in a flask of 50 ml
LB with chloramphenicol and kanamycin. The starter culture was then
allowed to grow overnight at 37°C, 200 RMP. The next day, 1 ml of the
starter culture (which was very cloudy at this point), was used to seed
a main culture of 1 L LB with the same concentration of
chloramphenicol and kanamycin. The main culture was grown at 37°C
with 225 RPM shaking until it reached a 600 nm OD of 0.6. On
reaching an OD of 0.6, expression was induced by adding 1 mM IPTG
to the culture, and the cells were left to express at 32°C for 4-6 hours.
When expression was completed, cells were sedimented
(Sorvall SLC-6000 rotor, 4000 RPM, 20 min, 4°C) and the supernatant
was discarded. The pellet was stored at -80°C until needed.
2.2.2 Labeled Expression
To produce
13
C and
15
N labelled Orb2 protein for NMR
experiments, a version of the protocol just described was used, with
modification based on (Marley, Lu, & Bracken, 2001). This modified
protocol was begun the same way as the previous protocol, up until the
point where the main culture achieved a 600 nm OD of 0.6. At this
point, the cells were sedimented at room temperature (Sorvall SLC-
6000 rotor, 4000 RPM, 20 min, room temperature) and then
resuspended in 1 L M9 wash buffer (50 mM Na2HPO4, 22 mM KH2PO4,
8.6 mM NaCl). Cells were spun down again, resuspended in M9
minimal media (Marley et al., 2001) and allowed to recover in the
26

incubator at 32°C for 1 hour. Following recovery time, cells were
induced with 1 mM IPTG and left to express protein overnight at 32°C.
Following expression, cells were spun down and stored at -80°C as
described in the previous protocol.
2.2.3 Protein Purification
To purify the protein, the cell pellet was resuspended in 50 ml
lysis buffer (50 mM Tris, 100 mM NaCl, 0.5% v/v Triton X-100, 0.05%
v/v β-mercaptoethanol, 1 mg/ml lysozyme and 1X Pierce Protease
Inhibitor at pH 7.6) and sonicated. The sonicated mixture was spun
down (Sorvall SS-34 rotor, 15 min, 10000 RPM) and the supernatant
was discarded. The pellet was then resuspended in another 50 ml lysis
buffer and sonication and spin down were repeated. The pellet was
then resuspended in denaturing buffer (6M GuHCl or 8M urea, 250 mM
NaCl, 100 mM Na2HPO4, 10% v/v Glycerol, pH 8.0) and sonicated.
Following sonication, remaining insoluble chunks were broken up
manually with a spatula, and the solution was allowed rest under slight
agitation for overnight at room temperature.
Following incubation, the solution was spun down (Sorvall SS-
34 rotor, 20,000 RMP, 4°C, 20 min) and the supernatant was poured
onto a gravity column that consisted of 5ml Ni-NTA resin (10 ml of
slurry) and had been equilibrated with denaturing buffer. The column
was mixed and incubated with the solution on a rocker for 1 hour. After
incubation, flow-thru was eluted and the column was washed with 25
ml of Denaturing buffer with 0.5% Triton X-100, denaturing buffer with
0.5 M NaCl, and denaturing buffer at pH 6.75. The column was then
eluted twice with 40 ml of denaturing buffer at pH 3.75. Elute purity
was confirmed by running the product on an SDS-PAGE gel.
2.2.4 Fibril Growth
Following purification of Orb2A or Orb2B in denaturing buffer,
fibrils were formed by dialyzing the Orb2 protein (with a concentration
27

of about 20 uM) into GT buffer (1 M urea, 100 mM KCl, 10 mM HEPES,
1 mM DTT, pH 7.6) at 4°C with slow stirring. Fibril formation was
monitored via electron microscopy or ThT fluorescence.
2.3 Results
2.3.1 Successful Protein Expression and Purification
The procedures described above gave an excellent yield of
purified Orb2A or Orb2B protein. With a typical yield of ~1 umol per a
liter of culture. Yield for labelled protein was typically a bit less, but still
more than enough.
28

2.3.2 Not All Protein Binds the Ni-Column

Figure 2. Extra Orb2 Protein Can be Found in the Flow-Thru.
Orb2A (top) and Orb2B (bottom) were purified and the flow-thru from the purifications
was stored. Orb2A elutes 1 and 2 had concentrations of 30 and 20 uM respectively.
Orb2B elutes 1 and 2 both had concentrations of ~15 uM. 5 days later, the flow-thru
was run on the same column using the same purification procedure and additional
protein was observed to bind to the column and come off in the elute fraction. The
new Orb2A elutes 1 and 2 were ~10 and ~5 uM respectively. The new Orb2B elutes
1 and 2 were ~18 and ~3 uM respectively. This effect was highly reproducible.
An odd, but highly reproducible, phenomenon was identified
when it was observed that ~1/3 of the Orb2 protein (A or B) being
purified would run through the Ni column being used and come out in
the flow-thru (see Figure 2). Prolonging the incubation time with the
29

resin, using more resin and changing the brand of resin failed to
prevent this from occurring. Further, it was found that if the flow-thru of
a column was run back thru the column again, the Orb2 protein
present would again split between the flow-thru and elution fractions at
the same ratio. However, this only occurred if the flow-thru was stored
and rerun on the column a day or more after the original column. An
attempt to rerun the flow-thru immediately after the first column
resulted in all remaining Orb2 ending up in the flow-thru fraction.
2.3.3 Orb2 Purified from the Non-Soluble Fraction Forms Fibrils
Rapidly

Figure 3. Orb2 Shows a Rapid Gain in ThT When Placed in Fibril Forming Conditions.
Orb2A and Orb2B in denaturing buffer were dialyzed in to GT buffer for 1 hour and
mixed with ThT. Fibril formation was monitored by recording ThT fluorescence using
an Eppendorf AF2200 PlateReader. PlateReader settings are available in the
appendix.
Using the fibril formation method described above, fibrils formed
rapidly, as indicated by ThT florescence (Figure 3) and the formation of
visible aggregates (confirmed by EM to be fibrils). Filtration and ultra-
centrifugation to remove possible seeds prior to fibril formation does
not seem to have any impact on this process.
30

2.3.4 Fibrils Grown from Non-Soluble Fraction Protein Are Large and
Highly Bundled

Figure 4. Fibrils Grown from Non-Soluble Fraction Protein Are Large and Highly Bundled.
EM images of Orb2A (a) and Orb2B (b) fibrils grown from protein purified from the
non-soluble fraction show large fibrils with extensive bundling. Images were taken on
a Jeol Transmission Electron Microscope and the procedure for preparing EM grids is
available in the appendix.
EM grids of Orb2A or Orb2B fibrils grown from protein purified
from the non-soluble fraction showed fibrils that were very large and
bundled extensively (Figure 4)
31

2.3.5 NMR Experiments that Highlight the Static Domains of the Fibril
Show Extensive Overlap and Poor Resolution

Figure 5. NMR Experiments that Highlight the Static Domains of Orb2A and Orb2B Fibrils
Grown from Non-Soluble Fraction Protein.
2D
13
C-
13
C DREAM (top), DARR (middle) and
15
N-
13
C NCA (bottom) spectra of
Orb2A (left) and Orb2B (right)
13
C and
15
N labeled fibrils grown from non-soluble
fraction protein. The low resolution of these spectra, combined with lack of any
residue types that only occur once in the Orb2A or B sequences, made connecting
any peaks to particular residues or gaining significant structural data from these
spectra impossible. However, these spectra do demonstrate an extensive structural
similarity between Orb2A and Orb2B, with almost all major peaks being found in both
samples (the spectra may appear slightly different due to signal to noise differences
between the samples). Spectra were recorded on a 600 MHz Agilent wide bore
spectrometer using a 1.6 mm sample rotor with 25 kHz MAS.
32

ssNMR experiments focused on the static domain of Orb2A and
Orb2B fibrils grown from non-soluble fraction protein tended to give
poor resolution spectra (Figure 5). This poor resolution, combined with
the lack of any residue types that only occur once in the Orb2A or B
sequences, made connecting any peaks to particular residues or
gaining significant structural data from these spectra impossible.
However, it is worth noting that the spectra of Orb2A and Orb2B fibrils
show near perfect overlap.
2.3.6 Fibrils Grown from Non-Soluble Fraction Protein Show Limited
Fibrilization Condition Impact

Figure 6.
1
H-
13
C Cross Polarization Spectra of Orb2A Fibrils Grown from Non-Soluble
Fraction Protein in Different Conditions.
In the standard fibril formation procedure,
13
C and
15
N labeled Orb2A was purified,
dialyzed in GT buffer and then left in a 15 ml tube on a rocker at room temperature
(Baseline). Alternatively, the sample was dialyzed against GT with isoxazole, zinc, no
urea, no KCl, 300 mM (extra) KCl, or pH 5.0 GT buffer. Also, after dialyzing into
normal GT buffer, we experimented with leaving the sample on a rocker at 4°C (low
temperature), not on a rocker (no shaking), or in a flask with a stir bar (stirring).
1
H-
13
C cross polarization spectra of fibrils from all samples were recorded on a 600 MHz
Agilent wide bore spectrometer using a 1.6 mm sample rotor with 25 kHz MAS.
In an effort to improve spectra resolution, we experimented with
changing the conditions of fibril formation. We purified a single batch of
13
C and
15
N labeled Orb2A and allowed it to fibrilize in a variety of
different conditions, including the standard fibrilization procedure
33

described above, and recorded 1D
1
H-
13
C cross polarization spectra of
fibrils from each condition. Most of the conditions we tried were slight
variations on our standard fibrilization procedure. However, we also
tried adding zinc to the buffer on the grounds that the Orb2 zinc finger
domain may require it to fold properly, and we tried adding isoxazole
on the basis of a study by Kato et al (Kato et al., 2012), which showed
that isoxazole can induce the formation of some amyloid fibrils.
We found that fibril formation conditions had little impact of on
the quality of the spectra of the fibrils finally obtained (Figure 6).
Interestingly, fibrils formed at pH 5.0 did show a change in visible
appearance and resembled a spider web clinging to the tube, but this
did not correspond to a significant change in spectra.
2.3.7 Fibrils Grown from Non-Soluble Fraction Protein Have Well
Resolved INEPT based Spectra

Figure 7. INEPT based Spectra of Orb2A and Orb2B Fibrils Grown from Non-Soluble Fraction
Protein.
34

INEPT-TOBSY (top) and
1
H-
13
C HETCOR (bottom) spectra of fibrils grown from non-
soluble fraction Orb2A (left) and Orb2B (right). The spectra are well resolved. As was
seen in static selective spectra, there is still a considerable overlap between Orb2A
and Orb2B. Spectra were recorded on a 600 MHz Agilent wide bore spectrometer
using a 1.6 mm sample rotor with 25 kHz spinning.
In contrast to the static domain highlighting spectra just
described, fibrils formed from non-soluble fraction Orb2A and Orb2B
gave much more well resolved INEPT based ssNMR spectra (Figure 7).
Based on this, Orb2A was analyzed further using liquid state, proton
detected, spectra such as HNCO, HNCA and HNcoCA. Using these
methods, peak assignments were obtained for residues G159-T169,
demonstrating that this region is highly dynamic. This success served
as the basis for the paper: Dynamic domains of amyloid fibrils can be
site-specifically assigned with proton detected 3D NMR spectroscopy
(Falk & Siemer, 2016). A complete copy of this paper can be found in
the appendix. It was also observed that these dynamic selective
spectra of Orb2A and Orb2B also have a significant overlap.
2.4 Discussion
Although this project was unable to determine the structure or
location of the fibril core or extensively study the kinetics of fibril
formation, several useful pieces of information can be obtained from
the data gathered.
First, the loss of Orb2 protein in flow-thru during the purification
procedure suggests that denatured Orb2 has at least two
conformations, one of which does not bind Ni despite possessing a
HIS-tag (a very unusual behavior). The fact that re-purification of the
flow-thru only works when the second column is only run at least a day
after the first column suggests that the exchange rate between these
two conformations is very slow.
Second, non-soluble fraction Orb2A or Orb2B seems to form
fibrils extremely rapidly, in a variety of conditions and without the need
35

of seeding. This suggests that the fibrils being formed have high kinetic
and thermodynamic favorability.
Third, spectra of Orb2A and Orb2B show near perfect overlap.
This indicates that the fibrils share a very similar structure, including in
their cores. Going further, this suggests that the fibril core is the same
for both proteins and therefore not in their unique N-termini. The lack of
strong glutamine signals in spectra that highlight the static regions of
the fibrils also suggests that the glutamine rich region is not part of the
fibril core.
Finally, fibrils grown from non-soluble fraction protein give
strong and non-overlapping signals in spectra that are both highly
selective for static regions (such as DREAM) and highly selective for
dynamic regions (such as HSQC). This shows that the fibrils have a
range of dynamics over their domains and likely contain both a fibril
core that is almost immobile and one or more other domains that move
very rapidly.

36

CHAPTER 3. PROJECT II: STUDIES
ON SOLUBLE ORB2 AND ΔR RM
FRAGMENTS
3.1 Introduction
Due to the difficulty of obtaining high resolution static enhancing
spectra from non-soluble fraction Orb2 protein, as discussed above,
we attempted to purify and work with Orb2 expressed in the soluble
fraction of E. coli. This purification was accompanied by size-exclusion
chromatography to ensure the protein was monomeric. Because
purifying Orb2A and Orb2B through this method gave low yields, we
frequently worked with fragment versions of the two proteins called
Orb2AΔRRM and Orb2BΔRRM, which had much better yields
(discussed below).
3.2 Methods
3.2.1 Protein Expression
Expression was performed using either the Orb2A and Orb2B
full length plasmids previously described, or a pair of custom ordered
plasmids, which contained genes for the Orb2AΔRRM or Orb2BΔRRM
fragments and a kanamycin resistance gene. Orb2AΔRRM and
Orb2BΔRRM are Orb2A and Orb2B without their RRM and zinc finger
domains (see the discussion of the Orb2 sequence in the introduction).
This corresponds to Orb2A residues 1-320 and Orb2B residues 1-473.
Transfection prior to expression was done in the same manner
as for the non-soluble fraction protein expression (see appendix).
Orb2A and Orb2B full length plasmids were still transfected into
Rosetta (DE3) cells. The Orb2AΔRRM and Orb2BΔRRM plasmids
were codon optimized for E. coli and were transfected into BL21 cells
37

from New England Biolabs that were then grown on a plate with only
kanamycin.
To grow cells for protein expression, a colony from the
transfection plate was used to seed a 50 ml LB starter culture with the
appropriate antibiotics. This starter culture was grown at 37°C with 225
RPM shaking until it was visibly hazy. At this point, the starter culture
was used to seed 8 500 ml LB main cultures with appropriate
antibiotics, with 5 ml of seed per a culture. These cultures were also
grown at 37°C with 225 RPM shaking until they reached 0.6 OD at 600
nm. On reaching 0.6 OD, each flask was induced with 1 mM IPTG and
transferred to a pre-cooled incubator at 16°C, the flasks were then left
to express at this temperature with 160 RPM shaking for 16 hours.
Following expression, cells were spun down and stored using
the same procedure as described for non-soluble fraction protein (see
2.2.1). For a labeled expression, cells were washed and transferred
into 1 L M9 minimal media on reaching 0.6 OD at 600 nm in the same
manner as was done for the non-soluble fraction expression, and then
left to express at 16°C with 160 RPM shaking for 24 hours.

3.2.2 Protein Purification
To purify protein, cell pellets were first resuspended in 100 ml
GT buffer (1 M urea, 100 mM KCl, 10 mM HEPES, 0.05% β-
mercaptoethanol, pH 7.6) with 20 mM imidazole, 40 mM NaCl, 1 mg/ml
lysozyme and 1X Pierce Protease Inhibitor. Following resuspension,
cells were sonicated on ice for a total of 9 minutes, using a QSonica
Q125 sonicator with amplitude set to 100%. Sonication was broken up
into 59 second pulses interspersed with 30 second pauses. After
sonication, the lysate was spun down (SS-34 rotor, 4°C, 10,000 RPM,
15 min) and the supernatant was run through a 0.22 um filter.
Protein was purified from this filtrate using a either a 5 ml FPLC
Ni affinity column (either a GE HisTrap HP column or a Bio-Rad Bio-
38

Scale Mini Profinity column) that was pre-equilibrated with GT buffer.
Filtrate was run thru the column at a rate of 2.5 ml/min. The column
was then washed at a rate of 5 ml/min with 25 ml of each buffer in the
following order:
• GT buffer with 0.5% Triton-X 100
• GT buffer with 0.5 M NaCl
• GT buffer with 40 mM imidazole and 80 mM NaCl
• GT buffer with 1 M NaCl
The protein was then eluted using 25 ml GT buffer with 500 mM
imidazole and 1 M NaCl. One ml fractions were collected. Following
elution, the 6 elute fractions with the highest 280 nm UV reading (while
accounting for the spike and settling in absorbance due to imidazole in
the elution buffer) were pooled. If fewer than 6 fractions showed
significant absorbance, then all the ones that showed significant
absorbance were pooled. This typically resulted in fractions 6th to 11th
1 ml fractions being pooled.
Pooled fractions were filtered using a .2 µm syringe filter. 5 ml of
filtrate (or whatever was available) was run on a HiLoad 16/60
Superdex 200 pg column (Orb2AΔRRM) or a HiPrep 16/60 Sephacryl
S-300 HR column that was pre-equilibrated with GT buffer at .5 ml/min
(Orb2B and Orb2BΔRRM). Usage of a high resolution Superdex
column was necessary for Orb2AΔRRM purification due to the
presence of an unidentified contaminating protein that needed to be
resolved from Orb2AΔRRM (see results). Orb2A showed the what
appeared to be the same contamination and has not yet been
successfully purified (purification with the HiLoad 16/60 Superdex 200
pg column has not been attempted yet). Orb2AΔRRM eluted at ~72 ml
after starting protein injection onto the size exclusion column, while
Orb2B and Orb2BΔRRM eluted at ~60 ml. Elute fractions could then
be used for experiments immediately or stored by flash freezing in
liquid nitrogen and transferring to -80°C. Concentration was raised or
39

lowered as needed by spin concentrating with Amicon Ultra-4
centrifugal filters (3 kDa cutoff) or dilution with GT buffer.

3.2.3 Droplet Formation and Analysis
Droplet formation in all proteins could be induced by exchanging
20 µM protein from GT buffer into a 10 mM HEPES, 0.05% β-
mercaptoethanol, pH 6.5 buffer (droplet forming buffer). This exchange
was done via a Sephadex G-25 PD-10, MidiTrap or MiniTrap column
depending on what volume was being exchanged. Droplets were
directly visualized via an Axio Imager Z1 microscope in DIC mode and
600 nm OD was used as a proxy for droplet density.
To better visualize droplets, the protein was sometimes labeled
with Alexa Fluor 488 C5 Maleimide dye prior to droplet formation. To
label the protein, it was first exchanged into GT buffer with 5 mM TCEP
in place of β-mercaptoethanol via the one of the Sephadex columns
just described. 55 µM dye was then added, and the solution was left in
a glass tube wrapped in aluminum foil on a rocker overnight. A droplet
forming buffer exchange, as described above, was carried out the next
day, again with TCEP in place of β-mercaptoethanol. In this procedure,
β-mercaptoethanol was replaced by TCEP so that it could not react
with the maleimide group of the dye. The second exchange both
removed unattached dye and induced droplet formation.
At this point, the protein could be visualized using the same
microscope in fluorescence mode, with 488/10 nm excitation and
525/50 nm emissions.

3.2.4 Growth of Fibrils from Droplets
EM and fluorescence microscopy showed that fibrils naturally
form in samples containing Orb2AΔRRM and Orb2BΔRRM droplets if
left on a rocker and given time (see results, fibril formation from
droplets for the full-length proteins has not yet been tested). In order to
40

improve fibril homogeneity for NMR experiments, droplets were formed
using the previous method, and then left in a small flask with very light
stirring.

3.2.5 Protein EPR Labelling
In order to analyze droplets and fibrils using EPR, we obtained a
mutant plasmid, (Q45C and C10S), of Orb2AΔRRM and developed a
method of labelling this protein for EPR. The protein was expressed
and purified as described above (although a HiPrep 16/60 Sephacryl
S-300 HR column was used), except during the Ni column purification
procedure, where the column was washed with 30 ml of GT buffer
containing 30 mg/ml MTSL following the imidazole wash. No β-
mercaptoethanol or any other reducing agent was used in the
imidazole wash or any step afterwards. This method proved necessary
because we were unable to label the protein following purification and
then separate the protein from the unattached spin label (see results).
3.3 Results
3.3.1 Successful Protein Expression and Purification

50
100
150
200
250
300
0 20 40 60 80 100
280nm Absorbance (mAU)
Volume
Orb2B Sephacryl S-300 Purification
41

Figure 8. Size Exclusion Chromatograms for Orb2B, Orb2BΔRRM, and Orb2AΔRRM
Purifications.
Proteins were purified on a Ni affinity column followed by a size exclusion column to
ensure they were pure monomers. The desired protein peak is marked with a black
arrow. Orb2B and Orb2BΔRRM eluted at nearly the same position on the same
column, indicating that the RRM and zinc finger domains of Orb2 are well folded and
have little impact on the radius of gyration. The elution profiles of Orb2B and
Orb2AΔRRM show large peaks in the exclusion limit of the column and smaller peaks
after the desired protein. SDS-PAGE gels show almost no protein content in the
exclusion limit peak, this is hypothesized to be due to the peak containing a low
concentration of large light scattering species (although it is also possible that
species in the peak was simply too large to enter the gel), and a smaller
contaminating species (hypothesized to be a cleavage product of the desired protein)
in the following peak.
The methods previously described allowed us to produce and
purify Orb2B, Orb2AΔRRM, and Orb2BΔRRM. Orb2B and
0
50
100
150
200
250
0 20 40 60 80 100
280nm Absorbance (mAU)
Volume (ml)
Orb2BΔRRM Sephacryl S-300 Purification
0
20
40
60
80
100
120
140
160
180
200
0 20 40 60 80 100
280nm Absorbance (mAU)
Volume (ml)
Orb2AΔRRM Superdex 200 Purification
42

Orb2BΔRRM elute off the HiPrep 16/60 Sephacryl S-300 HR column at
nearly the same position. This suggest that the RRM and zinc finger
domains of both Orb2 proteins are well folded and contribute little to
the radius of gyration compared to the N-terminus. Orb2B purification
had a yield that was considerably lower than that of Orb2AΔRRM and
Orb2BΔRRM. All of the protein products described here, besides
Orb2BΔRRM, seemed to have a significant contaminating protein,
believed to be a cleavage product, that can only be separated from the
protein using the size exclusion column (see Figure 8).

43

3.3.2 Formation of An Apparent Oligomer

Figure 9. DLS Readings of Orb2B Following Purification.
Monomeric Orb2B was purified from the soluble fraction of E. coli and DLS
measurements were taken shortly afterwards. These measurements indicate
progression from a monomeric to an oligomeric species.
Observation of purified Orb2B by dynamic light scattering (DLS)
showed ~3 and ~12 nm species. The ~3 nm species may correspond
to the Orb2B monomer, as the published structure of the CPEB4 RRM
motifs (with clear sequence homology to the Orb2 RRM motifs) has an
~2 nm radius of gyration and it is quite possible that the residues
0
5
10
15
20
25
0.01 0.1 1 10 100 1000 10000 100000
% Mass
Particle Diameter (nm)
Orb2B DLS Results at 3 Hours
0
5
10
15
20
25
30
0.01 0.1 1 10 100 1000 10000 100000
% Mass
Particle Diameter (nm)
Orb2B DLS Results at 24 Hours
44

outside of this motif may add another 1 nm (Afroz et al., 2014). The
~12 nm species is consistent with a 12 nm oligomer discovered by our
collaborators in the Si lab (via EM) when purifying Orb2 from
Drosophila brain extract (preliminary data). Further, when purified
Orb2B was left at room temperature for several days, a shift was
observed in its DLS spectrum, possibly indicating a progression from
the monomer to the oligomer (see Figure 9).
An attempt to spin down this oligomer was made (TLA-100.3
rotor, 25°C, 60,000 RPM, 6 h), and this gave a very small and clear
pellet.

3.3.3 Formation of Droplets and Fibrils
Note: Extended figure, see caption for Figure 10 below.

0 HR FLUORESCENCE
45

0 HR DIC
0 HR EM
46

6 HR FLUORESCENCE
6 HR DIC
47

6 HR EM
12 HR FLUORESCENCE
48

12 HR DIC
12 HR EM
49

24 HR FLUORESCENCE
24 HR DIC
50

25 HR EM
48 HR FLUORESCENCE
51

Figure 10. Droplet and Fibril Formation by Orb2BΔRRM.
When purified monomeric Orb2BΔRRM was exchanged into a salt free buffer, it was
observed to immediately form droplets. These droplets were visible using differential
interference contrast (DIC) and fluorescence (if the protein was previously labeled
with fluorescent dye) microscopy. When the droplet solution was left with on a rocker,
electron microscopy (EM) showed the accumulation of fibrils, although the initial
droplets did not seem to be visible in EM. The observation of large spider web like
meshes around bright puncta in fluorescence microscopy, combined with the
corresponding appearance of large fibrils apparently originating from extremely
48 HR DIC
48 HR EM
52

protein rich points in EM images, suggests fibrils directly grow out of droplets if given
time. Further, the large structures seen in fluorescence and DIC microscopy at 24
and 48 h are consistent with a naked eye observation of hydrogels in droplet
containing solutions at later timepoints.

0 HR FLUORESCENCE
0 HR DIC
53

Figure 11. Droplet and Fibril Formation by Orb2AΔRRM.
Like Orb2BΔRRM (see Figure 10), Orb2AΔRRM was observed to form droplets, that
appeared to progress into fibrils when exchanged into salt free conditions. A more
complete dataset (currently unavailable due to technical issues) is needed to
determine if Orb2AΔRRM follows all the patterns of droplet and fibril formation that
Orb2BΔRRM does.

48 HR EM
0 HR FLUORESCENCE
54

Figure 12. Orb2B Droplet Formation.
Orb2B (labeled with Alexa Fluor 488 C5 Maleimide dye) immediately formed droplets
on exchange into salt free buffer. Droplets were visible using both fluorescence and
DIC microscopy. Fibril formation from Orb2B droplets has not yet been tested.

Figure 13. Orb2BΔRRM Droplet ThT Fluorescence.
Samples containing Orb2BΔRRM droplets showed an increase in ThT florescence
after ~24 hours. This may correspond to the formation of fibrils in these samples,
although it should be noted that electron microscopy shows fibril formation
considerably before the 24 hour timepoint where ThT fluorescence starts increasing.
-5000
5000
15000
25000
35000
45000
55000
0 10 20 30 40 50 60 70 80 90
ThT Fluorescence (Au)
Time (hours)
Orb2BΔRRM Droplet ThT Fluorescence
0 HR DIC
55

ThT fluorescence was monitored using an Eppendorf AF2200 PlateReader.
PlateReader settings are available in the appendix.
Droplets were observed after using the formation procedure
described in methods (see Figure 10, Figure 11, and Figure 12). Fibrils
slowly formed in samples containing Orb2BΔRRM and Orb2AΔRRM
droplets without any additional procedures. Small hydrogels were also
observed in Orb2BΔRRM droplet samples after ~24 h via naked eye.
Fibril formation from samples containing Orb2B droplets has not yet
been tested. Microscopy and EM images strongly suggest that
Orb2BΔRRM fibrils grew out of the droplets (see Figure 10).
Orb2AΔRRM produces fibrils with a similar appearance that may form
the same way, but a more complete dataset that will allow confirmation
of this is currently lacking (see Figure 11). Samples containing
Orb2BΔRRM droplets also showed an increase in ThT fluorescence
that may correspond to fibril formation (see Figure 13).

3.3.4 NMR on Orb2BΔRRM Fibrils

Figure 14. Electron Microscopy of Orb2BΔRRM Fibril NMR Sample.
A sample of Orb2BΔRRM fibrils was prepared for NMR analysis by exchanging
13
C
and
15
N labeled Orb2BΔRRM into salt free buffer as described in Droplet Formation
and Analysis. The sample immediately formed droplets and the solution was left in
56

very small flask with slow stirring for 48 hours, as this was found to form a solution of
highly homogenous fibrils (see image) that could be more easily spun down for
packing into an ssNMR rotor than fibrils produced without stirring.
Orb2BΔRRM fibril NMR samples were prepared as described in
the methods section. Stirring during fibril formation created fibrils that
were highly homogenous and stuck to each other, which allowed them
to be easily spun down and packed into the ssNMR rotor (see Figure
14).

57

Figure 15. NMR Experiments on Orb2BΔRRM Fibrils.
PARIS
DREAM
58

Spectra were recorded on a 600 MHz Agilent wide bore spectrometer using a 1.6 mm
sample rotor with 30 kHz MAS for the DREAM spectrum and 25 kHz MAS for all
other spectra.
NMR experiments based on dipolar coupling (CP, PARIS,
DREAM) J-coupling (INEPT) and direct excitation all gave detectable
signal when being used to analyze the Orb2BΔRRM fibrils. Since these
different classes of experiments highlight static domains, dynamic
domains, and have no preference (respectively), this indicates that a
range of different dynamics are present in the fibril (see Figure 15).
Dipolar coupling based experiments gave very different spectra than
were obtained using the same or similar (PARIS and DARR)
experiments on fibrils grown from Orb2B purified from the non-soluble
fraction (see Figure 5). These experiments also showed strong alanine,
threonine, serine, proline, valine, isoleucine and glycine peaks with
chemical shifts that were compatible with either a β-strand or random
coil conformation (incompatible with an α-helix) (Wang & Jardetzky,
2002). This indicates that these residues contribute to the fibril core,
however poor
15
N resolution and insufficient signal to noise prevented
sufficient information being gained to connect any peaks to particular
residues. Peaks attributable to glutamine were relatively weak in these
spectra, despite the abundance of glutamine in the Orb2BΔRRM
sequence. This suggests that the glutamine rich region is not the fibril
core. No Cα-Cβ peaks could be assigned to glutamine with full
confidence so chemical shifts could not be used to predict its
secondary structure.
59

3.3.5 Successful EPR Labelling of Soluble Protein

Figure 16. EPR Spectrum of Cysteine Free Orb2AΔRRM Mutant.
Cysteine Free version of Orb2AΔRRM was expressed, purified normally before being
“labeled” with MTSL before free spin label was removed via buffer exchange with a
PD-10 column. Despite this buffer exchange and the lack of any cysteine residues for
MTSL to react with, an EPR spectrum with a strong biradical component was
observed.

Figure 17. EPR Spectrum of Orb2AΔRRM Q45C, C10S Mutant.
A Q45C, C10S mutant was expressed and labeled with MTSL during the Ni column
purification procedure (see 3.2.5). When purification was completed, the resulting
sample gave an EPR spectrum with no sign of free spin label.
We initially attempted to spin label our protein after purification
and then remove excess spin label using a Sephadex G-25 MiniTrap
60

column. However, this method resulted EPR spectra that included a
small shoulder consistent with free spin label in the form of a biradical,
indicating label removal was not completely successful. A control study
with a cysteine free Orb2AΔRRM mutant showed that this problem
occurred even when there was no covalent interaction between the
protein and the label (see Figure 16). To address this issue, we
performed spin labeling on the Ni column during purification as
described in the methods section. EPR spectra recorded from this
sample showed successful labeling without any free spin label
shoulder (see Figure 17).
3.4 Discussion
Although a number of additional experiments can be done on
the topics addressed in this section, several points can already be
drawn. First, Orb2 should be added to a growing list of proteins that
are known to form droplets. This list is already rich in proteins that also
form fibrils or bind mRNA and Orb2 does both. Many proteins on this
list progress from droplets into fibrils and the evidence indicates Orb2
does this as well. In fact, EM and light microscopy images of fibrils
apparently growing out of Orb2BΔRRM droplets are particularly
compelling evidence of the role of droplets in seeding fibril formation.
An interesting feature of the Orb2 droplets discussed in this
section is the low concentration at which they are formed. In all of the
droplet forming experiments in this section, the protein had a final
concentration of ~10 µM. This was often a necessity due to low yields
in the protein expression and purification process, but it presents an
interesting contrast to droplet formation experiments in other studies,
where ~100 µM concentrations are often used (Molliex et al., 2015;
Zhang et al., 2017). The ability of Orb2 to form droplets at such a low
concentration indicates that droplet formation is not only possible, but
also, at least in a salt free condition, energetically highly favorable.
61

Another interesting detail is the remarkable length of the
Orb2AΔRRM and Orb2BΔRRM fibrils that seem to form from droplets.
These fibrils are only ~10 nm wide, indicating that they are likely a
single strand, and show no apparent bundling. Yet, electron and light
microscopy indicates they can be up to 100 µm long. If each fibril is in
fact a single strand (one “stack” of monomers), then this strength
suggests that the binding between monomers in the fibril is very strong.
The biological relevance of the droplets and fibrils discussed in
this section is currently unclear. Our collaborator, Dr. Kausik Si, has
shown that Orb2 forms small concentrated puncta in cells, and that
these puncta are likely critical to memory formation, but the biophysical
structure of the puncta is unknown (Majumdar et al., 2012). It is
possible that the puncta themselves are droplets, fibrils, or droplets
somewhere along the way to becoming fibrils, similar to what we have
seen in our Orb2BΔRRM time series. FRAP experiments by Dr. Si’s
lab have shown that puncta have very slow recovery (indicating slow
dynamics) and this is consistent with the apparent “hardening” of
droplets into fibrils that I have observed (Majumdar et al., 2012).

62

CHAPTER 4. PROJECT III:
Computational Modelling of
Huntingtin C-terminus
4.1 Introduction
The Siemer and Langen labs recently performed a series of
NMR and EPR studies on the C-terminus of Huntingtin exon 1 (Httex1),
with 46 glutamine residues in its poly-glutamine domain, in amyloid
fibrils. NMR was measured on Httex1 with and without a 6 residue C-
terminal His-tag, while all EPR measurements were done on His-
tagged protein. The Langen lab measured a series of DEER EPR
spectra that gave distance distributions for the Q63-Q75, Q75-Q91,
Q91-G102 and Q63-G102 residue pairs in the C-terminus, while the
Siemer lab measured the R1 and R2 relaxation constants of several
residues via NMR. The NMR studies were done at 0°C, while the EPR
studies were done on a sample that was at room temperature and then
flash frozen in liquid nitrogen. I was given both sets of data and asked
to build a computational model of the C-terminus that could reproduce
both datasets. I made several attempts to build this model using
molecular dynamics and molecular modelling, with varying degrees of
success.
4.2 Methods
4.2.1 Modelling Attempts
4.2.1.a Simulation 1
The first attempt to simulate the protein was done with a single
monomer of the C-terminus, with a His-tag, starting in a polyproline II
helix conformation (φ=-64.9, ψ=135). The monomer consisted of
residues 62 (the last of the glutamine stretch) to 118 (the end of the
63

His-tag) with all histidine residues protonated. The N, Cα and CO
atoms of the Q62 residue were fixed in place throughout the simulation,
in order to recreate the effect of the C-terminus being attached to a
static amyloid core. The N atom of Q62 was placed at the (0,0,0)
position in space and the helix was extended in the z-axis. The
monomer was placed in a periodic water filled box that extended from -
10 to 170 in the z-axis and -40 to 40 in the x and y-axis (simulation
space was defined in Ångström). This box was large enough that the
monomer never came into contact with its periodic image throughout
the simulation. The helix was placed along the z-axis and the box
made rectangular to avoid using a larger box that would take more
time to simulate.
The simulation was run at 310 K, with 2 fs timesteps, for 36 ns.
The simulation was run in NAMD (Phillips et al., 2005) using the
CHARMM36 force field (Huang & MacKerell, 2013). The simulation
configuration (Cterm.conf), output, and log files are saved as
production_run, production_run2, and production_run3 (the simulation
was run in multiple parts) in the folder
Sandy_Data\Modelling_Attempt_1 on the Siemer lab external hard
drive. The simulation starting structure files are available in the folder
Sandy_Data\Starting_Model_1 on the same drive.
4.2.1.b Simulation 2
Since the first simulation was run at well over room temperature,
the second simulation was a repeat of first the simulation using a much
more reasonable temperature of 298.15 K. This repeat was run for 20
ns. The simulation configuration (Cterm.conf), output, and log files are
saved as 25Cproduction_run1 in the folder
Sandy_Data\Modelling_Attempt_2 on the same drive.
4.2.1.c Simulation 3
In the third attempt to simulate the protein, the system
temperature was set to 273.15 K and implicit solvent was used, but
parameters were otherwise the same. Implicit solvent was used
64

because simulations can be run much more quickly with it. This
simulation was run for 70 ns and the simulation configuration
(Cterm.conf), output, and log files are saved as ImpSolvRun1 in the
folder Sandy_Data\Modelling_Attempt_3 on the same drive. The
simulation starting structure files are available in the folder
Sandy_Data\Starting_Model_2 (please note that starting model and
simulation attempt numbers are out of sync because some starting
models were used for multiple simulation attempts).
4.2.1.d Simulation 4
In the fourth attempt to simulate the protein, the same set up as
the first simulation was used, but the protein was now simulated
without the His-tag (with the remaining histidine residue still protonated)
at 277.15 K. The simulation was run for 38 ns and the simulation
configuration (Cterm.conf), output, and log files are saved as
production_run and production_run_80 (the simulation was run in two
steps) in the folder Sandy_Data\Modelling_Attempt_4 on the same
drive. The simulation starting structure files are available in the folder
Sandy_Data\Starting_Model_3.
4.2.1.e Simulation 5
In the fifth attempt to simulate the protein, I sought to begin from
a more realistic starting point than just the extended conformation. I did
this by using Rosetta, with the RosettaEPR module, to generate an
ensemble of potential structures for the C-terminus (with the HIS-tag)
compatible with the EPR data provided by the Langen lab (Hirst,
Alexander, Mchaourab, & Meiler, 2011; Simons et al., 1997). One
structure was selected from this ensemble and MTSL labels were
added at the appropriate locations. The structure had unprotonated
histidine residues and a single sodium ion was added to balance the
system charge. This structure was then used as the starting point for a
simulation run in a periodic water box (110 Å in all directions) at 300 K
for 22 ns with a 2 fs timestep. The simulation was run in NAMD
(Phillips et al., 2005) using the CHARMM36 force field (Huang &
65

MacKerell, 2013) as before. In an attempt to force an accurate
structure, constraints matching data from the Langen lab were applied
to the simulation. All constraints were implemented as half-harmonic
potentials with a force constant of 0.1 kcal/mol/Å
2
. The ranges of the
constraints applied were:
Residue
Pair
Constraint Range
(Å)
Q63-Q75 32-42
Q75-Q91 45-55
Q91-G102 29-39
Q63-G102 60-300
Table 1. Constraints Applied in Simulation Attempt 5.
The massive upper bound applied in the Q63-G102 constraint was
implemented because NAMD requires an upper bound to be set, but
the EPR measurements performed by the Langen lab did not give one
for this atom pair. The simulation was run in two parts
(752mtsl_minimize and 752mtsl_step1). Configuration, output, and log
files are saved in the folder Sandy_Data\Modelling_Attempt_5 on the
Siemer lab external hard drive. The simulation starting structure files
are available in the folder Sandy_Data\Starting_Model_4 on the same
drive.
4.2.1.f Simulation 6
Because results from many of the previous attempts showed the
protein collapsing in a manner that was inconsistent with EPR data
(see results). An attempt was made to simulate the protein using a
different water model that had been specially calibrated for intrinsically
disordered proteins. Starting from another model generated by Rosetta
(Simons et al., 1997) (see previous attempt), the protein was simulated
in a periodic water box (140 Å in all directions) with neutral histidine
residues at 278.15 K. The AMBER ff99SB force field (Hornak et al.,
2006) was used along with the TIP4P-D water model (Piana, Donchev,
Robustelli, & Shaw, 2015). The simulation was run in OpenMM
(Eastman & Pande, 2015) for 80 ns with 2 fs timesteps. The python
66

script to start the simulation, output, and state report files are saved in
the folder Sandy_Data\Modelling_Attempt_6 on the Siemer lab
external hard drive. The simulation was run in two steps:
lowtempoutput followed by lowtemplongrun. The simulation starting
structure and the self-made OpenMM implementation of the TIP4P-D
water model used are available in the folder
Sandy_Data\Starting_Model_5 on the same drive.
4.2.1.g Simulation 7
In the final attempt to simulate the protein, OpenMM (Eastman
& Pande, 2015) was used with the same force fields and temperature
as described in the previous attempt. However, this simulation was
begun using the same extended configuration as was implemented in
the first simulation attempt (with protonated histidine residues). The
protein was placed in a periodic water box and constraints were used
to hold the N, Cα, and CO atoms of the Q62 residue in place. The
simulation was run for 21 ns with 2 fs timesteps. The python script to
start the simulation, output, and state report files are saved in the
folder Sandy_Data\Modelling_Attempt_7 on the Siemer lab external
hard drive.

4.2.2 Recovery of Physical Parameters from Models
For some analyses, it was necessary to determine if and when
the simulation reached equilibrium. This was important because the
simulations were begun from relatively arbitrary conformations that had
not been tested or proven. Equilibration in molecular dynamics studies
is commonly assessed by tracking the root-mean-squared deviation
(RMSD) of all atoms from their starting positions. This value typically
increases in a logarithmic manner with simulation time until it reaches
a flat plateau. This corresponds to the simulated structure moving
away from the starting structure until it reaches a point where moving
in this direction is no longer necessarily energetically favorable, which
67

is typically taken as indicating equilibrium (Sharma, Lynn, Sharma,
Rajnee, & Jawaid, 2009; Wells, Müller, Wrenger, & Louw, 2009).
Using the precise model described above in this study
presented some issues. First, effective use of this method requires the
structure for which RMSD is being calculated be superimposed on the
starting structure (in order to eliminate the effect of the entire molecule
tumbling or drifting). This is straightforward for a globular protein but
may be difficult for the long unfolded chain of the Httex1 C-terminus.
Second, with a long unfolded chain, a small move in a single dihedral
could cause a huge increase in the atom position RMSD, as numerous
atoms are “swung” away from their initial positions, which would not
really correlate to a significant change in conformational space.
Based on these problems, I decided to instead asses the
equilibration status of the simulated system by measuring the RMSD of
all atom pair distances (the change in the distance between any
possible pair of atoms) over the course of the simulation. This method
should be insensitive to any tumbling or drifting of the entire molecule
without superimposition and will more directly reflect the
conformational space being explored.
Simulation results were compared to EPR results by taking a
survey of the inter-residue distances (Cβ-Cβ distances if the MTSL
label was not simulated) recorded in EPR experiments over all points
in the simulation trajectory and comparing to the distance distributions
obtained via EPR.
To compare simulation results to NMR data, the simulation
trajectory (dcd) file was first split into a series of single frame pdb files.
Splitting the trajectory into pdb files, measuring EPR distances, and
assessing system equilibration were all done using scripts in the VMD
Tk Console (Humphrey, Dalke, & Schulten, 1996). These scripts are
available on the Siemer lab external drive in the folder
Sandy_Data\Scripts.
68

After the simulation trajectory was split into pdb files, a
correlation function for each residue was established using the formula:
( ) = 1 .5 ∗ ( ( ) ∙ ( + ) )
2
− 0 .5
Equation 1. The Correlation Function.
Where NH(x) and NH(x+t) are the normalized N-H vectors at time x
and time x + t. The overbar is because the value for C(t) is averaged
for all possible values of x with that t value (smaller values of t will
allow more values of x for a given simulation length) (Maragakis et al.,
2008). This was done by applying a C++ script (available with the Tk
Console scripts as “cor_function.cpp”) to all trajectory frame pdb files
past the timepoint where the simulation was deemed to have reached
equilibrium.
After the correlation function for each residue was obtained, it
was used to fit the model free decay function:
( ) = ∗ −
1
+ ( 1 − ) ∗ −
2

Equation 2. The Model Free Decay Function.
Where t is time and a, k1 and k2 are fitted constants. Parameters from
the decay function fit were then used to obtain spectral density function
(the Fourier transform of the decay function):
( ) = (
2
5
) (
∗ 1
1 + ( 1
)
2
+
( 1 − ) ∗ 2
1 + ( 2
)
2
)
Equation 3. The Spectral Density Function.
Where ω is frequency.
Finally, the spectral density function was used with other
parameters to calculate the predicted R1 and R2 values for each
residue using the formulas:
1
= (
4
)
2
( ( − ) + 3 ( ) + 6 ( + ) ) + 2
(
3
4
) ( )
Equation 4. R 1 Calculated from the Spectral Density Function.
69

2
= (
4
)
2
( (
4
3
) ( ) + (
2
3
) ( 2 ) + (
1
2
) ( − ) + 3 ( ) + (
3
2
) ( ) + 3 ( + ) ) + 2
( (
1
4
) ( ) + (
1
8
) ( 2 ) + (
3
8
) ( ) )
Equation 5. R 2 Calculated from the Spectral Density Function.
Where D is the dipolar coupling between
15
N and
1
H, I is the is the
1
H
Larmor frequency, S is the
15
N Larmor frequency, C is the chemical
shift anistropy of
15
N, and r is the MAS frequency (Schanda & Ernst,
2016). The results of this prediction were compared to the actual R1
and R2 values provided by the Siemer lab. All of these calculations
were done using a Mathematica script which is available with the other
scripts as both a text file and a Mathematica notebook “R1andR2”.

4.3 Results
4.3.1 Assessment of Simulation Accuracy
I reviewed the results of all simulations to see if any could
accurately reproduce the NMR and EPR results obtained by the
Siemer and Langen labs. I first inspected the ability of the simulations
to reproduce EPR results because this is a straightforward test that
gives a definitive answer to the question: does the simulation have any
potential to be realistic? Here, I first discuss the results of the third and
fifth simulations, which were dismissed with only visual inspections,
before going into the details of the other simulations.
70

Figure 18. Simulating the Huntingtin C-terminus with Implicit Solvent Gives a Collapsed
Structure.
The Httex1 C-terminus was simulated in NAMD (Phillips et al., 2005) using the
CHARMM36 force field (Huang & MacKerell, 2013) with an implicit water model.
Despite beginning in an extended conformation, the structure completely collapsed
soon after the simulation start.
A visual inspection of the third simulation (where implicit solvent
was used) revealed that it quickly collapsed into a globular structure
(see Figure 18) that was inconsistent with DEER distances, and
analysis of this structure was not pursued further.

Figure 19. Constrained Simulation of the Httex1 C-terminus Gives an Unrealistic
Structure.
The Huntingtin C-terminus (with MTSL labels) was simulated in NAMD (Phillips et al.,
2005) using the CHARMM36 force field (Huang & MacKerell, 2013) with constraints
to hold the MTSL labels to distance ranges known from EPR experiments
(constrained distances shown). The model fails to meet the distance constraint
71

applied to the Q63-Q75 residue pair (30.93 Å instead of 32-42 Å) and barely meets
the constraints applied to the Q75-Q91 and Q63-G102 residue pairs (46.28 Å in 45-
55 Å and 60.24 Å in 60-300 Å). For all of these pairs, the MTSL labels appear to be
straining away from each other, indicating the structure would likely collapse if the
constraints were removed.
Visual inspection of the fifth simulation, where artificial
constraints were applied to the MTSL labels, showed a semi-collapsed
conformation where the labels often failed to meet the set constraints.
Further, in some cases where the constraints were met, this seemed to
only be because the labels had been pulled away from each other and
made to point in opposite directions (see Figure 19). Because of this,
this model was considered to be unrealistic and was not pursued
further.

Figure 20. EPR Based Residue Pair Distance Distributions Provided by the Langen
Lab.
72

0
10
20
30
40
50
60
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103109115121127
% Frames
Distance (Å)
Simulation 1
63-75 75-91 91-102 63-102
0
10
20
30
40
50
60
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103109115121127
% Frames
Distance (Å)
Simulation 2
63-75 75-91 91-102 63-102
0
10
20
30
40
50
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103109115121127
% Frames
Distance (Å)
Simulation 4
63-75 75-91 91-102 63-102
73

Figure 21. Spin Label Distance Distributions for Simulations 1,2,4,6 and 7.
The C-terminus of Httex1 was simulated using various methods and results were
compared to EPR based inter-residue distance distributions (see Figure 20). A
histogram showing the distribution of distances over all frames of the simulation is
plotted for each simulation.
Analysis of all EPR residue pair distances for the remaining
simulations (see Figure 20 and Figure 21) revealed that simulations 4
and 6 seemed give inaccurate results, with the Q75-Q91 residue pair
having a shorter distance throughout the simulations than EPR data
indicated. Further, the Q91-G102 and Q63-G102 pairs seemed to
adopt distances that were too short in simulations 4 and 6 respectively.
Based on these results, simulations 4 and 6 were deemed inaccurate
and not analyzed further.
0
10
20
30
40
50
60
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103109115121127
% Frames
Distance (Å)
Simulation 6
63-75 75-91 91-102 63-102
0
10
20
30
40
50
60
70
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103109115121127
% Frames
Distance (Å)
Simulation 7
63-75 75-91 91-102 63-102
74

Figure 22. Simulation 2 RMSD and Inter-Residue Distance Progression with Time.
The Huntingtin C-terminus was simulated in NAMD (Phillips et al., 2005) using the
CHARMM36 force field (Huang & MacKerell, 2013). The simulation was begun in an
extended conformation, but at the ~12 ns timepoint the Q75-Q91 residue pair
adopted a distance below what was measured by DEER and stayed at this distance
for the duration of the simulation. This correspond to the point where the Cα-Cα
distance RMSD seemed to reach a plateau, suggesting the simulation had
equilibrated to an overly collapsed conformation.
The distance distributions for simulation 2 appeared to match
the provided EPR data. However, inspection of these distances over
simulation time revealed that the Q75-Q91 distance became too short
during the second half of the simulation. Further, this seemed to
correspond to the point where the simulation RMSD maxed out (Figure
20
40
60
80
100
120
0 5 10 15 20
Distance (Å)
Simulation Time (ns)
Simulation 2 Inter-Residue Distance Vs. Time
63-75 75-91 91-102 63-102
0
2
4
6
8
10
12
14
0 5 10 15 20
RMSD (Å)
Simulation Time (ns)
Simulation 2 RMSD Vs. Time
75

22), indicating the simulation had equilibrated at this conformation.
Because of this, simulation 2 was not analyzed further.

Figure 23. Cα-Cα Distance RMSD Time Progressions for Simulations 1 and 7.
Based on these results it was determined that both simulations had reached
equilibrium at 8 ns.
-4
1
6
11
16
21
0 5 10 15 20 25 30 35 40
RMSD (Å)
Simulation Time (ns)
Simulation 1 RMSD Vs. Time
0
2
4
6
8
10
12
0 5 10 15 20 25
RMSD (Å)
Simulation Time (ns)
Simulation 7 RMSD Vs. Time
76

Figure 24. Actual and Simulated R 1 and R 2 Values.
The Huntingtin C-terminus was simulated using two different methods and R1 and R2
values were calculated from the simulation and compared to actual values obtained
via NMR. Simulation 7, which used an extended starting conformation, lower
temperature and a water model specially made for disordered proteins was more
accurate than simulation 1, but both tended to show R1 and R2 values that were
considerably below the actual values.
Since simulations 1 and 7 both matched the EPR data very well
(Figure 21), I tested if they could also reproduce the NMR data
obtained by the Siemer lab. I first studied the time progression of their
RMSDs and found that both simulations appeared to have reached a
plateau in this graph (and thus equilibrium) by 8 ns (Figure 23). Only
0
1
2
3
4
5
6
100 105 110 115 120
R
1
(1/s)
Residue
Actual and Simulated R
1
Values
R1 Simulation 1 R1 Simulation 7 R1
0
5
10
15
20
100 105 110 115 120
R
2
(1/s)
Residue
Actual and Simulated R
2
Values
R2 Simulation 1 R2 Simulation 7 R2
77

frames from past this timepoint were used in the calculation of their
correlation functions (Equation 1). Further, when fitting the calculated
correlation functions to the model free decay (Equation 2), only
timesteps below 10 ns and 20 ns were used for simulations 1 and 7
respectively. This was because a larger timestep will have fewer
correlation values for a given simulation time and thus the average will
be more noisy, making the fitting of Equation 2 potentially less
accurate. Finally, the fitted model free decays were used to obtain R1
and R2 values as described above (Figure 24). Values predicted from
simulation 1 highly inaccurate and considerably lower than the actual
values (indicating faster relaxation). Values predicted from simulation 7
were more accurate and somewhat reproduced the patterns observed
in the actual measurements but were still noticeably lower than they
should have been.

Figure 25. Simulations of the Huntingtin C-terminus Suggest a “Nunchuk” Model.
Frames from simulations 1 (a) and 7 (b) are shown with proline residues colored
green and all other residues colored orange. Proline residues appeared to maintain
consistent poly-proline II helices while the center region flexed about more freely.
Simulation 7 tended to take on a more extended conformation, but both simulations
accurately reproduced EPR data.
Visual inspection of simulations 1 and 7 showed that the two
polyproline sequences were maintained in consistent poly-proline II
78

helices, while the center region flexed about (which we describe as the
“nunchuk” model) (Figure 25). This is consistent with the distance
distributions observed in both EPR experiments and the simulations,
where the Q63-Q75 and Q75-Q91 residue pairs (on opposite sides of
the polyproline sequences) show a narrow and fairly even distance
distribution, but the Q75-Q91 residue pair has a wider distance
distribution that is skewed toward smaller sizes (as the center region
folds onto itself). Although the simulations showed similar behavior in
general, simulation 7 tended to adopt a more extended conformation
than simulation 1.
4.3.2 Simulation 7 Analysis
Since simulation 7 proved to be the most accurate of all
simulations, with excellent recovery of EPR data and decent recovery
of NMR data. I decided to perform some additional analyses on the
simulation output to gain information about the “nunchuk” model it
presented.

-150
-100
-50
0
50
100
150
200
250
Q
63
PPPPP
68
PPPPP
73
PQL PQ
78
PPPQA
83
QP L L P
88
QPQPP
93
PPPPP
98
PPPGQ
103
AVAE E
108
P LHRP
113
HHHHH
118
ψ Angle (degrees)
Residue
ψ Angles
79

Figure 26. Dihedral Angles from Simulation 7.
The average ψ and φ angles (blue bars) for each residue in simulation 7 was
calculated for all frames after the 8 ns timepoint. The standard deviation for each
angle is plotted in the error bars.
I first studied the distribution of the ψ and φ angles for all
residues in all frames after the 8 ns equilibration time point (see Figure
26). I found that almost all proline residues had ψ and φ averages
consistent with a polyproline II helix (φ = -50 to -100° and ψ = 120 to
170°). The sole exception to this was P74, which adopted a
configuration most consistent with a 3-10 helix (φ = -70° and ψ = -17°).
Proline residues within the two polyproline stretches (excluding P74)
also tended to show ψ and φ standard deviations that stayed within the
bounds of the polyproline II helix. The only exceptions to this were P65
and P71, which displayed large ψ standard deviations (43 and 40°
respectively) that pushed slightly beyond the edges of the polyproline II
helix but were still centered on the structure (Adzhubei, Sternberg, &
Makarov, 2013). Proline residues outside of these stretches tended
show larger ψ standard deviations (~40-60°).
Almost all non-proline residues showed an average compatible
with a β-strand conformation (φ = -100 to -125° and ψ = 125 to 160°)
with large standard deviations (~30° for φ and ψ). L76, L86 and L110
seemed to prefer a polyproline II helix, likely due to neighboring proline
-175
-125
-75
-25
25
75
PPPPP
68
PPPPP
73
PQL PQ
78
PPPQA
83
QP L L P
88
QPQPP
93
PPPPP
98
PPPGQ
103
AVAE E
108
P LHRP
113
HHHHH
118
H
φ Angle (degrees)
Residue
φ Angles
80

residues. A106 (φ = -84°, ψ = -31°) and E107 (φ = 52°, ψ = 41°)
adopted the dihedrals of a right and left-handed α-helix respectively.
The average dihedrals of G102, the only glycine in the sequence,
roughly correspond to a β-strand but the residue shows a considerable,
160°, standard deviation in its ψ angle (Adzhubei et al., 2013).

Figure 27. Conservation of the G102-P113 Sequence in Httex1.
The G102-P113 sequence in Httex1 is extremely conserved across a range of
species. This is illustrated here by a comparison of the sequence in Homo sapiens
(humans) and its counterpart in Alligator sinensis (Chinese alligator).
Based on the large ψ flexibility displayed by G102, I
hypothesized that the residues following it may serve as a protein
docking site. The flexibility in G102 could help this site move freely and
thus be available for protein binding. To investigate this possibility, I
performed a BLAST search on the NCBI website and found that the
G102 to P113 sequence is almost perfectly conserved throughout the
mammal family. In fact, reptilian (alligator) huntingtin contains the
same sequence with only two mutations (A106P and P103_A104insA)
(see Figure 27).
81

Figure 28. Time Progression of C-terminus Length and Angles in Simulation 7.
The length of the entire C-terminus, the angle formed by the residues between the
two polyproline stretches (inner angle), and the angle formed by the two polyproline
stretches (outer angle) were tracked with respect to simulation time. (A) shows the
progression of these measurements. (B) illustrates the parameters being measured.
Following the analysis of dihedral angles, I assessed the
general shape of the C-terminus with respect to simulation time. I
monitored the length of the chain (formally defined as the distance
between the Cα atoms of Q63 and P113), the angle of flexible region
between the two polyproline stretches (formally defined as the angle
formed by the Cα atoms of Q75, A83 and Q91), and the angle formed
by the two polyproline stretches (formally defined as the angle formed
by the Cα atoms of Q63, A83 and G102) (see Figure 28).
110
115
120
125
130
135
140
145
150
155
160
110
120
130
140
150
160
170
180
0 5 10 15 20
Chain Length (Å)
Angles (Degrees)
Time (ns)
Httex1 C-terminus Length and Angles Vs. Time
Inner Angle Outer Angle Chain Length
(A)
(B)
82

Since analysis of dihedral angles and a visual inspection of the
simulation results indicated that the polyproline stretches stayed
almost entirely in polyproline II helices while a series of non-proline
residues flexed between them (like a nunchuk chain), I was curious
what angle this flexing region typically adopted and if that angle would
also be held by the polyproline helices. I also sought to determine how
stretched out the entire chain (excluding the non-biological his-tag)
was.
Analysis of the data (see Figure 28) showed that the two angles
stayed in the same general region but did not seem to be closely
correlated. Both angles tended to be highly obtuse, typically at least
140° and never going below 110°. Interestingly, the “inner” angle of just
flexible region between the two polyproline helices seemed to change
more and was sometimes smaller than the “outer” angle formed by the
helices. This likely corresponds to the flexing of the inner region while
the polyproline stretches hold relatively still in space. The length of the
chain stayed above 130 Å for the entire simulation, combined with the
highly obtuse outer angle, this indicates that the C-terminus mainly
adapts a stretched-out conformation.

Figure 29. Mass Density Map of Simulation 7.
The map was created using the VMD volmap feature (Humphrey et al., 1996) with 1
Å voxels and a 0.005 density rendering cutoff. Coloring is based on occupancy with
regions in purple being most occupied (at least 0.01 density) and regions in blue
being the least. The background grid is 10 Å.
Finally, I assessed the spatial exploration of the C-terminus by
mapping the distribution of its mass over all frames in the simulation
trajectory following the 8 ns equilibration timepoint (see Figure 29).
83

This map showed that the C-terminus remained outstretched while at
the same time sampling a large number of conformations, with very
little preference for any one in particular.
Because the structure of the Httex1 fibril core is currently
unknown, how C-termini of different monomers in the amyloid fibril are
placed relative to each other is unclear, but the extremely broad
sampling shown in this density map (+/- 20 Å in the y axis) means that
it is very likely possible for neighboring C-termini to occasionally come
into contact with other. However, the broad conformational sampling
observed, as well as the accuracy of simulation 7 at recovering
physical parameters when only using a single monomer, means these
contacts are probably transient and without a major impact.
The dihedral angle data and measurements of the C-terminus
size and internal angles used in these analyses were gathered using
scripts in the VMD Tk console (Humphrey et al., 1996). These scripts
are available on the Siemer lab external drive in the folder
Sandy_Data\Scripts. The density map was created in VMD using the
volmap feature.
4.4 Discussion
We have developed models that reproduce EPR data for the C-
terminus of Httex1, obtained by the Langen lab, very well and offer a
good understanding (the nunchuk model) of the structure and motions
of the C-terminus that underlie this data. Unfortunately, our
reproduction of NMR data obtained by the Siemer lab has not been as
successful.
The most accurate NMR relaxation value predictions were
obtained using simulation 7. This simulation was run at low
temperature (5°C), started in an extended conformation, and used a
water model specially calibrated for disordered proteins. It may be
possible to improve this model by running the same simulation at an
84

even lower temperature. It should also be noted that the simulation
was run for a relatively short time (~20 ns) and a longer runtime (and
thus more complete conformational sampling) may also improve the
accuracy of calculated relaxation values. It is also possible that the
NMR experiments and calculations used to obtain relaxation values
may have systemically overestimated them.
Despite the imperfect recovery of NMR relaxation values from
simulation 7, several interesting details with potential biological
relevance can be gained from its results.
First, the C-terminus seems to maintain a stretched
conformation that samples a large amount of space. It is possible that
the C-terminus serves as a sort of barrier, keeping proteins from
directly interacting with the polyglutamine domain, since doing so
would likely block this movement of the C-terminus and thus incur an
entropic penalty. This may be beneficial because the polyglutamine
stretch is demonstrated to be prone to aggregation.
Second, the proline residues form polyproline II helices and
seem to impose this structure on some other residues as well. Thus,
the excessive proline content may prevent the formation of additional
β-strand content that could extend the fibril core and make fibril
formation too energetically favorable.
Finally, the flexibility of G102 (see Figure 26) is very interesting,
as it effectively terminates the order imposed by the poly-proline
stretches. This is even more interesting when considering the unusual
dihedrals of the A106 and E107 residues that follow shortly after it (the
only α-helix compatible dihedrals in the entire C-terminus) and the
incredible sequence conservation of the G102 to P113 sequence. I
believe that all of this supports the hypothesis that the G102 to P113
domain presents a docking site for a critical protein that interacts with
huntingtin.
85

Through this project, a major issue I encountered was the
tendency of the simulated protein to collapse into a conformation that
was too short to match the EPR measured distances. The tendency of
intrinsically disordered proteins to unrealistically collapse in molecular
dynamics simulations is well known and was the reason behind the
development of the TIP4P-D water model that was used in simulation
attempts 6 and 7 (Chong, Chatterjee, & Ham, 2017; Henriques,
Cragnell, & Skepö, 2015; Piana et al., 2015; Stanley, Esteban-Martín,
& De Fabritiis, 2015). However, the failures of simulation attempts 5
and 6 indicates that starting conformation may also play an important
role in the simulation accuracy. It is possible that the starting structure
provided by Rosetta (Simons et al., 1997) for these simulations was
already on the verge of collapse.
An obvious way to try to compensate for force field bias in favor
of collapsed protein is to increase the simulation temperature. However,
this can lead to increased dynamics and thus undermine the accuracy
predicted NMR values (as may have been the case with simulation 1).
It is likely that a completely accurate model of the Huntingtin C-
terminus will require a careful balance between system temperature,
force field, and starting conformation.

86

CHAPTER 5. CONCLUSIONS
When reviewing all of the experiments discussed here, one
striking detail that emerges is the incredible versatility of Orb2. Orb2
seems to adapt at least two different monomeric conformations in
denaturing conditions and another, different, monomeric conformation
in non-denaturing conditions. It seems to capable of forming oligomers,
droplets, and two different types of fibrils, distinct from each other in
both appearance and NMR spectra. The presence of two fibril types is
even more interesting when considering fibrils grown from the first 88
residues of Orb2A (Orb2A88) by the Siemer lab (Cervantes et al.,
2016). Neither of the fibril types described in this study resemble
(visually or in NMR spectra) these fibrils. Thus, it is very likely that the
Orb2 protein contains no fewer than 3 separate domains capable of
independently forming fibrils in different circumstances.
The versatility of Orb2 in adopting different conformations
seems to be further complemented by the versatility across domains of
Orb2 within a given conformation. NMR experiments show a range of
dynamics in both fibril types, and it is quite possible that this dynamic
diversity extends to other conformations as well.
Our collaborator has already shown that Orb2 can change from
a repressor of target mRNAs to an activator (Khan et al., 2015). It is
hypothesized that in this activated role, Orb2 serves to mark newly
formed synapses and thus prevent their degradation (Hervás et al.,
2016; Khan et al., 2015; Li et al., 2016). For Orb2 to succeed in this
role, it would need to be both immobile and persistent. Fibrils, droplets
or some combination thereof would be well suited for this task.
Formation of this activated state would also need to be tightly
87

regulated in time and place and thus the structure of the monomer and
how it progresses into the active state is also important.
This study has identified a variety of states of Orb2 and studied
the progression between some. It is hoped that this knowledge of
Orb2’s states and behaviors in-vitro can lay the groundwork for
understanding the critical transitions that it undergoes in-vivo and their
regulation.

88

CHAPTER 6. FUTURE DIRECTIONS
There are a number of follow up studies that can be conducted
based on the research presented here. I believe that one of the most
important areas to focus on is droplets and the fibrils that seem to be
derived from them. Currently, a relatively complete picture of the
droplet to fibril transition only exists for Orb2BΔRRM. The formation of
droplets and related fibrils (if any) should be completely mapped out for
Orb2AΔRRM and the full length Orb2A and Orb2B proteins. Mapping
out this process should incorporate FRAP measurements to monitor
how much of a liquid-like state the droplets retain. Also, this study has
currently only formed droplets in salt free buffer. Other studies have
shown the ability to form protein droplets in the presence of salt by
compensating with increased protein concentration or RNA (a potent
droplet inducer). To see how these factors influence Orb2 droplet and
fibril formation would be very interesting.
The structure of the Orb2BΔRRM fibrils that originate from
droplets, as well as their Orb2AΔRRM counterparts, is currently
unknown. Although NMR studies on the Orb2BΔRRM fibrils present
some useful insights, matching studies on Orb2AΔRRM fibrils to study
how similar they are would be very interesting. Further, it may be
possible to improve spectrum resolution, and thus gain more structural
information, by performing repeated fibril seeding and growth cycles to
produce more homogenous fibrils. EM images suggest that
Orb2BΔRRM fibrils are remarkably strong and better structural data
from NMR might help uncover the root of this strength.
Any studies on droplets and fibrils should relate back to the
biological relevance (if any) of these structures. Thankfully our
collaborator, Dr. Kausik Si, already has an assay that can distinguish
89

the active and inactive states of Orb2. Samples of droplets and fibrils
made should be sent to him assessment in this assay.
For all of the potential experiments I have just described, as well
the experiments presented in this thesis, a major hindrance is the
expression and purification of clean and monomeric Orb2. As
described above, the purification process is frequently challenged by
low yields and what appears to be cleavage products that must be
carefully separated from the desired protein. I believe that time spent in
developing an expression and purification procedure that prevents the
formation of this cleavage product and gives better yields (this will
likely require a change of expression systems) would be extremely well
invested and will greatly boost the speed and reliability of all
experiments just described.
In addition to experiments on Orb2, a number of experiments,
computational and in lab, can be conducted based on the simulations
of the Httex1 C-terminus described above. First, although simulation 7
proved accurate and insightful, I believe it was far too short to have
reliably sampled all possibly conformations of the Httex1 and should be
rerun with a goal of simulating ~200 ns rather than just 21 as was done
here. Future simulations should also consider incorporating multiple
Httex1 C-terminus monomers to study the interaction between them
that likely occurs in fibril, although this would require the Httex1 fibril
core structure to be solved so that the monomers can be correctly
positioned relative to each other.
Finally, the simulations shown here, and an analysis of
sequence conservation, indicates that the G102 to P113 region is
extremely critical for Huntingtin functionality. It would be interesting to
express disease causing (extended polyglutamine domain) and normal
Httex1 with and without this region in mouse and cellular models to see
what sort of an impact it has. It would be also be interesting to conduct
90

a pull-down assay with a peptide matching this region and mammalian
cell extract to see what proteins it binds.
91

BIBLIOGRAPHY
Adzhubei, A. A., Sternberg, M. J. E., & Makarov, A. A. (2013).
Polyproline-II Helix in Proteins: Structure and Function. Journal of
Molecular Biology, 425(12), 2100–2132.
https://doi.org/10.1016/j.jmb.2013.03.018
Afroz, T., Skrisovska, L., Belloc, E., Guillén-Boixet, J., Méndez, R., &
Allain, F. H.-T. (2014). A fly trap mechanism provides sequence-
specific RNA recognition by CPEB proteins. Genes &
Development, 28(13), 1498–1514.
https://doi.org/10.1101/gad.241133.114
Ambadipudi, S., Biernat, J., Riedel, D., Mandelkow, E., & Zweckstetter,
M. (2017). Liquid-liquid phase separation of the microtubule-
binding repeats of the Alzheimer-related protein Tau. Nature
Communications, 8(1), 275. https://doi.org/10.1038/s41467-017-
00480-0
Barth, H. G., Boyes, B. E., & Jackson, C. (1994). Size exclusion
chromatography. Analytical Chemistry, 66(12), 595R–620R.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8092479
Bergman, P., Roan, N. R., Römling, U., Bevins, C. L., & Münch, J.
(2016). Amyloid formation: functional friend or fearful foe? Journal
of Internal Medicine, 280(2), 139–152.
https://doi.org/10.1111/joim.12479
Brooks, B. R., Brooks, C. L., Mackerell, A. D., Nilsson, L., Petrella, R.
J., Roux, B., … Karplus, M. (2009). CHARMM: the biomolecular
simulation program. Journal of Computational Chemistry, 30(10),
1545–1614. https://doi.org/10.1002/jcc.21287
Cervantes, S. A., Bajakian, T. H., Soria, M. A., Falk, A. S., Service, R.
J., Langen, R., & Siemer, A. B. (2016). Identification and Structural
Characterization of the N-terminal Amyloid Core of Orb2 isoform A.
Scientific Reports, 6(1), 38265. https://doi.org/10.1038/srep38265
Chiti, F., & Dobson, C. M. (2017). Protein Misfolding, Amyloid
Formation, and Human Disease: A Summary of Progress Over the
Last Decade. Annual Review of Biochemistry, 86(1), 27–68.
https://doi.org/10.1146/annurev-biochem-061516-045115
Chong, S.-H., Chatterjee, P., & Ham, S. (2017). Computer Simulations
of Intrinsically Disordered Proteins. Annual Review of Physical
Chemistry, 68(1), 117–134. https://doi.org/10.1146/annurev-
physchem-052516-050843
Chuang, E., Hori, A. M., Hesketh, C. D., & Shorter, J. (2018). Amyloid
assembly and disassembly. Journal of Cell Science, 131(8),
jcs189928. https://doi.org/10.1242/jcs.189928
Eastman, P., Friedrichs, M. S., Chodera, J. D., Radmer, R. J., Bruns, C.
92

M., Ku, J. P., … Pande, V. S. (2013). OpenMM 4: A Reusable,
Extensible, Hardware Independent Library for High Performance
Molecular Simulation. Journal of Chemical Theory and
Computation, 9(1), 461–469. https://doi.org/10.1021/ct300857j
Eastman, P., & Pande, V. S. (2015). OpenMM: A Hardware
Independent Framework for Molecular Simulations. Computing in
Science & Engineering, 12(4), 34–39.
https://doi.org/10.1109/MCSE.2010.27
Elbaum-Garfinkle, S., Kim, Y., Szczepaniak, K., Chen, C. C.-H.,
Eckmann, C. R., Myong, S., & Brangwynne, C. P. (2015). The
disordered P granule protein LAF-1 drives phase separation into
droplets with tunable viscosity and dynamics. Proceedings of the
National Academy of Sciences of the United States of America,
112(23), 7189–7194. https://doi.org/10.1073/pnas.1504822112
Falk, A. S., & Siemer, A. B. (2016). Dynamic domains of amyloid fibrils
can be site-specifically assigned with proton detected 3D NMR
spectroscopy. Journal of Biomolecular NMR, 66(3), 159–162.
https://doi.org/10.1007/s10858-016-0069-2
Fowler, D. M., Koulov, A. V., Balch, W. E., & Kelly, J. W. (2007).
Functional amyloid – from bacteria to humans. Trends in
Biochemical Sciences, 32(5), 217–224.
https://doi.org/10.1016/j.tibs.2007.03.003
Gade Malmos, K., Blancas-Mejia, L. M., Weber, B., Buchner, J.,
Ramirez-Alvarado, M., Naiki, H., & Otzen, D. (2017). ThT 101: a
primer on the use of thioflavin T to investigate amyloid formation.
Amyloid, 24(1), 1–16.
https://doi.org/10.1080/13506129.2017.1304905
Henriques, J., Cragnell, C., & Skepö, M. (2015). Molecular Dynamics
Simulations of Intrinsically Disordered Proteins: Force Field
Evaluation and Comparison with Experiment. Journal of Chemical
Theory and Computation, 11(7), 3420–3431.
https://doi.org/10.1021/ct501178z
Hervás, R., Li, L., Majumdar, A., Fernández-Ramírez, M. del C., Unruh,
J. R., Slaughter, B. D., … Carrión-Vázquez, M. (2016). Molecular
Basis of Orb2 Amyloidogenesis and Blockade of Memory
Consolidation. PLOS Biology, 14(1), e1002361.
https://doi.org/10.1371/journal.pbio.1002361
Hirst, S. J., Alexander, N., Mchaourab, H. S., & Meiler, J. (2011).
RosettaEPR: An integrated tool for protein structure determination
from sparse EPR data. Journal of Structural Biology, 173(3), 506–
514. https://doi.org/10.1016/j.jsb.2010.10.013
Hornak, V., Abel, R., Okur, A., Strockbine, B., Roitberg, A., &
Simmerling, C. (2006). Comparison of multiple Amber force fields
and development of improved protein backbone parameters.
Proteins, 65(3), 712–725. https://doi.org/10.1002/prot.21123
Huang, J., & MacKerell, A. D. (2013). CHARMM36 all-atom additive
93

protein force field: Validation based on comparison to NMR data.
Journal of Computational Chemistry, 34(25), 2135–2145.
https://doi.org/10.1002/jcc.23354
Humphrey, W., Dalke, A., & Schulten, K. (1996). VMD: visual
molecular dynamics. Journal of Molecular Graphics, 14(1), 33–38,
27–28. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/8744570
Isas, J. M., Langen, A., Isas, M. C., Pandey, N. K., & Siemer, A. B.
(2017). Formation and Structure of Wild Type Huntingtin Exon-1
Fibrils. Biochemistry, 56(28), 3579–3586.
https://doi.org/10.1021/acs.biochem.7b00138
Isas, J. M., Langen, R., & Siemer, A. B. (2015). Solid-State Nuclear
Magnetic Resonance on the Static and Dynamic Domains of
Huntingtin Exon-1 Fibrils. Biochemistry, 54(25), 3942–3949.
https://doi.org/10.1021/acs.biochem.5b00281
Ivshina, M., Lasko, P., & Richter, J. D. (2014). Cytoplasmic
Polyadenylation Element Binding Proteins in Development, Health,
and Disease. Annual Review of Cell and Developmental Biology,
30(1), 393–415. https://doi.org/10.1146/annurev-cellbio-101011-
155831
Kato, M., Han, T. W., Xie, S., Shi, K., Du, X., Wu, L. C., … McKnight, S.
L. (2012). Cell-free Formation of RNA Granules: Low Complexity
Sequence Domains Form Dynamic Fibers within Hydrogels. Cell,
149(4), 753–767. https://doi.org/10.1016/j.cell.2012.04.017
Keleman, K., Krüttner, S., Alenius, M., & Dickson, B. J. (2007).
Function of the Drosophila CPEB protein Orb2 in long-term
courtship memory. Nature Neuroscience, 10(12), 1587–1593.
https://doi.org/10.1038/nn1996
Khan, M. R., Li, L., Pérez-Sánchez, C., Saraf, A., Florens, L.,
Slaughter, B. D., … Si, K. (2015). Amyloidogenic Oligomerization
Transforms Drosophila Orb2 from a Translation Repressor to an
Activator. Cell, 163(6), 1468–1483.
https://doi.org/10.1016/j.cell.2015.11.020
Klare, J. P. (2013). Site-directed spin labeling EPR spectroscopy in
protein research. Biological Chemistry, 394(10), 1281–1300.
https://doi.org/10.1515/hsz-2013-0155
Krüttner, S., Stepien, B., Noordermeer, J. N., Mommaas, M. A.,
Mechtler, K., Dickson, B. J., & Keleman, K. (2012). Drosophila
CPEB Orb2A Mediates Memory Independent of Its RNA-Binding
Domain. Neuron, 76(2), 383–395.
https://doi.org/10.1016/j.neuron.2012.08.028
Kumar, E. K., Haque, N., & Prabhu, N. P. (2017). Kinetics of protein
fibril formation: Methods and mechanisms. International Journal of
Biological Macromolecules, 100, 3–10.
https://doi.org/10.1016/j.ijbiomac.2016.06.052
Levitt, M. H. (n.d.). Spin dynamics : basics of nuclear magnetic
94

resonance.
Li, L., Sanchez, C. P., Slaughter, B. D., Zhao, Y., Khan, M. R., Unruh,
J. R., … Si, K. (2016). A Putative Biochemical Engram of Long-
Term Memory. Current Biology, 26(23), 3143–3156.
https://doi.org/10.1016/j.cub.2016.09.054
Majumdar, A., Cesario, W. C., White-Grindley, E., Jiang, H., Ren, F.,
Khan, M. “Repon,” … Si, K. (2012). Critical Role of Amyloid-like
Oligomers of Drosophila Orb2 in the Persistence of Memory. Cell,
148(3), 515–529. https://doi.org/10.1016/j.cell.2012.01.004
Mangiarini, L., Sathasivam, K., Seller, M., Cozens, B., Harper, A.,
Hetherington, C., … Bates, G. P. (1996). Exon 1 of the HD gene
with an expanded CAG repeat is sufficient to cause a progressive
neurological phenotype in transgenic mice. Cell, 87(3), 493–506.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8898202
Maragakis, P., Lindorff-Larsen, K., Eastwood, M. P., Dror, R. O.,
Klepeis, J. L., Arkin, I. T., … Shaw, D. E. (2008). Microsecond
molecular dynamics simulation shows effect of slow loop
dynamics on backbone amide order parameters of proteins. The
Journal of Physical Chemistry. B, 112(19), 6155–6158.
https://doi.org/10.1021/jp077018h
Marley, J., Lu, M., & Bracken, C. (2001). A method for efficient isotopic
labeling of recombinant proteins. Journal of Biomolecular NMR,
20(1), 71–75. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/11430757
Merkel, D. J., Wells, S. B., Hilburn, B. C., Elazzouzi, F., Pérez-
Alvarado, G. C., & Lee, B. M. (2013). The C-Terminal Region of
Cytoplasmic Polyadenylation Element Binding Protein Is a ZZ
Domain with Potential for Protein–Protein Interactions. Journal of
Molecular Biology, 425(11), 2015–2026.
https://doi.org/10.1016/j.jmb.2013.03.009
Miniaci, M. C., Kim, J.-H., Puthanveettil, S. V., Si, K., Zhu, H., Kandel,
E. R., & Bailey, C. H. (2008). Sustained CPEB-Dependent Local
Protein Synthesis Is Required to Stabilize Synaptic Growth for
Persistence of Long-Term Facilitation in Aplysia. Neuron, 59(6),
1024–1036. https://doi.org/10.1016/j.neuron.2008.07.036
Molliex, A., Temirov, J., Lee, J., Coughlin, M., Kanagaraj, A. P., Kim, H.
J., … Taylor, J. P. (2015). Phase separation by low complexity
domains promotes stress granule assembly and drives
pathological fibrillization. Cell, 163(1), 123–133.
https://doi.org/10.1016/j.cell.2015.09.015
Nekooki-Machida, Y., Kurosawa, M., Nukina, N., Ito, K., Oda, T., &
Tanaka, M. (2009). Distinct conformations of in vitro and in vivo
amyloids of huntingtin-exon1 show different cytotoxicity.
Proceedings of the National Academy of Sciences, 106(24),
9679–9684. https://doi.org/10.1073/pnas.0812083106
Oliveira, C., & Domingues, L. (2018). Guidelines to reach high-quality
95

purified recombinant proteins. Applied Microbiology and
Biotechnology, 102(1), 81–92. https://doi.org/10.1007/s00253-017-
8623-8
Pang Benny Yiu, C., & Wai Chen, Y. (2017). From Disorder to Mis-
Order: Structural Aspects of Pathogenic Oligomerization in
Conformational Diseases. Protein & Peptide Letters, 24(4), 307–
314. https://doi.org/10.2174/0929866524666170220111930
Phillips, J. C., Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E., Villa,
E., … Schulten, K. (2005). Scalable molecular dynamics with
NAMD. Journal of Computational Chemistry, 26(16), 1781–1802.
https://doi.org/10.1002/jcc.20289
Piana, S., Donchev, A. G., Robustelli, P., & Shaw, D. E. (2015). Water
Dispersion Interactions Strongly Influence Simulated Structural
Properties of Disordered Protein States. The Journal of Physical
Chemistry B, 119(16), 5113–5123.
https://doi.org/10.1021/jp508971m
Raveendra, B. L., Siemer, A. B., Puthanveettil, S. V, Hendrickson, W.
A., Kandel, E. R., & McDermott, A. E. (2013). Characterization of
prion-like conformational changes of the neuronal isoform of
Aplysia CPEB. Nature Structural & Molecular Biology, 20(4), 495–
501. https://doi.org/10.1038/nsmb.2503
Sahu, I. D., McCarrick, R. M., & Lorigan, G. A. (2013). Use of Electron
Paramagnetic Resonance To Solve Biochemical Problems.
Biochemistry, 52(35), 5967–5984.
https://doi.org/10.1021/bi400834a
Schanda, P., & Ernst, M. (2016). Studying Dynamics by Magic-Angle
Spinning Solid-State NMR Spectroscopy: Principles and
Applications to Biomolecules. Progress in Nuclear Magnetic
Resonance Spectroscopy, 96, 1–46.
https://doi.org/10.1016/j.pnmrs.2016.02.001
Sharma, R. D., Lynn, A. M., Sharma, P. K., Rajnee, & Jawaid, S.
(2009). High temperature unfolding of Bacillus anthracis amidase-
03 by molecular dynamics simulations. Bioinformation, 3(10), 430–
434. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/19759865
Shin, Y., & Brangwynne, C. P. (2017). Liquid phase condensation in
cell physiology and disease. Science, 357(6357), eaaf4382.
https://doi.org/10.1126/science.aaf4382
Si, K., Giustetto, M., Etkin, A., Hsu, R., Janisiewicz, A. M., Miniaci, M.
C., … Kandel, E. R. (2003). A neuronal isoform of CPEB regulates
local protein synthesis and stabilizes synapse-specific long-term
facilitation in aplysia. Cell, 115(7), 893–904. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/14697206
Si, K., Lindquist, S., & Kandel, E. R. (2003). A neuronal isoform of the
aplysia CPEB has prion-like properties. Cell, 115(7), 879–891.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/14697205
96

Simons, K. T., Kooperberg, C., Huang, E., & Baker, D. (1997).
Assembly of protein tertiary structures from fragments with similar
local sequences using simulated annealing and bayesian scoring
functions. Journal of Molecular Biology, 268(1), 209–225.
https://doi.org/10.1006/jmbi.1997.0959
Soria, M. A., Cervantes, S. A., Bajakian, T. H., & Siemer, A. B. (2017).
The Functional Amyloid Orb2A Binds to Lipid Membranes.
Biophysical Journal, 113(1), 37–47.
https://doi.org/10.1016/j.bpj.2017.05.039
Stanley, N., Esteban-Martín, S., & De Fabritiis, G. (2015). Progress in
studying intrinsically disordered proteins with atomistic simulations.
Progress in Biophysics and Molecular Biology, 119(1), 47–52.
https://doi.org/10.1016/j.pbiomolbio.2015.03.003
Sun, S., Han, Y., Paramasivam, S., Yan, S., Siglin, A. E., Williams, J.
C., … Polenova, T. (2012). Solid-state NMR spectroscopy of
protein complexes. Methods in Molecular Biology (Clifton, N.J.),
831, 303–331. https://doi.org/10.1007/978-1-61779-480-3_17
Walker, F. O. (2007). Huntington’s disease. The Lancet, 369(9557),
218–228. https://doi.org/10.1016/S0140-6736(07)60111-1
Wang, Y., & Jardetzky, O. (2002). Probability-based protein secondary
structure identification using combined NMR chemical-shift data.
Protein Science : A Publication of the Protein Society, 11(4), 852–
861. https://doi.org/10.1110/ps.3180102
Wegmann, S., Eftekharzadeh, B., Tepper, K., Zoltowska, K. M.,
Bennett, R. E., Dujardin, S., … Hyman, B. T. (2018). Tau protein
liquid–liquid phase separation can initiate tau aggregation. The
EMBO Journal, 37(7), e98049.
https://doi.org/10.15252/embj.201798049
Weingarth, M., Demco, D. E., Bodenhausen, G., & Tekely, P. (2009).
Improved magnetization transfer in solid-state NMR with fast
magic angle spinning. Chemical Physics Letters, 469(4–6), 342–
348. https://doi.org/10.1016/J.CPLETT.2008.12.084
Wells, G. A., Müller, I. B., Wrenger, C., & Louw, A. I. (2009). The
activity of Plasmodium falciparum arginase is mediated by a
novel inter-monomer salt-bridge between Glu295-Arg404. FEBS
Journal, 276(13), 3517–3530. https://doi.org/10.1111/j.1742-
4658.2009.07073.x
Zhang, X., Lin, Y., Eschmann, N. A., Zhou, H., Rauch, J. N.,
Hernandez, I., … Han, S. (2017). RNA stores tau reversibly in
complex coacervates. PLoS Biology, 15(7), e2002183.
https://doi.org/10.1371/journal.pbio.2002183

97

APPENDIX
Transfection Procedure
The transfection procedure used in these studies was as follows:
A 200 ul aliquot of cells in an Eppendorf tube, was removed from -80°C
storage and placed on ice. Once the cell aliquot had thawed slightly, 1-
2 ul of 10-25 ug/ul plasmid DNA was added and mixed in while
keeping the cells on ice. Cells were then given a heat shock by holding
the tube in a 42°C bath for 30 seconds. Following heat shock, cells
were returned to ice for a 2 min recovery. 800 ul SOC Recovery media
was then added to the tube and the cells were incubated to 37°C, with
shaking (225 RPM), for 1-2 hours. Finally, transfection was completed
by plating 200 ul of the mixture on an LB plate with appropriate
antibiotics. The plate was left still, inverted and sealed with parafilm
overnight at 37°C and transferred to the fridge the next day. Colonies
were visible on the plate following overnight incubation.
Plate Reader Settings
Table 2. Plate Reader Settings for Figure 3
System AF2200-PC
User AF2200-PC\AF2200
Plate
Greiner 96 Flat Bottom Clear
Polystyrol [GRE96ft.pdfx]

List of actions in this measurement script:

Kinetic

Fluorescence

98

Shaking (Linear) Duration: 1 s
Shaking (Linear) Amplitude: 1 mm

Measurement: Measurement1

Kinetic Measurement

Kinetic duration 16:00:00

Interval Time 0:15:00

Mode Fluorescence Bottom Reading
Excitation Wavelength 440 nm
Emission Wavelength 484 nm
Excitation Bandwidth 20 nm
Emission Bandwidth 25 nm
Gain 100 Manual
Number of Flashes 25

Integration Time 20 µs
Lag Time 0 µs
Settle Time 0 ms

Table 3. Plate Reader Settings for Figure 13
System AF2200-PC
User AF2200-PC\AF2200
Plate
Greiner 96 Flat Bottom Clear
Polystyrol [GRE96ft.pdfx]

List of actions in this measurement script:

Kinetic

Shaking (Orbital) Duration: 3 s
Shaking (Orbital) Amplitude: 1.5 mm
Fluorescence

Measurement:
Measurement1

99

Kinetic Measurement

Kinetic duration 23:59:59

Interval Time 0:15:00

Mode Fluorescence Top Reading
Excitation Wavelength 440 nm
Emission Wavelength 484 nm
Excitation Bandwidth 0 nm
Emission Bandwidth 0 nm
Gain 50 Manual
Number of Flashes 25

Integration Time 20 µs
Lag Time 0 µs
Settle Time 0 ms
Mirror 50% Mirror
Electron Microscopy Grid Preparation Procedure
Electron microscopy grids were prepared by first leaving a 10 µl
drop of the sample being analyzed on parafilm. An electron microscopy
grid was placed face down on the drop and left for 5 min. The grid was
then dried by touching its edge to a piece of filter paper and placed
faced down on a 10 µl drop of uranyl acetate. After 2 min, the grid was
dried again and placed on another 10 µl drop of uranyl acetate. The
grid was then immediately dried and transferred to another drop of
uranyl acetate before being dried and transferred to 10 µl drop of dH2O,
followed by drying and storage.
100

Dynamic domains of amyloid fibrils can be site-
specifically assigned with proton detected 3D
NMR spectroscopy (Full Copy)

COMMUNICATION
Dynamic domains of amyloid fibrils can be site-specifically
assigned with proton detected 3D NMR spectroscopy
Alexander S. Falk
1
• Ansgar B. Siemer
1
Received: 17 August 2016/Accepted: 14 October 2016/Published online: 20 October 2016
Springer Science+Business Media Dordrecht 2016
Abstract Several amyloid fibrils have cores framed by
highly dynamic, intrinsically disordered, domains that can
play important roles for function and toxicity. To study
these domains in detail using solid-state NMR spec-
troscopy, site-specific resonance assignments are required.
Although the rapid dynamics of these domains lead to
considerable averaging of orientation-dependent NMR
interactions and thereby line-narrowing, the proton line-
widths observed in these samples is far larger than what is
regularly observed in solution. Here, we show that it is
nevertheless possible to record 3D HNCO, HNCA, and
HNcoCA spectra on these intrinsically disordered domains
and to obtain site-specific assignments.
Keywords Amyloid fibrils Intrinsically disordered
domains Resonance assignment Solid-state NMR
Many amyloid fibrils have disordered dynamic domains
framing a relatively static b-sheet rich amyloid core (Sie-
mer et al. 2006; Heise et al. 2005; Tompa 2009). The
conformational space and residual structure of these
dynamicdomainsisofinterestsincetheyareonthesurface
of the fibril and might play important roles in the toxicity
of pathological amyloids or the function of functional
amyloids (Isas et al. 2015). Although attached to a
relatively immobile amyloid core, these dynamic domains
exhibit a surprising degree of motions that narrow
1
H lines
to well below 100 Hz (Siemer et al. 2006). As a result J-
coupling based
1
H–
13
C and
1
H–
15
N correlation spec-
troscopy becomes possible and the amino acid composi-
tions of the dynamic domains of several amyloid fibrils
have been reported (Siemer et al. 2006; Isas et al. 2015;
Raveendra et al. 2013; Helmus et al. 2010; Loquet et al.
2009). Site-specific assignments of these dynamic domains
are more challenging because of the limited
1
H resolution
compared to proteins in solution and the signal loss that
occurs during the multiple coherence transfers in three-
dimensional (3D)
1
H–
15
N–
13
C spectra typically acquired
for protein resonance assignment in the liquid-state. In the
following, we show that it is nevertheless possible to use
3D liquid-state NMR
1
H–
15
N–
13
C ‘‘out and back’’ type
pulse sequences on the dynamic domains of an amyloid
fibril. These spectra allow the site specific assignment of
many residues found in the dynamic domain and, conse-
quently, a more stringent analysis of the residual structure
found in these domains.
Figure 1a shows an
1
H–
15
N HSQC and Fig. 1ban
HNCO spectrum recorded on fully protonated and uni-
formly
15
N–
13
C labeled amyloid fibrils formed by the A
isoform of the functional amyloid Orb2 (Orb2A) (Ma-
jumdar et al. 2012). Orb2A fibrils were fully hydrated and
prepared as described in the Supplementary Material. The
spectra were recorded on an Agilent DD2 600 MHz
spectrometer in a 1.6 mm magic angle spinning (MAS)
probe operating at 12 kHz MAS and 25 C. Considering
that Orb2A has 551 residues, the HSQC spectrum only
shows a small portion of the protein with varying intensity
and linewidth. The 1D slices on the peak at d
1
H=
8.01 ppm and d
15
N= 123.9 ppm in Fig. 1a illustrate the
quality of this spectrum. The
1
H linewidth of*60-Hz is
Electronic supplementary material The online version of this
article (doi:10.1007/s10858-016-0069-2) contains supplementary
material, which is available to authorized users.
& Ansgar B. Siemer
asiemer@usc.edu
1
Department ofBiochemistry andMolecular Medicine,Zilkha
Neurogenetic Institute, Keck School of Medicine of USC,
1501 San Pablo St, Los Angeles, CA 90033, USA
123
J Biomol NMR (2016) 66:159–162
DOI 10.1007/s10858-016-0069-2
well above what is regularly obtained for small globular
proteins in solution (\10-Hz) (Cavanagh et al. 1995) and
still larger than what was found for highly perdeuterated,
crystalline proteins in the solid-state (*25 Hz) (Agarwal
etal.2006).Thesameistrueforthe
15
Nlinewidthofabout
40 Hz. We found that higher temperatures changed the
efficiency but had little effect on the linewidth of our
HSQC spectra. Higher MAS frequencies had no effect on
the spectra besides theresultingchange intemperature (see
Figure S1).
We are confident that the signal observed in these
experiments originates from dynamic regions of the fibril
and not free monomers or oligomers based on the follow-
ing reasons: (1) the signal-to-noise ratio of the HSQC
recorded in about 1 h is good, (2) we washed our fibrils
multiple times before packing the NMR rotor, (3) our EM
images do not show any Orb2A oligomers, and (4)
monomers and oligomers would likely have more highly
dynamic residues than the small number of peaks observed
in the HSQC.
Despite the limited resolution, we were able to record a
3D HNCO experiment (Kay et al. 2011; Grzesiek and Bax
1992) on this sample. In the absence of a gradient coil in
our MAS probe, we obtained excellent water suppression
usingxandy
1
Hpurgepulses(MuhandiramandKay1994)
asillustratedinFigureS2.AscanbeseenfromFig. 2b,the
signal-to-noise ratio for this HNCO experiment is*67:1,
which given the acquisition time of about 70 h, is
excellent.
Due to the larger chemical shift dispersion of C
a
chemical shifts, HNCA and HNcoCA experiments (Kay
etal.2011;GrzesiekandBax1992)(seepulsesequencesin
Figure S2) are better suited for site-specific backbone
assignments. The HNCA experiment gives cross peaks that
correlate the amide
1
H and
15
N to the C
a
of the same and
preceding residue. The HNcoCA experiment, by contrast,
only shows the peak corresponding to the C
a
of the pre-
ceding residue. The combination of the two experiments
allows the backbone assignment of all H
i
N
–N
i
H
–C
a
reso-
nances. Although these experiments are less efficient than
the HNCO experiment, we were still able to record both
experiments on the same sample as can be seen from
Fig. 2.
The signal-to-noise ratios of the C
i
a
peak in the HNCA
spectrum and the C
i-1
a
peak in the HNcoCA spectrum
whichcorrespondtotheNHpeakhighlightedinFig. 1aare
23:1 and 18:1 respectively. Using these spectra, we
sequentially connected many of the cross peaks in Fig. 1a.
We then identified amino acid types using the 2D
13
C–
13
C
adiabatic TOBSY spectrum shown in Figure S3. The
C
i
a
–C
i-1
a
connectivities were confirmed using the
H
i
N
–N
i
H
–C
i-1
0
resonances from the HNCO spectrum and the
C
i
0
–C
i
a
peaks from the
13
C–
13
C CTUC-COSY (Chen et al.
2006) spectrum shown in Figure S2. The assignment was
further facilitated by (a) the fact that there are relatively
few peaks in these spectra, (b) that the HNCA and
HNcoCA spectra are dominated by very strong Gly peaks
that have the same C
i
a
and C
i-1
a
shift indicating a poly-Gly
structure, and (c) identifying a serine-threonine pair that is
unique in the sequence of Orb2A (S160-T161). We con-
sequentlydeterminedthe sequential assignment ofmanyof
the cross peaks in Fig. 1a and identified them as belonging
to the sequence highlighted in red in Fig. 3a. The Gly
(a) (b)
Fig. 1 The most dynamic domains of amyloid fibrils have a
linewidth narrow enough to allow HSQC and HNCO spectra in the
absence of perdeuteration. a
1
H–
15
N HSQC spectrum of fibrils
formedbyOrb2A.1Dslicesandlinewidthsforoneresidueareshown
to illustrate spectral quality. Site-specific assignments are indicated.
b Strip plot of a HNCO spectrum from the same cross peak as in a).
The1D slice throughthe
13
Cdimension isshown toillustratespectral
quality
Fig. 2 The combination of HNcoCA and HNCA experiments allows
the site-specific backbone assignment of dynamic amyloid fibril
domains. Strip plot of HNcoCA (blue) and HNCA (red) experiments
showingresiduesG159-G162.TheHNCAexperimentshowsboththe
H
i
N
–N
i
H
–C
i
a
and H
i
N
–N
i
H
–C
i-1
a
correlations whereas the HNcoCA
shows only the H
i
N
–N
i
H
–C
i-1
a
, allowing assignment of both the C
i
a
and
C
i-1
a
shift. The large peak associated with G159 is likely because
G155-G159 are all glycine
160 J Biomol NMR (2016) 66:159–162
123
repeats framing the assigned sequence are very likely part
of this dynamic domain, resulting in the intense Gly peaks
seen in Figs. 1a and 2. Due to their repetitive nature, these
Glys could not be assigned with certainty. The same is true
fortheintensepeakusedforthe1DslicesinFig. 1a,which
is likely D173 based on its side-chain assignment and its
N-terminal Gly neighbor. Other cross peaks in the HSQC
spectrum in Fig. 1a were not assigned because they did not
have corresponding peaks in all 3D experiments.
This assignment is compatible with the assumption that
Gly and Ser rich regions have a high propensity to be
intrinsically disordered (Krieger et al. 2003), and this
region is likely to form a flexible linker between the
N-terminal domain thathas been showntobeimportant for
amyloid formation (Majumdar et al. 2012) and the RNA
recognition motifs (RRMs) at the C-terminus of the pro-
tein. The relatively small C
a
secondary shifts of the
assignedresiduesindicatethatthemostdynamicdomainof
Orb2A has relatively little residual structure (see Fig. 3b).
The fact that a small region of Orb2A gives signal in our
amide
1
H detected 2D and 3D experiments does not mean
that these are the only dynamic domains in the protein, but
that other regions are not dynamic enough to be detected
using this approach. Some of these less dynamic residues
givesignalsintheadiabaticTOBSYspectrumofFigureS3
but no corresponding signal in the 3D spectra.
How dynamic are the assigned regions when compared
to globular proteins in solids and solution? The fact that
repeated INEPT transfers work for residues in the most
dynamic regions indicates that dipolar couplings are
motionally averaged to a degree that coherent dipolar
dephasing is minimal. Therefore, we estimate the order
parameter S= d
exp
/d
rigid
as described by the scaling of
effective dipolar interactions (d
exp
) below its rigid value
(d
rigid
)tobeS& 0 (Schanda and Ernst 2016). Assuming a
rotational diffusion on a cone (with angle h) model
S
2
¼ 1=2cosh ðÞð1þcoshÞ
2
, a vanishing order parameter
describes a motion that is essentially isotropic. We esti-
mated the correlation time of the motion via the transverse
relaxation of the backbone
15
N because it can be suffi-
ciently approximated by the H–N dipolar coupling and its
chemical shift anisotropy (CSA). Using the expression for
the transverse relaxation rate under MAS by Schanda &
Ernst and approximating the
15
N transverse relaxation rate
R
2
via the
15
N linewidth of *40 Hz, we estimate the
correlation time of the dynamic regions to be s
C
& 95 ns
(see Supplementary Material). Under these assumptions,
the motional model would be the same as in solution and
the correlation time corresponds to the rotational correla-
tion time of a globular protein with a mass of about
300 kDa. This correlation time is only an estimate because
factors such as dipolar dephasing, sample and field
heterogeneity also contribute to the linewidth and the order
parameter could be larger than S
2
= 0. Overestimating the
relaxation rate from the linewidth leads to larger correla-
tiontimesandunderestimatingtheorderparameterleadsto
smaller correlation times. Assuming non R
2
contributions
to the linewidth of up to 20 Hz and order parameters of up
to S
2
= 0.2, the correlation time is in the range of
47-119 ns.
Our results demonstrate that
1
H detected J-coupling
based 3D pulse sequences can be used to assign the most
dynamic domains of non-soluble proteins in the absence of
deuteration. We think that this approach will be generally
applicable to amyloid systems that have considerable dis-
ordered domains outside the fibril core (Siemer et al. 2006;
Isas et al. 2015; Raveendra et al. 2013; Heise et al. 2005;
Helmus et al. 2010; Loquet et al. 2009) and potentially
other insoluble protein complexes that retain intrinsically
disordereddomains.Manyintrinsicallydisordereddomains
adopt well defined 3D structures when interacting with
binding partners (Uversky 2013). Using this approach
allows site-specific monitoring of conformational changes
in intrinsically disordered domains that are present in
amyloid fibrils or larger protein complexes.
Additional
1
H detected, J-coupling based experiments
might be possible depending on the degree of dynamics
and amount of sample. Only the most dynamic domains of
the protein were detected using this approach and less
dynamic domains that give signals in straight-though
13
C
detected experiments such as the adiabatic TOBSY could
not be assigned sequentially. In addition, there is a regime
of protein dynamics in which neither dipolar based cross
(a)
(b)
Fig. 3 The most dynamic domain of the Orb2A amyloid fibril is
located in the Glycine and Serine rich region and is intrinsically
disordered. a Domain structure of Orb2A. The N-terminus, Gln-rich
domain (Q), Glycine and Serine rich domain (G/S), RNA recognition
motifs(RRMs),andzincfinger(Zn)arehighlighted.Theresiduesthat
we assigned sequentially are highlighted in red in the sequence
excerpt below. b Secondary C
a
shifts of sequentially assigned
residues. The relatively small values indicate the absence of
secondary structure for this domain
J Biomol NMR (2016) 66:159–162 161
123
polarization spectra nor J-coupling based spectroscopy is
applicable and only direct excitation can be used to detect
these domains. Sometimes even this is not possible if
intermediate exchange is broadening the lines below the
detection limit. A possible approach to extend this method
to less dynamic regions is to combine it with partial
deuteration, similar to approaches used to reduce linewidth
for large proteins in solution (Yamazaki et al. 1994)orto
allow
1
H based spectroscopy on proteins with little
dynamics in the solid-state (Reif 2012). Furthermore,
TROSY based methods might allow the extension of our
method to less dynamic regions as demonstrated by Linser
et al. (2010). Where full perdeuteration combined with
1
H
back exchange allows the use of
1
H based spectroscopy on
relatively static protein samples, partial deuteration and
TROSY might allow our experiments to be performed on
protein domains that are less dynamic than those identified
here. The current approach, however, has the advantage of
being selective to the most dynamic domains.
Acknowledgments The authors thank Tobias Ulmer and Matthias
Ernst for fruitful discussions. This work was supported by the
University of Southern California, the Whitehall Foundation, and the
National Institutes of Health: NIGMS Award R01GM110521.
References
AgarwalV,DiehlA,SkrynnikovN,ReifB(2006)Highresolution1H
detected 1H, 13C correlation spectra in MAS solid-state NMR
using deuterated proteins with selective 1H, 2H isotopic labeling
of methyl groups. J Am Chem Soc 128(39):12620–12621
Cavanagh J, Fairbrother WJ, Palmer AG, Skelton NJ (1995) Protein
NMR spectroscopy: principles and practice. Elsevier Science,
Amsterdam
Chen L, Olsen RA, Elliott DW, Boettcher JM, Zhou DH, Rienstra
CM, Mueller LJ (2006) Constant-time through-bond 13C
correlation spectroscopy for assigning protein resonances with
solid-state NMR spectroscopy. J Am Chem Soc
128(31):9992–9993
Grzesiek S, Bax A (1992) An efficient experiment for sequential
backbone assignment of medium-sized isotopically enriched
proteins. J Magn Reson 99(1):201–207
Heise H, Hoyer W, Becker S, Andronesi OC, Riedel D, Baldus M
(2005) Molecular-level secondary structure, polymorphism, and
dynamics of full-length alpha-synuclein fibrils studied by solid-
state NMR. Proc Natl Acad Sci USA 102(44):15871–15876
Helmus JJ, Surewicz K, Surewicz WK, Jaroniec CP (2010) Confor-
mational flexibility of Y145Stop human prion protein amyloid
fibrils probed by solid-state nuclear magnetic resonance spec-
troscopy. J Am Chem Soc 132(7):2393–2403
Isas JM, Langen R, Siemer AB (2015) Solid-state nuclear magnetic
resonance on the static and dynamic domains of Huntingtin
exon-1 fibrils. Biochemistry 54(25):3942–3949
Kay LE, Ikura M, Tschudin R, Bax A (2011) Three-dimensional
triple-resonance NMR spectroscopy of isotopically enriched
proteins. 1990. J Magn Reson 213(2):423–441
Krieger F, Fierz B, Bieri O, Drewello M, Kiefhaber T (2003)
Dynamics of unfolded polypeptide chains as model for the
earliest steps in protein folding. J Mol Biol 332(1):265–274
Linser R, Fink U, Reif B (2010) Assignment of dynamic regions in
biological solids enabled by spin-state selective NMR experi-
ments. J Am Chem Soc 132(26):8891–8893
Loquet A, Bousset L, Gardiennet C, Sourigues Y, Wasmer C,
Habenstein B, Schu ¨tz A, Meier BH, Melki R, Bo ¨ckmann A
(2009) Prion fibrils of Ure2p assembled under physiological
conditions contain highly ordered, natively folded modules.
J Mol Biol 394(1):108–118
Majumdar A, Cesario WC, White-Grindley E, Jiang H, Ren F, Khan
MR, Li L et al (2012) Critical role of amyloid-like oligomers of
Drosophila Orb2 in the persistence of memory. Cell
148(3):515–529
Muhandiram DR, Kay LE (1994) Gradient-enhanced triple-resonance
three-dimensional NMR experiments with improved sensitivity.
J Magn Reson Ser B 103(3):203–216
Raveendra BL, Siemer AB, Puthanveettil SV, Hendrickson WA,
Kandel ER, McDermott AE (2013) Characterization of prion-
like conformational changes of the neuronal isoform of Aplysia
CPEB. Nat Struct Mol Biol 20(4):495–501
Reif B (2012) Ultra-high resolution in MAS solid-state NMR of
perdeuterated proteins: implications for structure and dynamics.
J Magn Reson 216:1–12
Schanda P, Ernst M (2016) Studying dynamics by magic-angle
spinning solid-state NMR spectroscopy: principles and applica-
tionstobiomolecules.ProgNuclMagnResonSpectrosc96:1–46
SiemerAB,ArnoldAA,RitterC,WestfeldT,ErnstM,RiekR,Meier
BH (2006) Observation of highly flexible residues in amyloid
fibrils of the HET-s prion. J Am Chem Soc
128(40):13224–13228
Tompa P (2009) Structural disorder in amyloid fibrils: its implication
in dynamic interactions of proteins. FEBS J 276(19):5406–5415
Uversky VN (2013) A decade and a half of protein intrinsic disorder:
biology still waits for physics. Protein Sci Publ Protein Soc
22(6):693–724
Yamazaki T, Lee W, Arrowsmith CH, Muhandiram DR, Kay LE
(1994) A suite of triple resonance NMR experiments for the
backboneassignmentof15N,13C,2Hlabeledproteinswithhigh
sensitivity. J Am Chem Soc 116(26):11655–11666
162 J Biomol NMR (2016) 66:159–162
123

Abstract (if available)

Abstract Orb2 is a functional, amyloid forming, protein that plays a critical role in long term memory formation in Drosophila melanogaster, with homologs that play a similar role in other species, including humans. Orb2 has two isoforms, Orb2A and Orb2B, and their interaction to form functional amyloids is hypothesized to be a key step in the memory formation process. Huntingtin is a protein that can form pathological amyloids and cause Huntington’s disease when mutated. It also has critical functional roles in normal neural development when not mutated. ❧ In this thesis, I present a series of experiments on Orb2, with goal of understanding the amyloid structure and how it relates to Orb2’s biological role. I show that Orb2 is capable of forming protein droplets and two different types of fibrils (one of which seems to be derived from droplets). I also show that fibrils formed by Orb2A and Orb2B are highly similar and that within a single monomer in a fibril, residues range in dynamics from extremely static to almost liquid-like. I also present a series of molecular dynamics simulations of a domain of Huntingtin. I show that these simulations can reproduce EPR and NMR data gathered on Huntingtin fibrils, and also help to identify what appears to be a highly conserved protein-protein interaction site in the domain. ❧ Because this thesis covers a range of experiments on different proteins from different sources, I first present an introduction providing a general background for all topics and methods discussed (CHAPTER 1. INTRODUCTION) before delving into experiments on Orb2 derived from non-soluble protein fractions (CHAPTER 2. PROJECT I: STUDIES ON NON-SOLUBLE FRACTION ORB2), experiments on Orb2 derived from soluble protein fractions (CHAPTER 3. PROJECT II: STUDIES ON SOLUBLE ORB2 AND ΔRRM FRAGMENTS) and simulations of the Huntingtin exon-1 C-terminus (CHAPTER 4. PROJECT III: Computational Modelling of Huntingtin C-terminus).

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Regulating functional amyloid formation: the promiscuous behavior of the Orb2A N-terminal amphipathic region

PDF

Structural studies on functional amyloids and the mechanism of aggregation and disaggregation of Huntingtin Exon 1

PDF

Overexpression and interaction of Orb2 in S2 insect cells

PDF

Structural characterization of the functional amyloid Orb2A using EPR spectroscopy

PDF

Oligomer formation of functional amyloid protein - Orb2A

PDF

Facilitating unambiguous NMR assignment by solid–state NMR using segmental isotope labeling through split-inteins

PDF

Protein-protein interaction and protein-lipid interaction of membrane proteins

PDF

Huntington’s disease vaccine development using fungal prion HET-s

PDF

Interaction of monoclonal antibodies MW1 and PHP1 with huntingtin exon1 protein

PDF

Before they were amyloid: understanding the toxicity of disease-associated monomers and oligomers prior to their aggregation

PDF

Algorithm development for modeling protein assemblies

PDF

The structure and function of membrane curving proteins on different membrane shapes and their regulation by post-translational modifications

PDF

Molecular basis of CD33 receptor function, and effects of a destabilizing transmembrane motif on αXβ2 integrin

PDF

Role of mitochondrially derived peptides in the inhibition of IAPP misfolding

PDF

Structural features and modifiers of islet amyloid polypeptide: implications for type II diabetes mellitus

PDF

Uncovering the influence of N-terminal phosphorylation on conformational dynamics of huntingtin exon 1 monomer

PDF

Vitronectin misfolding and aggregation: implications for the pathophysiology of age-related diseases

PDF

Enhancing and inhibiting diabetic amyloid misfolding

PDF

Dual effects of transmembrane proline residues on integrin function

PDF

Structural studies of the IAPP membrane-mediated aggregation pathway

Asset Metadata

Creator Falk, Alexander Stover (author)

Core Title Structure and kinetics of the Orb2 functional amyloid

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Medical Biophysics

Publication Date 06/13/2019

Defense Date 03/07/2019

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

Tag Amyloid,CPEB,fibril,functional amyloid,huntingtin,long term memory,molecular dynamics,molecular modelling,OAI-PMH Harvest,Orb2,protein droplets,protein phase separation,ssNMR

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Langen, Ralf (committee chair), Farley, Robert (committee member), Haworth, Ian (committee member), Siemer, Ansgar (committee member)

Creator Email falka@usc.edu,sandyfalk@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-172103

Unique identifier UC11660853

Identifier etd-FalkAlexan-7473.pdf (filename),usctheses-c89-172103 (legacy record id)

Legacy Identifier etd-FalkAlexan-7473.pdf

Dmrecord 172103

Document Type Dissertation

Format application/pdf (imt)

Rights Falk, Alexander Stover

Type texts

Source 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 a...

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