Identifying determinants of conformational stability in GTPases
Vishwajith Sridharan
A thesis submitted in partial fulfillment of the requirements for graduation with
Highest Honors, Bachelor of Science, Chemistry (Biochemistry Track)
The University of North Carolina at Chapel Hill
Abstract
Cellular functionalities are contingent upon the ability of proteins to maintain or transform their
shape (conformation). The molecular forces that govern protein folding have long been known1. These
include hydrophobic interactions, hydrogen bonding, and disulfide bonds. However, how proteins
maintain their conformation remains largely unanswered. In this study, twenty G proteins from
Saccharomyces cerevisiae were cloned, purified from E. coli, and tested for conformational stability. The
proteins vary in Tm from 35 to 60°C. This is paradoxical considering their highly analogous crystal
structures and identical catalytic activity. We find that thermal stability correlates with percentage of Cys,
Met, or Pro residues in the primary sequence. We also find that the stabilities group by their respective
family members, with the following trend observed for least to most thermostable G proteins: Rho/Cdc42
< Ran < Large G-proteins < Ras < Rab < Sar/Arf. Both these results suggest that thermal stability has a
greater dependence on primary sequence than tertiary structure. Ultimately, a more detailed understanding
of protein stability can help in engineering enzymes for new molecular therapeutics.
Introduction
A detailed understanding of protein thermodynamics would enable us to predict protein stability
from genome sequences and protein three-dimensional structures. This capability would transform
approaches for studying disease, engineering industrial-strength enzymes, designing biosensors, and
re-designing proteins for specific therapeutic and biotechnological purposes. For example, cancer-related
mutations in the p53 oncogene often cause a reduction in protein stability that can be counteracted by a
drug that specifically restores conformational stability.2 Similar approaches can be used to improve the
stability of industrially used enzymes (e.g. cellulase) that are used to convert biomass to fuel.3 The use of
thermostable enzymes like Taq Polymerase for DNA amplification has already revolutionized how we do
molecular biology.
Many methods have been introduced to increase conformational stability, but it remains difficult
stability. Because it is currently not possible to measure the conformational stability of the entire
proteome, our approach is to focus on a large protein family with a well-documented evolutionary history.
We have chosen to focus on G proteins – one of the largest, most extensively studied protein families.8,9,10
The G protein family is comprised of large and small G proteins that are conserved from yeast to
humans (Figure 1). G-proteins, which exist as monomers, function as molecular switches that cycle
between inactive guanosine diphosphate (GDP) bound and active guanosine triphosphate (GTP) bound
states. Guanine nucleotide exchange factors (GEFs) promote the exchange of GDP for GTP and guanine
nucleotide activating proteins (GAPs) promote the hydrolysis of GTP. Signaling by large G proteins work
to translate environmental and chemical stimuli into appropriate cellular responses. In yeast, G protein
coupled receptors on cell surfaces are used for mating and glucose sensing, and in plants they are essential
for ion homeostasis and cell growth11. In humans, large G proteins participate in physiological processes
such as vision, olfaction, inflammation, immunity, cardiac function, and neurotransmission12. Most
importantly, more than half of all drugs (e.g. beta blockers, antihistamines, antidepressants) act on G
protein signaling pathways to mitigate pathologies such as depression, inflammation, or hypertension.13
Like large G proteins, small G proteins are also found in all eukaryotes, and also function as
molecular switches14. Small G proteins are divided into several different families, namely the Ras, Rho,
Rab, Sar/Arf, and Ran families. Despite phylogenetic similarities, small G proteins participate in a wide
variety of activities inside the cell. The Ras family regulates gene expression and is involved in
cAMP-mediated signaling. The Rho family regulates cytoskeletal organization, and in yeast, polarized cell
growth. The Sar/Arf and Rab families are involved in the priming of vesicle formation, and vesicle
docking, respectively. The Ran family participates in nucleocytoplasmic transport and microtubule
organization15.
Here, we determine the thermal stabilities of twenty GTPases from Saccharomyces cerevisiae.
Our results show that conformational stability correlates with the percentage of Cys, Met, or Pro residues
Results
Protein conformational stability (thermostability) is assessed by unfolding a protein as a function
of increasing temperature. The resultant unfolding curve is sigmoidal in shape and is characterized by
two plateaus (one each for the folded and unfolded states) that bracket a cooperative unfolding transition.
Thermostability is quantified as the midpoint of this cooperative transition (Tm), and represents the
temperature at which half of the protein population is unfolded. Higher Tm values indicate proteins that
are more thermostable16,17. In this study, two independent methods – fast quantitative cysteine reactivity
(fQCR) and thermofluor assays – are used to determine Tmvalues for candidate proteins. The fQCR
approach uses cysteine labeling as a proxy for measuring the extent of protein unfolding, and has been
described previously18. At low temperatures, cysteine residues that are protected by the fold of the protein
cannot be covalently labeled by a fluorogenic reagent. However, as temperature is raised, these protected
cysteines become increasingly exposed to solution (and react) as the protein unfolds. The ThermoFluor
assay uses another proprietary reagent (Sypro Orange) whose fluorescence is quenched in an aqueous
environment. However, as temperature rises, the protein exposes its hydrophobic core region, to which
Sypro binds and becomes unquenched19. In both fQCR and ThermoFluor, dye binding to the protein can
be spectroscopically detected and quantified.
Twenty GTPases, including two large and eighteen small G proteins, were included in this study
(Table 1). The criteria for selection of these GTPases included ease of biochemical purification,
well-studied phylogeny and evolutionary history, conservation in higher eukaryotes, relevance to human health
and disease, as well as diverse cellular functions and localization. All twenty of the GTPases tested here
are evolutionarily conserved from yeast to humans, and include members from each of the G protein
families previously mentioned. The GTPases also exhibit identical catalytic activity, and have highly
homologous crystal structures. Figure 2 presents the conserved ribbon diagram alignment of existing
yeast or homologous eukaryotic crystal structures of the G proteins tested (see Supplementary Data for
more varied, with about 16% identity when all twenty are aligned, and higher for individual family
sequence alignments (25% Cdc/Rho, 40% Sar/Arf, 36% Rab, 59% Ras).
Each GTPase was cloned out of the yeast genome, tagged with a 5x His sequence, overexpressed
and purified from E. coli at high yield and purity before thermal stability was measured. The following
example with Cdc42 presents generalized approach used to investigate conformational stabilities. Yeast
Cdc42 is small rho-type GTPase involved in actin cytoskeletal remodeling. It can substitute for human
Cdc42, which is involved in cell-cycle progression. Figure 3 presents the purification scheme for Cdc42.
The first two lanes represent E. coli lysates, the next two lanes represent the washes after binding to
Ni-metal affinity beads, and the last two lanes represent the eluted pure protein. In this case, Cdc42 was
purified as a full-length protein and with its terminal four amino acids (CAAX-box) deleted to improve in
protein solubility. We then performed thermal stability assays (Figure 4). As temperature was increased,
the protein progresses from a baseline folded through a cooperative unfolding transition to an unfolded
state. The Tm is determined as the midpoint (or inflection point) of this transition. The stability of the
GTP-bound state of Cdc42 (47°C) is significantly higher than the GDP-bound state (40°C). This is likely
due closing of the nucleotide binding pocket in the GTP-bound state and a more compact conformation.
However, there does not appear to be a difference between the full-length and CAAX-truncated versions
of Cdc42 in both the GDP and GTP bound states (< 1°C Tm difference). Similarly, purification gels and
unfolding curves for the remaining nineteen proteins can be found in the Supplementary Data.
We used two independent methods, fQCR and ThermoFluor, to determine the conformational
stability of the GTPases. To ensure that both assays yield similar results for individual proteins, the
average Tm values obtained from two methods were correlated. Figure 5 shows a strong, positive, linear
correlation between the data points from the two thermal stability assays. As both proxies to detect
protein unfolding and stability provide highly similar results, we can be confident in the fidelity of the Tm
measurements.
Using these two techniques, the thermal stabilities of twenty G proteins was systemically
stable. The Tm values range from 35°C to 60°C for the G proteins studied. This result is paradoxical
considering the extensive structural and catalytic similarities between these proteins. It also raises two
avenues to explore. First, the origin of these differences is unclear. Second, there may be patterns in the
distribution of Tm values observed. We explore both of these questions in turn.
Several parameters are already known about the proteins tested, including their abundance in
vivo, hydropathicity, pI, contact order, etc. It is possible that any of these metrics could serve as predictors
of conformational stability. So, we undertook a bioinformatics analysis to correlate Tm with other known
protein parameters. In Table 2, we present the results of this statistical analysis (Pearson correlation). The
only metrics that yielded statistically significant correlations with stability values were percentage of
cysteine (p = 0.0006), methionine (p = 0.032), and proline (p = 0.0139) residues. This result is surprising
and implies a relationship between primary sequence and conformational stability. It also raises questions
about what is unique to these three residues, which we explore in the discussion below.
We also sought to determine discernible patterns in the distribution of Tm values. We find that the
conformational stability values group together by family. The Rho/Cdc42 family with the lowest average
Tm (41°C), followed by the Ran (45°C), then the large G proteins (46°C), then Ras (50°C), Rab (52°C),
and finally Sar/Arf (54°C). This result is not evident a priori, because the different G protein families
share significant structural homology. However, as individual families show greater primary sequence
variation, we can again imply a relationship between primary sequence and stability. For example, there
are several different ways to make an α helix or a β sheet, and some helix-sheet combinations might be
more thermostable than others. Moreover, extensive evolutionary analyses have previously shown that
different G protein families diverged at different evolutionary times20. We speculate that the pattern
observed also indicates a link between the evolutionary phylogeny of these G proteins and their
Discussion
Since the origin of life on earth, the survival of organisms, and the proper function of their
constituent proteins, has been contingent on the ability to adapt to environmental challenges and stresses,
especially temperature. The development of more efficient cloning techniques21, higher yield purification
schemes22, and rapid thermal stability testing methods23 has finally allowed us to systemically test
differences in protein conformational stabilities. In this study, twenty G proteins from Saccharomyces
cerevisiae were cloned, purified from E. coli, and tested for conformational stability. The proteins tested
vary in Tm over a range of 25°C. This is paradoxical considering their highly analogous crystal structures
and identical catalytic activity. Our analyses indicate three metrics – the percentage of Cys, Met, or Pro
residues in the primary sequence that correlate with the thermal stability values obtained.
Cysteine and methionine are among the amino acids most sensitive to oxidation due to
sulfur-containing moieties. Oxidation is often accompanied by increased unfolding of the protein, loss of protein
functionality, aggregation, and serves as structural markers for recognition by the proteasome for
degredation24,25. Oxidatively damaged proteins have been shown to accumulate in aging and several
neurodegenerative diseases. Ras proteins also contain a solvent accessible cysteine prone to oxidation,
resulting in enhanced guanine nucleotide exchange and Ras activation. Aberrant Ras activation has been
observed in 30% of human tumors26. Thus, the decreased thermal stabilities observed with increasing
numbers of residues susceptible to oxidation can provide a biophysical explanation for the physiological
consequences of oxidation.
The data above indicate higher percentages of prolines also correlate with decreased
conformational stability. Proline is also prone to oxidation. However, it is unique is that its pyrrolidine
ring imposes a rigid constraint on the available conformational space, introducing a kink in the secondary
structure27. Abrupt changes in the direction of the polypeptide chain due to high numbers of proline
residues may have a detrimental effect on overall protein structure and explain decreases in stability.
Studies have shown that optimizing these β-turns in the secondary structure by coupling proline with an
Another striking observation from Figure 6 is that stabilities sort alongside family members,
which in turn is determined by function in vivo. The following trend is observed for least to most
thermostable G proteins: Rho/Cdc42 < Ran < Large G-proteins < Ras < Rab < Sar/Arf. Phylogenetic
analyses of all G proteins have shown that different families diverge at different points29 through
sequence mutations. This suggests a link between evolutionary ancestry and thermal stability. Ambient
temperature is among the most fundamental stressors for single-celled organisms. Over the last two
decades, much has been discovered about how yeast reprogram their transcriptional and metabolic
machinery to adapt to thermal stress30. Recent genomic screens have shown that deletion mutants with the
highest thermosensitivity are not the same as those transcriptionally activated in response to heat shock.
Instead, survival appears to depend on a small number of genes selected for gauging metabolic health in
an ever-changing environment31.
We speculate that as these G protein sub-families evolved to perform a wide variety of activities
within the cell – from signaling, to nuclear transport, to cytoskeletal remodeling – their molecular
machinery was tuned to withstand varying degrees of thermal stress. This would imply that as
temperature is increased, some G protein families will show loss of function, likely because their
functions are dispensable under thermal stress. Several studies lend credence to this hypothesis. In S.
pombe, thermal stress leads to a period of decreased Cdc42 activity and altered monopolar growth32. Heat
shock in S. cerevisiae leads to nuclear retention of most poly(A)-mRNA, likely due to loss of Ran/Gsp1
activity, but stress-associated mRNAs are efficiently exported through a Ran independent pathway33. This
alternate pathway is dispensable under normal growth conditions, but becomes essential under heat stress.
Both Cdc and Ran family GTPases are among the least thermostable of the G protein family members
tested.
If different G protein families have evolved with differential thermotolerance, there must also be
evidence among the more thermostable families tested. Indeed, both Arl1 and Ypt6 mutants have shown
higher sensitivity to increased temperature than wild-type cells34. Moreover, activation of heat shock
important to acknowledge, however, that in vivo stability of proteins may be distinct from that measured
in vitro, given the complexity of the cellular milieu, counting effects of macromolecular crowding, the
presence of chaperones, interaction partners, etc.
Our results suggest several future experiments. In the short term, recombinant constructs where
portions of a more thermostable GTPases are grafted onto less themostable ones can yield additional clues
on how specific sequence or structural elements confer conformational stability. Mutants lacking cysteine,
proline, or methionine residues (or supplemented with additional ones) can also be used to better discern
the roles these amino acids play in conferring conformational stability. Minimizing the presence or
optimizing the positions of these three residues may provide an avenue to engineer a protein’s
conformational stability.
In the long term, given the relationship between primary sequence and stability seen here, we
propose to use existing bioinformatics and phylogenetic mapping tools36,37 to reconstruct evolutionary
ancestor sequences for individual G protein families and compare ancestral Tm values with extant ones.
Evolution is often assumed to select for a protein’s biochemical function rather than stability. Function is
in turn dependent on a protein’s ability to fold into a thermodynamically stable state. Thus, stability too is
constrained. Indeed, studies have shown that increased protein stability promotes evolvability38.
Ultimately, identifying changes in protein stability from primary sequences or engineering greater
stability into enzymes could provide new molecular diagnostics or therapeutics.
Experimental Procedures
Genomic PCR. Standard procedures for growth, maintenance, and transformation of yeast and bacteria
and for the manipulation of DNA were used throughout. Yeast (BY4741) were grown in 5 mL of YPD
media overnight, centrifuged for 5 minutes at 4500rpm, washed twice with dH2O, and resuspended in 150
μL of breaking buffer (2% Triton, 1% SDS, 100 mM NaCl, 10 mM Tris-Cl pH 8, 1 mL EDTA). Cells
supernatant containing genomic DNA was diluted 1:15 in 10 mM Tris, pH 8.5. PCR was performed using
established protocols, and sequenced using an external vendor to ensure fidelity.
Ligation-Independent Cloning and Transformation. All enzymes were obtained from NEB, and NEB
protocols were followed with exceptions noted. pLIC-His was digested with Ssp1 for 1 hour at 37°C,
followed by calf-intestinal phosphatase treatment for another hour at 37°C to prevent self-ligation. The
treated vector was gel purified, and overhangs were created with T4 DNA Polymerase in NEBuffer 2
containing 25 mM DTT, 2.5 mM dGTP at 22°C for 30 minutes. The insert from genomic PCR was
phosphorylated with poly-nucleotide kinase at 37°C for 1 hour, and overhangs were created as described,
but with 2.5 mM dCTP. The treated vector (15 ng) was incubated with the insert (30 ng) for 1 hour at
37°C, and transformed into competent E. coli XL1 Blue and BL21 (DE3)–RIPL E. coli cells (Stratagene)
using established procedures.
Protein Purification. The proteins used in this study were overexpressed to high density using the
autoinduction method introduced by F.W. Studier (Studier, 2005), All purification steps were done in a
PBS-GMT buffer (25 mM KPO4, 100 mM KCl, 50 mM GDP, 50 mM MgCl2, 1 mM TCEP [pH 7.0]),as
adapted and described in the Supplemental Materials of Isom et al, 201339.
Quantitative Cysteine Reactivity. Thermal unfolding of proteins in this study was quantified by fQCR.
Protein stock (in PBS-GMT) was diluted to 1 μM in cold PBS (pH 7.0) with 50 μM GDP or GTPγS and
chilled on ice. For each unfolding reaction, 5 μL of 26 mM working ABD stock solution (1 mM final)
was combined with 125 μL of ice-cold protein sample. Samples were mixed on ice, and 10 μL aliquots
were distributed across a 12-well PCR strip-tube (USA Scientific, 1402-2408), also chilled on ice. The
samples were then placed in a gradient thermocycler (Biometra Professional Thermocycler), heated for 3
minutes at a gradient of temperatures, immediately transferred to ice, and quenched by adding 2 μL of
quenched samples were then transferred to a 384-well plate (Greiner, 788076) and their fluorescence
values were quantified in a BMG Labtech PHERAstar plate reader using excitation and emission
bandpass filters of 400 and 500 nm. The resultant thermal unfolding curves were fit using a two-state
model of protein unfolding to provide values for the midpoint of thermal unfolding (Tm).
Thermofluor Assay. Protein stock (in PBS-GMT) was diluted to 5 μM, and 125 μL of the diluted protein
was mixed with 1.5 μL of 500x SYPRO (Invitrogen, S-6650) so the final protein concentration is between
1 and 6 μM in 6x SYPRO. The mixture was distributed in 20 μL aliquots across a 384 white well plate
(Genesee #24-305W) and sealed with a plate film (Bioexpress #T-2417-8). Applied Biosystems RT-PCR
machine was programmed to start at 20°C with a 30 second pause and end at 85°C, also with a 30 second
pause, and ramp at 3% between the initial and final temperature (45 minutes), collecting data at each
temperature point. The resultant unfolding curves were fit using a two-state model of protein unfolding to
provide values for the midpoint of thermal unfolding (Tm).
Acknowledgements
The author would like to thank Henrik Dohlman, Ph.D. and Daniel Isom, Ph.D., and members of the
Dohlman Lab for their continuing support and guidance in developing and completing this thesis. V.S.
acknowledges support from the National Science Foundation REU Site Award 1156840 to UNC-Chapel
Cdc42 GDP and GTP
S Bound
20 30 40 50 60
0.00 0.25 0.50 0.75 1.00
Full Length (GDP) CAAX Truncated (GDP)
Full Length (GTPS) CAAX Truncated (GTPS)
Figure 4. fQCR assay for yeast Cdc42.
Temperature (C)
N
o
rm
a
li
z
e
d
I
n
te
n
s
it
Correlation between T
mdetermined by fQCR and ThermoFluor
30 40 50 60 70
30 40 50 60 70
Figure 5. Linear regression between thermal stability
values obtained from two independent methods
Slope Y-intercept X-intercept 1/slope
0.9637 ± 0.1078 4.239 ± 5.329 -4.398 1.038
r² 0.8693
P value < 0.0001
ThermoFluor Assay Tm
fQC
R
T
References
1
Foit L., Morgan G.J., Kern M.J., Steimer L.R., von Hacht A.A., Titchmarsh J. Optimizing protein stability in vivo. Mol. Cell. 2009;36:861–871.
2
Foster BA, Coffey HA, Morin MJ, Rastinejad F. Pharmacological rescue of mutant p53 conformation and function. Science. 1999;286:2507–2510.
3
Talluri, S. Advances in the engineering of proteins for thermal stability. International Journal of Advanced Biotechnology and Research 2.1. (2011): 190-200.
4
Asial I, Cheng YX, Engman H, Dollhopf M, Wu B, Norlund P, Cornvik T. Engineering protein thermostability using a generic activity-independent biophysical screen inside the cell. Nature Communcations. 2013 Dec 19;4:2901.
5
Yu H, Huang H. Engineering proteins for thermostability through rigidifying flexible sites. Biotechnology Advances. 2013 Nov 6.
6
Wijma HJ, Floor RJ, Janssen DB. Structure- and sequence-analysis inspired engineering of proteins for enhanced thermostability. Current Opinions in Structural Biology. 2013 Aug;23(4):588-94.
7
Van Noort V, Bradatsch B, et al. Consistent mutational paths predict eukaryotic thermostability. BMC Evolutionary Biology. 2013 Jan 10;13:7
8
Wittinghofer A, Vetter IR (2011) Structure-function relationships of the G domain, a canonical switch motif. Annu Rev Biochem 80: 943–971.
9
Takai Y, Sasaki T, Matozaki T (2001) Small GTP-binding proteins. Physiol Rev 81: 153–208.
10
Rojas AM, Fuentes G, Rausell A, Valencia A. The Ras protein superfamily: evolutionary tree and role of conserved amino acids. J Cell Biol. 2012;196:189–201.
11Versele M, Lemaire K, Thevelien JM. (2001). Sex and sugar in yeast: two distinct GPCR systems. EMBO Reports. 2 (7):574-79.
12
Pierce KL, Premont RT, Lefkowitz RJ. Seven-transmembrane receptors. Nat Rev Mol Cell Biol. 2002;3:639–650.
13
Same as 12
14
Garcia-Ranea JA, Valencia A. Distribution and functional diversification of the ras superfamily in Saccharomyces cerevisiae. FEBS Letters 434 (1998): 219-225.
15
Takai Y, Sasaki T, Matozaki T. Small GTP-binding proteins. Physiological Reviews. 2001. 81(1): 153-208.
16
17
Freire, E. Thermal denaturation methods in the study of protein folding. Methods in Enzymology, Volume 259, Academic Press, Inc, 1995.
18
Isom DG, Vardy E, Oas TG, Hellinga HW. Picomole-scale characterization of protein stability and function by quantitative cysteine reactivity. Proc Natl Acad Sci USA. 2010;107:4908–4913.
19
Niesen FH, Berglund H, Vedadi M. The use of differential scanning fluorimetry to determine ligand interactions that promote protein stability. Nat. Protocols 2007 2(9): 2212-2221.
20
Garcia-Ranea JA, Valencia A. Distribution and functional diversification of the ras superfamily in Saccharomyces cerevisiae. FEBS Letters 434 (1998): 219-225.
21
Haun, RS; Serventi, IM; Moss, J. Rapid, reliable ligation-independent cloning of PCR products using modified plasmid vectors". BioTechniques 13 (4): 515–8 (1992).
22
Studier FW. Protein production by auto-induction in high density shaking cultures. Protein Expr. Purif. 41(1): 207-234 (2005)
23
Same as 13.
24
Hohn A, Konig J, Grune T. Protein oxidation in aging and the removal of oxidized proteins. J. Proteomics. 2013 Oct 30;92:132-59.
25
Berlett BS, Stadtman ER. Protein oxidation in aging, disease, and oxidative stress. J Biol Chem. 1997 Aug. 15;272(33):20313-6.
26
Hobbs GA, Bonini MG, Gunwardena HP, Chen X, Campell SL. Glutathiolated Ras: Characterization and implications for Ras activation. Free Radical Biology and Medicine. 2013 April. 57: 221-229.
27
Macarthur MW, Thorton JM. Influence of proline residues on protein conformation. J Mol Biol. 1991 Mar 20;218(2):397-412.
28
Trevino SR, Schaefer S, Scholtz JM, Pace CN. Increasing protein conformational stability by optimizing β-turn sequence. J Mol Biol. 2007 Oct 12;373(1):211-8.
29
Same as 7.
30
Morano KA, Grant CM, Moye-Rowley WS. The response to heat shock and oxidative stress in Saccharomyces cerevisiae. Genetics 190 (April 2012): 1157-1195.
31
Gibney PA, Lu C, Caudy AA, Hess DC, Botsein D. Yeast metabolic and signaling genes are required for heat-shock survival and have little overlap with the heat-induced genes. Proc Natl Acad Sci. 2013 Nov 12;110(46):E4393-402.
32
Vjestica A, Zhang D, Liu J, Oliferenko S. Hsp70-Hsp40 Chaperone complex functions in controlling polarized growth by repression Hsf1-driven health stress-associated transcription. PLOS Genetics 9(10): 1-17 (Oct 2013).
33
34
Maresova L, Vydareny T, Sychrova H. Comparison of the influence of small GTPases Arl1 and Ypt6 on yeast cells’ tolerance to various stress factors. FEMS Yeast Res 12 (2012): 332-340.
35
Hou J, Osterlund T, Liu Z, Petranovik D, Nielsen J. Heat shock response improves heterologous protein
secretion in Saccharomyces cerevisiae. Appl Microbiol Biotechnol (2013) 97:3559–3568
36
Harms MJ. Thornton JW. Analyzing protein structure and function using ancestral gene reconstruction. Curr Opinion Structural Biol. 2010 Jun;20(3):360-6.
37
Hanson-Smith V, Kolaczkowski B, Thornton JW. Robustness of ancestral sequence reconstruction to phylogenetic uncertainty. Mol Biol Evol. 2010 Sep;27(9):1988-99.
38
Bloom, J.D., Labthavikul, S.T., Otey, C.R., and Arnold, F.H. (2006). Protein stability promotes evolvability. Proc. Natl. Acad. Sci. USA 103, 5869–5874.
39
