proteins
STRUCTURE OFUNCTION OBIOINFORMATICSUsing a strategy based on the concept
of convergent evolution to identify
residue substitutions responsible
for thermal adaptation
Yeong-Shin Lin*
Institute of Bioinformatics, National Chiao Tung University, Hsinchu, Taiwan
INTRODUCTION
The nature of thermostable proteins has drawn much attention in the past few decades. Many protein sequence and structure features characteristic of thermophilic proteins were revealed by comparison to counterpart mesophilic homologues. One of the most noticeable features are differences in amino acid composition between the thermophilic and mesophilic proteins, especially charged versus polar (noncharged) residues.1–9 This difference was further found to be preferentially located at the protein sur-face,2,10–13 and the helices, especially the N-terminal resi-dues.11,14–16 Study showed that helices of thermophilic proteins appear to be more stable than those of the mesophilic homo-logues.17 Other possible indicators of thermostability are asym-metrical substitution patterns for certain amino acid pairs18; amino acid coupling patterns19; local structural entropy20; tighter hydrophobic packing6,21; folds with high levels of contact trace22,23; a decrease in the entropy of unfolding6; fewer free cys-teine amino acids except those involved in disulfide bridges and metal binding, or those inaccessible to the solvent6,24,25; and more side-chain-side-chain hydrogen bonds and ion pairs.4,12,16,21,26
However, despite the many factors described earlier that are related to thermostability of proteins, it is still not clear how to identify specific residues or fragments responsible for thermal adaptation, which may be useful for rational design27 of thermal proteins. Many factors important for protein thermostability have varying and, sometimes, even opposing contributions to protein thermostability. For example, the position of amino acid changes in the structure may be even more important than a global increase in charged residues.13,28 Moreover, each protein family seems to adopt its own strategy or mechanism to adapt to high or low temperatures.4,6,29–31 Most importantly, not all differences among the thermophiles and mesophiles may be attributable to protein thermostability; some of them may be due to phylogenetic
Grant sponsor: National Science Council.
*Correspondence to: Yeong-Shin Lin, Institute of Bioinformatics, National Chiao Tung University, Hsinchu 30068, Taiwan. E-mail: yslin@faculty.nctu.edu.tw.
Received 8 September 2007; Revised 24 January 2008; Accepted 15 February 2008 Published online 2 April 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.22049
ABSTRACT
Factors that are related to thermostability of pro-teins have been extensively studied in recent years, especially by comparing thermophiles and meso-philes. However, most of them are global characters. It is still not clear how to identify specific residues or fragments which may be more relevant to protein thermostability. Moreover, some of the differences among the thermophiles and mesophiles may be due to phylogenetic differences instead of thermal adap-tation. To resolve these problems, I adopted a strat-egy to identify residue substitutions evolved conver-gently in thermophiles or mesophiles. These residues may therefore be responsible for thermal adaptation. Four classes of genomes were utilized in this study, including thermophilic archaea, mesophilic archaea, thermophilic bacteria, and mesophilic bacteria. For most clusters of orthologous groups (COGs) with sequences from all of these four classes of genomes, I can identify specific residues or fragments that may potentially be responsible for thermal adaptation. Functional or structural constraints (represented as sequence conservation) were suggested to have higher impact on thermal adaption than secondary structure or solvent accessibility does. I further com-pared thermophilic archaea and mesophilic bacteria, and found that the most diverged fragments may not necessarily correspond to the thermostability-determining ones. The usual approach to compare thermophiles and mesophiles without considering phylogenetic relationships may roughly identify sequence features contributing to thermostability; however, to specifically identify residue substitutions responsible for thermal adaptation, one should take sequence evolution into consideration.
Proteins 2008; 73:53–62.
V
VC 2008 Wiley-Liss, Inc.
Key words: thermostability; thermophile; mesophile; phylogeny; orthologue; protein structure; amino acid composition; function prediction.
differences.16,17,28,32,33 Most thermophiles are archaea, while most bacteria are mesophiles. The differentiation of amino acid compositions among phyla34 suggests that it would be informative to account for phylogenetic rela-tionships. In this study, I adopted a strategy to identify convergent evolution that allows us to recognize residue substitutions shared among thermophiles or mesophiles in a cluster of orthologous groups, which may be respon-sible for thermal adaptation.
MATERIAL AND METHODS
Dataset
Clusters of Orthologous Groups of proteins (COGs) were obtained from http://www.ncbi.nlm.nih.gov/COG/.35 I selected 12 thermophilic archaea genomes (TA; Archaeo-globus fulgidus, Methanococcus jannaschii, Sulfolobus solfa-taricus, Pyrococcus horikoshii, Pyrococcus abyssi, Methano-pyrus kandleri AV19, Pyrobaculum aerophilum, Aeropyrum pernix, Methanothermobacter thermautotrophicus, Thermo-plasma acidophilum, ThermoThermo-plasma volcanium, Halobacte-rium sp. NRC-1); 1 mesophilic archaea genome (MA; Methanosarcina acetivorans str.C2A); 2 thermophilic bacteria genomes (TB; Thermotoga maritima, Aquifex aeolicus); and 7 mesophilic bacteria genomes (MB; Nostoc sp. PCC 7120, Synechocystis sp., Fusobacterium nucleatum, Trepo-nema pallidum, Borrelia burgdorferi, Chlamydia trachoma-tis, Chlamydophila pneumoniae CWL029). A COG may include more than one sequence (paralogue) from one genome. COGs with a PDB homologue and with at least one sequence from each of the above four classes were retained for further analyses.
Phylogenetic relationships
It is possible that distinct organisms may evolve similar traits independently as they both adapt to similar envi-ronments. This process is named as convergent evolution. In view of the idea that protein thermostability may evolve independently in archaea and bacteria, I used this concept to develop a strategy to identify residue substitu-tions responsible for thermal adaptation. The general phylogenetic relationships for archaea and bacteria are shown in Figure 1(A). For most homologous residues, species from the same lineage (archaea or bacteria) should share similar evolutionary history and should be clustered together [Fig. 1(B)]. However, for some resi-dues that are responsible for thermal adaptation, that is, they are under strong selection, species living in similar environments (thermophiles or mesophiles) may be clus-tered together due to convergent evolution [Fig. 1(C)]. For this reason, I will define two types of tree topologies to describe the evolutionary relationships: PR tree [the species tree representing true phylogenetic relationships; Fig. 1(B)] and CE tree [a fictitious tree clustering species
under convergent evolution together; Fig. 1(C)]. Al-though thermophilic or mesophilic origin of bacteria is still under debate,36–41 PR tree and CE tree topologies hold for either case.
Amino acid composition as a character state
To identify residue substitutions responsible for ther-mal adaptation, I subdivided the protein sequences into residues or sliding windows and analyzed which tree topology (PR tree or CE tree) is preferentially supported by the substitutions occurred in the regions studied. Note that, thermal adaptation may not easily be achieved by single amino acid substitution at a specific site; in contrast, differences in amino acid compositions have been suggested to be the most noticeable characters dif-ferentiated between thermophiles and mesophiles.1–9 I therefore examined the amino acid composition in a sliding window (the number of amino acids belonging to certain types, e.g., charged or polar) and coded this as the character state for each organism. Figure 1D shows the multiple sequences alignment for COG0020 as an example. The sliding window is on the original amino acid sequence; therefore gaps in the alignments have no influence and are not considered.
I first used, Fcp, the difference of the numbers of
charged (Asp, Glu, Lys and Arg) and polar noncharged (Asn, Gln, Ser and Thr) amino acids9 as a continuous character state for each organism to construct the parsi-mony tree (as described below). However, it should be noticed that Asp is not as preferred by thermophiles as the other three charged amino acids are and, in contrast, that Ala and Val are highly preferred by mesophiles and thermophiles, respectively.9 I therefore defined another
character state, Ftm, the number of Glu, Lys, Arg and Val
versus the number of Asn, Gln, Ser, Thr, and Ala. To reveal whether these results could be obtained by chance, I permuted the amino acid compositions from the
pre-vious two character states as a new one, Fpermuted, the
number of Asp, Asn, Ser and Val versus the number of
Glu, Gln, Thr, and Arg. I also used Faromatic, the number
of aromatic residues (Phe, Trp and Tyr), and subdivided
Ftminto Ftm-c (charged: Glu, Lys, and Arg), Ftm-p (polar:
Asn, Gln, Ser, and Thr) and Ftm-np (nonpolar: Val vs.
Ala) for extensive analyses.
To obtain an accurate measure of amino acid composi-tion, a sufficient window size is required; on the other hand, when the window size is large, the selection pres-sure on the fragment may be diluted, and the difference of selection pressures between fragments may therefore be blurred. These two issues should be balanced. I used 15 residues as the windows size in this study after pre-liminary examinations. Slightly increasing or decreasing the window size does not change the results.
Figure 1
The strategy based on the concept of convergent evolution developed in this study. (A) The general phylogenetic tree. Red and blue indicate thermophiles and mesophiles, respectively. TA, thermophilic archaea; MA, mesophilic archaea; TB, thermophilic bacteria; MB, mesophilic bacteria. (B) In the tree topology showing true phylogenetic relationships (PR tree, represented as cyan), archaea and bacteria are separately clustered. (C) In the fictitious tree topology clustering species under convergent evolution together (CE tree, represented as orange), thermophilics and mesophilics are separately clustered. (D) A partial alignment of amino acid sequences collected in COG0020, undecaprenyl pyrophosphate synthase. Only 9 out of 28 sequences are displayed in this figure, including one PDB structure (1JP3:A). The numbers represent the corresponding residue positions of 1JP3:A. The arrow indicates the aligned column position corresponding to the 161st residue of 1JP3:A. Amino acids located in a 15 aa-sliding window centered on the column residues are displayed with the gray shadow. (E) Using the difference of the numbers of charged and polar residues (Fcp) in the sliding window showing in Figure 1(D) as the character state (the numbers shown in the parentheses), the length of PR tree (the summation of the changes between the numbers) is calculated as 17 based on the parsimony principle, that is assigning 3 and 3, or 4 and 4 for the two internal nodes. This is the minimized tree length. (F) Similarly, the length of CE tree is calculated as 14 (with 4 and 2 for the two internal nodes). (G) A hypothetical distribution of tree lengths for trees randomly regenerated from the PR tree. The green area indicates the probability that a randomly regenerated tree has a tree length equal to or shorter than the length of CE tree (Psmall-CE). (H) For each aligned column of COG0020, the tree lengths for PR tree and CE tree are calculated using all the 27 sequences in COG0020 (excluding 1JP3:A), and represented as cyan and orange lines, respectively. The Psmall-CEvalue is represented as the green line. The column positions with Psmall-CE< 1022(thermal adaptation sites) are represented as yellow.
Calculation of the tree length
For each aligned column, the character state (of the sliding window centered on the column residue) for each protein sequence was assigned to each branch in the two models [PR tree and CE tree; Fig. 1(E,F)]. The tree length for both models for each aligned column can therefore be calculated by summing all character state changes between nodes on a tree. To simplify the calcula-tion, I assumed a topology where all species radiate (i.e., form a polytomy) from one of the two internal nodes. Parsimony was used to calculate the tree length, that is, the character states of the two internal nodes were recon-structed to minimize the tree length [e.g., Fig. 1(E,F)].
Assessing statistical significance
For an aligned column where CE tree has a smaller tree length than PR tree, residues surrounding the col-umn (in the sliding window) are probably under conver-gent evolution, which means they might be responsible for thermal adaptation. To clarify whether the small length of CE tree can be derived by chance, I assumed PR tree is the real tree topology and regenerated various trees from PR tree to construct a distribution of tree length for all possible trees. It should be noticed that, the difference between PR tree and CE tree is the switching between MA and TB sequences [Fig. 1(E,F)]. Assuming
there are n1 MA sequences out of m1 archaea sequences,
and n2 TB sequences out of m2 bacteria sequences, a
regenerated tree is performed by switching n1 archaea
sequences and n2 bacteria sequences from PR tree. The
total number of possible regenerated trees is Cm1
n13C
m2
n2.
The distribution of tree length for all possible regenerated trees was thus obtained [Fig. 1(G)], unless the number of
possible regenerated trees is larger than 105. In that case,
to save the computational time, a switching between n1
randomly selected archaea sequences and n2 randomly
selected bacteria sequences from PR tree was repeated
105 times. The probability (Psmall-CE) that a randomly
regenerated tree has a tree length equal to or smaller than the length of CE tree can thus be calculated [Fig. 1(G)]. I therefore defined the aligned column with its CE
tree length significantly small (Psmall-CE < 1022) as the
‘‘thermal adaptation site.’’ It should be noticed that not all substitutions causing thermal adaptation could be detected using this method, especially when the number of thermal adaptation substitutions is much smaller than
that of the background mutations (Psmall-CE may not be
significantly small). COGs with their total number of possible regenerated trees less than 100 were discarded because their sample sizes were too small to have the power to identify thermal adaptation sites. Figure 1(H) is
an example showing Psmall-CE and the lengths of PR tree
and CE tree using Fcp as the character state for each
aligned column of COG0020.
Protein structures
The secondary structure assignment and the solvent accessible surface areas (ACC) for each residue of the PDB homologues were obtained from DSSP (database of secondary structure assignments for all protein entries in the PDB; http://swift.cmbi.ru.nl/gv/dssp/).42 The DSSP method defines eight secondary structures according to
their hydrogen bonding patterns: a-helix (H), 310-helix
(G), p-helix (I), extended b-strand (E), isolated b-strand
(B), turn (T), bend (S), and coil (U). Relative solvent accessibility (RelACC) was the ACC for each residue di-vided by the maximum value of ACC for the amino acid (represented in percentage), which is estimated from a Gly-X-Gly extended tripeptide conformation.
The secondary structure and RelACC value for each residue of the PDB homologue for each COG were com-piled. Residues in a sliding window centered on a ther-mal adaptation site were annotated. For each category subdivided based on the secondary structures and RelACC values (secondary structures G, I, and B were ignored due to their extremely small sample sizes), the probability that a residue locates in a sliding window
centered on a thermal adaptation site (Pthermal) can thus
be estimated. Residues in a category with a high Pthermal
are suggested to have higher chance to be potentially re-sponsible, or at least to have higher chance to be detected as being responsible for thermal adaptation using the proposed method.
Sequence conservation
Considering the sample size, conservations in a COG were only estimated using TA and MB sequences. The conservation at one aligned column was estimated using
sequence entropy, S¼ P20j¼1pjlnpj, where pj is the
frequency of each amino acid in this column. The two conservation values of thermophilic archaea and
meso-philic bacteria, STA and SMB, were calculated separately
and then averaged as S for each aligned column in a
COG. Columns with S < 1 were defined as conserved
sites. The proportion of conserved sites (f) in a sliding window was thus calculated.
RESULTS AND DISCUSSION
Protein structure and thermal adaptation
I subdivided residues in all PDB homologues based on their secondary structure assignments and solvent accessi-bility to investigate the tendency of these residues poten-tially being responsible for thermal adaptation (see Fig. 2). Consistent with previous studies,2,10–16 the result shows that helix residues and exposed resides are
impor-tant for thermal adaptation using Fcp as the character
state to perform the analysis. The Pthermal value for
secondary structures (v2
test, P < 1024). Residues with
RelACC 0.40.6 also have their Pthermal value
signifi-cantly higher than each category with RelACC less than
0.4 (v2 test, P < 10216). However, the extremely exposed
residues are not the most preferred ones as expected.
When RelACC is increased from 0.40.6 to 0.60.8,
Pthermal is slightly decreased; when RelACC is further
increased to 0.81, Pthermal is significantly reduced (v2
test, P< 1026).
It was suggested that introducing stabilization muta-tions on protein surface may stabilize the reversibly unfolded state without creating volume interferences in the core.6 This study further indicates that mutations at the extremely exposed regions may only provide limited contributions to protein thermostability.
Using different amino acid compositions as the character states
Although the difference of the numbers of charged
and polar amino acids (Fcp) has long been recognized as
an important difference between thermophiles and meso-philes,1–9 Figure 3(A) shows that more thermal
adapta-tion sites were recognized using Ftmas the character state
instead. The relationships between protein structure and thermal adaptation are generally consistent with that
obtained using Fcp (see Fig. 2). To illustrate whether
these results could be derived by chance, Fpermuted was
used as a control character state. Each secondary
struc-ture and RelACC category shows background Pthermal
consistently, that is, 0.01 [Fig. 3(B)], because Psmall-CE <
1022 was used as the criteria to define thermal
adapta-tion site. Although aromatic clusters on protein surface was suggested to correlate with thermostability,43 using Faromatic as the character state also only obtains
back-ground Pthermal[Fig. 3(C)]. This result implies that either
the aromatic clustering is not a general feature for ther-mostability, or the clustering is not derived by increasing the number of aromatic amino acids but just special rearrangements.
I further subdivided Ftm into charged amino acids
(Ftm-c), polar amino acids (Ftm-p), and nonpolar amino
acids (Ftm-np) to compare their contributions to thermal
adaptation. The obtained results indicate that the increas-ing of charged residues [Fig. 3(D)] has much higher impact than the decreasing of polar residues [Fig. 3(E)] for thermostability. The charged residue substitutions
dominate at the helices (v2
test, P < 1028). It is likely that the increased charged residues usually locate around the thermal adaptation sites; in contrast, the replace-ments of polar residues may distribute along the whole protein more dispersedly. Except for the exposed helices, Pthermal values estimated using Ftm-np as the character state are only slightly higher than the background value [Fig. 3(F)], which suggests that Ala and Val substitutions may also tend to occur globally.
Sequence conservation and thermal adaptation
It has been proposed that stabilizing interactions in thermophilic proteins tend to locate in the less conserved areas of the protein.6 I subdivided aligned columns of each COG based on their sequence conservation, that is, the proportion of conserved sites (f) in the sliding win-dow centered on the column. It was found that when most residues in the local environment are conserved,
their Pthermal values are consistently low regard less their
secondary structures and solvent accessibility; however, they are still higher than the background value, 0.01 [Fig. 4(A)]. Because exposed residues are usually less con-served,44 residues with high RelACC have comparatively high variance due to their small sample size. Note that I used TA sequences and MB sequences independently to estimate the conservation. Although conserved mutations between thermophilic and mesophilic subfamilies (con-served intra subfamily but diverged between subfamilies) were suggested to provide key residue differences that are potentially related to increased thermostability,7 these regions should have high f values, and may therefore not necessarily relate to thermal adaptation based on results in this study.
When the number of nonconserved sites in the
win-dow increases, Pthermal increases, especially for helix or
moderately exposed residues [Fig. 4(B,C)]. When most sites in the window are nonconserved, each category
shows high Pthermal[Fig. 4(D)].
Most residues in a protein are under constraints to maintain protein function or structural stability. Muta-tions of residues under weaker constraints may therefore have higher chance to obtain extra thermostability, that
is, they are more evolvable. The differentiation of Pthermal
Figure 2
The relationships between secondary structure, solvent accessibility, and Pthermal revealed using Fcpas the character state. The error bars represent standard error. The gray dotted line indicates background Pthermal5 0.01.
among different secondary structures or regions with dif-ferent solvent accessibility is only effective under certain selection pressures. Note that although exposed residues
tend to be nonconserved and have high Pthermal,
noncon-served buried residues have significantly higher Pthermal
than conserved exposed residues (f 0.25 with RelACC
5 0 vs. f > 0.75 with RelACC >0.4, v2
test, P< 1028). Based on these results, I suggest that the chance that a residue is important for thermal adaptation correlates with protein secondary structure, solvent accessibility, and functional or structural constraints (represented as sequence conservation) independently, where functional or structural constraints may have the most significant impact.
Thermal adaptation versus phylogenetic differences
To investigate whether my strategy can distinguish spe-cific substitutions responsible for thermal adaptation from the phylogenetic differences, I used TA and MB sequences to calculate the differentiation between ther-mophiles and mesophiles. These two groups of species compose the majority of thermophiles and mesophiles, respectively. For each aligned column, the averaged
char-acter state values (e.g., Ftm) for TA sequences and MB
sequences were calculated separately. The absolute value of their difference, d, was thus used to represent the dif-ferentiation, which may therefore include the differences attributed to both thermostability and phylogeny.
Figure 3
The relationships between secondary structure, solvent accessibility, and Pthermalrevealed using (A) Ftm; (B) Fpermuted; (C) Faromatic; (D) Ftm-c; (E) Ftm-p; (F) Ftm-npas the character states. The error bars represent standard error. The gray dotted line indicates background Pthermal5 0.01.
Residues with low Psmall-CE values (thermal adaptation sites) usually have high d values (they may be responsible for thermal adaptation); in contrast, residues with high d
values may not necessarily have their Psmall-CE values
sig-nificantly low (some of them may be attributed to phylo-genetic differences). The result shows that residues with
the most significant Psmall-CE values and residues with the
highest d values can be nonoverlapped (e.g., COG0057 in Fig. 5). COG0057 is a protein family of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Five fragments in the 3D structure differ significantly between archaea and bacteria.45 Fragment I (residues 39-47 on 1B7G:O) is an
a-helix corresponding to the pick with very low Psmall-CE
values (residues 41-49). Fragment II (56–78) includes
two a-helices existing in archaea but not in bacteria;
where parts of this region are also recognized as thermal adaptation sites (52–54, 62, and 65). These two regions may correlate with thermal adaptation based on results in this study. Fragment III (169–178) is the so-called
S-loop. A part of the S-loop and its upstream b-sheet have
very high d values, but with insignificant Psmall-CE values.
(160–173) Similar situation is applied for the down-stream region (273–286) of fragment IV (257–268) which
includes another extra a-helix in archaea structure. The
differentiation of these two regions may therefore be due to phylogeny. Fragment V (320–332) is the C terminal a-helix only existing in archaea, and is not analyzed here.
It is likely that those phylogenetic differences in GAPDH are mainly due to a relocation of the active-site residues within the enzyme’s catalytic domain between archaea and bacteria.45 Most regions neighbor to the active-site residues (Ser138, Cys139, Asn140, Arg166, Arg167, His192, His193 and His219) show such phyloge-netic differences (136–142, 160–173, corresponding to the upstream of fragment III, and 188–191). The binding-site
residues of the NADP1 adenosine 20 phosphate (Lys33,
Thr34 and Ser35) also have high d values and insignifi-cant Psmall-CE values.
Using the whole dataset to reveal the relationship
between the significance of Psmall-CEand the difference, d,
we can find that many aligned columns show either high
significance of Psmall-CE or high difference, d (see Fig. 6),
although a positive correlation between them still can be
recognized (R5 0.503). This result suggests that
compar-ing thermophiles and mesophiles without considercompar-ing phylogenetic relationships can roughly identify the char-acters contributing to thermostability. However, to specif-ically identify residue substitutions responsible for
ther-Figure 4
The relationships between secondary structure, solvent accessibility, and Pthermalrevealed using Ftmas the character state with (A) f> 0.75; (B) 0.75 f > 0.5; (C) 0.5 f > 0.25 (D) f 0.25; where f is the proportion of conserved sites in a sliding window. The error bars represent standard error. The gray dotted line indicates background Pthermal5 0.01.
mal adaptation, using the strategy proposed here may be more convincing.
SUMMARY
The strategy based on the concept of convergent evolu-tion developed in this study successfully identified partic-ular fragments potentially being responsible for thermal
adaptation for most COGs (with at least couple sequen-ces from each of the four classes) despite that the identi-fication power of this strategy may be limited when the number of sequences in a COG is small. Although differ-ent organisms may potdiffer-entially utilize differdiffer-ent mecha-nisms to adapt thermal environments,23,28,31 the mu-tual features shared among thermophiles or mesophiles still can be identified through this strategy. Unexpected gene exchange between members of archaea and bacteria were reported.46–49 However, that happens infrequently, and most proteome should be developed to adapt the environment by the organisms themselves.23 Neverthe-less, such gene exchange may misinterpret most regions in a protein as thermal adaptation sites, which was not seen in the analyzed dataset. The thermal adaptation sites identified here should therefore be ideal candidates for rational design of thermal proteins.
For a given protein without orthologous information, the general trends summarized from this study can pro-vide some suggestions to improve thermostability, for example the addition of charged residues at the noncon-served regions, or at the weakly connoncon-served and moder-ately exposed helices. However, even for these regions, the probability of a residue to locate surrounding thermal adaptation sites (being responsible for thermal adapta-tion) is still hardly more than 10%. Before a more gen-eral feature is discovered, using orthologous sequences and the strategy proposed in this study to identify ther-mal adaptation sites is still the most reliable way.
It is worth to notice that this strategy is not restricted to identify thermal adaptation. Residues responsible for protein functions or features evolved independently mul-tiple times in different organisms could also be identified
Figure 5
Using Ftmas the character state to analyze COG0057 (similar results were obtained using Fcpas the character state, data not shown). (A) The tree lengths for PR tree and CE tree represented as cyan and orange lines, respectively. (B) The Psmall-CEvalue. (C) The absolute value of the difference (d) between the mean of TA sequences and the mean of MB sequences. (D) The 3D structure of 1B7G:O. Thermal adaptation sites (Psmall-CE< 1022) are denoted as red, while residues with d> 5 are denoted as blue. Fragments I–IV are also indicated.
Figure 6
The relationships between Psmall-CE(the probability that a randomly regenerated tree has a tree length equal to or smaller than the length of CE tree) and the absolute difference value (d) between the mean of TA sequences and the mean of MB sequences using Ftmas the character state.
using it. With more and more genomes been resolved, this convergent evolution-aided strategy should become a useful tool for proteomic analyses.
ACKNOWLEDGMENTS
The author thank Joshua Rest and the anonymous reviewer for the valuable comments, and the Structural Bioinformatics Core Facility at National Chiao Tung Uni-versity for the hardware and software support.
REFERENCES
1. Haney PJ, Badger JH, Buldak GL, Reich CI, Woese CR, Olsen GJ. Thermal adaptation analyzed by comparison of protein sequences from mesophilic and extremely thermophilic Methanococcus species. Proc Natl Acad Sci USA 1999;96:3578–3583.
2. Cambillau C, Claverie J-M. Structural and genomic correlates of hyperthermostability. J Biol Chem 2000;275:32383–32386.
3. Chakravarty S, Varadarajan R. Elucidation of determinants of protein stability through genome sequence analysis. FEBS Lett 2000; 470:65–69.
4. Szilagyi A, Zavodszky P. Structural differences between mesophilic, moderately thermophilic and extremely thermophilic protein sub-units: results of a comprehensive survey. Structure 2000;8:493–504. 5. Kreil DP, Ouzounis CA. Identification of thermophilic species by
the amino acid compositions deduced from their genomes. Nucleic Acids Res 2001;29:1608–1615.
6. Vieille C, Zeikus G. Hyperthermophilic enzymes: sources, uses, and molecular mechanisms for thermostability. Microbiol Mol Biol Rev 2001;65:1–43.
7. La D, Silver M, Edgar RC, Livesay DR. Using motif-based methods in multiple genome analyses: a case study comparing orthologous meso-philic and thermomeso-philic proteins. Biochemistry 2003;42:8988–8998. 8. Nakashima H, Fukuchi S, Nishikawa K. Compositional changes in
RNA, DNA and proteins for bacterial adaptation to higher and lower temperatures J Biochem 2003;133:507–513.
9. Suhre K, Claverie J-M. Genomic correlates of hyperthermostability, an update. J Biol Chem 2003;278:17198–17202.
10. Fukuchi S, Nishikawa K. Protein surface amino acid compositions distinctively differ between thermophilic and mesophilic bacteria. J Mol Biol 2001;309:835–843.
11. Chakravarty S, Varadarajan R. Elucidation of factors responsible for enhanced thermal stability of proteins: a structural genomics based study. Biochemistry 2002;41:8152–8161.
12. Alsop E, Silver M, Livesay DR. Optimized electrostatic surfaces par-allel increased thermostability: a structural bioinformatic analysis. Protein Eng 2003;16:871–874.
13. Thorvaldsen S, Hjerde E, Fenton C, Willassen NP. Molecular char-acterization of cold adaptation based on ortholog protein sequences from Vibrionaceae species. Extremophiles 2007;11:719–732. 14. Querol E, Perez-Pons JA, Mozo-Villarias A. Analysis of protein
con-formational characteristics related to thermostability. Protein Eng 1996;9:265–271.
15. Das R, Gerstein M. The stability of thermophilic proteins: a study based on comprehensive genome comparison. Funct Integr Genom 2000;1:76–88.
16. Kumar S, Tsai C-J, Nussinov R. Factors enhancing protein thermo-stability. Protein Eng 2000;13:179–191.
17. Facchiano AM, Colonna G, Ragone R. Helix stabilizing factors and stabilization of thermophilic proteins: an X-ray based study. Protein Eng 1998;11:753–760.
18. McDonald JH, Grasso AM, Rejto LK. Patterns of temperature adap-tation in proteins from Methanococcus and Bacillus. Mol Biol Evol 1999;16:1785–1790.
19. Liang H-K, Huang C-M, Ko M-T, Hwang J-K. Amino acid coupling patterns in thermophilic proteins. Proteins 2005;59:58–63. 20. Chan C-H, Liang H-K, Hsiao N-W, Ko M-T, Lyu P-C, Hwang J-K.
Relationship between local structural entropy and protein thermo-stability. Proteins 2004;57:684–691.
21. Hasegawa J, Shimahara H, Mizutani M, Uchiyama S, Arai H, Ishii M, Kobayashi Y, Ferguson SJ, Sambongi Y, Igarashi Y. Stabilization of Pseudomonas aeruginosa cytochrome c551 by systematic amino
acid substitutions based on the structure of thermophilic Hydroge-nobacter thermophilus cytochrome c552. J Biol Chem 1999;274:37533–
37537.
22. England JL, Shakhnovich BE, Shakhnovich EI. Natural selection of more designable folds: a mechanism for thermophilic adaptation. Proc Natl Acad Sci USA 2003;100:8727–8731.
23. Berezovsky IN, Shakhnovich EI. Physics and evolution of thermophilic adaptation. Proc Natl Acad Sci USA 2005;102:12742– 12747.
24. Rosato V, Pucello N, Giuliano G. Evidence for cysteine clustering in thermophilic proteomes. Trends Genet 2002;18:278–281.
25. Schneider D, Liu Y, Gerstein M, Engelman DM. Thermostability of membrane protein helix-helix interaction elucidated by statistical analysis. FEBS Lett 2002;532:231–236.
26. Vogt G, Woell S, Argos P. Protein thermal stability, hydrogen bonds, and ion pairs. J Mol Biol 1997;269:631–643.
27. Hellinga HW. Rational protein design: combining theory and experiment. Proc Natl Acad Sci USA 1997;94:10015–10017. 28. Robinson-Rechavi M, Alibes A, Godzik A. Contribution of
electro-static interactions, compactness and quaternary structure to protein thermostability: lessons from structural genomics of Thermotoga maritima.. J Mol Biol 2006;356:547–557.
29. Jaenicke R. Stability and stabilization of globular proteins in solu-tion. J Biotechnol 2000;79:193–203.
30. Gianese G, Bossa F, Pascarella S. Comparative structural analysis of psychrophilic and meso- and thermophilic enzymes. Proteins 2002; 47:236–249.
31. Beeby M, O’Connor BD, Ryttersgaard CB, Daniel R, Perry LJ, Yeates TO. The genomics of disulfide bonding and protein stabiliza-tion in thermophiles PLoS Biol 2005;3:e309.
32. Arnold FH, Wintrode PL, Miyazaki K, Gershenson A. How enzymes adapt: lessons from directed evolution. Trends Biochem Sci 2001;26: 100–106.
33. McDonald JH. Patterns of temperature adaptation in proteins from the bacteria Deinococcus radiodurans and Thermus thermophilus. Mol Biol Evol 2001;18:741–749.
34. Pe’er I, Felder CE, Man O, Silman I, Sussman JL, Beckmann JS. Proteomic signatures: amino acid and oligopeptide compositions differentiate among phyla. Proteins 2004;54:20–40.
35. Tatusov RL, Koonin EV, Lipman DJ. A genomic perspective on pro-tein families. Science 1997;278:631–637.
36. Barns SM, Delwiche CF, Palmer JD, Pace NR. Perspectives on arch-aeal diversity, thermophily and monophyly from environmental rRNA sequences. Proc Natl Acad Sci USA 1996;93:9188–9193. 37. Galtier N, Tourasse N, Gouy M. A nonhyperthermophilic common
ancestor to extant life forms. Science 1999;283:220–221.
38. Bocchetta M, Gribaldo S, Sanangelantoni A, Cammarano P. Phylo-genetic depth of the bacterial genera Aquifex and Thermotoga inferred from analysis of ribosomal protein, elongation factor, and RNA polymerase subunit sequences. J Mol Evol 2000;50:366–380. 39. Daubin V, Gouy M, Perriere G. A phylogenomic approach to
bacte-rial phylogeny: evidence of a core of genes sharing a common his-tory. Genome Res 2002;12:1080–1090.
40. Gaucher EA, Thomson JM, Burgan MF, Benner SA. Inferring the palaeoenvironment of ancient bacteria on the basis of resurrected proteins. Nature 2003;425:285–288.
41. Iwabata H, Watanabe K, Ohkuri T, Yokobori S-I, Yamagishi A. Thermostability of ancestral mutants of Caldococcus noboribetus iso-citrate dehydrogenase. FEMS Microbiol Lett 2005;243:393–398.
42. Kabsch W, Sander C. Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 1983;22:2577–2637.
43. Kannan N, Vishveshwara S. Aromatic clusters: a determinant of thermal stability of thermophilic proteins. Protein Eng 2000;13:753– 761.
44. Bustamante CD, Townsend JP, Hartl DL. Solvent accessibility and purifying selection within proteins of Escherichia coli and Salmo-nella enterica. Mol Biol Evol 2000;17:301–308.
45. Isupov MN, Fleming TM, Dalby AR, Crowhurst GS, Bourne PC, Littlechild JA. Crystal structure of the glyceraldehyde-3-phosphate dehydrogenase from the hyperthermophilic archaeon Sulfolobus sol-fataricus. J Mol Biol 1999;291:651–660.
46. Ibba M, Bono JL, Rosa PA, Soll D. Archaeal-type lysyl-tRNA syn-thetase in the Lyme disease spirochete Borrelia burgdorferi. Proc Natl Acad Sci USA 1997;94:14383–14388.
47. Klenk H-P, Clayton RA, Tomb J-F, White O, Nelson KE, Ketchum KA, Dodson RJ, Gwinn M, Hickey EK, Peterson JD, Richardson DL, Kerlavage AR, Graham DE, Kyrpides NC, Venter JC. The com-plete genome sequence of the hyperthermophilic, sulphate-reducing archaeon Archaeoglobus fulgidus. Nature 1997;390:364–370. 48. Deckert G, Warren PV, Gaasterland T, Young WG, Lenox AL, Graham
DE, Overbeek R, Snead MA, Keller M, Aujay M, Huber R, Feldman RA, Short JM, Olsen GJ, Swanson RV. The complete genome of the hyper-thermophilic bacterium Aquifex aeolicus. Nature 1998;392:353–358. 49. Pennisi E. Genome data shake tree of life. Science 1998;280:672–674.
