Intimate Evolution of Proteins

Discerning the significant relations that exist within and among genome sequences is a major step toward the modeling of biopolymer evolution. Here we report the systematic analysis of the atomic composition of proteins encoded by organisms representative of each kingdoms. Protein atomic contents are shown to vary largely among species, the larger variations being observed for the main architectural component of proteins, the carbon atom. These variations apply to the bulk proteins as well as to subsets of ortholog proteins. A pronounced correlation between proteome carbon content and genome base composition is further evidenced, with high G+C genome content being related to low protein carbon content. The generation of random proteomes and the examination of the canonical genetic code provide arguments for the hypothesis that natural selection might have driven genome base composition. Discerning the significant relations that exist within and among genome sequences is a major step toward the modeling of biopolymer evolution. Here we report the systematic analysis of the atomic composition of proteins encoded by organisms representative of each kingdoms. Protein atomic contents are shown to vary largely among species, the larger variations being observed for the main architectural component of proteins, the carbon atom. These variations apply to the bulk proteins as well as to subsets of ortholog proteins. A pronounced correlation between proteome carbon content and genome base composition is further evidenced, with high G+C genome content being related to low protein carbon content. The generation of random proteomes and the examination of the canonical genetic code provide arguments for the hypothesis that natural selection might have driven genome base composition. Comparative analyses of complete genome sequences were anticipated to reveal the molecular bases of biodiversity as well as to increase our comprehension of the constraints that shaped protein composition and structure during the natural history of living organisms. However, whereas comparative genomics highlighted genome structure plasticity and strengthened the role of lateral gene transfer during the natural history of living organisms (1Ochman H. Lawrence J.G. Groisman E.A. Nature. 2000; 405: 299-304Crossref PubMed Scopus (2658) Google Scholar, 2Koonin E.V. Aravind L. Kondrashov A.S. Cell. 2000; 101: 573-576Abstract Full Text Full Text PDF PubMed Scopus (201) Google Scholar), less progress was accomplished in understanding adaptive evolution at a molecular level and how it contributes to changes in proteins. Studies devoted to protein evolution mostly use comparative analyses of protein primary sequences and attempt to identify which rules govern the conservations, substitutions, and deletions of amino acids that are observed between proteins (3Suekoa N. Cold Spring Harbor Symp. Quant. Biol. 1961; 26: 35-43Crossref PubMed Scopus (74) Google Scholar, 4Singer G.A.C. Hickey D.A. Mol. Biol. Evol. 2000; 17: 1581-1588Crossref PubMed Scopus (192) Google Scholar, 5Thorne J.L. Curr. Opin. Genet. Dev. 2000; 10: 602-605Crossref PubMed Scopus (56) Google Scholar). By focusing on amino acid composition, such studies do not address the possibility that evolution of proteins, and more generally of biopolymers, might have been shaped at a more intimate level: the atomic level. It is worthwhile to note that, among the constitutive elements of proteins, carbon, nitrogen, and sulfur atoms are subject to geochemical cycles at the surface of the earth (6Brock D. Madigan M.T. Biology of Microorganisms. 6th Ed. Prentice-Hall International, London1991: 634-648Google Scholar). As a consequence, large fluctuations of both form and abundance of elemental components of proteins are occurring in natural habitats. One would thus expect that specific molecular mechanisms might have evolved to allow living organisms to respond to the elemental variations of the environment that they inevitably faced during their natural history. These adaptative processes undoubtedly left traces in the chemical composition of biopolymers, and it could be anticipated that the advent of complete genomic sequences, combined with the use of straightforward statistical methods, would open the way to retrieve such imprints in macromolecule sequences. To approach this question, we previously investigated the atomic composition of protein subsets from both the bacteria Escherichia coli and the eukaryote Saccharomyces cerevisiae (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar). This study allowed us to discern that, in both organisms, the atomic composition of several protein families has been constrained in response to specific selective pressures. A pronounced correlation between atomic compositions and metabolic roles were deciphered in enzymes involved in essential nutriment assimilation in these microbial organisms. For instance, carbon assimilatory enzymes were found to contain significantly less carbon atoms than the total proteins of E. coli and S. cerevisiae (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar). Also, sulfur assimilatory enzymes of both organisms were shown to comprise less sulfur atoms that the bulk proteins. Such impoverishments of sulfur and carbon assimilatory enzymes in their respective element were interpreted as an imprint of variations in the nutritional availability of these elements during the natural history of E. coli and S. cerevisiae. These results thus suggested that the evolution of proteins could have been subjected to ecological constraints more than previously thought. In the next step, we wanted to know how broad the fluctuations of elemental protein composition might be, and more specifically, whether the atomic contents of proteins could display substantial variations among different organisms. Here we report the first results of the systematic analysis of the atomic contents of the proteins encoded by organisms whose genome has been entirely sequenced. Complete protein data sets from 46 organisms, representative of the bacteria, archaea, and eukaryote kingdoms were compiled and analyzed using quantile distributions. The results show that the atomic composition of proteins, especially their carbon content, widely differs among species and may differ more between species than within proteomes. A strongly significant correlation between the average carbon content of proteomes and the DNA base composition of the genomes was moreover uncovered. Randomization methods and analysis of the genetic code further demonstrate that the extant genetic code relates DNA G+C usage with the carbon content of proteins. Organisms Used for This Study—Five eukaryotic, ten archaeal, and thirty-one bacterial proteomes were downloaded from the GenBank™ FTP site (ftp.ncbi.nih.gov/genbank/genomes/). All of the proteins containing one or more undetermined amino acid (“X”) were discarded, except for Arabidopsis thaliana proteins in which only the X amino acids were discarded. The coding G+C contents were retrieved from Kazusa's Codon Usage Data Base (www.kazusa.or.jp/codon/), which is an extended WWW version of CUTG (Codon Usage Tabulated from GenBank™). The prokaryotic total G+C contents were found in the literature. The considered organisms are the following: A. thaliana (25,531 proteins; G+C coding = 44.17%), Caenorhabditis elegans (17,074 proteins; G+C coding = 42.59%), Drosophila melanogaster (14,335 proteins; G+C coding = 53,99%), Homo sapiens (24,493 proteins; G+C coding = 52.51%), S. cerevisiae (6,330 proteins; G+C coding = 39.72%), Aeropyrum pernix (2,694 proteins; G+C coding = 57.49%; G+C total = 56.3%), Archaeglobus fulgidus (2,420 proteins; G+C coding = 49.37%; G+C total = 48.5%), Halobacterium sp. (2,058 proteins; G+C coding = 65.26%; G+C total = 65.9%), Methanococcus jannaschii (1,773 proteins; G+C coding = 31.94%; G+C total = 31.09%), Methanobacterium thermoautotrophicum (1,869 proteins; G+C total = 49,50%), Pyrococcus abyssi (1,765 proteins; G+C coding = 45.16%; G+C total = 44.7%), Pyrococcus horikoshii (2,038 proteins; G+C coding = 42.32%; G+C total = 42.0%), Sulfolobus solfataricus (2,977 proteins; G+C coding = 36.30%), Thermoplasma volcanium (1,499 proteins), Thermoplasma acidophilum (1,478 proteins; G+C coding = 47.38%; G+C total = 46%), Aquifex aeolicus (1,521 proteins; G+C coding = 53.58%; G+C total = 43.4%), Bacillus halodurans (4,066 proteins; G+C coding = 44.33%; G+C total = 43.7%), Bacillus subtilis (4,100 proteins; G+C coding = 44.35%; G+C total = 43.5%), Borrelia burgdorferi (1,606 proteins; G+C coding = 29.33%; G+C total = 28.6%), Campylobacter jejuni (1,654 proteins; G+C coding = 30.98%; G+C total = 30.6%), Caulobacter crescentus (3,767 proteins; G+C coding = 67.42%; G+C total = 67.2%), Chlamidia muridarum (898 proteins; G+C coding = 40.64%; G+C total = 40.3%), Chlamidophila pneumoniae (1,092 proteins; G+C coding = 41.29%; G+C total = 40.6%), Deinococcus radiodurans (3,089 proteins; G+C coding = 67.24%; G+C total = 66.6%), E. coli (4,289 proteins; G+C coding = 51.11%; G+C total = 50.8%), Haemophilus influenzae (1,666 proteins; G+C coding = 37.52%; G+C total = 38.1%), Helicobacter pylori (1,565 proteins; G+C coding = 40.45%; G+C total = 39%), Lactococcus lactis (2,266 proteins; G+C coding = 35.69%), Mesorhizobium loti (6,752 proteins; G+C coding = 57.29%; G+C total = 62.47%), Mycobacterium leprae (1,605 proteins; G+C coding = 59.63%; G+C total = 57.8%), Mycobacterium tuberculosis (4,137 proteins; G+C coding = 65.79%; G+C total = 65.6%), Mycoplasma genitalium (489 proteins; G+C coding = 31.74%), Mycoplasma pneumoniae (689 proteins; G+C coding = 41.06%; G+C total = 40.0%), Mycoplasma pulmonis (782 proteins; G+C coding = 26.38%), Neisseira meningitidis (2,081 proteins; G+C coding = 51.49%; G+C total = 51.5%), Pasteurella meningitidis (2,014 proteins; G+C coding = 37.81%; G+C total = 40.4%), Pseudomonas aeruginosa (5,565 proteins; G+C coding = 66.63%; G+C total = 66.6%), Rickettsia prowazekii (834 proteins; G+C coding = 30.60%; G+C total = 29.1%), Staphylococcus aureus (2,594 proteins; G+C coding = 32.90%), Streptococcus proteins; G+C coding = G+C total = sp. proteins; G+C coding = G+C total = proteins; G+C coding = G+C total = proteins; G+C coding = G+C total = proteins; G+C coding = G+C total = proteins; G+C coding = G+C total = and proteins; G+C coding = G+C total = of proteins within different of proteins of proteins. genomes were retrieved The of to proteins and and to proteins and to these were to species or to the protein carbon were and the quantile were as for proteomes. of proteins. protein were first using the natural from random different G+C For each G+C content different protein were and their carbon quantile were protein is of random proteins, the of which are among from to to the of protein The were with the constraints of = and = is the of the base X of the and is to the wanted G+C content. The were discarded. To random protein with genetic were first the of and D. Mol. Evol. PubMed Scopus Google Scholar), which to one of the amino acids to each of coding for one amino acid in and to the each genomes coding for proteins of were as and G+C contents of and The correlation between G+C content and the carbon were for each The quantile of the of the were for of the to a correlation of DNA were from the DNA of an E. coli gene coding for a protein of L. one of the DNA was and the base was to be one different in the DNA the to be with each of the base was that would the carbon content of the encoded protein were The of the of the DNA and the carbon content of the protein were of of Protein in the atomic content of proteins encoded by entirely we quantile methods and S. P. Protein PubMed Scopus Google Scholar, S. Science. PubMed Scopus Google Scholar). For each the quantile of carbon, nitrogen, and sulfur within the of the encoded proteins were The quantile is the of proteins from the in which the of atoms is at The and the quantile to the quantile are major statistical of atomic contents S. Science. PubMed Scopus Google Scholar). As a first of how the atomic composition of proteins on a we the related of a bacteria that the well subtilis and a large of entirely genomes PubMed Google Scholar). The results of the analysis with different bacteria are in As previously for E. coli and S. cerevisiae (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar), the quantile that, for each the carbon, nitrogen, and sulfur contents of proteins that are the results show that, on the the atomic quantile are among whereas quantile for carbon, and to a for display substantial variations between species, the quantile are largely in of the analyzed bacterial The statistical of the observed between carbon be by using A that, for instance, the observed between the halodurans and S. aureus carbon are significant and not to by A hypothesis for the results in would have been that the of the quantile among species the of the of the considered within the canonical amino The of the nitrogen, not of the sulfur quantile rules this sulfur protein contents vary between species, whereas amino acids contain at one sulfur atom. In protein contents are or atoms are found in the of canonical amino the results in that, at the atomic the composition of the proteins encoded by each genome from one to and at a level the plasticity of protein atomic contents previously for proteins D. P. Nature. PubMed Scopus Google and microbial assimilatory enzymes (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar, L. P. A. Mol. Cell. Full Text Full Text PDF PubMed Scopus (201) Google Scholar). Protein between than within the variations in protein atomic content were observed for carbon, which is the main architectural component of proteins, we our analysis on this analyses be that the variations in protein carbon content observed for the for bacterial species as well as for the and eukaryote kingdoms. the carbon for kingdoms to be with and or only between the at the of the distributions. This that the of protein carbon contents largely differ between species, the of each is for of the species the of the carbon quantile is larger than the of each quantile the carbon of each species is than would be the of a species of the proteins of organisms not the atomic composition of proteins to be more different between species than within each The of the carbon quantile observed for both and bacterial species further that at each level of protein carbon content, the quantile are between the of the observed variations in protein carbon content their to have the of the proteins encoded by each To further this the carbon contents of several ortholog proteins by and bacterial species were of proteins within different E.V. Science. PubMed Scopus Google were retrieved from the data and the carbon contents were and with first that the carbon contents of ortholog proteins vary among species as the carbon content of proteomes not only data are on the variations of protein abundance at the level of organisms, we to the of protein abundance variations in our the study of S. A. N. Nature. PubMed Scopus Google Scholar), which the analysis of the abundance of proteins of S. and we that of proteins are in their were as one of the proteins of S. cerevisiae and representative in several bacteria or The proteins to these for to of the proteins in S. cerevisiae. of carbon usage of proteins to these were for each The results show that the carbon quantile of ortholog proteins are among species as the carbon quantile for the proteins of each In the quantile of ortholog and total proteins are among the carbon content of ortholog proteins was as by quantile were for each of the in of the species and with the quantile of the carbon content of proteomes. As in the of the quantile of carbon contents of ortholog proteins by are to the of the quantile of the carbon of total proteins. This that, within a the carbon content of ortholog proteins among species as the carbon content of proteomes. the results of the analysis of carbon contents that carbon content of ortholog proteins vary among species as the bulk proteins, of their These results are further the in protein carbon contents are reveal that the of carbon atoms for protein may differ from more than between This that, on the from to carbon atoms on to the of a protein of that, to the of and J.L. Escherichia coli and and for D. Scholar), such a may for more than of the carbon atoms to a canonical bacterial such as E. to Base the results that protein carbon content from constraints that were to on and of proteins and that are generally as the primary of protein evolution Google Scholar, Mol. Biol. Evol. 2000; 17: PubMed Scopus Google Scholar). thus for a for these atomic composition first to whether variations of protein carbon contents from the or of amino The amino acid contents of the proteins of organisms protein carbon contents were retrieved and This with that organisms with high protein carbon contents on more and organisms with low protein carbon content proteins in and The and the is that protein atomic variations from variations in the usage of a large of amino of the analyzed species were next and not a for the of the However, we that the of the carbon was related to and This us at the composition of genomic which both large and low variations as the protein carbon DNA G+C content is to vary from to between species, with related species generally DNA G+C contents A. S. S. A. PubMed Scopus Google Scholar). thus whether the average protein carbon content might be to DNA G+C content. As shown in a correlation between the genome composition and the of protein carbon contents was uncovered. The results are with the of the carbon quantile being with both the G+C content of the coding DNA and the G+C content of the total DNA In were for each of the bacteria, archaea, and that this for In is correlation between protein sulfur content and genome composition, protein sulfur widely vary among species To further that the correlation observed between the quantile of the carbon and the G+C content to the proteome and was not a of carbon usage in proteins or protein we the quantile between both the and the and the G+C DNA content were and as for the carbon content, significant were found at both we the carbon content of ortholog proteins to several and their to the G+C content of the genomes that As shown in the between low protein carbon content and high DNA G+C content as well for ortholog proteins. All of these results provide that the carbon content of proteins is strongly related to the base composition of their among organisms different DNA G+C of the of carbon for the of several ortholog protein protein coli crescentus = tuberculosis = leprae = coli = pneumoniae = genitalium = in a and Base in the proteome atomic composition is with genome base composition that this is in the structure of the canonical genetic To the hypothesis that the genetic code relates low protein carbon content to high DNA G+C content, we first a the natural random protein were from random with different G+C contents and the protein carbon quantile were This that the random protein from random the G+C content display the carbon quantile the random protein carbon are to the G+C content of the As shown in we found a correlation between composition of random genomes and the quantile of carbon contents of the proteins. the protein sulfur content of random proteomes do not with the base composition of random not as observed for organisms. To further how the extant genetic code relates low proteome carbon content to high G+C genome content, we next within the of the that one amino acid with and we how that or increase amino acid carbon content in the G+C content of demonstrate that among the that amino acid carbon content, do increase G+C content, do not the G+C content, and only the G+C content the was for amino acid carbon content with and the G+C content, Randomization show how the of the carbon content of a protein in the G+C content of gene we whether the for which we is specific to the extant genetic analyzed random protein with were to the of and D. Mol. Evol. PubMed Scopus Google Scholar). each genomes coding for proteins of and different G+C contents were The correlation between G+C content and the carbon were for each The quantile of the of the were for of the to a correlation of As shown in the random of the amino acids to the sets observed in the canonical genetic code show that less than of allow the protein carbon content to be with DNA G+C content and to vary with an or larger than is observed with the canonical genetic the of low protein carbon content to high G+C DNA content to be a by only genetic that could be from the sets found in the canonical This study the plasticity of the elemental composition of proteins, a previously of these large and between protein atomic compositions are observed among This for sulfur and carbon, the larger variations being for the atom. protein carbon contents differ more between species than within these are not to found in specific protein apply to the proteomes. The carbon contents of ortholog proteins vary as the carbon contents of total proteins, and these variations are observed for proteins as This that constraints different from on and during the evolution of protein atomic These results at a level the of protein atomic contents for several subsets of enzymes such as proteins D. P. Nature. PubMed Scopus Google and microbial assimilatory enzymes (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar). Such a protein evolution of the studies devoted to protein evolution on amino acid composition. protein carbon content do not in of one or amino acids the usage of of It is worthwhile to note that such protein carbon are of they could for more than of the carbon to a microbial the hypothesis of nutritional constraints as protein A significant correlation between protein carbon content and the base composition of the genomes is moreover evidenced, with low protein carbon content being with high DNA G+C content. This found in bacteria, archaea, and eukaryote not from carbon usage in proteins or protein families to proteomes. The analysis of several ortholog proteins, to different that the between low protein carbon content and high DNA G+C content as well for ortholog proteins. The correlation between DNA G+C and protein carbon contents strongly the possibility that the composition of one of might have been by the the canonical genetic the this the of the of DNA composition protein atomic composition or the The first composition protein atomic is in with the of evolution The of Google and more with the of (3Suekoa N. Cold Spring Harbor Symp. Quant. Biol. 1961; 26: 35-43Crossref PubMed Scopus (74) Google Scholar), to which genome composition are to between the and of the and N. S. A. PubMed Scopus Google Scholar). the molecular of is it was suggested to on DNA A. S. S. A. PubMed Scopus Google Scholar). was to in in amino acid composition in proteins G.A.C. Hickey D.A. Mol. Biol. Evol. 2000; 17: 1581-1588Crossref PubMed Scopus (192) Google Scholar, N. S. A. PubMed Scopus Google Scholar, PubMed Scopus Google Scholar). In this in DNA base usage to or carbon It has been that the amino acids were encoded by the and bases only S. A. PubMed Scopus Google Scholar), the that between protein carbon and DNA G+C contents might have been The composition DNA with a more of evolution and with the that metabolic and geochemical might be constraints that were on protein evolution (7Baudouin-Cornu P. Surdin-Kerjan Y. Marliere P. Thomas D. Science. 2001; 293: 297-300Crossref PubMed Scopus (112) Google Scholar, H. S. A. PubMed Scopus Google Scholar). In a the to nutritional in carbon could in the of the protein carbon content, which in to a of genomes in of the structure of the genetic code further arguments for such a and show that, the structure of the canonical the of the carbon content of proteins in to a genomic in The that constraints might during protein evolution has been previously in the of proteins in the a version of protein that is sulfur D. P. Nature. PubMed Scopus Google Scholar). sulfur by the of sulfur such as several enzymes that are by L. P. A. Mol. Cell. Full Text Full Text PDF PubMed Scopus (201) Google Scholar). our results that DNA G+C content might be a of the of protein atomic contents in response to carbon availability in natural habitats. G+C contents are to vary only in species with the genomes that show a average G+C content, from to N. S. A. PubMed Scopus Google Scholar, Mol. Evol. PubMed Scopus Google Scholar). This could be for by the that are to be subjected to specific elemental nutritional constraints less than Also, living bacteria to have genomes with significantly G+C and thus to proteins with less carbon than or bacteria, which on nutritional by their and thus are less subjected to carbon A. Genet. Full Text Full Text PDF PubMed Scopus Google Scholar). of the hypothesis that genome base composition is driven by nutritional constraints protein atomic contents would provide a way to the of genome composition a that at the of molecular are especially to for and