Diversity in G + C content at the third position of codons in vertebrate genes and its cause

Evolution of codon usage bias in Drosophila

We first review what is known about patterns of codon usage bias in Drosophila and make the following points: (i) Drosophila genes are as biased or more biased than those in microorganisms. (ii) The level of bias of genes and even the particular pattern of codon bias can remain phylogenetically invariant for very long periods of evolution. (iii) However, some genes, even very tightly linked genes, can change very greatly in codon bias across species. (iv) Generally G and especially C are favored at synonymous sites in biased genes. (v) With the exception of aspartic acid, all amino acids contribute significantly and about equally to the codon usage bias of a gene. (vi) While most individual amino acids that can use G or C at synonymous sites display a preference for C, there are exceptions: valine and leucine, which prefer G. (vii) Finally, smaller genes tend to be more biased than longer genes. We then examine possible causes of these patterns and discount mutation bias on three bases: there is little evidence of regional mutation bias in Drosophila, mutation bias is likely toward A+T (the opposite of codon usage bias), and not all amino acids display the preference for the same nucleotide in the wobble position. Two lines of evidence support a selection hypothesis based on tRNA pools: highly biased genes tend to be highly and/or rapidly expressed, and the preferred codons in highly biased genes optimally bind the most abundant isoaccepting tRNAs. Finally, we examine the effect of bias on DNA evolution and confirm that genes with high codon usage bias have lower rates of synonymous substitution between species than do genes with low codon usage bias. Surprisingly, we find that genes with higher codon usage bias display higher levels of intraspecific synonymous polymorphism. This may be due to opposing effects of recombination.

Precise switching of DNA replication timing in the GC content transition area in the human major histocompatibility complex

The human genome is composed of long-range G+C% (GC%) mosaic structures thought to be related to chromosome bands. We previously reported a boundary of megabase-sized GC% mosaic domains at the junction area between major histocompatibility complex (MHC) classes II and III, proposing it as a possible chromosome band boundary. DNA replication timing during the S phase is known to be correlated cytogenetically with chromosome band zones, and thus the band boundaries have been predicted to contain a switch point for DNA replication timing. In this study, to identify to the nucleotide sequence level the replication switch point during the S phase, we determined the precise DNA replication timing for MHC classes II and III, focusing on the junction area. To do this, we used PCR-based quantitation of nascent DNA obtained from synchronized human myeloid leukemia HL60 cells. The replication timing changed precisely in the boundary region with a 2-h difference between the two sides, supporting the prediction that this region may be a chromosome band boundary. We supposed that replication fork movement terminates (pauses) or significantly slows in the switch region, which contains dense Alu clusters; polypurine/polypyrimidine tracts; di-, tri-, or tetranucleotide repeats; and medium-reiteration-frequency sequences. Because the nascent DNA in the switch region was recovered at low efficiency, we investigated whether this region is associated with the nuclear scaffold and found three scaffold-associated regions in and around the switch region.

Codon usage tabulated from the international DNA sequence databases

Codon usage in 87 602 genes has been calculated using the nucleotide sequence data obtained from the GenBank Genetic Sequence Data Bank (Release 90.0; September 1995). The database is called the CUTG Database; the complete form of the database can be obtained by anonymous ftp from DDBJ and a part of the database, which lists the frequency of codon use in each organism, is made searchable through our World Wide Web server.

Human pseudoautosomal boundary-like sequences: expression and involvement in evolutionary formation of the present-day pseudoautosomal boundary of human sex chromosomes

The human genome is composed of long-range mosaic structures of G+C% (GC%), which are thought to be related to chromosome bands. We previously identified a boundary of Mb-level domains of GC% mosaic structures in the human major histocompatibility complex (MHC) and found in the domain boundary a sequence very similar to pseudoautosomal boundary (PAB) sequences of human sex chromosomes. We designated it 'PABL' and found many PABLs in the human genome. By analysis of six genomic and six transcribed PABLs, a core and consensus sequence of about 650 nt was defined; the 3'- and 5'-edges of the PABLs were strictly conserved. Northern blot analysis showed sizes of PABL transcripts to be 5-10 kb in length. Divergence time of PABLs was estimated to be 60-120 million years ago by analysis of human PABLs and PABXY1 of seven primates, and the evolutionary rates deduced showed PABLs to have been under selective constraints. A model for evolutionary formation of the present pseudoautosomal boundary was proposed by postulation of illegitimate recombination between two PABLs.

DNA sequence evolution: the sounds of silence

Silent sites (positions that can undergo synonymous substitutions) in protein-coding genes can illuminate two evolutionary processes. First, despite being silent, they may be subject to natural selection. Among eukaryotes this is exemplified by yeast, where synonymous codon usage patterns are shaped by selection for particular codons that are more efficiently and/or accurately translated by the most abundant tRNAs; codon usage across the genome, and the abundance of different tRNA species, are highly co-adapted. Second, in the absence of selection, silent sites reveal underlying mutational patterns. Codon usage varies enormously among human genes, and yet silent sites do not appear to be influenced by natural selection, suggesting that mutation patterns vary among regions of the genome. At first, the yeast and human genomes were thought to reflect a dichotomy between unicellular and multicellular organisms. However, it now appears that natural selection shapes codon usage in some multicellular species (e.g. Drosophila and Caenorhabditis), and that regional variations in mutation biases occur in yeast. Silent sites (in serine codons) also provide evidence for mutational events changing adjacent nucleotides simultaneously.

The tendency of lentiviral open reading frames to become A-rich: constraints imposed by viral genome organization and cellular tRNA availability

Human immunodeficiency virus type 1 (HIV-1) and other lentiviridae demonstrate a strong preference for the A-nucleotide, which can account for up to 40% of the viral RNA genome. The biological mechanism responsible for this nucleotide bias is currently unknown. The increased A-content of these viral genomes corresponds to the typical use of synonymous codons by all members of the lentiviral family (HIV, SIV, BIV, FIV, CAEV, EIAV, visna) and the human spuma retrovirus, but not by other retroviruses like the human T-cell leukemia viruses HTLV-1 and HTLV-II. In this article, we analyzed A-bias for all codon groups in all open reading frames of several lentiviruses. The extent of lentiviral codon bias could be related to host cellular translation. By calculating codon bias indices (CBIs), we were able to demonstrate an inverse correlation between the extent of codon bias and the rate of translation of individual reading frames in these viruses. Specifically, the shift toward A-rich codons is more pronounced in pol than in gag lentiviral genes. Since it is known that Gag synthesis exceeds Pol synthesis by a factor of 20 due to infrequent ribosomal frame-shifting during translation of the gap-pol mRNA molecule, we propose that the aminoacyl-tRNA availability in the host cell restricts the lentiviral preference for A-rich codons. In addition, less A-nucleotides were found in regions of the viral genome encoding multiple functions; e.g., overlapping reading frames (tat-rev-env) or in genes that overlap regulatory sequences (nef-LTR region).(ABSTRACT TRUNCATED AT 250 WORDS)

Codon bias in actin multigene families and effects on the reconstruction of phylogenetic relationships

Codon usage patterns and phylogenetic relationships in the actin multigene family have been analyzed for three dipteran species--Drosophila melanogaster, Bactrocera dorsalis, and Ceratitis capitata. In certain phylogenetic tree reconstructions, using synonymous distances, some gene relationships are altered due to a homogenization phenomenon. We present evidence to show that this homogenization phenomenon is due to codon usage bias. A survey of the pattern of synonymous codon preferences for 11 actin genes from these three species reveals that five out of the six Drosophila actin genes show high degrees of codon bias as indicated by scaled chi 2 values. In contrast to this, four out of the five actin genes from the other species have low codon bias values. A Monte Carlo contingency test indicates that for those Drosophila actin genes which exhibit codon bias, the patterns of codon usage are different compared to actin genes from the other species. In addition, the genes exhibiting codon bias also appear to have reduced rates of synonymous substitution. The homogenization phenomenon seen in terms of synonymous substitutions is not observed for nonsynonymous changes. Because of this homogenization phenomenon, "trees" constructed based on synonymous substitutions will be affected. These effects can be overt in the case of multigene families, but similar distortions may underlie reconstructions based on single-copy genes which exhibit codon usage bias.

Correlation between codon usage, regional genomic nucleotide composition, and amino acid composition in the cytochrome P-450 gene superfamily

The codon usage bias of 110 mammalian cytochrome P-450 genes has been determined and analyzed in relation to a variety of genetic, biochemical, and physiological parameters. In those P-450 genes exhibiting biased usage the preferred codons generally do not differ among the four species examined (rat, rabbit, man, and mouse) or from the predominantly used codons identified for all sequenced genes in a recent data base analysis (Wada et al. (1992) Nucleic Acids Res. 20 (Suppl.), 2111-2118). Codon usage bias does not correlate with evolutionary relationships, evolutionary age, or with the extent of evolutionary conservation of orthologous proteins; there is no obvious correlation with the level of expression of a given P-450, with its inducibility, nor with its physiologic role; and neither the preferred codons nor the degree of bias differ for P-450s expressed in different tissues. Codon usage bias does correlate with the C+G content at the codon third position, and thus preferred codons usually end in C or G; for those P-450s for which gene sequences are available this bias also correlates with the C + G content of the intronic and flanking regions of these genes. Moreover, a lesser increase in the C + G content at the codon first and second positions is also evident in genes located in regions of high C + G content; this leads to predictable differences in the amino acid compositions of P-450 enzymes that correlate with genomic nucleotide composition and the degree of bias in codon usage.

A boundary of long-range G + C% mosaic domains in the human MHC locus: pseudoautosomal boundary-like sequence exists near the boundary

The human genome is composed of long-range G+C% (GC%) mosaic structures related to chromosome bands. We found the human MHC locus to be an example of megabase-level GC% mosaic structures and predicted a possible boundary of the megabase-level domains within an undercharacterized 450-kb region harboring the junction of MHC classes II and III. Chromosome walking of the 450-kb region and base-compositional analysis precisely located the boundary of the mosaic domains, disclosing a sharp GC% transition. Near the transition point there was a 20-kb dense Alu cluster, a 30-kb dense LINE-1 cluster, and a sequence highly homologous with the pseudoautosomal boundary of the short arms of human sex chromosomes (PAB1X and PAB1Y); PAB1X and PAB1Y are the interface between sex-specific and pseudoautosomal regions. Many PAB1XY-like sequences (PABLs) were detected by hybridization against genomic DNA, and the new sequences defined the complete form of PABLs to be about 650 nt.

Prediction of function in DNA sequence analysis

Recognition of function of newly sequenced DNA fragments is an important area of computational molecular biology. Here we present an extensive review of methods for prediction of functional sites, tRNA, and protein-coding genes and discuss possible further directions of research in this area.

Frequencies of synonymous substitutions in mammals are gene-specific and correlated with frequencies of nonsynonymous substitutions

The frequencies of synonymous substitutions of mammalian genes cover a much wider range than previously thought. We report here that the different frequencies found in homologous genes from a given mammalian pair are correlated with those in the same homologous genes from a different mammalian pair. This indicates that the frequencies of synonymous substitutions are gene-specific (as are the frequencies of nonsynonymous substitutions), or, in other words, that "fast" and "slow" genes in one mammal are fast and slow, respectively, in any other one. Moreover, the frequencies of synonymous substitutions are correlated with the frequencies of nonsynonymous substitution in the same genes.

MHC polymorphism and parasites

The major histocompatibility complex (MHC) polymorphism is marked by the existence of allelic lineages that are extremely old, having been passed from one species to another in an evolutionary line of descent. Each species has several of these lineages and many of their more recent derivatives, the actual alleles. The lineages are separated by large genetic distances and are characterized by the presence of short sequence motifs which, at the protein level, have remained virtually unaltered for over 40 million years. Several explanations for the MHC polymorphism have been proposed. We argue that the only one consistent with the entire body of knowledge about the MHC is an explanation based on the immune response to parasites. Furthermore, we propose that parasites coevolving with their hosts have had a major influence on MHC polymorphism, whereas parasites that switched hosts recently and became very virulent have had little effect. The latter category includes micro- and macroparasites responsible for the major human infectious diseases. This hypothesis explains why no convincing association between human leucocyte antigen (HLA) alleles and resistance to infectious disease can thus far be documented, and indicates the direction in which the search for such associations should be taken.

Biased distribution of adenine and thymine in gene nucleotide sequences

We analyzed occurrences of bases in 20,352 introns, exons of 25,574 protein-coding genes, and among the three codon positions in the protein-coding sequences. The nucleotide sequences originated from the whole spectrum of organisms from bacteria to primates. The analysis revealed the following: (1) In most exons, adenine dominates over thymine. In other words, adenine and thymine are distributed in an asymmetric way between the exon and the complementary strand, and the coding sequence is mostly located in the adenine-rich strand. (2) Thymine dominates over adenine not only in the strand complementary to the exon but also in introns. (3) A general bias is further revealed in the distribution of adenine and thymine among the three codon positions in the exons, where adenine dominates over thymine in the second and mainly the first codon position while the reverse holds in the third codon position. The product (A1/T1)x(A2/T2)x(T3/A3) is smaller than one in only a few analyzed genes.

Distribution profiles of GC content around the translation initiation site in different species

We have analyzed the distribution of guanine-cytosine (GC) content around the translation initiation site in genomic DNA sequences of different species. A set of sequences belonging to one species is aligned at the translation initiation site, and the average GC content is calculated for 100 base windows over a range of 500 bases each for upstream and downstream region. Consistent with previous observations that coding regions are more GC-rich than non-coding regions, we observe a jump in the GC content at the translation initiation site, except for vertebrate sequences. It was also found that the overall shape of the GC content profile is similar within each organism group even though the average GC contents can be very different. Furthermore, by examining different profiles for different species, we have found a negative correlation between the average GC content and the jump size at the translation initiation site. Apparently, more AT-rich genomes, which tend to lack macroscopic mosaic structures, exhibit more marked differences in the GC content at the microscopic level.

Length variation in mitochondrial DNA of the minnow Cyprinella spiloptera

Length differences in animal mitochondrial DNA (mtDNA) are common, frequently due to variation in copy number of direct tandem duplications. While such duplications appear to form without great difficulty in some taxonomic groups, they appear to be relatively short-lived, as typical duplication products are geographically restricted within species and infrequently shared among species. To better understand such length variation, we have studied a tandem and direct duplication of approximately 260 bp in the control region of the cyprinid fish, Cyprinella spiloptera. Restriction site analysis of 38 individuals was used to characterize population structure and the distribution of variation in repeat copy number. This revealed two length variants, including individuals with two or three copies of the repeat, and little geographic structure among populations. No standard length (single copy) genomes were found and heteroplasmy, a common feature of length variation in other taxa, was absent. Nucleotide sequence of tandem duplications and flanking regions localized duplication junctions in the phenylalanine tRNA and near the origin of replication. The locations of these junctions and the stability of folded repeat copies support the hypothesized importance of secondary structures in models of duplication formation.

Selection on silent sites in the rodent H3 histone gene family

Selection promoting differential use of synonymous codons has been shown for several unicellular organisms and for Drosophila, but not for mammals. Selection coefficients operating on synonymous codons are likely to be extremely small, so that a very large effective population size is required for selection to overcome the effects of drift. In mammals, codon-usage bias is believed to be determined exclusively by mutation pressure, with differences between genes due to large-scale variation in base composition around the genome. The replication-dependent histone genes are expressed at extremely high levels during periods of DNA synthesis, and thus are among the most likely mammalian genes to be affected by selection on synonymous codon usage. We suggest that the extremely biased pattern of codon usage in the H3 genes is determined in part by selection. Silent site G + C content is much higher than expected based on flanking sequence G + C content, compared to other rodent genes with similar silent site base composition but lower levels of expression. Dinucleotide-mediated mutation bias does affect codon usage, but the affect is limited to the choice between G and C in some fourfold degenerate codons. Gene conversion between the two clusters of histone genes has not been an important force in the evolution of the H3 genes, but gene conversion appears to have had some effect within the cluster on chromosome 13.

Codon usage in Caenorhabditis elegans: delineation of translational selection and mutational biases

Synonymous codon usage varies considerably among Caenorhabditis elegans genes. Multivariate statistical analyses reveal a single major trend among genes. At one end of the trend lie genes with relatively unbiased codon usage. These genes appear to be lowly expressed, and their patterns of codon usage are consistent with mutational biases influenced by the neighbouring nucleotide. At the other extreme lie genes with extremely biased codon usage. These genes appear to be highly expressed, and their codon usage seems to have been shaped by selection favouring a limited number of translationally optimal codons. Thus, the frequency of these optimal codons in a gene appears to be correlated with the level of gene expression, and may be a useful indicator in the case of genes (or open reading frames) whose expression levels (or even function) are unknown. A second, relatively minor trend among genes is correlated with the frequency of G at synonymously variable sites. It is not yet clear whether this trend reflects variation in base composition (or mutational biases) among regions of the C.elegans genome, or some other factor. Sequence divergence between C.elegans and C.briggsae has also been studied.

The unusual nucleotide content of the HIV RNA genome results in a biased amino acid composition of HIV proteins

Extremely high frequencies of the A nucleotide are found in the RNA genomes of the lentivirus group of retroviruses. It is presently unknown what molecular force is responsible for this A-pressure. In this manuscript, we demonstrate a correlation between this 'A-pressure' and the amino acid-usage of the lentivirus family. We compared the amino acid composition of the Gag and Pol proteins of the human immunodeficiency viruses type 1 and 2 (HIV-1 and HIV-2) with that of the second group of human retroviruses; the human T-cell leukemia viruses type I and II (HTLV-I and HTLV-II). Differences in total amino acid content correlate with the preference for A-rich codons in the HIV genome. A pair-wise comparison of homologous amino acid positions in the Pol proteins indicates that both conservative and non-conservative changes can be accounted for by this A-bias. The putative molecular mechanism underlying this A-pressure and the evolutionary consequences are discussed.

Codon usage patterns among genes for lepidopteran hemolymph proteins

Patterns in codon usage were examined for the coding regions of the 23 known lepidopteran hemolymph proteins. Coding triplets are GC rich at the third position and a significant linear relationship between GC content of silent and nonsilent (replacement) sites was demonstrated. Intron GC content was significantly lower than in coding regions and no relationship between intron GC content and the same at silent and nonsilent sites was found. Though hemolymph proteins are all produced by the same tissue--fat body--significantly less bias was observed when all moth sequences were pooled than when sequences of the two major species were analyzed separately, as predicted by the genome hypothesis. In cases where no statistically significant bias was observed, polar or acidic/basic amino acids were almost exclusively involved. Calculation of codon adaptation indices (CAI) was of limited value in quantifying the degree of codon bias and probably reflects the complexity of multicellular-organism life cycles and the changing patterns of gene expression over different developmental stages.

Synonymous codon usage in Drosophila melanogaster: natural selection and translational accuracy

I present evidence that natural selection biases synonymous codon usage to enhance the accuracy of protein synthesis in Drosophila melanogaster. Since the fitness cost of a translational misincorporation will depend on how the amino acid substitution affects protein function, selection for translational accuracy predicts an association between codon usage in DNA and functional constraint at the protein level. The frequency of preferred codons is significantly higher at codons conserved for amino acids than at nonconserved codons in 38 genes compared between D. melanogaster and Drosophila virilis or Drosophila pseudoobscura (Z = 5.93, P < 10(-6)). Preferred codon usage is also significantly higher in putative zinc-finger and homeodomain regions than in the rest of 28 D. melanogaster transcription factor encoding genes (Z = 8.38, P < 10(-6)). Mutational alternatives (within-gene differences in mutation rates, amino acid changes altering codon preference states, and doublet mutations at adjacent bases) do not appear to explain this association between synonymous codon usage and amino acid constraint.

The GAPDH gene system of the red alga Chondrus crispus: promoter structures, intron/exon organization, genomic complexity and differential expression of genes

Our previous phylogenetic analysis based on cDNA sequences of chloroplast and cytosolic glyceraldehyde-3-phosphate dehydrogenases (GAPDH; genes GapA and GapC, respectively) of the red alga Chondrus crispus suggested that rhodophytes and green plants are sister groups with respect to plastids and mitochondria and diverged at about the same time or somewhat later than animals and fungi. Here we characterize the genomic sequences of genes GapC and GapA of C. crispus with respect to promotor structures, intron/exon organization, genomic complexity, G + C content, CpG suppression and their transcript levels in gametophytes and protoplasts, respectively. To our knowledge this is the first report on nuclear protein genes of red algae. The GapC gene is G + C-rich, contains no introns and displays a number of classic sequence motifs within its promotor region, such as TATA, CAAT, GC boxes and several elements resembling the plant-specific G-box palindrome. The GapA gene has a moderate G+C content, a single CAAT box motif in its promotor region and a single intron of 115 bp near its 5' end. This intron occupies a conserved position corresponding to that of intron 1 in the transit peptide region of chloroplast GAPDH genes (GapA and GapB) of higher plants. It has consensus sequences similar to those of yeast introns and folds into a conspicuous secondary structure of -61.3 kJ. CpG profiles of genes GapC and GapA and their flanking sequences show no significant CpG depletion suggesting that these genomic sequences are not methylated. Genomic Southern blots hybridized with generic and gene specific probes indicate that both genes are encoded by single loci composed of multiple polymorphic alleles. Northern hybridizations demonstrate that both genes are expressed in gametophytes but not in protoplasts where appreciable amounts of transcripts can only be detected for GapC.

Silent substitutions in mammalian genomes and their evolutionary implications

An analysis of silent substitutions in pairwise comparisons of homologous genes from different mammals has shown that, in spite of individual fluctuations, their frequencies (which are very strongly correlated with the frequency of substitutions per synonymous site calculated according to Li et al. 1985) do not vary, on the average, with the GC levels of silent positions. This holds in the general case, in which silent positions of pairs of homologous genes share the same composition, namely in the human/other primates, human/artiodactyls, and in the mouse/rat pairs, as well as in the special cases in which the composition of silent positions are different, namely in the human/rabbit and the human/rat (or human/mouse) pairs. A slightly lower frequency found for low GC values in the human/bovine and human/pig pairs seems to be due to the specific gene samples used. These results contradict the previously claimed existence of differences in mutation rates and of mutational biases in third codon positions of coding sequences located in different isochores of mammalian genomes. They also imply that the variations in nucleotide precursor pools through the cell cycle and the differences in replication timing, or in repair efficiency, which were reported for different isochores, do not lead, as claimed, to differences in mutation rates, not in mutational biases in mammals. The differences claimed appear to be due to using small gene samples when individual fluctuations from gene to gene are relatively large.

Mammalian gene evolution: nucleotide sequence divergence between mouse and rat

As a paradigm of mammalian gene evolution, the nature and extent of DNA sequence divergence between homologous protein-coding genes from mouse and rat have been investigated. The data set examined includes 363 genes totalling 411 kilobases, making this by far the largest comparison conducted between a single pair of species. Mouse and rat genes are on average 93.4% identical in nucleotide sequence and 93.9% identical in amino acid sequence. Individual genes vary substantially in the extent of nonsynonymous nucleotide substitution, as expected from protein evolution studies; here the variation is characterized. The extent of synonymous (or silent) substitution also varies considerably among genes, though the coefficient of variation is about four times smaller than for nonsynonymous substitutions. A small number of genes mapped to the X-chromosome have a slower rate of molecular evolution than average, as predicted if molecular evolution is "male-driven." Base composition at silent sites varies from 33% to 95% G+C in different genes; mouse and rat homologues differ on average by only 1.7% in silent-site G+C, but it is shown that this is not necessarily due to any selective constraint on their base composition. Synonymous substitution rates and silent site base composition appear to be related (genes at intermediate G+C have on average higher rates), but the relationship is not as strong as in our earlier analyses. Rates of synonymous and nonsynonymous substitution are correlated, apparently because of an excess of substitutions involving adjacent pairs of nucleotides. Several factors suggest that synonymous codon usage in rodent genes is not subject to selection.

Chloroplast DNA codon use: evidence for selection at the psb A locus based on tRNA availability

Codon use in the three sequenced chloroplast genomes (Marchantia, Oryza, and Nicotiana) is examined. The chloroplast has a bias in that codons NNA and NNT are favored over synonymous NNC and NNG codons. This appears to be a consequence of an overall high A + T content of the genome. This pattern of codon use is not followed by the psb A gene of all three genomes and other psb A sequences examined. In this gene, the codon use favors NNC over NNT for twofold degenerate amino acids. In each case the only tRNA coded by the genome is complementary to the NNC codon. This codon use is similar to the codon use by chloroplast genes examined from Chlamydomonas reinhardtii. Since psb A is the major translation product of the chloroplast, this suggests that selection is acting on the codon use of this gene to adapt codons to tRNA availability, as previously suggested for unicellular organisms.

Dinucleotides and G+C content in human genes: opposite behavior of GpG, GpC, and TpC at II-III codon positions and in introns

We have studied the behavior of the dinucleotide preferences under G+C content variation in human genes. The doublet preferences for each dinucleotide were compared between two functionally distinct zones in genes, the II-III codon positions, and the introns. The 16 dinucleotides have been tentatively classified in three groups: AA, AC, CC, CT, and GA, doublets showing no difference between introns and II-III codon positions in the full range of G+C variation TG and TA, which differ in the full range of G+C variation AT, AG, GT, TC, TT, GG, GC, CG, and CA, which show differences in regions over 50% G+C A remarkable pattern observed concerns the behavior of GG, GC, and TC, which showed opposite trends in II-III codon positions and in introns. If codon positions and introns are under the same structural requirements and the same mutational bias, our results indicate that the differences observed could be related to post-transcriptional constraints acting on mRNA.

Codon usage bias and base composition of nuclear genes in Drosophila

The nuclear genes of Drosophila evolve at various rates. This variation seems to correlate with codon-usage bias. In order to elucidate the determining factors of the various evolutionary rates and codon-usage bias in the Drosophila nuclear genome, we compared patterns of codon-usage bias with base compositions of exons and introns. Our results clearly show the existence of selective constraints at the translational level for synonymous (silent) sites and, on the other hand, the neutrality or near neutrality of long stretches of nucleotide sequence within noncoding regions. These features were found for comparisons among nuclear genes in a particular species (Drosophila melanogaster, Drosophila pseudoobscura and Drosophila virilis) as well as in a particular gene (alcohol dehydrogenase) among different species in the genus Drosophila. The patterns of evolution of synonymous sites in Drosophila are more similar to those in the prokaryotes than they are to those in mammals. If a difference in the level of expression of each gene is a main reason for the difference in the degree of selective constraint, the evolution of synonymous sites of Drosophila genes would be sensitive to the level of expression among genes and would change as the level of expression becomes altered in different species. Our analysis verifies these predictions and also identifies additional selective constraints at the translational level in Drosophila.

Relationship between G + C in silent sites of codons and amino acid composition of human proteins

We have investigated the relationship between the G + C content of silent (synonymous) sites in codons and the amino acid composition of encoded proteins for approximately 1,600 human genes. There are positive correlations between silent site G + C and the proportions of codons for Arg, Pro, Ala, Trp, His, Gln, and Leu and negative ones for Tyr, Phe, Asn, Ile, Lys, Asp, Thr, and Glu. The median proteins coded by groups of genes that differ in silent-site G + C content also differ in amino acid composition, as do some proteins coded by homologous genes. The pattern of compositional change can be largely explained by directional mutation pressure, the genetic code, and differences in the frequencies of accepted amino acid substitutions; the shifts in protein composition are likely to be selectively neutral.

Regional base composition variation along yeast chromosome III: evolution of chromosome primary structure

The recent determination of the complete sequence of chromosome III from the yeast Saccharomyces cerevisiae allows, for the first time, the investigation of the long range primary structure of a eukaryotic chromosome. We have found that, against a background G+C level of about 35%, there are two regions (one in each chromosome arm) in which G+C values rise to over 50%. This effect is seen in silent sites within genes, but not in noncoding intergenic sequences. The variation in G+C content is not related to differential selection of synonymous codons, and probably reflects mutational biases. That the intergenic regions do not exhibit the same phenomenon is particularly interesting, and suggests that they are under substantial constraint. The yeast chromosome may be a model of the structure of the human genome, since there is evidence that it is also a mosaic of long regions of different base compositions, reflected in wide variation of G+C content at silent sites among genes. Two possible causes of this regional effect, replication timing, and recombination frequency, are discussed.

Codon usage patterns in chromosomal and retrotransposon genes of the mosquito Anopheles gambiae

Codon usage was compiled for fourteen chromosomal genes and four retrotransposons from the mosquito Anopheles gambiae. Variation exists among chromosomal genes in the degree of bias. The genes showing the highest bias are probably most highly expressed. In these genes, the base composition at the third codon position is much richer in G + C than is the overall coding sequence. Thus, codon usage is biased toward G- or C-ending codons. Codon usage in each retrotransposon is quite different, not only from chromosomal genes but also from the other retrotransposons. Codon usage comparisons among homologous genes from An. gambiae and two other Dipterans, the yellow fever mosquito Aedes aegypti and the fruitfly Drosophila melanogaster, show that while there are similarities, particularly between An. gambiae and D. melanogaster in the preference for G- and C-ending codons, each species has evolved a distinct pattern of codon usage.

Codon preference of Aedes aegypti and Aedes albopictus

The codon bias of two Aedes mosquito species was examined using a sign test. In general, there appeared to be some preference for C + G at the third base position, although this was not statistically significant. While amino acids such as phenylalanine and tyrosine clearly displayed biases, others such as valine and serine appeared to have little or no bias for any particular codon. Three homologous genes of Aedes aegypti and Drosophila melanogaster were compared using the chi-square test and the codon bias of the two species compared. Drosophila melanogaster was found to have a much stronger bias for C + G at the third base position compared to Aedes. The implications and usefulness of the codon bias are discussed.

Cluster of fibronectin type III repeats found in the human major histocompatibility complex class III region shows the highest homology with the repeats in an extracellular matrix protein, tenascin

Walking and sequencing a genome portion centromeric of CYP21B in the human MHC class III region disclosed a cluster of fibronectin type III repeats in an approximately 50-kb DNA segment. Fibronectin type III repeats are known to consist of ca. 90 amino acid residues and exist in a wide range of protein species. Homology searches in protein databases showed that the repeats found had the highest homology with the repeats of human tenascin, an extracellular matrix protein. One cDNA sequence located immediately centromeric of CYP21B, the 3' portion of which is transcribed by the opposite strand of CYP21B, was found also to have six type III repeats followed by a fibrinogen domain. Pairwise homology comparison of these repeats in the MHC locus with those of human tenascin showed a general parallelism in their gene organization, indicating that the newly found repeats are elements of certain tenascin-like genea.

Codon usage in the G+C-rich Streptomyces genome

The codon usage (CU) patterns of 64 genes from the Gram+ prokaryotic genus Streptomyces were analysed. Despite the extremely high overall G+C content of the Streptomyces genome (estimated at 0.74), individual genes varied in G+C content from 0.610 to 0.797, and had third codon position G+C contents (GC3s) that varied from 0.764 to 0.983. The variation in GC3s explains a significant proportion of the variation in CU patterns. This is consistent with an evolutionary model of the Streptomyces genome where biased mutation pressure has led to a high average G+C content with random variation about the mean, although the variation observed is greater than that expected from a simple binomial model. The only gene in the sample that can be confidently predicted to be highly expressed, EF-Tu of Streptomyces coelicolor A3(2) (GC3s = 0.927), shows a preference for a third position C in several of the four codon families, and for CGY and GGY for Arg and Gly codons, respectively (Y = pyrimidine); similar CU patterns are found in highly expressed genes of the G+C-rich Micrococcus luteus genome. It thus appears that codon usage in Streptomyces is determined predominantly by mutation bias, with weak translational selection operating only in highly expressed genes. We discuss the possible consequences of the extreme codon bias of Streptomyces and consider how it may have evolved. A set of CU tables is provided for use with computer programs that locate protein-coding regions.

Recent evidence for evolution of the genetic code

The genetic code, formerly thought to be frozen, is now known to be in a state of evolution. This was first shown in 1979 by Barrell et al. (G. Barrell, A. T. Bankier, and J. Drouin, Nature [London] 282:189-194, 1979), who found that the universal codons AUA (isoleucine) and UGA (stop) coded for methionine and tryptophan, respectively, in human mitochondria. Subsequent studies have shown that UGA codes for tryptophan in Mycoplasma spp. and in all nonplant mitochondria that have been examined. Universal stop codons UAA and UAG code for glutamine in ciliated protozoa (except Euplotes octacarinatus) and in a green alga, Acetabularia. E. octacarinatus uses UAA for stop and UGA for cysteine. Candida species, which are yeasts, use CUG (leucine) for serine. Other departures from the universal code, all in nonplant mitochondria, are CUN (leucine) for threonine (in yeasts), AAA (lysine) for asparagine (in platyhelminths and echinoderms), UAA (stop) for tyrosine (in planaria), and AGR (arginine) for serine (in several animal orders) and for stop (in vertebrates). We propose that the changes are typically preceded by loss of a codon from all coding sequences in an organism or organelle, often as a result of directional mutation pressure, accompanied by loss of the tRNA that translates the codon. The codon reappears later by conversion of another codon and emergence of a tRNA that translates the reappeared codon with a different assignment. Changes in release factors also contribute to these revised assignments. We also discuss the use of UGA (stop) as a selenocysteine codon and the early history of the code.

Codon usage in Aspergillus nidulans

Synonymous codon usage in genes from the ascomycete (filamentous) fungus Aspergillus nidulans has been investigated. A total of 45 gene sequences has been analysed. Multivariate statistical analysis has been used to identify a single major trend among genes. At one end of this trend are lowly expressed genes, whereas at the other extreme lie genes known or expected to be highly expressed. The major trend is from nearly random codon usage (in the lowly expressed genes) to codon usage that is highly biased towards a set of 19-20 "optimal" codons. The G + C content of the A. nidulans genome is close to 50%, indicating little overall mutational bias, and so the codon usage of lowly expressed genes is as expected in the absence of selection pressure at silent sites. Most of the optimal codons are C- or G- ending, making highly expressed genes more G + C-rich at silent sites.

Essential factors determining codon usage in ubiquitin genes

Ubiquitin is ubiquitous in all eukaryotes and its amino acid sequence shows extreme conservation. Ubiquitin genes comprise direct repeats of the ubiquitin coding unit with no spacers. The nucleotide sequences coding for 13 ubiquitin genes from 11 species reported so far have been compiled and analyzed. The G + C content of codon third base reveals a positive linear correlation with the genome G + C content of the corresponding species. The slope strongly suggests that the overall G + C content of codons of polyubiquitin genes clearly reflects the genome G + C content by AT/GC substitutions at the codon third position. The G + C content of ubiquitin codon third base also shows a positive linear correlation with the overall G + C content of coding regions of compiled genes, indicating the codon choices among synonymous codons reflect the average codon usage pattern of corresponding species. On the other hand, the monoubiquitin gene, which is different from the polyubiquitin gene in gene organization, gene expression, and function of the encoding protein, shows a different codon usage pattern compared with that of the polyubiquitin gene. From comparisons of the levels of synonymous substitutions among ubiquitin repeats and the homology of the amino acid sequence of the tail of monomeric ubiquitin genes, we propose that the molecular evolution of ubiquitin genes occurred as follows: Plural primitive ubiquitin sequences were dispersed on genome in ancestral eukaryotes. Some of them situated in a particular environment fused with the tail sequence to produce monomeric ubiquitin genes that were maintained across species. After divergence of species, polyubiquitin genes were formed by duplication of the other primitive ubiquitin sequences on different chromosomes. Differences in the environments in which ubiquitin genes are embedded reflect the differences in codon choice and in gene expression pattern between poly- and monomeric ubiquitin genes.

Evident diversity of codon usage patterns of human genes with respect to chromosome banding patterns and chromosome numbers; relation between nucleotide sequence data and cytogenetic data

The sequences of the human genome compiled in DNA databases are now about 10 megabase pairs (Mb), and thus the size of the sequences is several times the average size of chromosome bands at high resolution. By surveying this large quantity of data, it may be possible to clarify the global characteristics of the human genome, that is, correlation of gene sequence data (kb-level) to cytogenetic data (Mb-level). By extensively searching the GenBank database, we calculated codon usages in about 2000 human sequences. The highest G + C percentage at the third codon position was 97%, and that of about 250 sequences was 80% or more. The lowest G + C% was 27%, and that in about 150 sequences was 40% or less. A major portion of the GC-rich genes was found to be on special subsets of R-bands (T-bands and/or terminal R-bands). AT-rich genes, however, were mainly on G-bands or non-T-type internal R-bands. Average G + C% at the third position for individual chromosomes differed among chromosomes, and were related to T-band density, quinacrine dullness, and mitotic chiasmata density in the respective chromosomes.

The compositional properties of human genes

The present work represents the first attempt to study in greater detail previously proposed compositional correlations in genomes, based on a body of additional data relating to gene localizations as well as to extended flanking sequences extracted from gene banks. We have investigated the correlations that exist between (1) the GC levels of exons of human genes, and (2) the GC levels of either intergenic sequences or introns associated with the genes under consideration. In both cases, linear relationships with slopes close to unity were found. The similarity of the linear relationships indicates similar GC levels in intergenic sequences and introns located in the same isochores. Moreover, both intergenic sequences and introns showed GC levels 5-10% lower than the corresponding exons. The above findings considerably strengthen the previously drawn conclusion that coding and noncoding sequences (both inter- and intragenic) from the same isochores of the human genome are compositionally correlated. In addition, we find linear correlations between the GC levels of codon positions and of the intergenic sequences or introns associated with the corresponding genes, as well as among the GC levels of codon positions of genes.

Mammalian DNA replication: mutation biases and the mutation rate

Experimental studies have shown that the fidelity of DNA replication can be affected by the concentrations of free deoxyribonucleotides present in the cell. Replication of mammalian chromosomes is achieved using pools of newly-synthesized deoxyribonucleotides which fluctuate during the cell cycle. Since regions of mammalian chromosomes are replicated sequentially, there is the potential for differences among mammalian loci in both the relative and absolute frequencies of the various transitional and transversional mutations which may occur. Where these mutations are effectively neutral, at silent sites in genes and in non-coding sequences, this may result in different rates of evolution and in different base compositions, as have been observed in data from mammalian genes. A simple model of the DNA replication process is developed to describe how the mutation rate could be affected by the G + C contents of the deoxyribonucleotide pools and of the replicating DNA. Mutation rates are predicted to vary from locus to locus; only in the particular case of identical G + C contents in the DNA locus and the deoxyribonucleotide pools, and no proofreading, will the mutation rate be uniform over all loci.

On the structure and chromosome location of the 72- and 92-kDa human type IV collagenase genes

The 72- and 92-kDa type IV collagenases are members of a group of secreted zinc metalloproteases. Two members of this family, collagenase and stromelysin, have previously been localized to the long arm of chromosome 11. Here we assign both of the two type IV collagenase genes to human chromosome 16. By sequencing, the 72-kDa gene is shown to consist of 13 exons, 3 more than have been reported for the other members of this gene family. The extra exons encode the amino acids of the fibronectin-like domain which has so far been found in only the 72- and 92-kDa type IV collagenase. The evolutionary relationship among the members of this gene family is discussed.

Codon context

The analysis of coding sequences reveals nonrandomness in the context of both sense and stop codons. Part of this is related to nucleotide doublet preference, seen also in non-coding sequences and thought to arise from the dependence of mutational events on surrounding sequence. Another nonrandom context element, relating the wobble nucleotides of successive codons, is observed even when doublet preference, codon usage and bias in amino acid doublets are all allowed for. Several phenomena related to protein synthesis have been shown in vivo to be affected by the nucleotide sequence around codons. Thus, nonsense and missense suppression, elongation rate, precision of tRNA selection and polypeptide chain termination are all affected by codon context. At present, it remains unclear how these phenomena may influence the evolution of nonrandomness in the context of codons in natural sequences.

Sequence analysis suggests that tetra-nucleotides signal the termination of protein synthesis in eukaryotes

An increasing number of cases where tri-nucleotide stop codons do not signal the termination of protein synthesis are being reported. In order to identify what constitutes an efficient stop signal, we analysed the region around natural stop codons in genes from a wide variety of eukaryotic species and gene families. Certain stop codons and nucleotides following stop codons are over-represented, and this pattern is accentuated in highly expressed genes. For example, the preferred signal for Saccharomyces cerevisiae and Drosophila melanogaster highly expressed genes is UAAG, and generally the signals UAA(A/G) and UGA(A/G) are preferred in eukaryotes. The GC% of the organism or DNA region can affect whether there is A or G in the second or fourth positions. We suggest therefore, that the stop codon and the nucleotide following it comprise a tetra-nucleotide stop signal. A model is proposed in which the polypeptide chain release factor, a protein, recognises this sequence, but will tolerate some substitution, particularly A to G in the second or third positions.

Three human homologs of a murine gene encoding an inhibitor of stem cell proliferation

The G0S19 genes are members of the "small inducible" family of genes, which have similar exon-intron organizations and encode secreted proteins with similar dispositions of cysteine and proline residues. G0S19-1 mRNA is increased shortly after the addition of lectin or cycloheximide to cultured human blood mononuclear cells. The cDNA sequence is homologous to that of a murine gene encoding an inhibitory cytokine (MIP1 alpha/SCI), which decreases hemopoietic stem cell proliferation. The homology extends to the 3' noncoding region, which contains two conserved elements: (i) GGGACTCTTC, a potential transcription factor NF chi B-binding site, and (ii) TTTTGTAATTTATTTT, which is found in some related genes (e.g., that encoding the immediate early protein ornithine decarboxylase). A similar but complementary sequence is present in human immunodeficiency virus. Two of the three human genes that hybridize to G0S19-1 cDNA were sequenced. G0S19-1 has 5' AP1-like recognition elements as found in some other phorbol ester-responsive genes (e.g., c-fos). G0S19-2 has a 5' Alu sequence, but is likely to be expressed because of the conservation of sections of the gene believed to be important for function. The 5' flanks of both genes contain the nucleotide motifs CK-2 and SRE, indicating cytokine-like genes with the potential to respond to growth factors. G0S19-1 is the main G0S19 gene expressed in adult T lymphocytes and may encode a homeostatic negative regulator of the size of cell populations (or subpopulations) which are derived ultimately from marrow stem cells. As such, it is a potential antioncogene.

Giant G+C% mosaic structures of the human genome found by arrangement of GenBank human DNA sequences according to genetic positions

To determine the overall variation in the G+C% distribution over long ranges of the human genome, DNA sequences of human genes, which were closely linked genetically or physically, were surveyed from the GenBank Data Bank. A total of 72 sequences longer than 2 kb, which were mutually linked within 500 kb, were identified. The sequences belonged to 17 linkage groups and were ordered in each group according to their genetic positions. Analyses of the G+C% distribution along the ordered sequences showed that sequences within each group almost always had similar G+C% levels, but those belonging to different groups often had different levels. Similar analyses of more distantly linked sequences (e.g., greater than 10 Mb) showed mosaic structures of G+C% distribution. These findings are consistent with predictions made from the "isochore" structures found by CsCl equilibrium centrifugation, in that the structures having homogeneous base compositions stretched over at least several hundred kilobases. A possible boundary of the giant G+C% mosaic structures was identified between X-linked G6PD and F8C.

Analysis of tyrosine hydroxylase and insulin transcripts in human neuroendocrine tissues

We have previously shown that the mRNA for human tyrosine hydroxylase (TH) exists in one of 4 forms as a result of alternative splicing of intron 1. In order to determine the tissue-specific expression of the multiple human transcripts we have utilized specifically primed polymerase chain reactions (PCR) in combination with reverse-transcribed RNA. Using PCR analysis we determined that many human neuronal tissues express all 4 forms of human TH (hTH-1-4); however, hTH-3 and -4 are generally expressed at less than 1% of the other two. Of all tissues examined, only pancreatic beta-islet cells expressed a single form of human TH mRNA (hTH-2). The mRNA for rat TH is present in only one form, the equivalent of hTH-1. Since we have shown that the human genes for TH and insulin are only 2.7 kb apart and several groups have reported insulin-like expression in neuronal tissues, we looked for insulin transcription in the same central and peripheral nervous system samples. Insulin mRNA was not present within the limits of detection.