Skip to main content
NCBI home page
Genome Biology and Evolution logoLink to Genome Biology and Evolution
. 2026 Apr 13;18(5):evag093. doi: 10.1093/gbe/evag093

Evidence for Non-optimal Codon Choice in Highly Transcribed Sex-Biased Genes of Drosophila melanogaster

Carrie A Whittle 1, Cassandra G Extavour 2,3,✉,b
Editor: Carolin Kosiol
PMCID: PMC13148536  PMID: 41974123

Abstract

Biases in synonymous codon use occur in many unicellular and multicellular organisms. Optimal codons, defined as those most commonly used in highly transcribed genes, are thought to arise from selection for cost-efficient translation, which would favor codons with abundant matching tRNAs. Such presumed selection is described as optimal codon choice. Non-optimal codons, defined as those least commonly used in highly transcribed genes, may in principle also play important roles, but the dynamics of their use remain understudied. Here, we examine non-optimal codon use using sex-biased genes expressed in the gonads of Drosophila melanogaster as a case study. We show that genes with sex-biased expression exhibit a preference for non-optimal codon use, especially testis-biased genes. Further, we show that the use of non-optimal codons is not random. Instead, specific non-optimal codons are favored, again especially in testis-biased genes. Non-optimal codon use is positively linked to elevated disorder of the encoded proteins. Remarkably, all 18 degenerate amino acids were associated with higher disorder when encoded by the identified primary non-optimal codon, than when encoded by its sister optimal codon. We hypothesize that selection may have promoted non-optimal codon choice for a subset of favored non-optimal codons to regulate translation. We discuss the putative roles of tRNA gene copy numbers, pleiotropy, and sex-biased expression in the evolution of this level of gene regulation.

Keywords: non-optimal codons, sex-biased expression, protein folding, tRNAs, Drosophila

Significance Statement.

Optimal codons, those codons most commonly used in highly expressed protein-coding genes, are thought to improve translational efficiency in a range of organisms. However, relatively minimal attention has been given to non-optimal codons, those least often used in highly transcribed genes, and their potential roles in translation. Here, using sex-biased gonadal genes of Drosophila melanogaster as a model system, we demonstrate that non-optimal codons are preferentially used in sex-biased genes, particularly within highly expressed testis-biased genes and are associated with disordered proteins. Moreover, the preference for a specific non-optimal codon (per amino acid) is ubiquitous across all degenerate amino acids. We propose that non-optimal codon use is non-random and may have evolved under selection for roles in translational regulation and protein folding in sex-biased genes, and in a manner associated with their tRNA abundances. The findings have significant implications for the understanding of pacing of translation, protein conformation, and protein functionality.

Introduction

The genetic code contains 18 amino acids encoded by synonymous codons (Ikemura 1981; Sharp et al. 1986; Shields et al. 1988; Plotkin and Kudla 2011; Liu et al. 2021). Optimal codons are defined as those codons most frequently used for amino acids in abundant transcripts and are thought to promote cost efficient or more accurate translation, a phenomenon often described as optimal codon choice (Ikemura 1981; Hershberg and Petrov 2009). To date, optimal codons have been identified in unicellular (Ikemura 1981, 1985; Akashi 2003) and some multicellular organisms, including insects such as flies, beetles, crickets, and mosquitoes (Ikemura 1985; Shields et al. 1988; Moriyama and Powell 1997; Hambuch and Parsch 2005; Behura and Severson 2011; Williford and Demuth 2012; Whittle and Extavour 2017; Galtier et al. 2018; Whittle et al. 2019, 2021). The optimization of codons in highly expressed genes is thought to have resulted from adaptation of the coding sequence (CDS) to favor use of the most common tRNAs in the cells, often measured as the number of matching tRNA genes in the genome, which gives rise to a high ratio of tRNA supply to codon demand (Percudani et al. 1997; Duret 2000; Akashi 2003; dos Reis et al. 2004; Gingold et al. 2012; Zhou et al. 2013; Du et al. 2017). However, a number of studies have found that some optimal codons have no tRNAs (wobble codons) in certain taxa, a scenario that may slow translation (Stadler and Fire 2011; Whittle et al. 2019, 2021). While studies in multicellular taxa have largely focused on optimal codons (Ikemura 1981; Duret and Mouchiroud 1999; Duret 2000; Akashi 2001; Cutter et al. 2006; Whittle et al. 2007; Ingvarsson 2008; Hershberg and Petrov 2009; Behura and Severson 2011), non-optimal codons, defined as those least commonly used in highly expressed genes, may also have translational roles (Tuller et al. 2010; Wang et al. 2011; Zhou et al. 2015; Fu et al. 2016; Brule and Grayhack 2017; Whittle et al. 2019, 2021; Liu et al. 2021). When non-optimal codons have few or no matching tRNA genes, some data suggest that their roles could include targeted slowing of translation for cotranslational protein folding or to prevent ribosome jamming (Letzring et al. 2010; Tuller et al. 2010; Pechmann and Frydman 2013; Yu et al. 2015; Zhou et al. 2015; Stein and Frydman 2019; Liu et al. 2021). In addition, under a somewhat paradoxical scenario, some non-optimal codons have been found to have abundant tRNAs and yet are not used as optimal codons. In these cases, this high tRNA abundance to codon use ratio could lead to rapid and preferential translation of specific mRNAs, namely, those mRNAs with unusually high use of the non-optimal codons (Torrent et al. 2018; Guimaraes et al. 2020; Whittle et al. 2019, 2021). At present, however, studies on non-optimal codon use and its potential molecular consequences remain sparse. Key outstanding questions include whether some non-optimal codons may be favored over others, that is, whether non-optimal codons may exhibit non-random use, and if so, whether favored non-optimal codons may be linked to specific properties of the proteins whose transcripts deploy them.

An attractive model system to study codon use, particularly non-optimal codon use, is sex-biased genes expressed in the gonads of multicellular organisms. Because these genes are thought to play an important role in reproductive success, factors influencing their expression and protein functionality may be especially relevant to fitness. Indeed, pioneering studies in the insect D. melanogaster and in certain plants have shown that optimal codon use differs between sexes and tends to be lower in male- than in female-biased reproductive genes (Zhang et al. 2004; Hambuch and Parsch 2005; Whittle et al. 2007). Thus, gonadal genes exhibit sufficient selection differences in codon use to be detectable on a gene-wide level (e.g. Fop, or frequency of optimal codons; Bahiri-Elitzur and Tuller 2021; Zhang et al. 2004; Hambuch and Parsch 2005; Whittle et al. 2007) and provide an informative system to fill gaps in our knowledge on the dynamics of non-optimal codon use. D. melanogaster as a study system is especially valuable given its vast genomic and gene expression resources available for study (Graveley et al. 2011; Li et al. 2014; Gramates et al. 2022).

Some previous observations suggest that the dynamics of non-optimal codon use in genes may be dependent on protein-folding properties (Zalucki and Jennings 2007; Yu et al. 2015; Liu et al. 2021). For example, studies of the circadian clock proteins FQR (in Neurospora) and PER in D. melanogaster have found that experimentally changing non-optimal codons to optimal codons in disordered regions impaired protein form and function, a pattern not found upon optimization of codons in structurally ordered regions (α-helixes) (Yu et al. 2015; Zhou et al. 2015; Fu et al. 2016). Thus, non-optional codons in those two genes may contribute to proper protein structure, particularly for disordered regions (Yu et al. 2015; Zhou et al. 2015). Further, changes in a single synonymous codon, including non-optimal codons, have been linked to altered protein folding and disease phenotypes (Kimchi-Sarfaty et al. 2007; Letzring et al. 2010; Fu et al. 2016; Goodarzi et al. 2016; Kirchner et al. 2017). Thus, the extent and accuracy of protein folding may be a factor linked to a gene's overall non-optimal codon use (Zhou et al. 2015) or potentially to the use of specific non-optimal codons. One approach to quantify the contribution to protein folding of individual amino acid sites (and thus, on a codon-by-codon level) and also on a gene-wide basis is to assess the relative solvent accessibility (RSA). The RSA is an index that quantifies the exposed surface area of an amino acid within the protein structure (that reflects secondary and tertiary structures), relative to its maximum possible surface area in Angstroms (Fraczkiewicz and Braun 1998; Tien et al. 2013). Low RSA values typically indicate deeply buried hydrophobic amino acids or protein regions, while exposed residues at the protein surface are typically hydrophilic with relatively high RSA values. RSA analyses have proven effective in revealing links between protein folding and the rate of protein sequence evolution, including slow protein evolution in highly ordered (low RSA) regions (Conant and Stadler 2009; Franzosa and Xia 2009; Ramsey et al. 2011; Scherrer et al. 2012; Echave et al. 2016; Moutinho et al. 2019; Bricout et al. 2023). In this regard, it is reasonable to infer that RSA may also provide an effective means to study protein folding properties with respect to optimal and non-optimal codon use.

Here, from the study of highly expressed sex-biased genes in D. melanogaster, we provide evidence of non-random non-optimal codon use, which is consistent with selection for the use of non-optimal codons. Specifically, we report a stepwise increase in non-optimal codon use from highly transcribed sexually unbiased, to ovary-biased, to testis-biased genes. We show that certain non-optimal codons are markedly preferred over other sister non-optimal codons, especially in the testis-biased genes. In addition, genes with the highest frequency of non-optimal codon use (>60% of amino acids) show significantly elevated gene wide RSA, and this pattern is most prevalent in testis-biased genes. We further reveal that for a single amino acid, the most strongly favored non-optimal codon per amino acid is consistently linked to elevated RSA as compared to its sister optimal codon, a pattern observed for 18 of 18 degenerate amino acids in the genetic code. The favored use of certain non-optimal codons in testis-biased genes includes both codons with few and with plentiful matching tRNA gene copy numbers. This suggests that selection for specific non-optimal codons in testis-biased genes could enhance at least two potential functions, namely slowed translation and targeted up-translation (Stein and Frydman 2019; Liu et al. 2021; Whittle et al. 2021). We hypothesize that the use of non-optimal codons in sex-biased genes, and especially in male-biased genes, may be mediated by very low cross-tissue gene pleiotropy (Otto 2004; Yanai et al. 2005; Mank and Ellegren 2009), and by weak or absent potential for sexual conflict (Ellegren and Parsch 2007; Wong and Holman 2023).

Results and Discussion

A Priori Classifications of Organismal Optimal and Non-optimal Codons in D. melanogaster

Prior to analyses of sex-biased gene sets and codon use, we first aimed to define the organism-wide optimal and non-optimal codon status for D. melanogaster for each of the 59 codons encoding the 18 degenerate amino acids. For this, we compared the relative synonymous codon usage (RSCU) (Sharp and Li 1987) of a set of highly expressed (ribosomal protein genes) versus all genes in the genome and defined this value as the ΔRSCURibosome-All (see Materials and Methods; Subramanian et al. 2022). This approach allows a means to identify the optimal (most common), and the non-optimal (least common) codons used in highly expressed genes in the genome (as compared to low or moderately expressed genes), which is the foundation for our subsequent analysis of codon use in highly transcribed sex-biased genes (Table S1). All codon classifications obtained with this method, and their tRNA abundance statuses, are provided in Table S1. We identified optimal codons as universally G3 and C3 codons, as previously suggested for D. melanogaster (Shields et al. 1988; Duret and Mouchiroud 1999; Hambuch and Parsch 2005; Heger and Ponting 2007). This pattern of codon use is not explainable by mutational biases but rather likely arose from selection pressures (Shields et al. 1988; Duret and Mouchiroud 1999; Heger and Ponting 2007). Herein, for our study, for each of the 18 degenerate amino acids, we explicitly defined the (primary) optimal codon (Whittle et al. 2019, 2021) as that with the largest positive ΔRSCURibosome-All value compared with its sister codons (Table S1). All other codons for a given amino acid (a total of 41) were classed as non-optimal codons. These typically had a negative or near-zero ΔRSCU value (and an RSCU value 1.5-fold to 6-fold smaller than that of the optimal codon). We then focused on further analysis of these non-optimal codons, which have been minimally considered or excluded in codon use studies to date (Shields et al. 1988; Duret and Mouchiroud 1999; Hambuch and Parsch 2005; Heger and Ponting 2007) (see Supplementary Text File S1 for details).

We classified all optimal and non-optimal codons based on the number of exact matching tRNA genes in the D. melanogaster genome (Table S1, GtRNA database; Chan and Lowe 2016), because tRNA abundances, often well reflected by gene copy numbers, underlie purported translational functions (Percudani et al. 1997; Duret 2000; Akashi 2003; Stadler and Fire 2011; Gingold et al. 2012; Du et al. 2017; Stein and Frydman 2019; Whittle et al. 2019). Twelve of the 18 optimal codons had the most abundant perfectly matching tRNA genes among its sister codons (e.g. GCC for Gly with 14 tRNAs; denoted as Opthigh-tRNAs status; high/low tRNAs were demarcated by the median of 5 tRNA genes). The other six optimal codons had no matching tRNAs and were classed as wobble codons (Optwobble; e.g. Pro-CCC). In turn, 26 of the 41 non-optimal codons had low tRNA numbers (Non-optlow-tRNAs, status; e.g. ATA for Ile), and the remaining 15 had abundant tRNA genes (Non-opthigh-tRNAs status; e.g. ACT for Thr) (Table S1). Hereafter, we focus on the study of sex-biased genes and their patterns of optimal and non-optimal and optimal codon use (Table S1). When pertinent, we also consider statuses of their tRNA gene abundances.

Identification and Characterization of the Sex-Biased Gene Sets

We identified highly expressed testis-biased (N = 916), ovary-biased (N = 258), and unbiased (N = 605) genes in D. melanogaster (Fig. S1a) using the strict criteria that a gene have ≥100 RPKM and at least 5-fold-bias in one sex versus the other to be classified as sex-biased (testis vs. ovaries [mated]), where genes classified as unbiased had <5 fold bias (and ≥100 RPKM in at least one sex), based on transcriptome data from modEncode (Table S2; Graveley et al. 2011; Gramates et al. 2022 see Materials and Methods). For stringency, we also compared gene expression in the male accessory glands and virgin ovaries (Table S2). We found that 86.9% of the identified testis-biased genes were also male-biased in accessory glands (using a cutoff of ≥ 5-fold sex-bias) and that 77.1% of the identified ovary-biased genes were also female-biased in virgin ovaries, indicating that strong male–female transcriptional differences were also present in those alternate reproductive tissue types. The predicted functions of these genes based on GO analysis showed testis-biased genes were strongly linked to testicular roles such as spermatogenesis, and ovary-biased genes were associated with female reproductive process roles such as the chorion and oogenesis. In contrast, unbiased genes were preferentially involved in the ribosome and other housekeeping functions (Table S3; chromosomal locations of genes are discussed in Supplementary Text File S1). In terms of characteristics of these genes with sex-biased expression, the CDS lengths were shorter for testis-biased (median = 273 codons) and unbiased genes (median = 269 codons) than for the ovary biased genes (median = 498 codons; MWU tests P < 0.05) (Fig. S1b). We then determined the tau value, an index of gene expression specificity with values from 0 to 1 (Yanai et al. 2005), across 59 tissue stages in D. melanogaster (Table S2). tau exhibited a stepwise decline from testis-biased genes (median = 0.95) to ovary-biased genes (median = 0.83) to unbiased genes (median = 0.74) (ranked ANOVA and Dunn's P < 0.05 for each paired contrast; median genome-wide tau  = 0.85, Fig. S1c). Thus, the testis-biased genes exhibited higher specificity, and lower cross-tissue pleiotropy (Mank and Ellegren 2009; Meisel 2011; Assis et al. 2012), than ovary-biased and unbiased genes.

We assessed the relationships between tRNA gene numbers, amino acid use, and amino acid size complexity (S/C) scores, which are thought to reflect biosynthetic costs of amino acids (provided in Table S4;  Dufton 1997). The tRNA gene copy number per amino acid (pooled across all codons) was strongly positively correlated to the amino acid use (percentage of each amino acid relative to all concatenated amino acids per gene set) for all genes (testis-biased genes: Spearman's R = 0.91, P < 10−7; ovary-biased genes: R = 0.85, P < 10−7; unbiased genes: R = 0.83, P < 10−7; Fig. S2a–c). Moreover, the S/C score per amino acid (Dufton 1997) was inversely correlated to the organismal tRNA gene numbers (R = −0.59, P < 10−7, Fig. S2d). In sum, the patterns suggest that all three gene sets have a history of selection on codon use (Akashi 2003; Du et al. 2017; Whittle et al. 2021), further affirming their utility as a model to study dynamics of codon use in this taxon.

By examining the codon adaptation index (CAI), an index that reflects the gene-wide codon frequencies relative to a highly expressed reference gene set (here, ribosomal genes) (Sharp and Li 1987; Subramanian et al. 2022), we found a stepwise decrease in CAI from unbiased, to ovary-biased, to testis-biased genes (ranked ANOVA and Dunn's P < 0.05, Fig. S3a; note that CAI was highly correlated to Fop across the genome, Spearman's R = 0.967, P < 10−7). In other words, as genes exhibited increased gonad-biased expression, they also presented increased non-optimal codon use, with the highest non-optimal codon use in testis-biased genes (see Supplementary Text File S1). Further, within our highly expressed sex-biased genes sets (the targeted gene sets herein), we subdivided expression into two classes, ≥200 RPKM (extremely high expression), and ≥100 to 200 RPKM (high expression) and subdivided genes by CDS length (short and long CDS lengths, with long defined as ≥272 codons; cutoff at the 33rd percentile of all gene lengths), which were linked to CAI within each of the sex-biased gene sets (Fig. S3b). The differences in CAI between gene sets (Fig. S3a) were robust to both transcription level and length cutoffs (Fig. S3a and b; see also Supplementary Text File S1). The high non-optimal codon use in testis-biased genes is not due to enhanced genetic interference, as those processes would be expected to ameliorate non-optimal codon use in longer rather than shorter genes, a pattern that was absent from our data (Fig. S3) (Comeron et al. 1999; Betancourt and Presgraves 2002; Hambuch and Parsch 2005; Loewe and Charlesworth 2007). The testis-biased genes herein were short (Fig. S1b), and even within the testis-biased gene set (Fig. S3b), shorter genes (<272 codons) had greater non-optimal codon use (lower CAI) than longer genes for both expression classes ((MWU-tests P < 0.05, Fig. S3b; see Supplementary Text File S1). Thus, other factors must explain the stepwise increase in non-optimal codon use from unbiased, to ovary-biased to testis-biased genes. Moreover, it is notable that the extremely highly expressed testis-biased genes tended to have an elevated degree of sex-bias relative to the highly expressed testis-biased genes (the median was 149.5 and 210 fold bias for long and short genes in the former, and 71.0 and 69.5 fold bias in the latter, consistent with high extent of specialization in the testis in each group), suggesting the more biased the expression in the testis (among genes with ≥100RPKM), the greater the use of non-optimal codons.

Non-optimal Codon Use is Non-random in D. melanogaster Sex-Biased Genes

To further characterize codon use in unbiased and sex-biased genes, we assessed codon use on a codon-by-codon basis for each gene set described above. We determined the ΔRSCU between testis-biased genes versus ovary-biased genes (ΔRSCUTestis-Ovary) and between testis-biased genes versus unbiased genes (ΔRSCUTestis-Unbiased), for each of the 59 codons with synonymous codons (cf. Whittle et al. 2007). In these contrasts, a positive ΔRSCU value indicates higher use of a codon in the testis-biased genes, and a negative value denotes a reduction in its use within testis-biased genes (and thus elevated use in the respective ovary or unbiased genes, see Materials and Methods). The ΔRSCUTestis-Ovary and ΔRSCUTestis-Unbiased values per codon are shown in Table S5. Remarkably, we found that testis-biased genes reduced their use of optimal codons, and this pattern was universal to all of the degenerate amino acids. Specifically, 16 of the 18 optimal codons in D. melanogaster had a statistically significant negative ΔRSCUTestis-Ovary (P < 0.05, only exceptions were for Ser and Thr P > 0.05), and 18 of 18 optimal codons had a negative value for ΔRSCUTestis-Unbiased. The magnitude of negative ΔRSCU values was consistently larger for the ΔRSCUTestis-Unbiased contrast than for the ΔRSCUTestis-Ovary contrast, consistent with a stepwise decline in the use of optimal codons from unbiased genes to ovary-biased genes to testis-biased genes (Table S5, Fig. S3). This suggests the possibility that there are weaker translational selection pressures for optimal codons in both types of sex-biased genes and especially in the testis-biased genes. However, another significant hypothesis worthy of consideration here is that the patterns may reflect a non-random displacement of optimal codons by the favored use of non-optimal codons in testis-biased genes. In other words, it could be the case that testis-biased genes reduce their use of optimal codons not arbitrarily, but rather in favor of preferential use of non-optimal codons.

To test this hypothesis, we asked whether there was any pattern to the non-optimal codons that were used in testis-biased genes. We found that for the nine amino acids encoded by two synonymous codons (Asn, Asp, Cys, Gln, Glu, His, Lys, Phe, Tyr), the universal reduction in use of the optimal codon in testis-biased genes was compensated by increased use of the sister non-optimal codon (as indicated by positive ΔRSCUTestis-Unbiased and ΔRSCUTestis-Ovary values, Table S5), consistent with a hypothesis of preferential use of the non-optimal codons in testis-biased genes. Nonetheless, the 2-fold degenerate amino acids cannot be used to fully differentiate whether there may also be codon choice among the non-optimal codons. Thus, we examined the nine amino acids that were encoded by three, four, or six synonymous codons (Ala, Arg, Gly, Ile, Leu, Pro, Ser, Thr, Val). If the low use of optimal codons in testis-biased genes were solely due to weaker selection for optimal codons, then one would expect a similar degree of enhanced use of all the synonymous non-optimal codons per amino acid in testis-biased genes (under a neutral model). In contrast, however, we found that the elevated use of non-optimal codons in testis-biased genes was nonarbitrary (Fig. 1a, Table S5). As an example, for the 4-fold degenerate amino acid Thr, a neutral model would predict similar and increased use of the three non-optimal codons (ACG, ACA, ACT) under reduced use of the optimal codon (ACC). However, we found that only ACT (Non-opthigh-tRNAs) had elevated use in testis-biased genes versus ovary-biased genes (ΔRSCUTestis-Ovary =+0.166; t-test P < 0.05. Figure 1b), while the other two non-optimal codons showed no elevation in use. This is not consistent with a neutral model of non-optimal codon use and instead suggests a preference for one specific non-optimal codon over the others (Fig. 1b). The same preference for ACT was also observed using the ΔRSCUTestis-Unbiased comparison (ΔRSCU=+0.201, Table S5), a codon that was detected even though the ovary-biased and unbiased genes were from fully non-overlapping gene sets, indicating that the preference for this non-optimal codon in the testis-biased genes is robust to the two contrasts. Another example is Ile (three-fold degenerate) where the reduced frequency of the optimal codon ATC in testis-biased versus unbiased genes contrast was accompanied by a strong preferential use of just one of the two non-optimal codons (ATA; ΔRSCUTestis-Ovary =+0.114; 9.-fold higher than that of its synonymous non-optimal codon ATT; Table S5), and the same non-optimal codon preference was also identified using the ΔRSCUTestis-Unbiased contrast (value was +0.266. t-test P < 0.05, Fig. 1c, Table S5). While the same primary non-optimal codons (largest positive ΔRSCU) tended to occur in both the ΔRSCUTestis-Ovary and the ΔRSCUTestis-Unbiased contrasts, the latter were more often statistically significant (Fig. 1, Table S5). This likely reflects not only the larger ΔRSCU values in the latter contrasts but also the larger N values and thus power for the unbiased gene set (N = 605 for unbiased, and 258 for ovary-biased). Overall, the pattern of the highly favored use of a specific non-optimal codon over others per synonymous codon family in testis-biased genes was also observed for Gly (GGA, Fig. 1d), Ala (GCA), Leu (TTG), Ser (AGT) and Val (GTA as primary non-optimal codon, GTT as secondary), Pro (CCT, particularly ΔRSCUTestis-Ovary contrast), and Arg (AGG as primary, AGA as a close secondary) (Table S5, Fig. 1a). All primary non-optimal codons ended in A3, T3, or G3 (Table S5), thus excluding a role of mutational biases for a particular nucleotide at the 3rd “silent” position and suggesting instead that selective processes have influenced the use of those codons. Notably, the identified primary non-optimal codon per amino acid always had a positive value for both ΔRSCUTestis-Ovary and for ΔRSCUTestis-Unbiased (P < 0.001 for all latter contrasts) but was typically smaller in value for the former contrast, as expected if non-optimal codons are preferred in sex-biased over unbiased genes (and also more favored in testis-biased than ovary-biased genes). Together, these patterns show that testis-biased genes exhibit lowered use of use optimal codons and preferentially use specific non-optimal codons, rather than randomly choosing among the available non-optimal codons.

Fig. 1.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

a) The values for ΔRSCUTestis-Ovary (left) and ΔRSCUTestis-Unbiased (right) per codon for amino acids with three or more synonymous codons (P-values in Table S5). The amino acid-codon and the codon status (left) with respect to tRNAs (right) are each indicated. Close-up examples of the ΔRSCU for: b) all four codons for Thr; c) all three codons for Ile; and d) all four codons for Gly (* indicates P < 0.05 in b–d). In panel a, ΔRSCUTestis-Unbiased for Leu-CTG Opthigh-tRNAs =0.818 and Arg-CGC Optwobble = 1.11 and these endpoints are not shown as they are outside the limit of X-axis.

Notably, the primary non-optimal codon per amino acid (Table S5, Fig. 1a) in testis-biased genes was not predicted by tRNA status. Specifically, for Arg, Leu, and Ser, the primary non-optimal codon had low tRNA abundance (Non-optlow-tRNAs status), while others such as Gly, Pro, and Thr favored the use of codons with highly abundant tRNAs (Non-opthigh-tRNAs status; Table S5). We speculate that this pattern suggests that selection of non-optimal codons with few tRNAs (Arg, Leu, and Ser) may typically be involved in slowing translation for protein folding or to limit ribosome jamming, while those with abundant tRNAs (Gly, Pro, and Thr) may often be involved in ensuring up-translation of mRNAs carrying those codons. These predictions are based on presumed supply-demand ratios of tRNAs:codons (Brackley et al. 2011; Zhou et al. 2015; Brule and Grayhack 2017; Stein and Frydman 2019; Whittle et al. 2021). Thus, we suggest that the primary non-optimal codons identified here could potentially regulate gene function by influencing translation. It should be noted that the tRNA populations for D. melanogaster closely match those of its sister species D. simulans, D. sechellia, D. yakuba, and D. erecta (Spearman's R ≥ 0.998 [P < 10−7]) and thus these tRNA populations have existed for a substantial period of time (at least 13 Mya since divergence; Tamura et al. 2004; see Supplementary Text File S1 for details).

We determined the percentage of genes in each gene set under study that exhibited extreme use of the primary non-optimal codons (defined as RSCU values ≥1.5). As shown in Fig. S4a, a larger percentage of testis-biased genes exhibited extreme use of the primary non-optimal codons than did the comparable ovary-biased and/or unbiased genes (Chi-square P < 0.05; shown as examples are Thr-ACT, Gly-GGA, Ile-ATA and Val-GTA; see Supplementary Text File S1 for parallel data for optimal codons, Fig. S4b). It is worth noting that the percent of genes with extreme use of optimal codons with plentiful tRNAs differed from those optimal codons with wobble status (Fig. S4b; in all gene sets, P < 0.05). This observation is important as it allows us to infer that these two classes of optimal codons may not be equivalent in functions (see Supplementary Text File S1) and thus in future studies should not be treated as equal optimal codons nor studied without including tRNA gene data.

Amino Acids Using Non-optimal Codons Have High Surface Exposure and Disorder

To determine whether these patterns of non-optimal codon use were linked to specific biophysical properties of the encoded proteins, we first assessed the solvent accessibility RSA value for every amino acid (and its associated codon) in the testis-biased genes (N = 299,817 amino acids), ovary-biased genes (155,282 amino acids), and unbiased genes (218,580 amino acids). We found that RSA value was strongly correlated to the disorder probability scores (from NetsurfP; Hoie et al. 2022) in all three datasets (Spearman's R ≥ 0.68 for each of three datasets, P < 10−7). Thus, higher RSA reflects both surface exposure and intrinsic disorder (Akdel et al. 2022; Ilzhöfer et al. 2022; Piovesan et al. 2022). In terms of major secondary structures of proteins (α-helices, β-strands and random coils; Hoie et al. 2022), a majority of amino acids in the proteins of all three gene sets were located within random coils (60.6% for testis-biased, 57.2% of unbiased, and 70.8% for ovary-biased genes, Chi-square tests P < 0.05). α-helices were the next most common structures in all three gene sets (31.1, 22.4% and 31.5%, respectively), and β-strands the least common (11.2, 8.3% and 6.8%, Fig. S5a). The lowest RSA values occurred for amino acids located in β-strands, while these values were intermediate for α-helices and highest for random coils for all three gene sets (Fig. S5b, median RSA values across all residues for testis-biased genes were 0.13, 0.34, and 0.53, for ovary-biased genes were 0.11, 0.33, and 0.60 and for unbiased genes were 0.11, 0.30, and 0.57 for β-strands, α-helices and random coils, respectively; MWU-tests P < 0.05 for all contrasts per gene set). These patterns are consistent with disordered regions having the greatest surface exposure.

For each gene under study, we then determined the percent of amino acids encoded by non-optimal codons, as a gene-wide direct measure of non-optimal codon use (denoted as Percent-Non-opt). The genes were placed into bins of low (≤50%), moderate (>50 to ≤60%), and high Percent-Non-opt codon use (>60%; note that we excluded Met and Typ from this analysis, discussed in Supplementary Text File S1). We found the average RSA per gene increased from the lowest Percent-Non-opt bin (median RSA = 0.40), to the moderate (median = 0.43) to high (median = 0.47) bins (MWU-tests each P < 0.05; note the variation within each group within the box plots) in testis-biased genes (Fig. 2a). In other words, the elevated non-optimal codon use in these testis-biased genes tended to be linked to greater disorder in their encoded proteins, whereas genes using more optimal codons tended to be more well-folded (Fig. 2a). For ovary-biased genes and unbiased genes, no statistically significant differences were observed in RSA between the highest Percent-Non-opt bin and the moderate and low bins (P > 0.05, Fig. 2b and c), likely because the highest non-optimal codon class was so uncommon in those gene sets for testing (N ≤ 23 genes in each case). Nonetheless, the moderate Percent-Non-Opt bins had elevated average RSA per gene as compared to the low bins for both the ovary-biased and the unbiased genes (MWU-tests P < 0.05). Overall, elevated non-optimal codon use is associated with greater protein surface exposure in testis-biased, ovary-biased and unbiased genes, with the most striking stepwise correlation observed for the testis-biased genes (Fig. 2a–c).

Fig. 2.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

a–c) The relationships between high non-optimal codon use (>60% of amino acids), moderate non-optimal use (>50 to ≤60%), and low non-optimal codon use (≤50%) and the relative solvent accessibility for: a) testis-biased genes; b) ovary-biased genes; and c) unbiased genes in D. melanogaster. Pie charts of the number of genes with d) high non-optimal; e) moderate non-optimal; and f) low non-optimal codon use, also showing the median values of tau across genes. Examples of AlphaFold2 predicted protein structures of a typical: g) testis-biased gene with high non-optimal codons and RSA; h) ovary-biased gene wth moderate non-optimal codons and RSA; and i) an unbiased gene with low non-optimal codons and RSA. In a–c, different letters below bars indicate a statistically signficant difference (MWU-test P < 0.05).

Importantly, of the 255 genes (out of 1,779 highly expressed genes under study; Fig. S1a) within the highest Percent-Non-opt class (>60% non-optimal codons, Fig. 2d), 83% were testis-biased (N = 212), while only 9.0% were ovary-biased (N = 23) and 7.8% were unbiased (N = 20; Fig. 2d). Thus, the highest degree of non-optimal codon use is most commonly observed in the testis-biased genes and is much less common among ovary-biased and unbiased genes. Given that non-optimal codons, when they occur, tend to be found in higher RSA regions (Fig. 2a–c), this implies a concentration of non-optimal codons in the high RSA regions of testis-biased genes (i.e. non-optimal codons less common overall in ovary-biased or unbiased genes). Even when taking into account that there were more testis-biased genes than ovary-biased or unbiased genes, we find that 23.1% (212 of 916) of the testis-biased gene set belongs to the highest class bin of non-optimal codon use, versus only 8.9% (23 of 258) and 3.3% (20 of 605) of the ovary-biased and unbiased gene sets, respectively. In terms of genes in the moderate Percent-Non-opt bin (Fig. 2e), testis-biased genes were most common in absolute numbers (N = 325; that 35.4% of all testis-biased genes), with less than a third as many ovary-biased genes (N = 99 of 258, 38.3% of the ovary genes), and even fewer unbiased genes (N = 78; 12.9% of unbiased genes). In the lowest Percent-Non-opt bin (with the highest optimal codon use), unbiased genes were most strongly represented with 509 genes (84.1% of the unbiased gene set), while 389 were testis-biased (42.5% of testis-biased gene set), and 136 were ovary-biased genes (52.7% of the ovary-biased gene set; Fig. 2f). Overall, it is evident that a major portion of testis-biased genes were linked to the highest non-optimal codon use bin (Percent-Non-opt > 60%), a group that also had very high protein-wide RSA (Fig. 2a–c). These RSA patterns add another layer of support (in addition to the results in Fig. 1) to the inference that the high non-optimal codon use in D. melanogaster genes is non-random, given that it is also negatively associated with the degree of protein folding in all gene sets under study (Fig. 2a–c) and that that the highest degree of non-optimal codon use (and thus high RSA) is especially common in testis-biased genes (Fig. 2a and d).

In Fig. 2g–i, we provide examples of the predicted structures of a testis-biased gene from the highest Percent-Non-opt bin, an ovary-biased gene from the moderate Percent-Non-opt gene bin, and an unbiased gene from the low Percent-Non-opt bin (note RSA variation within each bin, Fig. 2a–c). The testis-biased sperm/accessory gland-related gene protein Mst57 Da (FBgn0011668), a relatively short protein, encodes 73.0% of its amino acids with non-optimal codons, and is predicted to be highly unstructured (Fig. 2g) with an average CDS-wide RSA of 0.72 ± 0.09 (for an example of a similar weakly folded protein structure for a longer testis-biased gene, BG642163, in the highest Percent-Non-opt bin see Fig. S6). The ovary-biased gene protein Jabba (FBgn0259682), which is involved in early embryogenesis (Gramates et al. 2022), has 55.1% of its amino acids encoded by non-optimal codons, and shows more helical structures and a lower RSA value (0.50 ± 0.01; Fig. 2h) than the testis-protein (Fig. 2g). The unbiased gene encoding ribosomal protein RpL3 (FBgn0020910) has 73.2% optimal codon use and an average RSA of 0.36 ± 0.01 (Fig. 2i) and high levels of helical and beta-sheets illustrating its extensive degree of folding (NetsurfP predictions for all three proteins are in Fig. S7). In sum, these patterns exemplify the tendency for positive associations in non-optimal codon use and RSA.

Functional clustering shows that the 212 testis-biased genes in the highest Percent-Non-opt bin (>60% non-optimal codons, Fig. 2d) are enriched in genes involved in sperm-related functions (Table S6). Examples include genes encoding accessory gland proteins Acp54A1, Acp26Aa, and Acp36DE, the testis-enriched gene RpL10Aa, the seminal fluid protein Sfp87B, and spermatogenesis genes soti, Mst89B, Mst57Da, and other Mst genes (Table S7; gene functions from DAVID and FlyBase; Gramates et al. 2022; Sherman et al. 2022). Sperm and accessory gland genes (Tables S6 and S7; an example structure in Fig. 2g) are of particular interest to evolutionary biologists due to their key role in sexual reproductive success, and their often-rapid rates of protein sequence divergence, including in Drosophila (Pitnick 1996; Civetta and Clark 2000; Torgerson et al. 2002; Dorus et al. 2006). Overall, these testis-biased genes with the highest percentage of non-optimal codons were highly specialized to testicular functions, including accessory gland and sperm functions.

Tissue Specificity Suggests a Mechanism to Accumulate Non-optimal Codons

Analyses of gene expression specificity across 59 tissues/stages in D. melanogaster (Table S2) revealed that testis-biased genes had exceptionally elevated tau values, with a median near the maximum possible value of 1 for genes in the high, moderate, and low Percent-Non-opt bins (median values ≥ 0.96, Fig. 2d–f; see Materials and Methods). Thus, the expression of these genes is highly specific to the testis. In turn, ovary-biased genes, had intermediate tau values (median between 0.83 to 0.87) among low, moderate, and high Percent-Non-opt bins, while unbiased genes had the lowest tau (0.71 to 0.77) (Fig. 2d–f). Assuming expression breadth reflects a gene's cross-tissue pleiotropy (Mank and Ellegren 2009; Meisel 2011; Assis et al. 2012), then testis-biased genes comprise the least pleiotropic gene set under study. The low pleiotropy of these testis-biased genes may allow them to be under less severe functional constraint than those involved in a wide range of tissues or pathways (Otto 2004; Mank et al. 2008; Meisel 2011; Dean and Mank 2016), possibly allowing them the flexibility to evolve the observed high non-optimal codon use (Figs. 1 and 2). Moreover, sex-biased gene expression, including sex-specific expression (79.0% of testis-biased genes were sex-specific herein, Fig. S1 legend) may act as a means to limit sexual conflict (Ellegren and Parsch 2007; Ingleby et al. 2014; Wong and Holman 2023). Thus, we predict that non-optimal codon use in the testis-biased genes may typically be of little or no fitness consequence to the ovaries or other tissues, facilitating accumulation of non-optimal codons in testis genes to regulate gene expression for specialized testicular functions.

All 18 Degenerate Amino Acids Have Higher RSA Values When Encoded by the Primary Non-optimal Codon Than by the Optimal Codon

Finally, we assessed the RSA value for each individual type of amino acid when it was encoded by the primary non-optimal codon (Table S5) versus when it was encoded by the optimal codon (Table S1) in the testis-biased genes. Our analyses of all amino acids encoded by testis-biased genes showed that for all 18 degenerate amino acids, the primary non-optimal codon was linked to a higher average RSA than the primary optimal codon (Fig. 3; sign test P < 0.001; averages and standard errors are in Table S8). Thus, the elevated RSA of non-optimal codons included every degenerate amino acid, regardless of its maximum solvent accessibility and its polarity (which vary among amino acids; Tien et al. 2013). When encoded by a non-optimal codon, residues were consistently located in less deeply folded protein regions, than when that same amino acid was encoded by its optimal codon. This result includes the primary non-optimal codons Thr-ACT, Ile-ATA, and Gly-GGA (Fig. 2b–d) and spans all 2-fold, 3-fold, 4-fold, and 6-fold degenerate amino acids (Table S5). Thus, the favored use of the primary non-optimal codons versus the optimal codon in more disordered regions was universal across all 18 degenerate amino acids.

Fig. 3.

For image description, please refer to the figure legend and surrounding text.

Open in a new tab

The average RSA of testis-biased genes when encoded by the primary non-optimal codon (defined in Table S5) versus the optimal codon (Table S1) for all 18 degenerate amino acids. Results are from all amino acids and codons in testis-biased genes (N = 299,817 amino acids). In total, 18 of 18 degenerate amino acids had higher average RSA when encoded by the non-optimal codon; a sign test of the non-optimal versus optimal codon across all 18 amino acids had P < 0.001. The standard errors are provided in Table S8.

In addition, the primary non-optimal codons, which all exhibited elevated RSA compared to optimal codons, included those with few and those with plentiful matching tRNA genes (i.e. with Non-optlow-tRNAs and Non-opthigh-tRNAs statuses) (Fig. 3, Tables S1, S5; optimal codons included both Opthigh-tRNAs and Optwobble statuses). The rarity of exact matching tRNAs for Non-optlow-tRNAs codons (Tables S1, S5, Fig. 3) may make these types of codons poised to slow translation and allow proper co-translational protein folding (Brule and Grayhack 2017; Stein and Frydman 2019), including in disordered regions (Zhou et al. 2015). However, the high ratio of tRNA genes to codon use for Non-opthigh-tRNAs codons suggests that this class of codons likely has different functions, possibly in the rapid translation of mRNA segments containing the codons, or the up-translation of any mRNAs with high use of these (otherwise) rare codons within a cellular mRNA pool (Torrent et al. 2018; Whittle et al. 2019; Whittle et al. 2021). Thus, despite potentially having distinct functions in slowing translation (Non-optlow-tRNAs) or in targeted enhanced translation of mRNAs (Non-opthigh-tRNAs)) (Brackley et al. 2011; Zhou et al. 2015; Fu et al. 2016; Brule and Grayhack 2017; Stein and Frydman 2019; Liu et al. 2021; Whittle et al. 2021), the two types of non-optimal codons (Table S1) are each consistently favored over optimal codons in less-structured protein segments (Fig. 3). Based on these findings, we speculate that the combined use of primary Non-optlow-tRNAs and Non-opthigh-tRNAs codons (Fig. 3) may respectively act to slow down and speed up local rates of translation within specific types of mRNAs (for instance, in highly transcribed mRNAs that have ameliorated use these otherwise rare codons, such as testis and sperm genes; Table S5, Fig. 3, and/or in some regions of ovary-biased and unbiased genes; Fig. 2b and c), thereby providing a means to control translational pacing and potentially allowing proper protein conformation and functionality.

The Potential Role of Selection in Shaping Non-optimal Codon Use

Some of the non-optimal codon use patterns observed here may reflect a history of relaxation of constraint, particularly among the non-primary non-optimal codons per amino acid (Fig. 1, Table S5). However, several lines of evidence support a hypothesis that selection has at least partly contributed to the use of a subset of preferred (primary) non-optimal codons. First, we a priori identified optimal and non-optimal codons from analysis of the most highly expressed genes in the D. melanogaster genome (Table S1), which are most apt to experience translation-related selection pressures (including tRNA abundance pressures), using an approach parallel to that historically used to identify optimal codons (Duret and Mouchiroud 1999; Duret 2000; Akashi 2001; Wang et al. 2011; Whittle and Extavour 2016; Whittle et al. 2019, 2021). This a priori analysis automatically (in Table S1) controls for codon use in lowly and moderately expressed genes, wherein selection is weak or absent and neutral processes tend to be prevalent (Wang et al. 2011; Whittle et al. 2019, 2021). Second, a specific subset of non-optimal codons, typically one specific codon per amino acid, exhibited much higher use than the other sister non-optimal codons in highly transcribed testis-biased genes (Fig. 1, Table S5). We posit that this pattern, remarkably observed for 18 of 18 degenerate amino acids (Table S5), is unlikely to be explained by a fully neutral process, whereby all non-optimal codons would be expected to have similar, or arbitrary use patterns. Third, the same primary non-optimal codon was identified when using the ovary-biased genes and when using the unbiased genes as the reference dataset to compare with testis-biased genes (Fig. 1, Table S5); the ovary-biased and unbiased groups are mutually exclusive gene sets, but the same primary non-optimal codon per amino acid was nevertheless identified in in the testis-biased genes. Fourth, the primary (preferred) non-optimal codon for amino acids ended in different nucleotides (A, T, G; Fig. 1, Table S5) and thus a mutation-bias or gene conversion process (e.g. favoring G) associated with high gene expression in the testis is a poor explanation for this pattern. Fifth, our findings that 18 of 18 degenerate amino acids in D. melanogaster showed preferential location of the primary non-optimal codon in more disordered regions over its sister optimal codon (higher RSA, Fig. 3), infers a remarkably consistent pattern of non-arbitrary codon use among sister codons, across every degenerate amino acid, again unlikely to be the result of a neutral process. Further, we showed this pattern was observed for non-optimal codons that we a priori identified that had enhanced use in highly expressed genes (Table S1), which are most apt to experience translation-related selection pressures (as compared to low or moderately expressed genes, similar to optimal codons in highly expressed genes; Duret and Mouchiroud 1999; Duret 2000; Akashi 2001; Whittle et al. 2007; Wang et al. 2011; Whittle and Extavour 2016; Whittle et al. 2019; Whittle et al. 2021). While neutral processes are apt to play a role in shaping the use of the non-primary non-optimal codons per amino acid for highly expressed testis-biased and ovary-biased genes (for those amino acids with three or more codons and thus, more than one non-optimal codon, Fig. 1), we propose the collective evidence best supports a model whereby one (or rarely two) non-optimal codon is favored over others and that use of these codons tends to be preferentially located in genes and gene regions that encode disordered proteins (Figs. 2 and 3).

Conclusions and Future Directions

The collective evidence herein supports the hypothesis that selection has contributed to shaping the use of a specific subset of (preferred) non-optimal codons. Non-optimal codons are non-arbitrarily used in sex-related genes in D. melanogaster, and certain non-optimal codons are favored over others. This pattern is particularly prevalent in testis-biased genes and suggests that certain (primary, or preferred) non-optimal codons could have functions in regulating translation.

Future studies should assess the prevalence and evidence for selection for non-optimal codon use in other multicellular organisms, including a system whereby non-optimal codons are more common in the ovaries (Whittle and Extavour 2017) or in other specific organ systems. In addition, investigation of the placement of all four categories of codons studied herein, namely Opthigh-tRNAs, Optwobble, Non-optlow-tRNAs, and Non-opthigh-tRNAs, with respect to the starting ramp of translation, versus middle and terminal CDS regions and especially with respect to the RSA in disordered regions, will help reveal whether these codons may cause localized pacing (fast/slow) of translation (Quax et al. 2015; Miller et al. 2019; Stein and Frydman 2019) in D. melanogaster and other multicellular models. Experimental studies to quantify translation rates and protein folding of testis genes with extreme use of Non-opthigh-tRNAs codons (Fig. S4, Table S5; cf. Zhou et al. 2015; Allen et al. 2022) may further reveal their putative roles in the rapid translation of specific mRNAs in a cellular mRNA pool or of mRNA regions (Torrent et al. 2018; Whittle et al. 2021), while investigations of those genes with extreme use of Non-optlow-tRNAs codons (Fig. S4) may inform on their function in the slowed translation and precise protein conformation in testis-involved genes, as inferred for circadian genes (Zhou et al. 2015; Fu et al. 2016; Liu et al. 2021). Additional data on the tRNA populations in the testes and ovaries in this species as the technology improves (Guimaraes et al. 2020) (discussed in Supplementary Text File S1), will also help determine whether or how there is variation in tRNA populations among the gonads, beyond that predicted by tRNA gene frequency (Fig. 2). Future research should also evaluate whether younger testis-biased genes have greater flexibility to evolve a set of preferred non-optimal codons with functional roles (cf. Allen et al. 2022), and whether non-optimal codon use may have particular roles in shorter genes (as testis-biased genes tended to be short herein). Finally, in vitro or in vivo studies that replace non-optimal codons with optimal codons (Zhou et al. 2015; Fu et al. 2016; Allen et al. 2022), and vice versa, to evaluate their effects on protein structure and functions for testis-biased and ovary-biased genes studied here, may provide further insights into the potential specific protein biophysical roles of the non-optimal codons in this species. As part of such analyses, an evaluation of altered Non-optlow-tRNAs and Non-opthigh-tRNAs statuses within disordered protein regions may reveal their individual roles on the pacing of translation, protein conformation, and protein functionality.

Materials and Methods

Genomic Data and Expression Analyses

We downloaded the genome of D. melanogaster (13,986 protein coding genes) version 6.50 (Gramates et al. 2022). For our analysis, we extracted the longest CDS per gene (Sandmann et al. 2011; Whittle et al. 2019, 2021). To identify sex-biased genes, we used the modEncode database that contains expression data (Graveley et al. 2011; Gramates et al. 2022) for a wide range of D. melanogaster tissue types and developmental stages (N = 59 tissues/stages studied herein, Table S2;  Graveley et al. 2011; Li et al. 2014; Gramates et al. 2022), and that we also used to measure tau (Yanai et al. 2005) across all tissues. We identified highly expressed sex-biased genes as those with ≥100RPKM and ≥5-fold sex-bias in one sex versus the other in the contrast of 4-day mated testes versus 4-day mated ovaries, and unbiased genes as those with <5-fold bias (and ≥100RPKM in at least one gonad). For additional rigor, we also identified sex-biased genes using male accessory glands versus virgin ovaries (Table S2). Further details on identification of sex-biased genes and calculations of expression specificity using tau ((Yanai et al. 2005) (described previously in Whittle and Extavour 2024) using the data in Table S2, are provided in Supplementary Text File S1.

Codon Use and ΔRSCU

The organism-wide RSCU values (Table S1) for D. melanogaster were obtained from the Codon Statistics Database (CSD) (Subramanian et al. 2022). From these, we calculated the ΔRSCURibosome-All =RSCURibosomal genes -RSCUAll genes (using the concatenated genes per dataset), which is a more conservative version of the approach using the highest and least expressed genes (Duret and Mouchiroud 1999; Cutter et al. 2006; Whittle et al. 2007; Ingvarsson 2008; Wang et al. 2011). We identified the largest positive ΔRSCURibosome-All values as the optimal codon (and each was also identified as optimal at the CSD, with P-values < 0.05). All other codons per amino acid were classed as non-optimal, given their large disparity in use from the optimal codon (see Supplementary Text File S1). The tRNA gene counts in the D. melanogaster genome were determined using the GtRNA database (http://gtrnadb.ucsc.edu) (Chan and Lowe 2016), which utilizes tRNAscan-SE to extract tRNAs from the genome (Lowe and Chan 2016; Chan and Lowe 2019). CAI values that reflect similarity of codon use to the highly expressed ribosomal gene set, and the Fop values, were obtained from the CSD (Subramanian et al. 2022).

For the study of differences codon use among sex-biased genes, we determined herein the RSCU values and the codon raw counts per codon for every testis-biased, ovary-biased, and unbiased gene using Codon W (Peden 1999). For each of the studied 59 codons encoding a degenerate amino acid, the RSCU values were used to calculate ΔRSCUTestis-Ovary = average RSCUTestis-biased genes−average RSCUOvary-biased genes and the ΔRSCUTestis-Unbiased = average RSCUTestis-biased genes−average RSCUUnbiased genes. The RSCU values per codon were compared across all genes per contrast using t-tests to obtain P-values (Whittle et al. 2007). The codon counts per gene under study (for each testis-biased, ovary-biased and unbiased gene) were used to calculate the Percent-Non-opt codon use per gene (relative to all optimal and non-optimal codons per gene), based on the non-optimal and optimal codon identities that were defined in Table S1 (see also Supplementary Text File S1). The RSCU values used to calculate ΔRSCU for sex-biased and unbiased genes are shown in Tables S9 and S10.

Protein Folding, Relative Solvent Accessibility, and Codon Use in the Sex-biased Genes

To determine the RSA per amino acid, we used NetsurfP v3.0 set to default parameters (Hoie et al. 2022). This program is a machine-learning based tool proven effective for determining RSA and other structural data of proteins (Klausen et al. 2019; Hoie et al. 2022). NetsurfP v3.0 has been widely used to characterize protein folding and structures (Klausen et al. 2019; Hoie et al. 2022), and its RSA calculations have been shown to be equivalent to those predicted by other top predictor tools such as SPOT-1D-Single (Hoie et al. 2022). We also found strong correlations using protein structures from the Protein Data Base and RSA from GETAREA (Fraczkiewicz and Braun 1998) (data not shown). We determined the type of secondary structure that each amino acid was located within, namely a helix, string (β-string/sheet) or a random coil, and the disorder scores per amino acid, using NetsurfP 3.0 (Hoie et al. 2022).

Protein folding predictions for the visualization of structures in Fig. 2g–i were determined using Alphafold2 (Juniper et al. 2021), available for the D. melanogaster proteins (Uniprot proteome Identity UP000000803, taxon identity = 7227) from the AlphaFold2 database (https://alphafold.ebi.ac.uk) (Varadi et al. 2022). Visualization was conducted using Geneious Prime (Kearse et al. 2012).

For Fig. 3, using all concatenated testis-biased genes, the RSA was determined for each type of amino acid (18 degenerate amino acids) when it was encoded by the preferred non-optimal codon and when the same amino acid was encoded by the optimal codon, across all amino acid sites. Then, the average RSA was determined for all cases where the amino acid was encoded by the preferred non-optimal codon and when encoded by the optimal codon (averages and standard errors are provided in Table S8).

GO Functions

Gene ontology and functional analysis was performed using DAVID (Sherman et al. 2022).

Supplementary Material

evag093_Supplementary_Data
evag093_supplementary_data.pdf (1.1MB, pdf)

Acknowledgments

The authors thank members of the Extavour lab for valuable discussions. C. Extavour is an Investigator of the Howard Hughes Medical Institute, which supported this work. The valuable comments from two anonymous Reviewers are appreciated.

Contributor Information

Carrie A Whittle, Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, MA 02138, USA.

Cassandra G Extavour, Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, MA 02138, USA; Department of Molecular and Cellular Biology, Harvard University, Cambridge, MA 02138, USA.

Supplementary Material

Supplementary material is available at Genome Biology and Evolution online.

Author Contributions

C.A.W. and C.G.E. devised the study, conducted analyses, and wrote the manuscript.

Data Availability

All data used in this study are publicly available. The locations of the data are provided in the Materials and Methods.

Literature Cited

  1. Akashi  H. Gene expression and molecular evolution. Curr Opin Genet Dev.  2001:11:660–666. 10.1016/S0959-437X(00)00250-1. [DOI] [PubMed] [Google Scholar]
  2. Akashi  H. Translational selection and yeast proteome evolution. Genetics. 2003:164:1291–1303. 10.1093/genetics/164.4.1291. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Akdel  M, et al.  A structural biology community assessment of AlphaFold2 applications. Nat Struct Mol Biol.  2022:29:1056–1067. 10.1038/s41594-022-00849-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Allen  SR, et al.  Distinct responses to rare codons in select Drosophila tissues. eLife. 2022:11:e76893. 10.7554/elife.76893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Assis  R, Zhou  Q, Bachtrog  D. Sex-biased transcriptome evolution in Drosophila. Genome Biol Evol. 2012:4:1189–1200. 10.1093/gbe/evs093. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bahiri-Elitzur  S, Tuller  T. Codon-based indices for modeling gene expression and transcript evolution. Comput Struct Biotechnol J.  2021:19:2646–2663. 10.1016/j.csbj.2021.04.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Behura  SK, Severson  DW. Coadaptation of isoacceptor tRNA genes and codon usage bias for translation efficiency in Aedes aegypti and Anopheles gambiae. Insect Mol Biol. 2011:20:177–187. 10.1111/j.1365-2583.2010.01055.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Betancourt  AJ, Presgraves  DC. Linkage limits the power of natural selection in Drosophila. Proc Natl Acad Sci U S A.  2002:99:13616–13620. 10.1073/pnas.212277199. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Brackley  CA, Romano  MC, Thiel  M. The dynamics of supply and demand in mRNA translation. PLoS Comput Biol.  2011:7:e1002203. 10.1371/journal.pcbi.1002203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Bricout  R, et al.  Evolution is not uniform along protein sequences. Mol Biol Evol.  2023:40:msad042. 10.1093/molbev/msad042. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Brule  CE, Grayhack  EJ. Synonymous codons: choose wisely for expression. Trends Genet. 2017:33:283–297. 10.1016/j.tig.2017.02.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Chan  PP, Lowe  TM. GtRNAdb 2.0: an expanded database of transfer RNA genes identified in complete and draft genomes. Nucleic Acids Res. 2016:44:D184–D189. 10.1093/nar/gkv1309. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Chan  PP, Lowe  TM. tRNAscan-SE: searching for tRNA genes in genomic sequences. Methods Mol. Biol. 2019:1962:1–14. 10.1007/978-1-4939-9173-0_1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Civetta  A, Clark  AG. Correlated effects of sperm competition and postmating female mortality. Proc Natl Acad Sci U S A.  2000:97:13162–13165. 10.1073/pnas.230305397. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Comeron  JM, Kreitman  M, Aguadé  M. Natural selection on synonymous sites is correlated with gene length and recombination in Drosophila. Genetics. 1999:151:239–249. 10.1093/genetics/151.1.239. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Conant  GC, Stadler  PF. Solvent exposure imparts similar selective pressures across a range of yeast proteins. Mol Biol Evol.  2009:26:1155–1161. 10.1093/molbev/msp031. [DOI] [PubMed] [Google Scholar]
  17. Cutter  AD, Wasmuth  JD, Blaxter  ML. The evolution of biased codon and amino acid usage in nematode genomes. Mol Biol Evol.  2006:23:2303–2315. 10.1093/molbev/msl097. [DOI] [PubMed] [Google Scholar]
  18. Dean  R, Mank  JE. Tissue specificity and sex-specific regulatory variation permit the evolution of sex-biased gene expression. Am Nat.  2016:188:E74–E84. 10.1086/687526. [DOI] [PubMed] [Google Scholar]
  19. Dorus  S, et al.  Genomic and functional evolution of the Drosophila melanogaster sperm proteome. Nat Genet. 2006:38:1440–1445. 10.1038/ng1915. [DOI] [PubMed] [Google Scholar]
  20. dos Reis  M, Savva  R, Wernisch  L. Solving the riddle of codon usage preferences: a test for translational selection. Nucleic Acids Res. 2004:32:5036–5044. 10.1093/nar/gkh834. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Du  MZ, et al.  Co-adaption of tRNA gene copy number and amino acid usage influences translation rates in three life domains. DNA Res. 2017:24:623–633. 10.1093/dnares/dsx030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Dufton  MJ. Genetic code synonym quotas and amino acid complexity: cutting the cost of proteins?  J Theor Biol.  1997:187:165–173. 10.1006/jtbi.1997.0443. [DOI] [PubMed] [Google Scholar]
  23. Duret  L. tRNA gene number and codon usage in the,C. elegans genome are co-adapted for optimal translation of highly expressed genes. Trends Genet. 2000:16:287–289. 10.1016/S0168-9525(00)02041-2. [DOI] [PubMed] [Google Scholar]
  24. Duret  L, Mouchiroud  D. Expression pattern and, surprisingly, gene length shape codon usage in Caenorhabditis, Drosophila, and Arabidopsis. Proc Natl Acad Sci U S A. 1999:96:4482–4487. 10.1073/pnas.96.8.4482. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Echave  J, Spielman  SJ, Wilke  CO. Causes of evolutionary rate variation among protein sites. Nat Rev Genet.  2016:17:109–121. 10.1038/nrg.2015.18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Ellegren  H, Parsch  J. The evolution of sex-biased genes and sex-biased gene expression. Nat Rev Genet.  2007:8:689–698. 10.1038/nrg2167. [DOI] [PubMed] [Google Scholar]
  27. Fraczkiewicz  R, Braun  W. Exact and efficient analytical calculation of the accessible surface areas and their gradients for macromolecules. J Comput Chem.  1998:19:319–333. 10.1002/(SICI)1096-987X(199802)19:3<319::AID-JCC6>3.0.CO;2-W. [DOI] [Google Scholar]
  28. Franzosa  EA, Xia  Y. Structural determinants of protein evolution are context-sensitive at the residue level. Mol Biol Evol.  2009:26:2387–2395. 10.1093/molbev/msp146. [DOI] [PubMed] [Google Scholar]
  29. Fu  J, et al.  Codon usage affects the structure and function of the Drosophila circadian clock protein PERIOD. Genes Dev. 2016:30:1761–1775. 10.1101/gad.281030.116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Galtier  N, et al.  Codon usage bias in animals: disentangling the effects of natural selection, effective population size, and GC-biased gene conversion. Mol Biol Evol.  2018:35:1092–1103. 10.1093/molbev/msy015. [DOI] [PubMed] [Google Scholar]
  31. Gingold  H, Dahan  O, Pilpel  Y. Dynamic changes in translational efficiency are deduced from codon usage of the transcriptome. Nucleic Acids Res. 2012:40:10053–10063. 10.1093/nar/gks772. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Goodarzi  H, et al.  Modulated expression of specific tRNAs drives gene expression and cancer progression. Cell. 2016:165:1416–1427. 10.1016/j.cell.2016.05.046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Gramates  LS, et al.  FlyBase: a guided tour of highlighted features. Genetics. 2022:220:iyac035. 10.1093/genetics/iyac035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Graveley  BR, et al.  The developmental transcriptome of Drosophila melanogaster. Nature. 2011:471:473–479. 10.1038/nature09715. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Guimaraes  JC, et al.  A rare codon-based translational program of cell proliferation. Genome Biol. 2020:21:44. 10.1186/s13059-020-1943-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Hambuch  TM, Parsch  J. Patterns of synonymous codon usage in Drosophila melanogaster genes with sex-biased expression. Genetics. 2005:170:1691–1700. 10.1534/genetics.104.038109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Heger  A, Ponting  CP. Variable strength of translational selection among 12 Drosophila species. Genetics. 2007:177:1337–1348. 10.1534/genetics.107.070466. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Hershberg  R, Petrov  DA. General rules for optimal codon choice. PLoS Genet. 2009:5:e1000556. 10.1371/journal.pgen.1000556. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Hoie  MH, et al.  NetSurfP-3.0: accurate and fast prediction of protein structural features by protein language models and deep learning. Nucleic Acids Res. 2022:50:W510–W515. 10.1093/nar/gkac439. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ikemura  T. Correlation between the abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes: a proposal for a synonymous codon choice that is optimal for the E. coli translational system. J Mol Biol. 1981:151:389–409. 10.1016/0022-2836(81)90003-6. [DOI] [PubMed] [Google Scholar]
  41. Ikemura  T. Codon usage and tRNA content in unicellular and multicellular organisms. Mol Biol Evol.  1985:2:13–34. 10.1093/oxfordjournals.molbev.a040335. [DOI] [PubMed] [Google Scholar]
  42. Ilzhöfer  D, Heinzinger  M, Rost  B. SETH predicts nuances of residue disorder from protein embeddings. Front Bioinform.  2022:2:1019597. 10.3389/fbinf.2022.1019597. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Ingleby  FC, Flis  I, Morrow  EH. Sex-biased gene expression and sexual conflict throughout development. Cold Spring Harb Perspect Biol.  2014:7:a017632. 10.1101/cshperspect.a017632. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Ingvarsson  PK. Molecular evolution of synonymous codon usage in Populus. BMC Evol Biol. 2008:8:307. 10.1186/1471-2148-8-307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Juniper  T, et al.  Use of beta-D-glucan in diagnosis of suspected pneumocystis jirovecii pneumonia in adults with HIV infection. Int J STD AIDS.  2021:32:1074–1077. 10.1177/09564624211022247. [DOI] [PubMed] [Google Scholar]
  46. Kearse  M, et al.  Geneious basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics. 2012:28:1647–1649. 10.1093/bioinformatics/bts199. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Kimchi-Sarfaty  C, et al.  A “silent” polymorphism in the MDR1 gene changes substrate specificity. Science. 2007:315:525–528. 10.1126/science.1135308. [DOI] [PubMed] [Google Scholar]
  48. Kirchner  S, et al.  Alteration of protein function by a silent polymorphism linked to tRNA abundance. PLoS Biol. 2017:15:e2000779. 10.1371/journal.pbio.2000779. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Klausen  MS, et al.  NetSurfP-2.0: improved prediction of protein structural features by integrated deep learning. Proteins. 2019:87:520–527. 10.1002/prot.25674. [DOI] [PubMed] [Google Scholar]
  50. Letzring  DP, Dean  KM, Grayhack  EJ. Control of translation efficiency in yeast by codon-anticodon interactions. RNA. 2010:16:2516–2528. 10.1261/rna.2411710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Li  JJ, Huang  H, Bickel  PJ, Brenner  SE. Comparison of D. melanogaster and C. elegans developmental stages, tissues, and cells by modENCODE RNA-seq data. Genome Res. 2014:24:1086–1101. 10.1101/gr.170100.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Liu  Y, Yang  Q, Zhao  F. Synonymous but not silent: the Codon usage code for gene expression and protein folding. Annu Rev Biochem.  2021:90:375–401. 10.1146/annurev-biochem-071320-112701. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Loewe  L, Charlesworth  B. Background selection in single genes may explain patterns of codon bias. Genetics. 2007:175:1381–1393. 10.1534/genetics.106.065557. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Lowe  TM, Chan  PP. tRNAscan-SE on-line: integrating search and context for analysis of transfer RNA genes. Nucleic Acids Res. 2016:44:W54–W57. 10.1093/nar/gkw413. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Mank  JE, Ellegren  H. Are sex-biased genes more dispensable?  Biol Lett.  2009:5:409–412. 10.1098/rsbl.2008.0732. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Mank  JE, Hultin-Rosenberg  L, Zwahlen  M, Ellegren  H. Pleiotropic constraint hampers the resolution of sexual antagonism in vertebrate gene expression. Am Nat.  2008:171:35–43. 10.1086/523954. [DOI] [PubMed] [Google Scholar]
  57. Meisel  RP. Towards a more nuanced understanding of the relationship between sex-biased gene expression and rates of protein-coding sequence evolution. Mol Biol Evol.  2011:28:1893–1900. 10.1093/molbev/msr010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Miller  JB, Brase  LR, Ridge  PG. ExtRamp: a novel algorithm for extracting the ramp sequence based on the tRNA adaptation index or relative codon adaptiveness. Nucleic Acids Res. 2019:47:1123–1131. 10.1093/nar/gky1193. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Moriyama  EN, Powell  JR. Codon usage bias and tRNA abundance in Drosophila. J Mol Evol. 1997:45:514–523. 10.1007/PL00006256. [DOI] [PubMed] [Google Scholar]
  60. Moutinho  AF, Trancoso  FF, Dutheil  JY. The impact of protein architecture on adaptive evolution. Mol Biol Evol.  2019:36:2013–2028. 10.1093/molbev/msz134. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Otto  SP. Two steps forward, one step back: the pleiotropic effects of favoured alleles. Proc Biol Sci.  2004:271:705–714. 10.1098/rspb.2003.2635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Pechmann  S, Frydman  J. Evolutionary conservation of codon optimality reveals hidden signatures of cotranslational folding. Nat Struct Mol Biol.  2013:20:237–243. 10.1038/nsmb.2466. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Peden  JF. Analysis of codon usage. University of Nottingham; 1999. [Google Scholar]
  64. Percudani  R, Pavesi  A, Ottonello  S. Transfer RNA gene redundancy and translational selection in Saccharomyces cerevisiae. J Mol Biol. 1997:268:322–330. 10.1006/jmbi.1997.0942. [DOI] [PubMed] [Google Scholar]
  65. Piovesan  D, Monzon  AM, Tosatto  SCE. Intrinsic protein disorder and conditional folding in AlphaFoldDB. Protein Sci. 2022:31:e4466. 10.1002/pro.4466. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Pitnick  S. Investment in testes and the cost of making long sperm in Drosophila. Am Nat.  1996:148:57–80. 10.1086/285911. [DOI] [Google Scholar]
  67. Plotkin  JB, Kudla  G. Synonymous but not the same: the causes and consequences of codon bias. Nat Rev Genet.  2011:12:32–42. 10.1038/nrg2899. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Quax  T, Claassens  NJ, Söll  D, van der Ooost  J. Codon bias as a means to fine-tune gene expression. Mol Cell.  2015:59:149–161. 10.1016/j.molcel.2015.05.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Ramsey  DC, Scherrer  MP, Zhou  T, Wilke  CO. The relationship between relative solvent accessibility and evolutionary rate in protein evolution. Genetics. 2011:188:479–488. 10.1534/genetics.111.128025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Sandmann  T, Vogg  MC, Owlarn  S, Boutros  M, Bartscherer  K. The head-regeneration transcriptome of the planarian Schmidtea mediterranea. Genome Biol. 2011:12:R76. 10.1186/gb-2011-12-8-r76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  71. Scherrer  MP, Meyer  AG, Wilke  CO. Modeling coding-sequence evolution within the context of residue solvent accessibility. BMC Evol Biol.  2012:12:179. 10.1186/1471-2148-12-179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Sharp  PM, Li  WH. The codon Adaptation Index–a measure of directional synonymous codon usage bias, and its potential applications. Nucleic Acids Res. 1987:15:1281–1295. 10.1093/nar/15.3.1281. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Sharp  PM, Tuohy  TM, Mosurski  KR. Codon usage in yeast: cluster analysis clearly differentiates highly and lowly expressed genes. Nucleic Acids Res. 1986:14:5125–5143. 10.1093/nar/14.13.5125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  74. Sherman  BT, et al.  DAVID: a web server for functional enrichment analysis and functional annotation of gene lists (2021 update). Nucleic Acids Res. 2022:50:W216–W221. 10.1093/nar/gkac194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Shields  DC, Sharp  PM, Higgins  DG, Wright  F. Silent” sites in Drosophila genes are not neutral: evidence of selection among synonymous codons. Mol Biol Evol. 1988:5:704–716. 10.1093/oxfordjournals.molbev.a040525. [DOI] [PubMed] [Google Scholar]
  76. Stadler  M, Fire  A. Wobble base-pairing slows in vivo translation elongation in metazoans. RNA. 2011:17:2063–2073. 10.1261/rna.02890211. [DOI] [PMC free article] [PubMed] [Google Scholar]
  77. Stein  KC, Frydman  J. The stop-and-go traffic regulating protein biogenesis: how translation kinetics controls proteostasis. J Biol Chem.  2019:294:2076–2084. 10.1074/jbc.rev118.002814. [DOI] [PMC free article] [PubMed] [Google Scholar]
  78. Subramanian  K, Payne  B, Feyertag  F, Alvarez-Ponce  D. The Codon statistics database: a database of codon usage bias. Mol Biol Evol.  2022:39:msac157. 10.1093/molbev/msac157. [DOI] [PMC free article] [PubMed] [Google Scholar]
  79. Tamura  K, Subramanian  S, Kumar  S. Temporal patterns of fruit fly (Drosophila) evolution revealed by mutation clocks. Mol Biol Evol. 2004:21:36–44. 10.1093/molbev/msg236. [DOI] [PubMed] [Google Scholar]
  80. Tien  MZ, Meyer  AG, Sydykova  DK, Spielman  SJ, Wilke  CO. Maximum allowed solvent accessibilities of residues in proteins. PLoS One. 2013:8:e80635. 10.1371/journal.pone.0080635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  81. Torgerson  DG, Kulathinal  RJ, Singh  RS. Mammalian sperm proteins are rapidly evolving: evidence of positive selection in functionally diverse genes. Mol Biol Evol. 2002:19:1973–1980. 10.1093/oxfordjournals.molbev.a004021. [DOI] [PubMed] [Google Scholar]
  82. Torrent  M, Chalancon  G, de Groot  NS, Wuster  A, Babu  MM. Cells alter their tRNA abundance to selectively regulate protein synthesis during stress conditions. Sci Signal.  2018:11:eaat6409. 10.1126/scisignal.aat6409. [DOI] [PMC free article] [PubMed] [Google Scholar]
  83. Tuller  T, et al.  An evolutionarily conserved mechanism for controlling the efficiency of protein translation. Cell. 2010:141:344–354. 10.1016/j.cell.2010.03.031. [DOI] [PubMed] [Google Scholar]
  84. Varadi  M, et al.  AlphaFold protein structure database: massively expanding the structural coverage of protein-sequence space with high-accuracy models. Nucleic Acids Res. 2022:50:D439–D444. 10.1093/nar/gkab1061. [DOI] [PMC free article] [PubMed] [Google Scholar]
  85. Wang  B, et al.  Optimal codon identities in bacteria: implications from the conflicting results of two different methods. PLoS One. 2011:6:e22714. 10.1371/journal.pone.0022714. [DOI] [PMC free article] [PubMed] [Google Scholar]
  86. Whittle  CA, Extavour  CG. Expression-linked patterns of codon usage, amino acid frequency, and protein length in the basally branching arthropod Parasteatoda tepidariorum. Genome Biol Evol. 2016:8:2722–2736. 10.1093/gbe/evw068. [DOI] [PMC free article] [PubMed] [Google Scholar]
  87. Whittle  CA, Extavour  CG. Rapid evolution of ovarian-biased genes in the yellow fever mosquito (Aedes aegypti). Genetics. 2017:206:2119–2137. 10.1534/genetics.117.201343. [DOI] [PMC free article] [PubMed] [Google Scholar]
  88. Whittle  CA, Extavour  CG. Gene protein sequence evolution can predict the rapid divergence of ovariole numbers in the Drosophila melanogaster subgroup. Genome Biol Evol. 2024:16. 10.1093/gbe/evae118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  89. Whittle  CA, Kulkarni  A, Chung  N, Extavour  CG. Adaptation of codon and amino acid use for translational functions in highly expressed cricket genes. BMC Genomics. 2021:22:234. 10.1186/s12864-021-07411-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  90. Whittle  CA, Kulkarni  A, Extavour  CG. Evidence of multifaceted functions of codon usage in translation within the beetle model Tribolium castaneum. DNA Res. 2019:26:473–484. 10.1093/dnares/dsz025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  91. Whittle  CA, Malik  MR, Krochko  JE. Gender-specific selection on codon usage in plant genomes. BMC Genomics. 2007:8:169–179. 10.1186/1471-2164-8-169. [DOI] [PMC free article] [PubMed] [Google Scholar]
  92. Williford  A, Demuth  JP. Gene expression levels are correlated with synonymous codon usage, amino acid composition, and gene architecture in the red flour beetle Tribolium castaneum. Mol Biol Evol. 2012:29:3755–3766. 10.1093/molbev/mss184. [DOI] [PMC free article] [PubMed] [Google Scholar]
  93. Wong  HWS, Holman  L. Pleiotropic fitness effects across sexes and ages in the Drosophila genome and transcriptome. Evolution. 2023:77:2642–2655. 10.1093/evolut/qpad163. [DOI] [PubMed] [Google Scholar]
  94. Yanai  I, et al.  Genome-wide midrange transcription profiles reveal expression level relationships in human tissue specification. Bioinformatics. 2005:21:650–659. 10.1093/bioinformatics/bti042. [DOI] [PubMed] [Google Scholar]
  95. Yu  CH, et al.  Codon usage influences the local rate of translation elongation to regulate co-translational protein folding. Mol Cell.  2015:59:744–754. 10.1016/j.molcel.2015.07.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  96. Zalucki  YM, Jennings  MP. Experimental confirmation of a key role for non-optimal codons in protein export. Biochem Biophys Res Commun.  2007:355:143–148. 10.1016/j.bbrc.2007.01.126. [DOI] [PubMed] [Google Scholar]
  97. Zhang  Z, Hambuch  TM, Parsch  J. Molecular evolution of sex-biased genes in Drosophila. Mol Biol Evol. 2004:21:2130–2139. 10.1093/molbev/msh223. [DOI] [PubMed] [Google Scholar]
  98. Zhou  M, et al.  Non-optimal codon usage affects expression, structure and function of clock protein FRQ. Nature. 2013:495:111–115. 10.1038/nature11833. [DOI] [PMC free article] [PubMed] [Google Scholar]
  99. Zhou  M, Wang  T, Fu  J, Xiao  G, Liu  Y. Nonoptimal codon usage influences protein structure in intrinsically disordered regions. Mol Microbiol.  2015:97:974–987. 10.1111/mmi.13079. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

evag093_Supplementary_Data
evag093_supplementary_data.pdf (1.1MB, pdf)

Data Availability Statement

All data used in this study are publicly available. The locations of the data are provided in the Materials and Methods.

Articles from Genome Biology and Evolution are provided here courtesy of Oxford University Press

RESOURCES