Gene duplication is a fundamental genomic process that occurs on a small scale via segmental duplication and on a large scale via whole genome duplication (WGD). Gene duplication triggers biological innovation (neofunctionalization; 1), reduces pleiotropy by division of labor (subfunctionalization; 2), contributes to reproductive incompatibilities via divergent resolution [3], and generates redundancy [4]. More often than not, however, the ultimate fate of duplicate genes is the loss of one copy, usually within a few million years [5]. Persistence of duplicate genes occurs when natural selection favors duplicated protein function or expression, or when modifications soon after duplication distinguish the duplicates in some way (for example by causing paralogous expression domains to diverge). The targets of natural selection after gene duplication can be broadly divided into those that involve no change in function, gain of function, or loss of function [6]. For example, both copies of a duplicated gene may evade pseudogenization without changed function if having higher expression is advantageous, if disrupting the dosages of interacting duplicated proteins is disadvantageous, or if segregation of functionally distinct alleles that evolved prior to duplication is beneficial [5, 7–9]. Persistence of duplicates could also be triggered by changes after duplication if advantageous mutations occur in one or both paralogs, or if degenerative mutations make the paralogs non-redundant [1, 2, 10, 11]. Over time, increased divergence between duplicate genes reduces the chances of synfunctionalization, a phenomenon by which one of the paralogs convergently evolves the function or expression domain of the other, thereby re-establishing redundancy and leading to the loss of one paralog [12]. A challenge in evolutionary biology, then, is to identify the relative importance of different mechanisms for duplicate gene retention, and also to understand what genetic factors prior to duplication portend a high probability of both paralogs persisting as functional loci after duplication.

Various factors are correlated with the odds of duplicate gene persistence, including how paralogs are generated (by segmental duplication or WGD), the breadth and intensity of expression, dosage sensitivity, interactions with other proteins, the number of functional domains, and the rate of evolution [13–23]. In the case of ohnologs - duplicate genes generated by WGD, expression dosage is thought to frequently drive functional persistence, suggesting that the stoichiometry of expression among duplicated loci is important [21, 24–29]. However, many of these factors are correlated with each other, making it challenging to tease apart their relative impact. For example, genes with high expression intensities or broad expression patterns tend to evolve slowly [30–37] so it is not clear with which degree each variable influences duplicate gene persistence after factoring out the role of the other. Our focus here is to disentangle and quantify the relative contributions of several variables to duplicate gene persistence after WGD.

This study examines gene expression, molecular evolution, and gene structure information of singleton genes in the diploid species S. tropicalis that are orthologous either to a pair of ohnologs in the tetraploid species X. laevis or to a single-copy gene in X. laevis. We assume that all single-copy genes in X. laevis were once part of a pair of ohnologs but that one copy has been lost due to post-WGD pseudogenization. Our analyses therefore consist of two gene sets: (1) gene triads, which include a pair of X. laevis ohnologs and the corresponding S. tropicalis singleton ortholog, and (2) gene dyads, which include one X. laevis singleton and the corresponding S. tropicalis singleton ortholog.

Nucleotide sequences from 3,387 gene triads and 4,746 dyads were gathered from the NCBI UniGene databases using tBLASTx and a reciprocal best hit approach. UniGenes are a set of non-redundant clusters of transcript sequences that compose the expressed sequence tag (EST) libraries. In a gene triad, the two X. laevis UniGenes were reciprocal best hits within the X. laevis UniGenes, and they both returned the same S. tropicalis UniGene as the top hit. In a dyad, each putative X. laevis singleton has a single unique reciprocal top hit with S. tropicalis. UniGenes that had more than two X. laevis genes or more than one S. tropicalis gene with reciprocal best hits were excluded. To test whether our putative singletons were indeed singletons, we tried to amplify the other ohnolog of 17 X. laevis singletons using PCR primers designed to amplify both the X. laevis singleton and the S. tropicalis ortholog [41]; in all cases only one gene copy was amplified in X. laevis. Triads and dyads were aligned using MUSCLE [42], and Perl scripts were used to predict the beginning and end of coding regions by looking for the longest open reading frame in either direction. Overlapping alignments shorter than 201 nucleotides were discarded along with 5' and 3' untranslated regions and indels.

Non-normalized EST libraries were obtained from NCBI (dbEST Library IDs: 10829, 10830, 10895, 10896, 20954, 20886, 21298, 20560, 20561, 20562, 20892, 20891, 20911, 20931, 20912, 20947, 20953, 20901, 16856, 16857, 16858, 16871, 16854, 16853, 16863, 16862, 16870, 16864, 16872, 17807, 16868, 16869, 16867, 16865, 16866, 17804, 17805, 17806, 16855, 16873, 16876, 16875, 16877, 16878, 16874, 16801, 16859, 16860, 16861, 16880, 8773, 8701, 20682, 9909, 9665, 9908, 14603). We classified and combined 718,484 S. tropicalis ESTs (from a total of 1,271,375) into 14 distinct adult tissues and 4 embryological stages: brain, bone, eye, heart, kidney, liver, lung, thymus, pancreas, skin, spleen, fat body, ovary, testis, egg, gastrula stage, neurula stage, and embryo stage 62. Each library consists of at least 3,900 transcripts (on average 39,915) and each individual UniGene has its own set of unique ESTs. We used the proportion of transcripts of each gene divided by the number of transcripts in a given EST library as an estimate of its expression level to control for the different sizes of the EST libraries. For twenty-three outlier genes with high expression, we truncated the expression level to a value of 1% of the respective EST library to prevent these outliers from dominating the results. Of the set of genes for which sequence data were available, a total of 3,298 triads and 4,426 dyads also had expression data (at least one S. tropicalis EST read in at least one EST database).

We used logistic regression to test the relative impact of expression and evolution of genes from a diploid species, S. tropicalis, on the odds of functional persistence of duplicate genes generated by WGD in a tetraploid species, X. laevis. Our analysis employed EST databases and new molecular sequences from closely related outgroup species generated with next-generation sequencing of cDNA. We found that genes that are highly and broadly expressed, slowly evolving, and with more streamlined information packaging are preferentially retained as duplicates after WGD. Total level and evenness of expression have the largest impact of these parameters.

Higher and broader gene expression, both positively correlated with ohnolog retention, implies greater dosage sensitivity and pleiotropy, respectively [27, 36, 47–49]. The preferential retention of duplicate genes with these characteristics - stronger dosage constraints and more diverse interactions - supports the idea that expression dosage balance is important after WGD, and is consistent with other recent findings [13, 24–27, 45, 50]. In addition, broader gene expression reduces the likelihood of duplicate gene loss through synfunctionalization [12]. The prediction that breadth is related to the level of pleiotropy could be further tested with protein interaction information, which is currently lacking for Xenopus.

Having higher and broader gene expression also allows more opportunities for partitioning paralogous expression patterns by subfunctionalization [2, 51]. Here we did not test for evidence of expression subfunctionalization or neofunctionalization because comprehensive expression profiles in a closely related outgroup species are not available. But when one considers protein structure, higher information content in terms of longer genes and more exons could facilitate subfunctionalization, for instance because of a larger mutational target. In yeast, retained duplicates have longer proteins and more functional domains than singletons [52]. In contrast, we find that fewer exons and shorter intron length are more likely to be retained following WGD, after taking into consideration the effect of expression and molecular evolutionary rates. These findings suggest that subfunctionalized regions are generally not parceled by exons in X. laevis, or that protein subfunctionalization is subtle in this species.

It has been suggested that slowly evolving proteins may be preferentially subfunctionalized after WGD because they accumulate fewer substitutions over time than faster evolving genes, thus retaining their interchangeability for longer periods, and providing more opportunities for subfunctionalization to occur [23]. This is despite the observation that paralogs are often found to have higher evolutionary rates than singletons [53, 54]. Our findings do not directly address this proposal, but we find that slower evolving genes (genes with lower dN and dS in S. tropicalis) were preferentially retained after WGD, which supports the expectations of Sémon and Wolfe [23].

Thus, in addition to supporting mechanisms based on dosage balance and pleiotropy, our results are also consistent with mechanisms for duplicate gene retention that involve regulatory subfunctionalization. The observation that gene expression has a major impact on the odds of functional persistence of ohnologs may be in part due to the greater mutational opportunities at the regulatory level compared with the protein-coding region. Once initial retention is achieved, for example through dosage balance or subfunctionalization, this may subsequently decrease constraints in different parts of the gene and thus allow further modifications to occur for long-term duplicate gene persistence [55, 56].

Caveats exist in our interpretation of these data and results. We used genetic data from the most closely related diploid species available for comparison, which are not necessarily the same as the state of these genes in the diploid ancestor of the tetraploid species. However, we are only assuming that diploid characters are correlated with the ancestral state, and we argue that our interpretations are reasonable under the assumption that the relative rates of molecular evolution and expression between genes have generally remained similar since speciation (e.g. that broadly expressed genes in the extant diploid were also broadly expressed in the ancestral diploid). It is also possible that some duplicates were misidentified as singletons due to lacking data from one ohnolog or extensive divergence between ohnologs. We suspect that the proportion of misclassified singletons is small based on congruence between our database of triads and previous studies [23, 43, 57, 58]. Because WGD in clawed frogs is probably the result of allopolyploidization (reviewed in [39]), these ohnologs presumably diverged to some degree in the diploid ancestral species before their contact in the tetraploid species. We suspect this does not have a large impact on our designation of dyads and triads, however our findings might thus be less representative of duplicate gene persistence in autopolyploids. It is also possible that segmental duplication occurred in some S. tropicalis loci, adding noise to our analysis. Again, we speculate that this would have a negligible impact on our overall conclusions and note that, using the UniGene sequence database, the X. laevis genes were always better BLAST hits for S. tropicalis sequences than any other S. tropicalis genes. We also expect the dataset of S. tropicalis orthologs to be quite comprehensive because a complete genome sequence is available for this species [59]. Additionally, our molecular data from dyads and triads could be biased towards slowly evolving genes with discernable homology to outgroup sequences from P. carvalhoi and H. curtipes. Whether or not this is the case is difficult to know without more complete data from the outgroup species. Another concern is that cloning biases and small EST sample size could impact expression estimates for some loci. While this is possible, it seems doubtful that this would substantially affect the magnitude of the correlation coefficients.

We used logistic regression to evaluate the contributions of correlated genetic parameters in a diploid species on the odds of whether an ohnolog is retained in a closely related tetraploid species. Our results suggest that several parameters have a significant influence on the odds of duplicate persistence after taking other parameters into account, and that total expression and evenness of expression are most important. These findings are broadly consistent with retention mechanisms involving dosage balance and subfunctionalization. To further dissect apart the role of these retention mechanisms, one could examine patterns of expression and evolution of duplicate genes relative to an ortholog in an outgroup species. For example, studies of regulatory subfunctionalization in the tetraploid could determine whether the expression domain inferred for an ancestral gene are divided amongst the duplicate genes in the tetraploid [23, 60, 61]. A challenge in this type of study is the assessment of the ancestral expression state, because all orthologs have evolved since diverging from the most recent common ancestor. A possible alternative is to investigate protein interactions for reduced and partitioned pleiotropy among duplicates, but confirmation of subfunctionalization would also require inference of ancestral interactions.