Introduction

Duplicated genes are important contributors to molecular evolution (Ohno 1970; Conant and Wolfe 2008; Dittmar 2010; Innan and Kondrashov 2010). A gene duplication event creates a second (derived) copy of a gene via one of many molecular mechanisms, including non-allelic recombination and reverse transcription of mRNA (Zhang 2003; Kaessmann et al. 2009; Marques-Bonet et al. 2009). The derived copy can acquire novel functions and/or the ancestral and derived loci can each evolve a subset of functions present prior to duplication (Spofford 1969; Hughes 1994; Force et al. 1999; Lynch and Force 2000). When functions are partitioned between the paralogous copies, gene duplication can resolve pleiotropic conflicts present in the single-copy ancestor (Hittinger and Carroll 2007; Des Marais and Rausher 2008; Connallon and Clark 2011; Gallach and Betrán 2011; Abascal et al. 2013; VanKuren and Long 2018).

Gene duplication can give rise to a derived copy located in tandem to the ancestral copy or dispersed to another genomic location. The ancestral copy of a dispersed duplicate can further be lost from the genome, creating the appearance that the gene was “relocated” to the derived locus (Fig. 1). Comparative genomic analyses in animals and plants have revealed that gene relocation occurs frequently, and relocated genes may be as common as canonical dispersed duplications in which the ancestral copy is retained (Bhutkar et al. 2007; Meisel et al. 2009; Wicker et al. 2010; Han and Hahn 2012; Ciomborowska et al. 2013). Furthermore, gene relocation can promote reproductive isolation between species because some F2 hybrids lack the relocated gene (Masly et al. 2006; Bikard et al. 2009; Moyle et al. 2010).

Fig. 1
figure 1

Gene duplication and relocation. In the ancestral arrangement, a gene (white circle) is located on chromosome 1. After gene duplication, the derived copy (star) is located on chromosome 2. In the case of gene relocation, the copy at the ancestral locus is subsequently lost

Full size image

Despite the prevalence and evolutionary importance of gene relocation, the selection pressures acting on relocated genes have received considerably less attention than the evolutionary dynamics of canonical duplicated genes. The analyses that have been performed identified some important differences between relocated genes and canonical dispersed duplicates. For example, derived copies of duplicated genes in animal genomes tend to be narrowly expressed in reproductive tissues (Vinckenbosch et al. 2006; Meisel et al. 2009, 2010; Baker et al. 2012; Kondo et al. 2017). In contrast, Drosophila and human relocated genes tend to be broadly expressed across many tissues (Meisel et al. 2009; Ciomborowska et al. 2013). In addition, the derived copies of dispersed duplicates tend to experience positive selection or relaxed constraints (Kondrashov et al. 2002; Conant and Wagner 2003; Han et al. 2009; Han and Hahn 2012), while mammalian relocated genes appear to evolve under strong purifying selection (Ciomborowska et al. 2013). It has been hypothesized that positive selection on the derived copies of duplicated genes fixes mutations that improve testis-specific functions once pleiotropic constraints are relaxed by duplication (Betrán and Long 2003; Torgerson and Singh 2004; Betrán et al. 2006; Rosso et al. 2008; Meisel et al. 2010; Quezada-Diaz et al. 2010; Tracy et al. 2010; VanKuren and Long 2018). Gene relocation is unlikely to resolve pleiotropic conflicts because a second copy of the gene is not retained. To improve our understanding of the evolutionary dynamics of relocated genes, we combined population genetic, functional genomic, and experimental approaches to characterize the selection pressures acting on Drosophila relocated genes.

Materials and Methods

Identifying Duplicated and Relocated Genes

Drosophila genomes have six chromosome arms, known as Muller elements A–F (Muller 1940; Schaeffer et al. 2008). We analyzed previously annotated inter-chromosome-arm duplicated and relocated genes that occurred along the lineages leading to Drosophila melanogaster and Drosophila pseudoobscura (Hahn et al. 2007; Meisel et al. 2009), ignoring duplication and relocation events involving the minute element F. The lineage-specific duplicates were identified by examining phylogenetic reconstructions of gene families from the D. melanogaster, D. pseudoobscura, Drosophila willistoni, Drosophila virilis, and Drosophila grimshawi genomes. We selected gene families in which the phylogenetic reconstruction included a duplication event along the lineage leading to D. melanogaster or D. pseudoobscura after the divergence with all other lineages. From this group, we then curated a list of duplications in which one copy was on a different chromosome arm than the homologous genes across all species. The ancestral copy of a duplicated gene in one species’ genome was inferred to be the copy found on the same chromosome arm as in the other four species, and the derived copy is the one on a different chromosome arm. Relocated genes were identified as present in a single copy in D. melanogaster or D. pseudoobscura, with single-copy orthologs on a different chromosome arm in the other four species. As a control, we also analyzed single-copy non-relocated genes that are retained as 1:1:1:1:1 orthologs on the same chromosome arm across all five species (Meisel et al. 2009). We excluded genes on element F from our control set.

Sequence Divergence, Polymorphism, and Selection

We obtained estimates of polymorphism and divergence for relocated genes, non-relocated single-copy genes, and the ancestral and derived copies of inter-chromosome-arm duplicated genes in the D. melanogaster genome from published datasets. Two data sets were used to calculate divergence between D. melanogaster and Drosophila simulans orthologs. All of the duplications and relocations in our data set happened before the divergence of the D. melanogaster and D. simulans lineages, so that our estimates of divergence are specific to either the ancestral or derived copy. First, we obtained estimates of nucleotide sequence divergence along the D. melanogaster lineage after the split with D. simulans for all 1:1 orthologous genes between these two closely related species (Hu et al. 2013). In the results presented here, we analyzed substitutions per site for zero- and four-fold degenerate sites within protein coding regions. Second, we obtained estimates of the ratio of non-synonymous to synonymous substitutions per site (\(d_\text{N}/d_\text{S}\)) from a published analysis comparing D. melanogaster and D. simulans genes (Stanley and Kulathinal 2016).

We obtained the amount non-synonymous \((P_\text{N})\) and synonymous \((P_\text{S})\) polymorphic sites within D. melanogaster genes from the Drosophila Genetic reference panel (DGRP; Mackay et al. 2012; Ràmia et al. 2012). We only included polymorphic sites with a minor allele frequency > 5% to minimize the inclusion of segregating deleterious alleles (Fay et al. 2001). We also obtained the number of non-synonymous \((D_\text{N})\) and synonymous \((D_\text{S})\) substitutions between D. melanogaster and D. simulans from the DGRP data. We analyzed the polymorphism and divergence data for single-copy non-relocated genes, the ancestral and derived copies of inter-chromosome-arm duplicates, and relocated genes within the framework of McDonald and Kreitman (1991). First, we used a \(\chi ^2\) test of independence to identify genes with an excess or deficiency of non-synonymous substitutions. We assigned genes as evolving under positive selection if they have a significant excess of non-synonymous substitutions, and we assigned genes as evolving under strong negative selection if they have a significant deficiency of non-synonymous substitutions. Second, we calculated \({\overline{\alpha }}\), the fraction of non-synonymous substitutions fixed by selection (Smith and Eyre-Walker 2002), for each group of genes:

$$\overline{\alpha } = 1 - \frac{{\overline{{D_{{\text{S}}} }} }}{{\overline{{D_{{\text{N}}}}}}}\,\overline{{\left( {\frac{{P_{{\text{N}}} }}{{P_{{\text{S}}} }}} \right)}}.$$
(1)

We calculated \({\overline{\alpha }}\) separately for single-copy non-relocated genes, ancestral copies of inter-chromosome-arm duplicates, derived copies, and relocated genes. We performed 1000 bootstrapped replicate analyses to calculate a confidence interval (CI) for each \({\overline{\alpha }}\) estimate.

Gene Expression Profiles

We analyzed available microarray data to assess the expression across adult tissues of D. melanogaster relocated genes, duplicated genes, and single-copy genes. Expression measurements were taken from FlyAtlas, which includes 11 non-redundant adult non-sex-specific tissue samples (brain, crop, midgut, hindgut, Malpighian tubule, thoracicoabdominal ganglion, salivary gland, fat body, eye, heart, and trachea), two male-specific organs (testis and accessory gland), and two female-specific organs (ovary and spermatheca) (Chintapalli et al. 2007). Expression levels for spermatheca were averaged between mated and unmated females (Meisel et al. 2009). We used \(\tau\) as a measure of expression breadth for each gene:

$$\tau = \frac{{\sum\nolimits_{{i = 1}}^{N} 1 - \frac{{\log _{{10}} S_{{\text{i}}} }}{{\log _{{10}} S_{{\max }} }}}}{{N - 1}},$$
(2)

where N is the number of tissues (15); \(S_\text{i}\) is the expression level in tissue i, and \(S_{{\max }}\) is the maximum expression of that gene across all tissues (Yanai et al. 2005; Larracuente et al. 2008). All \(S_\text{i}\,{<}\,1\) were set to 1 for this analysis. Values of \(\tau\) range from 0 to 1, with higher values corresponding to more tissue-specific expression.

We also analyzed microarray data from D. melanogaster testis (Chintapalli et al. 2007) and RNA-seq data from D. pseudoobscura testis (Meisel et al. 2010) to infer the expression levels of relocated, duplicated, and single-copy genes. Finally, we analyzed sex-specific microarray data from D. melanogaster and D. pseudoobscura heads and whole flies to calculate “sex-biased” expression (Meisel et al. 2012), i.e., the relative expression of genes in males and females (\(\log _2{M}/{F}\)).

Viability and Fertility Effects of Knockdown

To assess if relocated and non-relocated single-copy D. melanogaster genes are essential for viability and male fertility, we used Gal4-UAS inducible RNA interference (RNAi) to knock down the expression of relocated and single-copy non-relocated genes. Flies carrying an inducible construct containing a hairpin sequence that silences the expression of a target gene via RNAi (UAS-RNAi) were obtained from the Vienna Drosophila Resource Center (VDRC; Dietzl et al. 2007). Knockdown was performed using two different sets of RNAi lines. The first set, known as “GD” lines, were produced by random integration into the D. melanogaster genome of a P-element construct carrying a pUAST vector with 10 copies of the UAS and a 300–400 bp inverted repeat targeting the gene of interest. The second set, known as “KK” lines, also carry 10 copies of UAS and a long inverted repeat, but they were inserted into specific sites in the genome using \(\phi\) C31 targeted integration (Groth et al. 2004; Bateman et al. 2006). Expression of the RNAi construct in some of the KK lines can be lethal because of mis-expression of the developmental gene tiptop (Green et al. 2014; Vissers et al. 2016), which can lead to false inference about the essentiality of duplicated genes (Kondo et al. 2017). We therefore performed analyses of our results from the GD and KK lines separately to assess the extent to which our results could be attributed to systemic effects of KK lines.

To assay the effect of knockdown on viability, individual males carrying a UAS-RNAi transgene were crossed to individual females carrying a Gal4 driver construct that is ubiquitously expressed under the tubulin 1\(\alpha\) promoter (P{tubP-Gal4}). P{tubP-Gal4} is expressed in many tissues and throughout development (Lee and Luo 1999), which causes constitutive knockdown of the target gene when combined in the same genotype with a UAS-RNAi construct. In addition, P{tubP-Gal4} is balanced over the TM3 chromosome, which carries the dominant Stubble (Sb) allele, allowing us to differentiate between knockdown and non-knockdown (control) siblings within each cross. The females were allowed to lay eggs for 3 days following mating on cornmeal media, and then all progeny that emerged were scored for their sex and bristle phenotype (stubble or wild-type). We assessed the viability of the knockdown flies by comparing the counts of knockdown progeny with their control siblings. We also performed control crosses in which the UAS-RNAi male is replaced with a male from the progenitor stock from which the RNAi lines were derived—GD lines were created by transforming w1118 flies (VDRC line 60000), and KK lines were created by transforming y, w1118; P{attP, y+, w3’} flies (VDRC line 60100). These control males do not carry a UAS-RNAi construct.

We used linear models to assess the effect of RNAi knockdown on viability. For each gene, we modeled the number of progeny recovered \((N_\text{ijk})\) with the phenotype associated with either knockdown (wild-type bristles) or non-knockdown (stubble bristles) from crosses involving a fly either carrying the UAS-RNAi construct or from the progenitor non-RNAi strain:

$$\begin{aligned} N_\text{ijk} \sim G_\text{i} + P_\text{j} + G_\text{i} {\times}\, P_\text{j} + L_\text{k(i)} + b, \end{aligned}$$
(3)

where \(G_\text{i}\) is a fixed effect indicating if the line carried a UAS-RNAi construct targeting the gene or if it was a control line; \(P_\text{j}\) is a fixed effect indicating the phenotype of the progeny (either knockdown or stubble control); \(L_\text{k(i)}\) is a fixed effect (nested within \(G_\text{i}\)) indicating the UAS-RNAi construct used to knock down the target gene; and b is a random effect indicating the replicate block in which the viability assay was performed. If only one UAS-RNAi construct was used to knock down the target gene, then \(L_\text{k(i)}\) was excluded from Eq. 3. The effect of knockdown on viability was estimated as the effect of the \(G_\text{i}\,{\times}\, P_\text{j}\) interaction for crosses in which the gene is knocked down and the progeny have wild-type bristles. If the \(G_\text{i}\,{\times}\, P_\text{j}\) interaction has a significant effect on the number of progeny recovered from the cross (\(N_\text{ijk}\)), then there is an effect of knockdown on viability. To test for significance of the interaction term, we used a drop in deviance test to compare the fit of the full model with a model excluding the interaction term.

To assess the effects of RNAi knockdown on male fertility, we crossed UAS-RNAi males to females carrying a Gal4 driver construct that is constitutively expressed under the bag of marbles (bam) promoter (P{bam-Gal4-VP16}) to create male progeny in which the target gene is knocked down the germline (Sartain et al. 2011). The bam promoter drives expression in germ cells after differentiation from the stem cells (Chen and McKearin 2003). We assessed the fertility of the knockdown male progeny by crossing individual 5-day-old males to single 4- to 6-day-old Orergon R (OreR) or Canton S (CanS) virgin females, and we observed all matings to ensure that copulation occurred. We only considered the results of matings in which we observed copulation between the male and the CanS/OreR female to ensure that the fertility assay was not confounded by behavioral effects that interfere with mating success. The females were allowed to lay eggs for 2 days after mating on cornmeal media, they were then transferred to a new vial for 2 additional days of egg laying, and the total number of adult progeny that emerged in both vials were added together as a measure of the fertility of the knockdown male. As a control for each batch, we assessed the fertility of males that were created by crossing bam-Gal4 females with males from the progenitor strains that do not carry the UAS-RNAi constructs.

For each gene, we modeled the number of progeny recovered (\(N_\text{ijk}\)) from matings involving either a male carrying the UAS-RNAi construct or a control male carrying a chromosome from the progenitor line:

$$\begin{aligned} N_\text{ijk} \sim G_\text{i} + L_\text{j(i)} + T_\text{k} + b, \end{aligned}$$
(4)

where \(G_\text{i}\) is a fixed effect indicating if the male carried a UAS-RNAi construct targeting the gene or if it was a control line; \(L_\text{j(i)}\) is a fixed effect (nested within \(G_\text{i}\)) indicating the UAS-RNAi construct used to knock down the target gene; \(T_\text{k}\) is a fixed effect indicating the genotype of the female used to assess fertility (either CanS or OreR); and b is a random effect indicating the replicate block in which the fertility assay was performed. If only one UAS-RNAi construct was used to knock down the target gene, then \(L_\text{j(i)}\) was excluded from Eq. 4. The effect of knockdown on fertility is quantified by \(G_\text{i}\). If \(G_\text{i}\) has a significant effect on the number of progeny (\(N_\text{ijk}\)), then there is an effect of germline knockdown on male fertility. To test for a significant effect of male genotype, we used a drop in deviance test to compare the fit of the full model with a model excluding the male genotype (\(G_\text{i}\)). For some of the genes, only one UAS-RNAi line and one female genotype were used in a single block, and we therefore could not use the drop in deviance to assess the effect of knockdown on fertility. In those cases, we assessed the effect of the male genotype using a single factor ANOVA (equivalent to a Student’s T test): \(N_\text{i} \sim G_\text{i}\). All analyses were performed in the R statistical programming environment (R Core Team 2015).

Data Availability

All divergence data, gene expression data, and results from RNAi experiments are available as Supplemental Files. File S1 contains a description of all Supplemental Data.

Results

Relocated Genes Evolve Fast Because of Relaxed Selective Constraints

We tested if the protein coding sequences of genes that were relocated to other chromosome arms along the D. melanogaster lineage evolve at different rates than inter-chromosome-arm duplicated genes or single-copy non-relocated genes. The ancestral copies of duplicated genes, derived copies, and relocated genes all evolve faster at zero-fold degenerate (amino acid changing) sites than single-copy non-relocated genes (Fig. 2a). Accelerated amino acid sequence evolution can be driven by positive selection, relaxed constraints, or higher mutation rates. The derived copies of duplicated genes evolve faster than single-copy genes at four-fold degenerate (silent) sites (Fig. 2b), suggesting that higher mutation rates could explain the faster evolution of derived copies at zero-fold degenerate sites. However, \(d_\text{N}/d_\text{S}\) is significantly elevated in the ancestral copies, derived copies, and relocated genes relative to single-copy non-relocated genes (Fig. 2c). We therefore conclude that mutational bias cannot entirely explain the faster amino acid sequence evolution of relocated genes.

Fig. 2
figure 2

Divergence and polymorphism-divergence statistics for single-copy non-relocated genes, the ancestral and derived copies of inter-chromosome-arm duplicated genes, and relocated genes are plotted. Divergence estimates are between D. melanogaster and D. simulans at a zero-fold degenerate sites, b four-fold degenerate sites, and c \(d_\text{N}/d_\text{S}\). The distribution of divergence values for single-copy genes is represented by a boxplot, while individual divergence values are shown for each of the other genes as a point (with the median indicated by a horizontal line). Significant differences in divergence when comparing single-copy genes with either ancestral copies, derived copies, or relocated genes are shown by red asterisks (*P < 0.05, **P < 0.005, ***P < 0.0005, and ****P < 0.00005 in a Mann–Whitney U test). d Point estimates of \({\overline{\alpha }}\) are plotted along with the 95% CI. The numbers in parentheses in the x axis labels indicate how many genes are included in each classification

Full size image

Other analyses have found that the derived copies of duplicated genes evolve faster and experience more positive selection than the ancestral copies (Kondrashov et al. 2002; Conant and Wagner 2003; Han et al. 2009; Han and Hahn 2012). We fail to detect significant differences between ancestral and derived copies of inter-chromosome-arm duplicated genes in divergence at zero-fold degenerate sites (P = 0.373), divergence at four-fold degenerate sites (P = 0.553), or \(d_\text{N}/d_\text{S}\) (P = 0.208; all P values from Mann–Whitney U tests). However, we have small sample sizes of ancestral and derived duplicates (14–30 depending on the divergence estimate), which likely limits our power to detect significant differences in evolutionary rates. There are substantially more relocated genes with divergence estimates (34–60), and we detect significantly elevated \(d_\text{N}/d_\text{S}\) in the derived copies of duplicated genes relative to relocated genes (\(P\,{=}\,7.5\,{\times}\,10^{-5}\) in a Mann–Whitney U test). Our results demonstrate that the protein coding sequences of relocated genes evolve faster than single-copy non-relocated genes, and the derived copies of duplicated genes evolve faster than relocated genes.

To distinguish between relaxed selective constraints (decreased purifying selection) and increased adaptive substitutions (positive selection) driving the rapid evolution of relocated genes, we analyzed polymorphism and divergence data. If accelerated evolutionary divergence is driven by positive selection, we expect the ratio of non-synonymous to synonymous substitutions to be greater than non-synonymous to synonymous polymorphisms (McDonald and Kreitman 1991). Only a handful of duplicated and relocated genes have an excess of non-synonymous substitutions (Table 1). In addition, the proportion of ancestral copies, derived copies, or relocated genes with evidence for positive selection is not greater than the proportion of single-copy non-relocated genes with evidence for positive selection (Table 1). Furthermore, the fraction of non-synonymous substitutions fixed by selection (\({\overline{\alpha }}\)) in relocated genes falls below the 95% CI for single-copy non-relocated genes (Fig. 2d). There is also not a significant difference in \({\overline{\alpha }}\) between relocated genes, ancestral copies of duplicated genes, or derived copies (Fig. 2d). In summary, our results provide no evidence that relocated genes experience a disproportionate amount of positive selection, and we conclude that the accelerated evolution of relocated genes is driven by relaxed selective constraints.

Table 1 Counts of D. melanogaster single-copy non-relocated genes, ancestral copies of duplicated genes, derived copies, and relocated genes with no evidence for strong selection, evidence for positive selection, and evidence for negative selection
Full size table

Relaxed constraints on duplicated genes may be present prior to duplication and not necessarily be a result of duplication (O’Toole et al. 2018). To test for relaxed constraints prior to duplication/relocation, we examined the evolution of the D. melanogaster single-copy orthologs of D. pseudoobscura duplicated and relocated genes. We find that \(d_\text{N}/d_\text{S}\) of the D. melanogaster orthologs of D. pseudoobscura duplicated genes is significantly higher than \(d_\text{N}/d_\text{S}\) of single-copy non-relocated genes, and \({\overline{\alpha }}\) of the orthologs of duplicated genes is within the 95% CI of \({\overline{\alpha }}\) of the single-copy genes (Supplementary Fig. S1). The rapid evolution of orthologs of duplicated genes without higher rates of positive selection suggests that duplicated genes are more likely to have single-copy orthologs that evolve under relaxed constraints, consistent with what has been observed in primates (O’Toole et al. 2018). The D. melanogaster orthologs of D. pseudoobscura relocated genes also have elevated \(d_\text{N}/d_\text{S}\) relative to single-copy non-relocated genes, and \({\overline{\alpha }}\) of the orthologs of relocated genes is below \({\overline{\alpha }}\) of non-relocated genes (Supplementary Fig. S1). We therefore conclude that both duplicated and relocated genes are more likely to arise from genes evolving under relaxed constraints.

Relocated Genes are Broadly Expressed, are Highly Expressed in Testis, and have Male-Biased Expression

The derived copies of D. melanogaster inter-chromosome-arm duplicated genes tend to be narrowly expressed in male reproductive tissues, whereas relocated genes are broadly expressed (Meisel et al. 2009). Using available microarray data from 15 adult D. melanogaster tissues, we confirmed that the derived copies of D. melanogaster duplicated genes in our data set are more narrowly expressed (higher \(\tau\)) than single-copy genes (Fig. 3a). In contrast, relocated genes do not significantly differ in their expression breadth from single-copy non-relocated genes (Fig. 3a) or the ancestral copies of duplicated genes (P = 0.900 in a Mann–Whitney U test).

Fig. 3
figure 3

Expression of single-copy non-relocated genes, the ancestral and derived copies of inter-chromosome-arm duplicated genes, and relocated genes are plotted. a Distributions of \(\tau\) for D. melanogaster genes are plotted. Distributions of gene expression in testis from b D. melanogaster microarray data and  c D. pseudoobscura RNA-seq data are plotted. Distributions of \(\log _2{M}/{F}\) in d D. melanogaster whole fly, e D. melanogaster head, f D. pseudoobscura whole fly, and g D. pseudoobscura head are plotted. The distributions for single-copy genes are represented by boxplots, while individual values are shown for each of the other genes as a point (with the median indicated by a horizontal line). Significant differences when comparing single-copy genes with either ancestral copies, derived copies, or relocated genes are shown by red asterisks (*P < 0.05, **P < 0.005, ***P < 0.0005, and ****P < 0.00005 in a Mann–Whitney U test). The numbers in parentheses in the x axis labels indicate how many genes are included in each classification

Full size image

The derived copies of duplicated genes in animal genomes are often testis-expressed (Vinckenbosch et al. 2006; Meisel et al. 2009, 2010; Baker et al. 2012). We indeed find that the derived copies of D. melanogaster duplicates in our dataset are more highly expressed in testis than single-copy non-relocated genes (Fig. 3b) and the ancestral copies of duplicated genes (\(P\,{=}\,2.8\,{\times}\,10^{-3}\) in a Mann–Whitney U test). In addition, D. melanogaster relocated genes are also more highly expressed in testis than non-relocated genes (Fig. 3b). Surprisingly, the derived copies of D. pseudoobscura duplicated genes are not more highly expressed in testis than either the ancestral copies or single-copy genes (Fig. 3c). D. pseudoobscura relocated genes, on the other hand, are more highly expressed in testis than non-relocated single-copy genes (Fig. 3c). We therefore conclude that Drosophila relocated genes are highly expressed in testis, but the testis-expression of the derived copies of duplicated genes is species dependent.

Testis expression is the primary driver of male-biased gene expression in Drosophila (Parisi et al. 2003). In addition, male-biased and testis expression are among the best predictors of evolutionary rates of protein coding genes (Meisel 2011). Because relocated genes evolve fast (Fig. 2a–c) and are testis expressed (Fig. 3b, c), we assessed whether relocated genes also have male-biased expression. D. melanogaster relocated genes do indeed have more male-biased expression than single-copy non-relocated genes in head, but not in whole fly (Fig. 3d, e). D. pseudoobscura relocated genes also have more male-biased expression than single-copy non-relocated genes in head, and in whole fly as well (Fig. 3f, g). Drosophila relocated genes therefore are broadly expressed across many tissues, are highly expressed in male-limited tissues, and have elevated expression in males (Fig. 3). However, unlike the derived copies of duplicated genes, relocated genes do not have limited expression in male-specific tissues.

To assess how the expression profiles of relocated genes affect their rates of evolution, we calculated Spearman’s non-parametric rank order correlation (\(\rho\)) between each of our divergence estimates and expression metrics for D. melanogaster single-copy genes, ancestral copies of inter-chromosome-arm duplicates, derived copies, and relocated genes. Consistent with previous results (Meisel 2011), faster evolution of single-copy non-relocated genes is associated with more male-biased expression in whole fly (higher \(\log _2{M}/{F}\)), narrower expression (greater \(\tau\)), and higher testis expression (Supplementary Fig. S2). Faster evolution of relocated genes is also positively correlated with narrower expression (Supplementary Fig. S2), even though relocated genes are not narrowly expressed (Fig. 3a). In contrast, testis expression levels are not positively correlated with evolutionary rate for relocated genes (Supplementary Fig. S2), even though relocated genes evolve fast (Fig. 2a–c) and are highly expressed in testis (Fig. 3b). We observe similar results for the derived copies of inter-chromosome-arm duplicated genes (Supplementary Fig. S2). We therefore conclude that higher testis expression could explain the faster evolution of relocated genes when compared to single-copy non-relocated genes, but expression breadth is the best predictor of evolutionary rates within relocated genes.

Relocated Genes are Not Disproportionately Essential for Viability

The broad expression of relocated genes suggests that they may be essential for viability. To test this hypothesis, we compared the effects of RNAi knockdown of relocated genes to knockdown of single-copy non-relocated genes in D. melanogaster. We first analyzed the effect of knockdown using randomly inserted UAS-RNAi constructs (GD lines) that do not have any known systemic effect on viability independent of RNAi knockdown of the target gene. Using those data, we find some evidence that relocated genes are less essential than non-relocated single-copy genes. Knockdown of less than a quarter (5/24) of relocated genes causes a significant decrease in viability (Fig. 4a), whereas over half (8/15) of the single-copy non-relocated genes have a significant viability effect when knocked down (Fig. 4b; P = 0.079 in Fisher’s exact test). In addition, we quantified the effect of knockdown on viability, with more negative values indicating a larger effect. The median effect of knockdown on viability is significantly more negative for single-copy non-relocated genes than relocated genes (\(P\,{=}\,1.8\,{\times}\,10^{-4}\) in a Mann–Whitney U test).

Fig. 4
figure 4

The effects of RNAi knockdown on viability and fertility are plotted. Knockdown was performed using a, b ubiquitous expression of Gal4 to assess viability and c, d germline expression of Gal4 to assess male fertility. Only data using GD lines are plotted. RNAi targeted a, c relocated genes or b, d single-copy non-relocated genes. Dots indicate the mean effect of knockdown across replicates, and the vertical bars show the standard error. Each point is a gene, and those colored red have knockdown effects significantly less than zero, indicating decreased a, b viability or c, d male fertility. (Color figure online)

Full size image

We also tested the effect of knockdown using site-specific UAS-RNAi construct insertions (KK lines) that have a known systemic effect on viability (Green et al. 2014; Vissers et al. 2016). Indeed, we observe that the median knockdown effect on viability is more negative using the KK lines (− 13.54; see Supplementary Fig. S3) than the GD lines (1.13; Fig. 4) for relocated genes (P = 0.005 in a Mann–Whitney U test). Therefore, the systemic effects of the KK lines on viability make them unsuited for inferring the essentiality of both relocated and duplicated genes (Kondo et al. 2017). Surprisingly, there is not a more negative viability effect of knockdown using KK lines (− 15.53) than GD lines (− 20.36) for single-copy non-relocated genes.

The broader expression breadth of relocated genes than duplicated genes (Fig. 3a) suggests that relocated genes are more likely to be essential for viability. To test this hypothesis, we compared our results examining the effect of knockdown of relocated and non-relocated genes with published data assessing if ubiquitous knockdown of the derived copies of D. melanogaster duplicated genes induces lethality (Chen et al. 2010). In this approach, if knockdown of a gene causes lethality, the gene is considered essential. We considered knockdown of relocated and non-relocated genes to induce lethality if there were no knockdown progeny recovered in at least 90% of replicate experiments we performed with at least one GD RNAi line, similar to the criteria for considering the lethal effect of knocking down duplicated genes in the published data (Chen et al. 2010). We only considered inter-chromosome-arm duplications, and we only analyzed results from GD lines because of the systemic effects of KK lines described above and previously reported (Green et al. 2014; Vissers et al. 2016; Kondo et al. 2017). We observe that similar proportions of relocated genes (16.7%) and derived copies of duplicated genes (15.0%) cause lethality when knocked down (Table 2). In contrast, 40% of single-copy non-relocated genes causes lethality (Table 2), but this is not significantly different from the fraction of essential relocated genes (P = 0.14 in Fisher’s exact test) or duplicated genes (P = 0.13 in Fisher’s exact test). We therefore conclude that, despite the increased expression breadth of relocated genes, they are not more likely to be essential for viability than the derived copies of duplicated genes. We also observe that, for both relocated and non-relocated genes, there is not a significant correlation between expression breadth and the effect of knockdown on viability (Supplementary Fig. S4). These results suggest that expression breadth is not a reliable proxy for gene essentiality.

Table 2 Counts of relocated genes, derived copies of inter-chromosome-arm duplicates, and single-copy non-relocated genes that are lethal or non-lethal to knockdown. The criterion for lethality is no knockdown progeny recovered in 90% of replicate experiments
Full size table

The genes included in the viability assays are a subset of all single-copy, duplicated, and relocated genes in the D. melanogaster genome. We tested if they are representative of the patterns of divergence, selection, and expression, we observe in the full set of single-copy, duplicated, and relocated genes. We confirmed that the derived copies of duplicated genes that were included in the viability assays do indeed evolve faster than the single-copy non-relocated genes included in the viability assays at all classes of sites (Supplementary Fig. S5). The relocated genes included in our viability assays also evolve faster than the single-copy non-relocated genes at four-fold degenerate sites (Supplementary Fig. S5). In addition, the derived copies of duplicated genes in our viability assays are narrowly expressed, and they have more male-biased expression (greater \(\log _2{M}/{F}\)) in whole fly than the single-copy genes (Supplementary Fig. S5). The relocated genes included in the viability assays also have more male-biased expression than the single-copy non-relocated genes (Supplementary Fig. S5). We therefore conclude that the genes included in the viability assays are in general representative of single-copy, duplicated, and relocated genes in the D. melanogaster genome.

Relocated Genes are not Disproportionately Essential for Male Fertility

Relocated genes are highly expressed in testis (Fig. 3b, c), suggesting that their products may perform essential roles in spermatogenesis. To test this hypothesis, we assessed the fertility of D. melanogaster males in which relocated genes were knocked down in the male germline using GD UAS-RNAi lines. We compared our results to germline knockdown of single-copy non-relocated genes. We also quantified the effect of knockdown on male fertility, with more negative values indicating a larger decrease in male fertility relative to controls.

Despite their higher testis expression, we do not find evidence that relocated genes are more essential for male fertility than single-copy non-relocated genes. Germline knockdown of only 3/23 relocated genes induced a significant decrease in male fertility (Fig. 4c), compared to nearly half (6/13) of single-copy non-relocated genes (Fig. 4d; P = 0.046 in Fisher’s exact test). In addition, the median knockdown effect on male fertility is more negative for non-relocated genes than relocated genes (Fig. 4c, d; \(P\,{=}\,1.8\,{\times}\,10^{-5}\) in a Mann–Whitney U test).

We also tested the effect of knockdown on male fertility using site-specific UAS-RNAi construct insertions (KK lines) that have a known systemic effect on viability (Green et al. 2014; Vissers et al. 2016). Unlike the viability effects, we do not detect a systemic effect of the KK lines on male fertility (Supplementary Fig. S3). In fact, the median effect of knockdown on male fertility using GD lines (− 9.21) is more negative than with KK lines (6.52) for relocated genes (P < 0.01 in both paired and unpaired Mann–Whitney U tests). The median knockdown effect on male fertility for single-copy genes is also more negative for GD lines (− 40.5) than KK lines (− 8.58; P = 0.039 in a paired Mann–Whitney U test). Our results therefore suggest that both GD and KK lines can be used to assess if genes are necessary for male fertility. Indeed, when we analyze data from GD and KK lines together, we observe the same general trends as each set of lines separately (Supplementary Fig. S6). Specifically, the average effect of knockdown on male fertility is greater for single-copy non-relocated genes (− 19.82) than relocated genes (− 2.02; P = 0.00086 in a Mann–Whitney U test). In addition, male fertility under germline knockdown is negatively correlated with testis expression level for both relocated and non-relocated genes (Supplementary Fig. S7), suggesting that testis expression is predictive of the effects of germline knockdown.

Finally, we tested if the genes included in the fertility assay are a representative subset of all non-relocated and relocated genes in the D. melanogaster genome. The relocated genes in our fertility assay evolve faster at four-fold degenerate sites than the non-relocated genes, and they have more male-biased expression than the non-relocated genes (Supplementary Fig. S8). However, the relocated genes in our fertility assay are not more highly expressed in testis than the non-relocated genes (Supplementary Fig. S8). This is because both the relocated and non-relocated genes included in our fertility assay have higher testis expression than the genes not included in the fertility assay (P = 0.024 and P = 0.0019 in a Mann–Whitney U test for relocated and non-relocated genes, respectively). It is therefore possible that the result of our fertility assay was biased by selecting relocated and non-relocated genes with higher testis expression than average.

Discussion

Gene relocation occurs frequently in eukaryotic genomes, and relocated genes may be as common as canonical inter-chromosomal duplicated genes (Bhutkar et al. 2007; Wicker et al. 2010; Han and Hahn 2012). In addition, relocated genes have been hypothesized to evolve under more selective constraints than the derived copies of dispersed duplicated genes, and some studies have suggested that they are as conserved as single-copy non-relocated genes (Meisel et al. 2009; Ciomborowska et al. 2013). We found that relocated genes evolve faster than non-relocated genes, and there is no evidence that this faster evolution is driven by positive selection (Fig. 2 and Table 1). We therefore conclude that Drosophila relocated genes evolve fast because they are under relaxed constraints. The derived copies of dispersed duplicates evolve even faster than relocated genes, consistent with the hypothesis that relocated genes are under more selective constraints than duplicated genes.

Relocated genes are broadly expressed, while the derived copies of inter-chromosome-arm duplicates are narrowly expressed (Fig. 3). Broad expression and high expression levels are associated with slower evolution of Drosophila genes (Larracuente et al. 2008; Meisel 2011). It is therefore not surprising that the derived copies of inter-chromosome-arm duplicates evolve faster than both non-relocated and relocated genes (Fig. 2). However, despite their broad expression, relocated genes also evolve faster than non-relocated genes (Fig. 2). Our results suggest that, even though expression level and breadth are predictive of evolutionary rates within categories of genes (Larracuente et al. 2008; Meisel 2011, Supplementary Fig. S2), expression differences between categories of genes are poorly associated with evolutionary rate differences between categories.

The broad expression and high testis expression of relocated genes led us to hypothesize that they are essential for viability and male fertility (Fig. 3). However, our RNAi experiments revealed that the relocated genes are less essential for viability and male fertility than single-copy non-relocated genes (Fig. 4 and Table 2). This is consistent with our results that suggest relocated genes evolve faster than single-copy non-relocated genes because relocated genes are under relaxed selective constraints (Fig. 2 and Table 1). These results also demonstrate that functional analyses that complement expression measurements are necessary to identify differences in selective constraints acting on different classes of genes.

Our inference of relaxed constraints comes, in part, from analyses of extant DNA sequences and effects of RNAi knockdown on extant relocated genes. It is possible that relocated genes (and duplicated genes) could have experienced strong positive or purifying selection immediately after duplication (or loss of the ancestral paralog), and the signatures of those selection pressures were lost over time. Analysis of a large panel of young duplications and relocations is necessary to test this hypothesis (e.g., Masly et al. 2006; VanKuren and Long 2018). However, our observation that the D. melanogaster orthologs of D. pseudoobscura relocated genes also evolve under relaxed constraints (Supplementary Fig. S1) suggests that relocated genes have evolved under relaxed constraints for most of their histories.

An excess of genes has been relocated from the X chromosome to the autosomes across the Drosophila genus (Meisel et al. 2009; Vibranovski et al. 2009b). Three hypotheses could explain this phenomenon. First, the female-biased transmission of the X chromosome may favor X-linked female-beneficial mutations and prevent the fixation of male-beneficial mutations on the X (Rice 1984). This sexually antagonistic selection could favor the X-to-autosome relocation of genes that perform male-beneficial functions (Wu and Xu 2003). Second, expression of the X chromosome is down-regulated in spermatogenesis (Vibranovski et al. 2009a; Meiklejohn et al. 2011), which could favor the X-to-autosome relocation of genes that have beneficial effects when highly expressed in spermatogenesis (Betrán et al. 2002; Emerson et al. 2004; Meisel et al. 2009). Third, there may be a mutational bias in favor of X-to-autosome duplications (Metta and Schlotterer 2010; Díaz-Castillo and Ranz 2012), but this is not supported by copy number polymorphisms (Schrider et al. 2011). We find no evidence that knockdown of relocated genes disproportionately affects male fertility (Fig. 4). However, relocated genes are more highly expressed in testis than non-relocated single-copy genes (Fig. 3), which holds true even if we only consider autosomal genes (P < 0.05 for both D. melanogaster and D. pseudoobscura). Our gene expression analysis therefore provides some evidence that the X-to-autosome relocation bias could be driven by selection in favor of higher testis expression on the autosomes, but additional work is necessary to fully test this hypothesis.

There are three important technical limitations of our experiments that could have reduced our ability to detect male-specific functions of relocated genes. First, we used ubiquitous and germline knockdown to assess if genes are essential for viability and male fertility. We chose to assay the effect of germline knockdown because relocated genes are highly expressed in testis (Fig. 3), and the derived copies of duplicated genes are hypothesized to be specialized for germline functions (Marques et al. 2005; Vinckenbosch et al. 2006; Potrzebowski et al. 2008; Meisel et al. 2010; Tracy et al. 2010). We demonstrated that, even though they are highly expressed in testis, relocated genes are not disproportionately essential for spermatogenesis (Fig. 4). However, our results may be biased by the Gal4 driver that we selected (bam), which is expressed early in spermatogenesis immediately after differentiation from the stem cell niche (Chen and McKearin 2003). Knockdown in later stages of spermatogenesis or in somatic testis tissue may reveal testis-biased functions for relocated genes. Second, germline knockdown of a panel of duplicated genes would allow for a comparison of male-specific functions between relocated genes and derived copies of inter-chromosome-arm duplicates. This may reveal additional insights into the causes of differences in the evolutionary rates of duplicated and relocated genes (Fig. 2). Third, we only measured male fertility in a non-competitive environment. Females of many Drosophila species mate with multiple different males, and sperm from the males compete to fertilize eggs within females (Parker 1970; Gromko et al. 1984). Seminal proteins produced by male accessory glands are considered to be the key modulators of male–male and male–female interactions in the female reproductive tract (Harshman and Prout 1994; Adams and Wolfner 2007; Ravi Ram and Wolfner 2007), but some duplicated testis-expressed genes also play important roles in this male–male competition (Rettie and Dorus 2012; Yeh et al. 2012). It is therefore possible that testis-expressed duplicated and relocated genes may have greater effects on male fertility than non-relocated single-copy genes if experiments are performed using multiply mated females. Additional work is necessary to further investigate each of these three limitations of our experiments.

In conclusion, we demonstrated that Drosophila relocated genes evolve fast, and this rapid evolution is likely the result of relaxed selective constraints (Fig. 2 and Table 1). This differs from mammals, where relocated genes evolve under strong purifying selection (Ciomborowska et al. 2013). Drosophila relocated genes are also less essential for viability and male fertility than single-copy non-relocated genes (Fig. 4), which is consistent with relocated genes evolving under relaxed constraints. In addition, the derived copies of inter-chromosome-arm duplicates appear to be under even more relaxed constraints than relocated genes, which allows them to evolve even faster. Additional work is necessary to determine the causes of differences in selection pressures acting on relocated genes in mammals and Drosophila.