Introduction

Convergent phenotypic evolution, which results from similar selection pressures in independent lineages, is a common, undisputed property of animal and plant evolution. The frequency of convergent adaptive molecular evolution, however, remains an open question. Convergent adaptive molecular variants may include both amino acid polymorphism [1]–[6] and amino acid divergence [7]–[9]. The relatively few examples of such convergent changes are based largely on molecular and functional analysis of proteins with well-defined structures and functions; consequently, the inference of convergent adaptive polymorphism or divergence allows plausible arguments to be made regarding the phenotypic adaptive effects of particular mutations in different lineages. Most gene products are not understood sufficiently well to use such an approach.

An alternative, statistical approach is to use molecular population genetic data to ask whether directional selection is repeatable over evolutionary time. For example, the McDonald-Kreitman test [10], which uses contrasts of polymorphic and fixed variants to test the neutral model of molecular evolution, requires no knowledge of protein structure or specific functions of residues or domains. This allows one to ask the general question of whether a gene with a history of recurrent adaptive protein evolution in one species is likely to have a similar selective history in other species; that is, is directional selection idiosyncratic or predictable? For example, the Drosophila seminal fluid protein gene Acp26Aa was first inferred to have a history of recurrent adaptive protein evolution in the melanogaster subgroup [11]. Acp26Aa was later shown to be under such selection in the obscura group of Drosophila [12], which diverged from the melanogaster subgroup tens of millions of years ago.

The Drosophila innate immune system transcription factor, Relish, is a potentially interesting gene for addressing the question of predictable versus idiosyncratic directional selection. Previous studies demonstrate that the innate immune system, a highly conserved pathway from insects to humans, is vulnerable to signaling disruption by both bacterial and viral pathogens. Moreover, Relish activation and/or signaling repeatedly emerges as a pathogen target. In a vertebrate system, Neish et al. [13] demonstrate that Yersina bacteria disrupts phosphorylation of the human Relish homolog, NF-κB. In an insect system (Drosophila melanogaster), Lindmark et al. [14] and Thoetkiattikul et al. [15] demonstrate Relish signaling disruption by various bacteria and a polydnavirus, respectively.

Compromised immune response in the presence of these pathogens, combined with documented Relish-pathogen interactions, makes this locus a likely target for repeated host-pathogen evolutionary interactions in distantly related taxa. Nevertheless, population genetic data for the Relish locus provided strong evidence of adaptive divergence in D. simulans, but no evidence of adaptive divergence in D. melanogaster [16]. Similarly, the termite Relish locus appears to be rapidly evolving in a subset of lineages [17]. Although Relish likely contributes to immune function in all species examined, the evolutionary dynamics associated with this locus are dramatically different across lineages. To further investigate the repeatability of directional selection at this locus in Drosophila, we characterized the evolutionary forces acting on Relish across three highly diverged sister species-pairs, D. mojavensis/D. arizonae, D. yakuba/D. teissieri, and D. pseudoobscura/D. miranda.

Results and Discussion

Levels of synonymous and nonsynonymous polymorphism at Relish (Table 1) were consistent with previous descriptions Drosophila mojavensis/D. arizonae [18], whereas lower than expected levels of variation were estimated for D. yakuba [19] and D. pseudoobscura [20]. Levels of Relish synonymous divergence in these species pairs were typical of those estimated at other genes. Levels of non-synonymous divergence (scaled to synonymous divergence), however, were highly heterogeneous across species, suggesting the protein evolutionary rates vary due to heterogeneous selection regimes (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Polymorphism and divergence for all species examined as well as previously published estimates for D. melanogaster and D. simulans.

https://doi.org/10.1371/journal.pone.0000442.t001

We used the McDonald-Kreitman test to determine whether synonymous and non-synonymous variation at Relish supports the hypothesis of adaptive protein evolution. All three species pairs failed to reject the null hypothesis of neutral evolution (Table 2). The D. simulans/D. melanogaster species pair is the only one associated with evidence of adaptive protein evolution at Relish [16].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. McDonald-Kreitman tests of Relish variation for four species pairs.

https://doi.org/10.1371/journal.pone.0000442.t002

Low levels of polymorphism at Relish in D. yakuba and D. pseudoobscura could be due to recent, strong directional selection at Relish or at linked sites. We used the HKA test [21] to determine whether the polymorphism-to-divergence ratios at Relish were unusual compared to those from the putatively neutral loci Xdh in D. yakuba/D. teissieri, (J. Comeron pers. comm.) and Adh in D. pseudoobscura/D. miranda [22]. Only the D. yakuba/D. teissieri data rejected the null (χ² = 6.39, p = 0.01), which is consistent with linked selection in this region of the D. yakuba genome. The Relish gene is near the middle of chromosome arm 3R in D. yakuba (D. yakuba genome assembly, v2), which suggests that this result is not due to sampling a large region of reduced polymorphism near centromeres and telomeres [23]. Further analysis of the regions flanking Relish is necessary to determine the extent of reduced polymorphism in this genomic region.

The Relish population genetic data from three, distantly related, Drosophila species pairs generally supports the idea that Relish evolution in the D. melanogaster/D. simulans pair is highly unusual. Previous analyses of D. melanogaster/D. simulans suggest that evidence of strong directional selection at Relish is most likely a D. simulans-lineage phenomenon [16]. This finding raises the interesting question of what D. simulans-specific biological or historical attributes caused the highly unusual history of a key component of the insect immune system.

Methods

Population samples of Relish were sequenced from inbred lines of D. yakuba (P. Andolfatto), D. tessieri (M. Long), D. mojavensis (W. Etges and Tucson Stock Center), D. arizonae (W. Etges), D. pseudobscura (M. Noor), D. miranda (Tucson Stock Center). Most data were obtained by direct sequencing. For the few lines with residual heterozygosity, PCR products were cloned in PCR-4 vector (Topo TA cloning kit, Invitrogen) and individual colonies were sequenced. Population genetic estimators and tests statistics were calculated in DnaSP v.4.0 (Rozas et al. 2003). Sequence data for this paper have been submitted to Genbank under accession numbers EF494515-EF494539.

Acknowledgments

The authors thank the two anonymous reviewers for their valuable suggestions.