From molecules to mating: Rapid evolution and biochemical studies of reproductive proteins

Abstract

Sexual reproduction and the exchange of genetic information are essential biological processes for species across all branches of the tree of life. Over the last four decades, biochemists have continued to identify many of the factors that facilitate reproduction, but the molecular mechanisms that mediate this process continue to elude us. However, a recurring observation in this research has been the rapid evolution of reproductive proteins. In animals, the competing interests of males and females often result in arms race dynamics between pairs of interacting proteins. This phenomenon has been observed in all stages of reproduction, including pheromones, seminal fluid components, and gamete recognition proteins. In this article, we review how the integration of evolutionary theory with biochemical experiments can be used to study interacting reproductive proteins. Examples are included from both model and non-model organisms, and recent studies are highlighted for their use of state-of-the-art genomic and proteomic techniques.

Significance

Despite decades of research, our understanding of the molecular mechanisms that mediate fertilization remain poorly characterized. To date, molecular evolutionary studies on both model and non-model organisms have provided some of the best inferences to elucidating the molecular underpinnings of animal reproduction. This review article details how biochemical and evolutionary experiments have jointly enhanced the field for 40 years, and how recent work using high-throughput genomic and proteomic techniques have shed additional insights into this crucial biological process.

1. Introduction

Sexual reproduction, while prevalent along every branch of the tree of life, remains a challenge for evolutionary biologists to explain [1]. Asexual reproduction offers the advantages of propagating twice the genetic material, lacks the costs associated with finding mates, and can more rapidly establish favorable epistatic effects [2]. Early mathematical models also supported asexual reproduction as the optimal reproductive strategy. However, in more realistic scenarios of dynamic ecosystems with changing environments and co-evolving symbiotes, frequent recombination is needed and natural selection favors sexual reproduction [3,4]. Given the breadth and diversity of sexually reproducing organisms, it is no surprise that various strategies have evolved to improve reproductive success. In animals, males often perform various courtship displays [5–7], deliver pheromones that affect female behavior and physiology [8–10], and regulate the contents of their ejaculate based on female quality [11,12]. Similarly, to improve mate fitness and quality, females must be able to discriminate between these cues for honest or dishonest signals of fitness [13]. Both the male and female characteristics involved can be modified by sexual selection. Under sexual reproduction, mates must be procured to provide complementary genetic material, much like a predator capturing prey for energy and nutrient acquisition. Just as predators and prey often evolve through arms race dynamics, the continual adaptation between elaborate male traits and female perception represents one of the most well characterized examples of rapid, exacerbated co-evolution [14].

The literature is rich with examples of co-evolving sexually selected traits [15]. For historical reasons, the majority of study systems have been visible characteristics such as body size, coloration, mating behaviors, and secondary sexual traits [16]. In recent decades, as molecular biology and biochemistry have advanced, research on sexually selected traits has broadened to include the study of reproductive proteins [14], which we broadly define as any polypeptide directly involved in reproduction. While all reproductive proteins may be subject to sexual selection, the most interesting examples are likely those that directly bind molecules derived from the other sex: examples include pheromones and their cognate receptors [17], interacting egg and sperm surface proteins [18], and seminal proteins that alter female physiology [19]. A recurring theme among reproductive proteins is rapid evolution. As sexual reproduction is an essential biological process for most animals, one might expect that the majority of the reproductive proteins would be under strong negative selection to maintain compatibility. However, the recurring pattern of rapidly evolving reproductive proteins has been observed in both vertebrates and invertebrates at several stages of reproduction [14]. Because selection is most likely to act on functionally important residues in a protein, signatures of positive Darwinian selection can often guide further investigation into their underlying biochemical mechanisms [20], with studies of reproductive proteins serving as exemplars of applying molecular evolutionary techniques to characterize protein function [21–25].

Near the turn of the century and following the completion of the human genome project, a surge of high throughput technologies emerged which have altered the size and scope of questions that biologists can now ask. Various next-generation sequencing (NGS) platforms permit de novo analysis of whole genomes and transcriptomes for both model and non-model organisms [26,27]. Likewise, advances in mass spectrometry (MS) now provide the opportunity to qualitatively and quantitatively characterize whole proteomes [3,28,29,30]. These techniques have additionally been adapted to a wide array of other specific “omic” applications (e.g., metabolomics, phosphoproteomics, pharmacogenomics), but both NGS- and MS-based approaches are quickly becoming the standard for the initial characterization of any biological system [23]. Here we review the biochemical investigations of many reproductive proteins that span various levels of reproduction: pre-copulatory behavior (pheromones), copulation (seminal proteins), and fertilization (sperm/egg proteins). Recent studies in the field have employed NGS- and MS-based approaches, and we discuss how applying such “omics” techniques to reproductive systems may be further integrated with detailed mechanistic and theoretical evolutionary models.

2. Molecular evolution and models of sexual selection

The molecular evolution of any given trait is shaped by neutrality or some form of selection (balancing, directional, or disruptive) (Fig. 1A). Balancing selection reduces genetic diversity and stabilizes a trait at some optimum phenotype. Disruptive selection is the opposite of balancing selection and favors individuals with extreme phenotypes. Finally, directional selection shifts a trait towards a single extreme. Two suites of statistical tests which use either allele frequencies or nucleotide substitutions have been developed, and each tests for selection on relatively different time scales. In the first suite of analyses, assumptions are made concerning the rates at which specific mutations are accumulated and distributed among alleles, often within and between populations. These tests are particularly valuable for identifying recent selection following selective sweeps, but are also heavily influenced by population demographics and bottlenecks (for more thorough review, see [31]). The second set of analyses compares the frequency of nucleotide substitutions at codons within genes – usually between species – and describes trends on relatively longer time scales. In the absence of selection, most nucleotide substitutions (and amino acid substitutions) are free to accumulate at the basal mutation rate. The rate of synonymous substitutions (d_S) provides an estimate of this mutation rate, and under neutrality, non-synonymous substitutions (d_N) should similarly occur, yielding a ratio of d_N/d_S ≈ 1. Because most non-synonymous substitutions alter the tertiary structure of a protein and negatively impact function, non-synonymous substitutions should occur more rarely (d_N/d_S < 1). Residues where non-synonymous substitutions are disfavored are described as under negative or purifying selection [20,32–34]. Unsurprisingly, the average d_N/d_S across the protein coding sequence for most genes is less than one, and in humans, the genome-wide average d_N/d_S ∼ 0.25 [35]. The purging of deleterious mutations by purifying selection often results in stabilizing selection of a trait. However, under situations where rapid mutation may be adaptive, non-synonymous substitutions can accumulate more quickly than the mutation rate (d_N/d_S > 1) and the trait is described as under positive selection [32]. The forces leading to positive selection often generate directional selection, but disruptive selection is also possible when nearly any deviation from the mean is similarly favorable.

Fig. 1.

An introduction to trait selection and a summary of the G matrix model of sexual selection described by Lande [41] (and graphics adapted from Mead and Arnold [210]). (A) When a trait is under selection, three different modes are used to describe the direction that the mean moves: stabilizing selection (same mean, less variation), disruptive selection (moves towards the extremes), or directional selection (moves away from the mean along a single trajectory). (B) A mathematical representation of the G matrix, where in the simplified form a change in the mean of a trait (Δz̄) is equal to its genetic structure (G) times some selective force (β) plus a mutation constant (u). In the expanded form, the two traits being observed are a male display (d) and the corresponding female preference for the display (p). The G matrix is composed of the genetic variation for both d and p (V_d and V_p), and covariance (or genetic coupling) between the two traits (C_dp). Both natural (β_N) and sexual (β_S) selective forces can act on both the male display (β_Nd, β_Sd) and the female preference (β_Np, β_Sp), and each trait is subject to independent mutations (u_d and u_p). When there is weak or no genetic coupling (C_dp ∼ 0), the matrix simplifies such that the change in male display is only dependent on the direct selection acting on the display, and similarly for female preference. Therefore, male traits will evolve until female preferences are matched, represented by different evolutionary trajectories (dashed lines) converging towards an optimum (solid diagonal line) (C). However, when there is genetic coupling (C_dp > 0), selection on the male display can “pull” the female preference in a particular direction (indicated by the cross multiplication of C_dp and β_d in B), and likewise for selection on the preference. Because of this deviation from independent selection towards the typical optimum, it is possible for male displays and female preferences to undergo “runaway selection” and evolve indefinitely, yielding ever increasingly more elaborate ornaments (D).

Since each residue differentially contributes to a given protein's structure and function, and it is likely that all three forces of selection are simultaneously acting on protein-coding genes to different degrees. For example, with serine proteases, most of the protein surface is covered in polar residues that are functionally neutral and highly interchangeable. Within the active site, purifying selection preserves the catalytic triad of serine, histidine, and aspartate that are critical for enzymatic activity, and mutations are rare save for cases of atypical function [36]. However, for select serine proteases that are involved in apoptosis, adaptive response to pathogens which cause cell death has promoted positive selection on active site residues that mediate inhibitor binding and substrate specificity [37]. While the identification of neutrally evolving sites and those under purifying selection can be advantageous for understanding protein function, both forces are common in maintaining protein function and do not necessarily reflect adaptation to specific stimuli. Hence, greater interest is often placed on sites under positive selection, and various statistical packages exist to compute d_N/d_S scores along phylogenetic trees for both whole genes and specific residues [20].

While positive selection and rapid evolution have been documented in a range of systems [33,38–40], they are practically hallmarks of interacting reproductive proteins [14]. Over the last few decades, various quantitative genetic models have been developed to address how elaborate male traits and female preferences may evolve. While qualitatively described by Fisher in the 1930s, Lande [41] was the first to formalize the theory using a genetic correlation matrix (Fig. 1B–D). To illustrate this, assume that females of a given species have a preference for some male ornament, such as large, colorful peacock tails. If there is some heritable component to both the male ornament and the female preference, female peacocks with strong preferences and males with bright tails should produce daughters and sons who carry both traits, leading to genetic correlation and linkage disequilibrium. This process may continue iteratively through generations, which can lead to greater genetic association, stronger preferences, and more pronounced ornaments as part of a model of runaway selection [16]. This has alternatively been dubbed “the sexy son” hypothesis [42]. However, various factors may restrain the characteristics from evolving indefinitely. Elaborate ornaments may be energetically costly, and the ornament and preference may evolve to an upper limit until they are balanced by natural selection (“sensory bias”) [43]. There are also instances where choosing to mate with males with large ornaments may confer a cost to the female, such as reduced viability and/or increased risk of predation, leading to antagonism between mating preference and survivorship (“sexual conflict”) [44]. Alternatively, the male ornament may itself be indicative of some underlying adaptive trait, such as increased offspring survival or fecundity (direct benefits) [45,46]. In this example, both male peacock tails and female perception are complex, polygenic traits. Techniques such as high resolution quantitative trait loci (QTL) mapping can approximate the relative contribution of individual genes to these traits; however, these methods are not sensitive to low-impact genes, require high quality genome sequences, and are highly labor intensive.

The archetype of male ornaments and female preferences provides a framework that also applies to interacting reproductive proteins. Elaborate male anatomical displays and their perception within the female nervous system are almost certainly pairs of complex, polygenic traits. However, reproductive protein dynamics may be as simple as a single male-derived molecule docking with a female receptor through protein–protein interactions [47,48]. This one-to-one stoichiometry permits a more tractable environment to study co-evolution, theoretically at single amino acid resolution given sufficient phylogenetic data [49]. The use of NGS and MS-based proteomics are rapidly increasing the feasibility of such studies, as will be described below.

3. Protein pheromones in vertebrates

One of the largest barriers to mating is often the search for potential mates, particularly when population densities are low. Volatile chemoattractants in the form of pheromone signals are widely used by both invertebrates and vertebrates. The term “pheromone” was originally coined by Karlson and Luscher in 1954, derived from the Greek words pherin (“to attract”) and hormon (“to stimulate”) [50]. The first described pheromone was bombykol, a long chain alcohol released by female silkmoths to attract conspecific males [51]. Since this original discovery, a wide range of pheromone molecules have been described that not only attract potential mates, but also serve as territory markers [52], signal species identity [53], coordinate migratory behavior [54], and prime the nervous system to sense additional pheromone signals [55]. While volatile odorants are the most well-documented and characterized type of pheromones, many vertebrates also utilize water-soluble peptide or protein pheromones [17,56]. In contrast to their volatile counterparts, which are generally synthesized as part of complex enzymatic cascades, peptide/protein pheromones are direct gene products whose evolutionary histories can be more readily probed by sequencing and other standard molecular biological techniques.

Differences in pheromone chemistry have resulted in the evolution of specialized anatomy for their detection. The olfactory system of terrestrial vertebrates includes two discrete sets of neuronal epithelia: the main olfactory epithelia (MOE) and vomeronasal epithelia (VNE) [57,58]. In general, the MOE senses generic odorants while the VNE is specialized to detect pheromones and similar heterospecific cues (although there is evidence for significant crosstalk between these systems and a few exceptions to the rule [59]). There is further partitioning within the VNE such that neurons express one of two main types of G protein coupled receptors (GPCRs): vomeronasal type 1 receptors (V1Rs) that detect volatile pheromones [60], or vomeronasal type 2 receptors (V2Rs) that bind peptide or protein pheromones [61]. One of the goals of modern pheromone research has been to precisely characterize the co-evolution between specific protein pheromones and their cognate V2Rs.

Despite significant advancements towards independently characterizing pheromone and receptor profiles, pairing specific receptors with their ligands has remained a challenge. Protein pheromones are generally purified using various chromatographic separations, with their identity ascertained through joint sequencing and mass spectral characterization. Receptor identification, by comparison, has relied predominantly on molecular and immunohistochemical techniques. Both V1R and V2R receptor types were identified in rodents through the preparation of single-cell cDNA libraries, subtractive hybridization against MOE cDNA to remove genes common to all olfactory neurons, and localization verified by in situ hybridization [60,61]. Thanks in part to these early studies, the majority of V2R sequences have been mined by homology search from completed genome sequences. Within the mouse and rat genomes, there are at least 209 and 168 V2R genes respectively [62]. As GPCRs, V2Rs contain seven helical transmembrane domains that anchor them to the plasma membrane, but there are also large N terminal extracellular domains that were postulated to be the site of pheromone binding [58]. The d_N/d_S ratio for the N-terminal extracellular domain is approximately three times higher than the mean d_N/d_S for the whole protein, and 27 positions were specifically detected to be under positive selection [62]. V2Rs do not readily express in cell culture and require specific chaperones to translocate to the plasma membrane [63]. The current approach to identifying receptor:ligand pairs involves co-labeling neurons with probes for specific V2Rs and for signals of activation upon pheromone treatment. Riboprobes with varying degrees of degeneracy can be designed to detect one or more V2R sequences by in situ hybridization [64]. Neuronal activation can be detected using a number of methods: calcium imaging (with neurons pre-loaded with fluorescent Ca²⁺probes suchasFura-2/AMorendogenously expressing transgenic reporters) [65], immediate-early gene activation (such as c-Fos or Ebr1, with either mRNA detected by in situ hybridization [64] or protein by immunohistochemistry [66]), or the amino acid derivative agmatine (which selectively enters activated neurons and can later be detected by immunohistochemistry [67]). In a recent study examining mouse MOE and VNE transcriptomes, there were few differences between male and female receptor profiles, suggesting that other neuronal features or changes in the central nervous system are responsible for sex-specific pheromone effects [68].

Only a few murine protein families have been demonstrated to activate V2R-containing neurons, including major urinary proteins (MUPs) [69], exocrine secreted peptides (ESPs) found in tears [70], and select MHC peptides [71]. Both the MUPs and ESPs are highly duplicated, multigene families that display variable isoform expression depending on genetic background, sex, and age. MUPs are part of the lipocalin superfamily of binding proteins, and the NMR solution structure revealed a large hydrophobic pocket which binds secreted metabolites [72]. While mouse urine has long been known to contain a repertoire of pheromones, the original sole function of MUPs was thought to be binding volatile pheromones and preventing their rapid evaporation upon release into the environment, extending the longevity of the signal. However, bioassays testing bacterially expressed recombinant MUPs (which lack the volatile component of mouse urine) revealed that the proteinaceous component itself can act as a pheromone; specifically, exposure to recombinant MUPs induced male aggressive behavior [69]. MUP expression is highly variable between different inbred mice strains; for example, mass spectral analysis identified MUP17 as highly expressed in both wild and C57BL/6 male mice, yet BALB/C males only expressed trace levels. MUP17 plays a key role in female sexual behavior, and is also referred to as “darcin,” in reference to the amorous male protagonist of Pride and Prejudice. Recombinant darcin alone increased female mating receptivity, but also promoted learning of associated volatile compounds. While females were not innately attracted to many volatile compounds in male urine, if darcin was exogenously added to the volatile mixture when applied to females, repeat exposure to the volatile components alone similarly increased receptivity. Hence, MUP bioactivity may rely on both the lipocalin-like binding site and additional protein–protein interactions with V2R receptors [73,74]. It is noteworthy that hamster vaginal secretions contain a similar lipocalin molecule, aphrodisin, which is also hypothesized to serve as a carrier for volatile compounds to the male VNE [75].

Similar to the MUP family, the exocrine secreted peptide (ESP) family is a highly duplicated multigene family that performs an array of pheromone activities in mice. The first identified member of the family, ESP1, was discovered in the extraorbital lacrimal gland as a male-specific component that stimulated female sexual behaviors. The bioactive component of this gland secretion was isolated by anion exchange chromatography, and the identity of ESP1 ascertained by protein sequencing using Edman degradation and mass spectrometry. A subsequent genome-wide search localized ESP1 (along with the other 23 tandemly repeated genes) to chromosome 17 [76]. Co-labeling of ESP1-activated neurons with specific V2R probes revealed V2Rp5 to be the specific receptor for ESP1 [70]. The 3D structure of ESP1 was determined by NMR, and when supplemented with biochemical data from both mutagenesis and affinity chromatography experiments, charge interactions were suggested to drive binding between ESP1 and V2Rp5 [77]. Surprisingly, the predominantly alpha helical structure of ESP1 resembles that of the sea hare pheromone attractin [78]. At least ten of the ESP genes (ESP2-11) are expressed in both male and female mice, usually in more than one type of tear gland [76]; however, ESP36 is at least one example of a female-specific isoform, and ESP22 is a juvenile-specific isoform [79]. When recombinant ESP22 was applied to female mice, males were less sexually active, and it was hypothesized that the function of ESP22 may be to discourage mounting attempts on juveniles [80]. The specific receptors and roles of the other ESP family members remain unknown, but presumably perform additional pheromone functions as part of rodent communication [81].

In addition to mammals, protein pheromones have also been well studied in salamanders [47,82,83]. As basal tetrapods, amphibians are particularly well-suited to understanding the early evolution of pheromones in terrestrial vertebrates. Amphibian protein pheromones evolved at least 350 million years ago [84], and have been studied in two different salamander families: Salamandridae (newts) and Plethodontidae (lungless salamanders). The first characterized peptide pheromone was the decapeptide sodefrin, which is released from the cloaca of male firebelly newts (Cynops pyrroghaster) to attract gravid females during the mating season. Sodefrin was originally identified by separating male abdominal gland extract using reversed phase liquid chromatography, screening the chromatographic fractions for ability to attract female newts, and analyzing the single bioactive fraction by gas–liquid phase protein sequencing [85]. The closely related species Cynops ensicauda uses a similar peptide termed silefrin that varies from sodefrin by only two substitutions (P3L and L8Q), yet the activity of each peptide is species-specific [86]. Recently, a variant of sodefrin was found in a single population of C. pyrroghaster with a highly conservative L8V mutation, and it was only effective at attracting local females [87]. These two natural sodefrin variants suggest that residue eight likely plays a critical role in receptor binding. Subsequent cDNA sequencing revealed that the decapeptide was cleaved from a larger ∼20 kDa precursor using a dibasic cleavage site, similar to many peptide hormones such as insulin and glucagon [88]. Homologs of the precursor known as sodefrin precursor-like protein (SPF) have been identified in additional newt species [89], and in Lissotriton helveticus, the precursor similarly facilitates mating by increasing female nudging of the male cloaca [84]. Phylogenetic comparison of SPF sequences revealed that the cleaved sodefrin decapeptide is unique to the Cynops genus, such that a second pheromone independently evolved from a previously co-opted pheromone gene [89]. Sodefrin signals through the female newt olfactory system, but the receptor(s) remain unidentified.

While adult newts are primarily terrestrial animals, their life cycle begins with an aquatic larval stage and they return to freshwater habitats to reproduce. Consequently, the release of water-soluble proteins and peptides into the aqueous environment is analogous to the release of volatile odorants into the air. In contrast, lungless salamanders of the family Plethodontidae are purely terrestrial, with females laying clutches of eggs on land and juveniles hatching already in their adult morph [47]. Plethodontid salamanders contain various skin glands that secrete pheromones which provide information on species [90, 91], body size [92], sex [90], female gravidity [92,93], diet [94,95], and parasite load [96]. The most well-characterized of these glands is a male-specific chin gland – the mental gland – that seasonally hypertrophies in response to elevated plasma androgens that coincide with the courtship season [97,98]. Previously termed the “hedonic gland” [99], the primary function of this gland seems to be the production of proteinaceous pheromones that modify female courtship behavior and reduce the length of courtship [24,100–103]. Once the gland fully develops, nearly all of its transcriptional and translational capacity is diverted to pheromone synthesis, such that >70% of the glandular mRNA codes for pheromone [104]. In contrast to the mammalian examples where pheromones are exclusively sensed by olfaction, female plethodontid salamanders receive pheromone by one of two delivery mechanisms. In the majority of plethodontid salamanders (∼300 species), mental gland development is accompanied by the development of hypertrophied premaxillary teeth which are used to “scratch” the dorsum of the female, and pheromones diffuse transdermally into the bloodstream. However, in a single clade of large eastern Plethodon salamanders (28 species), mental gland morphology has transitioned to a large pad-like structure that is “slapped” to the female nares and pheromones are delivered to the olfactory system [47].

To date, three major pheromone families have been transcriptionally, proteomically, and functionally characterized: SPF (a homolog of the newt pheromone), Plethodontid Receptivity Factor (PRF), and Plethodontid Modulating Factor (PMF). Found in both newts and plethodontid salamanders, the SPF family is likely >300 million years old and is possibly the oldest vertebrate pheromone family [89]. All three proteins are derived from separate gene families: SPF is related to phospholipase A2 inhibitors, PRF is related to the IL-6 helical cytokines, and PMF is a member of the three-finger protein (TFP) superfamily. RTPCR based analyses revealed that SPF and PMF are present in all plethodontid species, while PRF is a more recently derived family found only in eastern Plethodon species, including those with olfactory delivery (Fig. 2A). Determination of d_N/d_S ratios revealed strong signatures of positive selection and rapid evolution in all three pheromone families [105–108]. In addition to rapid evolution, the three gene families have been subjected to extensive gene duplication such that they are generally maintained as multi-isoform blends in the pheromone extract.

Fig. 2.

Summary of plethodontid pheromone evolution. (A) A reduced phylogenetic tree emphasizing species for which proteomic characterizations have been performed (P. shermani, P. cinereus, and D. ocoee). Approximate times in millions of years (MYA) for gain and loss of key morphological, behavioral, or genetic trait acquisitions are included, as well as representative reverse phase chromatograms for the pheromones of each species (using 70 minute gradients at 1% acetonitrile/min). While PMF is one of the oldest pheromone gene families, its expression varies dramatically between the three species, with D. ocoee expressing a single PMF isoform at extremely low abundance (<1% total pheromone, not visible by HPLC), P. cinereus expressing ∼4–6 isoforms, and P. shermani expressing >30 isoforms. (B) Alignment of PMF isoforms identified in the P. shermani pheromone extract by HPLC and MS-based proteomics (adapted from [109]). The bar graph shows consensus strength, the blue box highlights the signal peptide, and the few absolutely conserved residues (mostly the conserved eight cysteine core) are colored in black. (C) A putty model of the PMF 3D structure from [25], with backbone diameter representing residue heterogeneity (Shannon Weaver diversity index), and color signifying the mode of evolution (blue, purifying selection; yellow, neutral selection; orange, positive selection at >95% confidence; red, positive selection at >99% confidence).

More detailed proteomic analyses have been performed for three plethodontid species representing cases of transdermal delivery with no PRF (Desmognathes ocoee), transdermal delivery with PRF (Plethodon cinereus), and olfactory delivery (Plethodon shermani). Initial pheromone characterization was performed for P. shermani using multiple types of high performance liquid chromatography (HPLC) to isolate individual proteins, which were then identified by MS analysis using a shotgun EST library as a reference [109,110]. Subsequent studies with D. ocoee and P. cinereus instead used de novo transcriptomes as reference databases, with amplification of identified candidates by RT-PCR and validation by Sanger sequencing [111,112]. Despite detection of SPF mRNA by RT-PCR in all three species, it was only observed proteomically in D. ocoee, where it comprised ∼30% of the total pheromone. While D. ocoee may express up to ∼70 different SPF isoforms, six major classes of highly divergent isoforms shared <50% sequence identity. Protein abundance estimates by spectral counting suggest that isoform expression is relatively stable between individual male D. ocoee [112]. When applied to females during courtship, purified SPF decreased courtship time [102]. Proteomic analyses of the D. ocoee pheromone extract also led to the identification of several peptide hormone paralogs such as glucagon, insulin, leptin, and relaxin. It was hypothesized that when delivered to the female bloodstream, these hormone-like pheromones may directly influence female physiology by exploiting pre-existing endocrine receptors [112]. In contrast, both P. cinereus and P. shermani with transdermal and olfactory delivery, respectively, secrete pheromone mixtures of predominantly PRF and PMF, but with multiple notable differences. First, PRF was heavily glycosylated in P. cinereus and not P. shermani, with the loss of glycosylation resulting from mutation of two NXS/NXT N-glycosylation motifs and an alternative signal peptidase cleavage position which removes an O-glycan near the N-terminus. While glycosylation can affect various aspects of protein solubility, stability, and receptor specificity, its functional significance in this pheromone context remains to be determined [111]. Second, P. shermani expresses ∼5–10 times more PMF isoforms compared to P. cinereus (Fig. 2A). Specifically, individual P. shermani males express more than 30 different PMF isoforms. Most, if not all, of this diversity appears to be from gene duplication, and 99 different putative isoforms were identified through cDNA sequencing of a single population. HPLC-based analyses suggest that each male expresses a unique combination of isoforms [110], and behavioral studies show that female behavior changes upon exposure to different combination of isoforms [24]. Mass spectral characterization of P. shermani PMFs confirmed that at least 27 different PMFs are expressed, and are highly divergent with an average amino acid conservation of ∼30% (Fig. 2B) [109].

One possible hypothesis as to why olfactory delivery evolved as a replacement to transdermal delivery is signal amplification. Akin to normal endocrine cues, pheromones supplied to the bloodstream likely bind to their receptors in a concentration-dependent manner. Hence, male salamanders with transdermal delivery must deliver some critical dosage of pheromone to elicit a response. However, application of pheromone to the olfactory system offers the opportunity for signal amplification via neuronal activation. In P. shermani, V2Rs are highly expressed by VNE neurons [113], with different neurons separately detecting PRF and PMF [114,115]. Because V2Rs are tuned to recognize signals at nanomolar levels or lower [71], there is likely limited concentration dependence for olfactory pheromones, and less restriction on gene duplication, such that plethodontid males benefit from the increased PMF isoform diversity. More isoforms increases the likelihood of stimulating females with any possible receptor profile, and improves the chances of mating success. In a recent study, the 3D structure of PMF was determined by multidimensional NMR, and NMR relaxation analysis revealed that the residues under positive selection tended to be on the most flexible regions of the molecule (that result from a novel disul fide bonding pattern compared to all other three-finger proteins; Fig. 2C) [25]. Evolved flexibility may be yet another way to create pheromone diversity and stimulate additional receptors. However, as with the case of sodefrin, the receptors for plethodontid pheromones are unknown, and it remains difficult to explicitly test models of sexual selection.

4. Seminal fluid proteins

For animals that utilize internal fertilization, male ejaculate is a complex and highly variable mixture, containing cells (gametes), mucins, small metabolites (sugars and amino acids to nutrify spermatozoa), hormones, and various peptides and proteins — this last category simply being termed seminal fluid proteins (SFPs). The types and functions of SFPs vary tremendously between species, including structural proteins (mucins [116] and/or aggregate-forming proteins that result in copulatory plugs [116,117]), antimicrobial peptides [118], and proteins that directly affect female physiology and behavior by modulating female receptivity, longevity, egg production, feeding, and sperm storage potential [119]. Classical biochemical methods are generally not well-suited to studying SFP interactions. Since they are delivered as part of the male ejaculate only during mating, it is often difficult to purify compounds and artificially apply them within the appropriate biological context (with a few notable exceptions [120]). In contrast to the simpler case of pheromones being synthesized by single glands, seminal fluid is the cumulative product of several organ secretions (in humans, this includes the testes, prostate, seminal vesicle, and bulbourethral glands [121]), and disentangling the relative contributions from each of these tissues can be challenging. However, advances in gene expression analysis and proteomic methods have driven a recent explosion in our broader understanding of seminal fluid composition and its effects on female physiology.

The majority of SFP studies have utilized insect models (for thorough review, see [119]). A number of microarray, EST, and proteomic studies have been performed on honey bees [122], beetles [123], mosquitoes [124–126], butterflies [127,128], ticks [129], bed bugs [130], and field crickets [131,132];unsurprisingly as an established laboratory model, fruit flies (Drosophila spp.) have been the most extensively characterized [133]. Sperm competition is well-documented in Drosophila, and it was accurately predicted that SFPs would play a significant role in this process. A particular focus has been on accessory gland proteins (Acps), which constitute the majority of seminal fluid proteins [133]. Genome analysis, whole sperm proteomics, and tissue-specific microarray analysis from Drosophila melanogaster males allowed prediction of 112 Acps [134]. Two key Acps have been the subject of extensive biological and molecular characterization: sex peptide (SP, or Acp70Aa) and ovulin (Acp26Aa). Both proteins increase female egg production, but through different mechanisms: SP stimulates oocyte maturation and accumulation of yolk proteins [135,136], while ovulin stimulates ovulation by triggering octopamine signaling in the central nervous system (through an unknown receptor) [137]. SP also decreases female receptivity to re-mating, increases feeding behavior, and changes dietary preferences from carbohydrate-rich to protein-rich food. Energy acquisition and mating have been extensively linked in many systems [138,139], and this change in feeding behavior is likely a secondary response to increased yolk production [140]. SP-induced effects are observable less than 24 h after mating and can last up to two weeks. This long-term response results from SP binding to sperm and being released gradually over several days following proteolytic cleavage [141,142]. The SP receptor (SPR) was identified by screening D. melanogaster females generated as part of a genome-wide series of RNA-interference (RNAi) lines, with SPR knockdown females laying fewer eggs and re-mating at similar frequencies to virgin females. SPR is a GPCR expressed on neurons in both the female reproductive tract and central nervous system [143].

The evolution and interactions between SP and SPR are atypical compared to other studied reproductive proteins. First, there is minimal evidence that SP is rapidly evolving [144]. Second, in contrast to pheromone examples where there is likely one-to-one relationships between ligands and receptors, SPR is a promiscuous receptor both in expression profile and ligand specificity. Specifically, SPR is also expressed in the brains of larvae and adult males, and has similar affinity for both SP and myoinhibitory peptides (MIPs) [145,146]. A recent study showed that MIP-SPR interactions play a critical role in Drosophila sleep behavior [147]. Third, there is little evidence for co-evolution between the two genes. Evolutionary rate covariation (ERC) is a recently developed metric for correlating the normalized number of substitutions between protein-coding genes along a phylogenetic tree, such that continual co-evolution between proteins can be inferred from strong positive correlations [148]. When this metric was applied to annotated Drosophila reproductive genes, there was weak ERC between SP and SPR; however, 111 other genes had significant ERC with SP or one of five other SFPs that cooperate with SP in a network. Twenty-one candidates were selected from this list for RNAi knockdown, and five genes had a significant effect on female re-mating behavior [149]. SP and SPR seem to have evolved through sexual conflict, where males prioritize their own mating success over female fitness [16]. Because the binding constant between SP and SPR is a proxy for female preference, it is influenced by both the sequence and expression level of each protein. SPR sequence evolution has presumably experienced purifying selection to maintain its non-reproductive functions (e.g., MIP binding); conversely, SP has likely evolved to an optimum to maximize binding affinity. Without a changing SPR sequence, SP cannot “chase” SPR, leading to it also being under stabilizing selection (Fig. 1C). With respect to the gene expression and the Lande model (Fig. 1B), SP is likely under positive sexual selection (increases mating success) and negative natural selection (costly to produce, given that males have limited Acps that they strategically allocate during multiple matings [150]). Similarly, SPR is also probably under negative natural selection, as females suffer reduced longevity from increased mating and receiving excess Acps [151–153]. Sexual selection on SPR is difficult to infer, because of the challenges in ascertaining if reduced mating frequency in response to sex peptide is purely a byproduct of reduced fitness [154] (zero or weakly negative sexual selection), or also in part due to mate choice and increased female receptivity (positive sexual selection). Even in the absence of sexual selection, SPR expression could be maintained in the female reproductive tract through genetic coupling with SP (with regard to Fig. 1B, this implies V_p × — β_Np ≈ C_dp × (β_Sd — β_Nd)), leading to a stable sexual conflict scenario. Testing such hypotheses is likely difficult and would require estimating female mating receptivity, likely using carefully controlled mate choice experiments and/or sophisticated neurophysiological approaches.

While more than 100 putative Acps have been identified through male-specific tissue screens, identifying expression of these proteins and their actual transmission to the female reproductive tract remains challenging. In one study [22], the novel proteomics strategy of whole fly isotopic labeling was employed to selectively identify male versus female proteins. Flies were fed yeast grown on ¹⁵N labeled media, and after one full generation, no unlabeled peptides were detectable. Male flies fed unlabeled yeast were then mated with ¹⁵N labeled females, and soluble proteins extracted from the female reproductive tract for MS analysis performed with searches tuned to only identify unlabeled peptides. This design provided a zero background to detect only male-derived peptides, resulting in identification of 138 SFPs, only 75 of which were previously predicted based on EST sequencing. Highly conservative tests for molecular evolution showed that at least nine of the genes were rapidly evolving. Repeating the experiment on two additional Drosophila species (Drosophila simulans and Drosophila yakuba) revealed only 63 SFPs were commonly expressed in all three species, likely suggesting lineage-specific gene expression.

The same isotopic labeling strategy was adapted to the study of seminal fluid proteins from the house mouse (Mus domesticus) [155]. Again, females were isotopically labeled with a ¹⁵N-based diet to minimize background, and the female reproductive tract was investigated for unlabeled proteins post- mating. A total of 69 proteins were identified, 62 of which belong to the copulatory plug, and seven copulatory plug structural proteins constituted ∼37% of the total protein. A previous proteomic study documented 506 proteins secreted from six distinct regions of the male M. domesticus reproductive tract, and only a few of these genes contained an elevated d_N/d_S ratio relative to the genomewide average of ∼0.12 (when compared to the rat genome) [156]. In contrast, the 69 SFPs had a significantly higher mean d_N/d_S of 0.27 (with the non-SFP reproductive tract proteins having a mean d_N/d_S of ∼0.06). In addition to structural proteins, there was an enrichment for proteins related to protection from oxidative stress and protease inhibition. When the MS search parameters were adjusted to focus on proteins with ¹⁵N incorporation to identify female proteins expressed in response to mating, six proteins were identified including two proteases: kallikrein-related peptidase 14 and lactotransferrin. Interestingly, both proteins had elevated d_N/d_S ratios (0.32 and 0.74, respectively), and may be rapidly co-evolving with SFP protease inhibitors [155]. Notably, the same categories of genes are enriched among human SFPs [121], and further suggest that many protein families are common to the male ejaculates of most species, with variation in copy number, sequence, and expression level driven by female sexual selection [157].

5. Gamete recognition proteins

The birth of sexual selection arguably initiated with the differentiation of sperm and egg, and each gamete benefiting from a different reproductive strategy [16]. Consequently, the interaction of egg and sperm are one of the most essential biological processes, and has been of major research interest, due in part to the likely role such interactions play in human infertility [14,158]. The process of fertilization can be divided into various stages, and while the particular molecules vary between organisms, the core process is highly conserved. Oocytes are surrounded by glycoproteinaceous egg coat that serves as an initial barrier to sperm, termed the zona pellucida (ZP) in mammals and the vitel line envelope (VE) in other animals. Sperm are attracted to oocytes by chemotaxis, and upon reaching the ZP/VE, form an acrosomal process: an extension of the sperm head that is coated in proteins secreted from an acrosome granule. The acrosomal proteins create a hole in the ZP/VE, permitting sperm passage to the oocyte plasma membrane, where a second set of molecular interactions fuse the two membranes and result in zygote formation.

The most well-characterized fertilization systems are those of marine invertebrates, including sea urchins [159,160], abalone [161], sea snails [162], oysters [163], and sea stars [164]. Early biochemical studies were facilitated by the ample availability of gametes from such external fertilizers to prepare various biochemical fractions. In particular, sea urchins have been a model of fertilization for over a century [165], and the first fertilization protein was identified in sea urchins nearly forty years ago [160]. Early ultrastructural studies of sea urchin sperm suggested that the contents of the acrosomal process were associating with the oocyte membrane [166], and biochemical analysis revealed that the acrosomal granule contained a single protein, termed bindin [160]. Agglutination experiments using bindin from different species of sea urchin revealed greater efficacy when bindin was paired with conspecific oocytes, suggesting species-specificity [159]. It was approximately a decade later before bindin was first cloned from the species Strongylocentrotus purpuratus [167], and has since been documented in several orders of sea urchin [165]. The size of sea urchin bindin varies between 193–418 residues, but includes a highly conserved ∼60 residue domain with 29 invariant residues [168]. Part of this domain is an 18 residue peptide (B18) that promotes fusion of the gamete plasma membranes [169]. Much of the variation in bindin length stems from repeats of short glycine-rich motifs that may be either N- or C-terminal to the conserved core [165].

Models of molecular evolution for bindin were variable between genera, such that species from four of the six genera tested showed signatures of positive selection (d_N/d_S > 1). An interesting quality about the genera with rapidly evolving bindin is that they include sympatric species with overlapping habitats, while the two non-rapidly evolving genera contain only allopatric species [170]. These data are in support of the reinforcement hypothesis: overlapping habitat creates risk of species hybridization, and if hybrids are not viable and/or less fit, selection should favor mechanisms to reinforce species boundaries. However, some allopatric species still display signatures of rapid bindin evolution such that reinforcement alone is not sufficient explanation. Another attractive hypothesis is one of sexual conflict and polyspermy. Following fertilization, the sea urchin VE depolarizes and becomes impermeable to additional sperm; however, before this depolarization, if eggs are penetrated by multiple sperm they will become nonviable. Hence, natural selection may counter sexual selection by favoring oocytes whose VEs have lower affinity for bindin [171]. Population genetics and field experiments with wild sea urchin populations support this hypothesis, as higher fertilization rates have been observed for uncommon bindin genotypes under high population densities where polyspermy risk is increased [172]. It was not until ∼25 years after the discovery of bindin that its receptor EBR1 was cloned from two Strongylocentrotus spp. by using subtractive cDNA hybridization. EBR1 is a large ∼350 kDa protein that includes an N-terminal ADAMTS-like domain and ∼19 species-specific repeats [173]. Similar to bindin, EBR1 evolved under positive selection in Stronglyocentrotus spp., although fewer residues were under selection compared to bindin [174]. Only EBR1 sequences from Stronglyocentrotus are deposited in GenBank. The large transcript size and presumably high sequence variability have made amplifying full-length EBR1 difficult; however, the sea star EBR1 was recently characterized using a de novo transcriptome and contained a similar architecture to the sea urchin homolog [164]. More sea urchin genera will need to be sampled to assess how reinforcement versus sexual conflict may contribute to the co-evolutionary chase between bindin and EBR1.

In addition to sea urchin and related echinoderms, the marine gastropod abalone has been another valuable model for characterizing interacting gamete proteins. Early studies with abalone acrosome extract revealed its ability to “unravel” purified egg VEs, creating an ∼3 μm hole for sperm passage. This activity was attributed to a single component: the 16 kDa protein lysin [175]. Similar to bindin, lysin displayed species-specificity in that its more effective at dissolving con-specific VEs compared to those of other species [21]. Estimating d_N/d_S ratios for lysin revealed strong positive selection [161], and it has since become a textbook example of rapid evolution [176]. Crystal structures were determined for lysin from two species (red abalone, Haliotis rufescens, and green abalone, Haliotis fulgens), and the two proteins have nearly identical 3D structures, with backbone atoms varying on average by ∼0.28 Å (Fig. 3A) [177]. Modification of crystallization conditions and accompanying FRET-based experiments revealed that lysin predominantly exists as a dimer [178]. Lysin monomers consist of a disjointed triple helix whose surface is polarized and has two faces of concentrated with either basic or hydrophobic residues, with the hydrophobic face defining the dimerization interface (Fig. 3B). Notably, the majority of residues under positive selection were concentrated on the N- and C termini, which form a nexus on one side of the monomer (Fig. 3C). These regions were also seemingly flexible and not well-resolved in the crystal structure. Chimeric recombinant lysins were prepared by substituting subsequences of lysin from red and pink abalone (Haliotis corrugata), and exchanging the N- and C-termini from each species was sufficient to invert the species specificity of VE dissolution [179]. In contrast to bindin, which mediates both VE dissolution and membrane fusion, lysin seems to only interact with the VE. A second molecule, sperm protein 18 kDa (Sp18), likely facilitates membrane fusion [180]. Similar to lysin, Sp18 has also been subjected to pervasive positive selection and exhibits species-specific activity [181]. However, while there is little discernable sequence similarity between lysin and Sp18, the crystal structure of Sp18 revealed a highly similar alpha helical topology [182], and the two genes share nearly identical intron–exon boundaries [181]. Hence, Sp18 and lysin are likely paralogous genes, duplicating from a common ancestral protein that mediated both VE dissolution and membrane fusion similar to bindin. Upon gene duplication of this ancestor, the new genes that now represent lysin and Sp18 were free to specialize in order to more efficiently facilitate fertilization.

Fig. 3.

Structural models of abalone sperm lysin. (A) Backbone cartoon overlay of lysin crystal structures from two different species: red and green abalone (with ribbon color matching the species common name, and the N- and C-termini noted). The average backbone rmsd between the two models is ∼0.28Å.(B) Surface model of the red abalone lysin dimer, colored based on residue type (blue, basic; red, acidic; green, hydrophobic; white, polar), with an inlay of the backbone structure for reference. The majority of hydrophobic residues cluster on or near the dimer interface, with the two exposed monomeric faces containing long tracts of basic residues. (C) The same surface model overlaid with the estimated mode of selection (blue, purifying; yellow, neutral; red, positive; from [211]). The residues under positive selection centralize on the edges of the basic tract and the nexus at the N- and C-termini.

Analogous to EBR1, the vitelline envelope receptor for lysin (VERL) is a giant, highly repetitive glycoprotein of ∼1000 kDa that was identified through lysin affinity purification [21]. Combined sequencing of RACE cDNA and a genomic cosmid library revealed that the VERL gene codes for a 3722 residue protein (∼410 kDa), ∼90% of which results from 22 repeats of ∼153 residues [183]. This repetitive domain is homologous to the vertebrate ZP-N polymerization domain (which will be described in further detail below) [184]. As part of a long fibrous molecule, these repetitive polymerization domains are thought to form an interlocking network of hydrogen bonds that stabilize the VE supramolecular structure. Binding curves between lysin and VERL are consistent with positive cooperativity, and suggest a stoichiometry of ∼2–3 lysin molecules per VERL repeat. Lysin presumably dissolves the VE by competing for hydrogen bonds between VERL repeats, resulting in the splaying of the fibers that allows sperm passage [21]. The majority of repeats (3–22, ordered N-terminus to C-terminus) are ∼99% identical within a given species, and are hypothesized to be homogenized through unequal crossing over and concerted evolution [185]. In contrast, repeats 1 and 2 have evolved independently under positive selection [186], and the d_N/d_S ratios between lysin and these first two VERL repeats are correlated along branches of the phylogenetic tree [187]. The cumulative result of these findings is that lysin and VERL evolve by chase dynamics: the first two repeats of VERL move away from lysin's optimum, and lysin quickly adapts to maintain high affinity interactions. These N-terminal repeats are likely on the outermost surface of the VE, making them the first repeats lysin encounters upon secretion from the acrosome. After binding repeats 1 and 2, affinity for the homogenized repeats would increase by cooperativity. However, in contrast to the pheromone system where olfaction provides signal amplification, non-enzymatic VE dissociation requires that lysin have an extremely low K_A for VERL and/or be in large stoichiometric excess. Sexual selection seems to have favored both, as lysin binds VERL at nanomolar concentrations [21], and individual sperm produce tremendous quantities of lysin (up to one gram of lysin can be purified from a single male abalone; W.J. Swanson, personal observation).

In addition to the trio of lysin, VERL, and Sp18, recent proteomic studies have suggested greater complexity in both the male and female sides of abalone fertilization. Joint transcriptome sequencing and MS proteomics of abalone sperm revealed two isoforms of a small acrosomal protein (termed sperm protein 6 kDa, or Sp6) which together are more abundant than either lysin or Sp18. The two isoforms of Sp6 are highly similar, differing primarily in a polyaspartate region near the N-terminus of four or eight residues, such that the two isoforms are referred to as Sp6_4D and Sp6_8D. Comparison of Sp6 sequences between abalone species was again indicative of rapid evolution by positive selection [23]. The function of Sp6 remains to be determined, but given its high net negative charge in contrast to the positive charges of lysin and Sp18, one hypothesis is that it stabilizes one or both of the molecules (with positive selection as a consequence of lysin/Sp18 rapid co-evolution with female receptors). Sp6 may also interact with additional VE proteins. While VERL was the first VE protein characterized through lysin affinity chromatography, deep cDNA sequencing and shotgun proteomics permitted identification of more than a dozen additional ZP domain-containing proteins in the abalone VE (Fig. 4A) [188,189]. The ZP domainis an ∼250 residue motif common to all vertebrate egg coat proteins, consisting of two disulfide bonded halves which independently adopt modified immunoglobulin-like folds. The separate halves of this domain are termed ZP-N and ZP-C, respective to their orientation in the protein. While polymerization between ZP domains involves intermolecular contacts between both ZP-N and ZP-C [190], biochemical studies have demonstrated ZP-N alone is sufficient for polymerization [191,192]. Most vertebrates have three ZP proteins – simply numbered ZP1, ZP2, and ZP3. The ZP primarily consists of ZP2 and ZP3 in approximately equal quantities, with both proteins under positive selection [193]. In abalone, 32 genes expressed in the ovary were identified that contained full ZP domains, with expression of 29 validated by tandem MS. One of these proteins was VERL, such that its structure includes 22 separate ZP-N repeats followed by a complete ZP domain (Fig. 4C). Approximate protein quantification using the normalized spectral abundance factor (NSAF) suggested that VERL is a relatively minor component (∼1% by stoichiometry; Fig. 4B) [188,189]; however, it is the only domain with multiple tandem ZP-N repeats, and may be acting as a scaffolding protein that interlocks with the other ZP-containing proteins. This crosslinking function would also explain why it is the target of lysin and is key for disrupting VE supramolecular structure. Interestingly, vertebrate ZP2 has a similar architecture of serial ZP-N repeats before a complete ZP domain(Fig. 4E) and may be providing an analogous scaffolding function. Phylogenetic analysis of the abalone ZP proteins identified one component, VEZP14, as a more recent paralog of VERL, although without the multiple tandem ZP-N repeats [189]. Co-evolutionary rate analysis revealed the surprising pattern of inverted rates of evolution between VERL and VEZP14: when lysin and VERL were most rapidly evolving, VEZP14 was evolving slowly, and vice versa (Fig. 4D). Lysin binds VEZP14 with similar affinity to VERL, such that VEZP14 may have evolved as a competitive inhibitor to slow and/or restrict lysin-mediated dissolution of the VE in certain contexts (analogous to strategies which attenuate polyspermy risk in sea urchins) [194]. The role of molecular decoys is well-established in immunology and pathogen response (where rapid evolution is also common) [195], but VEZP14 is the first instance that this has been proposed for a sexually selected system. This potentially suggests that there may be a small subset of “tools” that have evolved multiple times for systems under arms race dynamics, regardless of the source of selection.

Fig. 4.

Evolution and proteomics of ZP-domain containing proteins in abalone. (A) Phylogenetic tree of the 32 ZP-domain containing proteins from red abalone (H. rufescens), with three Tegula ZP domains included as outgroups. Notably, VEZP14 and VERL are well-supported on a separate branch from Tegula VERL, suggesting that they result from a more recent gene duplication. (B) Estimate of protein expression for ZP domain-containing proteins based on normalized spectral abundance factor (NSAF), determined by multidimensional protein identification (MUDPit). Both red and green abalone have relatively similar expression levels of homologous genes, with VERL representing ∼1% expression (adapted from [189]). (C) The gene architecture of VERL and VEZP14. (D) Co-evolutionary rates of VERL and VEZP14 relative to lysin, with the positive correlation between lysin and VERL suggesting that as VERL rapidly evolves, so does lysin, with the inverse relationship for lysin and VEZP14 (adapted from [194]). (E) Gene architecture of human ZP proteins, adapted from [193]. It is noteworthy that ZP2 contains a shorter, but similar, tandem array of ZP-N domains like VERL.

While biomedical implications have led to great interest in characterizing mammalian and human fertilization, our understanding of the molecular components and their interactions remains limited. As previously mentioned, ZP2 and ZP3 are the major constituents of the mammalian ZP, comprising ∼80% of the total protein [193]. Humans and other primates contain a fourth ZP protein (ZP4) that is a recent gene duplication of ZP1 [196]. Both ZP2 and ZP3 are within the top 10% most rapidly evolving genes in the human genome [197], and similar trends were found for field mice and birds [198,199]. Mouse oocytes with gene knockouts for ZP2 or ZP3 fail to form a proper ZP, with knockouts of ZP1 having a loose, fibrous ZP (suggesting a possible crosslinking role for ZP1, likely through its extra ZP-N motif) [200]. It was initially postulated that the sperm binding motif was exclusively on ZP3 [201], but the current consensus is that the interactions are more complex and involve carbohydrate recognition, the ZP supramolecular structure, or both [190,193,202,203]. Additional gene deletion studies have identified sperm factors important for ZP adhesion, but no clear molecular mechanism has been proposed [204]. However, significant advancements were recently made towards understanding mammalian membrane fusion. Using a high-throughput ELISA-like screen for membrane protein/ligand interactions [205], folate receptor 4 (FolR4, now renamed “Juno”) was characterized as the egg cell surface receptor for sperm Izumo1 and determined to be critical for membrane fusion. Immunocytochemistry experiments also demonstrated that upon sperm binding, Juno is shed from the oocyte membrane through vesicle formation and exocytosis, acting as a block to polyspermy. This pathway seems to be conserved in all mammals [206], but the co-evolutionary dynamics of these proteins remain uncharacterized.

6. Conclusions and future considerations

The rapid co-evolution of male proteins and female receptors through arms race-like dynamics is a phenomenon that permeates all levels of complex animal reproduction, from pre-copulatory events until a sperm fuses with an egg, and is not exclusive to any specific lineage. The case has been presented here that studies of reproductive systems are strongly enhanced by jointly characterizing the molecular underpinnings through classical biochemistry and the likely targets of selection through models of molecular evolution. The availability of high-throughput next-generation sequencing and mass spectral proteomics allow simultaneous identification of both. In the plethodontid salamander system, proteomics permitted identification of the subset of translated PMF isoforms behind the dozens of gene duplication events [109]. In Drosophila, whole animal isotopic labeling and shotgun proteomics dramatically expanded the range of identified seminal fluid proteins, for which evolutionary rates could be rapidly calculated from whole genome sequence data [22]. In abalone, joint transcriptomics and proteomics together permitted identification of a highly abundant, rapidly evolving protein that – because of atypical biochemical properties – had gone unnoticed despite more than 20 years of research [23]. As argued in the introduction, the application of these methods will continue to expand the number of organisms for which reproductive proteins are characterized, and allow more broad range phylogenetic comparisons. At the present, these high-throughput technologies are most well-suited for candidate identification, and still do not have the precision to fully replace biochemical structure/function-driven experiments. Structures determined by NMR and X-ray crystallography are limited for reproductive protein, and nearly all examples were discussed here. The NMR-derived model of ESP1 and homology model of V2Rp5 are the current best example of any interacting pair of reproductive proteins, but these data do not directly address the challenge of how particular mutations in a receptor might drive adaptive co-evolution in a male ornament. Sequencing based methods also provide limited information on post-translational modifications, such as glycosylation, which are functionally critical in many reproductive proteins [207]. While time, cost, and labor may preclude future structural studies for many reproductive proteins, improved in silico docking protocols, sequencing of multiple homologs for residue covariance between interacting partners [208], and advancements in NGS-based high-throughput mutational screens [209] will permit more rapid cross-validation of molecular evolutionary models. Such studies should drive the testing of many of the available hypotheses regarding sexual selection, providing a unification between theory and empiricism that currently does not exist.