Comparative functional analyses of the prostate-specific KLK3 enzyme in primates reveal the impact of sexual selection

Abstract

Male reproductive proteins frequently evolve rapidly in animals, potentially due to adaptive evolution driven by sperm competition, polyspermy avoidance, or pathogen defense. Alternatively, elevated rates of protein change may be due to relaxed constraint. The prostate-specific protease KLK3 has experienced dynamic evolution since its origin stemming from a gene duplication in the ancestor of all Old World primates, with instances of rapid evolution, stasis, and pseudogenization. As we demonstrate with functional assays using recombinant proteins, these changes have resulted in a chimpanzee KLK3 ortholog with greater enzyme velocity and higher efficiency than other apes, including humans. Reduced enzyme efficiency was observed in gorillas and gibbons who both possess a chimeric KLK2/KLK3 enzyme resulting from independent genomic deletions. The relative efficiency of KLK3 homologs among these species correlates well with their presumed levels of sperm competition. Furthermore, the reconstructed protein of the human-chimpanzee last common ancestor has enzyme kinetics identical to modern humans, suggesting that the observed functional differences between humans and chimpanzees are derived in the latter and allowing us to tentatively speculate that their common ancestor did not possess a polygynandrous mating system similar to modern chimpanzees.

Introduction

Sexual selection is an important driver of phenotypic adaptation (Darwin, 1871). The mechanism of sexual selection (e.g., female choice or sperm competition) will vary depending on the social grouping and mating behaviors of each species. Primates, especially humans and the other apes (hominoids), present an interesting group because closely related species have very different mating systems and associated adaptations driven by sexual selection. The gibbons (family Hylobatidae), or lesser apes, are predominantly socially monogamous, living in family groups of one adult male and one adult female along with their dependent juvenile offspring. As with most socially monogamous animals, neither group structure nor mating in gibbons is completely restricted to single male-female pairs (Fuentes, 2000; Li et al., 2022). Among the great apes, gorillas (Gorilla gorilla and G. beringei) typically live in social groups with a single adult male and multiple adult females; in this polygynous system, each female only mates with the single silverback male (Inoue et al., 2013). The closely related chimpanzees (Pan troglodytes), in contrast, live in large social groups with many adult males and females. Females in this polygynandrous species routinely mate with multiple males in the days surrounding ovulation (Watts, 2007). The social and sexual behavior of humans varies within and among cultures, but we are generally considered polygynous or monogamous, including serial monogamy (Dixson, 2009). The labile nature of sociosexual behavior in the apes has resulted in changing sexual selection pressures and differing adaptations among species. The overt male-male competition associated with single-male polygyny has produced gorilla males that are over twice as massive as females, whereas the monogamous gibbons show essentially no sexual dimorphism in either body size or canine size (Dixson, 2012). The intense sperm competition induced by sexually promiscuous females has resulted in chimpanzee males possessing among the largest testes relative to body size of any primate (Møller, 1988). Chimpanzees show a number of other adaptations of the male reproductive system putatively driven by sperm competition, including large seminal vesicles, the production of a copulatory plug, high sperm counts, large sperm midpiece volumes, and sperm that swim with greater force than other apes (Anderson & Dixson, 2002; Dixson, 1998; Dixson & Anderson, 2002; Nascimento et al., 2008).

Male reproductive protein-coding genes are among the most rapidly evolving group of genes, in nearly every animal taxon studied (Haerty et al., 2007; Swanson & Vacquier, 2002; Torgerson et al., 2002). In primates and other mammals, this accelerated rate of amino acid substitutions is often ascribed to sexual selection, and sperm competition in particular (Clark & Swanson, 2005; Dorus et al., 2004; Ramm et al., 2008). However, primate male reproductive proteins do not universally evolve faster than other proteins, and not necessarily in species with higher levels of sperm competition, suggesting that other factors such as pathogen defense or relaxed selection may be driving rates of protein evolution (Bowman et al., 2024; Carnahan-Craig & Jensen-Seaman, 2014; Dapper & Wade, 2020; Ports & Jensen-Seaman, 2024; Wong, 2011). Along with the general pattern of elevated rates of protein evolution, a number of male reproductive genes have lost function in primates, including those that are otherwise highly expressed in other species or essential for fertility in mouse models. Among these is the semenogelin 1 gene (SEMG1), which has lost function in gorillas and gibbons (Isshiki & Ishida, 2019; Jensen-Seaman & Li, 2003); the semenogelin 2 gene (SEMG2), which has become a pseudogene in chimpanzees and gorillas (Jensen-Seaman & Li, 2003; Kingan et al., 2003); a prostate-specific transglutaminase (TGM4), which has lost function in gorillas (Carnahan & Jensen-Seaman, 2008; Clark & Swanson, 2005); and the prostate-specific proteases (KLK2 and KLK3) having been deleted or lost function several times throughout primate evolution (Marques et al., 2012). Numerous genes that code for male proteins involved in sperm-egg recognition have been pseudogenized in humans and other primates, including ADAM1A, ADAM3, ZAN, ZPAX, and ZPR3 (Bickell & Hart, 2022; Carlisle & Swanson, 2021; Carlisle et al., 2024; Meslin et al., 2012).

The Kallikrein-related peptidase 3 (KLK3) is a serine protease secreted by the epithelial cells of the prostate gland. KLK3 is also known as prostate-specific antigen (PSA) as it is routinely used for both prostate cancer diagnosis and monitoring the progress of the disease (Lilja et al., 2008). KLK3 belongs to the tissue kallikrein gene family, which represents the largest locus of serine proteases in the human genome, comprising 15 paralogs on chromosome 19q13, each consisting of five exons (Lundwall & Brattsand, 2008; Yousef & Diamandis, 2001). In humans, KLK3 plays an essential role in semen liquefaction by degrading the gel-forming semenogelin proteins in seminal plasma after ejaculation (Lilja, 1985). The site-specific cleavage of semenogelins, primarily mediated by KLK3, releases motile spermatozoa entrapped within the seminal coagulum, thereby facilitating fertilization (Lundwall & Brattsand, 2008; Robert et al., 1997). The synthesis of KLK3 occurs as a pre-proprotein possessing a 17-amino acid signal peptide, responsible for its targeting to the endoplasmic reticulum for secretion (Lundwall & Lilja, 1987). Subsequent cleavage of the signal peptide generates an inactive zymogen, bearing a 7-amino acid activation sequence in the extracellular environment (Lundwall & Lilja, 1987). In humans, KLK2, which has trypsin-like activity, is mainly responsible for cleaving the activation sequence of the zymogen of KLK3, generating the mature and enzymatically active form (Lövgren et al., 1997; Takayama et al., 1997). Additionally, KLK4, KLK5, KLK14, and KLK15 present in seminal plasma may also activate KLK3 (Emami & Diamandis, 2008; Michael et al., 2006; Takayama et al., 2001a, 2001b). The mature KLK3 is a 237-amino acid single-chain glycoprotein, exhibiting chymotrypsin-like activity (Lilja, 1985; Malm et al., 2000; Robert et al., 1997). KLK3 contains a catalytic triad composed of His57, Asp102, and Ser195 (chymotrypsin numbering), which is crucial for proteolytic activity (Carter & Wells, 1988) and highly conserved in other KLK3 orthologs. Upon activation, mature KLK3 effectively digests the SEMG1 and SEMG2 proteins, primarily at Tyr, Phe, and Leu residues, leading to the formation of smaller peptide fragments (Malm et al., 2000; Robert et al., 1997). KLK3 is inhibited in vivo by zinc and endogenous protease inhibitors.

KLK2 and KLK3 originated from a recent duplication event in the ancestor of the catarrhine primates (humans, apes, and Old World monkeys), after their split from the platyrrhines (New World monkeys), approximately 42 million years ago (Marques et al., 2012; Olsson et al., 2004; Pavlopoulou et al., 2010). While sequence analyses have identified specific codons under positive selection in the KLK2 and KLK3 genes, their dN/dS values remain relatively low compared to other seminal proteins, suggesting overall conservation in primates (Marques et al., 2012; Mubiru et al., 2014), but with significant variation among KLK3 branches, and a gene-wide dN/dS value of 0.38 in a human-chimpanzee pairwise alignment (Clark & Swanson, 2005). Conversely, the C-terminal region of KLK3 has been reported to be evolving under positive selection, with dN/dS values greater than two (Mubiru et al., 2014). Moreover, gene deletions, disruptions, or pseudogenizations resulting in the loss of KLK2 function have been observed in multiple primate lineages (Marques et al., 2012). A noteworthy finding is that a gene fusion event occurred between KLK2 and KLK3 via a genomic deletion that removed portions of both adjacent genes forming a single chimeric KLK2/3 gene (cKLK) independently in gorillas and gibbons (Marques et al., 2012). The first four exons of cKLK closely resemble KLK3, while the last exon shows greater similarity to KLK2. Considering the roles of KLK2 and KLK3 in seminal coagulum dissolution, the fusion of KLK2-KLK3 genes in gorillas and gibbons may be associated with reduced sperm competition in these species. Supporting this idea, Marques et al. (2012) found a positive correlation between the number of functional KLK2 and KLK3 genes and the number of semenogelin repeat units, proposing a model of coevolution of KLK2 and KLK3 with their substrates SEMG1 and SEMG2. Additionally, they reported a correlation between the number of functional KLKs and indicators of strong sperm competition associated with polyandrous mating, such as increased semen coagulation and larger relative testes size (Marques et al., 2012).

However, these studies suggesting sexual selection as a major driving force behind the evolution of KLK3 are based solely on sequence data, leaving the impact of amino acid substitutions on KLK3 enzymatic functions in different primates unexplored. Moreover, the functionality of chimeric KLKs in gorillas and gibbons remains untested, despite predictions of their resemblance to functional KLK3-like genes (Marques et al., 2012). To address these gaps and explore the potential influence of sexual selection on KLK3 evolution, we produced recombinant KLK3 proteins of several primate species and quantified their enzymatic activity, revealing the possible effects of positive selection in polygynandrous chimpanzees and relaxed constraint in gorillas and gibbons who both experience low levels of sperm competition. These results shed light on the molecular evolution of KLK3 in the context of sexual selection through sperm competition and provide valuable insights into the impact of reproductive strategies and sexual selection on other seminal proteins in primates.

Methods

Sequence and structural analysis

Amino acid sequences from the six KLK3 homologs were generated from their corresponding cDNA sequences. The species used with GenBank accession numbers or chromosomal locations are as follows: human: Homo sapiens (NM_001648.2), chimpanzee: Pan troglodytes (panTro5, chr19:53082837-53087983), gorilla: Gorilla gorilla (XM_019016144.2), gibbon: Nomascus leucogenys (nomLeu3, chr10:71144641-71149874), and macaque: Macaca mulatta (NM_001042776.1). The coding sequence of the human-chimpanzee ancestor KLK3 was predicted using the maximum likelihood method implemented in the PAML package, v4.9 (Yang, 2007) and converted to the amino acid sequence for alignment purposes. Multiple sequence alignments were created with Clustal-Omega (McWilliam et al., 2013).

KLK3 protein structures for human, chimpanzee, the human-chimpanzee ancestor, gorilla, gibbon, and macaque were predicted using AlphaFold Colab (Jumper et al., 2021). Pairwise structure alignments were then performed and visualized using PyMOL v2.5. Root mean square deviation scores were calculated, which serve as a metric to assess the level of structural similarity between two proteins, representing the average distance between the α-carbon (Cα) atoms in one structure and their corresponding Cα atoms in the other structure after alignment.

Phylogenetic analyses

Phylogenetic analyses and tests of selection were performed using homologous sequences obtained from annotated genomic data of various primates (Supplementary Table S1). FASTA multiple alignments for the coding sequences of KLK3 were obtained from the Multiz30 alignment accessible in the UCSC Genome Browser. The first four exons, comprising 630 nucleotides (210 codons), from a dataset encompassing 13 primates were used in analyses, with the unrooted consensus species tree (Kuderna et al., 2023). The fifth exon was excluded from most analyses because in gorillas and gibbons it is derived from KLK2, and as such would be inappropriate to include in phylogenetic analyses, which assume vertical transmission of genes. The codeml program in PAML was used to estimate the synonymous and nonsynonymous substitution rates (dS and dN) (Yang, 2007). To assess the direction and magnitude of selection acting on the KLK3 gene, dN/dS (ω) ratios were calculated under various codon-substitution models, with models compared using likelihood ratio tests. For branch models, species were categorized into those with low or high sperm competition, using relative testes size as a proxy. For this, branches leading to species with a positive residual from the correlation between body size and testes size (Kenagy & Tombulak, 1986) were designated as foreground branches and those with negative residuals as background branches, with the data taken from Supplementary File 2 of Ports & Jensen-Seaman, (2024).

Generation of cDNA clones

KLK3 is a secreted pro-protein, containing a 17-amino acid N-terminal signal peptide followed by a 7-amino acid activation sequence. To prevent any effect of differences in the maturation process of pre-proKLK3 among homologs when expressed in human cells, all KLK3 clones were designed to have the same signal peptide and activation sequences derived from the human sequence. cDNA sequences of the human (HumKLK3), chimpanzee (ChpKLK3), gorilla (GorKLK3), gibbon (GibKLK3), and macaque (MacKLK3) KLK3 genes were cloned into the pCMV-5 mammalian expression vector (Sigma-Aldrich) incorporating a C-terminal hexahistidine tag. This was accomplished through commercially synthesized DNA cloned into pUC19 plasmids (GenScript), followed by subcloning into the pCMV-5 vector. The human-chimpanzee ancestral sequence (Anc_HCKLK3) was created through site-directed mutagenesis (QuickChange, Agilent Technologies), using the human KLK3 construct as a template. Sequences of the inserts of all six KLK3 clones were confirmed by in-house Sanger sequencing using BigDye v3.1 (Life Technologies), and the complete plasmid sequence through outsourced Oxford Nanopure Technology (Plasmidsaurus). For the complete cDNA sequences of KLK3 used in expression constructs, refer to Supplementary Figure S1. For complete plasmid sequences, see Dryad depository (https://doi.org/10.5061/dryad.931zcrjv6).

Expression constructs were also made containing the coding sequence corresponding to repeats IIIa through Ia of human and chimpanzee SEMG1 (249 amino acids; 28.14 kDa) in the pMAL2c E. coli expression vector to be produced as a fusion protein with the bacterial maltose binding protein (MBP; New England BioLabs, Ipswich, MA). Please refer to (Lilja & Lundwall, 1992) for the nomenclature of semenogelin repeats. To produce these constructs, the region was amplified with PCR using as template either human genomic DNA or chimpanzee genomic BAC clone CH251-562O11 (BACPAC Genomics, Redmond, WA). Products were purified, digested, and ligated into the pMAL2c vector using standard protocols. Constructs were transformed into NEBExpress chemically competent cells (New England BioLabs) and the inserts confirmed with Sanger sequencing using BigDye v3.1 (Life Technologies). See Supplementary Figure S1 for construct and primer sequences.

Protein expression and purification

Adherent HEK-293T cells (ATCC) were maintained in DMEM supplemented with 10% FBS, 100 U/ml penicillin, and 100 µg/ml streptomycin in a humidified 37 °C, 5% CO₂ incubator under standard conditions. Cells were transfected with the above pCMV-5 expression vectors along with an empty vector as a negative control. Transfections were performed using a cell density of 200,000 cells/ml and 1 μg DNA/mL with 3 μL/ml of FuGENE 6 transfection reagent (Promega). The culture media was changed after 24 hr. The media was collected 96 hr post-transfection and centrifuged at 5,000 rpm for 15 min at 4 °C to remove cells. Recombinant proteins were purified with affinity chromatography using Proteindex Ni-Penta Agarose resin (Marvelgent Biosciences), following manufacturer’s instructions for batch purification, with 1 mL of equilibrated resins for every 35 ml of culture supernatant. To remove the imidazole from the eluted protein, the first two elutions were combined and dialyzed into activation buffer (recipe below) using 10K MWCO Slide-A-Lyzer MINI Dialysis Devices (ThermoFisher). Dialysis was performed in two steps: a three-hour dialysis, followed by an overnight dialysis with fresh buffer at 4 °C. Purification was assessed with SDS-PAGE gel, followed by Coomassie or silver staining. The purified and dialyzed proteins were then aliquoted and stored at −20 °C.

Fusion proteins of the bacterial MBP and segments of human and chimpanzee semenogelin 1 proteins, corresponding to repeats IIIa through Ia (SEMG1 MBP-IIIaIa), were expressed by growing 50 ml cultures of E. coli harboring these constructs in rich media supplemented with ampicillin at 37 °C and 200 rpm. Once the culture reached an OD₆₀₀ of 0.5, protein expression was induced by adding IPTG to a final concentration of 0.75 M, with continued growth for two more hours. Subsequently, cells were pelleted at 4,000 × g for 10 min, the supernatant was discarded, and the pellet was resuspended in column buffer (20 mM Tris-HCl, 200 mM NaCl, 1 mM EDTA). Cells were lysed chemically, and the fusion protein was purified using affinity chromatography with amylose resin. Purity and yield were assessed with SDS-PAGE and Coomassie staining.

For western blot analysis, proteins were transferred to PVDF membranes, using wet transfer. Processing and detection followed standard protocols, using primary antibodies (THE His-tag mouse monoclonal antibody at 1:2,000, GenScript; or a rabbit polyclonal anti-KLK3 at 1:2,000, Abcam catalog no. ab53774), and infrared dye-conjugated secondary antibodies (1:20,000, Li-Cor). Blots were imaged on a Li-Cor Odyssey XF imager.

Quantification of the purified and dialyzed recombinant HumKLK3 protein was performed using an ELISA (Human PSA ELISA kit, Abcam), with the standards run in duplicate and the test samples in triplicate with two dilutions. Since the antibodies used in the ELISA may not bind other species’ proteins with equal affinity, quantification of these recombinant proteins was achieved through comparison to HumKLK3. For this, semiquantitative western blots were used, with the samples run in quadruplicate with a His₆-tagged loading control and detected with the anti-His₆ mouse primary antibody described above. Protein band intensities were measured using the Li-Cor Image Studio software, and each homologous purified protein was diluted to equal concentrations.

The PNGase F Glycan Cleavage Kit (Gibco) was utilized to investigate if recombinant proteins from six species were N-glycosylated. PNGase F functions as an amidase, breaking the bonds between the innermost GlcNAc and asparagine residues in high mannose, hybrid, and complex oligosaccharides. Reactions were incubated at 37 °C for an hour and analyzed by western blot, alongside untreated controls. Native human seminal plasma was used as a positive control (Lee BioSolutions).

Enzyme kinetics

KLK3 is secreted from the cell as proKLK3, which carries an activation sequence of seven amino acids at the amino terminus. Purified and dialyzed recombinant proKLK3 proteins were activated by incubating the rKLK3 protein with 1 μg/ml thermolysin (Sigma-Aldrich) in activation buffer (50 mM Tris-HCl, 10 mM CaCl₂, 150 mM NaCl, 0.05% Brij-35, pH 7.5) at 37 °C for 5 min. To inhibit further thermolysin activity, 1,10 phenanthroline (Sigma-Aldrich) was added to a final concentration of 10 mM and incubated for five more min. Control experiments showed that thermolysin did not cleave the substrate (data not shown).

The short peptide His-Ser-Ser-Lys-Leu-Gln linked to the fluorophore amino-4-methyl-coumarin (AMC) was used as a substrate (Mu-HSSKLQ-AMC; Sigma-Aldrich). The protease activity of the purified and dialyzed rKLK3 from six species, along with native human KLK3, was measured immediately after adding different concentrations of substrate (final concentration of 0–500 µM) in assay buffer (50 mM Tris-HCl, 0.1 M NaCl, pH 7.8) and 5% DMSO, in a total volume of 100 μl (Denmeade et al., 1997). Substrate hydrolysis was measured in triplicate through fluorescence change over time using a 96-well plate reader (SpectraMax iD3). Fluorescence from six different concentrations of AMC (0–1,000 pmol) was measured under the same conditions in duplicate to create a standard curve. Measurements were taken every minute while incubating at 37 °C with orbital shaking before and between readings at excitation/emission wavelengths of 345 nm/445 nm. The initial enzyme velocity (V₀) was measured as the slope of the fluorescence over 30 min time and fitted to the Michaelis–Menten equations with nonlinear regression (GraphPad Prism 9) to estimate V_max, k_cat, and K_m. Differences among species were tested for significance with one-way ANOVA, followed by Tukey’s post-tests.

Recombinant human and chimpanzee SEMG1-IIIaIa-MBP segments were also used as substrates in proteolytic assays. Purified and dialyzed rKLK3 proteins were activated with thermolysin and 1,10 phenanthroline, as above, and then incubated for 30 min at 37 °C with these semenogelin fusion proteins. Samples were run on 10% SDS-PAGE gels in duplicate, followed by Coomassie staining.

Results

Sequence alignment and structural analysis of KLK3 homologs

To gain insights into the potential functional differences among the six KLK3 homologs used in the functional assays, their amino acid sequences were aligned (Figure 1A). The 17-amino acid-long signal peptide sequence was found to be fully conserved in four species, except for unique substitutions in human and gibbon. The activation sequence was also fully conserved across all species, except for chimpanzee. However, in our study, we designed the expression constructs for all six homologs to have the same signal peptide and activation sequences as those in humans. This approach ensures the elimination of any potential functional differences arising from the expression and secretion of these recombinant proteins in a human cell line, and in the activation process of proKLK3 itself.

Figure 1.

Sequence alignment and tertiary structure of KLK3. (A) Amino acid residues in dark gray denote the signal peptide, followed by the activation sequence in light gray. Residues forming the catalytic triad are highlighted in red, the kallikrein loop in yellow, substrate binding sites in green, and N-glycosylation sites in cyan. The black triangle shows the beginning of exon V at amino acid position 211. Fully conserved residues are indicated by asterisks, partly conserved residues by colons, and weakly conserved residues by periods. (B) Mature KLK3 pairwise 3D structure alignments (amino acids 25–261) predicted using AlphaFold Colab and visualized in PyMOL v2.5. The catalytic triad, kallikrein loop, substrate binding, and N-glycosylation sites are shown in the same colors as in the sequence alignment and represented as sticks, except for the kallikrein loop. Amino acid substitutions between species are mapped onto the aligned structures and displayed as sticks. For example, “Glu/Lys32” represents Glu in human and Lys in chimpanzee at position 32. Values in parentheses indicate the number of substitutions in the comparison. Human, chimpanzee, gorilla, and gibbon KLK3 structures are shown in black, hot pink, deep blue, and violet, respectively, with a transparent color.

The alignment of KLK3 sequences revealed complete conservation of the catalytic triad His65, Asp120, and Ser213 in all six homologs (His57, Asp102, and Ser195 in chymotrypsin numbering). The substrate binding sites Ser207, Ser228, and Gly230 were also fully conserved. The kallikrein loop is relatively conserved, with one conservative substitution in gibbons and two amino acid substitutions (one conservative and one nonconservative) in macaque (Figure 1A). Previous studies have reported the presence of a single N-glycan at Asn69 (Asn61 in chymotrypsin numbering) in native human KLK3 (Stura et al., 2011). Interestingly, this residue has been conserved in chimpanzee and gorilla, but not gibbon or macaque. However, a different site, Asn102, was predicted to be N-glycosylated in both gibbon and macaque by the NetNglyc server, with scores of 0.6483 and 0.5959, respectively (Gupta & Brunak, 2002). Humans and the great apes have an Asp at this site (Figure 1A). Furthermore, a single O-glycosylation site has been reported at Thr143 in human KLK3 (Stura et al., 2011), and this residue is conserved in other homologs. The flanking sequence around the O-glycosylation site (DLPTQEP) was fully conserved, except for macaque KLK3. However, the NetOGlyc4.0 server (Steentoft et al., 2013) did not predict any potential O-linked glycosylation sites with scores higher than 0.5, neither at this site nor across other sites in all six homologs.

Despite amino acid substitutions among species, the predicted structures of the mature active forms of the proteins are highly conserved. Structural alignments of the five homologs to human KLK3 resulted with root mean square deviation values of <1 Å. The maximum difference among all homologs is 0.1 Å in their distances between the heteroatoms forming the catalytic triad (Figure 1B; Supplementary Figure S2). Five out of seven amino acid differences between humans and gorillas occur in the fifth exon.

Phylogenetic tests of selection of KLK3 in primates

Across 13 catarrhine primate species, KLK3 appears to be evolving with a moderate level of purifying selection, with a dN/dS (or ω) ratio of 0.4 estimated across all branches under a one-ratio model (M0), encompassing the first four exons of KLK3 (Figure 2). However, a model allowing for each branch’s ω to vary freely (M1) is significantly more likely, suggesting variation in the mode or intensity of selection across lineages (p = 0.049). The fifth exon was not included because in gibbons and gorillas it is derived from KLK2 and therefore not evolving vertically from ancestor to descendant, as phylogenetic models assume. Foreground branches, defined as those leading to species with relatively large testes, have slightly but nonsignificantly higher dN/dS ratios than the background branches. As a whole these foreground branches are not evolving neutrally (ω = 1), but rather showing some level of purifying selection (p = 0.016; Figure 2B).

Figure 2.

Phylogenetic tests of selection of exons I-IV from KLK3 homologs. (A) Unrooted consensus species tree (cladogram) displaying dN/dS ratios (ω) in 13 primates obtained under a free-ratios model. In cases where the ω is infinity or zero, the dN/dS ratio is given in parentheses. Underlined names indicate species with relatively large testes. Circles at the tips indicate the species used in KLK3 functional assays below. (B) Model comparisons from likelihood ratio test results. “df” indicates degrees of freedom. ^†ω_bg, ω for background branches and ω_fg, ω for foreground branches. Foreground branches are shown in red in the tree. ^‡BEB results were used to detect positively selected sites, with the posterior probabilities for each site given in parentheses. Sites with posterior probabilities >0.80 are indicated in regular type, >0.85 are underlined, >0.90 are in bold. *Significant p < 0.05 **Significant p < 0.01 ***Significant p < 0.001. (C) Human KLK3 tertiary structure showing positively selected sites identified under site models. The complete mature protein structure of human KLK3 is displayed although the selection models were tested on exons I–IV.

When allowing for variation in selection over codon sites, we found a significant difference between the null hypothesis of the nearly neutral model M1a and the positive selection model M2a (Figure 2B). Under M2a, 85% of the codon sites in the KLK3 sequence were found to have 0 < ω₀ < 1, 10% of sites had ω₁ = 1, and 5% of sites had ω₂ > 1. Two sites, 189P and 203T, were identified as under positive selection with posterior probabilities higher than 95%. Using a less stringent model, comparing M7 (beta) against M8 (beta&ω), gave similar results. Both sites are located on the protein surface of KLK3 and are not involved in the active site, substrate binding site, or the kallikrein loop (Figure 2C). Comparisons involving branch-site models did not yield any significant differences, indicating that there are no codon sites under positive selection specifically along the foreground branches, defined as those leading to species with relatively large testes (Figure 2B). Although these analyses were performed excluding the fifth exon, all of them were repeated with the full-length coding sequence but excluding gorilla and gibbon with the full results given in Supplementary Figure S3.

Successful expression and purification of recombinant KLK3 homologs

Expression of all six rKLKs was confirmed with western blot in the crude media, appearing as a single prominent band of approximately 30 kDa (Figure 3A). We opted to employ an anti-His₆ antibody instead of an anti-KLK3 antibody, as the latter may exhibit varying sensitivities towards the KLK3 homologs from different species. Surprisingly, the gibbon cKLK3 protein exhibited a slightly smaller size than the other species, despite having the same amino acid length. The predicted molecular weight of the gibbon homolog is 27.6 kDa, excluding the signal peptide and any post-translational modifications but including the hexahistidine tag, whereas that of the human is 27.7 kDa. The observed size difference on the blot appears larger than 0.1 kDa. No His₆-tagged protein expression was detected from the empty vector control.

Figure 3.

Expression and purification of recombinant KLK3 proteins. (A) Western blot of crude media samples collected 96 hr after transfection of HEK-293T cells with pCMV constructs carrying cDNA of six KLK3 homologs, along with the vector control (see Methods section for species abbreviations). (B) Silver-stained SDS-PAGE of HumKLK3 protein collected during purification and dialysis (CM: crude media, FT: column flow through, W1–W3: column washes, E1–E3: elutions, Dial: dialyzed samples of combined elution 1 and 2). (C) Silver-stained gel after purification and dialysis. 2 μg of native human KLK3 (NatKLK3) was loaded as a size standard but observed as multiple fragments. Uncropped images are available as Supplementary Figure S4.

Following purification with metal affinity chromatography and buffer exchange via dialysis, the KLK3 homologs appear as a single band of high purity as assessed with SDS-PAGE and silver staining (Figure 3B and C). Different protein ladders used in the western blot (WB-Master, GenScript) than in the SDS-PAGE (PageRuler Plus, ThermoFisher), likely explain the apparent discrepancy in size between the blot and the gels. Notably, no protein was detected in the vector control after purification. The commercially purchased native human KLK3, purified from whole semen, contains a prominent band of a similar size as the recombinant human KLK3, along with lower molecular weight bands (Figure 3C).

All six recombinant KLK3 homologs appear to be post-translationally N-glycosylated, as evidenced by the band shift upon treatment with PNGase F, in a similar fashion to native KLK3 in human semen (Figure 4). Considering that each glycan contributes between 1.4 and 2.7 kDa to the molecular weight (Wang et al., 2017), it appears that all six rKLK3 homologs possess a single N-glycan, similar to native KLK3. The recombinant gibbon and macaque proteins are glycosylated at the Asn residue at position 102, confirmed through site-directed mutagenesis (Supplementary Figure S6). Importantly, neither the purification process nor dialysis disrupted this glycosylation pattern.

Figure 4.

Western blot of rKLK3 proteins before and after the PNGase F treatment. Recombinant proteins and 2.4 µg of human seminal plasma (HumSP) were incubated with PNGase F at 37 °C for one hour. Untreated samples were similarly incubated but in the absence of PNGase F. All KLK3 homologs are detected by the rabbit anti-KLK3 primary antibody (green); recombinant His-tagged proteins are also detected by the mouse anti-His₆ primary antibody (red). Uncropped image is available as Supplementary Figure S5.

Enzyme kinetics vary substantially among KLK3 homologs

Enzyme activity of recombinant KLK3 homologs was quantified with the widely used synthetic substrate Mu-HSSKLQ-AMC, which fluoresces upon proteolytic cleavage. This peptide was first identified as that with the greatest specificity for KLK3 and highest catalytic efficiency from a library based on empirically determined cleavage sites of the human SEMG1 and SEMG2 proteins, the major targets of KLK3 in vivo (Denmeade et al., 1997). The particular sequence is derived from SEMG2; its homologous sequence in SEMG1 differs by a single amino acid (HSSKVQ). Prior to the assay, the concentrations of rKLK3s were normalized using quantitative western blot analysis (Supplementary Figure S7). Remarkable variation in reaction velocities (pmol s⁻¹) was observed, with the chimpanzee homolog exhibiting the highest rate at all concentrations, as well as the highest V_max and k_cat (Figure 5). Significant differences in V_max and k_cat values were observed for all pairwise comparisons, except between HumKLK3 vs GorKLK3 and GorKLK3 vs MacKLK3 (one-way ANOVA followed by Tukey’s multiple comparisons test; Supplementary Figure S8). Human, chimpanzee, their ancestor, gorilla, and macaque KLK3s exhibited similar K_m values, indicating comparable affinity towards the synthetic substrate. Notably, the chimpanzee KLK3 homolog showed the highest catalytic efficiencies (k_cat/K_m), while there was no significant difference between that of the human and the human-chimpanzee ancestor (Figure 5C and D). The recombinant homologs of the remaining species—gorilla, gibbon, and macaque—had significantly lower k_cat/K_m than human and also significantly different from each other, although the magnitude of difference between gorilla and macaque was minimal. The gibbon homolog had by far the lowest velocity and catalytic efficiency. The gibbon had a significantly greater V_max than the vector control (no enzyme), and therefore was able to catalyze this reaction at some level (95% CI excluded a V_max of zero). The K_m value estimated for commercially purchased native human KLK3 was similar to that of the recombinant human KLK3, at 4.23 × 10^-4 M, respectively (Supplementary Figure S9).

Figure 5.

Enzyme kinetics of rKLK3 proteins. Protease activity of six rKLK3 was measured on the synthetic chymotrypsin substrate, Mu-HSSKLQ-AMC. (A) Michaelis–Menten curves using varying substrate concentrations (0–500 µM), each measured in triplicate. The average of the triplicate vector control for each substrate concentration was used for background subtraction. (B) Enzyme kinetic parameters (mean ± SEM) determined by nonlinear regression. (C) and (D). Catalytic efficiencies (k_cat/K_m) and p-values from the k_cat/K_m comparisons among six rKLK3 species. All comparisons among species are significantly different, with the exception of human.

In addition to investigating the functional differences of rKLK3s on a synthetic substrate, we assessed their activity on recombinant peptides corresponding to one of their native physiological substrates in seminal plasma, SEMG1. For this, we used human and chimpanzee segments of SEMG1 corresponding to repeats IIIa-Ia expressed as MBP fusion proteins. While human, chimpanzee, and their common ancestor demonstrated clear activity on these substrates, the chimpanzee KLK3 appears to possess the most efficient degradation of both human and chimpanzee SEMG1 IIIa-Ia, consistent with the results shown above although with the caveat that the Anc_HC protein seems to have higher activity than human in this semiquantitative assay (Figure 6). Furthermore, there was no proteolysis of the MBP carrier protein, indicating that any degradation of MBP-SEMG1 IIIa-Ia was solely due to the activity of rKLK3 on SEMG1. Qualitatively similar results are seen regardless of whether the SEMG1 substrate is derived from human or chimpanzee.

Figure 6.

Degradation of a segment of SEMG1 protein by recombinant KLK3s from human, chimpanzee, and the predicted human and chimpanzee ancestor. Coomassie staining showing the cleavage of the human (left) and chimpanzee (right) maltose binding protein (MBP)-SEMG1 IIIa-Ia substrates after incubation with purified and dialyzed mature rKLK3 at 37 °C for 30 min. The MBP-SEMG1 segments were also incubated with the purification blank from the empty pCMV vector and alone as negative controls. Additionally, MBP was incubated with the pCMV blank and HumKLK3 as no substrate controls, as shown on the left gel image. The bands at ~90 kDa and ~50 kDa represent the SEMG1 segment and the MBP carrier protein, respectively. The same assay was repeated for a biological replicate, and both gel images are available as Supplementary Figure S10.

Discussion

Sexual selection plays an important role in driving the evolution of seminal proteins, including the prostate-specific KLK3 protein, although previous investigations have primarily relied on sequence data (Clark & Swanson, 2005; Marques et al., 2012; Mubiru et al., 2014). However, these studies did not explore the functional consequences of amino acid substitutions within KLK3 across diverse primate species, or how the chimerization of KLK2 and KLK3 into a single gene in gorillas and gibbons affects its function. We hypothesized that the KLK3 gene has evolved more rapidly and adaptively in species with high sperm competition. We also predicted that primate species sharing comparable levels of sperm competition would exhibit similar KLK3 activity. To test these hypotheses, we conducted intensive functional analyses complemented by phylogenetic and structural approaches. Our codon-based selection analysis did not reveal any correlation between higher sperm competition and an increased number of nonsynonymous substitutions in KLK3. However, our experimental analyses found variations in KLK3 protease activity among six homologs, despite the overall conservation of their predicted protein backbone structures. These functional differences uncover patterns suggestive of positive selection in polygynandrous chimpanzees and a relaxation of selective constraints in gorillas and gibbons, both of which experience relatively low levels of sperm competition, thus supporting our hypothesis.

Sequence evolution in KLK3 not affected by sexual selection through sperm competition

In general, the KLK3 protein has been moderately conserved during the evolution of the Old World primates, evidenced by the dN/dS value obtained from the one-ratio model (Figure 2). This result aligns with previous research, including studies conducted on the first four exons (Mubiru et al., 2014), as we have done, and on the full-length KLK3 (Clark & Swanson, 2005; Dorus et al., 2004; Marques et al., 2012). However, the free-ratio model provided a better fit than the uniform model, implying that KLK3 is evolving heterogeneously among branches, which is also consistent with other studies (Clark & Swanson, 2005; Mubiru et al., 2014). These findings collectively suggest that adaptations at the protein are not the result of recurrent selective sweeps, but rather via more subtle changes occurring among different species. Additionally, the nearly neutral model exhibited a superior fit to the data compared to the uniform model, indicating that the highly variable selective pressures are acting among the amino acid sites in the KLK3 gene. Site models applied to exons I-IV identified two codons (189 and 203) evolving under positive selection (Figure 2). The same analyses conducted on the full-length KLK3, excluding chimeric KLK3s from the test, identified an additional four codon positions under positive selection (Supplementary Figure S3). Of note, codons at 189P and 203T (where the given amino acid residues correspond to human KLK3) had been previously reported as positively selected, as were two other sites, 238E and 248H, in the full-length analysis (Marques et al., 2012; Mubiru et al., 2014). Among these sites, the proline substitution at codon position 189 (P189F/S/Y) is particularly noteworthy, as it is believed to induce changes in the secondary and tertiary structure of the protein, potentially leading to functional alterations (Mubiru et al., 2014). Additionally, codon positions at 165F (F165Y/H/N) and 238E (E238Q/R) are near the substrate binding pocket. The newly identified phenylalanine substitution at position 165 may affect substrate binding efficiency by altering the surrounding hydrophobicity. Our two-ratio and branch-site models did not show any evidence of faster evolution in either the entire KLK3 gene or specific codons, particularly in species characterized by high levels of sperm competition, where we used relative testes size as a proxy for sperm competition. Instead, purifying selection was a better fit than neutral evolution for these species.

Functional experiments reveal high enzymatic activity in chimps, low in gorillas and gibbons

Comparisons with commercially purchased native KLK3 and human seminal plasma have validated the successful expression, purification, and post-translational modifications of our six recombinant KLK3 proteins (Figures 3 and 4). Enzyme kinetics, calculated using the synthetic KLK3 substrate, revealed significant differences among homologs (Figure 5). Before delving into these specific details, we aim to compare the kinetic parameters obtained from our recombinant human KLK3 with those from the native KLK3 and previously published data. It is important to note that we expect the activation sequence of native KLK3 to be naturally removed in vivo. Nevertheless, we still treated it with thermolysin to standardize the assay. Intriguingly, we observed that additional thermolysin treatment enhanced the activity of native KLK3, implying that a certain percentage of native KLK3 might still exist in its pro-protein form (data not shown). Furthermore, our K_m value (423 μM) for native KLK3, aligns closely with the K_m value of 470 μM reported in a prior study utilizing the same substrate (Denmeade et al., 1997). However, another study found a markedly lower K_m value (19 μM) for native human KLK3 at pH 7.5 (Tomao et al., 2014).

The most exciting result of our study was the significantly higher efficiency of KLK3 observed in chimpanzees compared to humans and the last common ancestor of humans and chimpanzees. After the divergence from this ancestor, both humans and chimpanzees each independently acquired a single amino acid substitution along their lineages. Considering the conservative replacement of isoleucine with leucine within the ancestor-to-human lineage, it is not surprising that we observed similar levels of protease function in both homologs. This result prompted us to attribute the enhanced activity in chimpanzees relative to humans to the glutamic acid to lysine substitution at position 32 (E32K), unique to chimpanzees. It is difficult to mechanistically connect this substitution to altered enzymatic activity as this site is on the protein surface outside of any organized secondary structure and not near the catalytic triad or substrate binding pocket, although we note that an adjacent residue (C31) is involved in a disulfide bond. Surprisingly, both the E32K substitution in chimpanzees and the human-specific leucine substitution exist as single nucleotide polymorphisms in the human population (Goard et al., 2007). The isoleucine allele at position 132 (dbNSP ID rs2003783) is found at a frequency of nearly 10% in the worldwide population. It is not clear if the isoleucine allele is a retention of the ancestral allele or the result of a parallel mutation. However, it has not been significantly associated with diseases, including prostate cancer and male infertility (Gupta et al., 2017; Végvári et al., 2013), consistent with our finding that the Anc_HC KLK3 is indistinguishable from the typical human KLK3 in its enzyme kinetics. The SNP-E32K variant (rs2271092) is very rare in humans, with a global allele frequency of less than 0.001 (dbSNP, accessed 17 July 2024), and no reported association with disease.

Our findings indicate that the chimeric KLK2/3 proteins in gorillas and gibbons exhibit significantly lower activity compared to those in humans, chimpanzees, and their common ancestor. This difference is likely due to the presence of KLK2-like exon V in gorillas and gibbons, where two out of three substrate binding site residues are located (Figure 1). The amino acid changes, specifically S231P, R239K, S241A, and L242V, common to both gorilla and gibbon cKLK are situated within or near the binding pocket and may contribute to the reduced enzymatic function. Among these alterations, we anticipate that the serine to proline substitution at position 231 plays a key role in decreasing functionality due to its nonconservative nature and close proximity to the substrate binding site of G230 (Figure 1). We have shown that both gorillas and gibbon cKLK retain some functionality as a KLK3-like protein, as previously suggested by Marques et al. (2012). However, their efficiency and velocity, particularly in the case of gibbon cKLK3, are significantly reduced. KLK2 is typically regarded as the primary activator of KLK3 in humans (Lövgren et al., 1997; Takayama et al., 1997). Without two distinct enzymes, it remains unclear whether the gorilla and gibbon cKLK enzymes are activated by a different protease or have self-activating capabilities. It is possible that the chimeric proteins have a different or wider range of substrates, while still retaining some amount of chymotrypsin activity as the catalytic triad is encoded by the first four exons.

We also noted low activity in macaque KLK3. Compared to humans, noteworthy differences in macaque include two substitutions in the kallikrein loop and a distinct N-glycosylation site. Furthermore, if human KLK3 indeed possesses a single O-glycosylation site at position T143, as reported (Stura et al., 2011), the nonconservative replacement of Q144W in macaque might affect this O-glycosylation (Figure 1), potentially altering function. It is worth mentioning that KLK2 in rhesus macaques has a premature stop codon, rendering it a nonfunctional pseudogene (Clark & Swanson, 2005; Marques et al., 2012). Consequently, the same questions regarding KLK3 activation, as discussed for gorillas and gibbons above, could be applied to macaques as well.

The recombinant human and chimpanzee KLK3 proteins also displayed activity on substrates mimicking their native substrate, SEMG1 (Figure 6). Most importantly, human KLK3 exhibited activity on both the chimpanzee SEMG1 segment and human SEMG1 segment. Similarly, chimpanzee KLK3 demonstrated activity on both human and chimpanzee SEMG1 segments. This indicates that the differences in activity are not substrate-dependent and that there is no bias when using a substrate previously optimized for humans when testing with other species. Qualitatively similar results to the fluorescent artificial substrates were observed, such that the chimpanzee KLK3 homolog had higher level of activity compared to humans or their last common ancestor.

N-glycosylation retained across homologs with switched sites in gibbons and macaques

We have demonstrated that nonhuman KLK3 homologs are N-glycosylated as has been reported for human KLK3 (Figure 4). However, the specific N-glycosylation site varies among species; whereas humans and presumably the great apes glycosylate position N69, gibbons and macaques instead glycosylate site N102, located near the kallikrein loop (Figure 1; Supplementary Figure S9). It is possible that along with a shift in location of the glycosylation, a smaller glycan moiety is attached in gibbon contributing to the lower molecular weight. Interestingly, in humans, the amino acid position at 102 has been identified as a SNP (rs61752561, D102N), which has been strongly associated with an increased risk of prostate cancer (Gupta et al., 2017; Srinivasan et al., 2019). The minor allele with asparagine results in the creation of a second N-glycosylation site in human KLK3 and has been reported to induce greater flexibility in the kallikrein loop, limiting its ability to facilitate substrate access to the binding pocket, consequently leading to reduced enzymatic activity (Srinivasan et al., 2019). This allele is found at a global frequency of about 0.03, and notably higher in European populations compared to all others (dbSNP, accessed 17 July 2024). A similar effect is also seen at the paralogous human KLK2, which has a glycosylated asparagine at position 102 in vivo; removal of this glycan yields an enzyme with higher activity (Guo et al., 2016). Therefore, the decreased protease activity observed in gibbons and macaques may, in part, be attributed to the altered position of N-glycosylation, which could affect protein stability, substrate binding efficiency, or other factors. Regardless, the conservation of N-glycosylation across species despite variation in the sites, suggest a crucial role in the protein’s function and regulation.

Protease function of KLK3 affected by sexual selection through sperm competition

Despite only two amino acid differences in the KLK3 protein between humans and chimpanzees, the latter shows significantly greater catalytic efficiency (Figure 5). Humans, in turn, have more efficient enzymes than gorillas or gibbons, who both possess chimeric KLK2/KLK3 proteins. These trends parallel the expected levels of sperm competition in these four ape species based on behavioral observations and relative testes size (Figure 7), and therefore we suggest that the functional differences observed reflect the intense sperm competition in chimpanzees and the relaxation of selection in gorillas and gibbons. Numerous authors have now reported correlations between inferred levels of sperm competition and various measures of sequence and structure evolution of KLK3 and the SEMG1/SEMG2 structural proteins, its major physiological substrate. Specifically, the number of 60aa repeats of the SEMG proteins is greater in species with high sperm competition (Jensen-Seaman & Li, 2003; Marques et al., 2012); the dN/dS ratio of SEMG2 is positively correlated with residual testes size, observed numbers of male partners of females per ovulatory period, and degree of semen coagulation (Dorus et al., 2004; Marques et al., 2012); and the number of functional KLK2 and KLK3 genes is positively correlated with the degree of semen coagulation and residual testes size (Marques et al., 2012). These correlations rely on several simplistic assumptions, notably that crude measures like the number of repeat units or number of functional genes reliably reflect selection intensity, that selection works mostly through recurrent selective sweeps, and that synonymous substitutions are neutral. Our empirical approach of testing the function of homologous proteins allows for the direct observation of functional differences among species.

Figure 7.

Catalytic efficiency (K_cat/K_m) of the KLK3/cKLK homologs is positively correlated with relative testes size in hominoids. The x-axis is the log of the residual from a primate-wide correlation of body size and testes size (Kenagy & Trombulak, 1986).

The KLK3 protein corresponding to the last common ancestor of humans and chimpanzees (Anc_HC) had enzyme kinetics very similar to modern humans (Figure 5). If, as appears to be the case, the evolution of KLK3 is tracking with the intensity of postcopulatory sexual selection, our results suggest that the human-chimpanzee ancestor had relatively low levels of sperm competition and that the extensive female promiscuity seen in chimpanzees may be derived in the Pan lineage. Although we hesitate to speculate too much from a single gene, this in turn suggests that modern human sociosexual behavior—characterized by strong male-female pair bonding and high levels of male provisioning (Burkart et al., 2009)—evolved from an ancestor already possessing low sperm competition typically seen in monogamous or single-male polygynous mating systems. This interpretation is at odds with previous models of human evolution that assumed the last common ancestor was chimp-like in its polygynandry (Lovejoy, 2009) but may be consistent with models proposing a gorilla-like ancestor or one in which humans went through several stages (Chapais, 2013; Swedell & Plummer, 2012). Since neither behavior nor testes fossilize, it is essential to seek alternative data to reconstruct the social and sexual lives of our ancestral species, such as the comparative functional approach used here.

Supplementary material

Supplementary material is available online at Evolution.

Data availability

Uncropped gel images and additional replicates of gels can be found in the Supplementary Materials. Complete sequences of plasmid constructs have been deposited in Dryad (https://doi.org/10.5061/dryad.931zcrjv6). All plasmids will be made available as purified DNA to any researcher requesting them without restriction or charge.

Author contributions

E.F.K., A.C.Z., and M.I.J.S. conceived the project. E.F.K., A.C.Z., W.J.G., C.A.M., R.M., H.Z.J., T.H.W., D.R.B., M.R.H., and N.R.R. collected the data and carried out all analyses. E.F.K. and M.I.J.S. created figures and wrote the manuscript.

Funding

This work was funded by grants to M.I.J.S. from the National Institute of General Medical Sciences (1 R15 GM1234407-01), the Wenner-Gren Foundation (8839), and the Duquesne University Faculty Development Fund.

Conflict of interest: The authors declare no conflict of interest.