Three genes expressing Kunitz domains in the epididymis are related to genes of WFDC-type protease inhibitors and semen coagulum proteins in spite of lacking similarity between their protein products

The mammalian semen coagulum proteins are a heterogeneous collection of proteins secreted at very high concentration by the seminal vesicles. There are two homologous semen coagulum proteins, denoted semenogelin I (SEMG1) and semenogelin II (SEMG2), in most primates [1–3]. Duplication of tandem repeats of 60 amino acid residues in both SEMG1 and SEMG2 is responsible for most of the size heterogeneity of semenogelin molecules in primates, but the frequent occurrence of premature stop codons is also a contributing factor [1, 4, 5]. It has also been reported that the structure of semenogelin molecules can be affected by gene conversion [6].

The murine seminal vesicles secrete six proteins, Svs1-Svs6, at high concentration [7–9]. It has been shown that the genes of Svs2-Svs6, and SEMG1 and SEMG2 are homologous, in spite of lacking similarity between their protein products [10, 11]. This apparent paradox is explained by rapid evolution of single exon that codes for most of the gene product. The homologous genes are composed of three exons, the first of which encodes the signal peptide and the very first residues of the secreted protein, the second codes for the remainder of the protein and also carries a few 3' non-translated nucleotides, whereas the third exon only has 3' non-translated nucleotides and holds the poly-adenylation signal. Sequence comparison shows that the first and last exons are conserved in these genes, whereas the second exon is not [10]. Differing selection of splice acceptor sites and species unique repeat expansion have been proposed to explain the diversity of the second exon [10].

The proposed evolutionary mechanisms suggest that the seminal vesicle-transcribed genes might be related to genes with similar organization that code for secreted, but structurally very different proteins. This hypothesis led to the discovery that the genes of the elastase inhibitors elafin and secretory leukocyte protease inhibitor (SLPI), and the predominating seminal vesicle-secreted proteins are homologous, in spite of lacking similarity between their protein products [12]. Elafin and SLPI belong to the whey acidic protein four-disulfide core (WFDC) family of small protease inhibitors, which interacts with cognate proteases by what is known as the standard mechanism, by which the protease-binding loop of the inhibitors interacts with the catalytic site of proteases in a similar way as substrates [13].

With the advent of nucleotide sequences from the Human Genome Project, it became possible to assign the chromosomal location of genes in detail. Analysis of the human semenogelin locus on chromosome 20q12-13.1 showed that PI3, the gene of elafin, and SLPI are flanking SEMG1 and SEMG2 and an extended analysis of the chromosomal region identified 12 additional, homologous genes encompassing WFDC motif [14, 15]. Most of the WFDC genes and Svs2-Svs6 are conserved at the homologous chromosomal region in the mouse, which suggests that the semen coagulum proteins probably originate from a WFDC gene [16]. Many of the WFDC genes are with certain specificity expressed in the epididymis and display accelerated evolution, which might indicate that they are of importance in male reproduction [16].

Three of the novel genes with WFDC motifs, WFDC8, SPINLW1, and WFDC6 also carry the motif of Kunitz-type protease inhibitors [14]. Similar to the WFDC domain, the conserved Kunitz domain is present in several proteins that inhibit proteases by the standard mechanism [13]. A well-known and much studied example is the bovine pancreatic trypsin inhibitor (BPTI), also known as aprotinin and until recently marketed as an anti-bleeding drug under the name of Trasylol^®. Kunitz domains are also present in important protease inhibitors, such as the tissue factor pathway inhibitor, which regulates blood coagulation, and the light chain, also known as bikunin, of the multifunctional inter-alpha trypsin inhibitor [17–19].

We have in several previous studies, as outlined above, demonstrated that many of the genes at the semenogelin/WFDC locus are expressed in the male reproductive tract and display accelerated evolution by a variety of mechanisms. Most compelling are the findings suggesting that a WFDC gene gave rise to the genes of the semen coagulum proteins, as it also rises the question whether there could be still more genes at the locus on chromosome 20 that have gone through a similar metamorphosis. Our finding that the protease inhibitor locus on chromosome 20 carries several genes that code for inhibitors containing WFDC motifs only, while other genes encode products containing both WFDC and Kunitz motifs, raised the question as to whether there also exists homologous genes encoding putative inhibitors with Kunitz motifs alone. In this report, we have searched the locus for novel genes encoding proteins with Kunitz motifs. The identified genes have been analyzed with respect to structure, expression and function.

In this report, we describe the transcripts of three novel genes encoding Kunitz motif. They were discovered by analysis of nucleotide sequences from a locus where we previously have identified homologous genes of semen coagulum proteins and putative WFDC-type serine protease inhibitors [14]. In earlier studies we have shown that genes of different semen coagulum proteins and WFDC-type protease inhibitors are homologous, in spite of lacking structural similarity of their protein products [10–12, 15]. By and large, these similarly organized genes carry the signal peptide, the secreted protein and 3' non-translated nucleotides on three separate exons. Sequence and size conservation of the first and the third exon suggests that the genes are homologous, whereas the second exons, encoding the secreted protein, in most instances display very little sequence conservation. Two mechanisms, de novo selection of splice sites and size expansion by duplication of repeats in combination with replication slippage, have been proposed to explain the structural variation of the second exon [10].

The novel genes of putative Kunitz inhibitors are organized in a similar way as the semen coagulum and WFDC genes. Another characteristic feature of the genes is that the splice acceptor of the second exon always is that of a phase 1 exon, which means that it starts with the second nucleotide of the codon. Of the novel genes, it is only SPINT4 that has all of the common features, whereas SPINT3 is lacking the third exon similar to WFDC10A and WFDC10B. At a first glance, human SPINT5 might appear to differ from the other genes at the locus, as the structure of the longer 3' RACE product suggests that it could be transcribed from 8 exons. However, the splice form that is orthologous with mouse Spint5 is very similar organized as SPINT4, except for the presence of the puzzling small second exon. In spite of the lacking relationship between the protein products, the similarities in gene organization, chromosomal location, and site of expression, suggests that the novel genes with Kunitz motif are homologous with genes encoding WFDC-inhibitors and semen coagulum proteins. A Kunitz inhibitor within the frame of a WFDC gene could have evolved by exon shuffling that created an intermediate encompassing both a WFDC and a Kunitz domain, as is present in SPINLW1 and WFDC6. A subsequent loss of the WFDC domain in a purifying process would have created a Kunitz gene with the same regulatory sequences as the original WFDC gene.

The problems we experienced in generating a full-length transcript of human SPINT5 were probably caused by low levels of transcript in the samples from our tissue panel. Furthermore, the in frame stop codon between the signal peptide and the Kunitz motif could indicate that the RACE products were emanating from either a not fully spliced nuclear transcript or a erroneously spliced cytoplasmic transcript. In order to address this, we generated the sequence of mouse Spint5 and compared it with those of the human transcripts. The failure to detect a mouse equivalent of the alternative human SPINT5 transcripts and the peculiar structure of them, with no coding information for a signal peptide or a 3' non-translated sequence derived from four exons, suggests that they could originate from not fully processed messages or cryptic transcripts generated by accidental transcription of the gene in a tissue where it is not normally expressed. The former could also be the case for the human transcript that is orthologous with the single mouse transcript, where the second exon might not have been fully processed, as it contains an in frame stop codon that prevents translation of the Kunitz motif. When the sequence of the second exon was shortened to the corresponding nine nucleotide of its mouse ortholog, the reading frame is kept and a Kunitz domain is translated from the third exon. The presence of an appropriately located CAG triplet that could function as splice acceptor in exon 2 raises the possibility that human SPINT5 could be processed the same way as mouse Spint5.

Semi-quantitative RT-PCR has previously shown that several of the WFDC genes at the locus on chromosome 20 are expressed in many different tissues [14]. However, the same study also showed that a number of the genes at the telomeric sublocus, e.g. WFDC8, WFDC9, WFDC10A and WFDC13, are transcribed in the epididymis with certain specificity. Epididymal specificity in expression has also been reported for SPINLW1 and WFDC2[20, 21]. In this study it is shown that SPINT3 and SPINT4 are highly specific for the epididymis as well. This conclusion is supported by qPCR that show expression levels in the epididymis that are of the same order of magnitude as for CSTB, a housekeeping gene with low to moderately high expression levels. More importantly, the transcript levels of SPINT3 and SPINT4 in the epididymis exceeds those of the same genes in other tissues that were analyzed by more than three orders of magnitude, which suggests that the expression in those tissues might be neglectable. The result is also in line with an earlier study aiming at identification of epididymis-specific genes in the mouse, in which Spint4 was found to be highly expressed in a narrow segment encompassing distal caput and proximal corpus of epididymis [22].

Transcripts of SPINT5 were only detected in the epididymis, but at a level that is at least four orders of magnitude lower than for SPINT3 and SPINT4. In analogy with the discussion above, the low transcript level would indicate that under normal conditions SPINT5 presumably is of less importance in the tissues that were analyzed in this study. From this follows that SPINT5 probably is expressed either at a relatively rare anatomical site or at a developmental stage that is not covered by this investigation. In ruminants, it has been shown that several proteins with Kunitz motifs are expressed and secreted by the conceptus prior to, or at the time of, trophoblast implantation [23, 24]. Although, there are major differences in the implantation process between ruminants and many other mammals, e.g. in the timing, it is not unlikely that Kunitz inhibitors could be expressed at the time of trophoblast implantation also in humans, and one interesting candidate to study would be SPINT5.

Protein domains are building blocks that usually fold to explicit three dimensional structures, which many times have a specific function. The Kunitz domain consists of 50-60 amino acid residues, six of which are Cys that form three disulphide bonds and have characteristic locations in the primary structure. Studies on recombinant Kunitz domains expressed in E. coli show that proper folding and disulphide formation can be achieved either by directing the recombinant protein to the periplasmic space or by refolding of the protein extracted from inclusion bodies [17, 25]. Our approach was to express the recombinant protein as a part of a larger fusion product with thioredoxin, that would generate high solubility and support the proper folding of the Kunitz domain, and by using the E. coli K-12 sub-strain Origami(DE3), formation of disulphide bonds in the cytoplasm would be favored. Studies with recombinant BPTI and BSTI showed that the proteins isolated from the cytoplasm were almost exclusively in a monomeric state, which suggests correct disulphide formation and folding of the Kunitz domain. Furthermore, the recombinant inhibitors were as functionally active as BPTI isolated from bovine pancreas, something that also indicates that they are correctly folded. Expression of SPINT3 yielded an exclusively monomeric protein in a similar way as BPTI and BSTI, which suggests that the Kunitz domain was properly folded. Recombinant expression of SPINT4 yields high molecular mass complexes that probably consist of misfolded peptide chains held together by disulphide bonds in an unorganized way, but also a substantial fraction of monomers that presumably are correctly folded. The structure of SPINT4 differs from that of the consensus Kunitz domain. The latter contains three conserved disulphide bridges, but in SPINT4 there will only be two, as one Cys is missing. The absent disulphide would have linked the C-terminal alpha-helix to one of the beta-strands at the core of the molecule. Earlier experiment has shown that the folding of the Kunitz domain is only minimally affected by removal of this bridge and would not affect the protease-binding loop that is situated at the opposite side of the molecule [26]. The location of a Cys amino-terminal to the Kunitz domain in SPINT4 suggests that the monomeric form will have either two free thiol groups, or more likely a disulphide bridge between them. The crystal structure of BPTI shows that the amino-terminus of the protein is located relatively close to the third Cys of the Kunitz motif. Therefore, it is not unlikely that a disulfide bridge could be formed in SPINT4 without affecting the protease-binding loop, which is located at the other end of the molecule. This would create a strand of 6 amino acid residues that links the protein's amino-terminal part to the third Cys of the Kunitz motif.

Despite several attempts, we did not succeed to isolate recombinant SPINT5 from the cytoplasm of Origami(DE3) cells. It is not clear whether the failure was due to inherent instability of the protein product or whether the problems are due to technical problems with our construct, but future studies will presumably give an answer.

The experiments with recombinant BPTI demonstrated, not only that the Kunitz domain can be correctly folded in Origami(DE3) cells, but also that the fusion product with thioredoxin could be an equally efficient protease inhibitor as BPTI isolated from a natural source. It is therefore likely that recombinant SPINT3 and SPINT4 would be functional as well. They were tested by a set of proteolytic enzymes that represent different catalytic specificities, but unfortunately, we did not succeed to identify a target protease. The amino acid residues in BPTI that interact with target proteases have been characterized in detail and homologous residues can be identified in other Kunitz domains by sequence alignment [27]. There are several residues that are important for protease-binding, but those flanking the scissile bond, particularly the residue at the P1 position, are vital for the target- protease specificity [28, 29]. The probable scissile bonds in SPINT3 and SPINT4 are between Gln and Thr, and Tyr and Glu, respectively (Figure 6). One might then have expected both of them to inhibit PSA, which has been demonstrated to cleave C-terminal to both Gln and Tyr in SEMG1 and SEMG2 [30]. There was no inhibition and perhaps the interaction with the inhibitor protein was prevented by the kallikrein loop in PSA, which restricts the access to the catalytic cleft [31]. In chymotrypsin there is no such structure as the kallikrein loop and SPINT4 could be predicted to interact with this enzyme, which cleaves after large hydrophobic residues. Under our experimental condition there was no inhibition, which suggests that there is no high affinity interaction between chymotrypsin and SPINT4. This emphasizes the importance of contact between amino acid residues other than those of the scissile bond, which is nicely shown with the inhibition of chymotrypsin by BPTI; the scissile bond in this case is between Lys and Ala, i.e. residues that would predict specificity for trypsin, but not chymotrypsin. In spite of this, chymotrypsin is inhibited by BPTI.

The WFDC genes at the locus on chromosome 20 could be important during sperm maturation, as several of them are highly expressed in the epididymis. There is this far no report on target proteases in the epididymis. It has also been speculated that the WFDC genes are important but not fully recognized components, of innate immunity; in which some modulate host response by regulating endogenous pro-inflammatory proteases, such as elastase and cathepsin G, whereas others regulate poorly characterized exogenous proteases used for tissue invasion by microbes [16]. The genes expressing Kunitz domains could have a similar function. In this report it is shown that SPINT3 and SPINT4 with high specificity are expressed in the epididymis, which suggests a role in the maturation of spermatozoa. An innate immune function can also inferred, as it has been shown that the Kunitz domain of SPINLW1 inhibits elastase and display antibacterial activity [32].

We have in this report described three genes that produce small secreted proteins, composed of single Kunitz domains that are relatively specific for the epididymis. The similar organization, chromosomal location and site of expression, suggests that the novel genes encoding Kunitz motifs are homologous with the genes of WFDC-type protease inhibitors and semen coagulum proteins, in spite of lacking similarity in primary structure of their protein products. A likely mechanism, by which the genes could have been formed, include an exon shuffling event, in which the Kunitz motif was taken up by a WFDC gene, and a following purifying process, in which the WFDC domain was lost. This would then be the third mechanism, beside de novo selection of splice site and repeat expansion, by which the regulatory frame of a conserved WFDC/semen coagulum gene could generate a new protein product. The restricted expression of the novel genes to the epididymis suggests that the secreted Kunitz domains are important in male reproduction. The recombinantly produced proteins are presumably bioactive, as demonstrated with similarly made BPTI, but may have a narrower spectrum of inhibition, as indicated by the lacking activity against eight proteases with differing specificity. Another possibility is that they have lost the protease inhibiting property, which is typical of Kunitz domains, in favor of new hitherto unidentified functions.