Accelerated evolution of small serum proteins (SSPs)—The PSP94 family proteins in a Japanese viper

Gene 426 (2008) 7–14

Contents lists available at ScienceDirect

Gene

j ourna l homepage: www.e lsev ie r.com/ locate /gene

Accelerated evolution of small serum proteins (SSPs)—The PSP94 family proteins in aJapanese viper☆

Narumi Aoki, Hisashi Matsuo, Masanobu Deshimaru ⁎, Shigeyuki TeradaDepartment of Chemistry, Faculty of Science, Fukuoka University, 8-19-1 Nanakuma, Jonan-ku, Fukuoka 814-0180, Japan

Abbreviations: cDNA, DNA complementary to RNA;protein; HPLC, high performance liquid chromatographopen reading frame; PLA2, phospholipase A2; PSP94, pramino acids; PAGE, polyacrylamide gel electrophorreaction; RACE, rapid amplification of cDNA ends; SDSsmall serum protein; UTR, untranslated region.☆ The nucleotide sequence data reported here havesequence data bank: SSP-1, AB360906; SSP-2, AB3609AB360909; SSP-5, AB360910.⁎ Corresponding author. Tel.: +81 92 871 6631; fax: +

E-mail address: [email protected] (M. Deshima

0378-1119/$ – see front matter © 2008 Elsevier B.V. Alldoi:10.1016/j.gene.2008.08.021

a b s t r a c t

Article history:

Keywords:cDNA cloningPhylogenetic treeSnake serumTrimeresurus flavoviridis

SSPs) with molecular masses of 6.5–10 kDa were detected in Habu (Trimeresurusflavoviridis) serum; this included two novel proteins SSP-4 and SSP-5. The amino acid sequences of theseproteins and of SSP-1, SSP-2, and SSP-3, which were reported previously, were determined on the basis of thenucleotide sequences of their cDNAs. Although these proteins exhibited only limited sequence identity tomammalian prostatic secretory protein of 94 amino acids (PSP94), the topological pattern of disulfide bondsin SSPs was identical to that of the mammalian proteins. SSP-3 and SSP-4 lacked approximately 30 residues atthe C-terminal. Each of the full-length cDNAs encoded a mature protein of 62–90 residues and a highlyconserved signal peptide. The evolutionary distances between SSPs estimated on the basis of the amino acidchanges were significantly greater than those of the synonymous nucleotide substitutions; these finding,together with results from analyses of nonsynonymous to synonymous rates of change (dN/dS) suggest thatsnake SSPs have endured substantial accelerated adaptive protein evolution. Such accelerated positiveselection in SSPs parallels other findings of similar molecular evolution in snake venom proteins and suggeststhat diversifying selection on both systems may be linked, and that snake SSP genes may have evolved bygene duplication and rapid diversification to facilitate the acquisition of various functions to block venomactivity within venomous snakes.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Proteins that inhibit snake venom hemorrhagic metalloproteinases(MPs) have been isolated from the sera of certain snakes; examplesinclude HSF from the Habu, Trimeresurus flavoviridis (Yamakawa andOmori-Satoh, 1992), and BJ46a from Bothrops jararaca (Valente et al.,2001). Previously, we isolated a novel 10-kDa protein, SSP-1, havingMP inhibitory activity from Habu serum (Aoki et al., 2007). Unlikeother antihemorrhagic proteins, however, this so-called small serumprotein (SSP) could only inhibit brevilysin H6, a weakly hemorrhagicMP isolated from Gloydius halys brevicaudus venom (Fujimura et al.,2000). Analysis of the amino acid sequence indicated that SSP belongsto the prostatic secretory protein of 94 amino acids (PSP94) proteinfamily, based on topological similarities of the cysteine residues,

CRISP, cysteine-rich secretoryy; MP, metalloproteinase; ORF,ostatic secretory protein of 94esis; PCR, polymerase chain, sodium dodecyl sulfate; SSP,

been deposited to the DDBJ07; SSP-3, AB360908; SSP-4,

81 92 865 6030.ru).

rights reserved.

although the sequence homology was low. Two additional similarproteins, SSP-2 and SSP-3, in the same serum from the T. flavoviridiswere also isolated; SSP-2 has strong binding affinity to triflin, asmooth muscle contraction blocker isolated from T. flavoviridis venom(Yamazaki et al., 2002). This suggests that several distinct SSPs exist insnake blood where they are likely used as a defense against the toxiceffects of their own venom which may circulate through the blood asthe result of accidental self-envenomation (Aoki et al., 2007).

PSP94, also called β-microseminoprotein, is a 10.7-kDa, nonglyco-sylated, cysteine-rich protein (Dube et al., 1987). Even though PSP94was first isolated as a major protein from human seminal plasma (Liljaand Abrahamsson 1988), it was later found to be present at the samelevel in both sexes (von der Kammer et al., 1990). PSP94 proteins havealso been identified in other mammals (Fernlund et al., 1994; Xuanet al., 1999; Mäkinen et al., 1999) and in ostrich (Lazure et al., 2001).However, the amino acid sequences of these PSP94 family proteinsshow a limited homology. Apart from the 10 cysteines that form fivedisulfide bonds, there are only 16 amino acids conserved in themammalian PSP94 family proteins. This suggests that these proteinsare quite divergent and may be functionally quite different than thehuman PSP94 (Mäkinen et al., 1999; Nolet et al., 1991).

Genes for some snake venom proteins from T. flavoviridis (Ogawaet al., 1995; Nakashima et al., 1995; Chijiwa et al., 2003), Naja naja(Chuman et al., 2000), and Ophiophagus hannah (Chang et al., 2001)have been reported to evolve rapidly by gene duplication followed by

8 N. Aoki et al. / Gene 426 (2008) 7–14

rapid accumulation of nonsynonymous nucleotide substitutions thatresult in amino acid replacements in the encoded proteins. This typeof molecular evolution results in the accelerated creation of toxinswith diverse functions (Ohno et al., 1998).

In this study, we isolated two additional SSPs, SSP-4 and SSP-5,from the serum of T. flavoviridis. We also isolated a series of cDNAclones encoding the five SSPs from the liver of the same species, anddetermined the nucleotide sequences of these cDNAs. Based onthese cDNA sequences, we investigated molecular evolutionarypatterns in these SSP genes. Our molecular evolutionary analysessuggest that unusually rapid evolution of the amino acid sequencesof snake SSP proteins has occurred, possibly driven by selection forfunctional diversification. Such radical rapid evolution of SSPproteins parallels numerous reports of similarly accelerated evolu-tion of snake venom proteins, and suggests a potential coevolu-tionary interaction between these two functionally related classes ofsnake proteins.

2. Materials and methods

2.1. Materials

The blood and liver of T. flavoviridiswere obtained from specimenscollected from the Amami Oshima Islands. The sera obtained from fiveanimals were pooled and stored at −20 °C. RNA was extracted from asingle snake. Enzymes used for RNA and DNA manipulation wereobtained from Takara Shuzo (Kyoto) except for PowerScript reversetranscriptase (Clontech). All other reagents were purchased fromWako Pure Chemicals (Osaka).

2.2. Purification of SSPs

The T. flavoviridis serum (30 ml) was loaded onto a SephacrylS-200HR column (5×90 cm). Elution was carried out at 4 °C with0.15 M NaCl–50 mM Tris–HCl (pH 7.4). Fractions containing SSPs werecollected, desalted by dialysis, and subjected to reverse-phase HPLC ona Cosmosil-5C8-AR-300 column (1×25 cm, Nacalai Tesque). Elutionwas carried out with an appropriate gradient of acetonitrile in 0.1%trifluoroacetic acid (TFA) at a flow rate of 3.0 ml/min, and the eluatewas monitored at 230 nm.

2.3. Electrophoresis

SDS-PAGE was carried out on a 16.5% polyacrylamide gel undernonreducing conditions (Schagger and Jagow, 1987). Ovalbumin(46,000), carbonic anhydrase (30,000), chymotrypsinogen (25,000),soybean trypsin inhibitor (20,500), lysozyme (14,300), aprotinin(6500), and insulin B-chain (3500) were used as the molecular weightmarkers. After running the gels under constant current, they werestained with 0.1% Coomassie brilliant blue R-250 and destained with10% acetic acid.

2.4. Mass spectrometric analysis

The mass spectrum was measured on a Voyager DE-STR matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) massspectrometer (PerSeptive Biosystems). The sample was dissolved in0.1% TFA-50% acetonitrile containing α-cyano-4-hydroxycinnamicacid (10 mg/ml) as the matrix. The spectrum was calibrated on thebasis of the molecular mass of apomyoglobin.

2.5. Sequence analysis

The protein was S-pyridylethylated (Friedman et al., 1970), and theamino acid sequence of the modified protein was determined by anautomatic protein sequencer PPSQ 21 (Shimadzu).

2.6. Cloning of cDNAs encoding SSP-4 and SSP-5

Total RNA was extracted from 0.5 g of T. flavoviridis liver using theacid guanidinium-phenol-chloroform method, and this RNA was thenreverse transcribed to synthesize the first cDNA strand using anadaptor-linked oligo(dT) primer (5′-GGCCACGCGTCGACTAGTAC-(dT)17-3′). The obtained cDNAs were used as templates for the 3′-RACE reaction. The cDNA cloning experiments for SSP-1, SSP-2,and SSP-3 have been described previously (Aoki et al., 2007). Syntheticoligonucleotides, i.e., ssp-4N (5′-TNCCNGARAAYGARGAYGGN-GARGAYGTNCCN-3′), ssp-5N (5′-NTGYTTYCARGGNWSNTTY-GARGCNAARMGNATG-3 ′ ) , and 3 ′ -adp (5 ′ -GGCCACGC-GTCGACTAGTAC-3′), were used for PCR amplification. The ssp-4Nand ssp-5N primers were designed on the basis of the N-terminalamino acid sequences of SSP-4 and SSP-5, and 3′-adp correspondedto the adaptor sequence within an adaptor-linked oligo(dT) primer.The amplification products were subcloned into the plasmid vector,and their nucleotide sequences were determined. The resultsconfirmed that they were partial fragments of cDNAs encodingSSP-4 and SSP-5. These fragments were radiolabeled with [32P]-dCTP (3000 Ci/mmol) using a random primer DNA labeling kit(Takara Bio) and then used for hybridization screening of a cDNAlibrary.

The T. flavoviridis liver cDNA library was constructed using theCreator SMART cDNA Library Construction kit (BD Biosciences)according to the manufacturer's instructions. Briefly, the first cDNAstrand was synthesized using 1 μg of total RNA, and this wasfollowed by five PCR cycles for nonspecific enrichment of full-length cDNAs. The cDNA fragments were then ligated to the pDNR-LIB vector. When the plasmid clones were used to transformEscherichia coli JM109, the resulting library contained 2.0×106

independent clones.Clones (5×104) from an unamplified cDNA library were plated on

LB agar. The bacterial colonies were transferred onto Hybond-NXmembranes (GE Healthcare Bio-Science) and fixed by UV irradiation.The resulting replica membranes were prehybridized in Church'shybridization solution at 65 °C for 30 min and then hybridizedovernight at 50 °C with either radiolabeled ssp-4 or ssp-5 cDNA inChurch's hybridization solution, respectively. The membranes werewashed twice for 15 min at 50 °C with 1× SSC containing 0.1% SDS,and the hybridization signals were visualized using a BioImageAnalyzer (Fuji Film). Finally, 30 bacterial colonies were isolated andcultured. Their plasmids were purified by the standard alkali-SDSmethod. The nucleotide sequences of the cDNA inserts weredetermined using the ABI PRISM 377 DNA Sequencing System(Applied Biosystems).

2.7. Comparison and phylogenetic analysis

The nucleotide sequences were aligned and compared usingDNASIS (Hitachi software engineering). Phylogenetic analysis wascarried out by using CLUSTALW on the DNA Database Japan (DDBJ)web page and MEGA4 (Tamura et al., 2007); phylogenetic trees weredrawn using TreeView (http://taxonomy.zoology.gla.ac.uk/rod/rod.html). Detection of positive selection along phylogenetic lineageswas performed using GA-Branch, a codon-based genetic algorithmavailable on the Datamonkey web server (http://www.datamonkey.org; Kosakovsky Pond and Frost, 2005).

Previously reported cDNA and genomic sequences encoding PSP94family proteins were obtained fromGenBank and used in comparativeanalyses: human (accession No. S67815), rhesus monkey (M92161),cotton-top tamarin (AJ010154, AJ010158, and AJ010156 for mspE1,mspA1, and mspJ1, respectively), pig (NM_213852), rat (U65486),mouse (J89840), chicken (XM_421645), African clawed frog(AW641318), zebrafish (XM_001332586), Japanese flounder(C23089), cow (XM_867982), and horse (XM_001493992).

9N. Aoki et al. / Gene 426 (2008) 7–14

3. Results

3.1. Purification of SSP-4 and SSP-5 from T. flavoviridis serum

In a previous report, we described the separation of SSP-1, SSP-2,and SSP-3 from T. flavoviridis serum by using reverse-phase HPLC(Aoki et al., 2007). In that study, we reported that when a largeamount of serumwas fractionated, some minor peaks of new proteinscould be detected around the SSP peaks. In order to characterize theseproteins, high molecular-mass proteins were removed by gel filtrationon a Sephacryl S-200HR column (data not shown). The SSP-containingfractions were then subjected to reverse-phase HPLC on a C8 column(Fig. 1A). SSP-1, SSP-2, and SSP-3 were eluted in peaks 1, 2, and 3,respectively. Two novel proteins were obtained in peaks 4 and 5. Themolecular weights estimated by SDS-PAGE were 8000 and 12,000,respectively, as shown in the inset of Fig. 1A. The molecular masses ofthe proteins in peaks 4 and 5were determined by mass spectrometricanalysis to be 7008.2 and 10,247.9 Da, respectively. These proteins alsoshowed significant sequence similarities to other SSPs (Fig. 1B). Theproteins in peaks 4 and 5 were therefore designated SSP-4 and SSP-5,respectively.

3.2. Cloning and analysis of cDNAs encoding SSPs

In addition to the cDNA clones of SSP-1, SSP-2, and SSP-3 analyzedpreviously (Aoki et al., 2007), those of SSP-4 and SSP-5 wereadditionally isolated and analyzed in this study. Partial cDNA fragmentswere obtained for SSP-4 and SSP-5 by 3′-RACE PCR of T. flavoviridismRNAusing the ssp-4N and ssp-5Noligonucleotides thatwere designedon the basis of the N-terminal amino acid sequences of the

Fig. 1. Purification of SSP-4 and SSP-5. (A) Reverse-phase HPLC of the SSP-containingfraction obtained from T. flavoviridis serum on a Cosmosil-5C8-AR-300 column(1.0×25 cm). The broken line shows acetonitrile (%) in 0.1% TFA. SDS-PAGE of peaks 4and 5 is shown in the inset. M represents the molecular weight markers. (B) The N-terminal sequences of the proteins in peaks 4 and 5.

corresponding SSPs (Fig. 1B). The DNA products (approximately550 bp) that were amplified from the respective reactionscorresponded to SSP-4 and SSP-5 and were employed as probes inthe hybridization experiments. When the SSP-4 cDNA fragment wasused for screening 5.0×104 clones of the T. flavoviridis liver cDNAlibrary, three full-length cDNAcloneswere isolated. Similarly,five cDNAclones were isolated when the SSP-5 cDNA fragment was used. ThecDNAs for SSPs are designated by lowercase letters, e.g., ssp-4 and ssp-5.No cDNA clone encoding additional SSPs was identified despiterepeated screening experiments on the basis of the sequence similaritybetween SSP isoforms; this suggests that the PSP94 family of proteins inT. flavoviridis is likely represented by a total of five members that areexpressed in the liver.

The complete nucleotide sequences of ssp-4 and ssp-5 weredetermined, and these are shown together with those of ssp-1, ssp-2,and ssp-3 in Fig. 2A. The open reading frame (ORF) search andcomparison with the amino acid sequences of SSPs revealed that thededuced amino acid sequences were completely identical to those ofthe purified SSPs. The 19-amino acid stretch at the N-terminal of alldeduced sequences was identified as the signal peptide (Fig. 2A).Comparison of their structural organization of the SSPs revealed thatSSP cDNAs could be classified into two groups (Fig. 2B). The ORFs ofssp-3 and ssp-4 were shorter than those of ssp-1, ssp-2, and ssp-5due to an early stop codon at positions 301–303, resulting intruncation of proteins at their C-terminus relative to SSP-1, SSP-2,and SSP-5.

3.3. Comparison of the amino acid sequences of PSP94 family proteins

The amino acid sequences of SSP-4 and SSP-5 deduced from thenucleotide sequences, together with those of mammalian PSP94proteins and other SSPs, are shown in Fig. 3. Human and porcinePSP94s are composed of two discrete domains (Ghasriani et al.,2006). The corresponding domains in SSPs were deduced on thebasis of the domain structure of human PSP94, i.e., the N-terminal(residues 1–52) and C-terminal (residues 55–94) domains. Thecorresponding N-terminal (residues 1–57) and C-terminal (residues58–90) domains are shown in Fig. 3. Although the disulfide bridgesin SSPs were estimated from the data of porcine PSP94 (Wang et al.,2003), two S–S bonds, between Cys22 and Cys47 and between Cys69

and Cys92, were determined by enzymatic digestion of intact SSP-1(data not shown).

SSP-3 and SSP-4 are quite unique because they lack the C-terminaldomain, and are composed of only 62 amino acids. The C-terminal-truncated SSP (SSP-3 or SSP-4) is designated the “short-chain SSP” asopposed to the full-length protein. Comparison of SSP-1 and SSP-2with the mammalian PSP94 family proteins suggests that 10 cysteineresidues in the mature proteins are well conserved, irrespective of thelimited sequence similarities. Similar to SSP-2, SSP-5 is composed of90 amino acids while it has only 8 cysteine residues because Cys69 andCys92 are substituted with Tyr69 and Ser92, respectively. Short-chainSSPs have only 6 cysteines because of an additional substitution ofCys42 with Ser (Fig. 3).

3.4. Statistical analysis of the cDNA sequence encoding SSPs

The percent identities of the nucleotide sequences in the matureprotein-coding regions and the amino acid sequences of five SSPs areshown in Table 1. The average identity for all combinations of thenucleotide sequences was more than 65% with particularly highidentities in the case of ssp-1/ssp-2 (86.5%) and ssp-3/ssp-4 (94.1%). Incontrast, the identities of amino acid sequences were lower than thoseof the nucleotide sequences (lower by 24.3% on average). This impliesthat the encoded amino acid sequences are more divergent thanwould be predicted based on the degree of nucleotide divergenceunder neutral patterns of evolution.

Fig. 2. (A) Nucleotide sequences of cDNA clones encoding SSPs, and (B) schematic presentation of the organization of cDNAs. The protein-coding region is shown in capital letters, andthe 5′- and 3′-UTRs are in lowercase letters. The initial and stop codons are boxed, and the polyadenylation signals are underlined. Conserved residues are indicated by asterisks underthe sequences. The signal peptide-coding region is shown by an arrow. Codons of Cys residues are highlighted in white-on-black type. Codons coding conserved amino acids areshadowed in grey.

10 N. Aoki et al. / Gene 426 (2008) 7–14

Another remarkable feature was that the 5′- and 3′-UTRs, and thesignal peptide-coding regions, were highly conserved among thecDNAs, while nucleotide substitutions were found predominantly inthe mature protein-coding regions (Fig. 2A). This is in contrast to thegeneral tendency observed in most genes where the protein-codingregions diverge more slowly than the UTRs due to purifying selectionto conserve protein function.

To further characterize the molecular evolution of these SSPs,both their cDNA nucleotide sequences and the inferred amino acidsequences were used to estimate the phylogeny and evolutionarydistances among cDNAs, via the neighbor-joining method in MEGA4

(Tamura et al., 2007). Comparisons of the evolutionary distances ofprotein-coding regions based on synonymous nucleotide substitu-tions versus amino acid changes suggests a lack of notable purifyingselection and possible evidence of positive or adaptive selection foramino acid change (Table 1). Unlike most proteins in whichsynonymous substitutions may outnumber amino acid changesseveral-fold, these snake SSPs exhibit nearly equal or even highernumbers of amino acid changes than synonymous changesconsistent positive selection. This is especially true in the case ofevolutionary distance between SSP-1 and SSP-2 where amino acidchanges appear to greatly outweigh synonymous changes (Table 1),

Fig. 3. Sequence alignment of SSPs. Human and porcine PSP94 are included for comparison. Cysteine residues are shown in white-on-gray type. Asterisks and dots under sequencesindicate invariant residues in all SSPs and the conserved residues in all molecules, respectively. Regions of secondary structure predicted by homology modeling using an automatedprotein modeling server Swiss-Model are shown with strands depicted by broad arrows. Disulfide bridges are indicated by lines.

11N. Aoki et al. / Gene 426 (2008) 7–14

and that nucleotide changes seem to have preferentially targetednonsynonymous rather than synonymous sites (Fig. 4). Intriguingly,such nucleotide substitutions have not been scattered over themature protein-coding sequence, but are found in several limitedregions (Fig. 4).

A phylogenetic tree constructed using nucleotide sequences ofPSP94 family proteins from diverse vertebrate species together withsnake SSPs (using Kimura's two-parameter distances [Saitou andNei, 1987]) is shown in Fig. 5. As expected, the branches nearlycorrespond to the current taxonomic classes, although in somespecies, further diversity has been apparently generated throughgene duplications (e.g., snakes and tamarins). Comparing the PSP94family members between the cotton-top tamarin and the snakerevealed that there is much greater intraspecies variation in thesnake SSPs than in the three tamarin genes (Fig. 5). The diversity ofthe snake SSP genes exceeded the diversity among species indifferent organismal classes in some cases, highlighting thesubstantial and abnormal diversity of snake SSPs.

To further confirm that the patterns of relatively rapid proteinevolution of snake SSPs represented positive selection, we used theprogram GA-Branch (Kosakovsky Pond and Frost, 2005) to detectbranches of the SSP tree that have apparently experienced positiveselection on amino acid sequences. The results of this analysis foraccelerated evolution (positive values of dN/dS) are in good agree-ment with the result obtained by comparing amino acid versusnucleotide distances; the results of GA branch analysis suggest thatalmost all the branches in the snake SSP tree have experiencedpositive selection (with dN/dSN1), except for the divergence of SSP4and SSP5 (Fig. 6).

4. Discussion

4.1. Structures of SSPs and comparison with mammalian PSP94family proteins

PSP94 family proteins are small, hydrophilic, nonglycosylatedproteins with a rigid tertiary structure with five disulfide bonds(Dube et al., 1987). Although many animals have a single

Table 1Sequence identities among five SSPs isolated from T. flavoviridis serum

ssp-1 ssp-2 Ssp-3 ssp-4 ssp-5

ssp-1 86.5 69.6 69.6 66.5 DNAssp-2 68.2 70.0 69.9 71.4ssp-3 45.0 45.0 94.1 70.3ssp-4 41.9 38.7 80.0 70.0ssp-5 44.3 50.0 41.7 40.3

Protein

Their amino acid sequences and mature protein-encoding nucleotide sequences werecompared respectively, and the percent identities are cited.

representative of the PSP94 gene family, snakes appear to haveseveral. Previously, we isolated three PSP94 family proteins (namedSSPs) from the serum of the venomous snake T. flavoviridis (Aokiet al., 2007). In this study, we have purified two additional similarproteins, SSP-4 and SSP-5, as minor components from the sameserum and cloned the cDNAs encoding all five SSPs. We have alsofound that another Japanese viper Gloydius blomhoffii also has fiveSSPs (unpublished data).

The primary structures of SSP-3 and SSP-4 were quite unique sincethese proteins were composed of only 62 amino acids and lacked thetypical PSP94 C-terminal domain (Fig. 3). Production of these short-chain SSPs could be explained by the change in the 63rd codon to atermination codon (TAA) in both SSP-3 and SSP-4 (Fig. 2), since theirnucleotide sequences are conserved to some extent even behind thetermination codons, especially at the positions corresponding to Cyscodons of full-length SSPs. A protein similar to PSP94, PSP57, consistsof 57 amino acids and is similar in size to SSP-3 and SSP-4 is knownfrom humans (Xuan et al., 1995). The nucleotide sequence for PSP57,however, demonstrates that it is produced by alternative splicing ofthe PSP94 gene as a result of skipping of exon 3 and this is not the casefor SSP-3 and SSP-4.

SSP-1 and SSP-2 are full-length proteins composed of 88 to 90amino acids, including 10 cysteine residues. Two different clones (ssp-1a and ssp-1b) were obtained from the cDNAs encoding SSP-1 (datanot shown). They differed only with regard to the presence of twoamino acids, i.e., Arg29-Lys30 in SSP-1a and Ser29-Arg30 in SSP-1b. SSP-5 also had 90 residues but a disulfide bridge in the C-terminal domainwas missing, i.e., Cys69 and Cys92 were replaced by Tyr69 and Ser92,respectively (Fig. 3).

4.2. Accelerated evolution and diversification of snake SSP proteins

Mammalian PSP94 family proteins show unique patterns ofmolecular evolution. Accumulation of nucleotide substitutionsoccurs mainly in the protein-coding regions, indicating that theseproteins have evolved at abnormally rapid rates (Mäkinen et al.,1999; Nolet et al., 1991; Fernlund et al., 1996). Alignment of thenucleotide sequences of the five SSP genes has revealed that, inspite of the high conservation of sequences in the 5′-UTR andsignal peptide-coding regions, the accumulation of nucleotidesubstitutions occurred mainly in the mature protein-coding region(Fig. 2). This is contrary to the case observed in most genes inwhich nucleotide substitutions in the protein-coding region areconsiderably suppressed because they can modify or hinder thefunction of the resultant protein. Two possible explanations mayaccount for such an abnormality in SSP gene evolution: suppressedsubstitutions at the UTRs and signal peptide-coding region oracceleration of mutations in the mature protein-coding region. Incrotaline snakes, including T. flavoviridis, similar characteristics havealso been observed in several gene families of venom enzymes,including phospholipase A2 (PLA2) isozymes, serine proteases, and

Fig. 4. Comparison of the ORFs of SSP-1 and SSP-2. Gaps are introduced to maximize alignment and are indicated by hyphens. The initial codons and stop codons are shown in boldface. Non-identical nucleotides are highlighted in grey, and non-identical amino acids are shown in reversed type.

12 N. Aoki et al. / Gene 426 (2008) 7–14

MPs (Ogawa et al., 1995; Ohno et al., 1998; Chijiwa et al., 2003). Inthese cases, rapid accumulation of nucleotide substitutions in theprotein-coding sequences has been proposed since the nucleotide

Fig. 5. The phylogenetic tree for the protein-coding regions of snake SSPs and PSP94 familyindicated with a box. The tree was constructed using the neighbor-joining method based o

substitution rates in the noncoding regions of the genes in questionwere proven to be similar to those for normally-evolved TATA box-binding proteins in Trimeresurus species (Nakashima et al., 1993).

proteins of other vertebrates. The cluster of branches containing the five snake SSPs isn nucleotide sequences. The scale bar represents 0.05 substitutions per site.

Fig. 6. Positive selection along the evolutionary pathway of SSPs. Protein-codingsequences of the cDNAs for SSPs were analyzed using GA-Branch. The phylogenetic treewas constructed by neighbor-joining method and rooted by including the cDNAsequence for zebrafish PSP94 family protein as the outgroup. Branches inwhich positiveselection (dN/dS value of 3.490 on average) and negative selection (0.507 on average)were detected are shown by solid lines and dotted lines, respectively. The scale barrepresents 0.1 substitutions per site. Numbers on the branches show the percentprobabilities of positive selection.

13N. Aoki et al. / Gene 426 (2008) 7–14

On the other hand, it was found that in every pair of SSPs, thedivergence of the amino acid sequences was substantially lowerthan the divergence of the underlying nucleotide sequences of thegene, suggesting that substitutions had targeted nonsynonymousnucleotide positions preferentially, consistent with positive selection(Table 1; Miyata et al., 1980; Nei, 1987). This result was furtherconfirmed by statistical analysis of the relative rate of nonsynon-ymous to synonymous rates (dN/dS) in snake SSPs, which suggestedthat almost all evolutionary branches in the snake SSP tree hadexperienced positive selection with dN/dSN1. These results suggestthat relatively strong selection has driven the diversification ofsnake SSPs following multiple duplications of this gene family insnakes. Furthermore, it is possible that SSPs have diversified alsotheir physiological functions by modifying the structures responsiblefor their respective activities via accelerated evolution.

Although there have been several reports on the acceleratedevolution of snake venom proteins (Ohno et al., 2002; Soto et al.,2006), this is the first study in which the abnormal evolution ofsnake serum protein has been described. Snake venom enzymesmay have gained diversified functions that are advantageous forfeeding and defense, and it is likely that the SSP isoforms haveevolved cooperatively to establish a self-defense system againstnewly acquired venom components such as hemorrhagic MPs,myotoxic PLA2-like proteins, and some ion-channel blockers (Peraleset al., 2005).

4.3. Correlation between the structure and function of SSPs

The three SSPs have different physiological functions as reportedpreviously (Aoki et al., 2007). SSP-1 inhibited brevilysin H6, an MPisolated from G. halys brevicaudus venom. Very weak inhibitoryactivity was also observed in SSP-3. On the other hand, SSP-2exhibited potent affinity to triflin, a smooth muscle contractionblocker isolated from T. flavoviridis venom (Yamazaki et al., 2002). Thestoichiometry of SSP-2 and triflin was 1:1. A preliminary experimentshowed that SSP-5 also bound to triflin (data not shown). At present,we have no information regarding the properties of SSP-4.

It is known that the reactive sites of many proteinase inhibitors arelocated in the loop structure, often at the ends of β-strands (McBrideet al., 2002; Hamze et al., 2007). Because residues 22–55 are relativelywell conserved in all SSPs, it is possible that the reactive site is presentin the 20 N-terminal residues and in the loop formed by residues 48–51 in SSP-1 (Fig. 3). Homology modeling experiments have suggestedthat residues Phe48-Lys49-Gly50-Gly51 of SSP-1 are present in a loopstructure formed by a disulfide bond between Cys45 and Cys54 at theedge of the β-strand. They are also oriented toward some N-terminal

residues (data not shown). Thus, the area involving the Lys49-Gly50

sequence may participate in the interaction of SSP-1 with the MP.On the other hand, the regions that participate in the binding of

SSP-2 and SSP-5 to triflin may be similar to each other. The moststriking difference among the full-length SSPs is the stretch ofresidues from Tyr58 to Lys68 in which SSP-2 and SSP-5 showed highhomology to each other (Fig. 3). This region may correspond to thelong stretch connecting the N- and C-terminal domains as expected onthe basis of the 3-D structures of human and porcine PSP94s(Ghasriani et al., 2006). It has also been identified as being the corearea where extensive nonsynonymous nucleotide substitutions werefound (Fig. 4). Although there were 20 N-terminal residues that werehighly variable between SSP-2 and SSP-5, no similar sequence elementwas found. Therefore, some distinct structural elements in this stretchmight determine the binding specificity of SSPs to triflin.

4.4. Conclusion

There have been many studies that have investigated the proteincomponents of snake venoms, although the ways in which venomousanimals defend themselves against their own venoms remain almostunknown. Here we reported the discovery of novel small serumproteins (SSPs) that seem to be involved in self-defense of thevenomous snake, T. flavoviridis, against accidental self-envenomation.Future work to more completely identify the venom components thatare the molecular targets of different SSPs would strongly contributeto the elucidation of molecular mechanisms of self-defense and theapplication of such SSPs to drug design for a treatment for snakebite.Our results suggest that snake SSPs evolved through gene duplicationand accelerated protein evolution and positive selection for diversifiedprotein function; this pattern of molecular evolution parallels reportsof similar adaptive evolution in many protein components of snakevenoms. Elucidation of physiological functions of a series of SSPsmight provide insightful links for understanding the apparentsimultaneous acceleration of snake venom proteins and SSPs byallowing joint analysis of molecular evolution of SSPs and their precisetarget venom proteins.

References

Aoki, N., Sakiyama, A., Deshimaru, M., Terada, S., 2007. Identification of novel serumproteins in a Japanese viper: homologs of mammalian PSP94. Biochem. Biophys.Res. Commun. 359, 330–334.

Chang, L., Chung, C., Huang, H.B., Lin, S., 2001. Purification and characterization of achymotrypsin inhibitor from the venom of Ophiophagus hannah (King Cobra).Biochem. Biophys. Res. Commun. 283, 862–867.

Chijiwa, T., et al., 2003. Interisland evolution of Trimeresurus flavoviridis venomphospholipase A2 isozymes. J. Mol. Evol. 56, 286–293.

Chuman, Y., et al., 2000. Regional and accelerated molecular evolution in group I snakevenom gland phospholipase A2 isozymes. Toxicon 38, 449–462.

Dube, J.Y., et al., 1987. Isolation from human seminal plasma of an abundant 16-kDaprotein originating from the prostate, its identification with a 94-residue peptideoriginally described as β-inhibin. Andrology 8, 182–189.

Fernlund, P., Granberg, L.B., Roepstor, P., 1994. Amino acid sequence of β-micro-seminoprotein from porcine seminal plasma. Arch. Biochem. Biophys. 309, 70–76.

Fernlund, P., Granberg, L.B., Larsson, I., 1996. Cloning of β-microseminoprotein of therat: a rapidly evolving mucosal surface protein. Arch. Biochem. Biophys. 334, 73–82.

Friedman, M., Krull, L.H., Cavins, J.F., 1970. The chromatographic determination ofcystine and cysteine residues in proteins as S-β-(4-pyridylethyl)cysteine. J. Biol.Chem. 245, 3868–3871.

Fujimura, S., Oshikawa, K., Terada, S., Kimoto, E., 2000. Primary structure andautoproteolysis of brevilysin H6 from the venom of Gloydius halys brevicaudus.J. Biochem. 128, 167–173.

Ghasriani, H., Teilum, K., Johnsson, Y., Fernlund, P., Drakenberg, T., 2006. Solutionstructures of human and porcine β-microseminoprotein. J. Mol. Biol. 362, 502–515.

Hamze, A.B., Wei, S., Bahudhanapati, H., Kota, S., Acharya, K.R., Brew, K., 2007.Constraining specificity in the N-domain of tissue inhibitor of metalloproteinases-1; gelatinase-selective inhibitors. Protein Sci. 16, 1905–1913.

Kosakovsky Pond, S.L., Frost, S.D.W., 2005. A genetic algorithm approach to detectinglineage-specific variation in selection pressure. Mol. Biol. Evol. 22, 478–485.

Lazure, C., Villemure, M., Gauthier, D., Naude, R.J., Mbikay, M., 2001. Characterization ofostrich (Struthio camelus) β-microseminoprotein (MSP): identification of homo-logous sequences in EST databases and analysis of their evolution during speciation.Protein Sci. 10, 2207–2218.

14 N. Aoki et al. / Gene 426 (2008) 7–14

Lilja, H., Abrahamsson, P.A., 1988. Three predominant proteins secreted by the humanprostate gland. Prostate 12, 29–38.

Mäkinen, M., Valtonen-Andre, C., Lundwall, A., 1999. New world, but not Old World,monkeys carry several genes encoding β-microseminoprotein. Eur. J. Biochem. 264,407–414.

McBride, J.D., Watson, E.M., Brauer, A.B., Jaulent, A.M., Leatherbarrow, R.J., 2002. Peptidemimics of the Bowman–Birk inhibitor reactive site loop. Biopolymers 66, 79–92.

Miyata, T., Yasunaga, T., Nishida, T., 1980. Nucleotide sequence divergence and functionalconstraint in mRNA evolution. Proc. Natl. Acad. Sci. U. S. A. 77, 7328–7332.

Nakashima, K., et al., 1993. Accelerated evolution of Trimeresurusavoviridis venom glandphospholipase A2 isozymes. Proc. Natl. Acad. Sci. U. S. A. 90, 5964–5968.

Nakashima, K., et al., 1995. Accelerated evolution in the protein-coding regions isuniversal in crotalinae snake venom gland phospholipase A2 isozyme genes. Proc.Natl. Acad. Sci. U. S. A. 92, 5605–5609.

Nei, M., 1987. Molecular Evolutionary Genetics. Columbia Univ. Press, New York.Nolet, S., St-Louis, D., Mbikay, M., Chretien, M., 1991. Rapid evolution of prostatic protein

PSP94 suggested by sequence divergence between rhesus monkey and humancDNAs. Genomics. 9, 775–777.

Ogawa, T., Kitajima, M., Nakashima, K., Sakaki, Y., Ohno, M., 1995. Molecular evolution ofgroup II phospholipases A2. J. Mol. Evol. 41, 867–877.

Ohno, M., et al., 1998. Molecular evolution of snake toxins: is the functional diversity ofsnake toxins associated with a mechanism of accelerated evolution? Prog. NucleicAcid Res. Mol. Biol. 59, 307–364.

Ohno, M., Ogawa, T., Oda-Ueda, N., Chijiwa, T., Hattori, S., 2002. Accelerated and regionalevolution of snake venom gland isozymes. In: Ménez, A (Ed.), Perspectives inMolecular Toxinology. John Wiley & Sons, New York, pp. 387–400.

Perales, J., Neves-Ferreira, A.G., Valente, R.H., Domont, G.B., 2005. Natural inhibitors ofsnake venom hemorrhagic metalloproteinases. Toxicon 45, 1013–1020.

Saitou, N., Nei, M., 1987. The neighbor-joiningmethod: a newmethod for reconstructingphylogenetic trees. Mol. Biol. Evol. 4, 406–425.

Schagger, H., von Jagow, G., 1987. Tricine-sodium dodecyl sulfate-polyacrylamide gelelectrophoresis for the separation of proteins in the range from 1 to 100 kDa. Anal.Biochem. 166, 368–379.

Soto, J., et al., 2006. Genetic variation of a disintegrin gene found in the Americancopperhead snake (Agkistrodon contortrix). Gene 373, 1–7.

Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary geneticsanalysis (MEGA) software version 4.0. Mol. Bio. Evol. 24, 1596–1599.

Valente, R.H., Dragulev, B., Perales, J., Fox, J.W., Domont, G.B., 2001. BJ46a, a snake venommetalloproteinase inhibitor. Isolation, characterization, cloning and insights into itsmechanism of action. Eur. J. Biochem. 268, 3042–3052.

von der Kammer, H., Krauhs, E., Aumuller, G., Scheit, K.H., 1990. Characterization of amonoclonal antibody specific for prostatic secretory protein of 94 amino acids(PSP94) and development of a two-site binding enzyme immunoassay for PSP94.Clin. Chim. Acta 187, 207–219.

Wang, I., Yu, T.A., Wu, S.H., Chang, W.C., Chen, C., 2003. Disulfide pairings and secondarystructure of porcine β-microseminoprotein. FEBS Lett. 541, 80–84.

Xuan, J.W., Chin, J.L., Guo, Y., Chambers, A.F., Finkelman, M.A., Clarke, M.W., 1995.Alternative splicing of PSP94 (prostatic secretory protein of 94 amino acids) mRNAin prostate tissue. Oncogene 11, 1041–1047.

Xuan, J.W., et al., 1999. cDNA, genomic cloning, and gene expression analysis of mousePSP94 (prostate secretory protein of 94 amino acids). DNA Cell Biol. 18, 26.

Yamakawa, Y., Omori-Satoh, T., 1992. Primary structure of the antihemorrhagic factor inserum of the Japanese Habu: a snake venom metalloproteinase inhibitor with adouble headed cystatin domain. J. Biochem. 112, 583–589.

Yamazaki, Y., et al., 2002. Cloning and characterization of novel snake venom proteinsthat block smooth muscle contraction. Eur. J. Biochem. 269, 2708–2715.