The aspartic proteinase family of three Phytophthora species

Archetypal pepsin-like Aspartic Proteinases (MEROPS subfamily A1A)

In clade 1 of the phylogenetic tree (Figure 1), three proteins designated PiAP1, PsAP1 and PrAP1, are encoded by orthologous single exon genes (Table 2). The PxAP1 polypeptides (Table 3) display all of the essential components i.e. signal peptide, propart segment (Additional File 4) and two-domain mature enzyme region (Figure 2), including all the necessary hallmark motifs and residues of archetypal pepsin-like APs described earlier (Additional File 1). These polypeptides are not extended at their C-terminus. The phylogenetic analysis revealed that the three PxAP1 polypeptides cluster closely with fungal APs (Figure 1) that are known to be located in the vacuole of the cell. This subfamily, with saccharopepsin from S. cerevisiae as the parent member, is assigned the identifier A01.018 in the MEROPS database. Analyses of all sequenced fungal genomes has revealed that the only AP that is omnipresent is the vacuolar enzyme [25]. Thus, it seems likely that PiAP1, PsAP1 and PrAP1 are the respective vacuolar APs from each Phytophthora species. PiAP1 has been produced in recombinant form and shown to have AP activity on haemoglobin (F. Govers, Wageningen University, personal communication).

The polypeptides of the PxAP2 in clade 2 (Figure 1) are atypical in that they do not contain a signal peptide or a non-classical secretion signal (Table 3 and Additional File 5). This is highly unusual in APs although precedents have been observed in fungi including Botrytis cinerea[25] and in bacterial species of marine origin [26]. It is not obvious how these polypeptides are able to become incorporated into a membrane through the (single) hydrophobic segment in their C-terminal extensions (Additional File 6) when there is no signal peptide to mediate transfer into the endoplasmic reticulum. Experiments using fusion constructs with reporter proteins will be necessary to resolve this conundrum.

Within clade 3, five clusters (PxAP3 - PxAP7) were resolved from one another (Figure 1). More profound phylogenetic analysis (not shown) indicates that this clade forms a monophyletic clade with three entries from two filamentous fungi, i.e. the zygomycetes Rhizopus oryzae and Phycomyces blakesleeanus. PxAP3-encoding genes were only detected in P. sojae and P. ramorum (gene models Ps_157552 and Pr_77872; Table 1, Table 2). Despite repeated searching, an orthologue of these two in P. infestans could not be detected. The PxAP3 genes are located in regions of co-linearity. Scrutiny of the five flanking genes (upstream and downstream) in the P. sojae and P. ramorum genomes revealed that orthologues are present in similar arrangement in P. infestans DNA, however, on separate contigs. The gap between the contigs corresponds to the likely location of the "missing" AP gene. Moreover, no unassembled reads were identified in P. infestans that matched PsAP3 or PrAP3 sequences.

The originally-annotated gene models for Ps_157552 and Pr_77872 both required considerable correction (Additional Files 2 and 4). The 5'-splice donor site of the third intron in Ps_157552 was identified to be a GC dinucleotide sequence. This dinucleotide has been found to be used as the 5'-donor junction at low frequency (~ 1%) in a number of fungi [27] but no information is available as yet on its use in oomycetes. The amendments to the Ps_157552 and Pr_77872 gene models (Additional Files 2 and 4) resulted in predicted polypeptide sequences that contained both signal peptide and propart segments (Table 3; Additional File 5). The sequences of the mature enzyme regions (Figure 3) showed a pairwise identity of 71%. Both polypeptides contained all three of the disulphide bonds that are common in vertebrate APs and they were only slightly extended at their C-terminal end relative to pepsin (Additional File 6). PsAP3 and PrAP3 can be predicted to display different substrate specificities. The region between residues 292 and 298 is known conventionally as the "polyproline loop" and has been demonstrated to be a major determinant for substrate specificity [22]. In PrAP3, the sequence LGDDLYW of this region has two extra (LY) residues compared to the LSDD--W sequence of PsAP3 (dashed box in Figure 3). The presence of these two hydrophobic residues is likely to influence specificity.

Like PxAP3, the PxAP4 and PxAP5 polypeptides from Phytophthora contained the conventional AP segments of a signal peptide and propart (Table 3; Additional File 5), prior to the mature enzyme region (Figure 3). Each trio of polypeptides was also extended at its C-terminus but, whereas the extensions in PxAP4 consisted of 136-138 residues with one putative membrane-spanning stretch, the PxAP5 extensions were substantially longer (Additional File 6). Indeed, the C-terminal extensions in PiAP5 (322 residues) and its orthologues encompassed no less than seven potential transmembrane segments (Additional File 6), an architecture highly reminiscent of the seven transmembrane stretches that constitute the super-family of G-protein coupled receptors (GPCR) [28]. Blast search using the sequence of the C-terminal extension of PiAP5 as query, identified the first non-self hit (with an E-value of 0.001) as RpkA, a G-protein coupled receptor family protein from Dictyostelium discoideum[29]. A 55.5% similarity in a 265 amino acid overlap was calculated between the two sequences using Smith-Waterman analysis [30]. The intriguing possibility thus arises that, in Phytophthora species, a fusion protein is produced in which the N-terminal partner is an AP connected to a putative GPCR as the C-terminal component. Self-processing to release the AP component would generate the mature GPCR. Such a self-processing phenomenon is unprecedented, although receptors activated by extrinsic proteinases are well-documented [31]. Further investigations into the Phytophthora AP-GPCR fusion-proteins would appear to be well-merited.

The final two sub-groups that were resolved within clade 3 consisted of PxAP6 and PxAP7 (Figure 1). No orthologue was detected in P. sojae for the proteins PiAP7 and PrAP7, encoded by PITG_05004 and Pr_73546 (Table 1; Table 2). The PiAP7- and PrAP7-encoding genes are located in co-linear orthologous regions of the P. infestans and P. ramorum genomes, as confirmed by the organization of flanking genes. In P. sojae, the corresponding genomic region had a similar co-linearity arrangement with the exception of two missing genes, one of them in the region where an orthologous AP-encoding gene should have been located. Moreover, no unassembled reads were identified in P. sojae that matched the PiAP7 or PrAP7 sequences. The simplest interpretation is that an ancestor of the PxAP7-encoding gene was present in Phytophthora, and deletion of the region resulted in loss of the orthologue in P. sojae, although it cannot be excluded that substitutions have accumulated to an extent that prevents recognition of the gene.

The PxAP6 and PxAP7 polypeptides showed an unusual variation in the component segments that constitute an archetypal AP in that a propart was absent (Table 3). In each case, the signal peptide (Additional File 5) immediately precedes the mature enzyme region (Figure 3) which, in turn, is followed by a C-terminal extension (Additional File 6). As stated above, almost all pepsin-like APs are synthesised in the form of a precursor. The propart segment makes a number of essential contributions, such as ensuring proper folding and intracellular sorting of the zymogen polypeptide, and facilitation of its activation into the mature enzyme form upon encountering the appropriate conditions [11, 32]. Very recently, however, APs lacking a propart segment have been described from marine bacteria [26] and fungi [25], hence this unusual architecture of the PxAP6 and PxAP7 polypeptides in Phytophthora species is not without precedent.

BACE-like Aspartic Proteinases

The PxAP8 and PxAP9 polypeptides were resolved by the phylogenetic analysis into clade 4 (Figure 1). These polypeptides also contained a signal peptide (Additional File 5), the mature enzyme region (Figure 4) and a C-terminal extension (Additional File 6) but they too lacked a propart (Table 3 and Additional File 5), just like PxAP6 and PxAP7.

The phylogenetic analysis (Figure 1) revealed that these proteins cluster within clade 4 with human BACE and its paralogue, BACE2 (sub-families A01.004/041 in the MEROPS database). The mature enzyme forms of BACE/BACE2 are initially synthesised in precursor forms that do contain propart segments. While the presence of the propart has little effect on the intrinsic proteolytic activity, its inclusion at the N-terminus of proBACE/proBACE2 ensures much more rapid folding of the polypeptide(s) than is observed when only the mature forms of each enzyme are produced [[33, 34] and references therein].

BACE and BACE2 were among the first APs to be identified as having a C-terminal extension which ensured they were membrane-bound. The mature enzyme regions (Figure 4) have hallmark features which distinguish A01.004/041 subfamily members from classical APs such as pepsin. Among these are: (i) the highly-conserved Trp39 (see Figures 2 and 3) is replaced by Ala and is itself relocated to a position (residue 80) just downstream from the essential Tyr75 residue on the β-hairpin loop that overhangs the active site cleft (see Background), (ii) disulphide bonds in characteristic locations that increase stability of the polypeptide. These additional Cys residues and the Trp39Ala replacement are all present in PxAP8 and PxAP9 proteins (Figure 4).

BACE has been identified as the β-site Alzheimer's Precursor Protein (APP) Cleaving Enzyme (or β-secretase) responsible for the cleavage that generates the free N-terminal end of the Aβ peptide which plays a central role in Alzheimer's disease [35]. For almost a decade, BACE has been the target of intense efforts to develop inhibitors to alleviate the dementia of Alzheimer's disease. Genes encoding BACE and its paralogue, BACE2, have been identified in species ranging from humans through elephant, opossum, armadillo, hedgehog, shrew, and fish, to platypus (http://www.merops.ac.uk). To the best of the authors' knowledge, the identification of PxAP8-encoding genes from three Phytophthora species provides the first indication that BACE-like, membrane-bound APs may not be limited to the vertebrates. Intriguingly, the Phytophthora genomes also contain genes encoding homologues of presenilin, nicastrin, anterior pharynx defective-1(APH-1) and presenilin enhancer-2 (PEN-2) (although some gene models require substantial correction; not shown). These are all components of the human γ-secretase complex which completes the proteolytic processing by releasing the Aβ peptide at its C-terminal end from the Alzheimer's Precursor polypeptide [36]. The presence of both β- and γ-secretase gene products in microbes such as Phytophthora suggests that their function is not limited to the processing of a brain polypeptide related to Alzheimer's disease. These proteins therefore seem to be very ancient features that are conserved between oomycetes and vertebrates, and they have been coopted in vertebrate species to cleave the brain APP polypeptide.

While the PxAP8 polypeptides exhibited the hallmark characteristics of BACE/BACE2 subfamily members, the PxAP9 proteins lacked one feature that is necessary for an AP to be active as an enzyme. The Tyr residue located at position 75 (pig pepsin numbering) is conserved in every pepsin-like AP that is known to be active, although it does not participate directly in the catalytic mechanism [11, 21]. In PxAP9, this Tyr residue is replaced by Phe (Figure 4) so the catalytic activity can be postulated to be compromised, at best. A protein engineering study revealed that the replacement of Tyr75 in a fungal AP by Asn or Phe permitted activity, albeit at a considerably reduced level, while all other replacements abolished activity [37]. However, a number of catalytically-inactive APs have been reported previously, including the pregnancy-associated glycoproteins (PAGs) found in the placenta of ungulates (subfamily A01.971 in MEROPS) [38] and perhaps most notably, the Bla g2 protein (subfamily A01.950 in MEROPS) which is a predominant allergen from cockroach [39]. The crystal structure of Bla g2 shows that it has the fold and three-dimensional structure typical of pepsin-like APs [40]. Bla g2 and the PAGs from Ovis aries and Capra hircus all have a Phe residue at position 75. These three proteins have additional substitutions at the two active site DT/SG motifs, cementing their catalytic inactivity. Their function has been postulated to include roles in peptide-binding and presentation, but without cleavage. If PxAP9 should also prove to be catalytically ineffective, the active site will still be accessible for the binding of peptides that cannot undergo cleavage and so a role similar to the PAGs/Bla g2 in peptide binding might be feasible. This intriguing possibility is rendered all the more fascinating with the realisation that, in contrast to the PAGs/Bla g2, these Phytophthora orthologues are likely to be membrane-bound through their C-terminal extensions (Additional File 6) and might function as a novel type of surface receptor. In this regard, it may well be noteworthy that the putative intracellular domain comprised of the residues downstream from the transmembrane segment in the C-terminal extension in PxAP9 is considerably longer than its PxAP8 counterpart (Additional File 6). Neither of the C-terminal extensions, however, contains a discernible Pfam domain or protein motif. Thus, all three Phytophthora species may produce one catalytically-active BACE-like enzyme (PxAP8) and one peptide-binding, membrane-bound enzymatically inactive paralogue with an intracellular extension of unknown function (PxAP9).

Nepenthesin-like Aspartic Proteinases (MEROPS subfamily A1B)

The remaining three AP genes in each Phytophthora genome (Table 2), encoding polypeptides PxAP10, PxAP11 and PxAP12 cluster into clade 5 with representative APs from MEROPS subfamily A1B including the patriarch of plant origin, nepenthesin, and plasmepsin V from Plasmodium falciparum (Figure 1). The PxAP10 - 12 polypeptides contrast with their PxAP6 - 9 counterparts in having the archetypal AP architecture of a signal peptide followed by a propart (Additional File 5), a mature enzyme region (Figure 5), terminating with C-terminal extensions (Additional File 6). PxAP10 - 12 differed, however, from all of the other Phytophthora APs (PxAP1 - 9) in having a small extra segment inserted (dotted box in Figure 5) in the region just upstream from Tyr75 (in pig pepsin numbering). No less than seven Cys residues are present between residues 39 - 75, with a further three Cys residues (shaded in Figure 5) conserved at characteristic locations elsewhere in the mature enzyme regions of the nine Phytophthora APs in clade 5. Together, these Cys residues are likely to form five specific disulphide bridges, paired as indicated in Figure 5. This arrangement of disulphide bridges has been deduced from biochemical analyses of the APs of plant origin [41, 42] that cluster with PxAP10 - 12 in clade 5. Indeed, most of the APs in plants [43] contain these Cys residues as well as the insert and they are grouped into subfamily A1B in the MEROPS database.

The plant APs identified in clade 5 are readily distinguished on the basis of their functional activity from one another and from others that are not included here for the sake of brevity, such as the nucellins (subfamily A01.073), regulators of the degradation of nucellar cells after ovule fertilisation [44], and the Chloroplast Nucleoids DNA-binding (CND41) proteinase (subfamily A01.050) which catalyses the degradation of RUBISCO in senescent leaves [45]. Nepenthesins (A01.040 subfamily) are extracellular enzymes secreted by the carnivorous plant, Nepenthes, to digest its prey [41] while CDR1 of Arabidopsis plays a crucial role in activating resistance of Arabidopsis against microbial pathogens [14]. Overexpression of CDR1 leads to constitutive disease resistance, while gene silencing compromises resistance. The CDR1 gene product is suggested to release an as yet unidentified peptide that activates a signalling cascade to induce disease resistance in a salicylate-dependent manner [14]. Recombinant CDR1 protein has been produced in E. coli and shown to be active as an AP in a dimeric form [42]. Two intermolecular disulphide bonds were proposed to be responsible for the covalent association of the two polypeptide chains in the proteolytically-active homodimer [42]. The Cys residues involved in the homodimer formation are highlighted in Figure 5 by a star ( ). Cys residues are present at identical locations within PxAP10 (Figure 5), suggesting that this enzyme may also function as a homodimer. In contrast, PxAP11 is devoid of cysteine residues in this region and so, based on this criterion, should function as a monomer. PxAP12 does contain two cysteine residues but at different positions within this region of the polypeptide (Figure 5). Whether these remain as free cysteines or are involved in disulphide bonds, intra- or inter-molecular, requires experimental investigation.

One feature does serve to distinguish between the plant APs and the nine PxAP10 - 12 polypeptides clustered in clade 5 (Figure 1). Most plant APs do not have C-terminal extensions (with NP_199124 from Arabidopsis included in Figure 1 as one of few exceptions). In contrast, all nine of the oomycete PxAP10 - 12 polypeptides do have C-terminal extensions, ensuring that all are membrane-bound enzymes (Additional File 6).

In addition to plant APs, plasmepsin V was resolved into clade 5 (Figure 1) as the entry most closely associated with the PxAP10 - 12 polypeptides. Plasmepsin V is one of ten APs encoded within the genome of Plasmodium falciparum[16], the causative agent of malaria. This enzyme (Figure 5) has a C-terminal extension with a hydrophobic stretch (not shown) that results in its anchoring as a membrane-bound polypeptide [46]. Very recently, plasmepsin V has been demonstrated to be the (aspartic) proteinase responsible for processing the precursors of Plasmodium effector proteins at the PEXEL motif (RxLxE/Q); this step is essential for the translocation of these polypeptides into the host red blood cell [18, 19]. The malaria parasite exports hundreds of such effector proteins across the parasitophorous vacuolar membrane into the human red blood cell, thereby reprogramming the host cell to meet the needs of the parasite. Plasmepsin V is conserved in Plasmodium species but is not produced in vertebrates susceptible to parasitic infection and thus serves as a target enzyme against which to develop inhibitors as novel therapeutic agents for treatment of malaria.

Intriguingly, Phytophthora species also secrete effector proteins that are important in modulating host defence and cell death mechanisms and many such effector proteins contain the internal sequence motif RxLR, similar to the Plasmodium PEXEL motif [8, 47, 48]. Phytophthora effector proteins are secreted after removal of their signal peptide, giving rise to a polypeptide that contains in its N-terminal region the RxLR-containing domain which is essential for subsequent delivery into the plant cell [47–50]. It is tempting to consider that a Phytophthora effector precursor might undergo further processing to remove the RxLR domain, thereby releasing the mature, biologically active effector domain. The similarities described above between plasmepsin V and its PxAP10, PxAP11 and PxAP12 orthologues in Phytophthora suggests a conserved biological role. By analogy to the role of plasmepsin V in processing of Plasmodium effector proteins [18, 19], one or more of its Phytophthora orthologues may reside in the translocon and act in the processing of Phytophthora effector precursors. There are obvious structural and genetic differences between the secretion mechanisms of the two parasites. The genome sequences of Phytophthora species lack genes homologous to the translocon complex components identified in Plasmodium[51]. Nevertheless, our hypothesis invites experimental investigation into targeting these particular APs in Phytophthora as a pathogenic Achilles heel through which the devastating losses from infection by Phytophthora might be reduced.

Data sources and analyses

The PCR products were purified and subsequently sequenced (Macrogen, Amsterdam, The Netherlands). The corrected nucleotide sequences were deposited in NCBI (PITG_09387 = PiAP1: HM588685; PITG_08190 = PiAP8: HM588686).

Expression of the eleven P. infestans AP genes was determined from analysis of the data derived from the NimbleGen microarray (Accession Number GSE14480), as described previously [4].

Identification of Phytophthora APs by PSI-Blast and HMMer profile alignment

A first HMMer search [24] was performed with default cut-off values of 10 using a profile based on the reference holozyme sequence set as provided by MEROPS. A second HMMer search was done with a profile that we generated based on over 5000 AP sequences identified by NCBI Entrez and verified by MEROPS batch analysis [52]. PSI-Blast was performed, using cut-off values of 10, with the PiAP1, PiAP8 and PiAP10 as queries against the NR database restricted to the taxid Phytophthora (which includes the complete proteome sequences of P. infestans but not of P. sojae or P. ramorum) using a Position Specific Substitution Matrix obtained by converging PSI-blast against the Swissprot database. PSSM generation using NR or Refseq databases failed due to high contamination in the first Blast iteration using the Blosum62 substitution matrix.

Protein sequence analyses

Once the open reading frames of relevant gene entries were established, the resulting protein sequences were analysed in various ways. The mature enzyme region was determined by using the Pfam protein families database [53]. The SignalP server (http://www.cbs.dtu.dk/services/SignalP/) was used to predict the signal peptidase cleavage site and the absence of non-classical secretion signals in the PxAP2 sequences was verified using the SecretomeP server (http://www.cbs.dtu.dk/services/SecretomeP/). The cleavage site between propart and mature enzyme region was predicted from detailed inspection of the sequences. The prediction of transmembrane domains was performed using the web-tools ConPred II [54] or TMHMM Server (http://www.cbs.dtu.dk/services/TMHMM/) [55]. GPCR relationship was analysed by PRED-GPCR (http://athina.biol.uoa.gr/bioinformatics/PRED-GPCR/) [56].

Sequence alignments in Figures 2, 3, 4 and 5 were essentially prepared using T-Coffee [57] at the EBI server (http://www.ebi.ac.uk), and manually adjusted in a limited number of places to better align structural features that are hallmarks of APs.

Phylogeny

The alignments shown in Figures 2, 3, 4 and 5 were combined using T-Coffee [57] with manual corrections. Non-homologous N- and C-terminal sequences were removed. A maximum likelihood (ML) tree was made using PHYML [58]. Bootstrap analysis was performed by combining the "seqboot" and "consense" commands from PHYLIP [59] and the ML algorithm from PHYML. Trees were analyzed using Dendroscope [60]. Bootstrap analysis indicated poor resolution of the PxAP2 clade. In order to improve the resolution of this clade we included the sequence of human gastricsin which was the first non-self hit detected in a Blast analysis using PiAP2 as a query. Similarly, in order to improve the resolution between Plasmepsin V and the PxAP10-12 clade, the sequence XP_002180775 from Phaeodactylum tricornutum (from the sister phylum of diatoms) was included in the phylogeny.