Background

The serpins form a family of more than 1,000 proteins found in plants, animals and viruses, but only rarely in fungi, bacteria, or archaea [1]. The majority are active SER ine P roteinase IN hibitors [2] with a unique cleavage mechanism. The serpin-fold consists of three β-pleated sheets with 8 or 9 short α-helical linkers. In the native state, serpins are in a metastable (stressed) configuration, which consists of a core structure with an exposed reactive centre loop (RCL) of 25-30 amino acids. The RCL presents an ideal bait to the target proteinase which cleaves between two residues known as P1 and P1'. Following cleavage, a covalent complex is formed with the target proteinase. The cut end of the RCL inserts between two β-strands within the serpin core and the molecule undergoes an extreme conformational transition to the stable (relaxed) configuration. The covalently attached proteinase is denatured by being crushed against the serpin core, which targets the complex for proteolytic degradation [3]. Such a "suicide cleavage" mechanism gives rapid turnover of both inhibitor and proteinase, with the result that a reduction in inhibitor activity gives explosive activation of proteolytic cascades. The cleavage mechanism places particular constraints on the serpin-fold, which are common to all serpin inhibitors. In particular, the suicide inhibitory mechanism requires a run of small side-chain amino acids in the proximal "hinge" region [1, 4], seven residues N-terminal to the scissile bond, to allow RCL insertion into β-sheet A. This flexible hinge region is absent in the non-inhibitory serpin-fold proteins. Twelve of the 29 D. melanogaster serpin genes lack the critical short-side chain residues necessary for an active proteinase inhibitor [5, 6].

Serpins have been extensively studied in mammals where they regulate many extracellular proteolytic cascades. The coagulation, inflammatory and complement pathways are controlled by α1-Antithrombin, α1-Antitrypsin and C1-Inhibitor, respectively [2, 7, 8]; while Plasminogen Activator Inhibitor-1 modulates angiogenesis, affecting both wound-healing and tumour growth [9]. There is also a related group of non-inhibitory serpin-fold proteins with diverse functions, including molecular chaperones [10, 11], hormone transport [12], chromosome condensation [13], tumour suppression [14] and storage proteins [15]. In many cases, the functional requirement for the serpin-fold in the activity of these non-inhibitory molecules remains unclear.

In contrast to the mammals, insect serpins have been relatively little studied. In Drosophila, the Toll-mediated innate immune response is controlled by Spn43Ac (synonym, necrotic) [16, 17]. The phenol-oxidase cascade is inhibited, at two different steps, by Spn27A [18, 19] and Spn28Dc [20]; while tracheal melanization is regulated by Spn77Ba [21]. Spn27A also controls dorso-ventral axis formation in the embryo [22]. The alternatively spliced transcripts of the Spn42Da (synonym, Spn4) gene encode four separate RCL peptides. These protein isoforms inhibit furin-, chymotrypsin- and subtilase-like serine proteinases as well as papain-like cysteine proteinases [23–25]. Other than this, little is known about the Drosophila serpins, or the target proteinases that they inhibit. As there are 206 annotated chymotrypsin-fold serine proteinases in Drosophila, identification of target proteinases is difficult. It appears that individual serpins may inhibit proteinases in different proteolytic cascades; while particular physiological responses might be modulated by serpins targeting different steps in the same proteinase cascade.

Serpin specificity is determined by the amino acid sequence of the RCL, in particular the P1 and P1' sites, between which the peptide chain is cleaved [26]. The identity of the P1 residue is critical, with a lesser contribution of the P1' site and adjacent residues [27, 28]. In general, this mechanism gives broad specificity, with each serpin inhibiting a range of related serine proteinases. This apparent "redundancy" has complicated studies of mammalian serpins, as most serpin null mutations give no discernable mutant phenotype. The lack of function of one serpin activity may be compensated by the activity of a related serpin. Furthermore, in a proteolytic cascade, the lack of inhibition of one proteinase could be compensated for by inhibition of another proteinase, by a serpin with a differing inhibitory profile. Such complex interactions are hard to analyse and are likely to have unpredictable effects on the evolutionary constraints acting on different serpin activities.

The vertebrate serpins have evolved through gene duplication and divergence, giving rise to a large number of transcripts encoding serpins with unique reactive centre loops and discrete physiological functions. These serpins have been divided into 9 clades of related proteins, based largely on sequence homology [1]. The number of serpins varies widely between species, with the mouse genome containing 64, compared to 35 in the human. Fitting non-vertebrate serpins within this structure is difficult, although an expanded set of 16 clades, with a number of orphan sequences, has been proposed [1]. The D. melanogaster serpins have been classified according to their chromosomal location, with many of them located in clusters of adjacent transcripts. This nomenclature is neutral with respect to functional similarities, but clusters of adjacent transcripts are immediately apparent. The serpins that map within polytene chromosome band 43A, for example, are identified as Spn43Aa, Spn43Ab, Spn43Ac and Spn43Ad. Of the 29 D. melanogaster serpin transcripts, 17 are located within 6 gene clusters. This aspect of genomic organisation has been given relatively little attention in vertebrates, but the mammalian Clade A and B serpins tend to be present in clusters and it is the relative expansion of these clusters that accounts largely for the increase of serpin transcripts in the mouse lineage [29–32]. In general, the D. melanogaster clusters contain widely diverged transcripts including inhibitory serpins and non-inhibitory serpin folds (Figure 1).

Figure 1
figure 1

The D. melanogaster serpin protein minimum evolution tree. The serpin genes are distributed within the genome as a mixture of single transcripts and complex gene clusters. The complex loci tend to contain widely diverged transcripts, but Spn28Da/Spn28Db; Spn77Ba/Spn77Bb/Spn77Ba and Spn88Ea/Spn88Eb represent clusters of closely related transcripts. Inhibitory serpins are indicated with a solid black square, e.g.: Spn43Aa.

Full size image

Although the tertiary structure of the core serpin-fold is strongly conserved, the primary sequence constraints are relatively loose [1]. The major function of the serpin core is to deliver energy stored in its metastable, stressed configuration, following cleavage of the RCL [6]. Relaxation of the serpin core distorts the active site of the proteinase and traps it in an inactive serpin/proteinase complex. This is quite different to the mechanism of action of Kunitz and Kazal-family inhibitors, which form a highly specific lock and key fit with their target proteinase. The loose sequence conservation of the serpin core presents particular problems when trying to identify the complete set of serpins in an organism. An initial homology search of the Drosophilid species available through FlyBase identifies 188 serpin genes, but even in the sibling-species, D. simulans, orthologues were identified for only 23 of the 29 D. melanogaster serpins. As the relative genomic location of putative orthologues was often conserved, we decided to use synteny as a way to check for other family members: the genes flanking annotated D. melanogaster serpins were used to identify the corresponding regions in the other species which were then closely examined for nearby serpin-like genes. This approach identified an additional 35 serpins. A final tBLASTx search of all 12 genomes with all 223 genes identified three further serpin genes. By analysing the sequence conservation of the serpin genes across the Drosophilidae, we address both evolutionary and structure-functional relationships within this protein super-family.

Results and Discussion

The tBLASTx search of the Drosophilid sequence-assemblies available through FlyBase identified 188 serpin orthologues. This number represents 59% (188/319) of the orthologues expected on the assumption that the different Drosophilid species average 29 serpin genes per genome, as in D. melanogaster. In general, fewer orthologues were found in the out-groups, than in species close to melanogaster (13 serpins in the Hawaiian D. grimshawi, compared to 23 in D. simulans), Table 1. The synteny search identified all the serpin genes annotated on FlyBase, plus an additional 36 genes, Table 1 (gene identifiers are given in Additional file 1). The further tBLASTx search, using all the previously identified genes, identified three additional serpins (Table 1). These additional genes bring our coverage up to 71% of the expected orthologues (with the least coverage being 16 serpins in D. grimshawi, compared to 28 in D. simulans). Following this, we trained a hidden Markov Model against the 16 D. grimshawi serpin genes and searched the 12 Drosophilid genomes. This search identified 139/255 (55%) of our set of D. melanogaster serpins plus Drosophilid orthologues, but no additional genes. A final search for homology to the consensus RCL hinge sequence of Drosophilid inhibitory serpins,

Table 1 D. melanogaster serpins and Drosophilid orthologues
Full size table

E [EKR]G [TSG] [ET] [AGS] [YAGS] [AGS] [VAGS] [TS] (see Additional File 2), identified 75/183 (41%) of the previous inhibitory serpins, but again no novel serpin genes.

Three genes identified as orthologues of Spn42Dd and also of either Spn28F, or Spn38F were assigned as Spn42Dd orthologues by synteny analysis: these three genes are so closely related in D. melanogaster (Figure 1), that distinguishing between their orthologues by homology alone is inconclusive. The structure of the orthologous Spn42D gene clusters in the 12 Drosophilid species is shown in Figure 2A. The synteny search strategy is illustrated in Figure 2B. The genes flanking D. sechellia show homology to genes adjacent to Spn42Dd in D. melanogaster, rather than those neighbouring Spn38F. Thus, GLEANR 3714 is syntenic to Spn42Dd, not Spn38F. Similarly, synteny analysis indicates that D. erecta GLEANR 8071 and D. pseudoobscura GLEANR 15081 are orthologous to Spn42Dd not Spn28F.

Figure 2
figure 2

The structure of the Spn42D gene cluster. (A) The top line shows the 5 Spn42D serpin genes in D. melanogaster: Spn42Da (violet), Spn42Db (blue), Spn42Dc (cyan), Spn42Dd (green), Spn42De (yellow). These genes are flanked proximally by coro (black) and CG9447 (magenta) and distally by CG30158 (red). The orthologous genes in the other Drosophilid species are shown in successive lines. The D. willistoni Spn42 complex lacks a Spn42Dd orthologue and contains an insertion of 3 transcripts that are not syntenic with the other species, GK20759-GK20758 (red). Putative Spn42Db orthologues for D. mojavensis, virilis and grimshawi are given in Table 1, but omitted from this Figure 2 as their scaffolds are not yet located within the genome assembly. (B) The D. sechellia serpin GLEANR3714 was flagged by homology as a possible orthologue of both Spn42Dd and Spn38F. Syntenic analysis, however, shows that the adjacent transcripts from the Spn42D cluster, Spn42Dc and Spn42De, are plausible orthologues of the adjacent D. sechellia genes GLEANR3713 and GLEANR3715; while the adjacent genes to D. melanogaster Spn38F (CG9333 and CG31676) show no homology to the adjacent D. sechellia genes. By these criteria, GLEANR3714 is the orthologue of Spn42Dd, not Spn38F.

Full size image

The distribution of serpin orthologues across the Drosophilidae

Orthologues of many of the inhibitory serpins were identified in all Drosophilid species, but five (out of 17) are restricted to the melanogaster group (Spn28B, Spn28Da, Spn28Db, Spn28 F and Spn88Eb) and one (Spn38F) has no identified orthologues in other species. In the case of the non-inhibitory serpin-fold proteins, five have identified orthologues across the Drosophilidae (Spn31A, Spn43Ab, Spn47C, Spn85F, Spn100A and Spn28B), while the other seven show a more restricted distribution (Spn28Db, Spn43Ad, Spn53F, Spn75F, Spn76A, Spn77Bb and Spn77B c) (Table 1).

Using synteny to find additional family members

Within multi-gene families, identifying the true orthologues of a particular gene becomes increasingly difficult in distantly related species. Here we have used conservation of gene order in the chromosome, both as a means to search for additional family members, and also as a way of assigning putative orthology in otherwise ambiguous cases. This approach is based on the finding that the order of the majority of the genes along the chromosome arms is conserved among the Drosophilid species [33–36]. This allows neighbouring transcripts to be used to identify putative serpin genes which may have been missed by homology-dependent searches. Using this approach, we identified 223 serpin genes in the Drosophilid lineage with conserved synteny to D. melanogaster, compared to 188 genes identified by sequence homology alone. A tBLASTx homology search using this expanded set of 223 genes as queries, identified 3 additional orthologues, bringing the total number of identified serpin genes to 226. The identity of candidate syntenic genes was confirmed by aligning their putative RCL sequences with that of the corresponding set of orthologues (Table 1 and see Additional file 2). By this criterion the synteny approach made no false assignments. In addition to increased identification of serpin transcripts, the method allows the assignment of putative orthology when considering closely-related genes.

There were three cases in which the tBLASTx and synteny searches failed to identify orthologues in a single species (D. sechellia Spn43Ab, D. willistoni Spn43Ad and D. erecta Spn85F). Examination of the sequence alignment of the syntenic genomic regions by eye identified the missing orthologues for D. sechellia Spn43Ab and D. erecta Spn85F, but the D. willistoni Spn43Ad transcript is deleted from the orthologous Spn43A gene complex.

If each of the Drosophilid species contained 29 serpin genes per genome, as in D. melanogaster, the synteny approach identifies 71% of the expected serpin genes. It is notable, however, that even in the sibling-species, D. simulans, only 23 orthologues were identified by sequence homology, with 5 more identified by synteny, Table 1. The number of identified serpin genes decreases uniformly with increasing evolutionary separation from D. melanogaster. Only 13 serpins are identified by sequence homology (and 16 by synteny) in D. grimshawi. While the exact number of serpins may vary among fruit-flies, it seems unlikely that this number should correlate with evolutionary separation from D. melanogaster. This progressive decrease implies that many of the "missing" serpin functions may be represented by transpositions of serpin genes to novel chromosome sites, followed by sequence divergence. Investigations aimed at identifying further serpin family members are described below.

Conservation of RCL sequences in serpin genes

Many of the RCL sequences are strongly conserved across the Drosophilid lineage, over 40 million years. This is particularly the case with putative inhibitory serpins, Additional file 2. The general RCL consensus hinge region sequence (P17-P9) for inhibitory serpins is E [EKR]G [TS].(AGS)4 [1], being the 17-9th residues N-terminal to the P1/P1'proteinase cleavage site. In the Drosophilid lineage the hinge consensus includes the P8 residue, which is constrained to [TS] and the P13 residue is predominantly [ET]. The presence, or absence, of the hinge-region consensus E [EKAQR]G [TSG] [ET] [AGS] [YAGS] [AGS] [VAGS] [TS] separates the Drosophilid serpins unambiguously into two classes (see Additional file 2), with the exception of Spn100A, see below.

The RCL sequences of the majority of inhibitory serpins are strongly conserved among species, with almost complete identity of the putative P1/P1' sites (Spn27A, Spn28B, Spn28Da, Spn28Dc, Spn42Dc, Spn43Aa, Spn43Ac, Spn55B, Spn77Ba, Spn88Ea and Spn100A), Table 1, Figure 3 and Additional file 2. There are three exceptions to this general rule, where the evolutionary constraints appear to be looser: Spn28F, Spn42Db and Spn42Dd. The orthologues of these three serpins show divergent RCL sequences, with variable P1/P1'sites affecting specificity for putative target proteinases. That these RCL sequences were able to diverge implies that their corresponding ancestral serpins were, at least partially, redundant genetic functions.

Figure 3
figure 3

Multiple sequence alignment of the RCL of Spn43Ac orthologues. Alignment annotation: * Identical residues: closely similar residues, broadly similar residues. The conserved residues in the flexible hinge region and the PF residues of the "shutter region" are underlined. The putative proteinase cleavage site (P1/P1') residues are marked in bold typeface.

Full size image

In general, the substrate specificity of many serpins is broad and loss of the activity of one serpin may be compensated by a related activity. Where this occurs, the constraints on RCL sequence divergence will be less stringent than for strictly non-redundant serpin functions. In addition, the loss of inhibition of a particular proteinase could be compensated by inhibition of a different proteinase within the same cascade, by a serpin with a different inhibitory spectrum. By the criterion that they have conserved P1/P1'sites, the majority of inhibitory serpins in the Drosophilid genomes are non-redundant. In D. melanogaster, visible or lethal phenotypes are associated with lack of function mutations (Spn27A [18, 19, 22], Spn28Dc [20]Spn42Da [23–25, 37, 38], Spn43Ac [16, 17], Spn77Ba [21]). The Japanese collection of RNAi knockdown strains includes eight of the seventeen inhibitory serpins (Table 1). Of these, 5 give a lethal phenotype (Spn27A, Spn38F, Spn43Aa, Spn43Ac and Spn55B) and three (Spn28F, Spn42De and Spn55B) are viable http://www.shigen.nig.ac.jp/fly/nigfly/index.jsp. These data sets are incomplete, but again indicate that many of the inhibitory serpins represent non-redundant genetic functions. There is no evidence for a general expansion of transcripts within complex loci as occurred within the vertebrate A and B serpin clades [29, 30, 32, 39].

In contrast, the RCLs of non-inhibitory serpin-fold proteins (Spn28Db, Spn31A, Spn43Ab, Spn43Ad, Spn47C, Spn53F, Spn76A, Spn77Bb, Spn77Bc, Spn85F and Spn100A) are weakly conserved, implying relaxed sequence constraints, Additional file 2. There are three exceptions, in which the RCL is extremely conserved (Spn31A, Spn43Ab and Spn85F). The RCLs of these three serpin-folds presumably bind to strongly-conserved partner molecules. Four (out of 12) of the non-inhibitory melanogaster serpin-folds have orthologues restricted to the melanogaster group.

The serpin core structure, of 3 β-pleated sheets and 8 or 9 α-helical linkers would be expected to be relatively unconstrained within the Drosophilid lineage. We find that the alignment scores between D. melanogaster and D. grimshawi, separated by about 40 Myr, are higher for the RCL sequences of inhibitory serpins than for their serpin cores (Table 2). In the non-inhibitory serpin-fold proteins, the alignment scores of the core and RCL regions are similar, with three exceptions where the RCL sequence is strongly conserved (Spn31A, Spn43Ab and Spn85F) and one where both core and RCL are moderately conserved (Spn100A) (Table 2).

Table 2 Sequence conservation in serpin core and RCL between D. melanogaster and D. grimshawi
Full size table

Structure of serpin gene clusters

The Spn42D gene cluster contains widely diverged transcripts in melanogaster (Figure 1) corresponding to ancient duplication events that are conserved across the most of the Drosophilidae. Within the Spn42D complex, the Spn42Db and Spn42Dd genes show particularly labile RCL sequences, however, the overall structure of the complex locus is maintained. The structure of the alternatively-spliced Spn42Da transcription unit is maintained within the 5 species of the melanogaster sub-group, but the RCL sequences diverge increasingly in the 7 outgroup species [38]. Similarly, the structure of the Spn43A complex is conserved, apart from the loss of the Spn43Ab gene in D. sechellia and the Spn43Ad gene in D. willistoni. In general, there is little evidence of variation of number of genes within complex loci, such as occurs in the vertebrate Clade A and Clade B lineages [39, 30–32].

The Spn77B and Spn88E gene clusters, on the other hand, contain transcripts with closely related sequences in melanogaster (Figure 1). The duplication/divergence events which gave rise to these clusters apparently occurred within the last 5 Myr in the melanogaster subgroup [40, 41]. The ancestral gene for the Spn77B cluster appears to be the inhibitory serpin Spn77Ba, which has orthologues in all the Drosophilid species. The Spn77Bb gene is unique to D. melanogaster, while the closely homologous Spn77Bc gene is found in D. melanogaster, D. simulans and D. sechellia. The Spn77Bb and Spn77Bc genes, therefore, appear to have arisen from recent duplication events; they are both non-inhibitory serpin-fold proteins, one putatively cell-autonomous (lacking an export signal) and the other secreted (carrying an export signal). Similarly, with the inhibitory Spn88Ea and Spn88Eb serpins, the ancestral gene would be Spn88Ea, with an RCL loop that is extremely conserved across the 12 species. The Spn88Eb gene appears to have arisen as a tandem duplication within the melanogaster sub-group and retains the S/A P1/P1' sites of Spn88Ea, except in melanogaster, where it is S/S. We have here an example of duplication of a serpin gene, divergence of the RCL sequence and, finally, modification of the putative proteinase cleavage site in melanogaster itself.

Similarly, the inhibitory Spn28Da and non-inhibitory Spn28Db serpins, are adjacent, closely related genes (Figure 1). They appear to represent duplications in the melanogaster group. The most likely ancestral sequence is the inhibitory Spn28B, which has a single intron, like Spn28Db. The Spn28Da duplication event was associated with loss of an intron and divergence of the inhibitory RCL sequence. The Spn28Db duplication event retained the single intron, but lost the conserved flexible hinge-residues characteristic of inhibitory RCL sequences. In general, it should be more frequent for non-inhibitory serpin-fold to evolve from an active inhibitor than the reverse; a single substitution within the conserved hinge-region motif can inactivate the inhibitory mechanism [4]; while a series of changes in a non-inhibitory serpin-fold would be required to generate the hinge-region consensus. We did not identify any sets of orthologues that contained both inhibitory and non-inhibitory serpins, which ought to have been identified by the synteny approach. The absence of such mixed sets of inhibitory and non-inhibitory orthologues again implies that the evolution of serpin function is constrained, in the absence of duplication/divergence events.

Sporadic loss and gain of individual genes

The D. melanogaster minimum evolution tree (Figure 1) indicates that the Spn28F, Spn38F, Spn28B, Spn42D, Spn28Db and Spn28Da genes are related. Within this group, Spn28F and Spn38F are closely related and, like Spn42Dd, contain 3 introns. The transcripts of the Spn28 cluster, on the other hand, contain one, or no, introns and, on this basis, represent a separate sub-group. The identified orthologues of all six of these genes are restricted to the melanogaster species sub-group, with the exception of Spn42Dd, which also has identified orthologues in ananassae and the pseudobscura sub-group. The most parsimonious interpretation of the relationship of the Spn28F and Spn38F genes is that they represent recent duplication/divergence events within the melanogaster sub-group lineage. Similarly, we identified two orthologues for Spn28F in D. ananassae, which represent a presumptive duplication in this species, and failed to identify a Spn28F orthologue in D. sechellia, indicating a presumptive loss. As mentioned above, the putative P1/P1' sites of the Spn28F orthologues are variable within the melanogaster group implying a degree of functional redundancy; which is consistent with the loss of this orthologues in D. sechellia.

Apart from these instances there is little evidence of loss and gain of serpin genes. As discussed in the previous section, the Spn77B gene cluster in melanogaster, represents a set of recent duplication events, the Spn88Eb gene corresponds to a recent duplication in the melanogaster sub-group and the Spn43Ad gene is missing in D. willistoni. In general, loss and gain of serpin genes is rare in the Drosophilid lineage.

That said, we might have failed to detect additional transposition/divergence events in the more distant species to D. melanogaster. If such events had occurred, the loose sequence constraints on the serpin core would allow transposed genes to diverge rapidly, so that they would be difficult to identify by homology. Furthermore, single amino-acid substitutions in the RCL of an inhibitory serpin can completely alter its substrate specificity [42]. The functions of the inhibitory serpins with orthologues restricted to the melanogaster sub-group could be replaced in the out-groups by serpins carrying similar P1/P1' residues in their RCLs attached to very distantly related serpin cores. Thus the melanogaster outgroups may well have additional serpin genes that we fail to identify by either homology or synteny.

A particular case of divergence/duplication of serpin inhibitory function can occur within the alternatively-spliced RCL-encoding exons of Spn42Da. Alternatively-spliced serpins were first identified in Bombyx mori [43], although the genetic mechanism was only later described in Manduca sexta [44, 45]. Similar alternative-exon cassettes have been described in Anopheles [46] and Caenorhabditis [47], although to date no mammalian examples are known. In the Drosophilid lineage, the furin-inhibitory RCL (exon 5) is maintained in all species as well as exons 6-8 in the melanogaster sub-group, but the corresponding exons duplicate and diverge in the 7 outgroup species [38]. It might be expected that some of the inhibitory serpin orthologues that we failed to identify in the outgroup species were complemented by inhibitory functions on the alternative-exon cassettes of the Spn42Da orthologues. This appears not to be the case by the criterion that the putative P1/P1'sites of these melanogaster serpins (i.e: K/K, L/S, Y/S, K/S, R/A, E/S and S/S; corresponding to Spn28B, Spn28Da, Spn28Db, Spn38F, Spn42Dd, Spn42De and Spn88Eb) see Table 1, do not correspond to the P1/P1'sites on the novel alternative-exon cassettes in the outgroup species (i.e: D. ana S/R, C/S, L/S; D. per C/S, S/R; D. pse C/S, S/R; D. wil R/S, S/A, S/Y, C/S, R/I; D. moj L/T, E/M; D. vir R/T, E/M, D/S and D. gri G/M, E/M, M/L, D/M, R/A, L/M), information extracted from Figures 2, 3, 4 [38].

Conservation of N-terminal and serpin core sequences of Spn43Ac and Spn100A

The serpins are generally between 350 and 400 amino-acids in length, with very few having either N- or C-terminal extensions to the core serpin domain. One of the few human serpins with an N-terminal peptide, Heparin cofactor II, shows reduced substrate specificity after N-terminal cleavage [48]. Similarly, the Drosophila melanogaster Spn43Ac (Nec) serpin has a long N-terminal extension, which is cleaved on immune challenge [49]. In the case of Spn43Ac, both the full length protein and the cleaved core serpin are active proteinase inhibitors in vitro, but the core serpin shows increased substrate specificity for porcine pancreatic elastase (PPE) [49]. Alignment of the Spn43Ac orthologues shows that the N-terminal extension is present throughout the Drosophilidae (see Additional file 3). This peptide varies in length between 69 amino acids (D. sechellia) and 145 amino acids in (D. grimshawi) with poor sequence conservation, other than the putative R/P scissile bond (see Additional file 3). A long, unstructured N-terminal extension appears to be critical for the function of the Spn43Ac orthologues, but its precise length and sequence are weakly constrained. Note the poor sequence conservation within Spn43Ac serpin core sequence, with the exception of the exposed RCL.

The Spn100A transcript encodes a putative 653 amino acid protein, which is outside the normal serpin size-range. The sequence shows two blocks of serpin homology: an N-terminal stretch of 44 amino acids and a C-terminal stretch of 235 amino acids, separated by an insertion of 294 amino-acids (Pfam alignment) (see Additional file 4). This structure is supported by comparison of genomic and cDNA sequences, which show a single intron of 87 bp in D. melanogaster. The novel internal segment of 294 amino acids can not, therefore, correspond to read-through of a nested transcript. The C-terminal of the Spn100 protein includes a putative RCL, including some of the flexible hinge region residues characteristic of inhibitory serpins. The alignment of the N- and C-terminal serpin domains is conserved in all 11 Drosophilid orthologues, and the amino-acid sequence conservation is similar across the whole length of the protein. The implication is that the Spn100A gene encodes a novel, serpin-derived, protein fold. Novel protein folds arise extremely infrequently and a mechanism involving chimeric transcripts formed from segments of pre-existing peptide-coding sequence has been proposed [50]. Database searches identify Spn100A orthologues in mosquitoes and other insects, but not in other organisms.

Comparison of Drosophilid and vertebrate serpins

Vertebrate genomes contain a large, but variable number, of serpins and proteinases. The human complement of 35 serpins and 176 chymotrypsin-fold serine proteinases is increased to 64 and 227 in the mouse, with many of the differences being in genes involved in reproduction and immunity [51]. Against this background, the D. melanogaster genome contains 29 serpins and 211 chymotrypsin-fold serine proteinases, with Spn42Da encoding 11 isoforms, with or without a signal peptide. A large proportion of these serpins have immune-related functions, or are expressed in the reproductive track. Comparison of human and mouse genomes shows that the serpins of nine of the vertebrate clades show conserved numbers and synteny, with the variations being restricted to clades A and B. In particular, the mouse clade A serpins have expanded, with human α1-antitrypsin being represented by a cluster of 5 genes in the mouse, and human α1-Antitrypsin being represented by a cluster of 14 genes. Similarly, 4 of the human clade B serpins are represented by clusters of 3 to 7 genes in the mouse [29, 31, 32]. A striking feature of this expansion in the mouse lineage, is that many of the RCL sequences are unique [29], novel sets of proteinases are being targeted. We find no evidence of a similar massive amplification of serpin genes in the Drosophilid lineage, although the melanogaster Spn77B and Spn88E clusters appear to represent recent tandem duplications. The one example of extensive duplication/divergence of serpin inhibitory functions that has been identified in the Drosophilid lineage is within the novel alternative-exon cassettes of the Spn42Da orthologues [38]. As in the mouse clade A and B expansions, novel sets of proteinases are being targeted by the diverged alternative-exon cassettes (see above).

An essential component of the serpin suicide-inhibition mechanism, is that the cut RCL sequence inserts between two strands of β-sheetA, as the metastable serpin structure relaxes. As a consequence, it is possible for the RCL of one serpin to insert between β-sheetA of another in the absence of enzymatic cleavage, to form an inert, stable polymer [52, 53]. Under normal physiological conditions, this tendency is low, but it represents a constraint on serpin evolution and mutations which alter the stability of the serpin core can favour polymer formation. Such mutations underlie a set of human genetic diseases, including the Z-allele α1-antitrypsin serpinopathy [54] and conformational dementias [55]. Similar constraints on the stability of the serpin core have been demonstrated in Drosophila, with the isolation of polymerogenic mutations of Spn43Ac (Necrotic) showing homologous amino acid substitutions to Z-allele α1-antitrypsin and other pathogenic serpin variants [56, 57].

Conclusion

We have developed a synteny search algorithm which is a powerful tool for identifying orthologous members of multi-gene families within related species. In the Drosophilid lineage, the majority of inhibitory serpin orthologues show a striking conservation of the critical the P1/P1' proteinase cleavage sites, indicating that these inhibitors do not readily change specificity for their target proteinases. Despite this, the distribution of orthologues is quite patchy: many D. melanogaster serpins have no identified orthologues in the more distant species. The balance between inhibitory serpins and the target proteinases that they control forms a strong evolutionary constraint. The majority of Drosophilid inhibitory serpin orthologues appear to represent non-redundant genetic functions, with strong constraints on the P1/P1'sites. Three complex loci in D. melanogaster appear to represent recent duplication/divergence events where the sequence constraints have been relaxed. Each complex contains a putative ancestral inhibitory serpin, with orthologues across the whole Drosophilid lineage, plus one or two genes restricted to the melanogaster group. Taken together our analyses imply that evolution within the Drosophilid serpin super-family has occurred predominantly by a gene duplication/divergence mechanism.

Methods

To assemble a curated set of serpin sequences within the Drosophilidae, we first compiled the orthologues from a tBLASTx search of the 12-species genes extracted from FlyBase http://flybase.bio.indiana.edu/. Additional serpins were identified using a synteny based approach, which relies on the conservation of the order of chromosomal transcripts between related species. The synteny search code is given in Additional file 5. Genes adjacent to serpin genes in Drosophila melanogaster were used to identify the corresponding region within the other Drosophilid genome sequence assemblies. Where the two adjacent genes on either side of a Drosophila melanogaster serpin could be uniquely resolved to orthologues in other species, the intervening sequence was examined for putative serpin genes. Where the adjacent gene orthologues could not be resolved in a given species, the process was repeated with more distant genes, up to a maximum of 3 genes from the original D. melanogaster serpin. Any putative matches were compared to the original serpin protein by tBLASTx alignment [58]. High-scoring matches were considered to be likely orthologues on the basis of both synteny and sequence conservation. A final tBLASTx search was made using all the Drosophilid serpin genes identified in the previous searches as queries against all 12 Drosophilid genomes. The sets of serpin amino-acid sequences were aligned using CLUSTALW. Minimum evolution bootstraps were built using MEGA4 [59] for each serpin within each Drosophilid species.