In order to identify open reading frames (ORFs) encoding putative Cyps, BLAST and TBLASTn analyses against GenBank^®, genomic sequence data and deduced coding sequences were performed. The Cyp proteins deduced from T. gondii, P. falciparum, T. annulata, T. parvum, B. bovis, and C. hominis are listed in Tables 1, 2, 3, 4, 5, 6, respectively. Two putative Cyps from Cryptosporidium muris were included in the analysis, because the orthologous Cyps could not be identified in the genome of C. hominis. Moreover, two Cyps deduced from the Plasmodium yoelii genome were included as the corresponding P. falciparum are quite unusual. Table S1 – in Additional file 1 in the supplemental online material – lists all Cyp proteins encoded in the genomes of S. pombe and Homo sapiens that were used for comparison with the apicomplexan Cyp repertoire.

The number of putative Cyp genes identified per genome ranges from 7 to 9 for C. hominis (whether or not orthologs for CmCyp44.6 and CmCyp48.8 are assumed to be present in C. hominis) to 14 for T. gondii, while the genomes of all four haemosporidia exhibit an intermediate number of 11 putative Cyps per genome. For T. annulata, an ortholog to TpCyp20.3 appears to be present on chromosome 1, however, its complete sequence could not be deduced from the genome data. Therefore, Table 2 lists only 10 Cyps for this organism although 11 Cyps are expected to be present. The number of Cyps in apicomplexan genomes is very similar to the 6 to 11 Cyp genes in the genomes of most fungi although it should be mentioned that there are fungi with extreme low (2 Cyps in the microsporidium Encephalitozoon cuniculi) and extreme high (16 Cyps in Rhizopus oryzae) numbers of Cyp genes 9. An extremely high number of 19 Cyp genes per genome can also be found in the kinetoplastid protozoan parasite Trypanosoma cruzi 30. Similar extremes cannot be found in the genomes of the currently sequenced apicomplexa.

In order to identify subfamilies within the Cyp repertoire and to analyze their phylogenetic relationship, the putative Cyp domains as identified by CD-BLAST 3132 were aligned by ClustalW2 33. Maximum likelihood analysis with PhyML 34 was used to calculate an unrooted tree shown in Figure 1. Statistical support values at the branches are calculated by a likelihood ratio test which produces values similar but not identical to those obtained by bootstrapping 34.

Although it may be assumed that small Cyps containing only a single Cyp domain have been present early in evolution before occurrence of Cyps with one or more additional domains, it is not possible to unequivocally identify a "primitive" Cyp protein subfamily in the apicomplexa from which all other subfamilies have derived, since there are several single- and multi-domain Cyps in the genomes of all protists analyzed so far. Obviously, many subfamilies of Cyps have already evolved before spread of the major lines of eukaryotic evolution.

According to their phylogenetic relationship, 16 different Cyp protein subfamilies were defined here (Figure 1) – many of them well known from other eukaryotes. All these subfamilies exhibit a statistical support in the likelihood ratio test implemented in PhyML of at least 85% and all families containing Cyps with multiple domains are also supported by their domain architecture. The only exception is the subfamily containing putative Cyps with a so-called SYF2 domain, a domain first described in the yeast splicing factor SYF2 35. One of these putative SYF2-containing Cyps, i.e. PfCyp80.9, has a very divergent sequence that does not fall into the same PhyML-deduced group as the other subfamily members (Figure 1B). The corresponding protein deduced from P. yoelii (PyCyp74) was therefore also included and the latter is apparently an ortholog to the SYF2 Cyps of other apicomplexa. Since the subfamily of Cyps with SYF2 is strongly supported by domain architecture and all Plasmodium species but P. falciparum posses putative SYF2 Cyps with high similarity to PyCyp74, it appears that the putative PfCyp80.9 was either not predicted correctly or has undergone dramatic alterations after separation of P. falciparum from P. vivax and the rhodent malaria species. Instead of clustering with other SYF2 Cyps, PfCyp80.9 forms a group together with a group of large putative Cyps that can only be identified in the genus Plasmodium (Plasmodium-spec. Cyps), represented in Figure 1 by PfCyp72.9 and PyCyp69.8.

The phylogram in Figure 1 also indicates the presence of two major groups of Cyps depending on whether they contain a Cyp domain related to the Cyp_ABH subtype (CD database accession number [cd01926], drawn on dark gray background in the lower half of Figure 1A) or any of the non Cyp_ABH-like domains (on light gray background in the upper half of Figure 1A). Within the Cyp_ABH group, it is noteworthy that several important groups of well-known Cyps are absent from apicomplexan genomes whereas there are new Cyp subfamilies that appear to be specific for apicomplexa. On one hand, there are apparently no orthologs of HsPPIB or HsPPIC (both involved in protein folding in the secretory pathway 8), PPID (function in mitochondrial permeability transition during cell death responses 3637), and PPIG (involved in splicing 38). On the other hand, there are several Cyp subfamilies that are specific at least for lower eukaryotes or even for apicomplexa but do not have orthologs in their mammalian hosts and might therefore be promising drug targets in the future. This includes in particular mitochondrial Cyps, Cyps with SYF2, Cyps with signal peptide, and a group of small, presumably cytosolic Cyps specific for apicomplexa.

The following sections will describe genomic organization and protein domain architecture of these subfamilies beginning with the Cyp_ABH-containing proteins. The different subfamilies will be described in the same order in which they are presented in the phylogenetic tree in Fig. 1.

The prototypical Cyps in humans and S. pombe, HsPPIA and SpCyp1, respectively, are closely related as shown in Figure 1. They form a cluster together with additional human paralogs such as PPIE, PPIF and PPIAL4A-G. The corresponding putative Cyps in apicomplexa, ChCyp17.9, BbCyp23.7, PfCyp19, ChCyp18.4, TgCyp18.8, and TgCyp18.9 (Figure 2A), form a related but separate cluster, i.e. SpCyp1 is significantly more closely related to HsPPIA, HsPPIE, and HsPPIF than to any of the apicomplexan Cyps. C. hominis and T. gondii encode two distinct putative members of this PPIA-like subfamily in their genomes. Due to their very high expression levels and their cytoplasmic localization, cytoplasmic Cyps containing a Cyp_ABH type domain (CD database accession-no [cd01926]) are considered to be the most important receptors for CsA leading to inactivation of the cytosolic calcineurins 39. Indeed, mutations in PfCyp19 appear to be sufficient to confer resistance to CsA to P. falciparum 25.

All these small cytosolic Cyps have no introns interrupting their putative ORFs (Figure 2A).

Conspicuously, the putative PPIA-like Cyps of the piroplasms, BbCyp23.7 (Figure 2A), TpCyp25.5 and TaCyp25.7 (Figure 2B), are predicted to contain an NH₂-terminal signal peptide (SP) for cotranslational transport into the ER and for the latter two the PATS algorithm 40 predicts an apicoplast transit signal (AP) following the SP. Such combinations of SP and AP are typical for proteins which are transported into the apicoplast 41. TpCyp25.5 has been described to be cotranslationally transported into dog pancreas rough microsomes in a wheat germ in vitro translation system 42 demonstrating that the SP is functionally active in this heterologous system. Since no removal of the signal peptide was detectable in this system, TpCyp25.5 was proposed to be anchored to the membrane of the ER via an uncleavable signal peptide. However, since transport of proteins to the apicoplast has been shown to require passage through the ER (for review see 41) and the properties of proteases responsible for removal of signal peptides might be quite different in apicomplexa and mammals, these results do not exclude that in vivo TpCyp25.5 is transported further from the lumen of the ER into the apicoplast. Moreover, it cannot yet be excluded that the putative BbCyp23.7 will turn out to be localized in the apicoplast as well since the sequence between the SP and the Cyp domain is long enough to function as an AP. Since the neural network analysis used in PATS has been trained only on AP of proteins from P. falciparum, it might well turn out to be less sensitive to functional AP in other apicomplexa such as B. bovis. However, TpCyp25.5 and TaCyp25.7 appear to be more closely related to HsPPIA and SpCyp1 than to the other apicomplexan Cyps of this group whereas BbCyp23.7 clusters together with the clearly cytoplasmic Cyps of other apicomplexa. Whether the position of apicoplast Cyps in the phylogram truly reflects different evolutionary origins from cytoplasmic Cyps or different selective pressures caused by localization in cytoplasm and apicoplast cannot be decided using the current dataset. In addition to the presence of a putative AP, this group also differs from the small cytosolic apicomplexan Cyps – including BbCyp23.7 – by the presence of three introns within the coding sequence. In order to evaluate whether these Cyps with AP are more closely related to Cyps from plants or algae, BLASTp analyses were performed against protein database entries from dinoflagellates, red, green, and brown algae, green plants, and Arabidopsis thaliana. However, highest similarity was always found to cytosolic PPIA-like Cyps and never to Cyps known to be localized in plastids (data not shown). Experimental evidence concerning the localization of BbCyp23.7, TaCyp25.7 and TpCyp25.5 might provide important information on the evolutionary history of these proteins as well, as they should be considered to be monophyletic if all three turn out to be localized in the apicoplast. Moreover, the fact that Theileria species do not have a prototypical cyctosolic PPIA-like Cyp might explain why T. annulata is resistant to CsA 21.

The dendrogram in Figure 1 reveals a group of putative small Cyps with SP that also contain a Cyp_ABH type domain and are relatively closely related to the PPIA-type Cyps. The domain architecture and genomic organization of these Cyps is schematically presented in Figure 3. One putative member of this Cyp subfamily could be identified in each apicomplexan genome. In contrast to the small cytosolic PPIA-like Cyps, the coding regions of all subfamily members are interrupted by introns. Whereas C. hominis and T. gondii show a very similar exon/intron structure with 4 introns, the putative genes of both Theileria species have only 2 introns – apparently due to fusion of exons 3 and 4 – and after further fusion of exons 1 and 2 only a single intron remains in B. bovis. In PfCyp21.7, loss of introns has resulted in a Cyp domain that is encoded by a single exon. However, a new intron has also appeared within the region encoding the SP, which is encoded by a single exon in the other Cyps of this subfamily.

Cyps within the secretory pathway of other eukaryotes are typically of the PPIB type. The phylogram in Figure 1, however, reveals that the Cyp subfamily with putative ER-localization in apicomplexa does not form any cluster with HsPPIB and SpCyp4 and is therefore proposed not to represent orthologs of PPIB-like Cyps. Apparently, Cyps in the secretory pathway evolved independently at least twice during evolution of eukaryotic Cyps.

For some members of this Cyp subfamily experimental evidence regarding their expression and function is available. First, the PfCyp21.7 protein has been shown to be expressed at extremely high levels in blood-stage parasites, constituting up to 0.5% of total cellular protein 24. Conspicuously, PfCyp21.7 has been been reported not to be confined to the secretory pathway but to be at least partially present in the cytosol as well 24. This raises the possibility that this Cyp subfamily might also be able to interact with cytosolic calcineurin-like phosphatases. Indeed, genetic analysis provides evidence that a mutation in PfCyp21.7 is sufficient to confer resistance to CsA even in the presence of intact PfCyp19 25.

Secondly, TgCyp19.6 has been shown to be secreted by the parasite and to trigger release of IL-12 from host dendritic cells. Moreover, a 19.4 kDa Cyp from Neospora caninum belongs to the same orthology group (data not shown). This protein has been described to be secreted by the parasite and to be present in large amounts in culture supernatants of cell infected with N. caninum tachyzoites 43. NcCyp19.4 from cell culture supernatants was shown to be a very potent inducer of IFNγ production by peripheral blood mononuclear cells and CD4⁺ T cells. Induction of IFNγ by NcCyp19.4 could be specifically inhibited by CsA in a dose dependant manner. These results indicate that apicomplexan Cyps with signal peptide are not only involved in protein folding in the secretory pathway but can fulfill important immunomodulatory functions in infected tissues.

Putative Cyps with a mitochondrial localization signal at their NH₂-terminus are schematically shown in Figure 4. The mitochondrial localization signal and a cleavage site were significantly predicted by MitoProt II 44 for TaCyp24.8 and TpCyp24.5. In contrast, cleavage site prediction was not possible for both PfCyp32.3 and BbCyp26.9. Nevertheless, MitoProt II predicts a high probability of mitochondrial localization and the algorithm PlasMit 45, which was specifically developed to predict mitochondrial proteins in Plasmodium, also suggest a mitochondrial localization of PfCyp32.3. Putative proteins of this Cyp subfamily could be detected only in the haemosporidia but neither in T. gondii nor in C. hominis. In T. gondii, mitochondrial PPIase activity might be achieved by the putative TgCyp31.8, a member of the subfamily of apicomplexa-specific Cyps (Figure 6). In contrast to all other members of this group, TgCyp31.8 is predicted to have an NH₂-terminal mitochondrial localization signal. However, mitochondrial PPIase activity might also be dispensable in apicomplexan mitochondria as it is completely absent from the genomes of both C. hominis and C. muris.

The genomic organization differs largely between the different genera with 5 exons in T. gondii, 3 or 4 in T. annulata and T. parva, respectively, and only a single large exon in P. falciparum (Figure 4).

A group of putative Cyp proteins that appears to be present exclusively in Plasmodium species is shown in Figure S1 in Additional file 2 in the supplemental online material. In order to demonstrate that these proteins represent a subfamily on their own, the putative PyCyp69.8 was included in the phylogenetic analysis shown in Figure 1. Figure S1 reveals that both PfCyp72.5 and PyCyp69.8 possess several nuclear localization signals and two coiled-coil domains, which are typically involved in protein-protein interaction. Moreover, PSORT II recognizes an RNA-binding motif typical for components of ribonucleoprotein particles 46 in PfCyp72.5 further suggesting that this subfamily might somehow be involved in RNA processing.

A multi-domain Cyp subfamily within the Cyp_ABH domain group are the predicted Cyps containing an SYF2 domain (PFAM accession-no.: [pfam08231]) (Figures 1 and 5). This subfamily does not form a monophyletic cluster in Figure 1B due to the fact that putative PfCyp80.9 is quite aberrant and therefore clusters together with the Plasmodium-specific Cyps described in the section above. However, since PfCyp80.9 contains a SYF2 domain and since its ortholog in P. yoelii, PyCyp74, is closely related to the other SYF2 domain containing Cyps, this family should nevertheless be considered to be monophyletic.

In the genome of C. hominis, a SYF2 Cyp could not be identified, presumably due to incomplete sequence information since a putative orthologues protein is encoded in the genome of C. muris, and this sequences was therefore included for further analyses (Figures 1 and 5). The predicted SYF2-Cyps are quite large proteins with predicted molecular weights between 48.8 kDa (C. muris) and 80.9 kDa (P. falciparum). The Cyp_ABH domain is located in the immediate NH₂-terminus of the proteins while the SYF2 domain can be found close to the COOH-terminus (Figure 5). The large region between these two defined domains does not exhibit any known sequence features, and homology between individual subfamily members is very low. Only a few scattered amino acids appear to be conserved throughout the subfamily. In accordance with a suspected role in RNA processing, PSORT II predicts at least one nuclear localization signal in all putative SYF2-Cyps but CmCyp48.8. The latter is also peculiar due to its small size and the presence of only an incomplete SYF2 domain. It must at least be considered that the prediction of the protein coding region of this protein from the genomic sequence is still only partially correct.

The genomic organization of SYF2-Cyps is again characterized by progressing loss of introns. While the coding sequence of TgCyp66.2 is spread across six small exons, there is a large first exon in all other members of the subfamily encoding the Cyp domain, the intervening region and the first part of the SYF2 domain. The remaining sequence is split in two exons in TpCyp59.8 and BbCyp57.3 which have further fused in TaCyp63 to give a two exon structure. Finally, the coding sequence of PfCyp80.9 is encoded by a single very large exon. For CmCyp48.8, only parts of the coding sequence could be identified in the available genomic sequences – further challenging the reliability of the predicted protein sequence. Therefore, schematic presentation of the genomic organisation of CmCyp48.8 is not shown in Figure 6.

SYF2 is a nuclear protein described to be involved in splicing of primary transcripts in the yeast Saccharomyces cerevisiae suggesting a contribution of SYF2-Cyps in RNA processing. Cyps with an SYF2 domain have not been identified in higher eukaryotes, however, the CDART tool 47 retrieves three non-apicomplexan putative protein sequences with a similar domain architecture from the ciliates Tetrahymena thermophila (accession-no.: [XP_001019212]) and Paramecium tetraurelium [XP_001423850] as well as from the primitive green algae Ostreococcus tauri (chlorophyta) [CAL53491].

The sequence of TgCyp36.7 has been predicted by the TwinScan algorithm 48, however, this sequence (TgTwinScan_3870) is no longer available in the predicted ORFs of the T. gondii genome. Since this is the only gene prediction from this region of chromosome VII containing the complete Cyp domain, it was nevertheless included in the analyses though the correct prediction of the ORF outside the Cyp domain is quite dubious. The putative TgCyp36.7 protein (Figure S2 in Additional file 3) is very perculiar in several aspects. First, it does not have orthologues in any of the other apicomplexan genomes (Figure 1). Secondly, despite its relatively large size, the only known protein domain recognized within its sequence is a Cyp domain. Though the latter clusters together with Cyp_ABH domains in Figure 1, it is not recognized as this domain subtype by CD-BLAST but only as general Cyp domain [cd00137]. Thirdly, TgCyp36.7 and HsPPIL6 form a very significant cluster in the phylogenetic analysis (Figure 1) indicating that they might be orthologs. Indeed, size and domain architecture of both proteins are similar. However, the huge evolutionary distance between TgCyp37.7 and HsPPIL6 – indicated by the long branch leading to TgCyp36.7 in the phylogram (Figure 1) – severely questions this hypothesis. Currently, no functional data are available for either PPIL6 or TgCyp36.7.

An additional subfamily of relatively small putative Cyps containing a Cyp_ABH domain can be identified in most apicomplexan genomes with the exception of C. hominis and T. annulata (Figures 1 and 6). Since there is also no evidence for an ortholog from C. muris or Cryptosporidium parvum sequences, the conclusion that this subfamily was lost in the genus Cryptosporidium appears to be valid. In contrast, BLAST analysis indicates the presence of an orthologous gene on chromosome I of T. annulata though the coding sequence could not be completely deduced – maybe due to insufficient sequence quality of the genome sequence. Domain architecture of this Cyp subfamily (Figure 5) reveals that there is a considerably larger heterogeneity than for the two groups described above. First, the putative TgCyp31.8 sequence contains an additional NH₂-terminal mitochondrial localization signal as predicted by MitoProtII 44. Due to the fact that this signal is only observable in a single species and would indicate a significant functional difference to its orthologs in other apicomplexa, careful experimental analyses are needed to compare localization and function of this group of Cyps in different apicomplexa. Secondly, TpCyp20.3 is very small and consists of little more than a Cyp domain, while BbCyp28.6, PfCyp26.4, and TgCyp31.5 have considerable COOH-terminal extensions. Functional data on this Cyp subfamily are completely missing yet. This group of Cyps has obviously no direct orthologs in mammalian genomes and appears to be specific for apicomplexa. In BLASTp analyses, the most closely related non-apicomplexan Cyps appear to be of plant origin (data not shown). The fact that most of these proteins are predicted to be cytoplasmic and that they have no orthologs in mammalian hosts makes them an attractive target to develop drugs such as non-immunosuppressive CsA derivatives that might specifically target this Cyp subfamily.

The PPIH-like Cyps represent another subfamily containing a Cyp_ABH domain that is predicted to be present in all analyzed apicomplexan genomes (Figures 1 and 7). In addition to their Cyp domain, these putative proteins have a short NH₂-terminal extension which does not contain any recognizable motifs or domains. Only in PfCyp24.9 this NH₂-terminal region is characterized by its richness in Asn residues. Though none of the putative apicomplexan PPIH-like Cyps contains any obvious subcellular localization signals, it should be mentioned that their human ortholog has been described to be located in the nucleus and to be associated with the splicing machinery 4950. Specifically, HsPPIH is able to interact independently with the factors HsPrp3 and HsPrp4 that both integrate into the U4/U6 di-snRNP particle. The binding-site of HsPrp3 and HsPrp4 for HsPPIH is highly homologous, and binding does not need enzymatic activity of PPIH since it is not impaired by the presence of CsA. PPIH-like Cyps are highly conserved between apicomplexa, fungi and mammals suggesting that the apicomplexan orthologs might carry out similar functions as well.

The next multi-domain Cyp subfamily to be described here are the recently identified FCBP proteins 28 which contain two phylogenetically unrelated PPIase domains, i.e. an FK506-binding domain (FKBP) at the NH₂-terminus and a Cyp_ABH type domain in the COOH-terminus (Figure 8). Between these two enzymatic domains, there are three tetratricopeptide repeat domains (TRP) [cd00189] which are typically involved in protein/protein interactions and might contribute to recruitment of specific substrates for FCBP proteins.

Isomerase and chaperone activity have been demonstrated for both PPIase domains of TgFCBP57.3 and the inhibitors FK506 and CsA can suppress activity of the FKBP and Cyp domain, respectively 28. Moreover, Adams et al. 28 could show that only the complex of the FKBP domain with FK506 but not the complex of the Cyp domain with CsA was able to inhibit T. gondii calcineurin protein phosphatase activity. Although a weakly synergistic inhibitory effect of FK506 and CsA on parasite growth was noted, this must not necessarily be due to action of TgFCBP57.3 but can also involve any of the other Cyp or FKBP proteins expressed by T. gondii. More convincing as a first hint for an important role of FCBPs in the physiology of apicomplexa is the fact that suppression of TgCyp57.3 expression by RNA interference results in severely decreased incorporation of [³H]uracil 28.

In addition to TgFCBP57.3, putative FCBP proteins can be found only in the genomes of T. parva, T. annulata, and B. bovis but not in any of the Plasmodium or Cryptosporidium species (Figures 1 and 8). In all four apicomplexan FCBP proteins, the enzymatically active domains are separated by TRP repeats. Conspicuously, BLASTp and tBLASTn analyses of protein and nucleic acid databases as well as the CDART tool reveal that putative proteins containing both a Cyp and a FKBP domain are present even in very distantly related organisms such as bacteria 528 (see Tables S2 and S3 in Additional file 4 in the supplemental online material for accession-no.). Furthermore, putative FCBP proteins can also be identified in the ciliophora T. thermophila and P. tetraurelia (Figure S3 in Additional file 5). Since ciliophora and apicomplexa are considered to be phylogenetically related and are usually placed together with dinoflagellates in the infrakingdom alveolata 51, this finding suggests that FCBP proteins were already present in their common ancestors. This hypothesis is also supported by the fact that the deduced FCBPs of ciliophora are also separated by TRP repeats. However, at least the putative TtFCBP131.6 appears to have evolved new or additional functions, since this protein exhibits the presence of an additional (though incomplete) NTPase domain in its very long NH₂-terminus. Such an NTPase domain can be found neither in its homologs in Paramecium nor in the apicomplexan FCBPs. An alternative explanation for the large NH₂-terminus might be an incorrect prediction of the intron/exon structure resulting in fusion of two adjacent but distinct genes in the database entry. An important argument for the latter hypothesis is the incompleteness and therefore presumably non-functionality of the NTPase domain in the predicted sequence of TtFCBP131.6.

Putative FCBPs can also be identified in the oomycete Phytophora capsici, the green algae O. tauri (chlorophyta) and in archaebacteria (Figure S3 in Additional file 5). Whereas PcaFCBP52.5 also contains a Cyp_ABH domain, the Cyp domains in O. tauri CPR7 is truncated and therefore only recognized as Cyp superfamily (accession-no.: [cl00197]). In both predicted archaebacterial FCBPs, CD-BLAST identifies only a Cyp domain without further specification (accession-no.: [cd00317]). In contrast to PcaCyp52.5, neither OtCPR7 nor the archaebacterial FKBPs do contain TRP repeats separating the two PPIase domains (Figure S3). Finally, it should be mentioned that the OtCPR7 sequence might be COOH-terminally truncated since the Cyp domain itself is truncated. In contrast to all other FCBP proteins identified here, OtCPR7 contains an NH₂-terminal mitochondrial localization signal as predicted with high significance by both PSORT II 52 and MitoProt II.

There are also several putative dual-family immunophilins with an NH₂-terminal Cyp and a COOH-terminal FKBP domain in proteo- and flavobacteria as well as in spirochaeta (Figure S3). Here, these proteins are called CFBPs, and they do not contain any TRP repeats. All these putative bacterial CFBPs are very similar in size and domain architecture, however, Borrellia hermsii CFBP38 has a prokaryotic membrane lipoprotein lipid attachment site (Prosite accession-no.: [PS51257]) at its immediate NH₂-terminus as identified by InterProScan suggesting that BhCFBP38 is exported by the bacterium. The domain architecture of all non-apicomplexan FCBPs and some representative CFBPs are shown in Figure S3.

The discontinuous distribution pattern of FCBPs and CFBPs in phylogenetically unrelated clades raises the question whether these proteins evolved multiple times independently. Alternatively, a common evolutionary origin of proteins with this domain architecture might be assumed followed by either loss from most genomes or horizontal gene transfer. In order to address this question, BLAST analyses were used to identify those Cyps and FKBPs in archaebacteria, eubacteria, and eukaryotes that show the highest similarity to the diverse FCBPs and CFBPs. All proteins used for these analyses are listed in Tables S2 and S3 in Additional file 4. Then, maximum likelihood analyses were performed independently on ClustalW2-built alignments of Cyp and FKBP domains. Results of these phylogenetic analyses are presented in Figure 9. The cyclophilin domains of all eukaryotic FCBPs are closely related (i.e. most of them are recognized as Cyp_ABH domain by CD-BLAST) and therefore form a highly significant cluster in Figure 9A (group in the dendrogram with blue background). However, they are clearly not monophyletic as there are several non-FCBP Cyps within this group and FCBP proteins have apparently evolved at least three times independently – i.e. in chlorophyta, oomycetes and alveolata. For OtCPR7 this conclusion is further supported by the fact that this FCBP does not contain any TRP repeats. Cyp domains of the putative archaebacterial FCBPs are not even closely related to this group and form a completely independent cluster. The Cyp domains of proteo-/flavobacterial CFBP proteins are monophyletic – in contrast to those of spirochaetes. However, for the latter group there are currently only members known from Treponema denticula and four Borrelia species. It is for instance possible that one of these two proteins is highly divergent from the average spirochaete CFBP due to secondary evolutionary changes. In particular, the presence of a lipoprotein anchor at the NH₂-terminus of BhCFBP38 suggests an extracellular localization of the mature protein and therefore a significantly altered function.

The Cyp domains of FCBPs of ciliophora and apicomplexa are closely related, surprisingly, however, a non-FKBP Cyp from the rhodophyte Griffithsia japonica is proposed to be a member of the same cluster as revealed by maximum likelihood analysis. Since red algae are frequently supposed to be the evolutionary origin of the apicoplast, one explanation for this result may be that the Cyp domain of FKBPs in alveolata was derived from the genome of a rhodophyte-related secondary endosymbiont.

Phylogenetic analysis of the deduced FCBP domains does also not support a monophylic origin of alveolate FCBPs. Overall, phylogenetic distances between FKBP domains are much larger than for Cyp domains (compare scale bar between Figures 9A and 9B) indicating that the latter are far better conserved. Moreover, the phylogram reveals poor sequence conservation even within groups containing a well conserved Cyp domain. For instance, the CFBPs of proteo-/flavobacteria do not form a monophyletic group when FKBP domains are analyzed (Figure 9B). The two spirochaete CFBPs are clearly separated and the green algal OtCPR7 does not show any close relationship with FKBP domains from other eukaryotic FCBPs but appears to be closely related to bacterial FKBPs suggesting that it might have been acquired from a cyanobacterial endosymbiont. Though the FKBP domains of all alveolat FCBPs can be found in the same highly significant cluster (group with blue background), this group also contains non-FCBP FKBPs. Conspicuously, however, all FKBP proteins within this group also contain TRP repeats (compare Figures 8, S3, 9B, and Table S3). In contrast to the results obtained for Cyp domains, not even the FKBP domains of FCBPs from ciliophora and apicomplexa appear to be monophyletic. On one hand, it is quite unlikely that the same structure of FKBP domain and Cyp domain connected by TRP repeats arose multiple times independently and it can therefore be suspected that this result is due to high and diversifying evolutionary pressure on FKBP domains in this protein family. On the other hand, there is a widely distributed monophyletic family of FKBP proteins with TRP repeats. These proteins might indeed have captured a Cyp domain several times independently. The strongest argument for the latter evolutionary pathway is the presence of putative FKBP proteins with TRP repeats and a concurrent absence of predicted FCBP proteins in all Plasmodium species. If FCBP proteins would be a monophyletic, ancient group in alveolata or at least in apicomplexa, secondary loss of the Cyp domain from FCBPs restoring the ancient FKBP/TRP protein has to be postulated for Plasmodium.

In summary, additional data from more species are needed to draw a conclusive pattern of evolution for FCBP proteins in protists. Currently, there are hints supporting either a monophyletic origin or multiple independent origins though the latter option appears to be slightly more likely.

The first group of putative Cyps with a non-Cyp_ABH domain is represented by the WD40 repeat-containing Cyps which are encoded in all apicomplexan genomes analyzed here (Figure 1). This subfamily contains multi-domain Cyps with a Cyp_WD40 domain (CD accession-no: [cd01927]) in its COOH-terminus (Figure 10). WD40 domains (CD accession no.: [cl02567]) are characterized by repeats of about 40 amino acids containing a characteristic Trp/Asp dipeptide. The length of the deduced WD40 domain ranges from 145 amino acids in P. falciparum to 321 and 328 amino acids in C. hominis and T. gondii, respectively. The WD40 repeat domain is either located close to the NH₂-terminus as in Theileria species and in B. bovis, or is preceded by a longer NH₂-terminal extension without identifiable domains or motifs as in T. gondii, P. falciparum, and C. hominis. There is no sequence similarity between the orthologs within this region. In T. gondii, however, a nuclear localization signal can be found here, and in P. falciparum there are two short stretches rich in Asn and Lys, respectively. An orthologous protein from E. tenella has recently been described to contain a very Ser- and His-rich NH₂-terminus 29.

The genomic organization differs largely between species with intronless genes in C. hominis and P. falciparum while there are between 6 (B. bovis) and 12 introns (T. gondii) in the other species.

Deckert et al. 53 showed that the human WD40-repeat Cyp is a component of the spliceosomal B complex which contains the complete set of U snRNAs in a precatalytic state. However, its precise role in splicing or regulation of splicing has not been addressed yet. Recent structural analyses show that – in the crystal – the NH₂-terminus of the protein binds to the active site of a neighboring molecule in a substrate-analogous manner 54. Binding of this sequence to the active center without subsequent isomerization was also confirmed by NMR solution studies.

For Cyps of the PPIL1/SpCyp2 subfamily, CD-BLAST does not recognize any special Cyp domain but only the Cyp superfamily in general (accession-no.: [cl00197]). Only two putative members of this subfamily can be found in the apicomplexan genomes analyzed here, i.e. TgCyp21 and PfCyp23.2 (Figure 1 and Figure S4 in Additional file 6). PfCyp23.2 is also predicted to have an NH₂-terminal coiled-coil region. However, there is no hint for such a domain in other PPIL1-like Cyps such as TgCyp21 or SpCyp2.

HsPPIL1 has been demonstrated to be a part of the spliceosomal machinery 38 and to directly interact with the highly conserved transcriptional cofactor SKIP 55. Although PPIL1-like Cyps are widely spread, they are for instance missing in a large number of fungal genomes 9 indicating that the function of this subfamily is not essential. Since SKIP is also involved in splicing and remains bound to the spliceosome throughout both trans-esterification steps 55, PPIL1-like Cyps and SKIP might be involved in the complex linkage of transcription and splicing during mRNA processing.

Within the non Cyp_ABH group, the PPIL3-like Cyps are the only subfamily of small single domain Cyps that is widely distributed among apicomplexa. PPIL3-like Cyps can be predicted in all the apicomplexan genomes and consist of little more than the Cyp_PPIL3 domain (accession-no.: [cd01928]) itself (Figures 1 and 11). Since apparent localization signals are missing, a predominant cytosolic localization of the putative proteins might be assumed. However, at least one splice form of human PPIL3 has been identified as part of the B complex of the spliceosome in the nucleus 38 though its precise role in the spliceosome B complex has not been analyzed yet.

In contrast to the PPIA-like subfamily, the major subfamily of small cytosolic Cyps, most of the PPIL3-like Cyps are encoded by genes consisting of four (P. falciparum) to six (B. bovis) exons (Figure 11). In this case, the only exception is ChCyp18.9, which has an intronless coding sequence.

Putative PPIL2-like Cyps form a very robust phylogenetic cluster (Figure 1) and are characterized by the presence of a so-called RING finger domain that has been reported to facilitate E3 ubiquitin-ligase activity 56. Cyps with RING finger motif in their NH₂-terminus are widely spread among different organisms including fungi, plants and mammals, and the Arabidopsis thaliana ortholog, AtPUB49, has been shown to be an active E3 ubiquitin-ligase and to exhibit PPIase and chaperone activity suggesting that it is involved in protein folding and degradation processes. The Cyp_RING domain (accession-no: [cd01923]) is present in the vicinity of the COOH-terminus (Figure 12). This subfamily is apparently missing in the genomes of P. falciparum and C. hominis. The deduced sequence of TgCyp72.9 is somewhat larger than its orthologs in Theileria and Babesia (between 58.8 and 59.4 kDa) due to a longer COOH-terminus. Moreover, TgCyp72.9 contains a putative nuclear localization signal which is not present in its haemosporidian orthologs.

The genomic organization of Cyps with RING finger domain again shows signs of intron-loss during evolution with eight introns in T. gondii, one intron in both Theileria species and no intron left in B. bovis.

In contrast to most other moderate to large size Cyps, the subfamily containing a Cyp-CeCyp16-like domain (accession number: [cd01925]) does not contain any additional domain that could be identified by CD-BLAST or InterProScan (Figure 13). However, there is a nuclear localization signal detectable in all putative apicomplexan CeCyp16-like Cyps, which is located in approximately the same distance from the Cyp domain in all subfamily members with the exception of TgCyp64.5 where it immediately follows the Cyp domain. Moreover, using PSORTII at least one coiled-coil protein-protein interaction domain can be identified in all these proteins but CmCyp43.1 (for CmCyp43.1 the score for a coiled-coil region is only slightly too small to be judged as significant by PSORTII). PfCyp51.8 is even predicted to contain two coiled-coil regions. Moreover, PfCyp51.8 has a large Lys-rich region encompassing both coiled-coil domains. A comparable but much smaller positively charged region consisting of a large number of Arg residues is present in TgCyp64.5.

The C. elegans ortholog CeCyp16 has been shown to be expressed predominantly in the intestine 57 and high-troughput RNAi screening experiments revealed strong phenotyps for CeCyp16 including lethality, embryonic lethality, morphological abnormalities and maternal sterility 5859 indicating that this subfamily of Cyps is very important at least in multi-cellular organisms. However, since CeCyp16 is much smaller than its apicomplexan orthologs and most of the latter miss a clearly definable positively charged domain in their COOH-terminus that can be found in nematode CeCyp16-like proteins 57, it is not yet possible to draw any conclusions regarding the function of CeCyp16-like Cyps in apicomplexa. Even certain biochemical observations made on recombinant CeCyp16-like proteins from the nematodes C. elegans and Onchocerca volvolus cannot simply be extended to their ortologs in apicomplexa. In particular, although both nematode and apicomplaxan CeCyp16-like Cyps reveal absence of an otherwise highly conserved Trp residue in the active center of the enzyme, there are acidic residues in this position in nematode CeCyp16-like proteins but a wide variety of different amino acids residues in CeCyp16-like Cyps of apicomplexa including Gln, Val, Tyr, Cys, and Phe. Since the conserved Trp residue has been shown to be crucial for CsA binding/sensitivity, it can therefore be assumed that CeCyp16-like proteins of apicomplexa are in general rather resistant to CsA. However, predictions about changes in substrate specificity cannot be made using the results of biochemical analyses made on C. elegans and O. volvolus orthologs.

Putative PPIL4-like Cyps are only identifiable in the genomes of C. hominis and T. gondii (Figures 1 and Figure S5 in Additional file 7). Although orthologs are present in other Cryptosporidium species (data not shown) and many but not all genomes of fungi 9, this subfamily is apparently absent from all other apicomplexan genomes analyzed. ChCyp34.5, just like its orthologs SpCyp6 and HsPPIL4, contains an RRM (RNA recognition motif) closer to its COOH-terminus and a Cyp_RRM domain (accession-no.: [cd01921]) in its immediate NH₂-terminus (Figure S5). In contrast to other PPIL4-like Cyps, the deduced sequence of ChCyp34.5 is missing a Ser/Arg-rich SR domain in its COOH-terminus. The predicted sequence of TgCyp66.3 is very unusual since it contains a Cyp domain that is interrupted by a large insertion which remains to be confirmed experimentally. However, in contrast to ChCyp34.5, TgCyp66.2 is more typical for PPIL2-like Cyps since it possesses a Ser-rich and highly positively charged domain in its COOH-terminus. Although it also contains multiple Lys in addition to Arg residues, it can be assumed that this domain fulfills a function similar to that of the SR domain of mammalian and fungal PPIL4-like Cyps. It is not unlikely that both ChCyp34.5 and TgCyp66.3 are not yet predicted accurately (e.g. a missing exon in ChCyp34.5 and a missing intron in TgCyp66.3 would explain the current results) and it will finally turn out that both possess normal Cyp domains and an SR domain.

PPIL4-like Cyps should not be confused with the PPIE-like Cyps, a subfamily that is missing in all apicomplexan genomes. PPIE-like Cyps contain an RRM motif in the NH₂-terminus and a Cyp_ABH domain in their COOH-terminus.

ChCyp34.5 contains a nuclear localization signal within its Cyp domain and PSORTII predicts a nuclear localization. Due to its high content of positively charged amino acid residues, the putative TgCyp66.3 is predicted to have a multitude of overlapping nuclear localization signals in its COOH terminus in addition to one signal about 100 amino acids away from its NH₂-terminus. Indeed, the orthologous AtCyp59 protein from A. thaliana has been described to be localized in the nucleus but outside of those nuclear speckles rich in SR domain proteins 60. Although interaction with other SR domain proteins implicated in RNA splicing could be demonstrated using yeast-two-hybrid and pull-down assays, the punctuate nuclear localization pattern and a measurable interaction with the COOH-terminal domain of RNA polymerase II suggest that AtCyp59 predominantly participates in transcriptional processes and that it is only marginally involved in splicing 60. It is still too early to speculate whether PPIL2-like Cyps of apicomplexa have similar functions as AtCyp59 or other PPIL2-like Cyps since the SR domain responsible for all known AtCyp59 interactions is missing in ChCyp34.5 and the Cyp domain of TgCyp66.3 might well be non-functional. In this context it is also noteworthy that the Cyp domain of this protein is less conserved than that of other PPIL4-like Cyps. Whereas this domain in SpCyp6 and HsPPIL4 is recognized as Cyp_RRM domain (accession no.: [cd01921]), CD-Blast only recognizes a domain belonging to the Cyp superfamily (accession no.: [cd00197]). It is also possible that the truncated or disrupted PPIL4-like Cyps in Cryptosporidia and Toxoplasma containing a slightly degenerated Cyp domain represent transition states that ultimately led to complete loss of this gene in other apicomplexa.

Loss of introns during evolution of organisms exhibiting a parasitic live mode such as Giardia 61, Trypanosoma 62, Trichomonas 63, and Encephalitozoon 64 has been recognized previously and has also been described for apicomplexan parasites on a genome wide scale 65. While Toxoplasma is known to have a genomic organization with a very high number of about five introns on average per gene, Theileria and Plasmodium species have approximately two and one intron on average per gene. Only a very small number of introns can be found in the genomes of Cryptosporidium species which have introns in only about 5% of their genes. However, most of the few remaining introns in Cryptosporidium genomes are not conserved to T. gondii introns and are therefore supposed to be not of ancient origin 65. Apparently, loss of introns occurred independently in the lineages leading to crypto- and haemosporidia.

As expectable from the differences in the general genomic organization, the exon/intron structure of Cyp genes varies widely between species. The number of introns per Cyp gene, i.e. 0.5 for C. hominis, 4.5 for T. gondii, 2.0 for P. falciparum, 3.1 for T. parva, 3.5 for T. annulata, and 2.8 for B. bovis is very similar to that observed on the genome scale. Despite this overall similarity, however, there are some remarkable exeptions from the general pattern: First, the putative ChCyp22.9 with SP in C. hominis shows a conserved position of introns in comparison to its ortholog in T. gondii. Secondly, some Cyp subfamilies have a high number of introns in both haemosporidians and T. gondii although only minimal conservation of introns can be observed. This group includes the small apicomplexa-specific cyclophilins, the Cyps with WD40 repeats (with exception of PfCyp87), and the PPIL3-like Cyps. Finally, there is one Cyp gene subfamily, the CeCyp16-like Cyps, showing a completely reversed trend with more introns in the haemosporidian genomes than in T. gondii. Since several of the introns in haemosporidia appear to be of ancient origin, it is most likely that TgCyp65.5 has lost its introns since divergence from the last common anchestor with the haemosporidia.

In summary, though the trend of moderate and nearly complete loss of introns observed on a genome-wide scale for haemo- and cryptosporidia, respectively, could also be observed for Cyp genes in general, there are exceptions to this rule in certain subfamilies that might be exploited in the future to decipher the selection forces that contribute to conservation of certain introns despite high overall frequency of intron loss. For instance, it would be highly intriguing to look for any functional roles for the three introns in ChCyp22.9 (e.g. on regulation of gene expression) that might explain counterselection against their deletion during evolution.

Initially, putative apicomplexan Cyps were identified using BLASTp and tBLASTn algorithms to search in GenBank^® protein and nucleic acid databases as well as in PlasmoDB, ToxoDB, CryptoDB, and in the Theileria parva genome database of TIGR. S. pombe Cyp1 and Cyp2 were used as query sequences. These Cyps were chosen because they are not closely related. If a Cyp subfamily member was not identified in one of the apicomplexan organisms, a Cyp of the same subfamily from a closely related apicomplexan parasite was used as query to search in protein, cDNA, EST and genome databases. This method ensures that no Cyps are missed in any of the taxa. In order to prevent that no complete subfamilies was overseen, BLAST analyses were also performed using the complete T. gondii Cyp repertoire as a query. However, no additional Cyp sequences could be identified.

In contrast to conventional nomenclature for many Cyps, molecular mass suffixes in the names were given with one position after the decimal point since otherwise identical names would have resulted in a few cases. It was decided not to use suffix letters to avoid a possible confusion with mammalian Cyps. For instance, a Cyp19A might have been confused with a human CypA/PPIA. In addition it should be mentioned that all molecular mass suffixes used have been derived from the predicted sequence of unprocessed proteins. Although this can currently be only a provisional nomenclature, consecutive naming with numbers or letters would result in different names for orthologues Cyps and identical names for unrelated Cyps of different apicomplexa. A more function based nomenclature of apicomplexan Cyps should be introduced later, when at least for one apicomplexan genome all Cyps have been verified experimentally. For human and S. pombe Cyps, names according to the entries in the ENSEMBL database were used.

Homologous putative protein sequences were aligned using ClustalW2 33. Maximum likelihood phylogenetic trees were then calculated with PhyML 34 using the approximate likelihood ratio test option and the JTT model 66 for amino acid substitution. The program was set to estimate the proportion of invariable sites and the gamma distribution parameter, while the number of substitution rate categories was set to four. The input tree was built using the BIONJ algorithm implemented in PhyML. The resulting trees in Newick format were visualized and processed using MEGA4 6768.

For identification of protein domains, CD-BLAST 3132 and InterPro Scan 69 were used. Moreover, protein sequences were scanned for subcellular localization signals with PSORT, SignalP 70, PATS 40, PlasMit 4571, and Mitoprot 44.