Identification of Physcomitrella patensMIPs

When this study was initiated only 11 out of the 23 PpMIPs had been described in the literature [20, 40]. Since then one more of the 23 PpMIPs (PpPIP2;1) has been published [41]. All 23 PpMIP sequences were categorized as belonging to an aquaporin euKaryotic Orthologous Groups (KOG) at the PpDB and most of these also had a suggested classification (Table 1). Based on the phylogeny of the PpMIPs together with the AtMIPs and Z. mays MIPs (ZmMIPs) a new and more systematic classification of the PpMIPs, that is consistent with the AtMIPs and ZmMIPs nomenclature [16, 18], is proposed (Table 1).

Phylogeny and classification

Using the full length protein alignments of all PpMIPs, AtMIPs and ZmMIPs [see Additional file 1] the neighbour joining (NJ) method resulted in one tree (Fig. 1) which was compared to trees from the maximum parsimony (MP) method and the Bayesian (Bay) method. Bootstrap support and Bayesian posterior probabilities were used to construct a "method-consensus" cladogram summarizing the results of the three methods and used to classify the PpMIPs (Fig. 2). The classification of AtMIPs and ZmMIPs in subgroups within subfamilies is similar for all MIPs except the NIPs. We named the PpNIPs according to the nomenclature used in classification of the NIPs in Z. mays and O. sativa since these four wider subgroups allow more sequence divergence and hence are more generic than the more narrow seven subgroups defined in A. thaliana. P. patens subgroups that failed to group with the previously classified subfamily groups were given consecutive higher indices (e.g. PpPIP3, PpTIP6, PpNIP5 or PpNIP6). In total 3 PpPIP1s, 4 PpPIP2s, 1 PpPIP3, 4 PpTIP6s, 1 PpNIP3, 3 PpNIP5s, 1 PpNIP6 and 2 PpSIP1s were categorized. Four PpMIPs failed to be classified into a subfamily, since they lack orthologs among the MIPs identified in A. thaliana and Z. mays. One of these was the MIP xenolog (homolog resulting from horizontal gene transfer) PpGIP1;1 previously identified as a GlpF-like MIP and named accordingly [20]. The remaining three were the PpHIP1;1 which shares similarities with both TIPs and PIPs but forms a separate distinct subfamily of its own, and the PpXIP1;1 and PpXIP1;2, two divergent MIPs that share some unique previously undescribed motifs.

To find orthologs of the three uncategorized PpMIPs (PpHIP1;1, PpXIP1;1 and PpXIP1;2) searches of databases at NCBI and embl were conducted. Hits representing a wide variety of species were selected and the corresponding protein sequences were aligned with the PpPIPs, the PpTIPs and either PpHIP1;1 or PpXIP1;1 and PpXIP1;2. The alignments were used in phylogenetic analyses to evaluate if the newly acquired sequences could help in categorizing the three PpMIPs. The PpHIP1;1 hits were mainly annotated as TIPs or AQP4s in the databases and the phylogenetic analysis resulted in three clusters (PIPs, TIPs and AQP4s) but PpHIP1;1 were still basal to all of these and could therefore not be assigned to any of these subfamilies (data not shown). As for PpXIP1;1 and PpXIP1;2, hits were mostly annotated as Plant MIP, TIP or AQP0 sequences. The phylogenetic analysis resulted in four different subfamilies, TIPs, PIPs AQP0s and a fourth clade consisting of unspecified plant MIPs and the PpXIPs (data not shown), see further analyses in next paragraph.

The XIPs – an unrecognized MIP subfamily in higher plants

Sequences belonging to this fourth clade have a weak overall sequence similarity to MIPs in general (about 30 % amino acid identity, data not shown), and could neither be assigned to any of the previously identified classes of plant MIPs (PIPs, TIPs, NIPs, SIPs and GIPs) nor be associated with the PpHIP1;1 sequence. However, some conserved motifs within this new subfamily (see discussion) were identified and based on these one representative sequence (the castor bean cDNA sequence [GenBank:EG656577]) was selected. This sequence was used in database searches in order to obtain more MIPs belonging to this novel subfamily. A handful of more sequences that all shared the same conserved motifs were identified. One of these sequences originated from Populus trichocarpa and therefore the P. trichocarpa genome at JGI were searched, identifying 4 more paralogs (Table 2). These sequences, together with the sequences retrieved from the castor bean cDNA and the PpXIP searches and all PpMIP sequences (except PpHIP1;1) were combined into one sequence alignment used in phylogenetic analysis. The resulting trees confirmed that the unclassified MIPs form a distinct monophyletic clade (with the PpXIPs as basal taxa), different from the other MIPs included in the analysis (Fig. 3). As shown in Table 3 there is considerable variation both at the first NPA box and the ar/R filter among the sequences in this clade. We propose that, awaiting further characterization, MIPs in the new subfamily should be referred to as X Intrinsic Proteins (XIPs) emphasizing that currently we have very little information on the function of these proteins.

Numbera	IDb	Typec	Organism	Descr.	Comments
1	DN837617	EST	Selaginella moellendorffii	-	cDNA from whole plant
2	BT014197	EST	Solanum lycopersicum d	-	cDNA from fruit
3	DY275505	EST	Citrus clementina	-	cDNA from mixed tissue
4	CO092422	EST	Gossypium raimondii	-	cDNA from whole seedlings
5	CK295158	EST	Nicotiana benthamiana	-	cDNA from mixed tissue
6	EG656577	EST	Ricinus communis	-	cDNA from seeds
7	EG666650	EST	Ricinus communis	-	cDNA from roots
8	CK746370e DT60037e	EST	Liriodendron tulipifera	-	cDNA from flower buds
9	DR936893e DT742029e	EST	Aquilegia Formosa × Aquilegia pubescens	-	cDNA from mixed tissue
10	AM455454	WGSS	Vitis vinifera	-	Exons between nucleotides 61100–61186, 61265–61354 & 61465–62185
11	AM455454	WGSS	Vitis vinifera	-	Exons between nucleotides 69471–69617 & 69685–70443
12	557139	Gene	Populus trichocarpa	PIP	no EST support
13	829126	Gene	Populus trichocarpa	PIP	EST support from cambium
14	767334	Gene	Populus trichocarpa	PIP	no EST support
15	759781	Gene	Populus trichocarpa	PIP	no EST support
16	821124	Gene	Populus trichocarpa	PIP	EST support from petioles
17	XM_639170	Gene	Dictyostelium discoideum AX4 f	MIP	Hypothetical protein

^aNumber used for identification in Fig. 3 ^b GenBank ID or Protein ID for Populus trichocarpa v 1.1 database at JGI ^c EST = Expressed Sequence Tag, WGSS = Whole Genome Shotgun Sequence, Gene = Annotated gene ^d Tomato, previously named Lycopersicon esculentum ^e Two overlapping sequences were used to construct a full length sequence ^f The only non-plant species and a very divergent sequence

Gene structure

The average PpMIP was found to have 2.6 introns with a size of 246.4 bp. This is about half the number of introns, but of approximately the same size as predicted for the average P. patens gene in a genome wide analysis [42]. The exon/intron patterns of the PpMIPs were found to be highly conserved within each subfamily, as shown in Figure 4. Comparison with the AtMIPs showed the intron positions to be conserved for both PIPs and NIPs, but not for TIPs (in P. patens the intron position is 35 base pairs further to the 5'-end) and SIPs (completely lacking introns in P. patens). The exon/intron pattern also supported that the PpHIP and the PpXIPs were to be classified neither as PIPs, TIPs, NIPs, SIPs nor GIPs, but rather as separate subfamilies on their own.

The identification of five P. trichocarpa XIP paralogs allowed comparison of gene structure across species. All five P. trichocarpa genes have the same pattern of exon-introns with two introns in the N-terminal sequence (data not shown). This is also true for the PpXIP1;2, but since the N-termini have a high degree of interspecies variation it is hard to make any conclusion on whether the intron positions are exactly conserved.

Physcomitrella patensMajor Intrinsic Proteins

Comparison of protein superfamilies of distantly related species can aid in our understanding of protein function and by annotating all MIPs in P. patens we have made such a comparison possible for the MIP superfamily of higher plants and mosses. Originally we hypothesised that mosses were to have a relatively small superfamily, due to them being simpler (for example lacking vascular tissue and therefore having a less complex water transport regulation). It was therefore much to our surprise that we found P. patens to have seven subfamilies containing in total 23 different MIPs, an unexpected large and divergent superfamily. One of these (PpGIP1;1) is analysed in detail by Gustavsson et al. [20], and is therefore omitted from this discussion. Half of the remaining 22 PpMIPs are previously described by Borstlap [40] and Lienard et al. [41] and the remaining 11 are previously not described in the literature. The gene structure of the PpMIPs supports the phylogenetic analyses and the resulting division into seven subfamilies. Comparison with AtMIPs shows that PIPs and NIPs have conserved intron positions whereas SIPs and TIPs do not. This is consistent with the conservation of individual groups of the NIP and PIP subfamily in both P. patens and A. thaliana (discussed further below).

PIPs – the most conserved MIPs in plants

PIPs are remarkably well conserved plant MIPs that can be further classified into PIP1s and PIP2s. Both PIP1s and PIP2s are highly conserved in P. patens indicating that these groups must have formed early on in the evolution of land plants and are of fundamental importance in plant physiology. The physiological relevance of PIP1s and PIP2s in water relations in higher plants is well established and recently also carbon dioxide has been added to the list of possible substrates [reviewed in [4]]. The ar/R filter is strictly conserved in PIPs including PpPIPs suggesting that all PIPs, irrespectively of subgroup, have the same substrate specificity (Table 3). It is likely that the evolution of PIP sequences is constrained also in many other ways. For example the PIPs reside in the plasma membrane and it is essential that they are impermeable for protons in order to maintain the proton gradient. Furthermore, the water permeability of PIPs can be regulated by phosphorylations, pH and Ca²⁺ via an intricate gating mechanism [11]. From our results presented here it is clear that the diacidic motif in the N-terminal region and the histidine in the D-loop responsible for Ca²⁺ binding and pH gating, respectively, are both conserved in all PpPIP1s and PpPIP2s. The phosphorylation site in loop B is also conserved in all PpPIPs whereas the PIP2 specific C-terminal phosphorylation motif is restricted to the PpPIP2s. This suggests that the gating mechanism is generic in all species and tissues where PIPs are expressed and that for instance pH gating is not limited to anaerobic conditions in roots of higher plants.

In P. patens there is also an odd PIP (PpPIP3;1), basal to both PIP1s and PIP2s. The PpPIP3;1 has a deletion of 11 amino acids after the second NPA-box (between helix E and helix 6) and this, together with the relatively high divergence from other PIPs (e.g. lack of the Ca²⁺ binding site at the N terminal region and a conserved cysteine at helix 2) and the absence of ESTs, makes it questionable if this MIP gene is at all functional.

TIPs specialization occurred later

It has already been suggested that P. patens is lacking the specific isoforms of TIPs observed in higher plants [40] and now, with this complete set of PpMIPs at hand, this is confirmed. Interestingly, it has been proposed that vacuole sub-types harbor specific sets of TIP isoforms [43] and it is easy to speculate that the TIP groups in higher plants evolved due to special functional requirements of different vacuoles. The identification of conserved proteins in P. patens, involved in the sorting of proteins to different types of vacuoles, suggests that there are most likely more than one type of vacuole in bryophytes [44]. This implies that TIPs are not conserved markers for subtypes of vacuoles as the presence of only one group of TIPs in P. patens indicates that either there is only one of the vacuole types in moss that has TIPs, or alternatively several different vacuoles in the moss cell all have the same type of TIPs. Both interpretations are consistent with recent experiments in higher plants that have challenged the idea of TIPs as valid markers for vacuole sub-types [45, 46].

Rather than forming a very distant subclass of TIPs, the PpTIP6s appears as a conserved mosaic of the different motifs that are found in the different TIP groups of higher plants. For example the first few amino acid residues at the N-terminus are similar to TIP2s, whereas the C-terminal region is most similar to TIP3s. The identities of the amino acid residues at the ar/R filter (HIAR) are shared with both some TIP3s and TIP4s suggesting a similar specificity. In fact exactly these residues are the most common, comparing the frequencies in the selectivity regions of all A. thaliana, Z. mays and O. sativa TIPs (H_0.81I_0.62A_0.72R_0.75; based on Table 4 in [47]). This makes it likely that PpTIP6s are similar to the TIPs present in the last common ancestor of bryophytes and vascular plants and that the other motifs found at these positions are derived characters that have appeared later as different groups of TIPs evolved in vascular plants. The expansion and formation of specialized groups in the TIP subfamily of higher plants might suggest that some of these TIPs have taken over the functions of the MIPs of subfamilies that are missing in higher plants (e.g. HIPs and XIPs).

NIP groups evolved early

In higher plants NIPs form a divergent subfamily with large variation between species. This is true also for NIPs in P. patens, but surprisingly one of the three NIP groups identified is present also in higher plants, indicating that this group of NIPs, NIP3, was present already in a common ancestor to P. patens and higher plants (Fig. 2). The conserved intron positions among NIPs in A. thaliana and P. patens indicate that this gene structure was also present in the ancestral NIP gene. NIPs are different from other MIPs in that they often have unorthodox NPA boxes. In many NIP3s of higher plants the first and second NPA boxes are replaced by NPS and NPV, respectively [47]. The corresponding motifs in PpNIP3;1 are NPA and NPV (Table 3), which is identical to AtNIP6;1 (one of the two NIP3s in A. thaliana according to the monocot classification), suggesting that NIP3s had these motifs before the split of bryophytes and vascular plants.

The two NIP groups specific for P. patens (PpNIP5 and PpNIP6), have a unique combination of amino acids at the ar/R filter (Table 3). In contrast the ar/R region of PpNIP3;1 conforms to the residues found in other NIP3s, supporting that they are orthologs with the same conserved function. Recently a NIP3 have been shown to have a role in boron uptake in roots of A. thaliana [27] and even though mosses lack roots it cannot be ruled out that PpNIP3;1 has a role in boron transport in the moss.

The N-terminal region of NIPs is relatively long compared to most other plant MIPs and is encoded on a separate exon. Due to the lack of generally conserved motifs in this region the first exon is often missing in annotations of NIP genes. However, within NIP3s of higher plants several motifs have been recognized in the N-terminal region [48] and some of these features are also conserved in PpNIP3;1. Similar to higher plants PpNIP3;1 has a high degree of proline and threonine residues and a sequence (AKCFP), corresponding to the conserved motif (C [KN]C [LF] [PS]) in higher plants.

Many NIPs in higher plants have a conserved potential phosphorylation motif in the C-terminal region corresponding to the phosphorylation site in Glycine max NOD26 (GmNOD26, S262) and Spinacia oleracea PIP2;1 (SoPIP2;1; S274) [5, 49]. A serine at this position is also present in a similar motif in NIP3s of higher plants ([RK]XXRSFXR) [48] but not in PpNIP3;1 where the serine is substituted to a valine. In PpNIP5;3 and PpNIP6;1 there are serines but some of the basic residues in the motif are not conserved. In contrast a corresponding serine in the motif (KXXKSF [HR]R) is present in PpNIP5;1 and PpNIP5;2 suggesting that at least some NIPs in a common ancestor of bryophytes and higher plants were regulated by phosphorylation.

It is interesting to see that there is no NIP2 type of MIP in P. patens, a NIP-group recently identified as a silicon transporter in rice [28]. Since bryophytes are known to accumulate silicon [50], the lack of PpNIP2s suggests that this function is carried out by a different isoform or class of proteins in P. patens.

Only SIP1s are found in Physcomitrella patens

In A. thaliana there are two classes of SIPs, SIP1s and SIP2s, both having the same gene structure with two introns at conserved positions [16]. In P. patens there are two SIPs but neither of them has an intron. Surprisingly both of the PpSIPs belong to the SIP1 group whereas SIP2s of higher plants form a basal clade. This suggests that either SIP2s were present already in early land plants but were subsequently lost in P. patens in which the remaining SIP1s were subject to intron loss, or that SIP2s have rapidly diverged from SIP1s after the split leading to mosses and higher plants. An intron loss in PpSIP1s or an intron gain in a common ancestor to SIP1s and SIP2s in higher plant is equally likely in this scenario. In most SIP1s the corresponding sequence to the first NPA box is NPT, interestingly this unusual motif is conserved also in PpSIP1s, implying that this is a structurally and functionally important feature of SIP1s. In addition the ar/R filter is consistent with the phylogenetic classification, suggesting a conserved function of SIP1s among terrestrial plants.

HIP a unique MIP with similarities to both PIPs and TIPs

There are three P. patens MIP sequences that cannot be classified into any of the five subfamilies previously described in plants [16, 20]. One of these, the PpHIP1;1, seems to be a rather rare MIP, since we were not able to identify any orthologs. The unique gene structure indicates that this protein belongs to a separate subfamily. In phylogenetic analyses PpHIP1;1 tend to cluster with PIPs and TIPs, although the support for this is not very strong as seen in Figure 2. Upon looking at the ar/R filter (Table 3) one could also speculate that the HIP is related to TIPs and PIPs, since it has histidines both at the H2 position, typical for TIPs and the H5 position, typical for PIPs. What effect having two large and basic amino acid residues in the filter will have on transport properties is however unclear, and since there are no ESTs of the gene it might even be that it is not expressed. According to a subcellular localization prediction (WoLF PSORT [51], data not shown) PpHIP1;1 is slightly more likely to reside in the tonoplast than the plasma membrane. Further studies are required to explore expression, localization and substrate specificity of the PpHIP.

The two other sequences belong to another group, the XIPs, further discussed in the next paragraph.

The XIP subfamily

A search for PpXIP orthologs resulted in the finding of many XIP sequences from a wide variety of species, including five paralogs from P. trichocarpa (probably the same five described as "putative aquaporins lacking in the Arabidopsis" by Tuskan et al. [52]). It is striking that no sequences are from monocots. Although most sequences were from dicots, no ortholog was found in A. thaliana, which may be explained by gene loss due to a relatively recent reduction of the genome size [53]. Phylogenetic analyses confirmed that these sequences are from a, to our knowledge, previously unrecognized MIP subfamily, different from PIPs, TIPs, NIPs, SIPs and GIPs. The only non-plant sequence included in the analyses was a protein encoded by the [GenBank:XM_639170] gene from the amoeba Dictyostelium discoideum AX4 and it should be pointed out that although this protein is clustering with the XIPs in phylogenetic analyses, it is annotated as a hypothetical protein and lacks some of the characteristics of the XIPs. For example the amoeba protein has NPA boxes and an ar/R filter different from all other XIPs and also an overall highly divergent MIP sequence, all which makes it questionable if this protein has the same function as other XIPs. There is also a sequence from a lycophyte, the spike moss Selaginella moellendorffii, which together with the two PpXIPs are the three most divergent sequences albeit all three are clearly categorisable as XIPs. Although most sequences were derived from ESTs, no general conclusion could be made on expression pattern, since XIP transcripts were isolated from many different tissues ranging from roots, seedlings, flower buds to seeds and fruits (Table 2). Based on a subcellular localization prediction XIPs are likely to be situated in the plasma membrane (WoLF PSORT [51], data not shown).

In the first NPA box of the XIPs, the alanine is replaced by a valine, leucine, isoleucine, serine or cysteine. All of these replacements, except isoleucine, have been observed in NPA boxes of other MIPs [47]. The most conserved feature of the new subfamily is located after the second NPA box, where a cysteine amino acid is thoroughly conserved in the motif NPARC. This cysteine is only a moderate change of the conserved serine or threonine found in many other subfamilies e.g. PIPs, TIPs, NIPs and in several mammalian AQPs. However, from the solved structure of SoPIP2;1 it is clear that residues at this position can stabilize the conformation of the C-loop by hydrogen bonds ([PDB:1Z98];S226 – N153, see Fig. 5) an interaction that seem to be structurally conserved and that also can be seen in BtAQP1 ([PDB:1J4N]; S198 – N129), BtAQP0 ([PDB:1YMG];S188 - N119) and, with the donor-acceptor interchanged, in EcGlpF ([PDB:1FX8];D207 - T137). This stabilisation is probably directly affecting the permeability of the pore since the orientation of the arginine of the ar/R filter is also stabilised by a hydrogen bond to the backbone of the C-loop (Fig. 5). Interestingly all the XIPs also have a conserved cysteine resulting in the motif LGGC in the C-loop at a position that can be aligned to N153 in SoPIP2;1. This suggests that a cysteine bridge may covalently fixate the C-loop relative to the arginine in the XIPs and that the extracellular entrance to the pore therefore might be more rigid than that of other MIPs.

There is also a highly conserved motif with a proline at the end of helix 2, 7 amino acids before the first NPA-box (PISGGHINP), also found in mammalian AQP5s. A corresponding motif can be found in helix 5 of many other plant MIPs, which is interesting as this reflects the symmetry of the MIP proteins, consisting of two direct repeats of sequence. It is also worth noting that, with the exception of PpXIPs, there is a lack of an otherwise highly conserved glycine in helix 5, allowing the close packing of helix 2 and 5 [54], which in most XIPs is replaced by either a leucine or an isoleucine. An alternative alignment that retains the conserved glycine, but introduces two extra amino acids between helix 5 and the second NPA box is possible, but not used in the analysis presented here. This alignment will also affect which amino acid is positioned in the H5 position of the ar/R filter (Table 3). In the chosen alignment a valine is the most frequent residue in the H5 position and in the alternative alignment threonine would be in the H5 position. At the H2 position most XIPs have an aliphatic amino acid, something that can also be found in some NIPs and SIPs [47]. This suggests that XIPs are not primarily water channels, although substrate specificity experiments have to be carried out to establish this. In the XIPs from P. patens and S. moellendorffii there is a glutamine at the H2 respectively H5 position of the ar/R filter, also found in TIP4s and TIP5s of higher plants, suggesting that maybe these TIPs have taken over some function of the XIPs in primitive plants. Further studies of localization, specificity and expression patterns are needed in order to determine the function of this novel MIP subfamily.

Gene identification and annotation

Physcomitrella patens MIP genes were identified by TBLASTN searches of the PpDB at the Joint Genome Institute [37] using the protein sequences of the complete set of 35 MIPs from Arabidopsis thaliana as queries [16]. Gene models overlapping with hits were manually inspected and kept based on subfamily sequence similarity or EST support. If no satisfying model existed, the genomic sequence was used to identify exons for the new or modified model (as specified in Table 1). The PpGIP1;1 sequence was also added to the sequences since it was previously identified as a PpMIP [20]. Protein sequences corresponding to the translation of the PpMIP genes were used in a second round of TBLASTN searches to identify more divergent MIP sequences in PpDB, but none were found. The resulting 23 PpMIPs were used in a multiple alignment of translated sequences, together with the 35 AtMIP and 33 ZmMIPs [18]. Alignments were manually inspected and adjusted and care was taken to keep the number of gaps low and to avoid gaps in functionally important features, such as the NPA-boxes and transmembrane regions. The alignment that forms the basis for all the phylogenetic analysis regarding the PpMIPs presented here is available as ALIGN_001168 in the EMBL-align database (which can be accessed either via the EMBL-EBI SRS homepage [55] or FTP [56]).

Orthologs of the unclassified PpHIP, PpXIP1;1 and PpXIP1;2 were searched for by TFASTX3 searches of the EMBL nucleotide sequence database [57] and TBLASTN searches of the nr/nt, est, gss and htgs databases at NCBI [58] using the translated sequence of the three PpMIPs. Translations representing hits from a wide variety of species were used in protein alignments together with either PpHIP1;1 or PpXIP1;1 and PpXIP1;2 and the PpPIPs and PpTIPs. The alignments were manually inspected and adjusted as mentioned above and used for phylogenetic analysis of PpHIP1;1 and the PpXIPs and are available in the EMBL-align database as ALIGN_001169 respectively ALIGN_001170.

The translated sequence of one of the PpXIP orthologs found [GenBank:EG656577] was used in additional TBLASTN searches of the nr/nt, est, gss and htgs databases at NCBI in order to find more homologs of this group. One ortholog found was from Populus trichocarpa and a translation of this sequence was used in a TBLASTN search of the P. trichocarpa genome at JGI to find paralogs. These paralogs together with a selection of homologs from the [GenBank:EG656577] and PpXIP searches were used in a multiple sequence alignment of translated sequences together with 22 PpMIPs (all except the PpHIP). The alignment was manually inspected and adjusted in the same manner as the PpMIP-AtMIP-ZmMIP alignment. This alignment forms the basis for all the phylogenetic analysis regarding the XIP group of MIPs and is available as ALIGN_001171 in the EMBL-align database.

Phylogenetic analysis

The PpMIP sequence alignment was analyzed by three different phylogenetic methods, Neighbour Joining (NJ), Maximum Parsimony (MP) and Bayesian inference (Bay). For all methods, gaps were treated as missing data. PAUP*4.0b10 [59] was used for the NJ and MP analysis. The default settings were used for both methods and bootstrapping with one thousand replicates for each method assessed the confidence of the best trees. Bayesian phylogenetic inferences were conducted using MrBayes 3.0.2 [60] using vague or uninformative prior probability distributions of the likelihood model under the JTT [61] +I+Γ model. Two sets of four parallel Metropolis Coupled Monte Carlo Markov Chains, of which three were heated with 0.2 temperature increments, were run for 2 million generations starting from random trees. Each 100th tree was sampled. The first 25 % of sampled trees was discarded as burn in, and stationary phase was empirically determined by looking at the likelihood scores of the kept samples. Robustness of the inferred tree was evaluated using Bayesian posterior probabilities. A "method consensus" tree was constructed as an overview, in this tree only branches that had a bootstrap or posterior probability support of more than 50 % in at least two of the methods were kept and all other were collapsed.

For the PpHIP1;1, PpXIPs and XIP-group alignments, PAUP*4.0b10 [59] was used for a NJ and MP analysis (gaps treated as missing data). The default settings were used for both methods and for the XIP-group alignment analysis, bootstrapping with one thousand replicates for each method assessed the confidence of the best trees. All trees from the PpMIP, PpHIP, PpXIPs and XIP family analyses are available in nexus format for viewing in Tree-View [62] [see Additional files 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14].