PPP Family of Protein Ser/Thr Phosphatases: Two Distinct Branches?

All known protein phosphatases specific for phosphorylated Ser and Thr are encoded by two unrelated gene families, termed PPM and PPP (Cohen 1997 ; Barford, Das, and Egloff 1998 ). The PPM family comprises Mg2+-dependent phosphatases structurally related to PP2C, while the PPP family has traditionally been subdivided into three subfamilies, related to PP1, PP2A, and PP2B (calcineurin). Investigation of these phosphatases has revealed their roles in such processes as glycogen metabolism, regulation of the cell cycle and RNA splicing, regulation of numerous protein kinases, T-cell activation, and memory formation (Cohen 1997 ; Klee, Ren, and Wang 1998 ; Millward, Zolnierowicz, and Hemmings 1999 ; Price and Mumby 1999 ; Aggen, Nairn, and Chamberlin 2000 ). These three extensively characterized “classical” PPP subfamilies have recently been supplemented by novel members from various species (Cohen 1997 ). In addition, a novel divergent group of PPP phosphatases called the PP5/rdgC subfamily has been recognized. It comprises enzymes related to mammalian and fungal PP5/PPT, protein phosphatases with EF-hand domains (PPEF, termed rdgC in Drosophila), and plant PP7 (Becker et al. 1994 ; Cohen 1997 ; Andreeva and Kutuzov 1999 ). Not only do phosphatases of the PP5/rdgC subfamily share relatively low sequence identity with the classical subfamilies, they are also more divergent when presumed orthologs from different species are compared. For example, catalytic domains of mammalian PP1 are 76%–88% and ∼90% identical to those from plants and fungi, respectively, while those of mammalian PP5 are only ∼65% and 50%–58% identical to their presumed counterparts from plants and fungi, respectively. If Trypanosoma species are compared with mammals, the catalytic domains of their PPEF phosphatases are only 39%–42% identical, while PP1 and PP2A enzymes are 71%–72% and 67% identical, respectively. This situation makes it difficult to ascribe novel divergent PPP phosphatases to the known PPP subfamilies and to anticipate their possible functions, especially when only relatively short incomplete sequences are available or there are no recognizable regulatory domains attached to the catalytic domain. For example, an expressed sequence tag (EST) from watermelon (accession number AA660123) has been annotated as similar to PPX, a PP2A-related phosphatase, while it in fact encodes a homolog of Arabidopsis thaliana PP7. The latest illustration of this complexity is the submission to the sequence databases of the primary structure of a novel protein phosphatase from Plasmodium falciparum (Q9U493), which shares, within its catalytic domain, 35%–39%, 37%–40%, 30%–38%, 38%–42%, 38%, and 36%–43% identity with PP1, PP2A, PP2B, PP5, PP7, and PPEF, respectively. This enzyme was classified by the authors of Q9U493 as “PP1-like.” However, phylogenetic analysis places this phosphatase (PP1Pf in fig. 1 ) in the PP5/rdgC subfamily. Phylogenetic analysis using either neighbor-joining (fig. 1 ) or parsimony-based (not shown) methods (Saitou and Nei 1987 ; Felsenstein 1999 ) indicates that the PP1/PP2A/PP2B subfamilies and the PP5/rdgC subfamily may represent two distinct branches of phosphatase evolution. Moreover, we suggest that several specific structural features distinguish the enzymes of the PP5/rdgC subfamily from those of the PP1, PP2A, and PP2B subfamilies. The validity of these observations, described below and illustrated in figure 2 , has been tested by BLAST-P (Altschul et al. 1997 ) searches at NCBI (http://www.ncbi.nlm.nih.gov/blast/blast.cgi) against the nonredundant protein database. Each of the eukaryotic PPP phosphatase sequences retrieved (ca. 250) was checked manually for the presence of the structural features described below. Entries annotated as “hypothetical protein,” “putative phosphoprotein phosphatase,” etc. were checked for their identity by BLAST searches to establish their relation to the known PPP subfamilies. If no closely matching (>65%–70% identity) phosphatase sequences were found, these entries were included in the phylogenetic analysis (fig. 1 ). We found the following distinctive structural features: A conserved motif, GlyAspPhe(Tyr)Val(Ala)AspArgGly, is preceded by Leu or Met in PP1, PP2A, and PP2B subfamilies and by an invariant Asn residue in all PP5/PPT, rdgC/PPEF (including PPEFTc and PP1Pf), and PP7 phosphatases (fig. 2 ). We were unable to find any exceptions to this rule. A charged residue is found in the middle of the αB helix (elements of the secondary structure are assigned according to Goldberg et al. [1995] ; see fig. 2 ) in the PP1/PP2A/PP2B-related phosphatases but not in the PP5/rdgC subfamily. The exceptions are PPT from Neurospora crassa (fig. 2 ), PP2A3 from Saccharomyces cerevisiae (P32345; Ronne et al. 1991 ), and several uncharacterized (pseudo)genes from Caenorhabditis elegans. An Ala residue (or Gly in S. cerevisiae PPQ, see fig. 2 ) found at the end of the β10 strand in PP1, PP2A, and PP2B subfamilies is substituted for Ser in all phosphatases of the rdgC/PP5 subfamily. This is an almost universal rule; the only exception is a PP2A-related phosphatase from Trypanosoma brucei (Q27787; Erondu and Donelson 1991 ) which has Ser in corresponding position. The distance between the conserved motifs GlyAspXHisGly and GlyAspPhe(Tyr)Val(Ala)AspArgGly is 24 residues in all PP5/PPEF/PP7 phosphatases (fig. 2 ). In most PP1/PP2A/PP2B-related enzymes, this distance is 23 residues. However, it may be longer in some cases: 25 residues in PP2B from Dictyostelium discoideum (Q27558; Dammann et al. 1996 ), 24 residues in Sit4Sc (P20604; Arndt, Styles, and Fink 1989 ), 29 or 34 residues in PPαPf (O96914; Li and Baker 1998 ) and related phosphatases from A. thaliana F21B7.27 (fig. 1 ), T4I9.4 (Q9ZTA4), T27G7.10 (Q9SJF0), and At2g27210 (Q9SHS7). Within this 23- or 24-aa region, position 13 is occupied by a conserved Gly. In most PP1/PP2A/PP2B enzymes, residue 12 is also Gly; the exceptions are yeast phosphatases PPQ and PPZ1 (see fig. 2 ), PPY (P11612; Dombradi et al. 1989 ) and PPN (O77294; Armstrong et al. 1998 ) from Drosophila melanogaster, a PP2A isoform from rice (Q9ZSS3), and a number of uncharacterized phosphatase (pseudo)genes and cDNAs from C. elegans. None of the PP5/PPEF/PP7 enzymes have Gly in this position. These sequence patterns are confined to the regions adjacent to the residues which coordinate metal ions in the active center (fig. 2 ). They may therefore reflect some differences in the active center structure between PP1/PP2A/PP2B and PP5/PPEF/PP7 phosphatases. Although the nature and significance of these differences will probably not become clear until the three-dimensional structure of a phosphatase of the latter group is solved, the above structural features can be used to tentatively assign new divergent sequences to one of these branches when complete structures are not available. The observations described in this report suggest that PP1/PP2A/PP2B and PP5/PPEF/PP7 groups are likely to represent two distinct branches which have diverged early in eukaryotic evolution, probably before the divergence of the PP1, PP2A, and PP2B subfamilies and before the acquisition of specific regulatory domains by different catalytic subunits (for diagrams of domain organization of different PPP phosphatases, see fig. 4B in Becker et al. [1994] , fig. 1C in Li and Baker [1998] , and fig. 2A in Andreeva and Kutuzov [1999] ). The latter notion is supported by the presence in both branches of phosphatases which have only catalytic but no regulatory domains and are probably regulated by separate regulatory proteins. In the PP1 and PP2A subfamilies, such phosphatases (e.g., PP1, PP2A, and PPX) have been known for a long time. In the PP5/PPEF/PP7 group, PP1Pf appears to have a similar “mono-domain” organization (which possibly led to its assignment to PP1), but a quite distinct primary structure of its catalytic domain, being most closely related to PPEF from T. cruzi (fig. 1 ). There have been at least two different routes to structural diversification of the PPP phosphatases in both branches. Many of them have acquired specific regulatory domains, probably by fusion of their genes with the genes of their cognate regulatory proteins. These fusion events should have occurred in phosphatases of different subfamilies after their structural divergence and functional specialization, since in all known cases regulatory domains in different subfamilies are unrelated. Detailed comparison of the putative regulatory domains in PPEF phosphatases suggests that the fusion of the catalytic domain of the ancestral form with an EF-hand Ca2+-binding protein occurred prior to acquisition of the N-terminal domains. Indeed, the N-terminal domains of PPEF from T. cruzi and those from animals are unrelated, while their C-terminal domains are more similar to each other than to any EF-hand proteins or EF-hand domains in any other enzymes (Andreeva and Kutuzov 1999 ). In addition to the acquisition of and/or C-terminal regulatory domains, there has been another trend in the evolution of PPP phosphatases of both branches: insertion of sequences of various lengths within the catalytic domains. Such inserts are found in some calcineurin isoforms (Barton, Cohen, and Barford 1994 ), in PPα from P. falciparum (O96914; Li and Baker 1998 ) and its uncharacterized plant homologs distantly related to the PP1 subfamily (represented by F21B7.27At in fig. 1 ), in PPEF phosphatases from animals (Montini et al. 1997 ; Sherman et al. 1997 ), and in PP7 from plants (Andreeva et al. 1998 ). To our knowledge, experimental evidence for a functional role of the inserts in the catalytic domains has only been reported for PP7 in A. thaliana. In this phosphatase, proteolytic cleavage of one of the inserts leads to its activation, suggesting that the insert may be an autoinhibitory region (Kutuzov, Evans, and Andreeva 1998 ). The evidence discussed above suggests that the protein phosphatase group called the PP5/rdgC subfamily is as heterogeneous as the combined PP1/PP2A/PP2B group and should rather be considered as consisting of at least three distinct subfamilies: PP5/PPT, probably ubiquitous in eukaryotes; rdgC/PPEF, not found so far in plants and fungi; and PP7, identified so far only in plants. Completion of genome sequencing projects should answer the question of whether two latter subfamilies are indeed kingdom-specific and clarify their origin. Claudia Kappen, Reviewing Editor Abbreviations: EST, expressed sequence tag; PP, protein phosphatase. Keywords: protein phosphorylation protein Ser/Thr phosphatases rdgC/PPEF, PP5/PPT, PP7 Address for correspondence and reprints: Research School of Biological and Molecular Sciences, Oxford Brookes University, Gipsy Lane Campus, Headington, Oxford OX3 0BP, United Kingdom. E-mail: p0071233@brookes.ac.uk. Fig. 1.—A dendrogram showing distance relationships between the catalytic domains of the representative sequences of phosphatases of the PP1, PP2A, and PP2B (calcineurin) subfamilies and most of the PP5/rdgC subfamily members. Multiple alignment was generated using CLUSTAL W (Thompson, Higgins, and Gibson 1994 ) at the European Bioinformatics Institute (http://www2.ebi.ac.uk/clustalw) and edited manually to remove inserts found in several phosphatases which had no counterparts in other sequences. The tree was constructed by the neighbor-joining method (Saitou and Nei 1987 ) using the PHYLIP package, version 3.573 (Felsenstein 1999 ). The categories model was used for calculation of protein distance. The tree was visualized using TreeView, version 1.5.2 (Page 1996 ). Bootstrap values (out of 1,000 replicates) are shown for the most important nodes. Scale bar = number of substitutions per site. Species designations: At, Arabidopsis thaliana; Ce, Caenorhabditis elegans; Dm, Drosophila melanogaster; Fr, Fugu rubripes; Hs, Homo sapiens; Mm, Mus musculus; Nc, Neurospora crassa; Pf, Plasmodium falciparum; Pt, Paramecium tetraurelia; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe; Tc, Trypanosoma cruzi; Xl, Xenopus laevis. SwissProt or TrEMBL accession numbers: F21B7.27At, AAF86526; PP2BPt, O15712; PPαPf, O96914 (Li and Baker 1998 ); PPEF2Mm, O35385 (Sherman et al. 1997 ); Sit4Sc, P20604 (Arndt, Styles, and Fink 1989 ). Accession numbers and references for other sequences are indicated in figure 2 . A PP5-related phosphatase from A. thaliana is designated PPU according to D. Horvath and N.-H. Chua (TrEMBL entry O22662) Fig. 2.—Distinctive motifs specific for the phosphatases of the PP1/PP2A/PP2B subfamilies and the PP5/rdgC subfamily. The residues which discriminate between these two groups of PPP phosphatases are boxed. Elements of the secondary structure are indicated according to Goldberg et al. (1995) . The residues which coordinate metal ions in the active center are marked with asterisks. References for the phosphatase sequences: Q08209—Muramatsu and Kincaid (1993) ; P23287—Cyert et al. (1991) ; P32945—Chen, Chen, and Cohen (1993) ; P26570—Posas et al. (1992) ; P30366—Nitschke et al. (1992) ; P08129—Song et al. (1993) ; P32598—Ohkura et al. (1989) ; P32838—Posas et al. (1993) ; P23594—Sneddon, Cohen, and Stark (1990) ; Q07098—Ariño et al. (1993) ; P05323—Stone et al. (1988) ; P48529—Perez-Callejon et al. (1993) ; P33172—Brewis and Cohen (1992) ; O00743—Bastians and Ponstingl (1996); ; O15920—Li and Baker (1997) ; O14428—Yatzkan and Yarden (1997) ; P53043—Chen et al. (1994) ; O42205—Ollendorff and Donoghue (1997) ; P53041—Chen et al. (1994) ; O49346—Andreeva et al. (1998) ; O61106—Andersson et al. (1998) ; P40421—Steele et al. (1992) ; O17047—Sherman et al. (1997) ; O15253—Montini et al. (1997) ; Q9W6R4—Brunner (1999) ; O14830—Sherman et al. (1997)