PTEN (phosphatase and tensin homologue deleted on chromosome ten) proteins are dual phosphatases with both protein and phosphoinositide phosphatase activity. They modulate signalling pathways controlling growth, metabolism and apoptosis in animals and are implied in several human diseases. In the present paper we describe a novel class of PTEN pro-teins in plants, termed PTEN2, which comprises the AtPTEN (Arabidopsis PTEN) 2a and AtPTEN2b proteins in Arabidopsis. Both display low in vitro tyrosine phosphatase activity. In addition, AtPTEN2a actively dephosphorylates in vitro the 3′ phosphate group of PI3P (phosphatidylinositol 3-phosphate), PI(3,4)P2 (phosphatidylinositol 3,4-bisphosphate) and PI(3,5)P2 (phosphatidylinositol 3,5-bisphosphate). In contrast with animal PTENs, PI(3,4,5)P3 (phosphatidylinositol 3,4,5-trisphosphate) is a poor substrate. Site-directed mutagenesis of AtPTEN2a and molecular modelling of protein–phosphoinositide interactions indicated that substitutions at the PTEN2 core catalytic site of the Lys267 and Gly268 residues found in animals, which are critical for animal PTEN activity, by Met267 and Ala268 found in the eudicot PTEN2 are responsible for changes in substrate specificity. Remarkably, the AtPTEN2a protein also displays strong binding activity for PA (phosphatidic acid), a major lipid second messenger in plants. Promoter::GUS (β-glucuronidase) fusion, transcript and protein analyses further showed the transcriptional regulation of the ubiquitously expressed AtPTEN2a and AtPTEN2b by salt and osmotic stress. The results of the present study suggest a function for this novel class of plant PTEN proteins as an effector of lipid signalling in plants.
In eukaryotes, PIs (phosphoinositides) have emerged as second messengers in many signalling pathways, controlling important aspects of cell physiology including growth, metabolism, apoptosis, membrane dynamics, organelle identity and cytoskeletal rearrangements . They are produced from the membrane lipid phosphatidylinositol and differ by the reversible phosphorylation of the inositol ring at its D3, D4 and D5 positions, giving rise to seven isomers with specific subcellular localizations and functions: PI3P (phosphatidylinositol 3-phosphate), PI4P (phosphatidylinositol 4-phosphate), PI5P (phosphatidylinositol 5-phosphate), PI(3,4)P2 (phosphatidylinositol 3,4-bisphosphate), PI(3,5)P2 (phosphatidylinositol 3,5-bisphosphate), PI(4,5)P2 (phosphatidylinositol 4,5-bisphosphate) and PI(3,4,5)P3 (phosphatidylinositol 3,4,5-trisphosphate) . PI signalling events are achieved by the binding of their negatively charged headgroup to specific proteins, modifying their subcellular localization and/or their activity.
Among the plethora of PI-metabolizing enzymes constituted by kinases, phospholipases and phosphatases, PTEN (phosphatase and tensin homologue deleted on chromosome 10; EC 188.8.131.52) has received massive attention since its discovery in 1997 for its function in human cancer (it is the most mutated gene after p53) . PTEN harbours the specific catalytic motif (I/V)HCXAGXXR of protein tyrosine phosphatases and possesses both protein and lipid phosphatase activities. PTENs tumour suppressor activity has been attributed mainly to its role as a negative regulator of the PI3K (phosphatidylinositol 3-kinase)/Akt signalling pathway [4,5], which is a major growth and survival pathway in animals. In this pathway, PTEN antagonizes PI3K by removing the 3-phosphate from PI(3,4,5)P3 , which is the activator of downstream effectors like the serine/threonine kinase Akt, a proto-oncogenic protein. In addition, a recent study revealed that PTEN protein phosphatase activity or PTEN interaction with protein partners may be crucial for several physiological processes, independently from its lipid phosphatase activity .
In plants, the first homologue of HsPTEN (human PTEN), AtPTEN (Arabidopsis PTEN) 1, has been characterized in Arabidopsis thaliana . Like HsPTEN, the recombinant AtPTEN1 displays in vitro PI(3,4,5)P3 dephosphorylation activity. The RNAi (RNA interference) silencing of AtPTEN1, which is specifically expressed in pollen grains, leads to pollen cell death and plant sterility. However, the physiological relevance of this severe phenotype has not been established yet. It is unlikely to be due to PI(3,4,5)P3 dephosphorylation since neither this substrate nor the class I PI3K responsible for its synthesis have been identified in plants. Although additional proteins with PTEN N-terminal phosphatase domains can be identified in plants, no characterization of their biochemical activity has been published to date. Advances in plant biology have triggered a renewed interest in plant PTENs. First, the crucial role played by lipid signalling in plant growth and development and responses to biotic and abiotic stress is now well established [9–14]. PTEN appears to be a likely effector of phospholipid signalling in plants, given its PI phosphatase activity in animals, which targets the D3 phosphate of the PI3P, PI(3,4)P2 and PI(3,4,5)P3 substrates [15,16]. Secondly, additional evidence of the function of PTEN domains has been obtained in plants, with the recent implication of such a domain in the apical localization of the moss For2 formin required for rapid elongation of actin filaments .
In the present paper we report the identification and characterization of new plant PTEN proteins from Arabidopsis containing PTEN phosphatase and C2 domains called AtPTEN2a and AtPTEN2b. We demonstrated the phospholipid activity of AtPTEN2a, which preferentially uses PI3P, in contrast with HsPTEN. We further showed that substitution of residues at the core catalytic site with those found in HsPTEN increased AtPTEN2a activity and modified its substrate specificities. Modelling of protein–PI interactions provided likely explanations for preferential substrate use. Promoter, transcript and protein analyses further indicated that AtPTEN2a and AtPTEN2b are differentially regulated in the plant in response to environmental constraints such as salinity and osmotic stress. Moreover, protein–lipid-binding experiments demonstrated that AtPTEN2a effectively binds the signalling lipid PA (phosphatidic acid). The results of the present study obtained in A. thaliana provide new insights into the evolution of PTEN genes from Arabidopsis and open the way for the study of their physiological function in planta.
Plant materials and growth conditions
Seeds from A. thaliana (L.) Heynh. (ecotype Columbia-0) were surface-sterilized, stratified and either plated on 0.5× MS (Murashige and Skoog) medium supplemented with 0.8% agar and 1% (w/v) sucrose, or placed in 0.5× liquid MS medium supplemented with 2% (w/v) glucose and 20 mM Mes (pH 5.7) on a rotary shaker (120 rev./min). Seedlings (15-day-old) grown in liquid culture were osmotically stressed by transferring the plantlets to a new flask containing fresh medium supplemented with 0–200 mM NaCl or mannitol. For cultivation of adult plants, seedlings were transferred from the 0.5× MS solid medium to a soil/vermiculite mixture. For all experiments, the plants were cultivated at 22°C under a daily 16 h light period at 150 μmol of photons·m−2·s−1.
Sequence analysis and phylogeny
PTEN sequences were recovered from several databases and mainly from sequenced genomes (database accession numbers are given in Supplementary Table S1 at http://www.BiochemJ.org/bj/441/bj4410161add.htm). TAIR (The Arabidopsis Information Resource; http://www.arabidopsis.org/) accession numbers for the Arabidopsis PTEN genes were At5g39400 (AtPTEN1), At3g19420 (AtPTEN2a) and At3g50110 (AtPTEN2b). Only the amino acid sequences of the highly conserved phosphatase domain (170–195 amino acids) were retained for analysis since the sequences of the C2 domain were too divergent. All sequences were edited and aligned manually using BioEdit . The final alignment included six indels of 1–30 amino acids that were encoded as additional character states (absent, 0 and present, 1), regardless of their length. They were included in subsequent sequence analyses to exploit their phylogenetic information following the procedure of Barriel . Phylogenetic relationships were estimated by MP (maximum parsimony) and BML (Bayesian maximum likelihood) methods. MP analysis was first conducted using PAUP* v4.0b10 . The data matrix was analysed with the heuristic search and the default search options. Support of internal nodes was assessed using bootstrap analysis of 10000 replicates. BML analysis of the same dataset was performed using the MCMC (Metropolis-coupled Markov chain Monte Carlo) simulation program MrBayes v3.1 . The indel events were included as a separate binary (restriction) data partition. Model parameters for each partition were estimated separately during the MCMC process. BML analyses were run for 5000000 generations with sampling every 100 generations to produce 50000 sampled trees. Trees sampled before stationarity was reached (burnin option=generation 5000) were discarded and a 50% posterior probability Bayesian consensus tree was generated from the remaining trees. The frequency of a clade in the consensus tree represents the posterior probabilities of that clade.
Expression in bacteria and purification of recombinant proteins
The Arabidopsis At3g19420 (AtPTEN2a) and At3g50110 (AtPTEN2b) genes were PCR-amplified from the RAFL09-38-D10 cDNA clone (RIKEN BioResource Center, Japan) and C105193 cDNA clone (The Arabidopsis Biological Resource Center, Ohio, U.S.A.) respectively. The primers used for amplification contained gene-specific sequences flanked by vector-specific sequences required for cloning into pET Ek/LIC expression vectors (Novagen; the primer sequences used are detailed in Supplementary Table S2 at http://www.BiochemJ.org/bj/441/bj4410161add.htm). The generated PCR fragments were purified, treated with T4 polymerase and ligated into pET Ek/LIC vectors following the manufacturer's protocol. The full AtPTEN2a open reading frame was cloned in pET-30 Ek/LIC to produce an in-frame N-terminus His-tag fusion giving rise to pLIC30-AtPTEN2a. The AtPTEN2b ORF (open reading frame) was cloned in pET-41 Ek/LIC to produce in-frame N-terminus His- and GST (glutathione transferase)-tag fusions giving rise to pLIC41-AtPTEN2b. The site-directed mutagenesis (pLIC-AtPTEN2aC263S and pLIC-AtPTEN2aM267K/A268G) was performed using the QuikChange® kit (Stratagene) according to the manufacturer's instructions. Pfu DNA polymerase (Stratagene) was used in all PCR reactions and all constructs were verified by sequencing. The constructs above were introduced into the BL21 (DE3) strain of Escherichia coli (Invitrogen) to express the corresponding recombinant proteins. Transformed cells were cultured in 0.5 litres of Terrific broth medium containing 30 μg/ml kanamycin at 37°C. When a D600 value of 0.6 was reached, IPTG (isopropyl β-D-thiogalactopyranoside) was added to a final concentration of 1 mM and incubation was continued for 4 h at 37°C. Subsequent steps were carried out at 4°C. After centrifugation (6000 g for 20 min), the cells were washed, resuspended in 15 ml of lysis buffer [50 mM Tris/HCl (pH 7.5), 1 mM DTT (dithiothreitol), 1 mM PMSF and 500 mM NaCl] containing 1 mg/ml lysozyme and EDTA-free protease inhibitors (Roche Applied Science) and incubated at 30°C for 1 h. After sonication, the soluble fraction was obtained by centrifugation (20000 g for 20 min at 4°C). The supernatant was then loaded on to a 1 ml HisTrapTM HP column using an Äkta Purifier (GE Healthcare). After sample loading, the column was washed with lysis buffer containing 20 mM imidazole and the protein was eluted with the same buffer containing 250 mM imidazole. The eluted fraction was then diluted (1:4 dilution) in 20 mM Tris/HCl (pH 7.5), 1 mM DTT and 10% (v/v) glycerol. The protein was further purified by ion exchange using a MonoQ column HR 5/5 (GE Healthcare) equilibrated with the dilution buffer described above. The protein was eluted with a NaCl gradient ranging from 100 to 500 mM over 20 column volumes. Finally, purified proteins were frozen in liquid N2 and stored at −80°C after adjusting the glycerol concentration up to 50%.
Lipid phosphatase activity using Malachite Green assays
PTEN hydrolysing activity towards the different PIs was determined in vitro by the Malachite Green-based assay that measures the released Pi . Briefly, 2 μg of recombinant protein were incubated with 100 μM of one of the seven di-C8 PI lipids [PI3P, PI4P, PI5P, PI(3,4)P2, PI(3,5)P2, PI(4,5)P2 or PI(3,4,5)P3] (Echelon Biosciences) for 40 min (AtPTEN2a and AtPTEN2aM267K/A268G proteins) or 180 min (AtPTEN2b protein) at 30°C in a final volume of 25 μl. The reaction buffer was 100 mM Tris/HCl (pH 8.0) containing 10 mM DTT. The reaction was stopped with 100 μl of Malachite Green reagent (Echelon Biosciences). The amount of phosphate released was determined by reading the absorbance at 660 nm and converted into a molar amount of product using a standard curve. The amount of protein used and the duration of the assay were within the linear range of the reaction. For the reaction with PI3P, 8% of the substrate was consumed at the end of the reaction.
The different models of the phosphatase domain of Arabidopsis proteins have been built using the sequence alignment in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/441/bj4410161add.htm) and the software Modeller 9 version 2 (http://salilab.org/modeller/) using the default options. In each case, 20 models were generated and the five lowest-energy conformers were kept for further investigation. The PI-binding assays were performed using InsightII software programs, mainly Discover and the CVFF (constant valence force field). Point mutations were built using the Homology Module of InsightII. The general protocol used throughout is the following: the starting point is the X-ray structure of HsPTEN (PDB code 1D5R) with PI(3,4,5)P3 (PIP3) docked as described by Lee et al.  by superimposing the phosphates of PIP3 on the carboxylates of tartaric acid and then deleting the latter. PIP3 was extracted from the X-ray structure of the pleckstrin homology domain of the kinase B/AKT (PDB code 1H10). The PI3P molecule was built from PIP3, by deleting the corresponding phosphate groups. The first step was a multi-step minimization: (i) the backbone of the protein and the whole PIP3 (or PI3P) are kept fixed and all side-chains are fully minimized; (ii) PIP3 (or PI3P) is then released and full minimization is resumed until the RMSD (root mean square deviation) is <0.01. Then, the complex PIP3 (or PI3P)–protein is submitted to a 100 ps molecular dynamics run at 1000 K (timestep=1 fs) with the protein backbone fixed. The complex is fully minimized first with the protein backbone fixed then with the protein backbone tethered with 200, 100 and 50 kcal/Å2 (1 Å=0.1 nm) constraints. Finally, the complex was submitted to a 200 ps molecular dynamics run at 500 K (timestep=1 fs) with the protein backbone tethered with a 50 kcal/Å2 constraint, and 400 conformers were stored. Following a cluster graph, the two or three best solutions are submitted to full minimization and are retained for further investigation. For the plant PTEN, the starting point is obtained by superimposition of the modelled structure of the unknown PTENs on the structure of HsPTEN with PIP3 (or PI3P), then the PIP3 (or PI3P) molecule is transferred between the two proteins and HsPTEN is deleted, leaving PIP3 (or PI3P) docked within the studied protein.
The 1.16 kbp sequence of the AtPTEN2a gene promoter and the 1.27 kbp sequence of the AtPTEN2b gene promoter were amplified from genomic DNA by using specific primers (Supplementary Table S2) and fused to a GUS (β-glucuronidase) reporter gene in the vector pGreenIIK-GUSORF-TerNos (http://www.pgreen.ac.uk) between the KpnI and SalI sites. A. thaliana plants were transformed with Agrobacterium tumefaciens GV3101 carrying the constructs by using the ‘floral-dip’ method. GUS-activity staining was performed according to Sessions et al. .
Plant material (0.2 g) was homogenized and extracted in 0.1 ml of 50 mM Tris/HCl (pH 8.0), 1 mM EDTA, 0.1% SDS and 0.1% Triton X-100. The extracts were cleared by centrifugation at 13000 g for 15 min at 20°C and conserved at −80°C. SDS/PAGE and Western blotting were performed using an ECL Western blotting Kit (GE Healthcare) according to the manufacturer's protocol. The membranes were hybridized with a 1:5000 dilution of primary antibodies. Rabbit polyclonal antibodies against the FASTQNDSGVENTEK peptide for AtPTEN2a and the EPPKRDDPSANRSVT peptide of AtPTEN2b were from Eurogentec. Specificity and absence of cross-reactivity of the antibodies were further assessed by probing available pten2a and pten2b Arabidopsis insertional T-DNA mutants (results not shown).
For protein–lipid overlay, membrane strips pre-spotted with 100 pmol of lipids (Echelon Biosciences) were first blocked with 3% (w/v) fat-free BSA in TBST [Tris-buffered saline plus Tween 20; 10 mM Tris/HCl (pH 8.0), 150 mM NaCl and 0.1% Tween 20) and incubated with 0.5 μg/ml AtPTEN2a protein solution in the blocking buffer overnight at 4°C. The strips were washed with TBST, and PTEN2a was subsequently revealed using immunological revelation as described above. To detect lipid binding in vesicles, lipid mixtures consisting of 45% PC (phosphatidylcholine), 45% PE (phosphoethanolamine) and 10% LOI (lipid-of-interest) were used to produce vesicles sized 200 nm using lipid extrusion, as described by Testerink et al. . Dipalmitoyl PIPs were from Echelon Biosciences and all other lipids were obtained from Avanti Polar Lipids. Two total lipid concentrations were tested, 4 mM and 1 mM, corresponding to 200 μM and 50 μM of the LOI theoretically available for binding (with headgroups facing outside of the vesicle) respectively. AtPTEN2a protein (200 ng) was added to the vesicles and incubated for 1 h. Samples were centrifuged at 16000 g for 30 min at 22°C to spin down the vesicles, and the pellets were washed once and analysed by SDS/PAGE (10% gel) to detect protein bound to the vesicles. The gels were stained with colloidal Coomassie Brilliant Blue. The recombinant 2×FYVE domain tagged to GST–YFP (yellow fluorescent protein)  was used as a control for PI3P binding.
Transcript level analysis by real-time quantitative PCR
Total RNA was extracted from Arabidopsis tissues with the RNeasy plant mini kit (Qiagen). Purified RNA was treated with DNAse I using the DNA-free kit (Ambion). First-strand cDNA was prepared from 1 μg of total RNA with the Superscript RT II kit (Invitrogen) and oligo(dT)18 according to the manufacturer's instructions. A 0.2 μl aliquot of the total reaction volume was used as a template in real-time RT (reverse transcription)-mediated PCR amplification. The PCR amplification was performed as described by Joubès et al.  with gene-specific primers (Supplementary Table S2). The relative abundance of ACT2, eIF-1α and eIF-4A-1 in each sample was determined and used to normalize differences of total RNA amount according to the method described by Vandesompele et al. .
Arabidopsis contains three PTEN-like genes
PTEN proteins are characterized by two domains: an N-terminal phosphatase domain with the catalytic core signature motif of protein tyrosine phosphatases (I/V)HCXAGXXR and a C-terminal C2 domain well-conserved among the metazoans which is critical for membrane binding [28,29]. Using HsPTEN and Drosophila PTEN protein as query sequences, we identified three PTEN homologues in the Arabidopsis genome and a higher number of proteins with PTEN domains, e.g. the For2 formin . The AtPTEN1 protein has been biochemically characterized , whereas the two other PTEN proteins found in Arabidopsis, named AtPTEN2a and AtPTEN2b (corresponding to the At3g19420 and At3g50110 genes respectively) have not yet been studied (Figure 1A). The AtPTEN2a and AtPTEN2b genes encode putative proteins of 611 and 632 amino acids respectively, which are much larger than the HsPTEN and AtPTEN1 proteins (Figure 1B). All three Arabidopsis PTEN proteins present a phosphatase domain containing the catalytic protein tyrosine phosphatase motif (InterPro accession number IPR014019 Phosphatase_tensin-typ) showing 67–69% similarity with its human relative, and a C2 domain (InterPro accession number IPR014020 Tensin_phosphatase_C2-dom). The organization of these domains is conserved between human, Drosophila and Arabidopsis.
Representation of the PTEN genes and proteins from A. thaliana
Phylogenetic analyses of PTEN
In order to obtain more insights into the functional relationships of PTEN proteins in plants, we performed phylogenetic analyses of all PTEN proteins found in plants. A BLAST homology search using the highly conserved phosphatase domain from HsPTEN (181 amino acids) allowed recovery of PTEN sequences from angiosperms and gymnosperms, which were compared with lycophytes (Selaginella moellendorffii), green algae (Chlamydomonas reinhardtii) and animal PTENs (Supplementary Table S1). The matrix used for phylogenetic analyses includes 182 characters, of which 30 were constant and 125 were phylogenetically informative. Parsimony analysis resulted in 96 equally most parsimonious trees of 555 steps, with a CI (consistency index) of 0.757 and a highest RI (retention index) of 0.866. The result of phylogenetic analysis using the Bayesian method is shown in Figure 2(A). The strict consensus tree, which resulted in similar tree topology, is presented in Supplementary Figure S2 (at http://www.BiochemJ.org/bj/441/bj4410161add.htm). Within these trees, the available sequences from animals are together, whereas those representing various lineages of tracheophytes (including angiosperms, gymnosperms and lycophytes) are divided into two distinct and strongly supported monophyletic groups (posterior probability of 1.0 and bootstrap support of 100%), hereafter named PTEN1 (including AtPTEN1) and PTEN2 (including AtPTEN2a and AtPTEN2b). Within the PTEN1 clade, one sequence of S. moellendorffii (lycophytes) is placed as sister to all available sequences from gymnosperms and part of the eudicot sequences. Within the PTEN2 clade, the other S. moellendorffii sequence is placed as sister to all remaining monocot and dicot sequences. The only sequence representing the green algae (C. reinhardtii) is more closely related to that of the PTEN2 than to that of the PTEN1 clade.
Phylogenetic analysis of the phosphatase domain of PTEN
Interestingly, in comparison with the catalytic core motif HCKAGKGR found in animal PTENs, the translated sequences from the moss S. moellendorffii and monocots showed the typical PTEN signature, i.e. Cys124 and Arg130 (HsPTEN numbering) for catalysis and the two lysine residues Lys125 and Lys128 for lipid substrate-binding. Sequences from eudicots presented notable changes (Figure 2B). For the PTEN1 clade, Lys125 was replaced by a methionine residue and for the PTEN2 clade, Lys128 and Gly129 were substituted by methionine and alanine residues respectively. Considering the crucial role played by these amino acids in PTEN activity and substrate specificity , we hypothesized that such changes may alter the biochemical properties of plant PTEN2 and we therefore undertook the biochemical characterization of the eudicot PTEN2 from Arabidopsis.
Purification of recombinant Arabidopsis PTEN2 proteins from E. coli
HsPTEN was first described as a protein tyrosine phosphatase. It was later acknowledged as a preferential lipid phosphatase using phosphatidylinositol substrates such as PI(3,4,5)P3 . Both activities are conserved in AtPTEN1 . To address the function of AtPTEN2a and AtPTEN2b, we expressed the recombinant proteins in E. coli as His-tag (AtPTEN2a) or His–GST tag (AtPTEN2b) fusion proteins. Initial attempts to purify the proteins using affinity columns were largely unsuccessful because large aggregates were formed with other proteins, especially for AtPTEN2b. Affinity chromatography followed by ion-exchange chromatography eventually succeeded in the purification of His6–AtPTEN2a (and two mutated AtPTEN2a proteins obtained by site directed mutagenesis) and His6–GST AtPTEN2b fusion proteins (Figure 3A). SDS/PAGE analysis of the purified proteins indicated a size of approximately 120 kDa for the AtPTEN2a and 130 kDa for AtPTEN2b proteins, whereas the calculated mass of their translated products is 72 kDa and 100 kDa respectively. This was attributed to unusual gel-migration patterns, possibly due to the structure of the AtPTEN2 protein domains, since MS analysis of the purified AtPTEN2a confirmed its PTEN identity (Supplementary Figure S3 at http://www.BiochemJ.org/bj/441/bj4410161add.htm).
Biochemical characterization of PI phosphatase activity of recombinant AtPTEN2a and AtPTEN2b proteins
Substrate specificities of AtPTEN2 phosphatases
Because the functional roles of PTEN are largely linked to their lipid phosphatase activity , we next investigated the phosphatase activity of AtPTEN2 towards various PI substrates as described in Merlot et al. . Among monophosphorylated phosphatidylinositol isomers (PIPs), the recombinant AtPTEN2a enzyme efficiently removed phosphate from the D3 position of the inositol ring, less from the D4 position and not at all from the D5 position (Figure 3B). In the PIP2 isomers, the presence of a phosphate group in the D5 position led to a consistent reduction in AtPTEN2a lipid phosphatase activity. Strikingly, in contrast with animal PTEN  and with Arabidopsis AtPTEN1 , AtPTEN2a showed a very weak activity against PI(3,4,5)P3 as substrate. The recombinant GST–AtPTEN2b protein displayed no detectable phospholipid activity, except a weak one towards PI3P. AtPTEN2b was therefore excluded from further detailed characterization.
We next analysed the impact of mutations on those amino acids that may play a role in PI dephosphorylation. As expected from previous studies on human PTEN , mutation of the catalytic Cys263 to a serine residue abolished AtPTEN2a activity independent of the substrate assayed (results not shown). In eudicot PTEN2, the Met267 residue replaces the lysine found in the core catalytic site of all other PTEN proteins, including PTEN2 from monocots. The Ala268 residue found in AtPTEN2 proteins replaces, in both eudicot and monocot PTEN2, the glycine found in other PTEN proteins. Both residues could be involved in the differential use of PI substrates by AtPTEN2a in comparison with that reported for human PTEN. We therefore selected Met267 and Ala268 for mutation, according to the phylogenetic and sequence analyses done previously (Figure 2). These mutations had a remarkable effect on AtPTEN2a phospholipid activity. The AtPTEN2a M267K/A268G protein lost its activity for PIP substrates, whereas activity for PI(3,4)P2, PI(3,5)P2 and most notably for PI(3,4,5)P3 was greatly enhanced, 5, 5 and 40 times respectively compared with the wild-type AtPTEN2a (Figure 3C).
In addition, because animal PTENs are described as DSPs (dual-specificity phosphatases) with the ability to dephosphorylate lipid substrates and also phospho-tyrosine, phospho-serine and phospho-threonine residues, we assayed AtPTEN2a tyrosine phosphatase activity using p-NPP (p-nitrophenyl phosphate) as a substrate . The recombinant AtPTEN2a hydrolysed p-NPP poorly, the affinity for this substrate being in the millimolar range (Km=8.5 mM; Supplementary Figure S4 at http://www.BiochemJ.org/bj/441/bj4410161add.htm). Likewise, the HsPTEN affinity for p-NPP is low and is in the millimolar range (Km=25.6 mM) .
AtPTEN modelling studies of the phosphatase domain
In order to provide additional information on the relationship between the Arabidopsis PTEN structure and function, different models of AtPTEN1, AtPTEN2a and AtPTEN2b proteins were built using the sequence alignment shown in Supplementary Figure S1 as described in Experimental section. In each case, 20 models were generated and the five lowest-energy conformers were kept for further investigation. Superimposition of the three models of AtPTEN on the X-ray structure of HsPTEN (PDB code 1D5R) revealed weak differences in the structure of the phosphatase domain (Figure 4A). AtPTEN2a and AtPTEN2b differ mainly from AtPTEN1 and HsPTEN by a single unstructured lipophilic loop of eight amino acids (Figure 4A and Supplementary Figure S1). The two amino acid deletion in AtPTEN1 (Supplementary Figure S1) seems to have a minimal effect on the structure of the domain. MLPs (molecular lipophilicity potentials) were calculated at the atomic level for HsPTEN and the three models of AtPTEN 1, 2a and 2b (see the Experimental section of the Supplementary online data at http://www.BiochemJ.org/bj/441/bj4410161add.htm). From this experiment, an increase in lipophilicity is clearly seen from AtPTEN1 (almost equivalent to HsPTEN) to AtPTEN2a and AtPTEN2b (Supplementary Figure S5 at http://www.BiochemJ.org/bj/441/bj4410161add.htm). This result corroborates perfectly the difficulties encountered during purification on affinity columns as stated above.
Modelling of AtPTEN phosphatase domains
Docking of PI(3,4,5)P3 was explored from models of HsPTEN, AtPTEN2a and the AtPTEN2a mutant M267K/A268G (Figure 4B). The interactions observed were the following: (i) for HsPTEN P1 with Lys125 and His93, P3 with Lys128 and Arg130, P4 with Lys128, and P5 with Lys128; (ii) for AtPTEN2a P1 no interaction, P3 with Lys264 and Arg269 (respectively equivalent to Lys125 and Arg130 in HsPTEN), and P4 with Lys303 and P5 with Lys303 (no equivalent residue for HsPTEN Lys128 present in this protein, as it contains Met267); (iii) for the mutated version AtPTEN2a M267K/A268G P1 with Lys264 and Lys269 (equivalent to Lys125 and Arg130 in HsPTEN), P3 with Arg269 (equivalent to Arg130 in HsPTEN), P4 with Lys303 and Lys267 (equivalent to HsPTEN Lys128), and P5 with Lys267 (equivalent to Lys128).
Even if the location is not perfectly identical, in the AtPTEN2a M267K/A268G mutant, PI(3,4,5)P3 is engaged in the same interactions as in HsPTEN. On the other hand, in wild-type AtPTEN2a, the absence of lysine at position 267 leads to a reorientation of PI(3,4,5)P3 and to an interaction with Lys303. The same experiment has been done with PI3P in AtPTEN2a and AtPTEN2a M267K/A268G models. The AtPTEN2a protein displays interactions between P1 and Lys264 (one interaction), between P3 and Arg269 (three interactions), and between P3 and Lys264 (one interaction). The M267K/A268G AtPTEN2a mutant displays interactions between P1 and Lys267 (one interaction), between P3 and Arg269 (two interactions), between P3 and Lys267 (one interaction), between P3 and Lys264 (one interaction). This clear result states that Lys267 has replaced Lys264 for P1 and has replaced one of the interactions with Arg269. Moreover (and even if there are two different amino acids) the energy of the wild-type protein is largely lower than the mutant. The consequence of the mutation is a large displacement of PI3P within the active site as seen in Figure 4(C): the distance between the respective P3 is only 1.5 Å, but reaches 4.75 Å for the respective P1 which may easily explain the absence of reactivity for the M267K/A268G mutated AtPTEN2a.
AtPTEN2a affinity for phospholipid membranes in vitro
As PIs are membrane components, we investigated the ability of AtPTEN2a to be recruited to membranes. Surprisingly, protein–lipid overlay assays with commercial pre-spotted pure lipids revealed a strong binding to PA (Figure 5A), whereas no binding was visible for the other lipids. In order to investigate further this binding property, AtPTEN2a was incubated with liposomes containing a mixture of PC, PE and 10% of LOI:PA, PS (phosphatidylserine), PI3P, PI4P or PI5P. The SDS/PAGE analysis of AtPTEN2a bound to vesicles confirmed the result obtained by protein–lipid overlay analysis with PA at both LOI concentrations (Figure 5B) and identified a weak but clear binding for PS, even visible at 50 μM LOI. No binding was obtained for liposomes containing only PC and PE or liposomes containing PC/PE and various PIs. The recombinant 2×FYVE domain fused to GST–YFP was used in parallel as a positive control for PI3P binding  and bound specifically the liposomes containing PI3P.
Lipid-binding assays of recombinant AtPTEN2a
Expression patterns of AtPTEN2 genes in planta
To analyse the activity of AtPTEN2 promoters in planta, we produced A. thaliana plants stably transformed with AtPTEN2a and AtPTEN2b promoter::GUS transcriptional fusions. GUS staining of plant organs at several stages of development revealed that both genes were expressed in a wide range of plant organs (Figure 6A) and that activity of the AtPTEN2a promoter was always higher. For both genes, young tissues were more intensively stained as observed in seedlings, leaves, flowers and siliques. GUS staining results were consistent with transcript accumulation patterns obtained by semi-quantitative RT–PCR (Figure S6A) and with the Genevestigator database (https://www.genevestigator.com/gv/index.jsp). The latter further indicated that AtPTEN2a transcript abundance is 3–5-fold higher than that of AtPTEN2b independent of the plant tissue studied (Figure S6B). In contrast, AtPTEN1 transcript accumulation is very low except in reproductive tissues where it is restricted to pollen as described previously .
Analysis of AtPTEN2a and AtPTEN2b expression in different organs during plant development
We next produced antibodies specific against either the AtPTEN2a or AtPTEN2b isoforms and analysed their protein expression patterns in plant tissues by Western blotting (Figure 6B). AtPTEN2a protein, like the AtPTEN2a transcript, was observed in all plant organs, but was not detectable any more in the mature leaves or siliques. Surprisingly, given its transcript accumulation pattern, the AtPTEN2b protein accumulated to higher levels in seedlings and in stem terminal nodes. Taken together, these results support a role for AtPTEN2 throughout plant development, whereas AtPTEN1 function would be more restricted to pollen. Discrepancies between promoter activity, transcript abundance and protein levels observed in several organs or physiological stages further suggest a tight post-transcriptional regulation of both AtPTEN2a and AtPTEN2b.
Since PI signalling is crucial for plant responses to stress [10,13], we next investigated whether AtPTEN2 expression was altered in response to various stresses including dehydration, salt (NaCl) and osmotic (mannitol) stresses. No significant changes were observed for 15-day-old plantlets submitted to 24 h of dehydration (results not shown), whereas only a weak response was observed to salinity (NaCl) stress (Figure 7A). Conversely, osmotic stress triggered by mannitol concentrations ranging from 100 to 200 mM considerably enhanced AtPTEN2a transcript abundance, and to a much lower level that of AtPTEN2b (Figure 7B). This was not followed by parallel protein accumulation since AtPTEN2a and AtPTEN2b protein levels remained stable in leaves and only showed a weak increase in roots after a 48 h exposure to mannitol at 150 mM (Figure 7C).
Expression of AtPTEN2a and AtPTEN2b in response to NaCl and mannitol
Animal PTENs are DSPs that can dephosphorylate both phospho-tyrosine and phospho-serine/phospho-threonine residues within one substrate. In addition, animal PTENs have lipid phosphatase activity that targets the D3 phosphate from PIs. Two domains are known to play important roles in PTEN functions: the phosphatase domain and the C2 domain. The protein also contains a short N-terminal PI(4,5)P2-binding domain which is crucial for the binding of PTEN to membrane lipids . We have identified in the plant model Arabidopsis three PTEN proteins displaying the canonical PTEN phosphatase domain, which includes the catalytic core signature motif of protein tyrosine phosphatases, (I/V)HCXAGXXR, and the C2 domain (Figure 1B). Phylogenetic analyses of PTEN sequences from plant and animal kingdoms suggest a duplication event of the PTEN gene early during the evolutionary history of plants. As a consequence, two distinct monophyletic groups can be clearly identified in the vascular plants (Figure 2A). The PTEN1 clade comprises sequences from the angiosperms, gymnosperms and lycophytes. The Arabidopsis AtPTEN1 gene, which belongs to the PTEN1 clade, has been recently shown to display a PI(3,4,5)P3-dephosphorylating activity and to strongly affect pollen viability when inactivated by RNAi silencing . However, neither the PI(3,4,5)P3 PI or the type I PI3K kinase which catalyses its formation from PI(4,5)P2 have ever been identified in plants, thus questioning the mode of action of PTEN1 in plants. The PTEN2 clade includes sequences from the angiosperms (eudicots and monocots) and from lycophytes. That clade is evolutionarily distant from the PTEN1 clade and includes two genes in Arabidopsis: the AtPTEN2a and the AtPTEN2b genes. Like human PTEN, both AtPTEN2a and AtPTEN2b proteins are able to catalyse dephosphorylation of a synthetic protein phosphatase substrate (p-NPP) and of PIs (Figure 3 and Supplementary Figure S4), suggesting that other sequences that are present in this clade may have similar activities.
The phylogenetic pattern obtained for PTEN suggests that the duplication giving rise to the PTEN1 and the PTEN2 clades occurred at least before the divergence between lycophytes and spermaphytes (approximately 400 million years ago), and most likely earlier, before the origin of the green algae. In this case, the latter would have lost the PTEN1 gene. However, further investigations in other representatives of algae would be needed to test this hypothesis. Subsequently to this duplication event, the two PTEN paralogues evolved separately. Interestingly, the duplicated genes appear to have experienced different evolutionary pathways according to the plant lineage. Lycophytes and angiosperm eudicots retained the two genes, whereas angiosperm monocots (cereals and sugarcane) seem to have lost the PTEN1 gene. The current sequencing of additional plant and eukaryote genomes will precisely determine the origin and evolutionary dynamics of both genes in plants in the near future. In addition to the deletions observed in some lineages, subsequent duplications were observed in A. thaliana and Populus trichocarpa by the occurrence of closely related PTEN2 sequences.
Most of the biological effects of animal PTEN are attributed to its lipid phosphatase activity that dephosphorylates the PI(3,4,5)P3 and other D3 phosphate PIs . Biochemical characterization of AtPTEN2 proteins shows their preferential use of 3-PIs as substrates (Figure 3B). However, PI(3,4,5)P3 was found to be a very poor substrate for the AtPTEN2s, unlike animal PTEN and AtPTEN1. AtPTEN2a, which is by far the most active Arabidopsis PTEN2, dephosphorylates mostly PI3P, PI(3,4)P2 and PI(3,5)P2, and also has a weak activity towards PI4P, whereas the only AtPTEN2b substrate identified is PI3P. Detailed analysis of the amino acid residues at the active catalytic core of plant PTENs revealed mutations at crucial positions for the PTEN1 and PTEN2 sequences (corresponding to Lys125 in HsPTEN for PTEN1 and Lys128 for PTEN2 respectively), especially in the recent eudicot lineages (Figure 2B). In PTEN2, the absence of reactivity towards PI(3,4,5)P3 and the preferential use of PI3P as substrate can be explained by amino acid changes within the PTEN domain, as shown by studies combining site-directed mutagenesis (Figure 3C) and homology modelling of PI(3,4,5)P3 and PI3P docking on native and mutated AtPTEN2a protein (Figures 4B and 4C). The question arises as to whether the PTEN2 lipid phosphatase activity has evolved to achieve specific functions in vascular plants, especially in eudicots, reflective of the specific diversity of PI in plants.
All known PI isomers, with the exception of PI(3,4,5)P3 and probably also PI(3,4)P2, have been identified to occur in plants . PI3P constitutes 5–15% of the plant PIs and is exclusively synthesized by the type III PI3K, which is the unique type of PI3K in plants, known as Vps34 (vacuolar protein sorting 34). Studies of Arabidopsis type III PI3K-deficient plants severely affected in both vegetative and reproductive development identified a role of PI3P in the response to salinity, as well as in endocytosis [33,34]. PI3P has also been found to be important for the control of stomatal movement . PI(3,5)P2 was found to be increased in response to osmotic stress , but its role is still unknown, similar to PI5P, which is probably a degradation product of PI(3,5)P2. In Arabidopsis, whereas a number of proteins called 5PTases (inositol polyphosphate 5-phosphatases) are able to dephosphorylate the D5 position of the inositol ring of PIs or soluble inositol polyphosphates , only PTEN and myotubularins are potentially able to dephosphorylate D3 PIs. Like their animal counterparts, which are involved through the PI3K/AKT signalling pathway in the regulation of a wide range of processes including cell proliferation and growth, cell metabolism and apoptosis , the PTEN2 proteins found in vascular plants are therefore likely to play a crucial role in the control of 3-phosphorylated PI signalling events and homoeostasis.
PTEN activity can be further modulated in vivo by post-translational modifications , interactions with protein partners  and localization in the nucleus or in the cytoplasm, where it can shuttle between cytosol and plasma membrane . Recruitment of animal PTEN to membrane requires a PI(4,5)P2-binding motif and phosphorylated residues in the C-terminal domain of the protein , which were not identified in plant PTEN2. The phosphatase domain of AtPTEN2 proteins differs mainly from HsPTEN and AtPTEN1 by a single unstructured lipophilic loop of 8 amino acids (Figure 4A), with no known functional significance. However, we identified in plant PTEN2 a C2 domain, also found in animal PTEN, which may help tether PTEN to the membrane . Unexpectedly, we also discovered that AtPTEN2a was able to bind PA and, to a lesser extent, PS (Figures 5A and 5B). Interestingly, C2 domains from other proteins have been shown to bind PA, e.g. tomato PLDα (phospholipase Dα) , and mammalian PKCϵ (protein kinase Cϵ), which has been shown to specifically respond to PA accumulation in vivo . Whereas HsPTEN is known to bind PS , the binding of PA has not been reported for animal PTEN proteins. Both PS and PA are signalling lipids with various cellular and physiological effects in animals . In plants, PS signalling properties have not yet been established. In contrast, PA has been implicated in a variety of processes including abiotic stress responses, pathogen defence and hormone signalling [14,46,47]. Little is known about how PA exerts its effects in plants. An increasing number of PA-interacting proteins have recently been isolated in plants [9,47,48]. Among these are several protein kinases and phosphatases with roles in stress and hormonal signalling pathways such as the Raf-1 kinase homologue CTR1, involved in ethylene signalling , and the ABI1 phosphatase involved in ABA (abscisic acid) signalling . Both act as negative regulators in their respective pathways and are inhibited by PA. Interestingly, the plant 3-PDK1 (PI-dependent protein kinase-1), a component of the PI3K/Akt signalling pathway, which is negatively regulated by PTEN in animals, is activated by PA binding in plants . In mammals, a major downstream target of this pathway, the TOR (target of rapamycin) kinase is also a target of PA, which further regulates the assembly of the mTOR (mammalian TOR) complexes . Whether the PA-binding properties of AtPTEN2a modify its activity in vivo, e.g. through protein stabilization, membrane tethering, activation or inhibition of phosphatase activity, remains to be established.
In summary, our phylogenetical results strongly suggest that the PTEN gene has been duplicated in vascular plants (PTEN1 and PTEN2). When compared with HsPTEN, the eudicot PTEN2 proteins display changes in charged amino acids in the core catalytic site, which likely affect their biochemical properties. As a consequence, PI3P, one of the major PIs found in plants, is the preferential substrate of the Arabidopsis AtPTEN2a protein. In contrast, the preferred substrate of HsPTEN, PI(3,4,5)P3, which has not been detected in vascular plants, is a very poor substrate for AtPTEN2a. The AtPTEN2a protein displays strong binding affinity for PA, a major lipid second messenger with roles in stress and hormonal signalling in plants. Taken together, these results suggest that the eudicot PTEN2 phosphatase has evolved to fulfil specific roles in lipid signalling. The possible regulatory role of PTEN2 in response to developmental and stress signals is consistent with the ubiquitous expression of Arabidopsis AtPTEN2 genes in the plant, especially in young organs, which undergo a high metabolic activity/growth rate. In addition, strong transcriptional regulation of AtPTEN2 genes in response to salt and osmotic stresses, and tight post-transcriptional regulation of AtPTEN2 protein levels, are observed. Elucidating the physiological function and signalling pathways in which PTEN2 is implicated in both eudicots and monocots constitutes a new major challenge in plant science.
Bayesian maximum likelihood
Metropolis-coupled Markov chain Monte Carlo
Murashige and Skoog
phosphatase and tensin homologue deleted on chromosome ten
Tris-buffered saline plus Tween 20
target of rapamycin
yellow fluorescent protein
Christophe Rothan, Anne Pribat and Véronique Germain designed the research, performed sequence analysis, purification and biochemical characterization of the PTEN proteins, and promoter activity and transcript analysis in planta. Rodnay Sormani conducted PTEN protein expression in planta and participated in promoter activity experiments. Mathieu Rousseau-Gueutin performed the phylogeny analysis. Magdalena Julkowska, Christa Testerink and Véronique Germain performed the lipid-binding assays. Jerôme Joubés conducted the expression analysis in stress conditions. Michel Castroviejo participated in purifying the PTEN proteins. Michel Laguerre performed the modelling studies of the phosphatase domain. Christophe Rothan wrote the paper, and Anne Pribat, Rodnay Sormani, Véronique Germain, Christa Testerink and Christian Meyer contributed to the writing.
We thank Dr Teun Munnik for providing the GST–2×FYVE construct. We would like to give our thanks to the Proteomic platform from Bordeaux for the liquid chromatography tandem MS analysis of recombinant AtPTEN proteins. We also thank Dr Sebastien Mongrand for critical reading of the manuscript prior to submission.
This work was supported by the Région Aquitaine (to A.P.) and the Agence Nationale de la Recherche (ANR Blanc) (to R.S.). Work in C.T.'s lab is funded by the Netherlands Organization for Scientific Research (NWO) [grant numbers Vidi 700.56.429, ALW 820.02.017, NGI Horizon 93511011 and STW Perspectief 10987]. Financial support from EU-COST Action FA0605 is also acknowledged.
These authors contributed equally to this work.
Present address: School of Molecular and Biomedical Science, The University of Adelaide, Adelaide, SA 5005, Australia.