Biochemical Journal

Research article

Identification and characterization of AtI-2, an Arabidopsis homologue of an ancient protein phosphatase 1 (PP1) regulatory subunit

George W. Templeton, Mhairi Nimick, Nicholas Morrice, David Campbell, Marilyn Goudreault, Anne-Claude Gingras, Atsushi Takemiya, Ken-ichiro Shimazaki, Greg B. G. Moorhead

Abstract

PP1 (protein phosphatase 1) is among the most conserved enzymes known, with one or more isoforms present in all sequenced eukaryotic genomes. PP1 dephosphorylates specific serine/threonine phosphoproteins as defined by associated regulatory or targeting subunits. In the present study we performed a PP1-binding screen to find putative PP1 interactors in Arabidopsis thaliana and uncovered a homologue of the ancient PP1 interactor, I-2 (inhibitor-2). Bioinformatic analysis revealed remarkable conservation of three regions of plant I-2 that play key roles in binding to PP1 and regulating its function. The sequence-related properties of plant I-2 were compared across eukaryotes, indicating a lack of I-2 in some species and the emergence points from key motifs during the evolution of this ancient regulator. Biochemical characterization of AtI-2 (Arabidopsis I-2) revealed its ability to inhibit all plant PP1 isoforms and inhibitory dependence requiring the primary interaction motif known as RVXF. Arabidopsis I-2 was shown to be a phosphoprotein in vivo that was enriched in the nucleus. TAP (tandem affinity purification)-tag experiments with plant I-2 showed in vivo association with several Arabidopsis PP1 isoforms and identified other potential I-2 binding proteins.

  • inhibitory protein
  • phylogenetics
  • protein phosphatase 1 (PP1)
  • nucleus
  • signal transduction
  • TAP-tag

INTRODUCTION

Protein phosphorylation, the addition of a phosphate group to predominantly serine, threonine and tyrosine residues, is a major form of regulation for many cellular processes in eukaryotes. It has been estimated that 70% of proteins are phosphorylated at some point in their lifetime [1], and as more eukaryote genomes are sequenced, it is apparent that 2–4% of eukaryotic genes encode protein kinases and phosphatases [2]. The Arabidopsis thaliana genome encodes over 1100 protein kinases [3], and approximately 150 protein phosphatases [4]. The apparent disparity in numbers reflects the mode of specificity: essentially, individual kinases have specific targets based on the kinase domain and its other domains or functional motifs, while phosphatases have little specificity on their own, and many form multiprotein complexes in order to generate specificity. Thus individual phosphatase catalytic subunits are able to form multiple complexes, through binding to a regulatory subunit [5].

There are three main groups of protein phosphatases in eukaryotes, the serine/threonine, tyrosine/dual-specificity and Asp-based [4]. The PPP family, of the serine/threonine group, contains the most studied members, and includes PP1 (protein phosphatase 1) and PP2A (protein phosphatase 2A), and is one of the most conserved families of proteins among eukaryotes [2]. In plants, some roles for PP2A have been suggested by the identification of PP2A regulatory proteins in Arabidopsis by forward genetic screens. The first to be identified was RCN1 (roots curl in naphthylphthalamic acid 1), which is involved in auxin transport, abscisic acid signalling, gravitropism and the biosynthesis of and response to ethylene, while the second, FASS/TON2 (tonneau 2), is involved in cytoskeletal organization [5,6]. Recently, PP1 was found to be involved in stomatal opening, but no further roles have yet been defined in plants [7,8], and only a single binding partner, or regulatory subunit, has been identified [9]. In mammals, over 140 PP1 regulatory subunits are known with clear functions in a host of diverse biological processes and a further 450–500 or more binding proteins are thought to exist [10,11]. A majority of these binding proteins interact with PP1 via a sequence termed the RVXF motif [12,13]. Two main consensus sequences for the RVXF motif exist, [RK]-X0,1-[VI]-{P}-[FW] and [HKR]-[ACHKMNQRSTV]-V-[CHKNQRST]-[FW] [14]. These motifs both emphasize the importance of the ‘V’ and ‘F’ positions of the motif, mutation of which leads to the loss of PP1-binding capability in known PP1 interactors [15].

One of the earliest discovered, and most highly conserved, PP1 regulatory subunits is I-2 (inhibitor-2) [16,17]. Although it was identified over 30 years ago as a heat-stable protein from rabbit skeletal muscle capable of inhibiting phosphatase activity, to date its role in the cell remains poorly understood. In mammals, I-2 is seen to be involved in the cell cycle, with protein expression peaking during mitosis and becoming phosphorylated during mitosis by CDK2 (cyclin-dependent kinase 2) at a conserved PXTP motif [1820]. Other binding partners in mammals include the centrosomal kinase Nek2 (NIMA-related kinase 2), Aurora A kinase, and the prolyl isomerase Pin1, all which implicate a role in mitosis [2123]. Characterization studies of the I-2 protein have yielded several functional motifs which are often conserved. Aside from the PXTP motif, three other functional regions were identified through biochemical studies and sequence comparisons. These regions, the [SG]ILK motif, the RVXF motif and an α-helical region approx. 40 amino acids long with a key HYNE motif at its centre, were confirmed when the structure of human PP1–I-2 was published [24]. A further site (44-SQ-45 in human I-2) was identified within the RVXF motif as the site of phosphorylation by the ATM (ataxia telangiectasia mutated kinase) in response to DNA damage, which disrupts its interaction with PP1 at the G2/M-phase checkpoint of the cell cycle [25].

Study of PP1 is complicated, as in most eukaryotic systems, due to the number of isoforms. Mammals have three PP1 isoforms (α, β and γ), plus a splice variant (γ1 and γ2), while Arabidopsis has nine genes, TOPP (type one protein phosphatase) 1–9. In mammals, the isoforms are known to have distinct cellular localizations [26], while in Arabidopsis the nine isoforms have relatively similar cellular localization when overexpressed [9]; their mRNA is found in most tissues (with little specificity in most cases) and at least four of the nine can be found expressed in the same tissue [27]. The work of Takemiya et al. [9] demonstrated that, like mammals, plant PP1 is enriched in the nucleus.

Through our efforts to identify PP1 regulatory subunits in Arabidopsis we found AtI-2 (A. thaliana I-2), a homologue of the I-2 family of PP1-regulatory proteins, and determined its activity, binding partners and cellular localization. We also examined the conservation of the I-2 family across eukaryotes and evolution of the I-2 motifs noted above.

EXPERIMENTAL

Chemicals

All chemicals were obtained from Sigma–Aldrich unless otherwise noted.

Plant material

The heterotropic cell suspension culture of A. thaliana was a gift from Professor Carol MacKintosh (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, U.K.), and was cultured in the presence of sucrose, auxin and cytokinin as described previously [27]. Cells were harvested by rapid suction filtering and stored at −80 °C.

Preparation of affinity matrices

Recombinant human PP1γ was cloned, expressed and purified as reported previously [28]. PP1γ and BSA, 50 μg of each, were separately coupled to 100 μl of activated CH-Sepharose 4B™ (GE Healthcare) according to the manufacturer's protocol, except that final washes (after blocking unbound sites) were performed using 5×1 ml of PBS buffer.

Identification of PP1-interacting proteins

All procedures were performed at 4 °C unless otherwise stated. 60 g of cultured A. thaliana was taken from storage at −80 °C, broken into small pieces and mixed with 0.5 vol. of Arabidopsis-extraction buffer [50 mM Tris/HCl (pH 7.5), 1 mM EDTA, 1 mM EGTA, 1% (w/v) PVP (polyvinylpyrolidone), 0.1% BMe (β-mercaptoethanol), 1 mM PMSF and 50 mM NaCl] in a blender until homogenous. Once mixed and thawed, cells were lysed by a single pass through a French Pressure Cell (Sim–Aminco) at 16000 psi (110400 kPa) exit pressure and centrifuged for 40 min at 100000 g. The supernatant was applied to 10 ml of Q-Sepharose™ Fast Flow equilibrated in QA buffer [50 mM Tris/HCl (pH 7.5), 0.1 mM EGTA, 0.1% BMe, 1 mM PMSF and 50 mM NaCl] and gently mixed at 4 °C for 30 min. The matrix was collected and washed with 250 ml of QA buffer, and protein was eluted in QA buffer and 700 mM NaCl. The eluate was diluted to the volume of the original crude extract with QA buffer and 50 mM NaCl, divided in two, and then incubated with the affinity matrices (human PP1γ or BSA bound to CH-Sepharose™) for 2 h at 4 °C. The matrices were then collected and washed with 1 ml of PBS, 1 ml of PBS and 1.5 M NaCl, and again with 1 ml PBS before boiling the matrices in SDS/PAGE loading buffer. SDS/PAGE (10%) gels were run, the proteins were visualized with colloidal Coomassie Blue staining, then bands were excised and processed as reported previously [15].

Bioinformatic analysis of the I-2 family

Using the sequence of PP1R2 (I-2; GenBank® accession code NP_006232) and AtI-2 (At5g52200; GenBank® accession code NP_568768), genomes representative of a variety of taxa were searched for homologues [29]. Databases used include the NCBI (National Center for Biotechnology Information, http://www.ncbi.nlm.nih.gov; for Mus musculus, Xenopus laevis, Danio rerio, Ciona intestinalis, Drosophila melanogaster, Hydra magnipapillata, Trichoplax adhaerens, Monosiga brevicollis, Saccharomyces cerevisiae, Dictyostelium discoideum, Ostreococcus lucimarinus, Oryza sativa, Physcomitrella patens, Guillardia theta, Hemiselmis andersenii, Theileria parva, Plasmodium falciparum, Phytophthora infestans, Thalassiosira pseudonana, Tetrahymena thermophila, Leishmania major, Euglena gracilis, Trichomonas vaginalis and Giardia lamblia), the DOE JGI (Department of Energy Joint Genome Institute, http://www.jgi.doe.gov; for Branchiostoma floridae, Helobdella robusta, Nematostella vectensis, Chlamydomonas reinhardtii and Phycomyces blakesleeanus) and the Broad Institute (http://www.broadinstitute.org; for Coprinus cinereus, Rhizopus oryzae and Trypanosoma brucei). Sequences were aligned with Clustal X2 [30] and hand optimized in GeneDoc [31].

Expression and purification of AtI-2

AtI-2 amplified from the pda02740 clone (RIKEN BioResource Center accession number AY050347; [32]) was subcloned into the pET101 vector (Invitrogen) using the manufacturer-provided protocols to retain the 6×His tag. For expression, BL21 STAR™ cells (Invitrogen) were transformed with the AtI-2–pET101 construct and grown overnight in LB (Luria–Bertani) medium [1% (w/v) tryptone, 0.5% yeast extract and 1% (w/v) NaCl (all from Bioshop Canada)] plus 100 μg/ml ampicillin. Overnight cultures were diluted 100-fold on the following morning in 4 litres of LB medium and ampicillin, and grown to a D600 of 0.5, when expression was induced by the addition of 0.1 mM IPTG (isopropyl β-thiogalactopyranoside; Gold Biotechnology). Three hours after induction, the cells were harvested by centrifugation (4000 g for 10 min at 4 °C), resuspended in lysis buffer [25 mM Tris/HCl (pH 7.5), 150 mM NaCl, 10 mM imidazole, 0.5 mM PMSF, 0.5 mM benzamidine and 5 μg/ml leupeptin) and stored at −80 °C. The cells were then subsequently thawed and lysed by two passes through a French Pressure Cell (Sim–Aminco) at 16000 psi (110400 kPa) exit pressure, and centrifuged at 4 °C for 40 min at 100000 g. The supernatant was placed into a beaker in a water bath at 80 °C and, while stirring, the bath was heated to 90 °C over 10 min, after which it was held at 90 °C for 5 min. The sample was then cooled to 4 °C and again centrifuged for 40 min at 100000 g. The supernatant was incubated with 1 ml of Ni-NTA (Ni2+-nitrilotriacetate) agarose (Qiagen) for 30 min with gentle mixing at 4 °C. The matrix was washed with 200 ml of wash buffer [25 mM Tris/HCl (pH 7.5), 1 M NaCl and 10 mM imidazole) before being eluted with 10 ml of elution buffer [25 mM Tris/HCl (pH 7.5), 150 mM NaCl and 250 mM imidazole]. The eluate was dialysed into PBS then concentrated, dialysed into PBS and 50% (v/v) glycerol, and stored at −20 °C. AtI-2 V9A/W11A mutant was generated by site-directed mutagenesis of the AtI–2–pET101 construct following the Quikchange™ (Stratagene) protocol, and expression and purification conditions as above. Protein concentration was determined by the Bradford assay [33]. Purified proteins were examined for purity by SDS/PAGE (10% gel) and confirmed as AtI-2 by MS. The purified recombinant protein was blotted on to PVM (parasitophorous vacuolar membrane) and Edman sequencing performed as described in [34].

Generation of the AtI-2 polyclonal antibody

Pure recombinant AtI-2–6×His was used for antibody production in a New Zealand White rabbit, as described previously [34]. The AtI-2 antibody was affinity purified by repeated binding/elution from a nitrocellulose membrane with immobilized full-length (32 kDa) AtI-2 [15]. Pre-immune serum IgG was purified using Protein A Sepharose™ (GE Healthcare).

Purification of the microcystin-sensitive phosphatase complexes

The soil portions of mature A. thaliana plants described above (containing rosettes, bolts, flowers and siliques in various stages of development) were harvested and flash frozen in liquid nitrogen. The harvested tissue (50 g) was ground in a mortar and pestle with acid-washed sand in liquid nitrogen. Arabidopsis extraction buffer (from the Identification of PP1-interacting proteins section; 50 ml) was added to the powdered tissue and mixed. The mixture was filtered through two layers of miracloth (Calbiochem) and centrifuged at 50000 g for 30 min at 4 °C. The supernatant was incubated with 1 ml of MC (microcystin–Sepharose) [35] for 1 h at 4 °C. After extensive washing in buffer [50 mM Tris/HCl (pH 7.5), 0.1 mm EGTA, 5% (v/v) glycerol, 0.1% BMe and 350 mM NaCl] with 0.05% Brij-35 (BDH), then without detergent, the protein was eluted by incubation for 20 min with the wash buffer plus 3 M NaSCN. The eluate was dialysed into PBS and concentrated for analysis.

AtI-2 and TOPP tissue specificity

The samples of roots, rosettes, shoots, flowers and siliques were dissected from A. thaliana plants and frozen with liquid nitrogen. The protein was extracted as described above, with the appropriate volumes of buffer. Extracts were clarified by centrifugation at 14000 g for 10 min at 4 °C, and quantified with the Bradford assay [33].

Expression and purification of the Arabidopsis TOPP isoforms

The nine TOPP isoforms [9] were subcloned into the pET101 vector (Invitrogen) using the manufacturer-provided protocols to create an untagged protein. For expression, BL21 STAR™ cells (Invitrogen) were transformed with the TOPP–pET101 constructs and grown overnight in LB medium plus 100 μg/ml ampicillin. The overnight cultures were diluted 100-fold on the following morning in 4 litres of LB medium, ampicillin and 1 mM MnCl2, and grown to a D600 of 0.5. At this point, expression was induced by the addition of 0.1 mM IPTG overnight at 37 °C. The following morning, the cells were harvested by centrifugation (4000 rev./min for 10 min at 4 °C), resuspended in lysis buffer [25 mM Tris/HCl (pH 7.5), 100 mM KCl, 1 mM EDTA, 2 mM MnCl2, 5% (v/v) glycerol, 0.1% BMe and 1 mM PMSF] and stored at −80 °C. The cells were subsequently thawed, lysed by two passes through a French Pressure Cell (Sim–Aminco) at 16000 psi (110400 kPa) exit pressure and centrifuged at 4 °C for 40 min at 100000 g. The supernatant was incubated with 0.2 ml of microcystin affinity matrix [35] for 30 min at 4 °C, then the matrix was washed with 300 ml of wash buffer [25 mM Tris/HCl pH (7.5), 1 M NaCl, 0.1 mM EDTA, 1 mM MnCl2, 5% (v/v) glycerol, 0.1% BMe and 0.05% Brij-35]. The matrix was then eluted by incubating in elution buffer [25 mM Tris/HCl (pH 7.5), 3 M NaSCN, 0.1 mM EDTA, 1 mM MnCl2, 5% (v/v) glycerol and 0.1% BMe] for 20 min before being collected and dialysed [25 mM Tris/HCl (pH 7.5), 250 mM NaCl, 250 mM KCl, 1 mM MnCl2, 50% (v/v) glycerol and 0.1% BMe], and stored at −20 °C. Each TOPP, with the exception of TOPP6 (see the Results section) was 80–95% pure after this single chromatography step.

pNPP (p-nitrophenyl phosphate) phosphatase activity assays

Assays were adapted from MacKintosh and Moorhead [36]. Phosphatases and inhibitors [microcystin-LR (Enzo Life Sciences) and AtI-2] were diluted in assay buffer [50 mM Tris/HCl (pH 7.5), 1 mM EDTA, 2 mM MnCl2 0.5 mg/ml BSA and 0.1% BMe]. Inhibitors (10 μl) and phosphatases (10 μl; 150 ng) were combined and incubated at 30 °C for 10 min. 30 μl of substrate [50 mM Tris/HCl (pH 7.5), 0.2 mM MnCl2, 20 mM MgCl2 and 20 mM pNPP (Fluka)] was then added. After incubation for 10–20 min (the exact time was dependent on the activity of the specific isoform), the reaction was quenched with 150 μl of 0.5 M EDTA, and the A405 was measured. All the assays were done in triplicate and results are means±S.D. for two experiments. The recombinant expressed human I-2 was provided by Professor Charles Holmes (Department of Biochemistry, University of Alberta, Canada). IC50 values were determined by the concentration of AtI-2 required to inhibit 50% of the pNPP phosphatase activity.

The localization of AtI-2

GFP (green fluorescent protein) and RFP (red fluorescent protein) fusion proteins were expressed in fava bean (Vicia faba) epidermal cell layers using the particle bombardment technique (PDS-1000, BioRad) [37]. The TOPP2–RFP construct was as described in [9], the PRH-75–GFP construct was as described in [38] and the AtI-2–GFP construct was generated by Gateway® cloning (Invitrogen) as per the manufacturer's protocol into pDONR201, then the pK7FWG2 vector [39]. Cells were visualized with a DM IRE2 spectral confocal and multiphoton microscope (Leica Microsystems) with a TCS SP2 acoustic optical beam splitter (Leica Microsystems). Images were captured using the Leica Confocal software included with the device.

For immunolocalization, all solutions were made with PBS. A dark-grown Arabidopsis cell suspension culture was fixed using 4% (w/v) paraformaldehyde, followed by cell-wall digestion [1% (w/v) cellulase Onozuka R-10 (Phytotechnology Laboratories) and 0.1% macerozyme R-10 (Phytotech-nology Laboratories)] and then cell-membrane dissolution [1% (v/v) Triton X-100]. The cells were then blocked with 2% (w/v) BSA, and probed with 2 μg/ml affinity-purified anti-AtI-2 antibody or rabbit IgG, followed by anti-rabbit Alexa Fluor® 488 (Invitrogen). For the blocking experiments, recombinant AtI-2 was added to the anti-AtI-2 antibody solution at 20 μg/ml and then pre-incubated for 10 min at room temperature (22 °C) before probing. The nuclei were stained with 1 μg/ml DAPI (4′,6-diamidino-2-phenylindole) and the cells visualized using a fluorescence microscope (DMR, Leica). Images were captured using a cooled CCD (charge-coupled device) camera (Retiga 1350 EX, QImaging). Image enhancement and deconvolution confocal algorithm manipulations were performed using the Volocity software (Improvision, version 5.0.2).

Generation of alternate TAPa (tandem affinity purification) cell lines

Using Gateway® Cloning (Invitrogen), as per the manufacturer's protocol, AtI-2 and GFP were inserted into pDONR201 and then the pN-TAPA and pYL436 vectors [40] to generate both N- and C-terminally tagged versions of the construct. The vectors were transformed by electroporation in Agrobacterium strain GV3101. The transformation of Arabidopsis cell culture was performed as previously reported [41], with the exception of MS (Murashige and Skoog) medium used in place of B5 medium, timentin (GlaxoSmithKline) used in place of meropenem and 100 μg/ml gentamycin as a selection marker. Western blotting of the cultured selected cells showed immunoreactive bands at the predicted sizes of both endogenous AtI-2 and the TAPa constructs. The cells were cultured and harvested as in the Plant material section above.

TAP purifications

The protocol was adapted from Rubio et al. [40]. All procedures were performed at 4 °C unless stated otherwise. Briefly, 50 g of harvested transformed cell culture was mechanically separated and thawed in TAP buffer [50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 10% (v/v) glycerol and 0.1% Nonidet P-40) plus inhibitors [1 mM PMSF, 1% (w/v) PVP and 1 μg/ml ProteCEASE-100 EDTA free (G-Biosciences)]. The cells were lysed by a single pass through a French Pressure Cell (Sim–Aminco) at 16000 psi (110400 kPa) exit pressure, and centrifuged for 40 min at 100000 g. The supernatants were incubated with 100 μl of IgG Sepharose (GE Healthcare) for 2 h, followed by washing of the matrix with 20 ml of TAP buffer plus 1 mM dithiothreitol. Complexes were eluted by digestion with 10 units of Prescission Protease (GE Healthcare) for 2 h, eluate was collected from the column flowthrough combined with a 5 ml wash with TAP buffer plus 10 mM of imidazole. This digested eluate was incubated with 100 μl of Ni-NTA agarose (Qiagen) and incubated for 30 min. The matrix was then washed with 30 ml of TAP buffer plus 10 mM imidazole and then split into samples for MS analysis and samples for peptide elution.

For the MS analysis, the cells were then washed with 8 ml of triethylammonium bicarbonate (pH 8.5). The matrix was then collected and incubated overnight with 750 ng of trypsin (Promega) at 30 °C, then for an additional 1 h with a further 250 ng of trypsin before elution with 50% (v/v) acetonitrile and 2.5% (v/v) FA (formic acid). The pooled eluates were then dried. AtI-2 results were from three independently transformed cell lines (two with a C-terminal tag and one with an N-terminal tag) and the controls were replicate experiments from GFP–TAPa cell lines, along with unrelated nuclear-, cytosolic- and chloroplast-localized TAPa-tagged protein constructs.

For the peptide-elution experiments, the column was sequentially eluted with 2 mM peptide containing a mutated RVXF motif (‘RARA’; peptide sequence GKKRARAADLE), 2 mM peptide containing an RVXF motif (‘RVXF’; peptide sequence GKKRVRWADLE) and finally TAP buffer plus 250 mM imidazole. This peptide-elution protocol was adapted from Moorhead et al. [42].

Whole-sample MS analysis

Samples were resuspended in 5% (v/v) FA. After centrifugation at 4 °C (10 min, 12000 g), the supernatant was directly loaded on to capillary columns packed inhouse with Magic 5 μm, 100 Å (1 Å=0.1 nm), C18AQ. The data acquisition protocol was modified from Breitkreutz et al. [43]. MS/MS results were acquired in data-dependent mode (over a 2 h period in an acetonitrile 2–40% gradient) on a ThermoFinnigan LTQ equipped with a Proxeon NanoSource and an Agilent 1100 capillary pump. Acquired RAW files were converted to mgf (Mascot generic format) format, and searched with the Mascot search engine (Matrix Sciences) against the A. thaliana complement of the RefSeq database (http://www.ncbi.nlm.nih.gov/RefSeq/; release 38) with a precursor-ion-mass tolerance window of 3.0 Da and a fragment-ion-mass tolerance of 0.6 Da. Methionine oxidation was allowed as a variable modification, and trypsin specificity (with one missed cleavage allowed) was selected. The search results were parsed into a relational database developed inhouse (ProHits; [44]). The initial filtering of the results was performed by removing the hits with a Mascot score <60 and only one unique peptide. In the case of TOPP isoforms, the unique peptide requirement was relaxed due to the highly similar sequences of the TOPP isoforms. Proteins identified in the AtI-2 TAP purifications were only considered to be true binding partners if they were not found in all or the negative controls (3× GFP–TAP, 6× nuclear-localized control protein–TAP, 2× cytosolic control protein–TAP, 2× chloroplast-localized control protein–TAP).

RESULTS

Identification of AtI-2 and bioinformatic analysis of the I-2 family

From an A. thaliana cell culture extract, PP1-binding proteins were enriched by affinity chromatography over a human PP1γ-affinity matrix. We chose PP1γ because of its superior recombinant expression and 65–70% identity with all Arabidopsis isoforms (Supplementary Figure S1 at http://www.BiochemJ.org/bj/435/bj4350073add.htm). Numerous proteins were enriched on the PP1 matrix in comparison with the control matrix (Supplementary Figure S2 at http://www.BiochemJ.org/bj/435/bj4350073add.htm). MS analysis of the excised bands resulted in the identification of an Arabidopsis homologue (At5g52200) to the human PPP1R2, also known as I-2, as well as homologues to other known human PP1-interacting proteins, such as NIPP1 (At5g47790) and SDS22 (At5g19680). The identified proteins were also examined for potential RVXF motifs, the presence of which would support the possibility of true direct interacting proteins.

The presence of an Arabidopsis homologue of I-2 is not surprising, given that homologues have been identified in species as divergent as S. cerevisiae [45] and Caenorhabditis elegans [16]. Some features of the human protein, however, are notably absent, such as the recently identified PP1-interaction motif [SG]ILK motif and the ATM-phosphorylation site (SQ) within the PP1-docking RVXF motif. To determine if this pattern of conservation was consistent across Plantae, we aligned I-2 homologues from a variety of plants, algae and mosses (representatives can be seen in Figure 1A). While the other conserved motifs (RVXF, PXTP and HYNE) stay largely intact, there is a definite lack of the [SG]ILK motif and SQ in all plant species examined (Supplementary Figure S3 at http://www.BiochemJ.org/bj/435/bj4350073add.htm).

Figure 1 Conservation of the PP1 I-2 among plants and greater eukaryota

(A) An amino acid sequence alignment of human I-2, along with homologues from A. thaliana (At5g2200), O. sativa (rice; NP_001055135), P. patens (moss; XP_001766362) and O. lucimarinus (green algae; XP_001420243). Motifs identified as important for the PP1–I-2 interaction are noted. The [SG]ILK, RVXF and HYDE motifs have been previously identified as being the three major regions of interaction between human I-2 and human PP1γ isoform [24], the PXTP site is known to be phosphorylated by CDK2 (cyclin-dependent kinase 2)/cyclin B1 in human cells [19] and the SQ site is known to be phosphorylated by ATM in a DNA-damage-dependent manner in human cells [25]. (B) A phylogenetic overview demonstrating the conservation of the I-2 protein and the known functional motifs across six main groupings of Eukaryota [adapted from 54]. The JEH group includes the jakobids, euglenozoans and heteroloboseans, while the POD group consists of the parabasalids, oxymonads and diplomonads. The uncertainty of the order of early branching events is designated by the dotted polytomy [54]. The evolutionary distances shown are for demonstrative purposes only. The presence or absence of a homologue and, when a homologue is present, the conservation of known functional motifs (listed above) is shown by a ‘+’ or ‘−’ respectively, ‘+/−’ indicates differences in conservation within a group and ‘?’ denotes a motif that is only partially intact. Sequences examined for the functional motifs are as follows, the sources of sequences are listed in the Experimental section; H. sapiens: NP_006232; M. musculus: NP_080076; X. laevis: NP_001091136; D. rerio: NP_991231; B. floridae: jgi|Brafl1|123964| C intestinalis: XP_002131008; H. robusta: jgi|Helro1|157299| D. melanogaster: NP_524013; H. magnipapillata: XP_002155049; N. vectensis: jgi|Nemve1|211544| T. adhaerens: XP_002107804; M. brevicollis: XP_001744204; C. cinereus: CC1G_01429.2; P. blakesleeanus: jgi|Phybl2|150708| R. oryzae: RO3G_16062.3; S. cerevisiae: NP_014042; D. discoideum: XP_643942; O. lucimarinus: XP_001420243; A. thaliana: At5g52200; O. sativa: NP_001055135; P. patens: XP_001766362; T. parva: XP_764857; P. falciparum: XP_001351270; P. infestans: EEY56100; T. pseudonana: XP_002293304; T. thermophila: XP_001024018; T. brucei: CBH10274 and L. major: XP_847676. Complete alignment showing motifs is shown in Supplementary Figure S3 (at http://www.BiochemJ.org/bj/435/bj4350073add.htm). n/a, not applicable.

To better understand the conservation of I-2 and its functional motifs, we examined the genomes of eukaryotes from representative groups for homologues. In contrast with the common belief that I-2 is conserved among all eukaryotes [46], we found several groups of eukaryotes lacking an I-2 homologue candidate (Figure 1B), including red algae and a majority of the excavates examined (with the notable exception of the kinetoplastids, where all members examined do have an I-2 homologue candidate). Of those homologues we did find, a distinct pattern emerges with regard to the conserved motifs (Figure 1B and Supplementary Figure S3). The RVXF and PXTP motifs are conserved between all homologues, while the HYNE motif is conserved in all but green algae and the kinetoplastids; the SQ motif is found only in vertebrates. The [SG]ILK motif is only found within amoebozoans and opisthokonts (with possible exceptions among the diatoms and ciliates, although, based on a recent study [47], these putative [SG]ILK motifs are probably non-functional). This suggests fundamental differences in either PP1 structure or I-2 function within non-vertebrate species. To examine the first possibility, we aligned the nine isoforms of PP1 in Arabidopsis with human PP1γ, and marked all residues known to be associated with [SG]ILK motif binding, as well as RVXF and HYNE binding, based on the human I-2–PP1 structure [24] (Supplementary Figure S1). All of the residues critical for interaction are present, either as identical residues or as conserved substitutions, suggesting that the [SG]ILK motif would bind the Arabidopsis PP1 isoforms. Unfortunately, aside from I-2, Arabidopsis contains no homologues to other proteins known or predicted to have functional [SG]ILK motifs [10,47], so the question of the methods of plant PP1 regulation including this motif remairs unanswered.

AtI-2 is a phosphoprotein in vivo

To establish the quality of our affinity-purified AtI-2 antibodies, we performed Western blotting and found that this antibody could easily detect as little as 10 ng of recombinant AtI-2 (Figure 2B) and noted a tight group of immunoreactive bands that are the correct size to be AtI-2, when probing a crude Arabidopsis suspension cell extract (Figure 2C). When crude extracts were then treated with phosphatase, we could collapse these five or six AtI-2 bands to a single lower-mass band, suggesting that these additional bands are multiple phosphorylated versions of the protein. This is supported by recent phosphoproteomic studies that identified three in vivo sites on rice I-2 that are conserved with the Arabidopsis protein [48]. This is in contrast with the recombinant bacterially expressed protein, which has multiple bands due to partial degradation in vivo (Figure 2B).

Figure 2 Recombinant AtI-2, the generation of the AtI-2 antibody and tissue expression of AtI-2

(A) Recombinant purified AtI-2 and V9A/W11A mutants were run on a SDS/PAGE gel and stained with colloidal Coomassie Blue. Asterisks indicate bands excised and sent for MS analysis and identified as AtI-2, and the arrow denotes the primary PP1-interacting band (see Figure 3C for PP1-overlay assay). (B) Sensitivity of the AtI-2 antibody. Purified recombinant AtI-2 was run on a SDS/PAGE gel, Western blotted (WB) and probed with affinity-purified antibody, easily detecting all three major bands previously identified as AtI-2 down to 10 ng. The two lower-mass proteins (25 and 20 kDa) are also weakly detected with antibody affinity purified against the full-length 32 kDa version of AtI-2 suggesting that they are fragments of AtI-2 as well. (C) Specificity of the AtI-2 antibody. Arabidopsis suspension cell crude extract with and without λ phosphatase treatment was run on a SDS/PAGE gel, blotted on to a membrane and probed with the affinity-purified AtI-2 antibody. (D) Western blot for TOPPs and AtI-2 in a variety of Arabidopsis tissues (30 μg of protein in each lane). TOPPs are found in all tissues examined, while AtI-2 is mostly excluded from flowers and siliques. (E) Arabidopsis crude extracts were enriched for PPP family phosphatases on MC. The eluate (elu) of the MC column is enriched not only for TOPP(s), but also for AtI-2. Blank is a control matrix run in parallel to the MC matrix. Left-hand panels are from a significantly longer exposure than the right-hand panels.

Examination of the crude tissue samples from Arabidopsis plants (roots, rosettes, shoots, flowers and siliques) with antibodies against both AtI-2 and a pan-antibody against all TOPP isoforms reveals a partially overlapping distribution (Figure 2D and Supplementary Figure S4 at http://www.BiochemJ.org/bj/435/bj4350073add.htm).

After enrichment for PPP-family phosphatases (PP1, PP2A, PP4, PP5 and PP6) through MC-affinity chromatography of Arabidopsis above-ground tissue extracts [35], we noted an enrichment for AtI-2, indicating the presence of an in planta complex of AtI-2 and one or more TOPPs (Figure 2E).

Inhibition of TOPPs by AtI-2

In order to demonstrate the functional equivalence of AtI-2, we recombinantly expressed and purified eight of the nine TOPP isoforms using MC for use in enzymatic assays. TOPP6, the isoform we were unable to purify, does not bind MC when recombinantly expressed. After partial purification by ion-exchange chromatography, TOPP6 exhibited AtI-2-sensitive activity and, intriguingly, was not inhibited by up to 100 nM microcystin (results not shown). The remaining TOPPs were purified and assayed using pNPP in order to their determine specific activity, as well as inhibition, by AtI-2 and the mutated version of the RVXF motif V9A/W11A (Figures 3A and 3B). Preliminary inhibition curves showed that several of the TOPPs had IC50 values of approx. 1 μM for AtI-2 (ranging from approx. 250 nM to slightly over 1 μM), thus this concentration was used to compare all isoforms.

Figure 3 Kinetic analysis of TOPP isoforms of A. thaliana and effect of AtI-2 on the activity of the TOPPs

(A) Specific activity of TOPPs and human PP1γ using pNPP as a substrate. (B) Inhibition of TOPPs by AtI-2 (dark bars) and the RVXF motif mutant V9A/W11A (light bars). Each of the phosphatases were assayed using 1 μM inhibitor protein. Error bars represent the S.D. (C) Recombinant AtI-2 and the RVXF motif mutant V9A/W11A were run on a SDS/PAGE gel, blotted on to a membrane and probed with labelled PP1γ (FWB) [55] or affinity-purified AtI-2 antibody (WB) as indicated.

Inhibition was sensitive to the presence of an intact RVXF motif, and when the key hydrophobic residues were mutated the majority of PP1 activity returned (V9A/W11A mutant). Nevertheless, as shown in Supplementary Figure S5 (at http://www.BiochemJ.org/bj/435/bj4350073add.htm), at higher concentrations the RVXF-mutated AtI-2 can still inhibit the plant PP1 enzymes. Further support that the N-terminal RVXF motif is the primary PP1-docking site is supported by overlay analysis (Figure 3C). PP1γ readily binds purified recombinant AtI-2, but will not associate with purified recombinant V9A/W11A AtI-2. Interestingly, a portion of the wild-type AtI-2 is truncated when expressed in bacteria and purifies because of its C-terminal tag (Figure 2A). Edman sequencing shows the lower mass 30 kDa band begins with the amino acid sequence EIESNKPV, missing the more N-terminal RVXF motif, and fails to bind PP1 in the overlay assay (Figure 3C).

Localization

In order to determine the cellular localization of AtI-2, we utilized particle bombardment of fava bean using AtI-2 and one of the isoforms of PP1 (TOPP2; Figure 4). Co-expression of TOPP2–RFP and AtI-2–GFP shows strong co-localization, with an enrichment in the nucleus (Figure 4A, bottom panels). TOPP2–RFP without AtI-2–GFP co-expression has a similar localization; however, it also gives a strong signal in the nucleolus, as seen by co-localization with the nucleolar marker PRH75 (Figure 4A top panels and Figure 4B). AtI-2–GFP localization does not appear to change with the presence or absence of TOPP2 (Figure 4).

Figure 4 Localization of AtI-2 and TOPP2 in V. faba leaf epidermal cells expressed through particle bombardment

(A) Leaves were bombarded with TOPP2–RFP and the nucleolar marker PRH75–GFP (top panels), AtI-2–GFP alone (middle panels) or with both TOPP2–RFP and AtI-2–GFP (bottom panels). When expressed without AtI-2, TOPP2 is found to be enriched in the nucleolus, in addition to nuclear and cytosolic localizations. When AtI-2 is co-expressed, TOPP2 is no longer found within the nucleolus. The localization of TOPP2 construct with another Arabidopsis PP1 interactor can be found in [9]. Scale bars represent 10 μm; arrows point to voids in nuclear staining indicating nucleoli. Chloroplasts were detected by chlorophyll autofluorescence. (B) Leaves were bombarded with TOPP2–RFP and the nucleolar marker PRH75–GFP (top and middle panels), and AtI-2–GFP alone (bottom, panel). When expressed without AtI-2, TOPP2 is found to be enriched in the nucleolus, in addition to nuclear and cytosolic localizations. When AtI-2 is expressed, it has similar localization to TOPP2, without the nucleolar enrichment. Scale bars represent 20 μm in the top panels, 5 μm in the middle panels and 10 μm in the bottom panel.

In addition to particle bombardment, we utilized immuno-fluorescence microscopy using our polyclonal AtI-2 antibody (Supplementary Figure S6 at http://www.BiochemJ.org/bj/435/bj4350073add.htm). We observed a distinct nuclear enrichment in fixed unsynchronized cells. In some cells, an enrichment was seen within the nucleolus. This signal was not observed with pre-immune serum IgG and could be blocked by pre-incubation of the antibody with AtI-2.

TAP purifications

TAP-tagged AtI-2 constructs were transformed into Arabidopsis suspension cells, in order to determine if AtI-2 binds specific TOPPs and other proteins in vivo. The TAP purifications identified specific peptides to TOPPs 1–5 and TOPP7 (Table 1). In each of the three TAP purifications examined, TOPP2 was the primary phosphatase identified, with between four and seven unique peptides per experiment (Supplementary Table S1 at http://www.BiochemJ.org/bj/435/bj4350073add.htm), suggesting that TOPP2 is the preferred binding partner of AtI-2 in vivo, or at least the most abundant PP1 isoform.

View this table:
Table 1 Proteins identified to co-purify with TAP–AtI-2

Arabidopsis suspension cell lines stably transformed to express AtI-2 with a TAP tag were purified and whole final lysates were examined using MS, as described in the Experimental section. All TOPP isoforms found are listed (a), and other protein interactors (b) are listed if found in at least two of three trials.* denotes TOPPs that were initially removed by the ‘unique peptides > 1’ filter.The proteins listed are only included if they were not identified in TAP purifications of four other proteins over 13 total negative control trials.

Examination of the non-phosphatases in the TAP results is less telling (Table 1). The best hit, obtained in all three purifications and none of the negative controls, was AT2G35040, an AICARFT (5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase)/IMPCHase (inosine monophosphate cyclohydrolase) bifunctional enzyme, which catalyses the final two steps in de novo purine biosynthesis.

In order to confirm the function of the AtI-2 RVXF motif, we performed peptide elutions from the TAP-purified complexes (Supplementary Figure S7 at http://www.BiochemJ.org/bj/435/bj4350073add.htm). Peptide containing an RVXF motif, but not peptide with a mutated RVXF motif (‘RARA’), was able to disrupt the interaction between AtI-2–TAP and TOPP(s) retained on the matrix, supporting the idea that these proteins form a complex in vivo.

DISCUSSION

In mammals, PP1 has been shown to be a primary regulator of phosphorylation in the cell, with over 140 known regulatory subunits to modify the specificity and localization of the phosphatase activity. In a screen to identify potential PP1-regulatory subunits in Arabidopsis, we identified numerous proteins that bind specifically to PP1 over BSA from an Arabidopsis-suspension cell extract. While not all of the identified proteins are expected to be true PP1 regulators, the presence of Arabidopsis homologues of known PP1-regulatory subunits, such as NIPP1, SDS22 and I-2, indicates that the enrichment for PP1-interacting proteins was successful.

I-2 was discovered over 30 years ago and is among the most studied of the PP1-interacting proteins, yet its precise function remains enigmatic. We identified the plant I-2 in a screen for PP1-binding partners using affinity chromatography. This was supported by TAP-tag experiments that found a selection of Arabidopsis PP1 isoforms as the major AtI-2 binding partners, as well as the enrichment of AtI-2 on microcystin–Sepharose, a matrix that specifically enriches PP1, PP2A, PP4, PP5 and PP6 along with any interacting proteins (Figure 2E). These results, partially supported by previous work [27], demonstrate the expression of at least six isoforms of PP1 (TOPPs) in the same cell type. This points to either extreme redundancy or the possibility that each isoform carries out specific tasks, as is the case in mammals [13]. Human I-2 is known to form a trimeric complex with PP1, either of the protein kinases Nek2 or Aurora A and the prolyl isomerase Pin1, and given the stringency we employed to identify AtI-2-interacting proteins it is probable that the non-PP1 proteins identified in this series of experiments (Table 1) are true partners. The possibility of a true interaction is strengthened by several high-throughput analyses in S. cerevisiae, indicating that the homologues of AT2G35040, ADE16 and ADE17, are both found in PP1-containing complexes [49]. Unfortunately, the localizations of AtI-2 and AT2G35040 are not predicted to overlap (mostly nuclear compared with predicted plastid) and this awaits further investigation. Interestingly, these enzymes control the synthesis of purine nucleotides and thus, as in humans, provide a link to DNA replication and cell division.

Our biochemical identification of PP1 as an AtI-2 interaction partner is supported by our informatics analysis of I-2 across plants and other eukaryotes (Figure 1). This revealed the extent of the conservation of five key regions within the I-2 protein, namely the RVXF, SQ, PXTP, HYNE and [SG]ILK motifs, with only the PXTP motif not playing a role in direct association with PP1. The threonine and proline residues of this latter motif are conserved in every I-2 homologue identified, and phosphorylation of the threonine residue correlates with relief of PP1-activity inhibition and localization in the nucleus in human cells [20,50]. A recent study of the phosphoproteome in plants revealed a phosphorylated threonine residue in the PXTP motif of rice I-2, further supporting the conservation of function of the motif [48]. In addition to that phosphorylation site, they identified two further sites conserved in Arabidopsis, which are located 11 and 13 amino acids C-terminal of the PXTP site. Our examination of Arabidopsis crude extracts indicates at least five distinct phosphorylation states, raising the possibility of other phosphorylation sites and potential further regulation of I-2 function.

The RVXF, HYNE and [SG]ILK motifs all directly bind PP1 in the human complex. Within human I-2, the RVXF motif is somewhat cryptic based on sequence analysis, but was confirmed KSQKW by the human PP1–I-2 co-crystal. The KSQKW sequence binds the hydrophobic cleft where other RVXF motifs dock PP1 [12,24]. Here the side chains of glutamine and tryptophan residues (QKW) play the roles of the valine and phenylalanine residues (VXF). As confirmation of this in the present study, the corresponding valine and tryptophan residues were mutated into alanine, and these two point mutations significantly reduced the inhibition of PP1 by AtI-2, and the vast majority of binding to a PP1–digoxigenin conjugate [15]. Also, incubation with a peptide containing an RVXF motif is sufficient to disrupt the interaction of AtI-2–TAP and PP1 (Supplementary Figure S7).

After the RVXF and PXTP motifs, the HYNE motif is the next most conserved interaction motif, found in all organisms with the exceptions of green algae and the examined kinetoplastids. This motif is actually at the centre of the largest PP1-interaction surface on the human protein, from residues 130 to 169. As the species without the motif still have an intact RVXF and PXTP motif, it would be intriguing to examine if these I-2 proteins are still able to bind to PP1, and what ramifications it has on PP1 regulation. Highly conserved within the vertebrate lineage is the SQ motif, which is phosphorylated by ATM in response to DNA damage and prevents association of PP1 with I-2. I-2 is regarded as one of the most ancient of PP1 interactors [16], but with the SQ motif only being present within vertebrates, this indicates the evolution of this motif and its function late in the history of I-2, but early in the vertebrate lineage. To date, the precise role of dissociating PP1 from I-2 during DNA damage is not clear.

More conserved than the SQ motif, but not nearly to the extent of the others, is the [SG]ILK motif. Found intact only in animals and amoebozoans, there is a question of the exact role this motif plays. Recent studies have indicated that it is a PP1-specific binding motif, of similar calibre to the RVXF motif, and that it, in concert with the RVXF, acts as a building block of interactions [47]. It has been noted that several human PP1-regulatory subunits contain a [SG]ILK motif N-terminal of their RVXF motif (I-2 included), and one protein is also known that only carries a [SG]ILK motif [10,47]. Interestingly, despite the lack of any characterized conserved [SG]ILK motifs in Arabidopsis proteins, the residues that are known to interact with [SG]ILK on human PP1 are conserved, indicating that the ability to bind the motif is inherent to the structure of the conserved PP1 subunit itself.

The specific activities of the isoforms tested towards pNPP are variable (Figure 3A); at the extreme, TOPP1 has nearly six times the activity of TOPP8. The high specific activities of TOPP2, TOPP4, TOPP5 and TOPP7 are supported by our previous work, where endogenous Arabidopsis PP1 activity was selectively purified and the resulting proteins identified. These proteins were the only isoforms unambiguously identified via mass spectrometry (note that TOPP6 was renamed TOPP7 since publication) [27]. This work demonstrates for the first time that all Arabidopsis PP1 isoforms (including TOPP6; see the Results section) are capable of phosphatase activity.

The IC50 values for TOPP inhibition by AtI-2 were variable, ranging from approx. 250 nM to over 1 μM. Although previous studies have observed IC50 values for the human protein between 1 and 50 nM, significantly lower than observed in the present study for AtI-2 [16,51], it must be noted that previous assays used 32P-labelled phosphorylase a, against which PP1 has a significantly greater activity [52]. The combination of a lower sensitivity of detection for the colorimetric assay with the lower activity towards the substrate (therefore more protein used) accounts for the majority of the difference, and a 25–50 fold increase in IC50 value for pNPP assays is common [53]. In direct comparisons using pNPP and human PP1γ, human I-2 has an IC50 value close to that of TOPP2 with AtI-2 (Supplementary Figure S5B).

Human I-2 is found in multiple pools, specifically localized to the centrosomes and diffusely in the cytosol in asynchronous cultures, but during mitosis the PXTP motif is phosphorylated and the centrosomal pool is increased [20,21]. The present study found AtI-2 to be located throughout the cytosol and nucleus in leaf epidermal cells, with an enrichment in the nucleus. PP1 (TOPP2) was found to be enriched in the nucleolus in cells not overexpressing AtI-2, but the nucleolar TOPP2 signal was lost when AtI-2 was also overexpressed (Figure 4). The change in localization of TOPP2 indicates that AtI-2 interacts with TOPP2 in vivo, and actively keeps TOPP2 from nucleoli. This localization is in contrast with actively dividing asynchronous suspension culture, where a specific enrichment in the nucleolus is observed (Supplementary Figure S6). The difference in compartment suggests possible differences in localization by tissue or growth state (suspension cells are actively dividing, whereas leaf epidermal cells are not). The differences in expression are seen in various tissues, and while TOPPs were found in all tissues examined, AtI-2 was found enriched in root tissue, with very little in flowers and siliques.

A. thaliana expresses a homologue of the I-2 family of PP1 inhibitors. This protein (AtI-2) was identified via affinity chromatography with PP1 from suspension cell extracts. Despite a lack of conservation of one of three known sites of PP1–I-2 interaction, AtI-2 inhibits the phosphatase activity of all PP1 isoforms of Arabidopsis (TOPPs) and human PP1γ. These interactions also occur in vivo, and six of the nine TOPPs were found to co-purify with TAP-tagged AtI-2. AtI-2 and TOPP2 co-localize, and AtI-2 overexpression is capable of retargeting TOPP2 out of the nucleolus.

AUTHOR CONTRIBUTION

George W. Templeton performed all experiments and with Greg Moorhead conceived the study and wrote the manuscript. Mhairi Nimick recloned all the TOPP genes into the pET101 vector, and expressed and purified the recombinant proteins. Nicholas Morrice and David Campbell performed the MS for Supplementary Figure S1 and did the N-terminal sequencing. Marilyn Goudreault and Anne-Claude Gingras did the MS and results analysis for the TAP-tag experiments and Atsushi Takemiya and Ken-ichiro Shimazaki supplied the original TOPP constructs.

FUNDING

This work was supported by the Natural Sciences and Engineering Research Council of Canada [grant number 216895].

Abbreviations: AtI-2, Arabidopsis thaliana inhibitor-2; ATM, ataxia telangienctasia mutated kinase; BMe, β-mercaptoethanol; CDK2, cyclin-dependent kinase 2; FA, formic acid; GFP, green fluorescent protein; I-2, inhibitor-2; IPTG, isopropyl β-thiogalactopyranoside; LB, Luria–Bertani; MC, microcystin–Sepharose; Ni-NTA, Ni2+-nitrilotriacetate; pNPP, p-nitrophenyl phosphate; PP1, protein phosphatase 1; PP2A, protein phosphatase 2A; PVP, polyvinylpyrolidone; RFP, red fluorescent protein; TAP, tandem affinity purification; TOPP, type one protein phosphatase

References

View Abstract