Water channel proteins, AQPs (aquaporins), of the PIP (plasma membrane intrinsic protein) subfamily, provide a means for fine and quick adjustments of the plant water status. A molecular model for gating of PIPs by cytosolic protons (H+) and divalent cations was derived from the atomic structure of spinach SoPIP2;1 (Spinacia oleracea PIP2;1) in an open- and a closed-pore conformation. In the present study, we produced the Arabidopsis AtPIP2;1 (Arabidopsis thaliana PIP2;1) homologue in Pichia pastoris, either WT (wild-type) or mutations at residues supposedly involved in gating. Stopped-flow spectrophotometric measurements showed that, upon reconstitution in proteoliposomes, all forms function as water channels. The first functional evidence for a direct gating of PIPs by divalent (bivalent) cations was obtained. In particular, cadmium and manganese were identified, in addition to calcium (Ca2+) and H+ as potent inhibitors of WT AtPIP2;1. Our results further show that His199, the previously identified site for H+ sensing, but also N-terminal located Glu31, and to a lesser extent Asp28, are involved in both divalent-cation- and H+-mediated gating. In contrast, mutation of Arg124 rendered AtPIP2;1 largely insensitive to Ca2+ while remaining fully sensitive to H+. The role of these residues in binding divalent cations and/or stabilizing the open or closed pore conformations is discussed.
Water is crucial for life, and in most organisms its diffusion across membranes is facilitated by AQPs (aquaporins), a conserved family of channel proteins . AQPs assemble as tetramers, each of the monomers defining an individual pore. These monomers exhibit six membrane-spanning α-helices tilted along the plane of the membrane and connected by five loops (A-E) . In all AQPs, loops B and D, as well as the N- and C-terminal tails, are cytoplasmic. To achieve constant adjustment of membrane water permeability in a fast, fluctuating environment, cells have developed multiple controls of AQP function. The regulation of AQP gating, i.e. the opening and closing of the pore, allows a rapid and reversible modulation of water transport. In plants, several cellular effectors can affect the intrinsic water permeability of plasma-membrane AQPs of the PIP (plasma membrane intrinsic protein) subfamily, which have emerged as a model for studying AQP gating. The water permeability of Arabidopsis plasma membrane vesicles was shown to be reversibly inhibited by protons (H+), with a half inhibition at pH 7.2 . A further structure–function analysis identified a histidine residue conserved in cytoplasmic loop D of all PIPs as central for pH-dependent gating . That study and the determination of the atomic structure of spinach SoPIP2;1 (Spinacia oleracea PIP2;1) in an open and a close state  also indicated a possible role for residues of loop D adjacent to the histidine residue  or for residues located in the N-terminal tail . In particular, structural models of SoPIP2;1 suggested that, at acidic pH, the channel is maintained in a closed state by an ionic interaction between the protonated His193 residue of loop D and the carboxylic side chain of Asp28 located in the N-terminus of the protein. During pH-induced closure, displacement of loop D by 16 Å (1 Å=0.1 nm) moves the side chain of Leu197 into the lumen of the pore, thereby creating an hydrophobic barrier and a pore constriction that prevents water permeation.
The activity of water channels in the Arabidopsis plasma membrane can also be inhibited by calcium (Ca2+), with an IC50 of 75 μM . A recent study on plasma membranes from Beta vulgaris storage roots confirmed these observations  and identified both a high (IC50 of 5 nM) and low (IC50 of 200 μM) apparent affinity component for inhibition of membrane water permeability by Ca2+. Recent crystallization of SoPIP2;1 in the presence of 0.1 M CdCl2 and determination of the structure of the protein in a closed state provided further insights into the gating mechanisms induced by divalent (bivalent) cations such as Cd2+ and presumably Ca2+ . In this structure, Asp28 and Glu31 side chains directly interact with Cd2+. Upon binding, Cd2+ mediates a conformational change of loop D and closure of the pore by a network of hydrogen bonding, involving the cytoplasmic side chain of Arg118 in loop B and backbone carbonyl groups of Arg190 and Asp191 in loop D. Although Ca2+ or other divalent cations may substitute for Cd2+ in this model, the defined effects of divalent cations in terms of apparent inhibition constants have remained undefined. As a consequence, the prevalent role proposed for Ca2+ in inhibiting water channels in vivo has remained uncertain.
In the present work we performed a structure–function analysis of PIP gating using water transport assays on a purified protein. AtPIP2;1 (Arabidopsis thaliana PIP2;1) represents one of the most highly expressed PIPs in the roots and leaves of A. thaliana. WT (wild-type) and site-directed mutants of AtPIP2;1 were produced in the yeast Pichia pastoris, purified and reconstituted into proteoliposomes. The effects of a set of divalent cations on the water transport activity of the various AtPIP2;1 forms provided insights into the molecular basis of cation-dependent gating. Moreover, we showed that pH- and cation-dependent gating can be mechanistically dissociated in some specific mutants.
Cloning of WT and site-directed mutants of AtPIP2;1 for P. pastoris expression
The AtPIP2;1 cDNA was amplified by PCR using 2.5 units of the high-fidelity Isis DNA polymerase (MP BioMedicals) and the forward EcoRI-AtPIP2;1 and reverse AtPIP2;1-XhoI primers (Table 1) containing the ATG and STOP codons respectively. Site-directed mutants of AtPIP2;1, where Asp28, Glu31, Arg124 or His199 were replaced by an alanine residue, were obtained by PCR-mediated primer extension  using the forward EcoRI-AtPIP2;1 and reverse AtPIP2;1-XhoI primers, and a pair of complementary mutated oligonucleotides as listed in Table 1. PCR products were cloned into the P. pastoris vector pPICZ-B (Invitrogen) at EcoRI and XhoI restriction sites. After cloning, the sequence of WT or mutated AtPIP2;1 was systematically verified by sequencing (Genoscreen, Lille, France). P. pastoris was transformed by homologous recombination at the chromosomal AOX1 locus. For this, the pPICZB vectors containing WT or mutated AtPIP2;1 cDNAs were linearized with SacI and transferred into the strain X-33 (Invitrogen) using the electroporation method. Transformants were selected on 100 μg/ml zeocin YPDS (1% yeast extract, 2% bactopeptone, 2% dextrose and 1 M sorbitol) plates (Invitrogen) according to the manufacter's instructions or on 500 μg/ml zeocin YPDS plates for isolation of ‘jackpot’ clones.
Production of AtPIP2;1 in P. pastoris
Yeast transformants were selected based on an initial small-scale protein production assay . Briefly, after an initial overnight preculture in a glycerol-containing BMGY (1% yeast extract, 2% bactopeptone, 1.34% yeast nitrogen base, 4×10minus5% biotin, 1% glycerol and 0.1 M potassium phosphate, pH 6.0) medium (Invitrogen), cells were inoculated at 22 °C (D600=1.0) in a methanol-containing BMMY (BMGY medium without the glycerol and replacing it with 0.5% methanol) medium (Invitrogen) in baffled flasks, with shaking at 225 rev./min, to induce recombinant protein expression. Cells were harvested after 24 h and total membrane proteins were extracted as previously described . Immunodetection of AtPIP2;1 in the extracts  was used to select transformants with the highest protein expression level. For large-scale production, membranes were extracted using procedures adapted from  and . All steps were performed at 4 °C with all buffers supplemented with 1% of a cocktail of yeast protease inhibitors (Sigma). Cells were resuspended (5 ml/g of cells) in an extraction buffer [500 mM sucrose, 10% (v/v) glycerol, 20 mM EDTA, 20 mM EGTA, 50 mM NaF, 5 mM β-glycerophosphate, 1 mM phenantroline, 0.6% PVP (polyvinylpyrollidone), 10 mM ascorbic acid, 5 mM DTT (dithiothreitol), 1 mM sodium orthovanadate and 50 mM Tris/HCl, pH 8] and homogenized by high-pressure cavitation at 120 MPa in a Cell Disruptor (Constant System Ltd, Warwick, U.K.). The lysate was centrifuged for 10 min at 10000 g and the resulting surpernatant was further centrifuged for 35 min at 57000 g. The final microsomal pellet was resuspended in 330 mM sucrose, 2 mM DTT, 10 mM NaF and 5 mM potassium phosphate buffer, pH 7.8. Membranes were stripped using urea/alkali treatment , and resuspended in 10% glycerol, 2 mM DTT, 10 mM NaF and 20 mM piperazine, pH 9.5, at a protein concentration of approx. 20 mg/ml, and stored at −80 °C until use.
Solubilization and purification of AtPIP2;1
Stripped membranes were diluted to a protein concentration of 2 mg/ml and solubilized in 3% OTG (octyl β-thioglucopyranoside; Anatrace, Maumee OH, U.S.A.). After 1 h at 28 °C, unsolubilized material was pelleted at 100000 g for 30 min. AtPIP2;1 was purified by anion-exchange chromatography using 1 ml HiTrap Sepharose HP columns (Amersham). After a first equilibration step in buffer A (0.5% OTG, 10% glycerol, 2 mM DTT and 20 mM piperazine, pH 9.5), solubilized proteins were injected into the column and unbound proteins were washed out by 5 ml of buffer A. The retained proteins were then eluted with a linear gradient from 0.1 to 0.65 M NaCl in 20 ml of buffer A. Fractions that were the most enriched in AtPIP2;1 were pooled and AtPIP2;1 was further concentrated by ultrafiltration on 50 kDa filters (Microcon; Millipore). Protein concentration was determined by a BCA (bicinchoninic acid) titration assay (Pierce).
Liposome and proteoliposome reconstitution
Purified AtPIP2;1 was reconstituted into vesicles of Escherichia coli polar lipid extract (Avanti Polar Lipids, Alabaster AL, U.S.A.). The lipids were dissolved in a reconstitution buffer (30 mM KCl and 20 mM Tris/Mes, pH 8.3) to a concentration of 2.5 mg/ml and OTG was added at a concentration of 1% (w/v) to the preformed liposomes. Proteins were added to the liposome solution, at an LPR (lipid to protein ratio) between 16 and 66. After incubation at 28 °C for 30 min, proteoliposomes were formed at room temperature (20 °C) by a detergent removal procedure using SM2 polystyrene beads (Bio-Rad). The mass of beads was 30 times the mass of OTG present in the lipid solution . Vesicle size was normalized at 40 °C to a mean diameter of 0.2 μm by serial extrusion of the proteoliposome suspension through a polycarbonate filter (Mini-extruderi; Avanti Polar Lipids). Control liposomes were prepared in the same way as the proteoliposomes, except that a pure reconstitution buffer instead of a protein sample was added. Probably due to variable experimental conditions (room temperature during liposome formation, quality/age of the Bio-Beads used for detergent removal, quality/age of detergent and lipids etc.) we were not able to obtain the same specific activity for every reconstitution, even when working at the same LPR and with the same protein extract (see Supplementary Table S1 at http://www.BiochemJ.org/bj/415/bj4150409add.htm). This variability yielded an apparent discrepancy in calculated Pf (osmotic water permeability) values. We checked on a set of six independent reconstitutions of AtPIP2;1, all at LPR 33, with Pf measurements performed at pH values of 8.3 and 6.0, that the calculated percentage of AtPIP2;1 Pf inhibition is not dependent on Pf, that is on the reconstitution efficiency of the protein into proteoliposomes. Therefore, we decided to express in the present paper the results as percentage of initial Pf.
Because divalent cations are not diffusible through lipid bilayers, a modified procedure was required to study their effects on vesicle water transport. In brief, the newly reconstituted vesicles were loaded with 0.6 M glycerol for 2 h and diluted 20-fold in a reconstitution buffer complemented with the indicated concentration of divalent cations. The hypo-osmotic shock associated with vesicle dilution induces a transient opening of vesicles and equilibration of their interior with the extravesicular solution . Resealed vesicles were then extruded at 0.2 μm as described above.
Kinetics of vesicle volume adjustment were followed by 90° light scattering at λexcitation=515 nm. Measurements were performed at 15 °C in a SFM3 stopped-flow spectrophotometer (Biologic, Claix, France) essentially as previously described . Briefly, membranes were diluted 10-fold into the reconstitution buffer (105 mosmol·kg of water−1). Alternatively, membranes were diluted in the same medium but adjusted at a modified pH and incubated for 2 h at room temperature. Vesicles were then loaded in the stopped-flow device and mixed (dead time <3 ms) with an equal volume of the reconstitution buffer but with a concentration of 270 mM mannitol (392 mosmol·kg of water−1). This resulted in a 144 mosmol·kg−1 H2O inward osmotic gradient. The traces from ≥10 individual stopped flow acquisitions were averaged and the curves were fitted to single exponential equations to determine an exponential rate constant, Kexp. The Pf was computed from the light scattering time course and the size of membrane vesicles according to the following equation:
where V0 is the initial mean vesicle volume, Av is the mean vesicle surface, Vw is the molar volume of water and Cout is the external osmolality . Because they were extruded, all vesicle preparations were considered to have a mean diameter of 0.2 μm.
Statistical analyses were performed using a Statistica 7 software (StatSoft). Data were analysed using parametric (ANOVA followed by Newman–Keuls tests) and, in case of small samples (n≤6), non-parametric (Mann and Whitney) statistical tests. Results are means±S.E.M. Differences were considered significant at P<0.05.
Production of AtPIP2;1 in P. pastoris
In order to get sufficient quantities of purified AtPIP2;1, production in P. pastoris had first to be optimized. We first observed that, when P. pastoris transformed strains were grown in the presence of methanol, maximum AtPIP2;1 levels as monitored by Western blots on total protein extracts were reached after 24 h of growth while biomass still increased for an extra day (results not shown). In addition, standard selection of P. pastoris strains on 100 μg/ml zeocin led to 60–70% clones being positive for protein expression. Yet, when we selected “jackpot” strains at 500 μg/ml zeocin, although much rarer, we found a much higher level of AtPIP2;1 expression than clones selected in standard conditions (results not shown).
Because of our interest in the mechanisms of AtPIP2;1 gating, we produced a native, untagged form of the protein. Treatment of stripped membranes with OTG was able to fully solubilize AtPIP2;1 and yielded a protein extract with very few apparent contaminant proteins, as shown by Coomassie Blue staining (results not shown). The solubilized protein sample was then fractionated by anionic exchange, and fractions eluted with 0.36–0.42 M NaCl contained most of the AtPIP2;1 protein, which was further concentrated (Figure 1). MS analysis of a tryptic digest of the 28 kDa band shown in Figure 1 revealed that the eight most abundant peptides corresponded to AtPIP2;1 (results not shown). An overall yield of ∼65 μg of AtPIP2;1/litre of culture was obtained using this optimized expression procedure.
SDS/PAGE profile of purified AtPIP2;1
Functional reconstitution of AtPIP2;1 in proteoliposomes
AtPIP2;1 was reconstituted in proteoliposomes at LPRs of 16, 33 and 66 and the water transport properties of the proteoliposomes were compared with those of control E. coli liposomes by stoppedflow spectrophotometry. Figure 2 shows light-scattering recordings of the kinetics of liposome volume adjustment in response to a hypertonic challenge imposed at t=0 in a buffer at pH 8.3. The Kexp (rate constant) of the exponential curve fitted to the experimental light scattering recordings was 89.1±2.6 s−1, 60.6±4.0 s−1 and 21.3±0.9 s−1 for proteoliposomes with LPRs of 16, 33 and 66 respectively (Figure 2). Corresponding Pf values of 691.6±20.2 μm·s−1, 470.2±31.0 μm·s−1 and 165.0±7.0 μm·s−1 were calculated for the three types of proteoliposomes. Control liposomes were reconstituted, either with the reconstitution buffer alone or with proteins tentatively purified, in the same conditions as AtPIP2;1, but from untransformed control yeasts. The two control preparations displayed similar Kexp values of 3.9±0.1 s−1 and 4.3±0.2 s−1 respectively (Figure 2). These correspond to Pf values of 30.6±0.5 μm·s−1 and 33.7±1.3 μm·s−1 respectively.
Stopped-flow light scattering recordings of water transport in liposomes and proteoliposomes containing AtPIP2;1 at different LPRs
Temperature dependence of water transport indicated Ea (activation energy) values of 17.4 kcal·mol−1 and 4.0 kcal·mol−1 for control liposomes and for AtPIP2;1 proteoliposomes reconstituted at LPR=33 respectively (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/415/bj4150409add.htm). The low Pf and high Ea of control liposomes are typical of a lipid-mediated water transport. In contrast, the increased Pf associated with a lower Ea of proteoliposomes containing AtPIP2;1 establish the water channel activity of the reconstituted aquaporin.
pH regulation of AtPIP2;1 water transport activity
To evaluate the H+-dependency of AtPIP2;1-mediated water transport, proteoliposomes reconstituted at pH 8.3 were diluted 10-fold with an iso-osmotic buffer at pH 6.0, and vesicles were osmotically challenged using an hypertonic buffer also equilibrated at pH 6.0. While a short incubation time (∼1 min) in the iso-osmotic buffer at pH 6.0 led to a 50% decrease in Pf, longer incubations (up to 2 h) gradually decreased Pf to 8% of its initial value (Supplementary Figure S2, at http://www.BiochemJ.org/bj/415/bj4150409add.htm). These results suggest that AtPIP2;1 was inserted in the proteoliposomes with about one half of the proteins in an inside-out (cytoplasmic loop D outside) configuration which therefore could be instantaneously blocked by an extravesicular acidification. The time-dependent inhibition of Pf was interpreted as reflecting the time required for H+ diffusion through the lipid bilayer and blockade of AtPIP2;1 proteins in an outside-out configuration. pH-dependency of AtPIP2;1 activity was then tested after a 2 h incubation at 25 °C to allow H+ equilibration on both sides of the membrane. Figure 3 shows the effect of pH ranging from 3.8 to 9.5 on water transport of vesicles, either control or containing AtPIP2;1. Pf of AtPIP2;1 proteoliposomes was maximal at pH>8 and was decreased by 99% at pH 3.8, with a half inhibition at pH 7.15 (±0.15; n=2). Hill coefficients deduced from fitted experimental curves were close to unity (h=1.22±0.23), suggesting a lack of co-operativity between individual monomers during pH-induced channel closure. The pH dependency of the reconstituted AtPIP2;1 is consistent with measurements realized on plant plasma membrane vesicles or on Xenopus oocytes expressing AtPIP2;2 [3,4,6].
Effects of pH on Pf of liposomes and proteoliposomes containing AtPIP2;1
Regulation of AtPIP2;1 water transport activity by divalent cations
To evaluate the ability of divalent cations, and Ca2+ in particular, to directly gate AtPIP2;1, 1 mM CaCl2 was added directly on proteoliposomes in an iso-osmotic solution 3 min prior to the water transport assay. This treatment resulted in a ∼50% inhibition of Pf (results not shown). To ensure that Ca2+ had access to the interior of the vesicles and could block AtPIP2;1 in either orientation, proteoliposomes were submitted to a sudden hypo-osmotic shock in the presence of 1 mM CaCl2. This shock induces a transient opening of the vesicles and equilibration of their interior with the extra-vesicular solution . Under these conditions, the Pf of AtPIP2;1 vesicles was reduced by 86% (results not shown). We also checked that a similar hypotonic treatment in the presence of Ca2+ did not alter the Pf of control liposomes (results not shown). In experiments where the concentration of free Ca2+ was varied between 0 and 5 mM (Figure 4), we observed a monophasic dose–response curve with an apparent IC50 of 42±25 μM (n=5). The fitted Hill coefficient was close to unity (h=0.83±0.11), suggesting a lack of co-operativity during the process of Ca2+-dependent Pf inhibition. These data provide evidence that Ca2+ can act directly on AtPIP2;1 to induce pore closure.
Effects of Ca2+ on Pf of liposomes (control) and proteoliposomes containing AtPIP2;1 (AtPIP2;1)
Effects of a series of divalent cations on Pf of AtPIP2;1 vesicles were compared at a fixed concentration of 150 μM (Figure 5). We tested divalent cations that belong to the same alkaline earth metal series as Ca2+ (Mg2+, Ba2+ or Sr2+), display a similar atomic radius (Mn2+), are known to bind to plant SoPIP2;1 (Cd2+) or to regulate activity of animal AQPs (Ni2+) [5,14]. In addition, we tested the effects of mercury ions (Hg2+) which have been reported as common AQP blockers . Surprisingly, Hg2+ was not able to inhibit the Pf of AtPIP2;1 vesicles (Figure 5). Mg2+ did not significantly alter water transport in AtPIP2;1 vesicles either, whereas Ca2+ inhibited Pf by 60%. Sr2+, Ba2+ and Ni2+ reduced the Pf by 40%, 37% and 48% respectively. Mn2+ induced an inhibition of Pf by 60%, similar to Ca2+. Cd2+ was the most effective cation with an inhibition of Pf by 70% (Figure 5). Dose–response experiments for inhibition of Pf in AtPIP2;1 vesicles by Cd2+, Mn2+ and Ni2+ indicated apparent IC50 values of 27±13 μM, 85±37 μM and 178±46 μM respectively (Figure 6). This set of results clearly indicates that AtPIP2;1 can be blocked by a wide range of divalent cations with distinct sensitivities, Ca2+ being one of the most efficient blockers.
Effects of a series of divalent cations on the Pf of liposomes and proteoliposomes reconstituted with AtPIP2;1
Dose-dependent effects of Cd2+, Mn2+ and Ni2+ on Pf of proteoliposomes containing AtPIP2;1
Cytoplasmic determinants for regulation of AtPIP2;1 by divalent cations and pH
To examine the molecular bases of AtPIP2;1 sensitivity to H+ or divalent cations, we investigated the effects of point mutations of four residues that may be involved in these regulations. We selected Asp28, Glu31, Arg124 and His199, based on the possible role of the corresponding residues of SoPIP2;1, as was proposed from the atomic structures of the protein in open and closed states . The four residues were individually mutated to alanine and the resulting D28A, E31A, R124A and H199A forms were expressed in P. pastoris and purified as described for WT AtPIP2;1 (Supplementary Figure S4, at http://www.BiochemJ.org/bj/415/bj4150409add.htm). In this series of experiments, functional reconstitution of the mutant or WT proteins sometimes failed at the highest protein content (LPR range 16–33). Therefore, we analysed the Pf of proteoliposomes reconstituted at a lower protein content (LPR range 33–40), which ensured a very significant contribution of AQPs to Pf (Supplementary Figure S3, at http://www.BiochemJ.org/bj/415/bj4150409add.htm). When compared at a similar LPR, the increased water channel activity of proteoliposomes containing the WT or mutant proteins indicated that no major change in intrinsic water permeability had been induced by either one of the introduced mutations (results not shown). The sensitivities of mutant and WT proteins to H+ were compared at pH 6.0 (Figure 7A) and under intermediate blocking conditions by divalent cations (Ca2+, Cd2+ and Mn2+; all at 150 μM), to discriminate more easily between weakly and strongly affected mutants (Figure 7B).
Effects of H+ and divalent cations on Pf of proteoliposomes containing AtPIP2;1 WT or mutants (D28A, E31A, R124A and H199A)
The R124A mutant showed an inhibition of Pf by H+ (84.4±1.5%) not significantly different from that of WT (83.7±2.8%) (Figure 7A). By contrast, the H199A mutant was almost insensitive to H+ (5.5±4.0%), whereas the D28A and E31A forms showed intermediate Pf inhibitory responses, by 54.3±4.4% and 56.5±4.6% respectively. These results confirm a central role for His199 in H+-induced pore closure and provide the first functional data for a role of the N-terminal tail in this process.
In parallel experiments, the WT form showed a Pf inhibition of 39.8±8.1%, 61.0±5.8% and 50.0±7.7% in response to 150 μM Ca2+, Cd2+, and Mn2+ respectively. With respect to the WT form, the D28A mutant showed a tendency to lower inhibition by the three divalent cations (27.1±8.8%, 43.8±9.8% and 41.3±6.8%). By contrast, the E31A mutation had significant effects on Ca2+ and Cd2+ sensitivities, with Pf inhibitions of 6.4±7.1% and 21.6±14.2% respectively. Yet the inhibition of Pf by Mn2+ (36.8±7.9%) was similar to that of the WT form. The R124A form was significantly affected in its Pf sensitivity to all divalent cations, and the percentages of inhibition by Ca2+ (10.0±5.7%), Cd2+ (18.7±11.2%) and Mn2+ (26.3±7.0%) were all reduced by >2-fold with respect to those of the WT form. Finally, the H199A mutant also showed a very poor sensitivity to divalent cations with a Pf inhibition of 3.9±7.9%, 11.6±13.7% and 4.9±14.0% in response to 150 μM of Ca2+, Cd2+ and Mn2+ respectively. Dose-dependent effects of Ca2+, Cd2+ and Mn2+ on the Pf of liposomes (control) and proteoliposomes reconstituted with AtPIP2;1 WT or mutant (D28A, E31A, R124A and H199A) forms are shown in Supplementary Figure S4. These experiments showed a relative insensitivity of the four mutants to high Ca2+ concentration (1 mM). In contrast, a similar concentration of Cd2+ and Mn2+ was able to fully inhibit the R124A mutant form while being poorly active on D28A, E31A, and H199A. In conclusion, these analyses showed that all tested residues (Asp28, Glu31, Arg124 and His199) are involved in divalent cation-dependent gating of AtPIP2;1, although to different extents. Furthermore, replacement of Glu31 by an alanine residue revealed a striking ability of the mutant protein to discriminate between distinct divalent cations.
The present work reports on the reconstitution in proteoliposomes of AtPIP2;1, one of the most highly expressed PIPs in Arabidopsis. The protein conferred on liposomes extremely high water permeability, with Pf levels exceeding 1000 μm·s−1. Noticeably, these values are similar to those obtained with animal AQPs  and are the highest obtained for a reconstituted plant aquaporin [8,17]. PIP aquaporins have been described to be post-translationally regulated by several factors or modifications, such as H+, free Ca2+ or serine phosphorylation, all acting on the cytosolic side of the protein [3,4,6,18]. Whereas functional expression of PIPs in Xenopus oocytes was instrumental for deciphering the regulation of water channel activity by H+ and phosphorylation, structure–function studies addressing the regulation of PIPs by divalent cations have been lacking.
In the present study we have demonstrated that the water channel activity of AtPIP2;1 can be fully blocked by H+, with a half-inhibition at pH 7.15. In addition, AtPIP2;1 was shown to be inhibited by divalent cations, with the highest efficiency shown by Ca2+, Cd2+ and Mn2+. With respect to previous studies on plasma membrane vesicles purified from plants, the present work represents the first description of the direct inhibition of an individual PIP isoform by divalent cations. In particular, AtPIP2;1 showed an IC50 of 42 μM (±25 μM) for Ca2+-mediated Pf inhibition, which is in good agreement with the IC50 of 75 μM observed for plasma membrane vesicles from Arabidopsis suspension cells . In contrast, Ca2+-dependent inhibition of Pf in plasma membranes from Beta vulgaris storage roots  showed both a high (IC50∼5 nM) and a low (IC50∼200 μM) apparent affinity component. Because of their conserved structure, all PIP isoforms are likely to be blocked by Ca2+ but may differ in their sensitivity. The response of Beta vulgaris plasma membrane, whose composition of AQPs is unknown, may reflect this heterogeneity. Because PIPs can be divided in two subclasses, PIP1 and PIP2, it will be interesting to determine the sensitivity of PIP1s to Ca2+ inhibition. Surprisingly, AtPIP2;1 was not inhibited by mercury ions, although it has the same conserved cysteine residues as most other PIPs. Nevertheless, our observations are consistent with previous studies in Xenopus oocytes showing that AtPIP2;1 was insensitive to 1 mM HgCl2, whereas four other A. thaliana PIP isoforms (AtPIP1;1, AtPIP1;2, AtPIP1;3 and AtPIP2;2) were significantly blocked .
As the in vitro inhibition of PIPs by divalent cations such as Ca2+ or Cd2+ exhibits a relatively low sensitivity, the physiological significance of this phenomenon has been questioned. Peak free Ca2+ concentrations in the low micromolar range have been measured in the plant cytosol . However, free Ca2+ may locally reach higher concentrations, especially in the close vicinity of Ca2+ channels. In these respects, PIPs were found in detergent-resistant membranes [21,22], where they may cluster with other membrane transport proteins. Exposure of plants to trace concentrations of Cd2+ leads to accumulation of this highly toxic ion in plant cells and can affect the plant–water relations [23,24]. Our results suggest that one basis of these effects may be a direct gating of PIPs by the ion.
The atomic structures of spinach SoPIP2;1 in an open and a closed conformation  have provided a landmark advance for investigating the molecular determinants of PIP gating. The present structure–function approach provides a step forward and allows one to confirm and refine the proposed model . The atomic structure of SoPIP2;1 in the closed state showed a Cd2+ ion in close vicinity to the Asp28 and Glu31 residues . Although the D28A mutant shows a tendency to a lower inhibition by divalent cations, our results establish a clear functional role for Glu31 only. The results suggest that the latter residue is the main actor of this process, probably due to a higher flexibility of the Glu than the Asp side chain . The role of Glu31 in primarily binding the ion is further supported by the finding that its mutation selectively altered Ca2+, and to a lesser extent Cd2+, inhibition of Pf, whereas it did not significantly alter Mn2+ inhibition. This is in agreement with the notion that Ca2+ usually displays higher co-ordination numbers than Cd2+ or Mn2+ . Similar to Glu31, replacement of Arg124 drastically altered gating of PIPs in response to Ca2+ and Cd2+. Here, our results support the proposed role of this residue in bridging the N-terminal binding site of ions with loop D, in order to stabilize the closed conformation of the pore (Figure 8). We note, however, that, since Glu31 and Arg124 likely form hydrogen bonds , each of the two residues may actually contribute to both ion binding and stabilization of the closed-pore conformation. More unexpectedly, our analysis uncovered a role for His199 in loop D in divalent cation dependent-gating. Here, the analysis of the open and the closed conformations is not giving a direct clue to explain this result. We propose that the mutation of this or neighbouring residues may alter loop D local structure and unstabilize the closed conformation, leading to a constitutively open pore.
Interpretative model of PIP2 gating by H+- and divalent cations
The present work also provides insights into the mechanism of PIP gating by H+. Firstly, we confirmed by an independent approach that a conserved histidine residue in the loop D of all PIPs is a major H+ sensor . In view of the results discussed above, we cannot, however, rule out the hypothesis that replacement of His199 alters the loop D local conformation, thus leading to an apparent H+-insensitivity. However, the half inhibition of Pf at pH 7.15 supports the hypothesis that a residue with a near neutral pKa, which is a histidine residue, acts as the sensor. From the atomic structure of SoPIP2;1 in a closed conformation, it was proposed that the protonated histidine residue would interact through a salt bridge with the carboxylic group of residue 28, found as a glutamate or an aspartate, in PIP1s and in PIP2s respectively . Our structure–function analysis confirms this interaction, but also unravels a role for Glu31. Therefore, the present study indicates a flexibility for the salt bridge that again was not deduced from the atomic structure. Interestingly, conserved charged residues adjacent of the loop D histidine (namely Arg196 and Asp197 in AtPIP2;1) can confer, after mutation to alanine, H+-insensitivity to the protein . A closed-conformation model of AtPIP2;1 suggests that Asp197 could make a salt bridge with the N-terminal Lys33, conserved in most PIP2s (Figure 8). Noticeably, this positively charged residue is not found in PIP1s, and these AQPs may possibly exhibit a sensitivity to H+ slightly different from that of PIP2s.
Changes in H+ and free Ca2+ concentrations in the cytosol are used by plant cells, independently or in combination, to transduce developmental, hormonal or stress signals [27–30]. One well identified downstream effect of Ca2+ is the stimulation of Ca2+-dependent protein kinases, which have been reported to act on PIPs . Therefore, PIP gating is the converging point of multiple cell signalling mechanisms. The atomic structure of PIPs together with the present structure–function analysis show that the gating of PIPs by H+ and divalent cations relies at least in part on common molecular mechanisms. The present study also identified mechanisms that are specific for divalent cation-dependent gating. In particular, we showed that, although Arg124 is not directly involved in ion binding, a R124A mutant was largely insensitive to Ca2+ while remaining fully sensitive to H+. The expression in transgenic plants of this type of mutant will be instrumental to test for the physiological significance and specific contribution of divalent cations in AQP gating and regulation of the whole plant water status. Phosphorylation of PIP2s in loop B and in the C-terminal tail may also exert a control on pore opening and closing [18,31]. Therefore, a next step to add to our approach will be to understand the interplay between the multiple phosphorylated states of PIPs and their direct gating by H+ and Ca2+.
We are grateful to Dr Véronique Santoni for MS analyses performed during the study. This work was supported in part by a grant from the French Ministry of Research (ACI2003 ‘Biologie du Développement et Physiologie Intégrative’).