H2O2 is a relatively long-lived reactive oxygen species that signals between cells and organisms. H2O2 signalling in plants is essential for response to stress, defence against pathogens and the regulation of programmed cell death. Although H2O2 diffusion across membranes is often considered as a passive property of lipid bilayers, native membranes represent significant barriers for H2O2. In the present study we addressed the question of whether channels might facilitate H2O2 conduction across plasma membranes. The expression of several plant plasma membrane aquaporins in yeast, including PIP2;1 from Arabidopsis (where PIP is plasma membrane intrinsic protein), enhanced the toxicity of H2O2 and increased the fluorescence of dye-loaded yeast when exposed to H2O2. The sensitivity of aquaporin-expressing yeast to H2O2 was altered by mutations that alter gating and the selectivity of the aquaporins. The conduction of water, H2O2 and urea was compared, using molecular dynamics simulations based on the crystal structure of SoPIP2;1 from spinach. The calculations identify differences in the conduction between the substrates and reveal channel residues critically involved in H2O2 conduction. The results of the calculations on tetramers and monomers are in agreement with the biochemical data. Taken together, the results strongly suggest that plasma membrane aquaporin pores determine the efficiency of H2O2 signalling between cells. Aquaporins are present in most species and their capacity to facilitate the diffusion of H2O2 may be of physiological significance in many organisms and particularly in communication between different species.
Water channels, aquaporins, have been identified in most organisms, including plants, and have specific physiological functions related to water facilitation in many tissues [1–3]. Aquaporins are tetramers and each subunit forms a pore. Many plant aquaporin homologues are believed to specifically conduct water, whereas several homologues also conduct other small neutral solutes, such as glycerol, and are, therefore, called aquaglyceroporins [1–3]. The ar/R (aromatic/arginine) region and the highly conserved NPA (for asparagine, proline, alanine) region have been implicated in the selectivity of aquaporins. In the model plant Arabidopsis thaliana, 35 aquaporin homologues are encoded in the genome, and several of them conduct urea or ammonia in addition to water [4,5]. For simplicity, we will designate the proteins of this family, which are also called major intrinsic proteins, as aquaporins throughout the present paper, despite the ability of several homologues to conduct other solutes.
In several organisms, including mammals and plants, the expression of water channels is highly regulated at the transcriptional level. In addition, the conductance of several aquaporins can be post-transcriptionally regulated by different stimuli. Regulation by phosphorylation [6,7], calcium , extracellular pH  or intracellular pH  has been reported previously. Crystal structures have unambiguously proven that the plant plasma membrane aquaporin SoPIP2;1 (where PIP is plasma membrane intrinsic protein) exists in two different, phosphorylation-dependent conformational states, an open and a closed state . These conformations mostly differ in their cytoplasmic pore openings, but the size and geometry of the narrowest part of the pore, the ar/R region, is nearly identical.
H2O2 is a relatively long-lived ROS (reactive oxygen species) and can function as a signalling molecule in eukaryotes, leading to specific downstream responses [12,13]. H2O2 is a small neutral solute and membranes appear to play an important role in protecting cells against its toxicity, which may derive from H2O2 itself or from the more toxic hydroxyl radical derived from H2O2. This implies that native membranes restrict free H2O2 diffusion and that significant H2O2 gradients across membranes exist. Cellular H2O2 redox chemistry is mainly restricted to compartments surrounded by membranes such as peroxisomes and mitochondria and may also be found in other endomembrane compartments .
In plants, H2O2 is generated and released upon various stresses, e.g. during nutrient starvation . H2O2 is a signal that controls plant programmed cell death . Specific H2O2 generation in plants often involves plasma membrane NADPH oxidases that produce the superoxide anion, by enabling the transfer of electrons from the cytosolic NADPH across the membrane to external molecular oxygen. The spontaneous dismutation of the short-lived superoxide anion (that is very effective at the low apoplastic pH) will produce external H2O2. This is followed by H2O2 influx and the subsequent activation of Ca2+ channels . Such NADPH-dependent processes have been identified in the plant defence against attack by pathogens (‘oxidative burst’), root hair development  and the regulation of abscisic acid signalling and stomatal movements [12,17]. A recent study has shown that the human aquaporin 8 and the plant TIPs (tonoplast intrinsic proteins) AtTIP1;1 and AtTIP1;2 conduct H2O2 when heterologously expressed in yeast . Previous work had suggested that algae plasma membranes may contain pores for H2O2 [19a]. In that work, theoretical calculations combined with measurements of the membrane permeability at very high concentrations (up to 350 mM) and the relatively unspecific aquaporin inhibitor HgCl2 were used .
In the present study, we show evidence that several plant plasma membrane aquaporin homologues conduct the signalling- and defence-related small neutral solute H2O2. Furthermore, molecular simulations identify channel residues critically involved in H2O2 conduction. The conservation of these residues suggests that many plasma membrane aquaporins from plants facilitate the diffusion of H2O2. Although many aquaporins have a well-defined physiological function in water conduction, aquaporins may thus also participate in regulating and controlling H2O2 signalling.
Yeast strains, media and growth conditions
The Saccharomyces cerevisiae strains BY4741 (MATa, his3Δ1, leu2Δ0, met15Δ0, ura3Δ0) (Euroscarf), Δdur3 (MATα, dur3Δ, ura3Δ) and 23344c (MATα, ura3Δ)  were employed for growth tests. Uptake studies were performed in the yca1 yeast mutant (MATa, his3Δ1, leu2Δ0, met15Δ0, ura3Δ0, yor197w::kanMX4) (Euroscarf, Y02453). Yor197w encodes a putative cysteine protease that is similar to mammalian caspases and is involved in the regulation of apoptosis upon H2O2 treatment. For growth tests, cells were grown up to stationary phase in liquid YNB (yeast nitrogen base) medium (1.7 g/l yeast nitrogen base, 3.0% glucose and 1% arginine, supplemented with histidine, leucine and methionine when appropriate) at 28 °C and 160 rev./min. The cells were spotted as a 5-fold dilution series on freshly prepared plates containing H2O2 ranging from 0.25 mM up to 2.5 mM. Pictures of cells were taken after 96 h.
Plasmid constructs for yeast expression
The open reading frames of AtPIP1;1 (At3g61430), AtPIP2;1 (At3g53420), AtPIP2;4 (At5g60660), AtNIP1;1 (where NIP is NOD26-like intrinsic protein) (At4g19030) and AtNIP1;2 (At4g18910) were amplified from an Arabidopsis thaliana (Col-0) cDNA library, cloned into the pGEM-T Easy vector (Promega) and then subcloned into pDR195. AtTIP2;3 has been described previously . The point mutations were inserted using the QuikChange® kit from Stratagene. All constructs were verified by sequencing at GATC (Gesellschaft für Analyse-Technik und Consulting, Konstanz, Germany). The sequence encoding the N-terminal c-Myc epitope was generated by PCR. SDS/PAGE analysis and Western blot analysis followed standard protocols using a monoclonal anti-c-Myc antibody (Sigma). The yeast oxysterol-binding protein Kes1 was used as a normalization control and visualized with a monoclonal antibody [a gift of Vytas Bankaitis (Department of Cell and Developmental Biology, University of North Carolina School of Medicine, Chapel Hill, NC, U.S.A.)].
To reduce endogenous production of H2O2, the yca1 yeast mutant was employed. This mutant is partially defective in H2O2-induced apoptosis . Transformants were grown in selective medium to midlog phase, and successively supplied with the oxidant-sensitive fluorescent dye H2DCFDA (2′7′-dichlorodihydrofluorescein diacetate; Molecular Probes) to a final concentration of 20 μM and incubated at 28 °C and 550 rev./min for 45 min to allow uptake of the dye. Cells were then washed in water and resuspended in an experimental solution consisting of 40 mM sodium phosphate (pH 5.8) containing 3% (w/v) glucose. The fluorescence was measured before and after the addition of H2O2 (final concentration 2 mM or 10 mM). The fluorescence was measured in 10 min steps using a TECAN 96-well spectrofluorimeter, set at an excitation wavelength of 493 nm and an emission wavelength of 524 nm.
Structure of the open SoPIP2;1 aquaporin
A model of the ‘open conformation’ SoPIP2;1 aquaporin was obtained by combining the open and closed pore crystal structures of SoPIP2;1. The closed structure (PDB accession number 1Z98; Ala24–Val179 and Pro201–Ser274) and the open structure (PDB accession 2B5F; Phe180–Leu200) were combined , and a structural alignment was performed using VMD (see Supplementary Figure S2 at http://www.BiochemJ.org/bj/414/bj4140053add.htm). Serine residues 115 and 274 were exchanged to phosphoserines in order to keep the channel in an open state . To avoid an artificial negative charge at the C-terminus (as in the partially resolved C-terminus in the structure), a phenylalanine residue, which is not resolved in the crystal, was added to this critical region. For the simulations, this ‘open conformation’ SoPIP2;1 was embedded into a pre-equilibrated 70 Å×70 Å (1 Å=0.1 nm) (as a monomer; Supplementary Figure S3 at http://www.BiochemJ.org/bj/414/bj4140053add.htm) or 130 Å×130 Å (as a tetramer) spanning POPE (16:0/18:1c9-palmitoyloleyl-phosphatidylethanolamine) lipid bilayer.
The system was hydrated with TIP3P water molecules and counter ions were added to obtain a neutral net charge, using the module LEAP from the AMBER8 suite. The total tetrameric system included approx. 111.200 atoms. One substrate molecule was placed above each monomer (>−30 Å in z-direction from the Cα of Val206; Figure 6) and a fifth above the central cavity of the tetramer (Figure 5). Within a tetramer, the starting position of the substrates differed by ±3 Å in the lateral direction from the central axis of each monomer. These four different starting positions were identical for all substrates (i.e. for the centre of mass of each substrate).
MD (molecular dynamics) simulations
MD was carried out using the NAMD2 program [19a] using the Amber03 force field. Force-field parameters for H2O2 and the POPE lipid bilayer were estimated using the module ANTECHAMBER and the general AMBER force field (GAFF) of AMBER8. The bond angles, dipole moment and other properties matched the well-known properties of H2O2 . The force-field parameters for urea were provided with AMBER8 and the force-field parameters for phosphoserine were obtained from the AMBER Parameter Database (http://pharmacy.man.ac.uk/amber/) . The system was initially minimized for 1000 steps with the protein atoms fixed. Then another 1000 minimization steps were applied without any atoms fixed. The system was then equilibrated for 2 ns using a timestep of 2 fs with PBCs (periodic boundary conditions). A pressure of 1 atm (101.3 kPa) and a temperature of 310 K were applied during equilibration and were kept constant using the Langevin Piston method. Non-bounded interactions were calculated with a cutoff of 12 Å, and the long-range electrostatic forces where calculated using the PME (Particle Mesh Ewald) method. To keep the protein in position, the Cα atoms of the protein backbone were constrained with a harmonic force of 0.5 kcal·mol−1·Å−2.
For monomeric simulations, PBCs were used, but the individual MD simulations on tetramers were carried out without PBC, PME and Langevin dynamics. To simulate the permeation across the aquaporin channels, the system was confined by applying a harmonic constraint of 200 kcal·mol−1·Å−2 to the 3.5 Å outer layer of water. A second water layer of the same thickness below was constrained with 5 kcal·mol−1·Å−2. To keep simulation times reasonable, a constant force vector in the z-direction was applied to the permeant molecule, either water, H2O2 or urea . In initial trials we found that the constant force of ∼41 pN was required so that solutes entered the channel entrance within a few ns. The trajectories along the z-axis were calculated relative to the Cα atom of Val206 as a reference. The entrance point into the channel along the z co-ordinate was defined as Cα of Lys144 (on average −20.5 Å±1.5 Å from valine) and the internal channel exit was defined as Cα of Pro33 (19.2 Å±1.3 Å from the Val206).
Residency time and energy profiles
To construct the overall histogram of the time at each position along the reference axis, the residence time measured for 0.2 Å intervals of each simulation was accumulated in a weighted sum, such that the biased potentials due to the applied force were discounted. The energy profiles were then estimated in 1 Å intervals along the channel axis from transition state theory as in :
where τ is the residence time, ΔG‡ is the Gibbs free energy of activation, T is the absolute temperature, R is the universal gas constant, kb is Boltzmann's constant and h is Planck's constant.
Expression of AtPIP2;1 increased H2O2 sensitivity and H2O2 accumulation in yeast
We considered the possibility that aquaporins could mediate the conduction of H2O2 across membranes, since some channel homologues have been shown to conduct several low-molecular-mass molecules. Initial studies focused on the Arabidopsis protein AtPIP2;1. When the Arabidopsis AtPIP2;1 was heterologously expressed in yeast, growth and doubling times were unaffected in standard liquid culture or solid medium. However, expression of AtPIP2;1 increased sensitivity of yeast cells to H2O2 (Figure 1A). Growth of cells expressing AtPIP2;1 was affected in a dose-dependent manner and impaired at 400 μM H2O2, while non-expressing controls, transformed with the vector pDR195 without the aquaporin coding insert, were not sensitive up to 1.6 mM H2O2 (Figure 1A). In a similar growth assay, the plant tonoplast aquaporin TIP1;1 was recently shown to conduct H2O2 [13,19]. In principle, reduced or impaired growth may be due to factors which perturb the metabolism in an unknown way.
S. cerevisiae is sensitive to H2O2 and sensitivity is increased by aquaporin expression
To further test whether the increased sensitivity of cells expressing AtPIP2;1 was due to increased uptake of H2O2, a carboxylated derivative of the oxidant-sensitive probe, DCF (2′,7′-dichlorofluorescein), was employed. When cells were incubated with a di-ester derivative of the probe, it entered the cytosol by passive diffusion. Once inside the cell, the esters are cleaved by endogenous esterases, rendering the probe impermeable to the membrane. The DCF can be oxidized by ROS such as H2O2, resulting in a fluorescence increase . To reduce the production of internally generated ROS by H2O2-stimulated apoptosis, we used the yeast yca1 mutant that is partially defective in undergoing apoptosis . Upon exposure to H2O2, the overall intracellular accumulation of ROS that was measured with the fluorescent dye was higher in AtPIP2;1-expressing yeast cells compared with cells transformed with empty vector alone. Consistent with these results, the c-Myc-epitope-tagged AtPIP2;1 was stable in yeast and detected as a monomeric protein of ∼30 kDa (Figure 2A) and as an oligomeric protein (results not shown).
Mutations that affect water conduction also affect sensitivity to H2O2
Sensitivity of yeast to H2O2 is altered by a mutation that reduces PIP2 water conductance
A conserved cytosolic histidine residue has been shown to be involved in PIP2 regulation of water flux by cytosolic pH, a phenomenon that was initially demonstrated using the homologue AtPIP2;2 . Replacement of this histidine residue by a lysine residue in AtPIP2;2 reduced water conduction and mimicked a closed channel . The relevant histidine residue at position 199 in AtPIP2;1 (Figure 2B) was replaced by a lysine residue (H199K) by site-directed mutagenesis. As expected, the sensitivity of the H199K-mutant-expressing yeast to growth on H2O2 was similar to vector-transformed controls (Figure 2C), although a similar protein level was detected from this mutant in Western blot analysis (Figure 2A).
Sensitivity of yeast to H2O2 is altered by pore mutations that alter selectivity
Residues in the constriction region of the pores can affect the selectivity properties of aquaporins [28,29]. Alignments and homology models indicate that the constriction region of all PIP-type aquaporins from Arabidopsis is identical. In AtPIP2;1, the constriction (ar/R) region is composed of the four residues Phe87, His216, Thr225 and Arg231. In contrast, the ‘selectivity filter’ residues in NIP-type major intrinsic proteins differ, which leads to a larger pore diameter . A mutant, AtPIP2;1NIP1;2, in which the ar/R residues in AtPIP2;1 were exchanged to the corresponding residues in AtNIP1;2 (F87W, H216V, T225A) also conducted H2O2, as was deduced from its ability to confer increased sensitivity on H2O2 when expressed in yeast (Figure 3C). The ar/R constriction region in this mutant is wider, as can be easily deduced from homology models of this mutant (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/414/bj4140053add.htm). The residues of the ar/R region of AtPIP2;1 were also exchanged to the corresponding residues of AtNIP5;1, which conducts hydrophobic, non-charged solutes as large as boric acid, but not water, at a high rate . By introducing the three relevant mutations in AtPIP2;1 (F87A, H216I, T225G), the ar/R constriction region was even wider in this mutant channel (designated AtPIP2;1NIP5;1) (Supplementary Figure S1).
Pore mutations alter sensitivity to H2O2 and various other substrates
Yeast growth is sensitive to high concentrations of toxic boron. AtPIP2;1 expression affected neither the boric acid sensitivity of yeast, nor enhanced urea uptake on selective plates (Figures 3B and 3D). The NIP1;2-like aquaporin mutant (AtPIP2;1NIP1;2) did not exhibit novel properties compared with AtPIP2;1. In contrast, the expression of AtPIP2;1NIP5;1 enhanced its growth sensitivity to boric acid, indicating boron uptake (Figure 3B). This mutant form of AtPIP2;1 dramatically increased the sensitivity of yeast growth to H2O2 beyond that of wild-type AtPIP2;1 (Figure 3C). When expressed in a yeast mutant that lacks endogenous urea transporters (Δdur3; ) this construct enabled yeast to grow on urea as a sole nitrogen source, indicating urea conduction across the plasma membrane (Figure 3D). Thus an AtPIP2;1 aquaporin with a selectivity filter as in AtNIP5;1 apparently facilitates the conduction of larger solutes, including H2O2. These observations are in agreement with earlier studies  that showed that residues in the ar/R region are critical for substrate selectivity.
H2O2 conduction by some, but not all, aquaporin homologues
Representatives of individual sub-groups of the plant aquaporins were tested for their capacity to impair yeast growth on plates containing H2O2. Although cells expressing either AtPIP1;1 or AtNIP1;1 were similar to controls, the yeast expressing AtNIP1;2 or the tonoplast-intrinsic AtTIP2;3 exhibited an intermediate phenotype (Figure 4). It is interesting to note that AtNIP1;1 had been found to conduct glycerol across the plasma membrane when expressed in yeast . Despite strong expression of the c-Myc-tagged AtNIP1;1 in yeast (Figure 2A), it did not increase the sensitivity to H2O2 (Figure 4). We conclude that H2O2 conduction is not supported by NIP1;1, at least when heterologously expressed in yeast. The AtPIP2;4-expressing cells were the most sensitive to H2O2 (Figure 4). Whether these differences in H2O2 sensitivity are due to differences in the open probability of the channel, (mis-)targeting of the membrane proteins to other membranes in the heterologous system or to protein stability, remains unclear at this point. It is, however, notable that AtPIP1;1, AtPIP2;1 and AtPIP2;4 have identical residues at the selectivity filter, which differ in other family members such as AtNIP5;1. It has been repeatedly noted that PIP1-type aquaporins are not active in heterologous systems . This may explain why AtPIP1;1 expression did not alter sensitivity to H2O2 (Figure 4). The tonoplast intrinsic AtTIP2;3, which is known to facilitate ammonia and urea conduction when heterologously expressed in yeast , also conducts H2O2. This homologue is known to facilitate solute fluxes across the plasma membrane of yeast, despite its intracellular localization in planta . Other TIP-type aquaporins, with further different selectivity filter residues in the ar/R region, have been shown to conduct H2O2 .
Several plant aquaporins increase H2O2 sensitivity
Molecular simulation of conduction through PIP2;1
The plasma membrane-localized aquaporin AtPIP2;1 from Arabidopsis is highly similar to other PIP2-type aquaporins, such as the crystallized SoPIP2;1 channel from spinach. The closest Arabidopsis homologue of SoPIP2;1 is AtPIP2;7, but all PIP2-type aquaporins share a very high conservation within the pore and have the identical selectivity filter residues. Open and closed conformations of the SoPIP2;1 structure are available . We performed MD studies on the open SoPIP2;1 channel to compare H2O and H2O2 conduction through plant plasma membrane aquaporins. We initially observed that the closed pores of the tetrameric SoPIP2;1 were highly stable during minimization, equilibration and further simulation, when embedded in a lipid bilayer. In addition, the pores readily filled with a single-file of water during equilibration, but water conduction was impaired by loop D, which closes the pore from the cytosolic side. In contrast, the structures of the lower-resolution open monomers, or homology-modelled AtPIP2;1, were less stable during similar simulations and no continuous, stable water line formed, as described in previous simulations [29,32]. This was probably mainly due to the lower structural resolution (results not shown).
We therefore constructed open-pore monomers based on the high-resolution structure of the closed pore, where the backbone came from the closed pore, but loop D was taken from the open conformation (Supplementary Figure S2). MD simulations on a modified ‘open’ model of the closed pore structures have already been published . The ‘open’ pore tetramer embedded in a POPE lipid bilayer is shown in Figure 5(A).
Setup of the MD simulations on tetrameric open SoPIP2;1
We first tested whether steered MD  can reproduce the well-established features of H2O conduction in aquaporins from other species [29,32]. In accordance with previous reports, a stable water line formed in this ‘high-resolution open’ monomer structure during a MD simulation (Figure 5B). Furthermore, the dipole orientation of water molecules reversed at the NPA region, similar to aquaporins from other species [29,32]. To increase the probability that a specific water molecule enters a pore, it was positioned close to the external pore vestibule and a constant force was applied to steer the solute into the z-direction (see Figure 5A) . Since solutes might also pass through the tetramer centre, rather than the four channel pores , we positioned a fifth water on top of the tetramer center (Figure 5A and Supplementary Figure S3).
The trajectories from several independent simulations showed how H2O molecules traverse the pores (Figure 6A). We observed co-ordinated water conduction, similar to what has been described in earlier simulations on other aquaporins [29,32,34,35]. After 4 ns, seven out of 11 (64%) water molecules had crossed the pore entirely (Figure 6A). Individual water molecules showed two preferred residency positions close to the ar/R region and the NPA region (Figure 6A). Only one single water molecule that had entered the outer channel vestibule did not cross the entire pore during the full simulation time of 16 ns. This molecule moved away from the central pore axis and got stuck in a cavity close to Glu168 in transmembrane helix 4. Furthermore, the continuous water line was broken in some cases at the ar/R region, a feature that was also observed when pressure was applied to accelerate the transit of solutes . In particular cases, the labelled water molecules did not enter a pore. The trajectories of these waters were excluded from further analysis.
MD simulations of H2O conduction by SoPIP2;1
H2O2 conduction in PIP2;1
Similar simulations were performed with five H2O2 molecules replacing the waters at identical starting positions. H2O2 competed with water for hydrogen bonds, but it was evident from the trajectories that the entrance of H2O2, as well as the conductance, was slightly less favourable compared with water (Figure 6B). After 4 ns, only two full conduction events were observed (15%), but all H2O2 molecules finally crossed the pore during the 16 ns simulations. The major residency positions were similar to those of water, but further sites of increased residency close to the internal pore vestibule were detected (Figure 6B). A closer look at individual simulations showed that the residency at these positions resulted from the exact position of the flexible cytoplasmic loop D that is responsible for closing the pore. The maximal residency time for H2O2 in the ar/R region (indicating a major energy barrier for conduction) is in line with our finding that increasing the pore diameter in this region increased H2O2 conduction/sensitivity (Figure 3).
Comparison of solute conduction simulations in SoPIP2;1
Similar simulations were also performed with urea, a solute that was not conducted by AtPIP2;1, as indicated by our growth assays (Figure 3D). The trajectories for urea indicated that this solute was also not conducted in simulations. No full conduction event was observed within 16 ns. Only one single urea molecule had traversed most of a pore (Figure 6C). The trajectories identified the ar/R region as an insuperable barrier for urea conduction (Figures 6C and 6D). This result is strongly supported by the growth assays in Figure 3, which indicate that mutations in the ar/R region can convert PIP2;1 into a urea-conducting channel. Mutations in the ar/R region result in a wider pore diameter, which agrees with the conduction of urea in these mutants .
The energy profile of each substrate along the z-axis was calculated from the trajectories and further illustrates the differences in the conduction properties (Figure 6D). The energy barriers for H2O2 along the pore axis was at most positions slightly higher than for water, but the maximal barrier was of similar height. All three tested substrates have a dipole moment, which was preferentially oriented within the pore. This is well observed at the sites of longer residency, where snapshots of H2O2 and urea in their preferential positions and orientation are given in Figures 6(E) and 6(F).
Neither H2O2 nor urea permeated through the tetramer centre; the solutes did not enter this central cavity. In contrast, the solutes initially positioned above the tetramer centre nearly always entered and permeated a pore. For this reason, we also tested whether smaller simulations with monomers gave a similar outcome. Such smaller simulations are advantageous, since they saved 39% of resource time. The monomer was again embedded into a small patch of lipid bilayer (Supplementary Figure S4 at http://www.BiochemJ.org/bj/414/bj4140053add.htm). Interestingly, these simulations yielded essentially the same results (Supplementary Figure S5 at http://www.BiochemJ.org/bj/413/bj4140053add.htm). The trajectories of water, H2O2 and urea resembled those from tetramer simulations. This suggests that monomers are sufficiently rigid to computationally simulate conduction in isolated, monomeric pores.
The plasma membrane is an important protective barrier against extracellular H2O2. Yeast strains with impaired ergosterol biosynthesis and, thus, higher membrane permeability to lipophilic substances, were more sensitive to H2O2 . At low millimolar concentrations, H2O2 induces apoptosis in yeast [27,39]. Gradients of H2O2 have been reported across the membranes from mammalian cell lines , bacteria  and yeast , suggesting that native biological membranes generally prevent the free diffusion of H2O2. Very recently, TIPs from plants have been shown to specifically conduct H2O2 by different assays, including growth sensitivity on H2O2-containing plates . However, the physiological role of H2O2 permeability across the tonoplast remains somewhat unclear. In the present study we confirmed H2O2 conduction by a TIP (Figure 4). TIPs have a relatively weak selectivity and conduct many solutes, including urea, ammonium and water. Our analysis indicates that not only intracellular aquaporins, such as TIPs, but also plasma membrane PIP2-type aquaporins, which were previously believed to be highly selective channels for water, conduct H2O2.
The Saccharomyces genome encodes four aquaporin genes, two of which are highly homologous with specific water channels. Many laboratory yeast strains, however, contain function-abolishing mutations in these aquaporins . Therefore two different laboratory strains were tested for their H2O2 growth sensitivity, but only minor differences were observed and all were unable to grow at ≥3 mM H2O2 (results not shown). In both strains, the sensitivity to H2O2 increased several-fold with the heterologous expression of AtPIP2;1 aquaporins.
The permeability of the plasma membrane to non-ionic substances is generally determined by the membrane fluidity that depends on the content of, e.g. sterols . Higher plant plasma membranes are known for their high sterol content, which suggests that native plant membranes are relatively impermeable to non-electrolytes . In the simulations we used a simple POPE-lipid membrane without sterols, which does not reflect the complex properties of a native plant membrane. This simple membrane is expected to confer only a relatively weak diffusion barrier for solutes. In agreement with this, we occasionally observed full conduction events in simulations with water (three full events from all simulations) and H2O2 (two full events) across the synthetic lipid membrane.
In the present study we identified a novel substrate of channels by a combination of functional assays and computer simulations. The combined data suggest that PIP2-type plasma membrane aquaporins are efficient H2O2 channels (Figure 6). The aquaporin-mediated conduction of H2O2 may explain previous observations that algal membranes exhibit permeability to H2O2 at high concentrations . The molecular analysis shown in the present study suggests that the ar/R region is critically involved in selectivity towards urea, boron and H2O2. Widening the pore allows larger solutes to pass through the channel (Figures 3B and 3D), which is in agreement with previous reports [28,37]. The simulations and functional data suggest that the ar/R selectivity filter region is the most critical determinant of H2O2 conduction in aquaporins. Since all plasma membrane PIP2 aquaporins from plants have identical residues at these positions, it appears reasonable to predict that all eight PIP2-type aquaporins from Arabidopsis conduct H2O2. The yeast growth tests further imply that some aquaporins, with a different selectivity filter region, also conduct this solute (Figures 3 and 4).
Interestingly, resource-saving simulations on monomeric SoPIP2;1 yielded an overall result similar to full simulations, although the transit time along the channel seems to be reduced for monomeric simulations. This may be exploited in the future to reduce calculation times and costs. Indeed, simulations on monomers have already been used to probe the conduction properties of aquaporins . Computer simulations on channels also appear to be valuable in predicting novel substrates; this may be used in the future to predict physiologically important solutes that cannot be easily identified or measured by functional assays.
Although the pores were readily filled with a line of waters, the ar/R region was only partially occupied in several independent simulations, and the flexible side chain of the arginine (Arg224) bent briefly into the pore. The flexibility of this arginine residue had been observed in earlier simulations on other aquaporins [34,35], and was therefore fixed in some simulations . However, the flexibility appears to reflect free arginine rotamers, since this arginine residue also adopted different conformations within subunits of the AqpZ tetramer . No restraints were applied to any of the side chains during our simulations and the arginine residue behaved similarly in simulations with water or H2O2.
When expressed in yeast, a high permeability of aquaporins to H2O2 results in susceptibility to H2O2, indicating the putative physiological importance of the H2O2 conduction by aquaporins. It is known that developmental processes, such as root hair growth and stomatal movements, require NADPH oxidase activity and probably subsequent H2O2 influx to activate Ca2+ channels .
Whether H2O2 conduction across plasma membranes is altered in Arabidopsis mutants that lack individual aquaporins remains an unanswered question. However, the large number of plasma membrane aquaporins and their partially overlapping expression pattern and redundant function may require plants in which multiple PIP2s are knocked-out . Arabidopsis mutants with strongly reduced water conduction are not yet available . In line with this, we did not find H2O2-related phenotypes with plant lines harbouring single T-DNA (transfer DNA) insertions in PIP2 genes in preliminary experiments (results not shown).
It appears likely that plasma membrane aquaporins may be far more than just mediators of water conduction. Recent observations suggest that other small solutes, such as CO2, are also channeled by at least some aquaporins [1,2]. That excess H2O2 influx kills yeast is supported by the growth tests (Figure 1) and mutant yeast strains . Aquaporins belong to a multi-gene family in plants and it remains a future challenge to identify the physiological contribution of each individual member to H2O2 conduction.
The ability of pathogens to resist oxidative stress is crucial for their infectiousness and pathogenesis in the host. Aquaporins have also been identified in microbes and pathogens. Pathogens might increase virulence by minimizing the influx of the H2O2 generated by the plant. Protection against influx of H2O2 may be mediated by pore closure, e.g. by de-phosphorylation, transcriptional down-regulation during infection or even genomic loss of aquaporins.
We thank Junpei Takano for plasmids and discussions and Vytas Bankaitis for providing reagents. This work was supported in part by a grant of the Landesstiftung Baden Württemberg to U.L. Simulations were carried out at the HPC centre of the University of Tübingen, the Höchstleistungsrechenzentrum of the University of Stuttgart (HLRS) and the Scientific Supercomputing Center (SSC) Karlsruhe, Germany.
NOD26-like intrinsic protein
asparagine, proline, alanine
periodic boundary condition
plasma membrane intrinsic protein
Particle Mesh Ewald
reactive oxygen species
tonoplast intrinsic protein
These authors contributed equally to the present study.
Present address: Department of Cell and Developmental Biology, University of North Carolina, School of Medicine, Chapel Hill, NC 27599-7090, U.S.A.
Present address: Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Mail Stop 977-152, Berkeley, CA 94720-8205, U.S.A.