Identification of a Residue in Helix 2 of Rice Plasma Membrane Intrinsic Proteins That Influences Water Permeability*

Molecular selection, ion exclusion, and water permeation are well known regulatory mechanisms in aquaporin. Water permeability was found to be diverse in different subgroups of plasma membrane intrinsic proteins (PIPs), even though the residues surrounding the water holes remained the same across the subgroups. Upon homology modeling and structural comparison, a conserved Ala/Ile(Val) residue difference was identified in helix 2 that affected the conformation of the NPA region and consequently influenced the water permeability. The residue difference was found to be conservative within the two subgroups of PIPs in rice as well as in other plants. Functional tests further confirmed the prediction via site-directed mutagenesis where replacement of Ala103 or Ala102 in respective OsPIP1;1 or OsPIP1;3 with Val yielded 7.0- and 2.2-fold increases in water transportation, and substitution of Ile98 or Val95 in respective OsPIP2;3 or OsPIP2;7 with Ala resulted in 73 or 52% reduction of water transportation. Based on structural analyses and molecular dynamics simulations, we proposed that the difference in water permeability was attributed to the orientation variations of helix 2 that modified water-water and water-protein interactions.

Aquaporins (AQPs) 7 are small integral membrane proteins that facilitate water transport across the membranes and are widely distributed in animals, plants, and microbes. It is unique in plants that AQPs form a large family with 35 and 33 members in Arabidopsis and rice, respectively. According to the amino acid sequence, the AQPs could be classified into four subgroups, i.e. the plasma membrane intrinsic proteins (PIPs), the tonoplast intrinsic proteins, the NOD26-like intrinsic protein, and the small basic intrinsic proteins (reviewed in Maurel et al. (1) and references cited therein). Plant PIPs are divided into two phylogenic subgroups, PIP1 and PIP2 (2). The AQP monomer consists of six transmembrane ␣-helices tilted along the plane of the membrane and connected by five loops (A-E). Loop B and D as well as the N and C termini are cytoplasmic (3). Two highly conserved constrictions within the pore of AQPs were proposed by structural analysis of AQP proteins. One is the central constriction formed by two Asn-Pro-Ala (NPA) motifs located on two short ␣-helices and the other is the outer constriction or aromatic/arginine (ar/R) constriction formed by spatial arrangement of four aromatic amino acids. Recently it was found that in addition to facilitating the transport of water, AQPs also can transport other small solutes. Size exclusion at the two main constrictions is one of the mechanisms for water transport and substrate selectivity by AQPs (3,4). Extensive investigation of the plant PIPs revealed that PIP2 paralogues induced water permeation when expressing on Xenopus oocyte or yeast membranes, whereas PIP1 paralogues were almost inactive (5)(6)(7)(8)(9)(10). The sequence difference between the two subgroups, however, is quite subtle, as they share the same residues in main pore constrictions and almost the same hydrophobic residues around the aqueous pathway (6,11).
The first attempt to elucidate the mechanism of distinct water transport activities between the two subgroups was performed in radish PIPs (12). Based on the sequence difference, the authors identified a residue that discriminates PIP1s and PIP2s. They further showed that substitution of Ile 244 (at the pore entrance) in RsPIP1-3 with Val enhanced the activity to 250% of WT. Replacement of Val 235 with Ile (corresponding to Ile 244 in RsPIP1-3) resulted in a remarkable activity reduction to 45% of WT RsPIP2-2. This work provided evidence that punctual structural differences between the two PIP subgroups can to some extent explain their distinct water transport activity.
Crystal structures of mammalian AQP1 (13,14) and AQP0 (15)(16)(17), bacterial GlpF (4,18) and AqpZ (19), archaeal AqpM (20), and plant SoPIP2 (21) have provided important informa-tion related to water permeability (P f ). Combined with structure-function analyses and molecular dynamics (MD) simulations, these structures have provided detailed pictures for water transportation and gating mechanisms of AQPs. Specific features of pore radius, residue side chain orientation/ fluctuation, and energy barrier were proposed to affect water transport. For example, a narrow constriction at the NPA region was found to be essential for AQP1 water selectivity and permeability (22). Point mutations in the ar/R region changed the substrate specificity of the AQP1 and allowed the mutant protein to permeate urea, glycerol, ammonia, and proton (23). AQP0 has a specific constriction near the cytoplasmic region in addition to those at the ar/R and NPA regions (15)(16)(17). The Tyr 23 residue at the NPA region and the Tyr 149 residue at the cytoplasmic constriction were shown to play important roles in restraining water permeation through AQP0 (24). The differences of the key residues at three constriction regions resulted in lower permeability of AQP0 than AQP1 (25). In addition, the side chain dynamics were found to be a prerequisite for water permeation in AQP1 (26). In fact, water permeation via AQPs appears to be a dynamic and statistical process, i.e. the instantaneous pore radius or energy barriers do not directly restrict bulk water motion. This may account for the apparent discrepancy between channel radius and P f in typical AQPs (27). Osmotic permeability matrix analysis was proposed as a way to explain P f with respect to structural characteristics, where the water channel was decomposed into small local regions, and the channel permeability contributed to those from the regions themselves (diagonal elements of P f matrix) and those from the cross-correlation of different regions (offdiagonal elements of P f matrix) (27). Comparisons of the P f matrix of five aquaporins (AQP0, AQP1, AQP4, AqpZ, and GlpF) at the three constriction regions indicated that the NPA region hinders the collective motion of the water chain. The weakest correlation across the NPA region is observed in the P f correlation matrix of AQP0, which indicated that water motion at one end of the channel does not propagate to the other end (27,28). These reports prompt us to unravel the details of interactions of channel waterwater and water-protein in the two plant PIP subclasses, because the former promotes water permeability and the latter has a contrary effect.
Here, we integrated structural comparison, functional measurements, and MD simulations to address this issue. We show that an Ala/Ile(Val) residue at the NPA region dominates the permeability difference between OsPIP1s and OsPIP2s. Our results indicate that this residue contributes to the different water transportation activities of the two subgroups via modulating the molecular conformation around both the NPA and the ar/R regions.

EXPERIMENTAL PROCEDURES
Homology Modeling of OsPIPs and Sequence Alignment of Plant PIPs-Target structure of each WT or mutated OsPIP protein was generated by GeneAltas software (Accelrys Inc.) (29) using the default setting. For each PIP protein, 2-3 structure models were generated by homology modeling, and only the one listed at rank 1 with the highest score was selected ( Fig. 1) and used for further analysis. The algorithm PB90, the sequence profile-based searching protocol utilizing optimized PSI-BLAST, and the algorithm THM, the high throughput modeling protocol with subsequent verification with Profiles-3D and PMF Verify, were used by the software to select reference template from the template data base and to make the structure. The x-ray crystallography of bovine AQP1 (Protein Data Bank code 1J4N) was automatically selected as the best reference template (13) from three candidates (RCSB template 1J4N, 1H6I, and 1DLF). All the structures were visualized in VMD (30), and the channel radius of each protein was calculated using HOLE software (31). All PIPs were aligned using MUSCLE software (32), and the detailed sequence data source are listed in supplemental material.
Osmotic Permeation Assay-cRNA of the WT and mutant OsPIPs were produced by using mMESSAGE mMACHINE capped RNA transcription T3 kit (Ambion, Austin, TX). The cRNA was quantified by UV spectrum and stored as 1 g/l at Ϫ80°C. Xenopus laevis oocytes were prepared and injected with cRNAs as described previously (33). Oocytes were injected with 23.5 nl of sterile water (control) or an equal volume of cRNA solution. Oocyte swelling was measured directly after transferring from 200 to 40 mosM ND96 solution as described previously (34). The section area (S) of oocytes was calculated using LabWorks3.0 (UVP, United Kingdom). Water permeability per cell was calculated by P f ϭ V 0 (d(V/ V 0 )/dt)/(S 0 ϫ V w ϫ (OsM in Ϫ OsM out )) where the initial oocyte volume, V 0 , is 9 ϫ 10 Ϫ4 cm 3 ; the initial oocyte area, S 0 , is 0.045 cm 2 ; and the molar volume of water, V w , is 18 cm 3 /mol (35). At least 20 oocytes were tested for each protein. The statistical significance of the difference in P f between WT and mutated OsPIPs was assessed using Student's t test.
Membrane Expression of WT and Mutated OsPIPs in Xenopus Oocytes-For GFP-fused proteins, the cDNA fragments of WT and mutant OsPIPs without a stop codon were inserted between the EcoRI and XbaI site of expression vector pGEMHE pre-inserted with a GFP sequence. The cRNA of each GFP-fused protein was synthesized in vitro by mMESSAGE mMACHINE capped RNA transcription T7 kit (Ambion).
The fluorescence signal was visualized by laser confocal microscopy (Leica, Germany). Five oocytes were tested for each protein where cross-membrane fluorescence intensity for each cell was measured from eight positions along one-quarter of the membrane. The mean fluorescence intensity was then obtained by averaging the intensities from five cells in each case. The Student's t test was used to analyze statistical significance of the fluorescence intensity difference between WTs and correspondent mutants.
Molecular Dynamics Simulations of OsPIP1;1 and OsPIP2;7-All simulations for OsPIP1;1, OsPIP1;1 A103V, OsPIP2;7, and OsPIP2;7 V95A were carried out using the respective monomeric homology model generated using Protein Data Bank code 1J4N as template. Each system was composed of a channel tetramer, 1-palmitoyl-2-oleoyl-sn-glycero-3phosphatidylethanolamine lipid bilayer, a 25-Å thick slab of water neutralized by 100 mM chloride and sodium ions. All simulations were performed using NAMD2 (36) and CHARMM27 force field (37) in periodic boundary conditions with time steps of 1 fs. A smooth (10 -12 Å) cutoff and the Particle Mesh Ewald method were used to calculate van der Waals and electrostatic interactions. All simulations were done at constant temperature (300 K) and pressure (1 atm) and analyzed using the VMD program.
A collective diffusion model, proposed to characterize the osmotic permeability of a channel upon equilibrium MD simulations (38), was used here to evaluate water permeation of simulated OsPIP1;1, OsPIP2;7, and their respective mutants. Briefly, in Equation 1 the osmotic permeability P f of the channel gives where v w is the average volume of a single water molecule, and D n is defined as the diffusion coefficient of channel waters at equilibrium. By taking advantage of MD simulations that the movement of every water molecule is able to be monitored, a collective coordinate, n, is defined to quantify channel water translocation as shown in Equation 2, where L is the channel length along Z direction; s(t) denotes the set of channel water molecules at time t, and dZ i is the displacement of ith water molecule along Z direction at the interval dt. At equilibrium, the net amount of water permeation through the channel is zero on average, i.e. ͗n(t)͘ ϭ 0. Mean square displacement of n, ͗n 2 (t)͘, obeys the Einstein relation when t is much longer than the velocity correlation time of n as shown in Equation 3, Thus, one can determine P f from Equation 1 using D n value calculated from the collective coordinate n (Equations 2 and 3).

Identifying Important Residues for P f Difference in Two
OsPIP Subgroups-We generated three-dimensional structures of all OsPIPs using GeneAtlas and analyzed their pore radii by HOLE. Each OsPIP presented a narrow radius at the ar/R constriction region ( Fig. 1, red lines). OsPIP1;1 and OsPIP1;3, but not OsPIP1;2, had an extra constriction at the NPA region with narrow pore radius. All the OsPIP2s, except OsPIP2;1 and OsPIP2;8, did not have such extra constriction at the NPA region ( Fig. 1, dark green lines). Structural comparison of OsPIP1;1, OsPIP1;3, OsPIP2;3, and OsPIP2;7 at the NPA and ar/R constriction regions indicated that the orientation change of an Ile side chain at the NPA region might induce the difference in pore radius between the two subgroups because the hydrophobic Ile 101/100 side chains in OsPIP1;1/1;3 point to the pore, and the Ile 96/93 side chains in OsPIP2;3/2;7 orient away from the pore ( Fig. 2A, green licorice). As a conservative pore-forming residue, the orientation change of Ile side chain at the NPA region could possibly induce the low (in OsPIP1) or high (in OsPIP2) water permeation efficiency across the membrane. To identify the key residues responsible for the orientation change, 15 neighboring residues around the Ile site were analyzed within the maximum distance (6 Å) of van der Waals force (Fig. 2B). Comparisons of the neighboring residues around the Ile 101 of OsPIP1;1 and Ile 93 of OsPIP2;7 revealed the discriminating residues between PIP1s and PIP2s that is Ala 103 in OsPIP1;1 and Val 95 in OsPIP2;7. The Ala residue presented in three PIP1s and the Ile residue in all the OsPIP2s except OsPIP2;7 and OsPIP2;8, which had a Val residue at the corresponding site (Fig. 2B). Therefore, the difference at the Ala/Ile(Val) site was conservative in OsPIP1s and OsPIP2s. This finding suggested that the conserved site might be a potential candidate to induce the diverse orientation of the Ile residue between the two OsPIPs, which in turn affected the P f .
Structural Analyses of Ala/Ile(Val)-mutated OsPIPs-To test the hypothesis, the impact of the Ala/Ile(Val) site on side chain orientation of the Ile residue was analyzed by comparing the pore radii and structures of OsPIPs with their corresponding interchanged mutants. The virtual structures of four mutated OsPIPs, OsPIP1;1 A103V, OsPIP1;3 A102V, OsPIP2;3 I98A, and OsPIP2;7 V95A, were generated, and the corresponding pore radii were calculated. A nascent narrow NPA region was formed inside the pore of OsPIP2;3 I98A and OsPIP2;7 V95A mutant proteins (Fig. 3A, dark green lines), which was in accordance with the orientation change of Ile 96 side chain in OsPIP2;3 I98A or Ile 93 side chain in OsPIP2;7 V95A (Fig. 3B, green licorice), respectively. In addition, the pore radius at the ar/R region of OsPIP2;7 V95A was also getting bigger. By contrast, this constriction region became wider in OsPIP1;1 A103V, but no radius change was observed in OsPIP1;3 A102V (Fig. 3A, dark green lines), which was in agreement with the minor orientation change of Ile 101 side chain in OsPIP1;1 A103V and the invisible change of the Ile 100 side chain in OsPIP1;3 A102V (Fig. 3B, green licorice), respectively. In addition, an orientation change of the long side chain of the Phe 102 residue was also found in OsPIP1;1 A103V (Figs. 2A and 3B, blue licorice).
Functional Confirmation on Site-directed Mutagenesis-To test the functionality of the Ala/Ile(Val) site on P f , two sets of measurements were performed in water transport activity and membrane localization using site-directed mutagenized OsPIPs.
Membrane Localization-To further test if the site-directed mutation affected the membrane localization of OsPIP proteins, GFP was fused at the C-terminal end of WT OsPIP1;1, OsPIP1;3, OsPIP2;3, and OsPIP2;7 and of their corresponding mutants, respectively. The expression and localization of the proteins in Xenopus oocytes membrane were visualized using laser confocal microscopy (Fig. 4E). AQP2 was used as a marker protein to show the membrane localization (Fig. 4F). The fluorescence intensity was measured, and Student's t test analysis showed that no significant difference in fluorescence intensity was found between WTs and mutated OsPIPs with one exception for OsPIP1;3A/V that yielded 50% lower intensity than that of OsPIP1;3 (Fig. 4G), indicating that the membrane localization of OsPIPs was not altered by site-directed mutagenesis and that the membrane targeting efficiency was declined in OsPIP1;3A/V. Even with relatively low membrane expression, high permeability was also found for OsPIP17; 3A/V. These results further supported the hypothesis that the Ala/Ile(Val) difference between the four WT OsPIPs and their respective mutants OsPIPs, rather than interfering with membrane localization, induced a real change in intrinsic water transport activity.   showing the localization of AQP2 on the plasma membrane (F). G, estimation of expression levels of OsPIP constructs in injected oocytes. One-quarter of an oocyte was imaged by laser confocal microscopy and digitized using ImageJ software. Eight transmembrane boxes in a size of 100 ϫ 1 pixels were selected along the membrane, and the gray values inside the box were measured and averaged to be the relative fluorescence intensity (RFI) of protein expression for that cell. The relative fluorescence intensity of the each WT and correspondent mutant PIP was statistically analyzed by Student's t test. Significant value relative to the WT is indicated: **, p Ͻ 0.01. DECEMBER 31, 2010 • VOLUME 285 • NUMBER 53

Dynamic Structural Bases for Ala/Ile(Val) Site-related
Functionality-MD simulations were performed to understand the dynamic structural bases of Ala/Ile(Val) functionality using WT and mutated OsPIP1;1 and OsPIP2;7.
Correlation of Water Permeability with Water-Water and Water-Protein Interactions-Monomeric P f for WT OsPIPs and their mutants was estimated using the collective diffusion model described previously (38), where the diffusion coefficient of channel water was calculated in a time interval of 200 ps for the channel region of Ϫ6 Յ Z Յ12 Å with Z ϭ 0 at Ile 101 -CD atom (the fourth carbon atom of Ile residue that is consistently named with CHARMM force field, similar tokens were used in the following) for OsPIP1;1 or of Ϫ6 Յ Z Յ15 Å with Z ϭ 0 at Ile 93 -CD atom for OsPIP2;7. These analyses yielded P f values (in 10 Ϫ14 cm 3 /s) of 2.18 Ϯ 0.41 and 2.72 Ϯ 0.52 for OsPIP1;1 and OsPIP1;1 A103V, respectively, and of 3.53 Ϯ 0.58 and 2.91 Ϯ 1.02 for OsPIP2;7 and OsPIP2;7 V95A, respectively. These results were in agreement with the above functional measurements that OsPIP1;1 A103V enhanced but OsPIP2;7 V95A reduced the P f as compared with their respective WTs.
Water-water interactions promote water permeation, whereas water-protein interactions have the opposite effect. As exemplified in Fig. 5, A and B, stronger hydrogen bonding (H-bonding) in water-water interactions and weaker H-bonding in water-protein interactions were exhibited by OsPIP1;1 A103V (red lines), as compared with its WT (black lines). This further supported the observation that, as compared with OsPIP1;1, P f was enhanced in OsPIP1;1 A103V. By contrast, OsPIP2;7 V95A exhibited, by comparison with WT OsPIP2;7, reduced water-water interactions and fostered water-protein interactions, in agreement with the reduction in P f induced by the point mutation (Fig. 5, C and D, red lines). These results provided a global interpretation of structural bases for mutagenesis-induced changes in P f .
Orientation and Stability of Side Chain of Key Residues-Differences in H-bond interactions appeared to localize at a narrow region between the 0 to ϩ10 Å coordinates (Fig. 5), especially at two positions corresponding to the NPA (around ϳ0 Å) and ar/R regions (around ϳ10 Å) (Fig. 5, gray stripes). Residue orientation analyses for the two regions indicated that two key residues contributed to the differences. One was Ile 101/93 in OsPIP1;1/2;7 that pointed to the channel pore at the NPA region ( Fig. 2A, green licorice) and the other was Phe 97/89 in OsPIP1;1/2;7 that constituted the ar/R region ( Fig.  2A, purple licorice). The angle between Ile 101/93 vector (from atom Ile 101/93 -CD to atom Ile 101/93 -CG2 or CD 3 CG2 in short) or Phe 97/89 vector (from Phe 97/89 -CD2 to Phe 97/89 -CD1 or CD2 3 CD1 in short) and the z-axis vector (parallel to water channel and pointing from the cytoplasmic to extracellular domain) was calculated to evaluate the orientation of side chain of the two key residues. It was found that Ile 101 orientation was stable at ϳ127°for OsPIP1;1 A103V in three monomers (supplemental Fig. S2B) as compared with its unstable orientation for OsPIP1;1 with varied angles (supplemental Fig. S2A). In contrast Ile 93 orientation in OsPIP2;7 fluctuated up to ϳ68°(supplemental Fig. S2C) as compared with its stable orientation at ϳ127°for OsPIP2;7 V95A (supplemental Fig. S2D). Such orientation fluctuation affected P f because both Ile 101 and Ile 93 are pore-forming residues. A typical evolution analysis (Fig. 6A) indicated that the side chain of Ile 101 residue in OsPIP1;1 pointed to the channel pore at 1.8 ns (ϳ68°) to interrupt the continuity of the water molecule file, moved away at 3.0 ns (ϳ127°) to favor water file formation, and confined its orientation at 4.2 ns (ϳ150°) to induce an intermediate opening (Fig. 6A, insets). This was in agreement with different orientations of Ile side chain at different pore radii found in homology modeling (Fig. 2). This orientation instability hindered water molecule movements in the vicinity of Ile 101 residue and resulted in weaker water-water and stronger water-protein interactions.
Another region presenting different H-bond interactions around ϳ10 Å corresponded to the ar/R region (Fig. 5, gray  stripes). The side chain of Phe 97/89 residue in OsPIP1;1/2;7, one of the four residues constituting the ar/R region, oriented more stably at ϳ38°for OsPIP1;1 or at ϳ46 -88°for OsPIP2;7 than that at ϳ38 -120°for OsPIP1;1 A103V or at ϳ7-141°for OsPIP2;7 V95A (supplemental Fig. S3). The distinct nature of orientation and stability between the two OsPIPs affected the conformation of the ar/R region and resulted in a P f difference. Typically, the Phe 97 side chain in OsPIP1;1 A103V pointed to the channel pore (ϳ38°) followed by an open orientation (ϳ62°and even ϳ80°), and finally moved away from the ar/R region (insets) (Fig. 6B and supplemental Fig. S3, A  and B). This reorientation of hydrophobic Phe 97 residue favored the passage of water molecules and resulted in stronger water-water and weaker water-protein interactions, as compared with those in OsPIP1;1. High orientation fluctuation of Phe 89 side chain in OsPIP2;7 V95A indicated that the instability interrupted the continuity of the water file and resulted in weaker water-water and stronger water-protein interactions, FIGURE 5. Differences in channel water-water and water-protein interactions. Channel water-water and water-protein interactions between WT (black) and mutant (red) OsPIPs are compared. Average number of H-bonds between a water molecule with its oxygen atom located along the z axis and its adjacent water molecules (A and C) and between the molecule and channel protein (B and D) are illustrated for OsPIP1;1 (A and B) and OsPIP2;7 (C and D). A hydrogen bond was defined as having donor-acceptor distance of Ͻ3.5 Å and a donor-hydrogen-acceptor angle of Ͻ45°. Data were collected from all channel water molecules during the entire equilibration process for every monomer and presented as group average with a bin size of 0.2 Å. The reference origin was denoted as the Z coordinate of the Ile 101 -CD atom of OsPIP1;1 or Ile 93 -CD atom of OsPIP2;7. The NPA motif and ar/R region are highlighted in semitransparent gray stripes. although no significant difference in orientation angles was found (supplemental Fig. S3, C and D). It was also observed that the Ile 93 orientation in OsPIP2;7 (supplemental Fig. S2A) as compared with that in OsPIP2;7 V95A (supplemental Fig.  S2B) seemed not to be favorable for water permeation. In particular, slightly weaker water-water and stronger water-protein interactions were found around in the NPA region in OsPIP2;7 as compared with OsPIP2;7 V95A (Fig. 5, C and D). In fact, orientation stability of another residue (Phe 89 ) was involved, and in combination with Ile 93 orientation, it could account for the fact that the P f was reduced in OsPIP2;7 V95A. Taken together, the analyses of side chain orientation and stability confirmed the above observations that P f was enhanced for OsPIP1;1 A103V but reduced for OsPIP2;7 V95A.
Orientation of Helix 2-The two key residues discussed above, Ile 101 and Phe 97 of OsPIP1;1 and Ile 93 and Phe 89 of OsPIP2;7, are both located in the same helix 2 (Gly 92 -Ile 111 for OsPIP1;1 or Gly 84 -Val 103 for OsPIP2;7). They are, however, separated by several residues from the mutated site (Ala 103 in OsPIP1;1 and Val 95 in OsPIP2;7) ( Figs. 2A and 3B). One possible mechanism for orientation changes induced by punctual mutations at a distant site is that the two mutations (A103V or V95A) triggered a re-orientation of helix 2 thereby driving a distant reorientation of Ile 101/93 and Phe 97/89 side chains. To test this, the angle between helix 2 vector (Leu 104 -CA 3 Ala 94 -CA for OsPIP1;1 or Leu 96 -CA 3 Ala 86 -CA for OsPIP2;7) and the Z-axis vector was calculated. The mean angle was enhanced for OsPIP1;1 A103V (36.8 Ϯ 0.8°, red line) as compared with that for OsPIP1;1 (34.8 Ϯ 0.8°, black line) (Fig. 7A) but reduced for OsPIP2;7 V95A (37.7 Ϯ 1.0°, red line) when compared with that of OsPIP2;7 (39.9 Ϯ 0.8°, black line) (Fig. 7B). This indicated that the orientation change of helix 2 (Fig. 7, C and D) was positively correlated with those of Ile 101/93 and Phe 97/89 residues for both WT and mutated OsPIPs (supplemental Figs. S2 and S3). Specifically, the enhancement or reduction of the angle was induced by reducing the motion freedom of the Phe 102 residue in Os-PIP1;1 A103V or increasing the freedom of Phe 94 in OsPIP2;7 V95A (supplemental Fig. S4). Thus, the analysis confirmed the prediction that the orientation of helix 2 played a role to transfer the conformational change from the mutated site to the constraint regions.
In summary, we proposed a model to interpret how sitedirected Ala-Val mutation regulated the P f of OsPIPs. Typically, OsPIP1;1 A103V mutation drove a helix 2 orientation change (Fig. 6D, yellow for WT and red for A103V), which induced stable and favorable orientations of Ile 101 residue at the NPA region and of Phe 97 residue at the ar/R region (Fig. 6,  C and D). Such orientation change opened the channel pore, widened the radius profile, and combined with a well formed water molecule file, enhanced the P f . By contrast, the converse effect for OsPIP2;7 V95A mutation was demonstrated by opposite pathways (supplemental Fig. S5). . Also plotted are typical snapshots at specific moments (insets in cyan CPK or sticks). Background channels are presented in blue as control. C and D, typical monomer snapshot was illustrated for OsPIP1;1 alone (C) and for both OsPIP1;1 and OsPIP1;1 A103V at 4.86 ns (D). Channel structure was partially represented as a ribbon with the helix 2 (yellow), two NPA motifs with the corresponding half-helices, and loops (cyan), and channel radius profile. Two Asn from the NPA motif, an Arg from the ar/R region, and the residues (Ala 103 , Phe 102 , Ile 101 , and Phe 97 ) pertinent to the proposed model are illustrated as licorice in the same color as in Figs. 2A and 3B. Also plotted in D are the orientation changes in helix 2 (red ribbon) and the corresponding residues (CPK presentation) with the channel radius profile for OsPIP1;1 A103V.

DISCUSSION
AQPs are crucial for the transport of water and/or other small solutes across cell membranes, and their functionality is defined by specific gating, selectivity, and permeability properties. Several mechanisms have been proposed to determine the activity of AQPs in plant cells via co-and post-translational modification, gating, trafficking, as well as the combinatorial regulations (within two different types of AQPs) (1). Plant AQPs can be gated by protons, divalent cations, and phosphorylation, and several recent studies have provided novel insights into the molecular mechanisms involved. Structure-function analyses in plant and animal AQPs have also shown that their distinct molecular selectivity and ion exclusion are determined in large part by the ar/R and NPA constrictions. For rat AQP1, replacement of the ar/R components Phe 56 and His 180 together may enlarge the maximal diameter of the ar/R constriction by 3-fold and enable glycerol and urea to pass (23). The goal of this study was to identify the key residue(s) for the P f difference between OsPIP1s and OsPIP2s and to understand their corresponding structural bases at a single amino acid level.
In this study, selection of candidate residue(s) was achieved by identifying the conserved residue(s) from three-dimensional structural comparison between OsPIP1s and OsPIP2s (Fig. 2). The impact of the selected Ala/Ile(Val) residue on P f of OsPIPs was then tested by structural differences and distinct pore radii using the reconstructed three-dimensional structure via virtual homology modeling, which was further validated by interchanging the residue between OsPIP1s and OsPIP2s (Fig. 3). PIP channel analysis allowed identification of structure difference for PIP1s and PIP2s, i.e. the extra constriction at the NPA region of PIP1s. Exceptions were found for OsPIP1;2, which exhibits PIP2 pore feature, and OsPIP2;1, OsPIP2;8 showed PIP1 pore feature (Fig. 1). Because the P f of the OsPIP1;2 and OsPIP2;1 was consistent with their own subgroup (9), the Ala/Ile(Val) residue might be one of the important residues that influence the pore structure of the PIPs and hence the water permeation. We cannot rule out that other structure characters influence water permeation besides the major constriction at the NPA and ar/R regions, which was consistent with the result that OsPIP2;3 I98A and Os-PIP2;7 V95A mutant proteins were not impermeable to water as OsPIP1s (Fig. 4, C and D). Specific pore feature, e.g. the cytoplasmic constriction in AQP0 (24,25), and the key residues, e.g. the Ile or Val in RsPIPs located on loop E close to the extracellular space, may influence the water permeation through AQPs (12). On the other hand, homology modeling of the specific proteins need to be refined manually to fit the biological character. In addition, Ala/Ile(Val) site difference showed extensive conservation not only in rice PIPs but also FIGURE 7. Helix 2 orientation difference between WT and mutant. A and B, tetramer average of helix 2 orientation evolved during the equilibration process of OsPIP1;1 (A) and OsPIP2;7 (B). The helix 2 orientation was denoted by the angle between helix 2 vector (Leu 104 -CA 3 Ala 94 -CA for OsPIP1;1 or Leu 96 -CA 3 Ala 86 -CA for OsPIP2;7) and z-axis vector. C and D, referred visualization of conformational comparison of helix 2 was illustrated by structurally superimposing of the wild type (blue) and mutant (orange) for OsPIP1;1 (C) and OsPIP2;7 (D). Full channel structure was shown with blue transparent presentation for demonstrating channel pores.
in PIPs from other plant species (supplemental Fig. S1B), suggesting that it is physiologically important for the large numbered, diverse AQP family in plants.
The existence of a large number and divergent AQPs in plant genomes might be a result of adaptation to the different water and ionic environments from the sea during the landing process in plant evolution. The Ala/Ile(Val) residue difference was found to be conservative in all higher plant PIPs (supplemental Fig. S1B) as well as in PIPs from Physcomitrella patens (supplemental Fig. S1C), which is Ala in PpPIP1s and Val in PpPIP2s with one exception of Thr in PpPIP1;3. P. patens was diverged from the lineage leading to higher plants ϳ443-490 million years ago just before the evolution of vascular plants (10,39). It can be imagined that the development of a subgroup of PIPs with low water permeability (i.e. PIP1s) might be responsible for fine-tuning the cellular water transport and protecting the terrestrial plants from water loss. The functionality of Ala/Ile(Val) residue in regulating P f was then confirmed using an osmotic permeation assay, and it was concluded that the interchanging of Ala/Ile(Val) residue between OsPIP1 and OsPIP2 reversed the P f (Fig. 4, A-D). An alternative possibility, however, was that the various AQPs under investigation differed in their ability to traffic to the oocyte plasma membrane. Such a mechanism was determined to explain the apparent lack of activity of maize PIP1 isoforms after expression in Xenopus oocytes (40). Here, our results excluded the possibility by indicating that the OsPIP molecules were functionally located on the cell membrane for both WT and mutated OsPIPs and that the membrane localization of the protein was not significantly altered by site-directed mutagenesis (Fig. 4, E-G). In agreement with the previous report, OsPIP1;2 was found to be expressed and trafficked to the yeast membrane with similar efficiency as OsPIP2s, but the water transport activity of OsPIP1;2 was smaller than OsPIP2s (9). Our data further affirmed that OsPIP1s are less permeable to water than OsPIP2s partly due to the Ala/ Ile(Val) difference.
MD simulations are advantageous in elucidating the dynamics of AQP activities in an accessible time scale of water transport across the channel (ϳ1 ns) and in unraveling the evolution of conformational stability and the water-water and water-protein interactions. In this study, a model was proposed from MD simulations that the Ala/Ile(Val) mutation induced the orientation change of helix 2 followed by conformational change of Ile 101/93 residue at the NPA region and of Phe 97/89 residue at the ar/R region. This, in turn, enhanced water-water interactions and reduced water-protein interactions in OsPIP1;1 A103V or vice versa in OsPIP2;7 V95A (Figs. 5-7 and supplemental Figs. S2-S4). Our simulations correlated P f value with water-water and water-protein interactions on the viewpoint of conformational diversity and stability of key residues, which bridged the structural changes and functional measurements (27,28). This is because the orientation determines the energy barrier, and the instability interrupts the constancy of water motion, thereby affecting the dynamic process of water transport and the statistics of water-water and water-protein interactions. Overall, our ana-lyses provided structural bases for understanding the role of Ala/Ile(Val) site in P f .
It has been known that ar/R region determines the selectivity of AQP, and the mutation of its residue modified the channel property (18). In this study, the Ala/Ile(Val) mutation induced the orientation change of Phe 97/89 residue in the ar/R region and, in turn, enhanced or reduced P f in OsPIP1s or OsPIP2s. It still remains unknown whether the mutation affects the selectivity of OsPIP1s and/or OsPIP2s or whether there is a synergetic or competing mechanism between the selectivity and the permeability. It is possible that PIP1s are less permeable to water because they can transport other molecules, e.g. CO 2 (41). It should also be pointed out that there are several other amino acids that are conserved within PIP1 and PIP2 groups but not between the isoforms belonging to the two different subgroups. The mutations in any of these amino acids might also influence the function of the isoforms within these groups. Further investigations are still required to address these issues, which are beyond of the scope of this study.