Coordinated Post-translational Responses of Aquaporins to Abiotic and Nutritional Stimuli in Arabidopsis Roots*

In plants, aquaporins play a crucial role in regulating root water transport in response to environmental and physiological cues. Controls achieved at the post-translational level are thought to be of critical importance for regulating aquaporin function. To investigate the general molecular mechanisms involved, we performed, using the model species Arabidopsis, a comprehensive proteomic analysis of root aquaporins in a large set of physiological contexts. We identified nine physiological treatments that modulate root hydraulics in time frames of minutes (NO and H2O2 treatments), hours (mannitol and NaCl treatments, exposure to darkness and reversal with sucrose, phosphate supply to phosphate-starved roots), or days (phosphate or nitrogen starvation). All treatments induced inhibition of root water transport except for sucrose supply to dark-grown plants and phosphate resupply to phosphate-starved plants, which had opposing effects. Using a robust label-free quantitative proteomic methodology, we identified 12 of 13 plasma membrane intrinsic protein (PIP) aquaporin isoforms, 4 of the 10 tonoplast intrinsic protein isoforms, and a diversity of post-translational modifications including phosphorylation, methylation, deamidation, and acetylation. A total of 55 aquaporin peptides displayed significant changes after treatments and enabled the identification of specific and as yet unknown patterns of response to stimuli. The data show that the regulation of PIP and tonoplast intrinsic protein abundance was involved in response to a few treatments (i.e. NaCl, NO, and nitrate starvation), whereas changes in the phosphorylation status of PIP aquaporins were positively correlated to changes in root hydraulic conductivity in the whole set of treatments. The identification of in vivo deamidated forms of aquaporins and their stimulus-induced changes in abundance may reflect a new mechanism of aquaporin regulation. The overall work provides deep insights into the in vivo post-translational events triggered by environmental constraints and their possible role in regulating plant water status.

The absorption of soil water by roots is crucial in order for plants to maintain their water status. Studies in various plant species have shown that the root water permeability (root hydraulic conductivity Lp r ) is constantly adjusted depending on the developmental stage of the plant, its nutritional or hormonal status, or multiple environmental stimuli (1). Despite their importance in plant growth and adaptation, these multiple responses have not been investigated in a single plant species yet. Aquaporins form a large class of channel proteins that facilitate the diffusion of water and small neutral solutes across cell membranes and, among many other functions, contribute to root water uptake (2,3). Aquaporins are 25-to 30-kDa proteins with six membrane-spanning domains and five connecting loops (A-E), with N-and C-terminal tails exposed to the cytosol (4,5). Plant aquaporins show a high multiplicity of isoforms. Thirty-five homologs belonging to four homology subclasses have been identified in Arabidopsis. The plasma membrane intrinsic proteins (PIPs) 1 (with 13 isoforms further subdivided into the PIP1 and PIP2 subgroups) and the tonoplast intrinsic proteins (TIPs) with 10 homologs are the most abundant aquaporins in the plasma membrane and the tonoplast, respectively (6,7). Two other families include nine homologs of the nodulin-26-like proteins and three homologs of small basic intrinsic proteins (6 -8). Several lines of functional analysis, including the use of aquaporin chemical blockers or phenotypic analysis of transgenic plants with antisense inhibition or insertional mutagenesis of PIP genes, have demonstrated that PIP aquaporins mediate a significant part of root water uptake (9 -12). However, the mechanisms that underlie PIP regulation in roots under abiotic constraints appear composite and remain partially understood. In addition to the long-term transcriptional control of aquaporin functions, multiple post-translational mechanisms that affect aquaporin activity, targeting to their destination compartment, or stability seem to be involved in the response of plant roots to environmental and hormonal stimuli.
Post-translational modifications (PTMs) are central for regulating protein structure and function and thereby modulating protein catalytic activity, subcellular localization, stability, and interaction with other partners. Protein phosphorylation is one of most important and best characterized PTMs. Virtually all cellular processes are regulated in one or multiple ways by reversible phosphorylation. PIPs are no exception and show a conserved phosphorylation site in their first cytosolic loop (loop B) and multiple phosphorylations in adjacent sites of their C-terminal tail (13)(14)(15), as listed in the PhosPhAt database. Aquaporin phosphorylation seems to be a significant target in plants under stress: following exposure of Arabidopsis roots to salt (NaCl) or hydrogen peroxide (H 2 O 2 ), AtPIP2;1 phosphorylation is decreased and increased, respectively (15). Abscisic acid treatment of Arabidopsis seedlings also decreases the phosphorylation of several AtPIPs (16). A transient increase in phosphorylation of AtPIP2;6 was observed upon sucrose resupply to starved Arabidopsis plantlets (17), whereas nitrogen resupply to starved plants results in a transient dephosphorylation of AtPIP2;6 and AtPIP2;7 aquaporins (18). The phosphorylation of PIPs is thought to act on both gating and subcellular localization (5,15). In particular, the atomic structure of spinach SoPIP2;1 indicates that the phosphorylation of Ser115 in loop B directly contributes to opening of the pore (5). In addition, phosphorylation of C-terminal Ser274, a residue homologous to Ser280 of AtPIP2;1, acts on an adjacent SoPIP2;1 monomer to prevent its transition to a closed-pore conformation (5). Finally, phosphorylation of Ser283 was shown to be necessary for targeting of newly synthesized AtPIP2;1 to the plasma membrane and interfered with its internalization in response to salt stress (15). Plant aquaporins are subjected to many other PTMs. Methylation has been described in AtPIP2;1, but its functional significance is as yet unknown (1,19,20). Drought-induced PIP ubiquitination in the endoplasmic reticulum was shown to prevent its trafficking to the plasma membrane and triggered its degradation (21).
Qualitative and quantitative information about PTMs and their dynamic changes is now critically needed in order to address the general principles and complexity of aquaporin regulation during plant responses to stresses. To this end, we investigated in Arabidopsis a set of representative environmental stimuli, including water, osmotic, ionic, oxidative stress, and nutritional constraints. The effects of these stimuli on Lp r were characterized. The results provided a well-defined time frame for investigating the dynamics of aquaporin PTMs using dedicated quantitative approaches. The overall study points to multiple regulatory processes that can be shared between treatments. In particular, coordinated phosphorylation events and a putative role for deamidation were uncovered.

EXPERIMENTAL PROCEDURES
Plant Culture-Arabidopsis thaliana plants, ecotype Columbia-0 (Col-0), were grown in hydroponic conditions as described previously (22). Briefly, sterile seeds were sown on an MS/2 medium (23)  Water Transport Measurements-Measurements were performed essentially as described elsewhere (22). Briefly, the root system of freshly detopped 21-day-old Arabidopsis plants was inserted into a pressure chamber filled with the culture solution. The hypocotyl was carefully threaded through the soft plastic washer of the metal lid and connected to a flow meter using a low-viscosity dental paste (PRESIDENT microSystem, Coltene, Altstä tten, Switzerland). Pressure was then slowly applied to the chamber using nitrogen gas. Ten minutes of pre-pressurization were required in order to achieve a stabilized sap flow (J v ), which was then determined over 20-min periods for imposed hydrostatic pressures between 0.16 and 0.72 MPa. Root and shoot dry weight (DW) was measured after 48 h of drying. The hydraulic conductivity (Lp r , ml g Ϫ1 h Ϫ1 MPa Ϫ1 ) of each individual root system was calculated from the slope of a plot of J v versus driving pressure, divided by the root DW.
Protein Digestion and Phosphopeptide Enrichment-In-solution reductions/alkylations were performed simultaneously with detergent removal via the filter-aided sample preparation protocol (28). These steps were followed by two types of double digestions. One of these subsequently used Lys-C (Roche Applied Science) for 4 h at 37°C and trypsin (Sequencing Grade Modified, Promega, Madison, WI) overnight at 37°C. The other used chymotrypsin (Sequencing Grade Modified, Promega) for 5 h at 25°C and then AspN (Sequencing Grade Modified, Promega) overnight at 25°C. Peptides were eluted via step elutions with 50 mM ammonium bicarbonate followed by 50% acetonitrile and then 0.5 m NaCl. Peptides were desalted on C18 columns (Sep-Pak® Vac tC18 cartridge, 3 cc, Waters, Guyancourt, France) before strong cation exchange separation. Dried peptides were resuspended in 10 mM ammonium formate with 30% acetonitrile at pH 2.7 and loaded on a Famos-Ultimate HPLC (Dionex, Courtaboeuf, France) at 200 l min Ϫ1 on a column (PL-SCX 1000A, 5 m, 50 ϫ 2.1 mm, Varian, Agilent, Massy, France). A salt gradient (10 mM ammonium formate, 800 mM NH 4 Cl, 30% acetonitrile, pH 2.7) allowed us to obtain nine fractions. In the case of chymotrypsin/AspN digestion, two peptide fractions were obtained and quantified. In the case of Lys-C/trypsin digestion, nine fractions were obtained and treated as follows: 10% of the volume of fractions 4 (F4) to 9 (F9) were kept out to direct LC-MS/MS analyses. Fractions F2, F3, and F4 and fractions F5, F6, and F7 were pooled before the phosphopeptide enrichment step. The F1 fraction was set aside. After solvent evaporation, peptides were resuspended in a loading buffer (25% lactic acid, 60% acetonitrile, 0.75% trifluoroacetic acid (TFA)) for titanium dioxide (TiO 2 ) enrichment according to Refs. 29 and 30. Briefly, a C8 plug was packed on-tip with metal oxide bulk beads and equilibrated with 1% TFA and 80% acetonitrile including 25% lactic acid as selectivity chelating enhancers (buffer A). Two successive columns were built with 1 to 2 mm and 4 to 5 mm of TiO 2 beads, respectively. After loading, washes were performed with buffer A and buffer B (1% TFA in 80% acetonitrile) and peptides were eluted twice with NH 4 OH at 0.5% and then 5%. Eluates were pooled and dried.
Identification of Peptides and Modified Peptides-Digested peptides were concentrated with a pre-column (Dionex, C18 PepMap100, 300 m ϫ 5 mm, 5 m, 100 A) and separated with a reversed-phase capillary column (Dionex, C18 PepMap100, 75 m ϫ 250 mm, 3 m, 100 A) over a 140-min gradient. Peptide fragmentation was carried out with a Q-TOF-MS/MS system equipped with a nano-electrospray ionization source (Maxis, Bruker Daltonics, Bremen, Germany) and performed in the positive ion mode. The corresponding data were interrogated with Mascot 2.2.07 (Matrix Science) via ProteinScape software (Bruker Daltonics) against the Arabidopsis proteome (Tair10, 31,960 entries). Parameters of interrogation accepted one putative missed cleavage and 15-ppm and 0.05-Da mass ranges for the parent peptide and the MS/MS fragment, respectively. The selected enzymes were trypsin or chymotrypsin/AspN. Because proteins were reduced and alkylated, carbamidomethylation was taken as a fixed modification. Variable modifications were N-terminal acetylation; deamidation of asparagines and glutamines; methionine oxidation; phosphorylation of serines, threonines, and tyrosines; and methylation of aspartic and glutamic acids and of serines. Using the above criteria for protein identification, the rate of false peptide sequence assignment (false discovery rate) as determined by the "decoy database" function implemented in Mascot version 2.2.07 was 1%. All modified aquaporin peptides were checked manually via comparison of experimentally identified masses with the theoretically accepted masses for the fragments. Modified peptides were accepted if the PTM score was Ͼ20 and/or site-determining b or y ions were present.
Label-free Peptide Quantification and Statistical Analysis-Microsomal samples prepared from three independent plant cultures were all analyzed with two technical replicates corresponding to independent injections in the mass spectrometer. Data corresponding to peak intensity were computed using the ProfileAnalysis software (Bruker Daltonics). Fractionated peptides obtained after digestion with Lys-C/trypsin or chymotrypsin/AspN were quantified with a label-free approach using a direct comparison of the peak intensity of each peptide ion in multiple LC-MS. Bruker software (DataAnalysis, Pro-fileAnalysis) provided a workflow including peak detection, peak matching/alignment, normalization, detection of differential peptides, and statistical analyses of data. In particular, the technical variability was reduced via the "quantile" method (Bruker Daltonics), the goal of which is to homogenize the distribution of compound intensities for each LC-MS run in a common data set containing all acquired LC-MS runs. ProfileAnalysis (Bruker Daltonics) was used to discriminate peptides with quantitative variations, which were further identified via collision-induced dissociation. Three independent biological experiments and two technical repeats by experiment were used to calculate quantitative ratios and corresponding p values. An aquaporin peptide was considered for data interpretation if a quantitative ratio could be calculated in at least half of the studied treatments (i.e. seven) with a p value below 0.1 for at least three of the treatments. When a peptide met these criteria, its quantitative variations were considered in all treatments. For proper comparison of all treatments, quantitative data of each selected peptide were centered and reduced. Hierarchical clustering was performed using Multi-Experiment-Viewer 4.0. Euclidian was used as a distance matrix.

Modulation of Lp r by Multiple Abiotic
Treatments-In a previous study, we showed that exposure of Arabidopsis roots to 100 mM NaCl induces a rapid (half-time: 45 min) and marked (Ϫ70%) decrease in Lp r (22). In the present study, plants were treated for up to 6 h with an equivalent hyperosmotic solution containing 200 mM mannitol (Fig. 1A). This treatment induced a decrease in Lp r by up to 63%, with a half-time of 60 min. The application of H 2 O 2 to excised Arabidopsis root systems was previously shown to induce a timeand dose-dependent decrease in pressure-induced sap flow (J v ), an indicator of Lp r in these conditions (31). In particular, 0.5 mM H 2 O 2 induced a 40% inhibition of J v after 15 min (31). Here, we probed the effects on J v of NO, another oxidative treatment, using DEA-NO as a NO donor molecule. The application of varying concentrations of DEA-NO (50 to 500 M) induced a dose-dependent and progressive inhibition of J v over time (Fig. 1B, supplemental Fig. S1A). Upon prolonged treatment with 100 M DEA-NO, J v spontaneously reversed from its maximum (60%) inhibition until it eventually recovered to its initial level (supplemental Fig. S1B). Pre-treatment of the root system with an NO scavenger (100 M cPTIO) (24,25) fully counteracted the inhibitory effects of 100 M DEA-NO (supplemental Fig. S1C), indicating that NO specifically inhibits Arabidopsis Lp r .
The effect of plant nutritional status on Lp r was also investigated. After plants had been maintained for 1 day in darkness to deplete plant carbon, Lp r was inhibited by 30% ( Fig.  2A). The plants were then supplied with 30 mM sucrose or 30 mM mannitol (used here as an osmotic control). Whereas an additional 20% Lp r inhibition was observed after a 4-h treatment with 30 mM mannitol, Lp r was increased by 40%, with respect to control plants in darkness conditions, after 4 h of sucrose treatment ( Fig. 2A). Moreover, plants cultivated in the light showed inhibition of Lp r by 34% and 25% in response to treatments with 30 mM sucrose for 2 h and 4 h, respectively ( Fig. 2B). This inhibition was similar to that obtained by treating plants with 30 mM mannitol over 4 h (Fig. 2B). These results suggest that sucrose can act as an osmoticum or as a metabolite to inhibit or enhance Lp r , depending on whether it is supplied to plants cultivated in light or darkness conditions, respectively.
Phosphate and nitrate availability can dramatically alter plant growth and development (32). In our experimental conditions, different morphological alterations were observed according to the type of nutrient deficiency: a 6-day-long phosphate starvation induced a relative decrease in shoot DW with respect to replete plants, whereas a 6-day-long nitrate starvation induced a relative increase in root DW (supplemental Fig. S2). After 6 days of phosphate starvation, Lp r was decreased by 77% (Fig. 2C). This reduced value was increased by 50% and 100% after 2 h and 24 h of phosphate resupply, respectively (Fig. 2C). Following nitrate starvation, Lp r remained unchanged for 4 days (Fig. 2D) and was reduced by about 50% after 5 or 6 days (Fig. 2D). The resupply of plants with 4.25 mM nitrate for up to 24 h did not reverse Lp r inhibition (data not shown). These different morphological and physiological responses suggest that distinct molecular mechanisms are involved in plant responses to nitrate and phosphate starvation.
The overall results show that Arabidopsis root water permeability is responsive to osmotic (NaCl, mannitol), oxidative (H 2 O 2 , NO), and nutritional (sucrose, phosphate, and nitrate) stimuli. As aquaporins contribute to Ͼ60% of Lp r in this species (33), this root system appears to be a suitable model for addressing the modes of aquaporin regulation in roots under various abiotic constraints. For this purpose, proteomic analyses were performed on roots exposed to the abovedescribed treatments for durations that induced half or full Lp r inhibition. In brief, membrane proteins were prepared from roots of plants treated with 100 mM NaCl for 45 min (NaCl t1/2 ) and 2 h (NaCl tmax ), 200 mM mannitol for 1 h (man t1/2 ) and 4 h (man tmax ), 500 M H 2 O 2 for 8 min (H 2 O 2t1/2 ) and 18 min (H 2 O 2tmax ), and 125 M DEA-NO (NO t1/2 ) for 8 min and 18 min (NO tmax ). We also investigated (i) plants grown in darkness (dark) for 1 day and treated with 30 mM sucrose (dark suc ) or 30 mM mannitol (dark man ) for 4 h and (ii) plants starved in nitrate (N starv ) or phosphate (P starv ) for 6 days with, in the latter case, an optional phosphate resupply (P resup ) for 2 h.
Identification of Aquaporin Peptides and Their Modified Forms-Efficient identification of aquaporins and their PTMs requires optimal peptide sequence coverage. For this, we developed two types of double digestions of microsomal proteins, employing LysC followed by trypsin and chymotrypsin followed by AspN, respectively. Additionally, the released peptides and phosphopeptides were separated according to their charge on a strong cation exchange column, and phosphopeptides were further enriched with TiO 2 microcolumns. Using the two types of double digestions, we were able to identify 1291 proteins from a decoy database set to a false discovery rate of 1% with a 95% confidence level (supplemental Table S1). 1251 proteins were identified with LysC/ trypsin digestion and 162 proteins with chymotrypsin/AspN digestion, with 40 of the 162 proteins being specific for the latter digestion. In addition, 32% and 16% of proteins showed at least one phosphorylation and one deamidation event, respectively (supplemental Table S1). This global analysis revealed the presence of 12 PIP isoforms out of a total of 13 (supplemental Table S1). The only lacking PIP isoform (PIP2;6) was detected and shown to play a role in leaves (34). In addition, we detected 4 out of the 10 TIP isoforms (TIP1;1, TIP1;2, TIP2;2, TIP2;3), with 2 of them (TIP1;1 and TIP1;2) specifically identified through chymotrypsin/AspN digestion. Consistent with a lack of detection in roots, TIP3;1 and TIP3;2 are known to be specifically expressed in seeds, and TIP1;3 and TIP5;1 in pollen (35,36). MS analyses also showed that numerous PIP isoforms were modified by acetylation, phosphorylation, methylation, or deamidation (Table I, supplemental Fig. S3). Among the 14 phosphorylation sites identified in this study, 4 were located in the N-terminal tail of PIP1;1/PIP1;2 (Ser24, Ser27), PIP2;2 (Thr22), and PIP2;4 (Ser10) and were novel according to the PhosphAt database. The mass variation induced by deamidation is 0.98 Da. Thus, the monoisotopic ion was carefully examined to avoid any mistake in the selection of the parent (supplemental Fig. S4). Deamidation of Asn and Gln residues (Table I) S3). We previously showed that Lys (Lys4) and Glu (Glu6) residues can be methylated in the N-terminal tail of PIP2;1 (20). Although these modifications were not observed in this study's samples, methylation of a Glu residue was observed at another position in the N-terminal tail of PIP2;2 (Table I, supplemental Fig. S3). Combinatorial deamidation and phosphorylation was also observed in some peptides. This revealed a particularly complex modification profile in the N-terminal tail of PIP1;1-1;2 and in the C-terminal tail of PIP2;1-2;3 and PIP2;7 (Table I, supplemental Fig. S3). Although not exhaustive, these analyses establish that PIP aquaporins are highly modified proteins and point to a multiplicity of posttranslational regulation mechanisms acting on a same protein (Fig. 3). 2. Lp r response to light and carbon (A, B), phosphate (C), or nitrate (D) availability. A, Lp r was measured in plants grown in the light (light), exposed to a 24-h period of darkness (control), or exposed to a 24-h period of darkness followed by a treatment with 30 mM sucrose (S 2h and S 4h ) or 30

Classification of Aquaporin Peptides and Their Modified
Forms-A label-free MS approach was developed for comparative quantification of aquaporin peptides between treatments. Hierarchical clustering was then used to classify both unmodified and modified PIP peptides according to treatments (Fig. 4). We considered 55 aquaporin peptides, the quantitative behavior of which was significant in at least 3 treatments (i.e. with p Ͻ 0.1) and quantified in at least 7 treatments out of 14 so as to avoid misinterpretation due to missing data (Fig. 4, supplemental Table S3). For mutual comparison of all treatments, quantitative ratios of peptide abundance referring to specific control conditions were normalized to the general control condition. This overall analysis indicated that all treatments clustered with complementary kinetic points of the same or related stimuli (Fig. 4), with the exception of the man tmax treatment, which was surprisingly related to the H 2 O 2 treatments. The good biological consistency of this clustering attests to the quality of the overall proteomic data. The data yielded, however, a very intricate classification of the 55 peptides. A cluster of phosphorylated peptides including the phosphorylation of the N-terminal tail of PIP1 and of the C-terminal tail of PIP2;1-2;3 and PIP2;7 could be distinguished (Fig. 4). The phosphorylation of these peptides increased in darkness, H 2 O 2t1/2 , H 2 O 2tmax , and man tmax treatments and decreased in all other treatments except the N starv treatment, which showed contrasted phosphorylation behavior (Fig. 4). These data extend previous work showing the role of PIP2;1 phosphorylation during salt and H 2 O 2 responses (15). A general relationship was tentatively established between Lp r and aquaporin phosphorylation (Fig.  5). For each treatment, the mean relative abundance of all phosphorylated PIP peptides was reported to the normalized Lp r value (Fig. 5A). A significant correlation could be established (p ϭ 0.007; Fig. 5A), suggesting that aquaporin phosphorylation positively contributes to Lp r variations.
Aquaporin Abundance According to Treatment-Because variations of Lp r could also be due to primary changes in aquaporin abundance, we then focused our analysis on PIP and TIP expression levels in plants exposed to the osmotic, oxidative, or nutritional constraints described above. For this, we considered the abundance of proteotypic aquaporin peptides that can be considered as indicators of the abundance This table lists all aquaporin peptides harboring PTMs. The first column indicates the name of the peptide, as referred to in Fig. 4. The second and third columns contain the name of the aquaporin and the peptide sequence, respectively. The last two columns indicate the type and number of PTMs and the position of the modified residue(s). The fragmentations of peptides are detailed in supplemental Fig. S3. Acet, acetylation; Dea, deamidation; Met, methylation; P, phosphorylation; PIP1;1-1;2-1;3-1;4, PIP1;1 and/or PIP1;2 and/or PIP1;3 and/or PIP1;4; PIP1;1-1;2, PIP1;1 and/or PIP1;2; PIP2;1-2;2-2;3, PIP2;1 and/or PIP2;2 and/or PIP2;3. a A comma indicates the coexistence of different phosphorylation or deamidation sites within a peptide (supplemental Fig. S3).
of the entire protein, provided that they are not modifiable. Five PIP1, five PIP2, and eight TIPs could be selected according to these criteria (supplemental Fig. S5). Because not all PIP aquaporins could be quantified through this approach, we also made use of an additional quantification method (i.e. ELISA) using anti-PIP1 antibody that recognized PIP1;1 to -1;4 isoforms and anti-PIP2 antibody that recognized PIP2;1 to -2;3 isoforms (27). Supplemental Fig. S5A indicates that the PIP1 peptides showed decreased abundance after NO tmax treatment, and that this decrease was not as complete after NO t1/2 . This result was confirmed using an ELISA assay with anti-PIP1 antibodies (supplemental Fig. S5C). All PIP1 peptides also showed decreased abundance in response to a Na tmax treatment (supplemental Fig. S5A), as already shown by ELISA measurements (22). We also noticed a global decrease of PIP2 peptides in these conditions (supplemental Fig. S5B). Regarding nutritional constraints, the decrease in PIP2;1, PIP2;2, PIP2;4, PIP1;2, and PIP1;3 abundance in response to N starv and P resup treatments was confirmed by ELISA assays (supplemental Fig. S5C). Interestingly, the abundance of TIP1;1 and TIP1;2 showed marked variation in response to several treatments. The abundance of all detected TIP1;1 and TIP1;2 decreased in response to Na tmax , dark, and N starv treatments (supplemental Fig. S5D). In contrast, all TIP1;1 and TIP1;2 peptides showed increased abundance upon P resup (supplemental Fig. S5D). In a global analysis across all treatments, we considered normalized Lp r values and the mean relative abundance of peptides proteotypic for PIP1 (Fig. 5B), PIP2 (Fig. 5C), or TIPs (Fig. 5D). These analyses did not reveal any significant correlation between Lp r and PIP1 or TIP abundance (Figs. 5B, 5D). In contrast, a  Table I. FIG. 4. Heat map representation of aquaporin peptide abundance according to treatment. Aquaporin peptides that were unmodified or carrying phosphorylation (P) or deamidation (Dea), alone or in combination, were quantified via label-free MS and clustered using a Euclidian correlation distance metric. Green and red colors indicate down-and up-regulation, respectively (see scale). Gray indicates missing data. Corresponding peptide sequences are described in Table I and supplemental Table S2. Quantitative ratios and corresponding p values before centering and reduction of ratio values are indicated in supplemental Table S3. A cluster of phosphorylated peptides is framed. significant correlation could be detected between PIP2 abundance and Lp r , but with a low regression coefficient value (R 2 ϭ 0.20) (Fig. 5C). In summary, this study indicates that changes in Lp r are not commonly determined by changes in PIP1 or TIP abundance. They may depend, albeit to a low extent, on changes in PIP2 abundance. Such changes may contribute to more significant Lp r variations in a restricted number of treatments such as NaCl tmax , NO tmax , and N starv (supplemental Table S4).

DISCUSSION
The central aim of this work was to search for general molecular mechanisms involved in the regulation of root aquaporin function. For this, we characterized a large set of physiological contexts known to possibly modulate Lp r . These contexts were then used to develop a comprehensive proteomic analysis of root aquaporins.
Arabidopsis Root Is Highly and Specifically Responsive to Different Abiotic Treatments-One early achievement of this work was the identification of nine physiological contexts that modulate root hydraulics in the same plant species, in time frames of minutes (NO and H 2 O 2 treatments), hours (mannitol and NaCl treatments, exposure to darkness and reversal with sucrose, phosphate supply to phosphate-starved roots), or days (phosphate or nitrogen starvation). All treatments induced inhibition of root water transport except for sucrose or phosphate resupply to dark-grown or phosphate-starved plants, respectively. Abscisic acid is another well-known Lp r regulator in several plant species (1). However, this hormone had no effect on Arabidopsis root water transport in our experimental conditions (data not shown). The inhibition of Lp r in response to hyperosmotic treatment or reactive oxygen species has already been reported in many species, including FIG. 5. Relationship between Lp r and abundance of native and modified aquaporins. The mean relative abundance of phosphorylated PIP peptides (A) and of proteotypic peptides for PIP1s (B), PIP2s (C), and TIPs (D) was related to Lp r in NaCl tmax , man tmax , H 2 O 2tmax , NO tmax , dark, dark suc , dark man , N starv , P starv , and P resup conditions. All Lp r values were normalized with respect to Lp r in corresponding control conditions. A, phosphorylated PIP peptides are PIP11 12 (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29) Table I). B, PIP1 proteotypic peptides are PIP12 (loop E), PIP13 , PIP14 , and PIP15 . C, PIP2 proteotypic peptides are PIP21 Arabidopis (22,31), but the present work provides the first evidence of effects of active nitrogen species on root water transport. A role for NO, an important molecule in plant stress signaling, in aquaporin regulation during rice seed germination has been proposed (37). A thorough examination of root hydraulic responses to major plant nutrients or metabolites (phosphate, nitrate, and sugars) revealed highly specific responses. In the case of nutrient deprivation, this could reflect distinct morphological responses. For instance, phosphate starvation did not modify root mass (supplemental Fig. S2A) but decreased shoot mass by 55% (supplemental Fig. S2B), whereas nitrate starvation enhanced root DW by 100% (supplemental Fig. S2D) without inducing any change in shoot DW (supplemental Fig. S2E). The accompanying adjustments of root hydraulics may be somewhat related to these differential shoot-to-root alterations. For instance, the decreased Lp r of nitrate-starved plants can be interpreted as maintenance of the whole root hydraulic capacity (conductance) while the root absorption surface is increased to match the unchanged water demand for shoot growth and transpiration. In contrast, the decreased Lp r of phosphate-starved plants may represent a direct adjustment to a reduced water demand of growthinhibited shoots.
Although most abiotic stresses result in the inhibition of root water transport, nutrient supply to nutrient-starved plants was also investigated as a possible context for studying rapid stimulus-induced up-regulation of Lp r . In contrast to phosphate resupply, which had such effects, no change in Lp r was observed after a 1-day resupply of nitrate to nitrate-starved plants. Here, we typically face a species-specific response, as a similar nitrate treatment stimulates Lp r in birdsfoot (38 -40), tomato, cucumber (41), maize (40), or sunflower (38).
Carbon metabolites were investigated as still another type of physiological stimulus. Interestingly, Lp r was increased by 40% after sucrose resupply to plants cultivated in prolonged darkness. By contrast, a similar treatment in plants cultivated in the light inhibited Lp r by ϳ30% after 4 h (Fig. 2B). This inhibition was similar to that obtained by treating plants with 30 mM mannitol. Thus, sucrose exerts an inhibitory or a stimulatory effect when it is supplied to plants cultivated in light or darkness conditions, respectively.
Relationships between Aquaporin Abundance and Root Water Permeability-Aquaporin abundance is thought to be a critical determinant of Lp r . In support of this, we recently showed that a double pip2;1 ϫ pip2;2 knock-out mutant shows a 40% decrease in Lp r in control culture conditions (42). In addition, we recently quantified PIP aquaporins from roots using a multiple-reaction monitoring approach and showed that PIP1;1, PIP1;2, PIP2;1, PIP2;2, PIP2;4, and PIP2;7 aquaporins are the major aquaporin isoforms expressed in roots, in amounts from 70 to 280 pmol/mg of membrane proteins (43). Thus, the regulation of their abundance, as native or modified forms, should highly contribute to the observed changes in Lp r . Based on the physiological studies of the present work, we developed a thorough proteomic approach with peptide fractionation to quantify the abundance of specific PIP and TIP isoforms. In these respects, quantitative proteomics revealed that several treatments resulting in Lp r inhibition induced a prominent decrease in the abundance of some PIP1 and PIP2 isoforms (supplemental Fig. S5). In particular, the decreased abundance of PIP1 in NO tmax appears as a pretty unique context in which a rapid degradation (Ͻ20 min) of PIP isoforms can be observed. A weak overall dependence of Lp r on PIP2 abundance was also detected (Fig. 5C). We note, however, that an inhibition of Lp r was not systematically associated with a decreased abundance of PIP, and vice versa (Fig. 5). Thus, PIP abundance is not the only determinant of Lp r . TIPs represent other interesting aquaporin candidates for the regulation of root water transport. Parallel changes in Lp r and TIP abundance were punctually observed upon Na tmax , N starv , and dark conditions that inhibited Lp r and upon P resup , which stimulated Lp r  Table I and supplemental  Table S1. p values Ͻ 0.1 are indicated.
(supplemental Fig. S5D). TIPs may contribute to transcellular water transport or, alternatively, cytoplasm osmoregulation in concert with PIPs. Although genetic evidence supporting a role of TIPs in root water transport is lacking, a few studies have suggested a role for TIPs in salt stress responses. For instance, the overexpression of a soybean TIP2 isoform in transgenic Arabidopsis resulted in an increased plant sensitivity to salt stress (44). By contrast, Arabidopsis plants overexpressing a ginger TIP1 isoform became able to germinate and grow at high NaCl concentrations (150 mM) (45). However, the present study indicated that TIP abundance is not a major determinant of Lp r (Fig. 5D).
In summary, our study indicates that the modulation of Lp r cannot be accounted for exclusively by modulations in PIP or TIP abundance. In view of the sensitivity of our MS analysis, it is very unlikely that an aquaporin isoform that would play a major role in root water transport was missed in our study. These results prompted us to investigate PTMs as another mode of aquaporin regulation in roots under abiotic and nutritional constraints.
Coordinated Post-translational Events to Regulate Aquaporin Function-In line with previous studies (13)(14)(15)(16)(17), the present work highlights the diversity of aquaporin phosphorylated forms. It also points to combinations of various PTMs within a single aquaporin, including phosphorylation, methylation, deamidation, and acetylation (Fig. 3). The hierarchical clustering of PIP peptides shown in Fig. 4 indicated that modified peptides showed variations profiles that were somewhat more coordinated than those of their unmodified counterparts. A previous study using an absolute quantification MS methodology showed that 2-or 4-h treatments with 100 mM NaCl induced an increased abundance of the (PIP21 (277-287) 2P) peptide form, whereas a 15-min treatment with 2 mM H 2 O 2 induced a decrease of the (PIP21 (277-287) 1P) form (15). The present label-free quantification data are in full agreement with this previous study (15), which validates the label-free strategy used here.
The present work further revealed coordinated changes in phosphorylation in the N-terminal tail of PIP1;1-1;2 and in the C-terminal tail of five PIP2 aquaporins (PIP2;1-PIP2;4 and PIP2;7) and showed that, in contrast to aquaporin cellular abundance, the extent of aquaporin phosphorylation is a predominant parameter for controlling Lp r (Fig. 5A). Sequence alignment showed that two phosphorylation sites were conserved among the five PIP2 aquaporins (supplemental Fig.  S6A). The first site, corresponding to Ser280 in AtPIP2;1, is homologous to Ser274 of SoPIP2;1, the phosphorylation of which enhances the opening of the aquaporin pore (5,13). Phosphorylation of the second site, Ser283 in AtPIP2;1, was shown to promote targeting of this aquaporin to the plasma membrane and to interfere with its intracellular sorting after stress-induced internalization (15). Therefore, we propose that a decrease in Lp r associated with decreased PIP2 phosphorylation involves both gating and regulated trafficking of PIPs. Conversely, the increased phosphorylation of the same two sites may contribute to the Lp r increase induced by sucrose resupply to plants cultivated in prolonged darkness. We note, however, that Ser280 and Ser283 phosphorylation and Lp r showed opposite variations under some other treatments (man tmax , H 2 O 2 treatments, dark, dark man , N starv , P starvation). This suggests the involvement of other post-translational regulatory mechanisms.
Deamidation appeared as another prominent PTM in Arabidopsis root aquaporins. This modification was also identified in 16% of microsomal proteins (i.e. 206 out of 1291 proteins; supplemental Table S1). Given that 8 out of 12 PIPs were found to be deamidated, this modification seems to more frequently affect PIP aquaporins than other membrane proteins. Deamidation possibly interferes with protein stability, protein activation, and protein-protein interactions (46 -51). In mammalian AQP0, several sites of N-terminal truncation have been identified as sites of Asn deamidation (52,53). This age-related PTM, together with other PTMs, including phosphorylation, is supposed to serve as a molecular clock (54) to regulate AQP0 function. Nonenzymatic deamidation of Asn (N) and Gln (Q) residues may also occur spontaneously during sample preparation (55). In this case, Asn deamidation tends to preferentially occur at NG, NS, and ND sequences, whereas Gln deamidation occurs at QV, QL, QG, and, to a lesser extent, QA and QE sequences (56). Among the deamidated Asn residues identified in this study, only one corresponded to an NG sequence (Table I). In addition, the abundance of the identified deamidated peptides showed quantitative variation according to the treatment (Figs. 4 and 6), reinforcing the notion that the deamidated residues observed in PIPs reflect genuine PTMs occurring in planta.
Interestingly, we observed that deamidation was closely associated with phosphorylation in C-terminal PIP2 peptides. This observation may provide a clue to explain how an apparent increase in PIP2 C-terminal phosphorylation could be associated with a decrease in Lp r (Fig. 4). A typical case is PIP2;7, for which the abundance of the monophosphorylated form ((PIP27 (270 -280) 1P) and that of the same form with an additional deamidation ((PIP27 (270 -280) 1P1Dea) showed opposite behaviors. It remains unknown whether or how neighboring deamidation and phosphorylation could interfere to alter protein function (57). Deamidation was also identified at a conserved Asn residue in loop C of several PIPs (Table I, supplemental Fig. S6B). In the case of AtPIP2;7, molecular modeling according to the previously described SoPIP2;1 structure (5) indicated that the deamidated Asn152 residue is located at the entrance of the pore, close (Ͻ5 Å) to Arg224 (supplemental Fig. S7). The latter residue contributes to one of the main pore constrictions (aromatic residue/Arg) involved in substrate specificity. Deamidation-that is, the replacement of an amide with a carboxyl group-introduces a negative charge at physiological pH and may thereby interfere with aquaporin action. CONCLUSION In the present study, we used a large panel of abiotic and nutritional stimuli to search for aquaporin peptides that show quantitative variations consistent with the regulation of root water transport. We describe a high diversity of PIP response profiles concerning both aquaporin abundance and PTMs. Thus, PIP and TIP aquaporins undergo a combination of several regulatory mechanisms, including (i) the regulation of protein abundance, mainly involved in NaCl, NO, and N starvation treatments; (ii) the modulation of the phosphorylation state of the N-terminal tail of PIP1;1-1;2 and the C-terminal tail of PIP2;1-2;3 and PIP2;7, which appears to be a predominant factor controlling Lp r ; and (iii) the deamidation of Asn and Gln residues. The possible effects of deamidation on aquaporin function, which clearly deserve further experimental work, may provide new perspectives in membrane protein research.