Changes in the Phosphoproteome and Metabolome Link Early Signaling Events to Rearrangement of Photosynthesis and Central Metabolism in Salinity and Oxidative Stress Response in Arabidopsis

Rearrangement of central metabolism and photosynthesis is linked to the dynamic phosphoproteome in abiotic stress signaling in Arabidopsis.

Abstract

Salinity and oxidative stress are major factors affecting and limiting the productivity of agricultural crops. The molecular and biochemical processes governing the plant response to abiotic stress have often been researched in a reductionist manner. Here, we report a systemic approach combining metabolic labeling and phosphoproteomics to capture early signaling events with quantitative metabolome analysis and enzyme activity assays to determine the effects of salt and oxidative stress on plant physiology. K⁺ and Na⁺ transporters showed coordinated changes in their phosphorylation pattern, indicating the importance of dynamic ion homeostasis for adaptation to salt stress. Unique phosphorylation sites were found for Arabidopsis (Arabidopsis thaliana) SNF1 kinase homolog10 and 11, indicating their central roles in the stress-regulated responses. Seven Sucrose Non-fermenting1-Related Protein Kinase2 kinases showed varying levels of phosphorylation at multiple serine/threonine residues in their kinase domain upon stress, showing temporally distinct modulation of the various isoforms. Salinity and oxidative stress also lead to changes in protein phosphorylation of proteins central to photosynthesis, in particular the kinase State Transition Protein7 required for state transition and light-harvesting II complex proteins. Furthermore, stress-induced changes of the phosphorylation of enzymes of central metabolism were observed. The phosphorylation patterns of these proteins were concurrent with changes in enzyme activity. This was reflected by altered levels of metabolites, such as the sugars sucrose and fructose, glycolysis intermediates, and amino acids. Together, our study provides evidence for a link between early signaling in the salt and oxidative stress response that regulates the state transition of photosynthesis and the rearrangement of primary metabolism.

Abiotic stress reduces plant growth, limits crop productivity, and plays a major role in determining plant ecotypes. In the field, a plant may experience several distinct abiotic stresses either concurrently or at different times throughout the growing season. Salinity is a common example of the abiotic stresses that plants may encounter. Plants have the potential for adaptation to improve salt tolerance by regulating the ionic homeostasis pathways (Quan et al., 2007). In Arabidopsis (Arabidopsis thaliana), the Salt Overly Sensitive (SOS) pathway is known to be activated by salt stress through molecular and genetic analyses, and it is defined by three protein components: SOS1–SOS3. Salt stress is known to increase the calcium concentration in the cytoplasm. SOS3, a myristoylated calcium binding protein, is proposed to perceive this signal, and it physically interacts with and activates a Ser/Thr protein kinase, SOS2. One of the downstream targets of the SOS3-SOS2 complex is SOS1, a plasma membrane-localized Na⁺/H⁺ antiporter. Overexpression of SOS1 improves salt tolerance in transgenic plants (Qiu et al., 2002), and genetic evidence indicates that mutations in SOS1, SOS2, or SOS3 reduce the Na⁺/H⁺ exchange activity. Furthermore, phosphorylation of SOS1 by the SOS2-SOS3 complex could increase salt tolerance in yeast (Saccharomyces cerevisiae; Quintero et al., 2002).

Hydrogen peroxide (H₂O₂) plays a dual role in plants as the toxic byproduct of normal cell metabolism and a regulatory molecule in stress perception and signal transduction. Under salt stress, the accumulation of reactive oxygen species (ROS), particularly H₂O₂, is induced. Ample evidence suggests that ROS function as signals, inducing ROS scavengers and other protective mechanisms as well as damaging agents contributing to stress injury in plants. The activation of two mitogen-activated protein kinases, MPK3 and MPK6, and mitogen-activated protein kinase signal transduction are connected to perception of H₂O₂ (Kovtun et al., 2000). An oxidative-induced kinase acts as an ROS sensor in plants to activate MPK kinases, and oxidative stress-activated MAP triple-kinase1 has been postulated to function as a mitogen-activated protein kinase scaffold activating H₂O₂-induced cell death in plants (Nakagami et al., 2004).

Both salinity and H₂O₂ trigger the phytohormone abscisic acid (ABA)-dependent and -independent transcriptional regulation pathways. ABA accumulation activates ABA-dependent signaling by increasing the expression of transcription factors, such as dehydration-responsive element binding factors and abscisic acid-responsive element binding (AREB) proteins, as well as stress-responsive genes, such as Responsive to Desiccation Protein 29A (RD29A) and RD22 (Xiong et al., 2002; Yamaguchi-Shinozaki and Shinozaki, 2006). AREB protein is also known to be activated by ABA-dependent phosphorylation events (Furihata et al., 2006).

Genetic experiments have revealed that transcriptional control in response to salt or oxidative stress is a major regulatory mechanism in plants (Gadjev et al., 2006). Many molecular components of stress-induced signaling processes remain unknown. Furthermore, most efforts so far to characterize signaling components of salt or H₂O₂-induced responses have monitored changes in gene expression (Saijo et al., 2000). However, a significant part of the early response occurs through posttranslational modifications of proteins. Reversible phosphorylation of proteins is an important mechanism by which organisms modulate cellular processes in response to environmental cues. Because individual phosphoresidues can have distinct functions, to dissect the complex responses of signaling pathways, identification and quantification of protein phosphorylation at specific sites are necessary.

In this study, we used an ¹⁵N metabolic labeling strategy to quantitatively assess site-specific protein phosphorylation and metabolite levels in response to salt and oxidative stress over a time course of 1 h in Arabidopsis cell culture. To robustly quantify protein phosphorylation, paired reciprocal labeling experiments were conducted. Site-specific quantitative changes in protein phosphorylation were compared between the two stress types, so that proteins phosphorylated in response to both salt and oxidative stresses could be identified. The same quantitative approach was used to recognize changes in primary metabolites and intermediates upon abiotic stress perception. To link protein phosphorylation to the control of enzyme activity and the changes that we observed in primary metabolism under salt and oxidative stress, we directly measured the activities of key metabolic enzymes shown to be differentially phosphorylated in our phosphoproteomics experiments. Our study presents a multifaceted picture of the molecular mechanisms of the early response phase to salt and oxidative stress perception and uncovers a link between protein phosphorylation, activity, photosynthesis, and primary metabolism.

RESULTS

Reciprocal Metabolic Labeling Strategy

A metabolic labeling strategy with the stable heavy nitrogen isotope (¹⁵N) was used for accurate quantification of site-specific changes of protein phosphorylation and metabolite abundance under salt or oxidative stress over time (Supplemental Fig. S1). Cell cultures grown with ¹⁵N as the sole nitrogen source were treated with salt or H₂O₂ for 5, 15, 30, or 60 min, whereas cells cultured with the abundant naturally occurring nitrogen isotope (¹⁴N) were left untreated as controls. We performed paired reciprocal experiments, where cells grown in ¹⁴N-containing medium were treated and cell cultures supplemented with ¹⁵N were left untreated as controls. The cells grown in the ¹⁵N-supplemented medium exhibited greater than 98% heavy isotope incorporation into proteins, which was determined as described before (Engelsberger et al., 2006). Cells from ¹⁴N- and ¹⁵N-supplemented medium were combined at a 1:1 ratio after cell harvest. Soluble proteins and plasma membrane proteins were isolated from the cells and digested with trypsin. Metabolites were extracted from the cells as previously described (Scherling et al., 2009). Phosphopeptides from the soluble and plasma membrane fractions were enriched using TiO₂ and measured by LC-LTQ-Orbitrap mass spectrometry (MS). The labeled and unlabeled counterparts of a phosphopeptide pair produced distinct isotopic envelopes that were integrated, and their area was compared. This experimental strategy was repeated twice for a total of six biological replicates (three forward and three reciprocal labeling experiments), in which plasma membrane and soluble phosphopeptides and metabolites were measured at the respective time points after salt and oxidative stress treatment.

Early Changes in Protein Phosphorylation in Response to Stress

We identified 3,659 peptides mapped to 898 proteins and 410 phosphopeptides mapping to 389 phosphoproteins from the soluble and membrane fractions in all experiments. The phosphorylation of 352 phosphoproteins was robustly quantified with the metabolic labeling strategy described above, considering only phosphopeptides quantified at least at two of four sampling points after stress treatments (Supplemental Tables S1 and S2). An increase in protein phosphorylation was considered a stress response if the abundance of the respective phosphopeptide was changed at least 2-fold at least at one time point compared with untreated cells. We quantified 173 unique phosphopeptides based on these criteria, mapping to 165 proteins that, therefore, showed changes in phosphorylation upon stress treatment. Furthermore, there are seven phosphoproteins identified by more than two unique phosphopeptides; because a single phosphopeptide could uniquely identify a specific protein, there are 158 phosphoproteins identified by a single phosphopeptide. Phosphorylation occurred on 86% Ser, 16% Thr, and 0.2% Tyr (two sites) residues (Supplemental Table S1), which is comparable with previous studies (Chen et al., 2010; Engelsberger and Schulze, 2012; Wu et al., 2013; Supplemental Table S3).

To gain a better impression of the functional extent of the induced changes in protein phosphorylation and the effects of the stress treatments on cellular processes, we performed gene ontology (GO) classification of the differentially phosphorylated proteins and analyzed enrichment of GO categories (Fig. 1). We found protein kinases and transporters to be significantly enriched together, constituting nearly one-half of all of the proteins with altered phosphorylation (Fig. 1A). Kinases were enriched, because they are one of the main components of signal transduction cascades, which are highly active in the early stages of an adaptive stress response. Notably, we detected a significant enrichment of phosphoproteins in the nucleus without using any enrichment or isolation procedures for this organelle. The nucleus is often the terminus of signal transduction, where gene expression is reprogrammed to mount an effective response to the perceived stress. The classification of the biological process revealed a significant enrichment of phosphoproteins involved in the response to abiotic stress as well as protein modification and transport reflecting their molecular function and compartmentalization (Fig. 1B). Interestingly, phosphoproteins involved in postembryonic development and photosynthesis were also significantly enriched. In general, proteins showing the most rapid changes in phosphorylation were mainly involved in immediate adaptation to stress, such as transporters. Proteins showing altered phosphorylation at 15 min poststress treatment included transcription factors, Sucrose Non-fermenting1-Related Protein Kinase2 (SnRK2) kinases, and photosynthesis subunit proteins In contrast, proteins with relatively slow changes in phosphorylation beginning 30 min after application of the stress mainly included proteins involved in central metabolism, vesicle transport, and protein ubiquitylation and growth regulating factor.

Figure 1.

GO analyses of all phosphoproteins together (A) and phosphoproteins showing changes in site-specific phosphorylation in response to salt and oxidative stress separately (B). All shown GO categories were significantly enriched compared with the Arabidopsis genome (TAIR9). P values are indicated (Fisher’s exact test with Hochberg false discovery rate correction).

Stress-Induced Modulation of Transport Proteins

Thirty transport proteins showed significant changes in site-specific phosphorylation mapped by 32 phosphopeptides after stress treatment (Fig. 2). As expected, phosphorylation of the well-known sodium sensor Na⁺/H⁺ antiporter SOS1 (AT2G01980.1) was rapidly induced in the cytoplasmic domain after 5 min of salt stress. SOS1 has Na⁺/H⁺ exchanger activity, and this transport activity is essential for Na⁺ efflux from Arabidopsis cells. Consistently, the sodium/hydrogen exchangers NHX1 (AT5G27150.1) and NHX2 (AT3G05030.1) showed increased phosphorylation at the C terminus 15 min after application of salt stress, peaking at 1 h, indicating the existence of multiple phosphorylation responses of Na⁺/H⁺ antiporters in Arabidopsis. Three members of the water-transporting aquaporin family, plasma membrane intrinsic proteins (PIPs), were found to be affected by salt and H₂O₂ stresses. The phosphorylation of PIP2;2 (AT2G37170.1) was decreased after 1 h of both stresses, whereas PIP3 (AT4G35100.1) showed a 2-fold decrease in its phosphorylation after 15 min of salt treatment. The PIP2;5 (AT3G54820.1) phosphorylation pattern over time, decreasing after 15 min and increasing slightly after 1 h of salt stress, suggests that addition of external salt leads to a rapid closing of the water channel and a subsequent adaptive response. These results are consistent with decreased phosphorylation of PIP preventing water loss in response to stress (Zhu, 2003). Phosphorylation of plasma membrane proton ATPases, AHA1 (AT2G18960.1) and AHA2 (AT4G30190.1) was increased by both types of stress. Plasma membrane proton ATPases are known to generate the driving force for Na⁺ transport by SOS1 (Zhu, 2003); therefore, the increased phosphorylation of AHA1 and AHA2 could increase salt tolerance in Arabidopsis.

Figure 2.

Changes in site-specific protein phosphorylation of transport proteins over time after stress treatment. Treated to untreated ratios of phosphopeptide abundance are shown. Red indicates oxidative stress, and green indicates salt stress. The same symbol behind a protein annotation indicates that phosphorylation at the same site is plotted. Ratios were not mean subtracted or scaled. Hierarchical clustering was performed using Canberra distance.

Induced Phosphorylation of SnRK Protein Kinases

The abundance of eight phosphopeptides from SnRK kinases changed upon stress treatment, including seven from SnRK2 family kinases, one from Sucrose Non-fermenting1-related protein kinase regulatory subunit γ1 (AtKIN-γ), one from Arabidopsis SNF1 kinase homolog 10 (AKIN10), and one from AKIN11. Four of the respective phosphorylation sites have not been described previously (Heazlewood et al., 2008). All SnRK2 kinases identified in our study were differentially phosphorylated at multiple Ser/Thr residues in their kinase domains after stress treatments (Fig. 3), confirming previous in vitro results (Fujii et al., 2009). Phosphorylation site Ser 158 of SnRK2.4 peaked at 15 min after salt stress and declined 1-fold compared with the unstressed condition after 1 h. Phosphorylation at the same site was induced 30 min after application of oxidative stress, increasing sharply and peaking after 1 h. This suggests that different dynamics underlie salt and oxidative stress signaling at the same site on SnRK2.4. Stress-responsive changes in phosphorylation were identified for two sites of SnRK2.8. Phosphorylation at Thr 158 was induced by salt 30 min after treatment, remaining constant for up to 1 h. H₂O₂ treatment, however, led to a constant level of phosphorylation at Ser 154 15 min after treatment. A Ser residue at position 158 is conserved in members of the SnRK2 family, and only SnRK2.8 has a Thr residue at this position. Plant responses to salt and H₂O₂ may, therefore, be distinguished by site-specific phosphorylation of SnRK2.8. These results suggest that SnRK2 family members are components of diverse signaling pathways and show the functional divergence of proteins, despite similar primary structure.

Figure 3.

Changes in site-specific protein phosphorylation over time after stress treatment. Treated to untreated ratios of phosphopeptide abundance are shown. Line width and point size are not uniform in some cases to improve clarity. The same symbol behind a protein annotation indicates that the phosphorylation at the same site is plotted.

Receptor-Like Kinases

Ten phosphopeptides corresponding to eight isoforms of Receptor-Like Kinases (RLKs) were identified in our experiments. Changes in the phosphorylation of at least one isoform were found at each time point, indicating a significant role of RLKs in stress-responsive signal transduction. RLKs constituted a large portion of the activated kinases identified in our study; some were specifically phosphorylated under salt or oxidative stress, whereas others were phosphorylated under both conditions. Several kinases showed similar induction profiles to both H₂O₂ and salt stress, with phosphorylation increasing 15 to 30 min after treatment and remaining constant for the duration of 1 h or peaking relatively late at 60 min poststress (Fig. 3)

Stress-Induced Changes of Transcriptional Coactivators

We identified 12 transcriptional coactivators with altered phosphorylation after both stress treatments (Fig. 3). The transcription factors myb type-related transcription factor2 (AtMYB2) and WRKY transcription factors (WRKYs) were previously shown to play a role in drought tolerance in Arabidopsis (Bae et al., 2003; Ren et al., 2010). In our study, WRKY1/Zinc-Dependent Activator Protein1 (AT2G04880.1), Virulence E2-interacting protein1 (VIP1; AT1G43700.1), and several other transcription factors, including early response to dehydration14 (ERD14; AT1G76180.1), showed a rapid increase in protein phosphorylation in response to salt stress as early as 5 min after treatment, remaining constant or increasing slightly for up to 1 h. VIP1 is well known for its role in pathogen resistance, and it has recently been implicated as a central factor in the osmotic/salt stress response (Tsugama et al., 2012). ERD7 (AT2G17840.1) also responded rapidly to salt stress, with phosphorylation increasing dramatically 5 min posttreatment and then declining to basal levels at 15 min. High-mobility group B (HMGB; AT3G51880.1) showed similar early response profiles for both salt and oxidative stress, and it is a notable exception, because generally, the response to oxidative stress was delayed to about 30 min poststress for ERD7 and ERD14. This was especially observable for three transcription factors responsive to oxidative stress alone: jumonji domain-containing protein (AT1G78280.1), Nck-associated protein1 (AT2G35110.1), and nuclear factor y subunit B1 (AT2G38880.1). In addition, 25 DNA/RNA-related proteins were modulated by stresses. Nine of them are involved in RNA splicing (Supplemental Table S1), such as splicing factor, ribosomal RNA processing protein RimM, and others. These results suggest that the plant mRNA splicing machinery is a major target of phosphorylation and that a considerable number of proteins involved in RNA metabolism may be targeted by abiotic stress-activated kinases.

Stress Responses in Photosynthesis

Interestingly, we detected site-specific changes in phosphorylation of several proteins involved in photosynthesis after salt and oxidative stress treatment (Fig. 3). We found that phosphorylation at Thr 537 and Thr 541 of State Transition Protein Kinase7 (STN7) was affected by stress. In Arabidopsis, STN7 functions in the phosphorylation of the light-harvesting system of PSII and for the state transition, a process that balances the light excitation energy between PSII and PSI and optimizes the photosynthetic yield upon changes in light quality and quantity (Lemeille et al., 2010). Changes in phosphorylation at the known C-terminal STN7 sites Thr 537, Thr 539, and Thr 541 were also detected, and previous studies reported that phosphorylation at the C terminus of STN7 kinase is required for its stability under state 2 conditions (Willig et al., 2011). Three previously unknown phosphorylation sites on light-harvesting II complex (LHCII) proteins, namely light-harvesting complex protein B1B2 (LHB1B2; AT2G34420.1), LHCb4.2 (AT3G08940.1), and LHCb1.2 (AT1G29910.1), were induced upon stress as well as known sites on one PSII core protein H (PsbH; ATCG00710.1) and one PSI core protein P (AT2G46820.1). The phosphorylation of LHCII proteins generally decreased either 15 or 30 min after stress treatment but increased strongly after this dip. A notable exception was LHCB1.2, which increased steadily after both salt and H₂O₂ treatment.

Changes in Protein Phosphorylation of Metabolic Enzymes

The abundance of 18 phosphopeptides mapped to 13 proteins involved in primary metabolism changed significantly over time after application of salt and/or oxidative stress (Fig. 3). The highest phosphopeptide abundance was recorded for the sucrose phosphate synthase (SPS) isoform with the accession number AT5G20280.1 in our experiments (SPS1F). The abundance of two phosphopeptides covering the phosphorylation sites Ser 125 and Ser 152 increased with time after salt stress treatment. Phosphorylation at the sites Ser 180 and Ser 717 mapping to the SPS isoform AT4G10120.1 was induced by both salt and oxidative stress, whereas phosphorylation at Ser 148 decreased exclusively after 15 min of H₂O₂ treatment and increased slightly after 1 h, suggesting site-specific control of protein function by phosphorylation. The abundance of two phosphopeptides mapping to nitrate reductase2 (NIA2) was altered by oxidative stress, whereas the phosphorylation pattern did not change after the application of salt stress. Phosphorylation of trehalose-phosphatase synthase7 (TPS7; AT1G06410.1) was increased after oxidative stress treatment. Two isoforms of UDP-Glc 6-dehydrogenase (UGD) showed altered levels of phosphorylation by both stress treatments. Increased phosphorylation of Arabidopsis monodehydroascorbate reductase2 (ATMDAR2) was observed 1 h after salt stress treatment, whereas phosphorylation of ATMDAR1 was increased by 30 min of H₂O₂ treatment, possibly reflecting changes in NAD⁺ status. The phosphorylation of carbon fixation, starch and sugar metabolism, and glycolytic and tricarboxylic acid cycle enzymes, such as phosphoenolpyruvate carboxylase1 (PPC1), UGD (UGD3 and UGD4), phosphoglyceromutase, phosphoenolpyruvate carboxykinase1 (PEPCK1), Fru-2,6-bisphosphatase, and acetyl-CoA synthetase, were all also affected by salt and oxidative stress.

Stress-Induced Metabolite Changes

In addition to measuring site-specific changes in protein phosphorylation, we monitored changes in metabolite abundance to determine adaptive processes inherent in a shift in primary metabolites and intermediates. The changes in primary metabolism were measured using gas chromatography (GC)-MS, and in total, 65 metabolites were identified after comparison with reference standards (Supplemental Fig. S2; Supplemental Table S4). More than one-half of the studied metabolites showed significant (P < 0.05) differential accumulation at least at one sampling time point, showing that stress led to a reprogramming of major metabolic pathways in Arabidopsis. The abundance of the disaccharides Suc and Fru increased after 1 h of salt stress, and Glc abundance increased 30 min after H₂O₂ treatment. Stress has been shown to lead to the accumulation of starch or other sugars, which represent not only energy and storage products but also, carbon precursors, transport compounds, and signaling molecules. Myoinositol increased 2-fold after 30 min of salt stress. Malic acid, citric acid, 2-ketoglutaric acid, and succinic acid increased and peaked at 1 h of salt stress, Val, ethanolamine, Leu, and phosphoric acid increased after 30 min of salt stress. The levels of malic acid, 2-ketoglutaric, and citric acid decreased 15 min after application of oxidative stress. We found that oxidative stress exerted pronounced effects on glycolysis. These effects include a rapid decrease of Glc-6-P, glyceric acid-3-P, and ribose-5-P at 30 min of H₂O₂ treatment. Noteworthy was the decrease of ribose-5-P, which is a precursor in nucleotide biosynthesis. The levels of these metabolites were also decreased under salt stress. The amino acids Ser, Thr, and isobutanoic acid increased upon application of oxidative stress, whereas Met levels decreased substantially in agreement with Met synthase being sensitive to oxidation. Pro and Gly, which are common stress markers in plants, were decreased under both stresses. To gain an impression of all of the data, we conducted a principal component analysis (PCA) on the combined phosphoprotein and metabolite data sets (Weckwerth, 2008; Fig. 4). It showed that the effects of the stresses were apparent (separated from untreated plants on principal component 1) and clearly distinct (separated from one another on principal component 2) on the metabolite as well as the phosphoprotein level.

Figure 4.

Sample pattern recognition in the PCA plot shows sample separation in the lower dimensional space spanned by the first two principal components (PCs) gained by eigenvalue decomposition of the combined metabolite-protein covariance matrix. M, Stress-responsive metabolites; P, stress-responsive phosphopeptides.

Relationship between Changes of Enzyme Phosphorylation, Enzyme Activities, and Metabolites

We measured the activity of those enzymes involved in central metabolism that showed changes in site-specific phosphorylation and that catalyze the reactions, presumably leading to the changes in metabolite levels that we observed in the metabolomics experiments over time under both stress conditions. The activity assays showed that SPS activities are negatively correlated with changes in protein phosphorylation after salt and oxidative stress (Fig. 5). It has been reported that SPS activity is inhibited by phosphorylation (Huber and Huber, 1996). Therefore, the accumulation of Suc and Fru can be explained by the decreased phosphorylation of SPS and Fru-6-P 2-kinase. The activity of PPC1 and UGD3 showed changes highly coordinated with the changes in protein phosphorylation levels (Fig. 3; Supplemental Fig. S2). These results indicate major rearrangement of metabolism, particularly for Suc synthesis and glycolysis pathways, in adaptation to abiotic stress dependent on control of enzymatic activity by reversible protein phosphorylation. PPC1 activity and phosphorylation both increased 30 and 60 min after stress. PEPCK1 activity was highly correlated with phosphorylation, reaching a minimum in the very early adaptive phase 5 min after application of stress and then increasing in later stages. These enzymes are involved in the production of C skeletons for the synthesis of amino acids and other metabolites that are derived from the tricarboxylic acid cycle. Decreased activity was measured for phosphoglucomutase and phosphoglycerate kinase after application of both types of abiotic stress.

Figure 5.

Central metabolism and proposed involvement of identified phosphoproteins for the stress-induced STN7-dependent phosphorylation signaling pathway. The STN7 kinase is assumed to be activated by stress-induced changes of the redox level and phosphorylate the mobile pool of LHCII proteins (LHB1B2, LHCb4.2, and LHCb1.2) and PsbH, thus favoring its migration to PSI and activating the state transition from state 1 to state 2. The stress-responsive phosphoproteins and metabolites are marked in red and blue, respectively. Changes of enzyme activities are shown in the green boxes, and the unit of enzyme activity is nanomoles gram⁻¹ fresh weight minute⁻¹. Quantitative analyses of stress-responsive metabolites are shown with the heat maps. Values represent log₁₀-transformed data of peak areas. The top of the heat map shows salt-responsive changes. The bottom of the heat map shows oxidative-responsive changes. Bars represent the relative levels at 0, 5, 15, 30, and 60 min.

DISCUSSION

Reversible phosphorylation of proteins plays a central role as a molecular switch in intracellular signaling pathways (Olsen et al., 2006; Chen et al., 2012). Stress-induced protein phosphorylation is an important regulator of plant growth. Common stress responses in plants are changes in the activity of light-harvesting complexes, tricarboxylic acid cycle activity, accumulation of sugar alcohols, and increased photorespiration (Zhu, 2001; Hua et al., 2012). Many previous studies were reductionist in nature, focusing on individual mechanisms of abiotic stress response in plants. Here, we report a systemic analysis, integrating ¹⁵N metabolic labeling with phosphoproteomics and metabolomics measurements to characterize salt- and oxidative stress-induced changes quantitatively in the early response phase over time. We measured the activity of enzymes implicated by the phosphoproteomics results as potentially regulated by protein phosphorylation under salt and oxidative stress to link protein phosphorylation with their activity and changes in metabolite levels.

A TiO₂-based strategy was used to enrich the very low-abundance phosphopeptides of the signaling proteins that otherwise would have escaped detection. Therefore, it is not entirely clear if the measured site-specific changes in protein phosphorylation are caused by changes in phosphorylation of this protein pool, changes in gene expression and hence, changes in the protein pool, or both. Nevertheless, the absolute levels of phosphorylated proteins are changing in response to the abiotic stresses, allowing us to infer a causal connection between stress perception, protein phosphorylation, protein function, and effects on metabolite levels.

The maintenance of K⁺ and Na⁺ homeostasis is crucial to the physiology of living cells under salt stress. Phosphorylation of sodium-proton antiporters at the vacuolar (NHX1 and NHX2) and plasma membrane (SOS1) was induced under increased salinity. NHX proteins function to drive the efficient compartmentalization of Na⁺ into the vacuolar lumen, and recent research reported that NHX1 and NHX2 also control K⁺ homeostasis (Barragán et al., 2012). In addition to maintaining ion homeostasis in the cytosol, plants under salt stress also need to establish water or osmotic homeostasis. Both salt and H₂O₂ are known to strongly decrease the water permeability of plants. As expected, three water channel protein aquaporins showed altered phosphorylation levels at a conserved regulatory site in the protein’s C-terminal domain. In vitro analysis in spinach (Spinacia oleracea) has shown that gate closure of PIP2 is triggered by dephosphorylation at the homologous site (Ser 274) in the carboxyl-terminal region (Johansson et al., 1996, 1998). Furthermore, dephosphorylation of PIP2s was previously shown to occur after different stress treatments (Prak et al., 2008). Phosphorylation of PIP2.2, PIP2.5, and PIP3 changed significantly in this study. The decreased phosphorylation of the PIP water transporters under our applied stress regimes is consistent with a possible decrease in catalytic activity, which was suggested by the phosphorylation-dependent gating of the PIP channel observed in spinach at this conserved Ser residue. These results are also consistent with a role of PIP phosphorylation in controlling the speed of water flux across cellular membranes under both salt and oxidative stress.

SnRK2s are important protein kinases in plants with a multitude of functions, including a role in stress response. In Arabidopsis, the SnRK2 family consists of 10 members. We measured changes in site-specific phosphorylation of all SnRK2 kinase family members with the exceptions of SnRK2.5, SnRK2.7, and SnRK2.9 upon application of salt and/or oxidative stress. The phosphorylation site of all SnRK2s was located in the activation loop, presumably representing a conserved regulatory structural element within the kinase domain and suggesting that the phosphorylation has an effect on SnRK2 activity. All three members of the SnRK2 subclass III (SnRK2.2, SnRK2.3, and SnRK2.6) were differentially phosphorylated under salt stress. These proteins are known to be activated in an ABA-dependent manner, and in turn, they phosphorylate and activate downstream transcription factors, such as AREB transcription factors (Fujii et al., 2011). We found that phosphorylation of AREB was increased after 30 min of salt stress. In addition, a distinct temporal phosphorylation pattern was observed for the SnRK kinase subclasses I and II activation domains. This may indicate the differential mechanisms of activation and modulation of SnRK kinase activity that influence downstream phosphorylation and signal transduction. Taken together, our results indicate that SnRK kinase phosphorylation is modulated in at least two different ways corresponding to subclasses I and II.

The connection between the stress perception, downstream signaling, and metabolic energy pathways can lead to extensive changes in the transcriptome. AKIN10 and AKIN11 (SnRK1 family) are central integrators of stress response and energy metabolism, modulating the expression of more than 1,000 genes through phosphorylation of various transcription factors (Robaglia et al., 2012). The activity of AKIN10 was reported to be regulated by reversible phosphorylation (Shen et al., 2009). Furthermore, it is well known that primary metabolism is repressed in Arabidopsis by AKIN10/AKIN11 but activated in AKIN10/AKIN11 knockdown plants (Baena-González et al., 2007). The silencing of AKIN10 was also related to an activation of SPS and concomitant repression of transcription factors and some protein degradation regulators (Fragoso et al., 2009). Consequently, our data indicated that short-term metabolic adaptation is performed by AKIN10/AKIN11 through direct modulation of the activity of some key metabolic enzymes.

Modification of gene expression by phosphorylation-dependent activation of transcriptional coactivators provides a useful strategy for salt tolerance in plants. Basic leucine zipper (bZIP) transcription factors activate the expression of stress-induced genes involved in ion homeostasis, such as SOS1 and aquaporin PIP2.1 (Yang et al., 2009). ABA plays an important role in the adaption of abiotic stresses, such as drought and salinity. In the Arabidopsis genome, most of the AREB bZIP proteins are involved in ABA-responsive signal transduction pathways (Yamaguchi-Shinozaki and Shinozaki, 2006). bZIP transcription factors are known to be phosphorylated by the ABA-activated protein kinases SnRK2. Thus, AREBs modulate ABA-mediated AREB-dependent gene expression that enhances tolerance to abiotic stress in plants. VIP1 has recently been shown to play a central role in osmosensory signaling, and increased phosphorylation of VIP1 upon salt stress treatment was detected here. Other transcription factors, such as ERD7, ERD14, HMGB, and WRKY1, were also differentially phosphorylated under both stress treatments. The possible role of phosphorylation of these proteins in abiotic stress signaling is not well characterized.

Phosphorylation of PsbH proteins modulates the rearrangement of the entire membrane network of plant thylakoids. The N-terminal phosphorylation of PsbH must have a major role in stress-induced photosynthetic damage. Phosphorylation of LHCII proteins was reported for the adaptation to environmental changes (Fristedt et al., 2009). Moreover, LHCII proteins have been shown to be substrates of STN7 protein kinase (Pesaresi et al., 2011). Phosphorylation of LHCII was shown to be involved in the energy distribution between the two photosystems, and signaling between light reception and phosphorylation of LHCII was found to be related to the redox state of the plastoquinone pool, whereby reduced plastoquinol leads to kinase activation (Vener et al., 1998). Therefore, chloroplast protein phosphorylation affects the redox state (Kitajima and Butler, 1975). The differentially phosphorylated proteins identified here show that stress-activated protein kinases possibly phosphorylate the mobile pool of LHCII proteins as summarized in Figure 5.

The temporal changes of STN7 phosphorylation suggest that STN7 plays an important role in the adaptation of photosynthesis to changes in the cellular metabolic state. Because the phosphorylation of LHCII proteins and photosynthesis core proteins is STN7 dependent (Bellafiore et al., 2005), the phosphorylation sites identified on these proteins suggest that plants have signaling mechanisms that facilitate the rapid adaptation of chloroplast function to the prevailing abiotic environmental conditions. In a situation of decreased photosynthesis, the constant supply of carbon to central metabolism should be guaranteed. Changes in carbon metabolism were observed upon salt and oxidative stress treatments, with varying effects on the major pathways of primary metabolism. We found that the phosphorylation and activity of the enzymes SPS, UGD, and PPC were concurrent with the accumulation of Suc and Fru during stress, showing a rapid adaptive metabolic response. The measured increase in amino acids levels 30 to 60 min after stress may be linked to increased phosphorylation and activity of PPC1 and PEPCK1 at these time points. There is an observed decrease in abundance of tricarboxylic acid cycle intermediates, such as citric acid, 2-ketogluratic acid, succinic acid, malic acid, and glycolysis. Intermediates, such as Glc-6-P and glycerate-3-P, under oxidative stress can be linked to the decreased activities of enzymes, such as phosphoglucomutase and phosphoglycerate kinase, and are in agreement with the general energy conservation strategy previously reported based on transcriptomics results (Xiong et al., 2002). Interestingly, the tricarboxylic acid cycle intermediates were increased under salt stress, suggesting that plants may activate these pathways in the salt stress response and that there are congruent as well as differential aspects to the abiotic stress response pertaining to primary metabolism. Our results suggest a model that links STN7 phosphorylation and downstream phosphorylation of PsbH and LHCII complex proteins, modulating photosynthetic activity to phosphorylation and activity of metabolic enzymes, increase of sugars, such as Glc and Fru, and regulation of central metabolism under abiotic stress (Fig. 5).

In conclusion, this study has revealed the identity of a number of potential early protein targets of the phosphorylation/dephosphorylation machinery involved in abiotic stress response in vivo. Many proteins previously not implicated in salt or H₂O₂ tolerance had significantly altered temporal phosphorylation patterns under both stresses, including proteins of unknown function. The combination of phosphoproteomic and metabolomic data with measurements of enzyme activity presents a multifaceted picture of the early response phase to two different types of abiotic stress. Conserved and specific aspects of the response to both types of stress became evident on all three levels. Both salt and oxidative stress have a global impact on cellular metabolism, which is evident by the effects on photosynthesis, primary metabolism, particularly sugar and amino acid levels, glycolysis, and tricarboxylic acid cycle. The stress response is more specific when looking at changes in site-specific protein phosphorylation than in metabolite abundance, especially during the very early response phase.

MATERIAL AND METHODS

Metabolic Labeling of Arabidopsis Suspension Cell Cultures and Stress Treatments

A heterotrophic Arabidopsis (Arabidopsis thaliana) Columbia-0 cell suspension was cultured in JPL medium as previously described (Chen et al., 2010) at 23°C in an orbital shaker (120 rpm) under constant light. For quantitative proteomics experiments, one-half of the cell cultures was metabolically labeled with ¹⁵N by growing them with K¹⁵NO₃ as the only nitrogen source for 2 weeks, yielding a fully labeled proteome. Labeled and unlabeled Arabidopsis cell cultures were treated with 50 µm NaCl and 50 µm H₂O₂ on the sixth day after subculturing. Appropriate controls were treated with an equal volume of water. The cells were harvested 5, 15, 30, and 60 min after exposure to the stress treatments. Harvesting was performed directly from the flask onto a metal filter plate, and the medium was removed by suction. Cells were washed using a double volume of distilled water, immediately frozen in liquid nitrogen, and ground to a fine powder. Equal amounts of fresh weight of unlabeled and heavy isotope-labeled cells were mixed.

Preparation of Total Proteins and Plasma Membrane Proteins

For preparation of plasma membrane proteins, microsomal membranes were isolated from the tissue powder using extraction buffer as described previously (Chen et al., 2010). Total proteins were extracted with 50 mm HEPES-KOH (pH 7.5), 0.25 m Suc, 10% (w/v) glycerol, 0.6% (w/v) Polyvinylpyrrolidone K-25, 5 mm EDTA, 1 mm phenylmethylsulfonyl fluoride (PMSF), 5 mm dithiothreitol (DTT), 1 mm ascorbic acid, 50 mm NaF, 0.1% (v/v) Proteinase Inhibitor Cocktail, 0.1% (v/v) Phosphatase Inhibitor Cocktail 1, and 0.1% (v/v) Phosphatase Inhibitor Cocktail 2. Tissue homogenate was filtered through Miracloth (Merck) and centrifuged at 8,000g for the removal of the cellular debris. The tissue pellet was resuspended in microsomal suspension buffer containing 5 mm K₂PO₄/KH₂PO₄ (pH 7.8), 0.33 m Suc, 0.1 mm EDTA, and 1 mm DTT, and microsomes were obtained by ultracentrifugation at 32,000g. Plasma membrane was separated from the microsomes over a two-phase system mixture prepared with 5 mm K₂HPO₄/KH₂PO₄ (pH 7.8), 0.3 mm Suc, 6.4% (w/w) dextran T-500, 6.4% (w/w) polyethylene glycol 3350, and 5 mm KCl. The central microsomal phase was diluted 5 times with microsomal suspension buffer and centrifuged for 60 min at 38,000g at 4°C. The supernatant was discarded, and plasma membrane pellets were resuspended in buffer R of 5 mm K₂PO₄/KH₂PO₄ (pH 7.8), 0.33 m Suc, 0.1 mm EDTA, 1 mm DTT, 1 mm PMSF, and 5 µm leupeptin as described above plus 0.02% (w/v) Brij-58 for inverting vesicles. The inverted vesicles were subsequently washed three times with 100, 500, and 50 mm ice-cold NH₄HCO₃. Finally, the membrane protein parts were resuspended in 50 mm NH₄HCO₃ and stored at −80°C until further digestion. For soluble protein extraction, proteins were precipitated from the supernatant after initial centrifugation of the tissue homogenate using methanol-chloroform (Wessel and Flügge, 1984). The protein pellets were dissolved in 8 m urea and 100 mm NH₄HCO₃.

In-Solution Protein Digestion

For in-solution digestion, proteins were predigested for 3 h with endoproteinase Lys-C (1:100, w/w) at room temperature. After 4-fold dilution with 10% (v/v) acetonitrile and 25 mm NH₄HCO_3, proteins were digested overnight with Poroszyme immobilized trypsin (1:100, w/w) at 37°C.

Enrichment of Phosphopeptides

Phosphopeptides were enriched with titanium dioxide as described previously (Chen et al., 2010) with some modifications. TiO₂ tips were purchased from Glygen Inc. Before sample loading, the beads were equilibrated with 80% (v/v) acetonitrile and 0.1% (v/v) trifluoric acid (TFA) containing 20 mg mL⁻¹ 2,5-dihydroxybenzoic acid as a selectivity enhancer. The peptides were incubated with 5 mg of beads for 30 min. After successively washing with 80% (v/v) acetonitrile and 0.1% (v/v) TFA and then 10% (v/v) acetonitrile and 0.1% (v/v) TFA, the bound peptides were eluted from the beads with 200 μL of 0.3 m NH₄OH in 30% (v/v) acetonitrile (pH > 10). The eluted fraction was acidified with TFA and dried down completely in a vacuum concentrator.

Phosphopeptide Analysis and Identification by Liquid Chromatography/MS/MS

The peptide mixtures were analyzed by nanoscale C18 reverse-phase liquid chromatography (LC; Eksigent) coupled online to an LTQ-Orbitrap Mass Spectrometer (Thermo Fisher Scientific) using a 50-μm i.d. monolithic column (Merck KGaA) together with a 15-cm fused silica emitter. Peptides were eluted at 400 nL min⁻¹ with a three-step gradient from 5% to 25% (v/v) organic solvent (80% [v/v] acetonitrile, 2.5% [v/v] isopropanol, and 0.5% [v/v] formic acid) in 165 min followed by 25% to 80% (v/v) organic solvent in 5 min and 10% (v/v) organic solvent for 15 min. The eluate was electrosprayed online into the mass spectrometer by a nanoelectrospray ion source (Thermo Fisher Scientific). Peptide fragmentation was carried out by data-dependent acquisition of fragment ion spectra of multiple-charged precursor ions using collision-induced dissociation in the LTQ. Up to five data-dependent MS/MS spectra were acquired in the linear ion trap for each full-scan spectrum acquired at 30,000 full width at one-half maximum resolution in the Orbitrap, with an overall cycle time of approximately 1 s. Phosphopeptides were analyzed with multistage activation as previously described (Chen et al., 2010).

Phosphopeptide Identification

MS/MS spectra were converted to mascot generic file format with the DTASuperCharge version 1.17 software using the default settings with a precursor ion tolerance of 10 ppm. Peak lists were searched against a nonredundant Arabidopsis protein database with commonly observed contaminants amended (human keratin, trypsin, and lysyl endopeptidase; The Arabidopsis Information Resource 9 [TAIR9]; 35,386 entries) using the Mascot algorithm (version 2.2.0; Matrix Science). The following search parameters were applied: two missed cleavages were allowed, mass tolerances of 10 ppm for MS and 0.8 D for MS/MS were set, the enzyme was set to trypsin, carbamidomethyl was set as a fixed modification, and Met oxidation and phosphorylation of Ser, Thr, and Tyr were set as variable modifications. The false discovery rate was calculated using a concatenated target/decoy database and indicates the percentage of false-positive peptide identifications in the entire data set. In addition to a statistically significant MASCOT score, an ion score threshold of 30 was required for peptide identification. ¹⁵N metabolic labeling was chosen as a quantitative method for Mascot database searching, allowing the identification of labeled and unlabeled peptides in one database search.

Quantitative Phosphopeptide Analysis

Peptide pairs were quantified using the open-source software MSQuant. The precursor ion signal intensity of MS/MS spectra identifying the labeled (heavy) and unlabeled (light) counterparts of coeluting phosphopeptide pairs was extracted. Only phosphopeptide pairs identified in the forward and reciprocal counterparts of an experimental set were used for quantification. Formally, the precursor intensity counts of each MS/MS spectrum i identifying the light (MS2 I light) and heavy (MS2 I heavy) forms of each phosphopeptide in an LC-MS analysis j were normalized to the total precursor intensity counts of all MS/MS spectra k in the analysis. The normalized precursor intensity counts of the MS/MS spectra of the heavy and light forms of each phosphopeptide were then scaled to their average intensity counts in all LC-MS analyses for each respective condition (average of m spectra per LC-MS analysis in six analyses; three forward- and three reverse-labeling experiments per condition). Subsequently normalized and scaled precursor intensity ratios for each pair of MS/MS spectra identifying a phosphopeptide pair (MS2 Ratio_ij) were calculated in MSQuant:

Only phosphopeptides that showed opposite changes in the reciprocal experiments were considered to be correct and used further. The ratios obtained from the reciprocal experiments were converted to correspond to the ratios obtained from the forward experiments (i.e. treated to untreated ratios). The average ratio for all MS/MS spectral pairs for each phosphopeptide ion (p Ratio_l) was then used to determine its abundance ratio:

Subsequently, ratios of all phosphopeptide ions identifying the same phosphorylation site were averaged to arrive at site-specific fold changes in protein phosphorylation (P Ratio) in the experimental conditions:

Precision (sd and relative sd) was calculated accordingly. Ten to one ratios were inferred (0.1 [¹⁴N form only] or 10 [¹⁵N form only]) for peptides pairs for which only one counterpart was identified (either labeled or unlabeled) and no ratio could be calculated automatically. For estimation of the sample to sample variation of stress-induced phosphorylation within each time point, differences between the sample distributions were assessed by an R² test using the statistical programming environment R (Supplemental Fig. S2). All peptide sequences and phosphorylation sites were confirmed by manual inspection of the raw data to verify the peptide sequence and phosphorylation site assignment.

GC-MS Measurement and Data Analysis

Metabolites were extracted with a mixture of methanol:chloroform:water (2.5:1:0.5, v/v/v); 10 μL of internal standard solution (Leu-2,3,3-d₃, Asp-2,3,3-d₃, and d-sorbitol-¹³C₆) was added, and after fractionation, the upper polar phases were dried down completely in a vacuum concentrator and used for GC-MS analysis. The precipitate was dissolved and derivatized at 40 mg mL⁻¹ in dry pyridine, and it was incubated with 90 μL of N-methyl-N-trifluoroacetamide at 37°C for 30 min. The retention time index marker was added to the samples to determine the retention time index.

GC-MS analysis was performed on an HP6890 Gas Chromatograph (Agilent) with deactivated standard split-splitless liners. Sample volumes of 1 μL were injected in splitless mode at an injector temperature of 230°C, and GC was operated on a VF-5ms Capillary Column at a constant helium flow of 1 mL min⁻¹. The scan rates were set to 20 s⁻¹, and the mass range was set from mass:charge of 71:500. The raw data were analyzed with the LECO ChromaTof software (3.41), and the data processing includes mass-spectral correction for coeluting metabolites and retention index calculation based on n-alkanes. The metabolites were annotated and defined in a reference chromatogram. Signal peak areas for each metabolite were normalized to the total peak area in the measurement. Peak areas corresponding to each metabolite were normalized to the total peak area in the sample.

Metabolite and Phosphopeptide Data Analysis

ANOVA corrected for multiple testing (Benjamini Hochberg correction) was performed with a significance threshold α = 0.05 to determine statistically significant differences between the treatment and control groups. We used PCA to compare metabolite/phosphoprotein responses dependent on the respective stress supply. We aligned all of the normalized intensities of the phosphopeptides and metabolites of stress-treated samples in one data matrix, in which we combined all of the measurements, including the different fractionations and replicates of phosphoproteomics/metabolomics per time point and treatment (Wienkoop et al., 2008). GO analysis was done using the agriGO GO Analysis Tool Kit with the singular enrichment analysis with the Arabidopsis genome as background (TAIR9), and Fisher’s exact test corrected for multiple testing (Hochberg false discovery rate correction) with a significance threshold of 0.01 was used to determine significantly enriched GO categories.

Enzyme Activity Assays

For enzyme extraction, 20 mg of powdered cells were extracted with 1 mL of extraction buffer (50 mm HEPES/KOH, pH 7.5, 10 mm MgCl₂, 1 mm EDTA, 1 mm EGTA, 1 mm benzamidine, 1 mm ε-aminocapronic acid, 20 μm leupeptin, 500 μm DTT, 1 mm PMSF, 1% [v/v] Triton X-100, 17% [v/v] glycerol, 0.01% [v/v] phosphatase inhibitor mixture 2, and 0.01% [v/v] phosphatase inhibitor mixture 3; Sigma-Aldrich). The supernatant was the raw extract and used for the assays. Enzyme activities were assayed using the method as described before (Gibon et al., 2004). Reactions were started with the addition of a substrate or cofactor and incubated at 30°C for 20 min. Reactions were then stopped using 0.5 m HCl, 0.5 m NaOH, or 80% (v/v) ethanol. The concentrations of the products of these stopped reactions, such as glycerol-3-P, were then determined using cycling assays. The individual stopped assays and the subsequent cycling assays were described previously (Gibon et al., 2004). Accordingly, the cytosolic enzyme has maximum activity at pH 6.7, whereas the isoform is inactive. All experiments were done at 4°C.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Workflow of the reciprocal labeling experiments.

• Supplemental Figure S2. Quantitative analysis of stress-responsive metabolites.

• Supplemental Table S1. Phosphopeptides identified from Arabidopsis with a statistically significant change in abundance in response to salt and oxidative stresses.

• Supplemental Table S2. Quantitative information of all phosphopeptides identified in the stress-treated Arabidopsis cells.

• Supplemental Table S3. Full list of all peptide identifications under different conditions.

• Supplemental Table S4. Quantitative information of metabolite data identified in the stress-treated Arabidopsis cells.

Glossary

ABA

abscisic acid

DTT

dithiothreitol

GC

gas chromatography

GO

gene ontology

H₂O₂

hydrogen peroxide

LC

liquid chromatography

MS

mass spectrometry

PCA

principal component analysis

PMSF

phenylmethylsulfonyl fluoride

ROS

reactive oxygen species

TAIR9

The Arabidopsis Information Resource 9