ROS Homeostasis Regulates Somatic Embryogenesis via the Regulation of Auxin Signaling in Cotton

Somatic embryogenesis (S.E.) is a versatile model for understanding the mechanisms of plant embryogenesis and a useful tool for plant propagation. To decipher the intricate molecular program and potentially to control the parameters affecting the frequency of S.E., a proteomics approach based on two-dimensional gel electrophoresis (2-DE) combined with MALDI-TOF/TOF was used. A total of 149 unique differentially expressed proteins (DEPs) were identified at different stages of cotton S.E. compared with the initial control (0 h explants). The expression profile and functional annotation of these DEPs revealed that S.E. activated stress-related proteins, including several reactive oxygen species (ROS)-scavenging enzymes. Proteins implicated in metabolic, developmental, and reproductive processes were also identified. Further experiments were performed to confirm the role of ROS-scavenging enzymes, suggesting the involvement of ROS homeostasis during S.E. in cotton. Suppressing the expression of specifically identified GhAPX proteins resulted in the inhibition of dedifferentiation. Accelerated redifferentiation was observed in the suppression lines of GhAPXs or GhGSTL3 in parallel with the alteration of endogenous ascorbate metabolism and accumulation of endogenous H2O2 content. Moreover, disrupting endogenous redox homeostasis through the application of high concentrations of DPI, H2O2, BSO, or GSH inhibited the dedifferentiation of cotton explants. Mild oxidation induced through BSO treatment facilitated the transition from embryogenic calluses (ECs) to somatic embryos. Meanwhile, auxin homeostasis was altered through the perturbation of ROS homeostasis by chemical treatments or suppression of ROS-scavenging proteins, along with the activating/suppressing the transcription of genes related to auxin transportation and signaling. These results show that stress responses are activated during S.E. and may regulate the ROS homeostasis by interacting with auxin signaling.

During somatic embryogenesis (S.E.), differentiated somatic cells reentering the cell cycle undergo dedifferentiation and redifferentiation, followed by the formation of embryogenic cells and somatic embryos, and eventually develop into new plants (1). Such developmental restructuring involves the orchestration of various signal networks and reprogramming gene expression patterns to alter the original development program (2,3). The observed totipotency of somatic cells, allowing them to initiate embryogenic development under appropriate conditions, has been exploited to explore regulatory mechanisms and morphogenetic events occurring during the initiation and development of plant embryogenesis and represents a powerful tool for the propagation of plants combined with genetic engineering (1,3,4).
Recent analysis of the genes implicated in S.E. or exhibiting differential expression during S.E. were performed to uncover the molecular events of S.E (5)(6)(7). Although the identification of these genes increases the current understanding of embryogenic processes, the detailed mechanisms of S.E. largely remain unclear. Proteomics has emerged as a powerful tool for the systematic analysis of protein expression during particular biological processes (8). The proteomic analysis of S.E. by comparing embryogenic calluses (ECs) and nonembryogenic calluses (NECs) in different plant species has been carried out, and several differentially expressed proteins (DEPs) have been identified during S.E (9 -11).
S.E. induction is a complex process affected by many factors. Progressively, research has shown that stress factors play an important role during S.E (12,13). Reactive oxygen species (ROS) were recognized as pivotal regulators of plant growth and development (14,15). ROS display a dual function in many developmental processes, which depend on the level and subcellular distribution of these molecules (16). High concentrations of ROS are toxic, leading to oxidative damage. However, at appropriate concentrations, ROS also act as signaling molecules regulating many developmental and physiological responses (15,17). The modulation of ROS levels is involved in the control of cell proliferation, cell death, and senescence, particularly the destruction of subcellular organelles (16,18). The ROS concentration and subcellular distribution in plants is carefully regulated, as imbalances cause redox state disturbances that have crucial effects on the cell fate (16). The steady state of ROS in cells is maintained through ROS-generating enzymes, such as NADPH oxidases and ROS-scavenging enzymes, including superoxide dismutases (SOD), ascorbate peroxidases (APX), catalases (CATs), glutathione peroxidase (GPX), glutathione transferase (GST), and antioxidant molecules, such as glutathione and ascorbic acid (17,18).
ROS-mediated redox signal-regulated development is often associated with hormonal reactions and responses during plant development (19,20). Auxin plays key roles in plant growth and developmental processes and often cross talk with ROS to modulate diverse aspects of plant growth and development (18,21). It has been reported that indole-3acetic acid (IAA) could be metabolized by horseradish peroxidase, and overexpression of such peroxidase in tobacco enhanced defense responses and impaired growth because of increased IAA degradation activity (22,23). During S.E., many auxin-responsive genes are differentially expressed, and auxin-induced S.E. in cotyledons has been associated with oxidative stress and defense gene activation (6,24). ROS generation and auxin are both required for the cell cycle progression from the G0 to the G1 phase (25). In root gravitropism, auxin induces ROS production, which requires activation of phosphatidylinositol 3-kinase, and ROS may act downstream of auxin (26,27). Auxin also regulates the counterbalance of APX1 S-nitrosylation/denitrosylation activity to modulate APX1 activity to fine control of root development and determination of root architecture (28). Besides, ROS homeostasis can be directly modulated by auxin through inducing ROS detoxification enzymes, such as glutathione (GSH)-S-transferases, or indirectly by affecting the stability of DELLA (74) proteins (20). Reports have also shown that several redox-related genes modulate ROS crosstalk through auxin signaling (29 -31). The accumulation of mitochondrial ROS in the abo6 mutant mediated the crosstalk between ABA and auxin signaling (29). Chen et al. (30) revealed that AtAPX6 mediates the crosstalk between ROS, ABA, and auxin to protect desiccating and germinating Arabidopsis seeds from stress. Studies on the rbohD and rbohF double mutant have shown the involvement of ROS in activating Ca 2ϩ signaling and decreasing auxin sensitivity in Arabidopsis roots (32). The disruption of the NADP-linked thioredoxin and glutathione systems in a triple mutant elevated ROS levels and perturbed auxin transport and metabolism (33).
Cotton, as a main source of textile fiber, is one of the most important economic crops worldwide. Therefore, a reproducible and highly efficient regeneration scheme is greatly important for cotton genetic engineering (34). However, the regeneration of cotton species through S.E. was inclined to specific varieties, reflecting a genotype-dependent response (35). Thus, the underlying biochemical and molecular events during cotton S.E. remain an important research area for developmental biology. In a previous study, we identified an elite genotype, exhibiting a higher regeneration frequency than Coker lines (34). Despite the complex regulation of auxin signaling, transcription factors, miRNAs, and stress-related responses at the transcription level were investigated through next-generation sequencing of this genotype during S.E (6,12,36). The proteins and complex mechanisms underlying the development of cotton S.E. remain largely unknown. In the present study, we performed a proteomics analysis and revealed that the differential expression of specific proteins involved in various biological processes was associated with cotton S.E. Among those proteins, several redox-related proteins were identified. The suppression of GhAPXs and GhG-STL3 influenced cotton S.E. Data support the view that ROS homeostasis is crucial for initiating and maintaining dedifferentiation, while mild oxidative conditions promote redifferentiation, and there is an interplay between ROS and auxin homeostasis to modulate S.E. in cotton plants.

EXPERIMENTAL PROCEDURES
Plant Materials and Tissue Culture-The sterilized seeds of YZ1 (Gossypium hirsutum L.) were cultured on 1/2 MS (1/2 strength macro salts plus 15 g glucose, pH 6.0) at 28°C in the dark for 7 days. The hypocotyls were excised from aseptic seedlings, dissected into ϳ8 mm segments, and used as initial explants. The explants were subsequently cultured on MSB medium (MS medium plus B5 vitamins) containing 1.0 mg/l indol-3-ylbutyric acid (IBA) and 0.1 mg/l kinetin. After cultivation for 40 days, all explants were transferred to fresh MSB media for induction of embryogenic callus (EC) in which the medium contained twice the concentration of KNO 3 , with NH 4 NO 3 free, and supplemented with 3% (w/v) glucose, 0.25% (w/v) Phytagel, 0.5 mg/l IBA, 0.15 mg/l kinetin, 1.0 g/l glutamine, and 0.5 g/l asparagine for embryo induction and maturation as previously described (6). Cultures were maintained at 28 Ϯ 2°C under a 14-h photoperiod (irradiance of 135 mol/ms). As previously described (6), different time points/stages of explants (0 h, 2 d, 40 d, defined as nonembryogenic calluses, NECs), ECs and somatic embryos (globular embryos, GEs; torpedo embryos, TEs; and cotyledon embryos, CEs) were sampled and frozen at Ϫ70°C until required for further analysis.
To generate GhAPXs and GhGSTL3 suppression lines, gene-specific primers with attB1 and attB2 adaptors (listed in Table S1) were used, and the PCR products, respectively, cloned into pHellsgate4. RNAi vectors were introduced into G. hirsutum YZ1 plants by Agrobacterium tumefaciens using strains LBA4404. Two representative suppression lines for each of the genes GhAPXs and GhGSTL3 were selected for further experiments. The culture conditions for the transgenic plants were conducted as described above.
Two-dimensional gel electrophoresis (2-DE) was performed according to the manufacturer's instructions (Bio-Rad). Total proteins (1.0 mg) from each sample were individually loaded onto immobilized pH gradient strips (17 cm, pH 4 -7 nonlinear, Bio-Rad) with 300 l rehydration buffer (7 M urea, 2 M thiourea, 4% CHAPS, 1% DTT, and 2% v/v immobilized pH gradient buffer, pH 4 -7). The strips were rehydrated for 12 h at room temperature. Isoelectric focusing was performed using the following protocol: 50 V for 1 h, 500 V for 1 h, 1000 V for 1 h, and 10,000 V for 4 h, with a final step of 10,000 V for a total of 90 kVh. Each focused strip was equilibrated for 15 min with equilibration buffer (50 mM Tris-HCl, pH 8.8; 6 M urea; 30% glycerol; and 2% SDS; and 1% DTT), followed by a second equilibration step for an additional 15 min with equilibration buffer in which DTT was substituted with 2% (w/v) iodoacetamide. After equilibration, the immobilized pH gradient strips were fixed on top of vertical 12% acrylamide gels. Electrophoresis was performed using the Bio-Rad system (protean II XL) at 15 mA/gel for 1 h, followed by 45 mA/gel for 6 h until the bromphenol blue dye front reached the bottom of the gel.
The 2-D gels were stained with Coomassie Brilliant Blue solution. The stained gels were scanned using a GS-800 Calibrated Densitometer (Bio-Rad), and the protein spots were calculated using PDQuest software (Bio-Rad). After volumetric quantification and matching, spots of various intensities at different time points/stages of S.E. were analyzed using the Student's t test and calculated as a fold ratio with a threshold of p Յ 0.05 and fold-change of Ն 2 or Յ 0.5. Three biological repeats were performed for each 2-DE image to find steadily repeatable DEP spots.
DEPs were excised from the gels, followed by sequential treatments, including destaining with 25 mM NH 4 HCO 3 in 50% Acetonitrile (ACN) until the coomassie brilliant blue (CBB) disappeared, digestion with trypsin (Promega, Madison, WI) overnight at 37°C, and extraction with extraction buffer (67% ACN and 5% TFA), as indicated by our previous study (37). The treated spots were then analyzed using an ABI 5800 MALDI-TOF/TOF Plus mass spectrometer (Applied Biosystems, Foster City, CA). Both the MS and MS/MS data were integrated using GPS Explorer V3.6 software (Applied Biosystems). Successfully identified proteins were characterized with a 95% or higher confidence interval using the MASCOT V2.3 search engine (Matrix Science, London, UK). The Gossypium expressed sequence tag (EST) database (release data 20120128; 2,476,590 sequences; 555,009,942 residues) was used for searching. The other search parameters were as our previous study (37), including the enzyme trypsin; partial modifications of cysteine carbamido methylation and methionine oxidization; fixed and variable modifications; one missed cleavage site; peptide tolerance of 100 ppm; and fragment mass tolerance of 0.5 Da.
Redox Disturbance and Auxin Effect on S.E. Process-To determine the effect of redox homeostasis on S.E., explants were cultured on MSB medium supplemented with diphenyleneiodium (DPI) at various concentrations (0, 1, 2, and 2.5 M), or H 2 O 2 (0, 25, 50 M, 0.1, 0.5, and 1 mM), or 2 M DPI ϩ H 2 O 2 (25 M, 0.1 mM, and 0.5 mM). Explants were also cultured on MSB medium containing 0.1 mM buthionine sulfoximine (BSO), or 0.1 mM reduced glutathione (GSH), or 0.1 mM BSO ϩ 0.1 mM GSH to investigate the effect of redox disturbance on dedifferentiation process. Besides, explants were also cultured on MSB medium containing ASA 1 with different concentrations (0, 50, 100 M) to detect whether ASA could complement the effect of GhAPXs suppression. The fresh weights of the explants were recorded at different time points/stages during S.E., and the proliferation rate was indicated as the increased fresh weight of the explants per gram. To investigate the effect of redox disturbance on the redifferentiation process, homogenized ECs were also treated with BSO and GSH, both at concentrations of 0.1 mM for 20 d, and the number of somatic embryos was counted.
To  (38). Samples were ground into powder in 1.6 ml precooled acetone, followed by shaking for 20 min and centrifuging at 4°C (15,871 g for 15 min). The supernatant was added to a fresh centrifuge tube for further detection and H 2 O 2 determination was performed according to the manufacturer's instructions.
Auxin and Auxin Metabolites Extraction-Approximately 0.1-0.2 g of tissue from selected stages of S.E. were sampled and frozen at Ϫ70°C until further analysis. The measurement of endogenous free IAA and IAA metabolites were performed as previously described (40), with some modifications. Approximately 200 mg of each sample were ground to a fine powder in liquid nitrogen, followed by extraction with 800 l precooled 80% methanol solution containing 1% acetic acid, vigorous shaking in the dark overnight at 4°C, and centrifugation at 15,871 g at 4°C for 15 min. The supernatant was carefully collected, and the pellet was resuspended with additional 0.4 ml extraction buffer, extracted for 4 h at 4°C, followed by centrifugation. The supernatants were combined and loaded onto an HLB column (Waters) and washed with 70% methanol solution containing 2% acetic acid. The filtrates were dried through evaporation under the flow of nitrogen gas and dissolved in 60 l 10% methanol.
RNA Extraction, RT-PCR and qRT-PCR-Total RNA was isolated from selected samples as previously described (12). About 2 g RNA were used for reverse transcription. The first-strand cDNAs were synthesized with SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA) and used as templates. qRT-PCR was performed using the ABI Prism 7000 system (Applied Biosystems, Foster City, CA), and the expression levels of selected genes were normalized to GhUB7 (GenBank accession number: DQ116441) using the 2 -⌬Ct calculated method as previously described (41). qRT-PCR was conducted with three biological replicates and three technical replicates for each biological replicate. The primers for qRT-PCR and RT-PCR are listed in Tables S1 and S2, respectively. All primers were designed using Primer Premier software version 5.0.
Statistical Analysis-All graphical data were generated from three biological replicates, and the values are presented as the means Ϯ S.D. Statistical significance was determined using one-way ANOVA analysis, and p values Ͻ 0.05 were considered statistically significant.

Proteomics Analysis and Identification of DEPs During Cotton S.E-S.E. initiation and development in cotton plants is followed by sequential morphological and dynamic changes.
No visible morphological changes were observed on the explants cultured for 2 d (Fig. 1A), and cell expansion was observed through histological observation as previously de- scribed (6). After culturing for 7 d, both ends of the explants expanded and formed callus. Subsequently, mass calluses proliferated and developed into ECs, followed by differentiation into GEs, TEs, and CEs (Fig. 1A).
Based on the morphological events, the samples from 0 h, 2 d, NEC, EC, GE, TE, and CE were collected for comparative proteomics. The numbers of independent protein spots detected varied for different developmental stages, ranging from 838 to 1360 spots observed on 2-D gels. A total of 155 significantly expressed DEPs were successfully identified during S.E. using MALDI-TOF/TOF and the parameters described above (F Ն 2 or Յ 0.5 and p Յ 0.05) (Fig. 1B, Fig. S1). After removing the redundant proteins, 149 unique proteins were identified (Fig. 1C). These identified DEPs were categorized according to functional categories as indicated by molecular function (Table S3). A large proportion of these proteins have binding ability, such as general regulatory factors (14-3-3 proteins), glutamine synthetase, ATP synthase, and translation elongation factors. Besides, a large amount of the DEPs possess oxidoreductase activity, including several ROS-related proteins as well as HSP70 proteins and ribulose-bisphosphate carboxylases. DEPs have transferase activity, such as glutathione S-transferase family proteins and sedoheptulosebisphosphatase, were also characterized. And DEPs confer hydrolase activity and isomerase activity and several unknown proteins were also identified (Table S3). Further detailed information for these proteins was annotated and described in Table S3.
Spatial analysis was conducted on DEPs to confirm the degree of overlap during dedifferentiation, the transition from NEC to EC, and somatic embryo development during cotton S.E. as previously described (6). Approximately 104, 91, and 138 differentially expressed proteins were identified, representing the three different developmental stages (Fig. 1C). Among these, 37.6% (56) of DEPs were detected during all three developmental stages, of which 2.01% (3) were modulated during dedifferentiation, 0.67% (1) during the transition from NEC to EC, and 11.4% (17) during the somatic embryo development (Fig. 1C). Among the 149 identified proteins, 18.8% (28) (48) down-regulated at CE (Fig. 2A).
Gene ontology analysis was performed using Blast2GO. Under the level 2 biological process category, cellular and metabolic processes were most commonly represented, accounting for nearly one-quarter of all proteins (Fig. 2B), followed by the "single-organism process" (16.6%) and "response to stimulus" (14.2%). Other important biological processes, such as development (4.3%), reproduction (2.6%), and growth (0.3%) were also identified (Fig. 2B). Based on the molecular function category at level 3, a number of the pro-teins were assigned as having binding properties and a category of proteins possessing oxidoreductase activity were also well represented (Fig. 2C).
Plants possess a series of ROS scavenging enzymes for protection against ROS damage (15), and it is interesting to note that protein levels did not always correlate with mRNA abundance (42). As shown in Fig. 3, the mRNA expression levels of the three cytoplasmic APXs (SSP7332, 6304, 5316) were consistent with the protein levels, while the transcript level of stromal ascorbate peroxidase (SSP6417) showed an opposite expression pattern compared with the protein level (Fig. 3). The transcript levels for three identified GST proteins were also consistent with the expression patterns at the protein level (Fig. 3), and the transcript levels of the identified SOD enzymes (SSP6222, 7228, 8004) were consistent with the protein expression profiles observed during S.E. (Fig. 3). Similarly, the transcript profile of thioredoxin (SSP4540) was also consistent with the protein profile, but the transcript level of a second thioredoxin (SSP2028) showed an opposite protein expression pattern (Fig. 3). The results indicated that these ROS-related proteins might be involved in cotton S.E.
Except for ROS-related proteins, the expressions of DEPs involved in amino acid metabolisms, such as diaminopimelate epimerase (SSP3401), glutamine synthetase 1 (SSP4546), spermidine synthase 2 (SSP1521), arginosuccinate synthase (SSP6609), 3-isopropylmalate dehydratase (SSP3212), ketolacid reductoisomerase (SSP4760), and S-adenosylmethionine synthetase (SSP5644, SSP5414), were detected in the transcript level and the protein level. However, the expression profiles of these genes on transcript level were not correlated with the protein level (Fig. S2). The expression profiles of aldolase-type TIM barrel family protein (SSP1620, SSP1624, SSP1630) and pfkB-like carbohydrate kinase family protein (SSP2421, SSP2436, SSP3406), which assigned to the pentose phosphate pathway, were also investigated in both transcript level and protein level. The mRNA expression levels of genes involved in the pentose phosphate pathway were also not consistent with the protein levels, except for SSP2436 and SSP3406, whose expression profiles were consistent in transcript level and protein level (Fig. S2).
Among the DEPs, proteins involved in flavonoid biosynthesis such as chalcone synthase (SSP7619) and chalcone isomerase (SSP1226) were also characterized (Fig. S3A). The mRNA expression levels of chalcone synthase (CHS) (SSP7619) and chalcone isomerase (CHI) (SSP1226) were consistent with the protein levels (Fig. S3B). As well, the transcript profiles of other main genes involved in flavonoid biosynthesis were detected. The expression patterns of all these flavonoid genes were nearly similar during S.E. processes (Fig. S3B). The endogenous contents of flavonoids during SE processes were also consistent with the genes expression profile. The endogenous flavonoids were abundantly accumulated in 2 d but were less during subsequently time points/stages (Fig. S3C).
Defective Cytoplasmic GhAPXs-Mediated ROS Accumulation Regulates Cotton S.E-APXs play pivotal roles in maintaining intracellular ROS homeostasis (30). Considering the differential expression of APXs during S.E. (Fig. 4A), a disruption of these GhAPXs was conducted in cotton to clarify the function of ROS scavenging enzymes on S.E. Two representative interference lines (Ri167 and Ri24) were obtained. The expression of the three identified cytoplasmic GhAPXs (SSP7332, 6304, 5316) was down-regulated in the two interference lines (Fig. 4B), and S.E. initiation and dedifferentiation were significantly inhibited in these lines compared with wild type (Fig. 4C). The cell proliferation rates of interference lines were sharply decreased during dedifferentiation (Fig. 4D), accompanied by the higher endogenous H 2 O 2 content in the down-regulated lines than in wild-type plants during S.E. (Fig.  4E). The contents of reduced ASA were elevated in GhAPXs interference lines as compared with wild type during the dedifferentiation process. As well, the contents of total ASA (ASAϩDHA) were altered in GhAPXs suppression lines (Fig.  4F). The effect of exogenous application of ASA on dedifferentiation process was also detected in both GhAPXs interference lines and wild type. There were no significant differences between ASA treatment and control on the cell proliferation rate of both GhAPXs interference lines and wild type after culturing for 15 d (Fig. 4G). As culturing for 40 d, the cell proliferation rate was retarded by ASA treatment both in GhAPXs interference lines and wild type (Fig. 4G), which were consistent with the disturbance of ASA homeostasis in GhAPXs interference lines. Furthermore, ECs and somatic embryos were observed earlier in GhAPXs interference lines than in wild type (Fig. 4H). After culturing for 120 d, the differentiation rates were 81.9% and 59.2% in down-regulated lines compared with 34.4% in wild-type controls (Fig.  4I). These results suggested that GhAPXs play important roles during S.E., and the alteration of S.E. process through APX suppression might be associated with the elevated endogenous H 2 O 2 content.
Suppressing the Expression of GhGSTL3 Promotes the Redifferentiation during Cotton S.E-Other than APXs, several GST-related proteins were also identified as differentially abundant throughout embryogenesis (Fig. 5A, Table S3). In the present study, GhGSTL3 (SSP2324) interference was generated in cotton plants and two representative lines were selected (Fig. 5B). Dedifferentiation was marginally altered in GhGSTL3 interference lines relative to wild-type plants. There were no apparent distinction between interference lines and wild-type plants at the preliminary stage of dedifferentiation. However, callus production was greater in interference lines than in wild-type tissues cultured for 40 d (Fig. 5C). As the culture period progressed, the transition from NEC to EC was obviously accelerated, and somatic embryos were observed in GhGSTL3 suppression lines compared with wild-type plants (Fig. 5D). The differentiation rates were, respectively, 40.8% and 38.8% in interference lines and 18.5% in wild-type plants, and 75 and 55% in interference lines but only 22.3% in wild-type plants after culturing for 60 d and 100 d, respectively (Fig. 5E). The endogenous H 2 O 2 contents were significantly higher in interference lines compared with wild type (Fig. 5F).
ROS Homeostasis Is Crucial for Cotton S.E-To further investigate whether ROS signaling is involved in cotton S.E., explants were cultured on medium supplemented with different concentrations of DPI, a widely used inhibitor of NADPH oxidase, a ROS-generating enzyme (43). The dedifferentiation process was significantly retarded with DPI treatment, more severe as the concentration increased (Figs. 6A and 6B and Fig. S4A), indicating that ROS was necessary for dedifferentiation during cotton S.E. Similar results were obtained after H 2 O 2 treatments (Figs. 6A and 6B). The initiation and dedifferentiation process in cotton plants during S.E. was seriously

FIG. 3. Comparison of the expression profiles between protein and transcript levels for several redox-related proteins.
The quantification of the protein intensities and corresponding relative transcript levels are represented as bar and line charts, respectively. qRT-PCR was performed to determine the transcript levels of the selected proteins. The expression levels were normalized to GhUB7, and the relative expression was calculated as 2 -⌬CT , ⌬CT ϭ (C T, Target -C T, GhUB7 ) Time x . The error bars represent Ϯ S.D. of three biological replicates. APX, four DEP spots (SSP7332, 6304, 5316, 6417) were identified as ascorbate peroxidases; GSTL3, DEP spot 2324 was identified as glutathione transferase L3; GSTF9, DEP spot 5225 was identified as glutathione transferase Phi 9; GSTU5, DEP spot 5210 was identified as glutathione transferase Tau 5; MSD1, two specifically identified protein spots 6222 and 7228 were identified as MnSOD; Cu/Zn SOD, DEP spot 8004 was identified as Cu/Zn SOD; Thioredoxin, two DEP spots 4540 and 2028 were identified as thioredoxins. GSH and BSO (specific inhibitors of GSH biosynthesis) were used to mimic ROS perturbation in vivo. Explants were cultured on medium containing 0.1 mM BSO (Figs. 6C and 6D). Consistent with H 2 O 2 treatments, the initiation and dedifferentiation of explants cultured on medium containing BSO were nearly inhibited (Figs. 6C and 6D). With GSH treatment, the dedifferentiation process was slightly retarded (Figs. 6C and 6D). To modify the effect of BSO on cotton S.E., GSH was applied to medium containing 0.1 mM BSO. The negative effect of BSO on S.E. was partially restored by GSH (Figs. 6C and 6D). These results suggest that ROS is a key factor for dedifferentiation and ROS homeostasis is critical for initiating dedifferentiation.
Endogenous levels of H 2 O 2 in BSO-or GSH-treated samples were determined during cotton S.E. There were no significant differences found between treatments and control conditions at the preliminary stage of cotton S.E.; however, the H 2 O 2 content was significantly higher in BSO-treated samples compared with samples cultured under normal conditions during the late stages of dedifferentiation (Fig. 6E). The H 2 O 2 level in explants cultured on medium containing GSH was slightly elevated in 2 d cultures. Thereafter, no significant changes were observed until 40 d after culturing (Fig. 6E), consistent with the phenotype observed. The increased H 2 O 2 content in response to BSO treatment was reduced after applying 0.1 mM GSH to medium containing 0.1 mM BSO (Fig. 6E).
The ROS accumulation during the development of cotton S.E. was also investigated under both control and treated conditions, using 2Ј, 7Ј-dichlorofluorescein diacetate staining. After culturing for 2 d, the fluorescence intensity was significantly elevated under both BSO and GSH treatment conditions compared with control (Fig. 6F). During S.E., BSO treatment induced stronger fluorescence intensity compared with controls, while no significant differences were observed between GSH treatment and the controls (Fig. S5). These results indicated that inhibition of GSH activity by BSO treatment altered the redox state and elevated ROS accumulation in vivo.
The differentiation process was also examined by BSO and GSH treatments. With embryogenic callus cultured on medium containing 0.1 mM BSO for 20 d, the number of total somatic embryos was higher compared with the control (Figs. 6G and 6H). No significant difference was observed in the number of globular embryos following BSO treatment compared with controls, but significantly more torpedo and coty- ledon embryos were observed with BSO treatment (Figs. 6G  and 6H). For GSH treatment, no significant differences in the number of total somatic embryos, and the different development stages of somatic embryos were observed compared with control (Figs. 6G and 6H). The accelerated differentiation through BSO application was modified through additional GSH supplement (Figs. 6G and 6H). These results suggest that an oxidative environment promotes differentiation and accelerates somatic embryo development. treated cultures compared with untreated controls after 2 d and 7 d culturing (Fig. 7A). After culturing for 15 d, the content of free IAA was decreased in DPI-and H 2 O 2 -treated cultures, but there was no significant differences found in the content of IAA-Asp, IAA-Val, and oxIAA between the treatments and control at that stage (Fig. 7A). Free IAA and IAA metabolite contents were also examined in GhAPXs-defective transgenic lines during S.E. The metabolic profile and free IAA content were comparable between interference lines and wild-type plants during S.E., with a slight accumulation for explants at 40 d. No significant differences in oxIAA content were observed between GhAPXs-defective transgenic plants and wild-type plants (Fig.  7B). However, the IAA-Asp content was lower in suppression lines when cultured for 7 d but higher in suppression lines when cultured for 40 d compared with wild-type plants. There were no obvious differences in IAA-Val content with wild-type plants at different time points/stages, but the IAA-Val content was continuously elevated in suppression lines (Fig. 7B), indicating the metabolism patterns of IAA-Asp and IAA-Val was dependent on GhAPX expression during S.E.

ROS Perturbation Negatively Impacts Auxin Homeostasis during Cotton S.E-Transcriptional analysis revealed that ROS can influence plant development by modulating auxindependent gene expression in Arabidopsis
Transcription analysis was carried out to elucidate the effect of the disruption of ROS homeostasis on the expression of genes involved in auxin transport and signaling (AUX/IAAs, SAUR-like genes, GhABCB19, GhLAX1, GhPILS2, GhAUX1, and GhPIN1). Expression levels of these genes increased after both H 2 O 2 and DPI treatment (Fig. 7C). The transcription levels of several auxin transport and response genes were also altered in GhAPXs defective transgenic lines compared with wild-type plants at different time points/stages during cotton S.E. At 0 h, no obvious differences were observed between down-regulated lines and wild-type plants. However, the expression of these genes was differentially altered during S.E. development. By 7 d of culture, expression of GhPILS2, GhAGG1, and GhIAA19 was relatively increased in suppression lines, though the other genes showed no obvious changes (Fig. 7D). By 40 d of culture, expression of the auxin transport genes GhABCB19 and GhLAX1 was increased in the suppression lines compared with wild-type tissues, while expressions of the auxin response genes GhAGG1, GhIAA26, and GhDRM1 were reduced in down-regulated lines compared with wild-type plants, while GhIAA19 expression was relatively increased (Fig. 7D). These results revealed that auxin signaling and transport were altered in GhAPXs interference lines.
Consistent with S.E. inhibition through DPI or H 2 O 2 treatment, the recovery of growth could be observed by application of auxin to the medium containing 2 M DPI (Fig. 7E). A similar restoration was also observed with the addition of auxin to medium supplemented with H 2 O 2 (Fig. 7F). The degree of restoration was correlated with the auxin concentration, increasing as the level of auxin application increased under both DPI and H 2 O 2 treatment (Fig. 7G). DISCUSSION Early plant development is characterized by changes in the levels of key cellular proteins, as revealed by proteomics studies (8,42). S.E. is considered as both an efficient pathway for plant propagation and a feasible model for investigating the early regulatory and morphogenetic events in plant embryogenesis (1,3). Transcriptome analysis of cotton S.E. have been previously described (6,12). Given the potentially variable correlation between mRNA and protein levels for individual genes (42), proteomics studies are essential to understand the regulatory mechanisms underlying cotton S.E. During dedifferentiation and redifferentiation processes of cotton S.E., 5076 differential expressed genes (DEGs) were identified through transcriptome analysis (6). There were 67 common DEGs identified both by proteomics and transcriptome analysis (Fig. S6A). Gene Ontology (GO) analysis indicated that the commonly identified genes were mostly involved in apoplast, ROS metabolic process, glycolytic/gluconeogenesis process, and hexose biosynthetic process (Fig. S6B). Notably, several of these common genes were also participated in glutathione metabolism as indicated by Kyoto Encyclopedia of Genes and Genomes analysis (Fig. S6C). Both the two approaches clarified that ROS-related genes were differentially expressed during cotton S.E. process, indicating that ROS might be involved in cotton S.E. process.
Ascorbate peroxidases (APXs) play important roles in ROS scavenging and are involved in many development programs (30,44). Changes in endogenous ascorbate redox status have an effect on somatic embryo development and an oxidized environment benefit somatic embryo maturation (45)(46)(47). In the present study, several differentially expressed GhAPXs were identified (Fig. 3, Fig. 4A). The transgenic suppression of three identified APX proteins (SSP7332, 6304, 5316) changed endogenous ascorbate metabolism and induced endogenous H 2 O 2 accumulation (Figs. 4E, 4F). The dedifferentiation process on culturing tissues for S.E. induction was retarded, while redifferentiation was promoted in GhAPXs suppression lines compared with wild-type tissues (Fig. 4). This suggests the involvement of ROS-scavenging enzymes in cotton S.E. Plant GSTs also act as ROS-scavenging enzymes associated with responses to stresses, hormone signaling, and developmental changes (48 -50). The suppression of GhGSTL3, which was identified as being differentially expressed during S.E., had no significant effect on dedifferentiation, but redifferentiation was accelerated, and increased H 2 O 2 levels were observed compared with wild-type tissues (Fig. 5). The elevated growth rate of GhGSTL3 suppression lines during the late stage of dedifferentiation might reflect the accelerated redifferentiation in GhGSTL3 suppression lines (Figs. 5D and 5E), as EC and NEC showed contrasting density (34).
The changing abundances of ROS-scavenging enzymes during S.E. suggests that ROS signaling might be important for regulating S.E. development. Dedifferentiation and cell proliferation were consistently retarded following treatment with DPI (Fig. 6A, Fig. S4A), which is an inhibitor of flavoenzymes, particularly NAD(P)H oxidase. NADPH oxidase plays a vital role in root development through the generation of ROS, which regulates cell expansion through the activation of Ca 2ϩ channels (51). The inhibition of ROS production by blocking the activity of NADPH oxidase using DPI can phenocopy the rhd2 mutant defective in NADPH oxidase (51). Localized ROS production is detectable at the pollen tube apex, and the inhibition of pollen tube elongation was observed after treatment with DPI or the suppression of endogenous NADPH oxidase expression (52). This suggests that NADPH oxidasegenerated ROS and ROS signaling are critical for various plant developmental processes. The application of H 2 O 2 to cotton tissues also treated with DPI was able to partially rescue this ROS inhibition and cell proliferation (Fig. 6A), further confirming that ROS is necessary for the initiation and development of tissue dedifferentiation.
Redox regulation is an elaborate mechanism utilized by plants to perceive and respond to perturbations in the ROS concentration (16). ROS can serve as signaling molecules involved in diverse metabolic processes, but at high concentrations, ROS can induce cell cycle arrest and apoptosis (15,25). Also, ROS could cause oxidative modifications of proteins and affect their functionality, arising from modification of a wide range of amino acids (53)(54)(55). Protein carbonylation was regarded as a marker of protein oxidation and such posttranslational modification can result in the loss of function of target proteins (53,56). Protein carbonylation has shown to be target specific and targets enzymes of photosynthesis and energy, as well as amino acid metabolism and other proteins, including stress proteins, HSP70 chaperones, and translation elongation factors (54,56). Protein carbonylation plays roles in metabolic control and acts as a signal in physiological transitions in plants (56). The dose-dependent retardation of dedifferentiation in explants cultured on medium containing different levels of H 2 O 2 revealed that excess ROS negatively impacts dedifferentiation (Fig. 6, Fig. S4). The application of BSO effectively decreased the biosynthesis of GSH, a key antioxidant in ROS scavenging and redox homeostasis maintenance, leading to a disturbance of the endogenous redox balance (57). Consistent with the negative effect of excess H 2 O 2 on dedifferentiation, the initiation and development of dedifferentiation were largely blocked by BSO treatment (Fig.  6C). Although the H 2 O 2 content in BSO-treated explants was comparable with that in control plants during the early stage of dedifferentiation (Fig. 6E), endogenous ROS levels became elevated during BSO treatment, as indicated through 2Ј, 7Јdichlorofluorescein diacetate staining (Fig. 6F, Fig. S5). In wheat, the inhibition of callus regeneration and overall efficiency of transformation were also observed after BSO treatment or the silencing of either of the GSH biosynthesis genes GSH1 and GSH2 (58). Our results suggested that redox status plays important roles in regulating dedifferention. As well, several homologues of the targets for posttranslational modifications were characterized in our identified DEPs, such as HSP70, ribulose-bisphosphate carboxylases, rubisco activase, ATP synthase, 14 -3-3 proteins, glutamine synthetase, glutathione S-transferases (GSTs), and sedoheptulose-bisphosphatase (Table S3). Therefore, we speculated that posttranslational modifications of target proteins might be involved in S.E. metabolic control.
Consistent with these observations, the addition of GSH to BSO-containing medium could partially reverse the inhibitory effect of BSO on S.E. (Figs. 6C and 6D). Unexpectedly, the application of GSH slightly increased the H 2 O 2 content during dedifferentiation and slightly suppressed the dedifferentiation process (Figs. 6C and 6D). It has been reported that GSH is recruited into the nucleus during cell proliferation, and the sequestration of GSH in the nucleus is accompanied by H 2 O 2 accumulation (59). Thus, we speculate that the exogenous application of GSH might alter the endogenous GSH pool in the cytoplasm and nucleus, thereby influencing the normal cell proliferation cycle.
The roles of ROS as important regulators in plant development are potentially as diverse as hormonal signal transduction (14,15). An oxidative environment plays a positive role in somatic embryo development in plants through enhanced cell division (18). Consistent with this, the transgenic suppression of GhAPXs increased the endogenous H 2 O 2 content and accelerated the redifferentiation process (Figs. 4E-4G). Similar results were observed through the suppression of GhGSTL3 (Figs. 5D-5F). The application of BSO has also been found to increase white spruce somatic embryo yield and quality, and these effects could also be phenocopied through alterations in the glutathione redox state through the experimental manipulation of endogenous reduced (GSH) and oxidized (GSSG) glutathione levels (57,60). Similarly, BSO treatment positively affected cotton somatic embryo formation and development (Figs. 6G and 6H).
Stress responses were involved in S.E. process as demonstrated by previous work (12). N compounds, particularly glutamine, were important for the proliferation and maturation of somatic embryos in various species (61,62). Enhanced cell proliferation and inhibited differentiation were observed as ammonium supply was blocked, which were reversed by resupply of glutamine as the form of nitrogen source, because nitrogen supply intrigued oxidative stress response (61). Therefore, a balanced nitrogen supply and metabolism might be critical for plant embryogenesis. Glutamine synthetases play specific roles in nitrogen metabolism during embryogenesis (63). In our work, two glutamine synthetase proteins were identified, which were involved in amino acid metabolism. We speculated that amino acid metabolism might be involved in S.E. process through modulating nitrogen metabolism, which correlated with oxidative stress response. Glutamine synthetase was a target for carbonylation; we also wonder whether enzymes in amino acid metabolism were referred to post-translational modifications and then to play roles in metabolic control.
Proline synthesis from glutamate plays an important role in modulating plant cellular redox potential under stress conditions (64). Enhanced proline synthesis was considered to improve oxidative pentose phosphate pathway activity. Oxidative pentose phosphate pathway activity activation is traditionally correlated with high rates of cell division and differentiation (64). The activation of proline-linked pentose phosphate pathway plays a positive role in embryo formation that might be linked with endogenous cytokinin, auxin, and phenolic biosynthesis required for somatic embryogenesis (65). As we indicated, redox signaling was involved in SE process and proteins participating in the pentose phosphate pathway were identified during SE process. There might be a similar mechanism involved in cotton somatic embryogenesis. Carbonylation of sedoheptulose-bisphosphatase leads to enzyme inactivation, thus inhibiting carbon assimilation efficiency, resulting in growth and development retardation (66). Sedoheptulose-bisphosphatase was a critical enzyme in the pentose phosphate pathway, and differentially identified in our work, but its links to S.E. needs further experiments.
Flavonoids are polyphenolic compounds and have great potential to inhibit the generation of ROS and scavenge excess ROS once they are formed (67). Flavonoids are also natural inhibitors of auxin transport (68). Silencing the flavonoid pathway inhibits root nodule formation in Medicago truncatula due to blocked auxin transport (69). Flavonoids modulate the activity of auxin-transporting p-glycoproteins and likely modulate regulatory proteins activity (68). Flavonoid compounds participate in auxin transport arising from scavenging of ROS (68). Both the expression of genes in flavonoid biosynthesis and endogenous flavonoid compounds were significantly accumulated at the initiation stage of SE and might play roles in scavenging ROS and modulate auxin transport.
There is now good evidence to show that ROS signaling interacts with hormone networks to integrate extrinsic signals into developmental programs and stress tolerance responses (20). Auxin is important for both zygotic embryogenesis and S.E. induction, and development and auxin responses are concentration dependent (6,21). Optimum endogenous auxin levels should be rigorously controlled through auxin homeostasis-maintaining mechanisms, including biosynthesis, conjugate formation, degradation, and transport (21). APX6 protects Arabidopsis seeds to properly execute the germination program through the modulation of ROS crosstalk with hormonal signals (30). In our GhAPXs suppression lines, auxin homeostasis was altered during S.E. compared with wild type (Fig. 7), suggesting that GhAPXmediated ROS signaling interacts with auxin signaling in this developmental process.
Reduced catalase activity results in the accumulation of H 2 O 2 and attenuation of auxin levels at high light intensities (70). Mutation of the mitochondrial FtSH4 gene significantly elevated H 2 O 2 content and decreased the concentration of IAA and expression of auxin-related genes (71). These examples are consistent with attenuated levels of free IAA and IAA metabolites observed after H 2 O 2 treatment (Fig. 7A). Localized auxin accumulation could cause ROS generation and increase oxIAA formation by IAA oxidase to attenuate auxin signaling (72). However, the content of oxIAA was dynamically altered during the dedifferentiation process by H 2 O 2 treatment (Fig. 7A). It might be a complicated feedback regulation of ROS and auxin during the S.E. process.
The expression of some auxin-related genes, including those involved in auxin metabolism, auxin signaling and auxin transport, was elevated after H 2 O 2 treatment (Fig. 7C), a result that is not consistent with the negative effects of ROS on the expression of at least some auxin-related genes (71). A detailed analysis of the transcriptional regulation of the auxin signaling pathway through apoplastic ROS revealed that ROS transiently decreases the expression of auxin-related genes (31). The transcript levels of GH3 genes, including some AUX/ IAA and ARF genes, was both up-and down-regulated through ROS, with no consistent trend, and the expression of the auxin-responsive marker gene HAT2 was differentially altered during O 3 treatment (31). Furthermore, the expression of several AUX/IAA and SAUR genes and the auxin efflux carrier PIN3 was regulated through ROS or auxin in an inverse pattern (31), indicating the dynamic and intricate regulation mechanisms of ROS on auxin-related genes.
In agreement with this, the expression of several auxinrelated genes was differentially altered in GhAPXs suppression lines (Fig. 7E). Expression levels of auxin homeostasisrelated genes in GhAPXs suppression lines were different from those observed with H 2 O 2 treatment, likely reflecting the different roles of specific GhAPXs. It has been reported that the auxin-regulated counterbalance of APX1 with S-nitrosylation/denitrosylation to regulate APX1 activity also regulates root development in Arabidopsis (28). The level of auxin homeostasis has been found to increase in APX6-deficient Arabidopsis seeds, further suggesting ROS signal crosstalk with hormone signaling (30).
We found that auxin homeostasis was also perturbed during S.E. after DPI treatment (Fig. 7A). AtrbohD and AtrbohF have previously been shown to negatively regulate the auxin response in the root tip in the presence of ABA (32). Knocking down the expression of rbohB has shown to decrease ROS production and elevated the expression of some auxin signaling-related genes, suggesting that the generation of ROS through rboh genes impacts root development via auxin signaling (73). Consistent with this, the expression of auxin response genes, such as GhAUX/IAA and GhSAUR, and the auxin polar transport genes GhABCB19, GhLAX1, GhPILS2, GhAUX1, and GhPIN1, was increased after DPI treatment compared with normal control tissues after culture for 15 d (Fig. 7C).
Interestingly, the application of auxin partially reversed the inhibitory effects of DPI and H 2 O 2 (Figs. 7E and 7F), and the perturbation of dedifferentiation through 2,3,5-triiodobenzoic acid was partially recovered by application of moderate H 2 O 2 concentrations during late-stage dedifferentiation (Fig. S7). These results further indicate an interaction between ROS and auxin in cotton S.E. In general, for the SE process correlated with oxidative stress, moderate ROS was necessary for dedifferentiation, while excess ROS inhibited the dedifferentiation process, indicating ROS homeostasis was important for dedifferentiation, which was maintained by the activity of ROS-related proteins, such as APX, GST, SOD, and thioredoxin. ROS generation inhibited or enhanced both attenuated free IAA content and altered auxin metabolism, which then perturbed the dedifferentiation process. ROS homeostasis was dynamically integrated with auxin homeostasis to regulate dedifferentiation process, while enhanced ROS production promoted redifferentiation, which might be due to the attenuated free IAA (Fig. 8). On the other hand, flavonoid biosynthesis might be coupled with oxidative stress to modulate ROS homeostasis and function in auxin transport inhibition to regulate dedifferentiation. Oxidative stress might also activate the pentose phosphate pathway, which then stimulates redifferentiation. Amino acid metabolism might be involved in the SE process through modulating nitrogen metabolism, which connects with oxidative stress response, to regulate the SE process. Several enzymes involved in the pentose phosphate pathway and amino acid metabolism as well as stress-related proteins are targets for posttranslational modifications; therefore, the posttranslational modification mechanism might be involved in metabolic control of S.E. process (Fig. 8). Further and deeper studies are needed to confirm whether the complex interaction between ROS and auxin is integrated with other signal networks to regulate the initiation and development of cotton S.E.