Introduction Melatonin (N-acetyl-5-methoxytryptamine) is a well-known animal hormone that is involved in many biological processes including sleep, mood, circadian rhythms, retinal physiology, seasonal reproductive physiology, temperature homeostasis, sexual behaviour, antioxidative activity, and immunological enhancement (Galano et al., 2011; Venegas et al., 2012; Calvo et al., 2013). However, melatonin is not only found exclusively in animals, but is ubiquitously present in almost all forms of life including protists, prokaryotes, eukaryotic unicells, algae, fungi, and plants (Dubbels et al., 1995; Hattori et al., 1995; Kolář and Macháčkova, 2005; Arnao and Hernández-Ruiz, 2006, 2007; Tan et al., 2007a, b, 2012). In 1995, two reports first identified melatonin in higher plants (Dubbels et al., 1995; Hattori et al., 1995). In the last 20 years, additional research found that melatonin is universally distributed in leaves, roots, stems, fruits, and seeds of all plant species examined (Manchester et al., 2000; Reiter et al., 2001; Kolář and Macháčkova, 2005; Hernández-Ruiz and Arnao, 2008; Murch et al., 2009; Zhao et al., 2013). Interestingly, remarkably high concentrations of melatonin have been identified and quantified in popular beverages (beer, tea, coffee, and wine), crops (barley, corn, grape, wheat, rice, tobacco, and oats), and Arabidopsis in comparison with those in animals (Manchester et al., 2000; Kolář and Macháčkova, 2005; Arnao and Hernández-Ruiz, 2006, 2009b , 2013a , b; Tan et al., 2007a, 2012; Ramakrishna et al., 2012). Additionally, the melatonin content of tomato and rice has been modified by genetic engineering (Okazaki and Ezura, 2009; Okazaki et al., 2009, 2010; Byeon et al., 2012, 2013, 2014; Byeon and Back, 2014a , b). The well-known beneficial effects of melatonin on human health and the abundance of melatonin in popular beverages and crops may encourage the daily consumption of these products (Tan et al., 2012). To date, melatonin has also been found to be a ubiquitous modulator in multiple plant developmental processes and various stress responses (Kolář and Macháčkova, 2005; Arnao and Hernández-Ruiz, 2006; Tan et al., 2007a, 2012). Changes in melatonin in plants may be involved in circadian rhythms, flowering, promotion of photosynthesis, preservation of chlorophyll (Arnao and Hernández-Ruiz, 2009a ; Tan et al., 2012), stimulation and regeneration of root system architecture (Hernández-Ruiz et al., 2005; Pelagio-Flores et al., 2012; Zhang et al., 2014), delayed senescence of leaves (Byeon et al., 2012; Wang et al., 2012, 2013 a, b), and alleviation of oxidative damage mediated by reactive oxygen species (ROS) and reactive nitrogen species (RNS) burst (Tan et al., 2012) Moreover, melatonin protects against multiple abiotic processes such as cold stress (Posmyk et al., 2009a ; Kang et al., 2010; Bajwa et al., 2014), copper stress (Posmyk et al., 2008, 2009b), high temperature (Byeon and Back 2014b ), salt stress (Li et al., 2012), osmotic stress (Zhang et al., 2013), drought stress (Wang et al., 2014), and pathogen infection (Yin et al., 2013). The mechanisms were partially characterized after the direct exogenous application of melatonin to plants (Posmyk et al., 2008, 2009a, b; Zhao et al., 2011; Li et al., 2012; Pelagio-Flores et al., 2012; Wang et al., 2012, 2013a , b; Yin et al., 2013; Bajwa et al., 2014) or the creation of transgenic plants that produced either more or less melatonin through modulating its metabolic pathway (Kang et al., 2010; Byeon et al., 2013, 2014; Park et al., 2013; Byeon and Back, 2014a ; Wang et al., 2014). Finally, the recent studies which showed the protective roles of melatonin in response to abiotic stress indicate that this indole might be a potentially ideal target for future genetic engineering technology to improve abiotic stress resistance in plants. Thus, transgenic plants with higher melatonin concentration might lead to breakthroughs to improve crop production in agriculture as well as the general health of humans (Tan et al., 2012). Bermudagrass [Cynodon dactylon (L). Pers.] is a warm-season turfgrass for lawns, parks, and sport fields cultivated worldwide (Shi et al., 2012, 2013a, b , 2014 b, c, d). In response to global changed environmental stresses, improvement of abiotic stress resistance is very important for grass engineering (Shi et al., 2012, 2013a, b , 2014 b, c, d). As mentioned above, melatonin might be an ideal target for future genetic engineering of some plant species. However, the endogenous melatonin concentration and the possible role of melatonin in response to abiotic stress in bermudagrass is largely unknown. In this study, endogenous melatonin was examined after abiotic stress treatments in bermudagrass plants, and exogenous melatonin treatment was applied to investigate the in vivo role of melatonin in the response of bermudagrass to abiotic stress. In addition, the effects of exogenous melatonin treatment on ROS accumulation and cell damage, as well as underlying antioxidant responses, were determined. Moreover, comparative metabolomic and transcriptomic analyses were performed to identify differentially expressed metabolites and genes after exogenous melatonin treatment. This study provided some insights into the physiological and molecular mechanisms of melatonin in bermudagrass responses to multiple abiotic stresses. Materials and methods Plant materials and growth conditions Newly harvested common bermudagrass seeds were used in this study. After stratification in deionized water at 4 °C for 4 d in darkness, the bermudagrass seeds were sown in soil in the growth room, which was controlled at 28 °C, with an irradiance of ~150 μmol quanta m–2 s–1, 16h light and 8h dark cycles, and ~65% relative humidity. Plant abiotic stress treatment To test the effect of exogenous melatonin on plant physiological responses and abiotic stress resistance, 21-day-old bermudagrass plants were irrigated with water or with different concentrations of melatonin solutions for 7 d, respectively. After melatonin pre-treatment, all control and melatonin-pre-treated 28-day-old bermudagrass plants were subjected to subsequent salt, drought, or cold stress treatments. For salt stress treatment, 28-day-old bermudagrass plants were irrigated with NaCl solutions for 25 d; the NaCl concentration was increased stepwise by 50mM every 5 d to a final concentration of 300mM. For drought stress treatment, 28-day-old plants were subjected to a drought condition by withholding water for 21 d and then re-watered for 4 d. For cold stress treatment, 28-day-old bermudagrass plants were subjected to 4 °C treatment for 21 d, and then transferred to –10 °C for 8h. The freezing stress-treated plants were then recovered overnight at 4 °C and transferred to a standard growth room (28 °C) for 4 d. In each independent experiment, three pots with ~40 plants in each pot were used for each treatment in one concentration of melatonin, and at least three independent experiments were performed to obtain the results. The survival rate of the salt-, drought-, or freezing-stressed plants was recorded at 25 d after stress treatments. The plant leaf samples from melatonin-pre-treated 28-day-old plants were collected at the indicated time points after salt, drought, or cold treatment for the assays of multiple of physiological parameters. Quantification of melatonin by enzyme-linked immunosorbent assay (ELISA) Melatonin from plant leaves was extracted using the acetone–methanol method as described by Pape and Lüning (2006). Briefly, 1g of plant leaf samples was ground in liquid nitrogen, and then transferred to 5ml of extraction mixture (acetone:methanol:water=89:10:1) and homogenized extensively on ice, and the homogenate was centrifuged at 4500 g for 5min at 4 °C. The supernatant was moved to a new centrifuge tube containing 0.5ml of 1% trichloric acid for protein precipitation. After centrifugation at 12 000 g for 10min at 4 °C, the extract was used for quantification of melatonin using the Melatonin ELISA Kit (EK-DSM; Buhlmann Laboratories AG, Schonenbuch, Switzerland) according to the manufacturer’s instruction as described in Shi and Chan (2014a ). Quantifications of chlorophyll content Plant leaf chlorophyll was extracted using 80% (v/v) acetone for 6h with shaking in the dark. The concentration of chlorophyll was then calculated by examining the absorbance at 645nm and 663nm of the centrifuged supernatant. Quantification of electrolyte leakage (EL) The EL of plant leaves under control and abiotic stress conditions was assayed using a conductivity meter (Leici-DDS-307A, Shanghai, China) as previously described (Shi et al., 2012, 2013a, b , 2014 b, c, d). The relative EL was expressed as the ratio of initial conductivity to the conductivity after boiling. Determination of malondialdehyde (MDA) content The MDA content was extracted using chilled thiobarbituric acid (TBA) reagent, and was quantified via determining the absorbance of the supernatant at 450, 532, and 600nm as previously described (Shi et al., 2012, 2013a, b , 2014 b, d). Determination of ROS accumulation and antioxidants As two major indicators of ROS accumulation, hydrogen peroxide (H2O2) and superoxide radical (O2·–) contents were quantified using the titanium sulphate method and the Plant O2·– ELISA Kit (10-40-488, Bejing Dingguo, Beijing, China) as previously described (Shi et al., 2012, 2013a, b, 2014 b, d). The activities of three antioxidant enzymes, namely superoxide dismutase (SOD; EC 1.15.1.1), catalase (CAT; EC 1.11.1.6). and peroxidase (POD; EC 1.11.1.7), were assayed using a Total SOD Assay Kit (S0102; Haimen Beyotime, Haimen city, China), a CAT Assay Kit (S0051; Haimen Beyotime), and a Plant POD Assay Kit (A084-3; Nanjing Jiancheng, Nanjing city, China), respectively, as described by Shi et al., 2012, 2013a, b , 2014 b, d). The concentrations of reduced glutathione (GSH) and oxidized glutathione (GSSG) were determined using the GSH and GSSG Assay Kit (S0053; Haimen Beyotime) as described by Shi et al. (2014b , d ), and the GSH redox state was calculated as the ratio of GSH concentration to the concentration of GSH plus GSSG. Extraction, identification, and quantification of metabolites Extraction, identification, and quantification of metabolites from plant leaves were performed as in Shi et al. (2014d ). Briefly, the metabolite extraction and sample derivatization were performed as in Lisec et al. (2006), then the derivatizated extract was injected into a DB-5MS capillary cloumn (30 m×0.25 mm×0.25 μm; Agilent J&W GC column, California, USA) using gas chromatography time-of-flight mass spectrometry (GC-TOF-MS) (Agilent 7890A/5975C) according to the protocol described by Shi et al. (2014d ). After the GC-TOF-MS assay, the various metabolites were identified via comparing every retention time index-specific mass with reference spectra in mass spectral libraries (NIST 2005, Wiley 7.0). The numerous metabolites were then quantified based on the pre-added internal standard (ribitol) in the process of metabolite extraction. Hierarchical cluster analysis The hierarchical cluster analysis of several metabolites was performed using the CLUSTER program (http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/) and Java Treeview (http://jtreeview.sourceforge.net/) as in Shi et al., 2012, 2013a, b , 2014 b, d). For cluster analysis, all metabolites were quantified as fold change relative to the wild-type bermudagrass plants under control conditions, which was set as 1.0. RNA extraction, library construction, and sequencing For RNA extraction, 28-day-old bermuagrass plants in pots that were irrigated with water or 20 μM melatonin for 7 d were used. Each treatment was represented by two replicate leaf samples, and each sample contained leaves from at least 30 seedlings. Total RNA was extracted with TRIzol (Invitrogen) and was quantified as previously described (Shi et al., 2013c ). RNA quality was determined using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocol. The cDNA libraries were constructed with the mRNA-Seq Sample Preparation Kit™ (Illumina, San Diego, CA, USA) and the DNA yield and fragment insert size distribution of each library were determined on the Agilent Bioanalyzer. The cDNA libraries were then sequenced on an Illumina HiSeq2500 sequencing instrument using the 100bp single end protocol. Quantitative real-time PCR The above RNA samples were also used for synthesis of first-strand cDNA with reverse transcriptase (BIO-RAD, Hercules, CA, USA), and the cDNAs were used for quantitative real-time PCR using a CFX96™ Real Time System (BIO-RAD) as previously described (Shi et al., 2013c ). The specific primers of the analysed genes for real-time PCR are listed in Supplementary Table S1 available at JXB online, and the housekeeping genes have been described in Hu et al. (2012). Bioinformatics analyses of RNA-Seq data Raw RNA-Seq reads were first trimmed for low quality regions using clean reads with length longer than 25bp, obtained after trimming low quality bases (Q 1e-5). GO annotation was performed using the Blast2GO pipeline, and 18 701 (66%) transcripts were assigned with at least one GO term. Among the three GO categories, 13 402 transcripts were annotated in Biological Process, 14 052 transcripts in Molecular Function, and 14 685 in Cellular Component. Using fold change >2 and false discovery rate (FDR) 16-fold after melatonin treatment (Table 2). GO enrichment analysis in the biological process domain suggested that genes related to the cysteine biosynthetic process, response to light signal, and the photosynthetic process were down-regulated. In particular, the studies of Wang et al. (2012) showed that melatonin can lower ROS damage of many photosynthetic components. Therefore, the expression of genes involved in the photosystem might been suppressed through a negative feedback mechanism. The up-regulated genes were greatly enriched with the GO terms involved in gene expression regulatory process, such as protein phosphorylation, DNA-dependent transcription, regulation of circadian rhythm, etc. (Fig. 7). Table 1. Pathway enrichment analysis of genes whose expression was significantly affected by exogenous melatonin pre-treatment in bermudagrass Group MapMan pathway Up-regulation Down-regulation NF P-value NF P-value I N metabolism 5.71 0.0000 4.00 0.0140 Major CHO metabolism 3.30 0.0000 2.31 0.0100 TCA/org transformation 2.58 0.0023 2.63 0.0076 Transport 2.05 0.0000 2.42 0.0000 Hormone metabolism 2.02 0.0000 1.44 0.0110 Metal handling 1.78 0.0440 2.19 0.0260 Redox 1.59 0.0160 2.60 0.0000 Secondary metabolism 1.42 0.0094 2.75 0.0000 II Gluconeogenesis/ Glyoxylate cycle – – 24.02 0.0000 Glycolysis – – 2.30 0.0210 Photosystem – – 8.21 0.0000 Tetrapyrrole synthesis – – 4.88 0.0001 Fermentation 6.63 0.0005 – – Minor CHO metabolism 2.99 0.0000 – – Signalling 2.53 0.0000 0.95 0.0550 C1 metabolism 2.32 0.0430 – – III RNA 1.84 0.0000 1.04 0.0340 Cofactor and vitamin metabolism 1.83 0.0400 – – Amino acid metabolism 1.71 0.0036 1.70 0.0120 Lipid metabolism 1.64 0.0009 1.69 0.0023 Nucleotide metabolism 1.54 0.0280 1.87 0.0120 Stress 1.46 0.0000 – – Development 1.41 0.0016 – – Miscellaneous 1.27 0.0015 1.71 0.0000 Protein 1.11 0.0031 0.73 0.0000 IV Cell 0.80 0.0260 – – Not assigned 0.45 0.0000 0.76 0.0000 Mitochondrial electron transport/ATP synthesis 0.24 0.0085 – – DNA 0.12 0.0000 0.18 0.0000 MicroRNA, natural antisense, etc 0.00 0.0000 0.00 0.0000 Twenty-one-day-old bermuagrass plants in pots were irrigated with water or 20 μM melatonin for 7 d, then the 28-day-old plants (without and with 7 d of pre-treatment) were used for transcriptomic analysis. Differentially expressed genes (i.e. with P-value ≤0.05 and log2 fold change ≥1 or log2 fold change ≤ –1) were annotated using the Classification SuperViewer Tool and with MapMan. MapMan was used as the classification source. Group I indicates highly enriched pathways of both up- and down-regulated genes; group II indicates highly enriched pathways of either up- or down-regulated genes; group III indicates slightly enriched pathways; and group IV indictes under-represented pathways. Scales of normalized frequency (NF) are as follows: ≥4 2–4 1–20. 5–1 ≤0.5 Table 2. List of getnes highly induced (>16-fold) by exogenous melatonin pre-treatment in bermudagrass Seq_ID logFC P-value FDR Putative annoation in Arabdopsis E-value a Putative annoation in rice E-value a comp58508_c2_seq12 8.47 5.35E-247 2.50E-245 CBF4, DREB1D 2e-07 DRE 1e-35 comp58508_c2_seq7 7.49 0 0 DREB1A, CBF3 9e-15 DRE 2e-25 comp57631_c0_seq7 2.98 0 0 DREB2A 2e-23 AP2 domain containing protein 2e-62 comp55697_c1_seq4 4.92 0 0 DNA-binding protein 1e-23 AP2 domain containing protein 6e-34 comp49674_c0_seq1 4.49 0 0 HRE1 1e-14 AP2 1e-44 comp52672_c0_seq8 5.57 0 0 ERD15 5e-17 Expressed protein 2e-46 comp51935_c1_seq1 5.89 0 0 Cold-regulated gene 27 3e-13 Expressed protein 4e-74 comp59530_c3_seq1 5.50 4.31E-51 4.22E-50 HSFA6B 5e-61 HSF 4e-133 comp51719_c2_seq14 4.61 0 0 HSFC1 1e-57 HSF 1e-113 comp56606_c0_seq9 5.48 9.30E-52 9.25E-51 HSP20-like 2e-17 hsp20 3e-27 comp60859_c0_seq13 5.13 1.51E-14 5.65E-14 SLT1 | HSP20-like 4e-172 SLT1 protein 0 comp53718_c0_seq3 6.51 0 0 DnaJ 2e-36 Heat shock protein DnaJ 1e-78 comp51393_c1_seq14 5.36 2.40E-58 2.67E-57 DnaJ 6e-16 Heat shock protein DnaJ 1e-36 comp23443_c0_seq1 7.89 6.84E-42 5.67E-41 DNAJ heat shock protein 1e-35 Expressed protein 4e-61 comp53904_c0_seq1 4.32 0 0 WRKY25 4e-35 WRKY53 4e-51 comp55751_c0_seq1 6.18 0 0 WRKY40 2e-53 WRKY71 1e-132 comp55960_c0_seq3 4.57 1.84E-212 7.05E-211 WRKY46 4e-35 WRKY74 1e-93 comp54790_c0_seq12 5.79 1.72E-246 7.99E-245 WRKY51 1e-31 WRKY67 6e-37 comp57307_c2_seq3 5.57 0 0 AZF1 8e-12 C2H2 zinc finger protein 1e-23 comp54697_c5_seq1 4.22 1.63E-163 4.59E-162 ATL6 3e-42 Zinc finger, C3HC4 type 1e-55 comp57307_c1_seq2 5.61 0 0 STZ, ZAT10 4e-25 C2H2 zinc finger protein 3e-36 comp55694_c0_seq3 5.27 0 0 NIP2 1e-51 RING-H2 finger protein ATL2B 2e-90 comp57199_c1_seq6 4.07 1.01E-155 2.69E-154 AtMYB78 6e-14 MYB 2e-14 comp60805_c1_seq6 6.20 0 0 RVE1 5e-36 MYB 7e-97 comp60805_c1_seq4 5.70 0 0 LHY 2e-13 MYB 2e-92 comp58066_c0_seq2 4.59 8.75E-312 5.22E-310 BHLH 3e-38 Ethylene-responsive protein 1e-95 comp54812_c1_seq1 6.33 0 0 BHLH92 3e-19 BHLH 2e-79 comp56171_c3_seq1 5.02 0 0 anac032, NAC032 6e-07 NAC 3e-54 comp54298_c0_seq8 4.56 0 0 AF2 7e-73 NAC 2e-145 comp54697_c3_seq1 4.60 2.23E-132 5.10E-131 RING/U-box protein 1e-28 Zinc finger 4e-49 comp53888_c2_seq1 4.15 9.66E-50 9.21E-49 CMPG2 1e-15 U-box 7e-20 comp53888_c0_seq1 4.74 1.02E-122 2.15E-121 PUB29 2e-31 U-box 9e-90 comp48740_c0_seq1 6.07 0 0 Calcium-binding EF-hand protein 3e-35 EF hand family protein 8e-68 comp48880_c0_seq1 5.33 0 0 Calcium-binding EF-hand protein 1e-28 OsCML31 4e-60 comp61173_c2_seq4 5.64 0 0 Calcium-binding protein 3e-34 OsCML10 1e-57 comp42867_c0_seq1 5.25 6.59E-310 3.93E-308 Calcium-binding protein 1e-43 OsCML14 7e-72 comp50275_c0_seq5 4.36 1.47E-284 7.97E-283 Calcium-binding protein 5e-25 EF hand 1e-40 comp60797_c0_seq1 5.12 0 0 Calmodulin-binding protein 2e-160 Calmodulin-binding protein 0 comp55169_c0_seq2 4.50 0 0 Calmodulin-binding protein 8e-85 Calmodulin-binding protein 3e-145 comp51845_c0_seq1 4.67 0 0 CML43 3e-39 OsCML27 6e-78 comp55740_c0_seq1 4.20 0 0 TCH2, CML24 2e-38 OsCML16 1e-80 comp57565_c0_seq8 9.08 7.50E-220 3.05E-218 CKA1 0 Casein kinase II 0 comp54827_c0_seq1 6.78 6.62E-48 6.09E-47 Kinase 2e-20 Kinase 3e-26 comp55548_c0_seq1 4.39 0 0 Kinase 6e-148 Phosphotransferase 0 comp55703_c1_seq1 4.29 7.18E-159 1.96E-157 Kinase 7e-65 Tyrosine protein kinase 2e-124 comp60546_c0_seq1 4.11 1.75E-75 2.41E-74 Kinase 0 Leucine-rich repeat protein 0 comp54489_c2_seq1 5.52 3.60E-185 1.19E-183 JMJD5 8e-139 jmjC protein 5 9e-146 comp48478_c1_seq1 4.45 1.09E-76 1.52E-75 HDA18 2e-64 Histone deacetylase 2e-71 comp57103_c1_seq1 4.45 0 0 ERF-1 3e-22 ERF 1e-43 comp51949_c0_seq2 4.49 0 0 JAZ2, TIFY10B 2e-10 ZIM protein 8e-31 comp59913_c1_seq8 7.52 0 0 CYP707A1 8e-50 Cytochrome P450 1e-64 comp50455_c0_seq1 3.42 0 0 PYL5, RCAR8 7e-55 Cyclase/dehydrase 1e-87 comp55459_c2_seq1 4.20 0 0 PP2C 6e-101 PP2C 1e-164 comp55059_c0_seq4 4.33 2.77E-269 1.43E-267 SHY2, IAA3 2e-47 OsIAA24 2e-64 comp56859_c0_seq6 5.51 3.14E-169 9.21E-168 Alcohol dehydrogenase 8e-100 Dehydrogenase 9e-100 comp55475_c1_seq1 4.41 0 0 WCRKC thioredoxin 1 4e-45 Thioredoxin 4e-80 comp49522_c0_seq2 4.02 9.96E-36 7.38E-35 Oxidoreductase 2e-88 Dehydrogenase 4e-136 comp57222_c0_seq2 5.09 0 0 FMO1 2e-91 Monooxygenase 0 comp56296_c0_seq6 5.02 2.03E-49 1.92E-48 Copper transport protein 2e-09 Heavy metal-associated protein 3e-26 comp46851_c0_seq1 4.92 5.38E-117 1.07E-115 Heavy metal transport 1e-06 Expressed protein 4e-21 comp51430_c0_seq1 4.45 7.94E-278 4.22E-276 Heavy metal transport 1e-15 Heavy metal-associated protein 3e-41 Twenty-one-day-old bermuagrass plants in pots were irrigated with water or 20 μM melatonin for 7 d, then the 28-day-old plants (with and without 7 d of pre-treatment) were used for transcriptomic analysis. logFC, log2 fold change; FDR, false discovery rate. e E-value, expected value for putative annotation in Arabidopsis or rice. Fig. 7. The Biological Process GO terms enrichment of down-regulated (A) and up-regulated (B) genes between control and melatonin-pre-treated bermudagrass. The horizontal axis shows –log10 of the P-value. Twenty-one-day-old bermuagrass plants in pots were irrigated with water or 20 μM melatonin for 7 d, then the 28-day-old plants (with and without 7 d of pre-treatment) were used for transcriptomic analysis. (This figure is available in colour at JXB online.) To confirm the reliability of the RNA-Seq data, the expression of 18 genes (nine up-regulated and nine down-regulated by exogenous melatonin treatment) that were differentially expressed between control and melatonin-treated plants was assessed via quantitative real-time PCR. Consistently, the results of the real-time PCR assay exhibited the same trend and were correlated with the RNA-Seq data (Supplementary Fig. S1 at JXB online), confirming the reproducibility of RNA-Seq data. Pathway and GO term enrichment analyses The transcriptome data were submitted to the Mercator web tool to align with the public protein database, and 14 288 transcripts were located in at least one point in plant biological pathways. As shown in Table 1, pathway analysis revealed that melatonin affected the expression of many genes involved in N metabolism, minor carbohydrate metabolism, TCA/org transformation, transport, hormone metabolism, metal handling, redox, and secondary metabolism (Table 1, group I). Other transcripts involved in stress response and metabolism were extensively changed after melatonin treatment (Fig. 6B; Supplementary Fig. S2 at JXB online). These results indicated that melatonin treatment might induce a stress response in bermudagrass. The pathway analysis results were consistent with the study carried out by Weeda et al. (2014), which showed that melatonin altered many genes involved in plant defence (Supplementary Fig. S2), and these changes might contribute to the enrichment of stress-related GO terms (Fig. 7). Discussion As sessile organisms, plants have developed sophisticated strategies to respond to diverse environmental stresses. The stress signals are perceived by several receptors at the cell membrane level, followed by their transduction to multiple second messengers such as abscisic acid (ABA), H2O2, nitric oxide (NO), etc. These activate downstream stress-responsive genes and physiological responses, eventually leading to protective responses at the whole-plant level (Shi et al., 2012, 2013a, b , 2014 b, c, d; Shi and Chan, 2014a , b). Although no direct evidence had indicated that melatonin could serve as a second messenger, the induction of melatonin by multiple stress treatments (Fig. 1) indicates an in vivo role for melatonin in bermudagrass response to abiotic stress (Arnao and Hernández-Ruiz, 2006, 2009b , 2013a , b; Tan et al., 2007a, 2012). In this study, the protective role of melatonin on the response of bermudagrass to abiotic stress was revealed. Under control conditions, melatonin had no significant effect on bermudagrass growth or its physiological responses (Fig. 2). Under abiotic stress conditions, however, melatonin-pre-treated plants exhibited significantly higher chlorophyll content, lower EL, and higher survival rate than did non-treated bermudagrass plants (Fig. 2C–E). After recovery from abiotic stress treatments, melatonin-pre-treated plants exhibited better growth status than non-treated plants, with higher biomass (plant height and weight) (Fig. 2F, G). These results indicate that exogenous melatonin application improved salt, drought, or freezing stress resistance in bermudagrass, in accordance with the enhanced resistance to cold stress (Posmyk et al., 2009a ; Kang et al., 2010; Bajwa et al., 2014), copper stress (Posmyk et al., 2008, 2009b), high temperature (Byeon and Back, 2014b ), salt stress (Li et al., 2012), osmotic stress (Zhang et al., 2013), drought stress (Wang et al., 2014), and pathogen infection (Yin et al., 2013) due to melatonin in various plant species. As an antioxidant in animals, melatonin scavenges free radicals directly, stimulates the activities of antioxidant enzymes including CAT, SOD (both MnSOD and CuSOD), glutathione reductase (GR), and glutathione peroxidase (GPX), and increases the efficiency of mitochondrial oxidative phosphorylation (Tan et al., 1993, 1999, 2003, 2007a, b; Reiter et al., 2000; Kolář and Macháčkova, 2005). In plants, melatonin is also an important antioxidant and a radical scavenger (Arnao et al., 1996, 2001; Cano et al., 2003, 2006). Li et al. (2012) also found that exogenous melatonin modulates salinity-induced oxidative damage by directly scavenging H2O2 and enhancing the activities of antioxidative enzymes in Malus hupehensis. Consistently, exogenous application of melatonin significantly activated ROS detoxification of antioxidants, including enzymatic antioxidant enzymes (SOD, CAT, and POD) and non-enzymatic glutathione (GSH redox state), to maintain cellular ROS (mainly including H2O2 and O2·–) at a relatively low level. This results in the alleviation of abiotic stress-induced oxidative damage and further conferred improved abiotic stress resistance (Figs 3, 4). Consistently, RNA-Seq found that many genes involved in redox, many POD genes and glutathione S-transferases (GST) genes were significantly modulated by exogenous melatonin treatment (Table 1; Supplementary Fig. S2 at JXB online). In summary, the positive modulation by exogenous melatonin of the ROS detoxification system might contribute greatly to enhanced abiotic stress resistance in bermudagrass. Moreover, comparative metabolomic analysis showed the actions of melatonin treatment on carbon metabolites and amino acid metabolism under abiotic stress conditions. Notably, melatonin-pre-treated plants exhibited higher concentrations of 54 metabolites compared with non-melatonin-treated plants (Fig. 5; Supplementary Table S2 at JXB online). Among these metabolites, proline and some carbohydrates (fructose, sucrose, glucose, maltose, cellobiose, trehalose, galactose, and galactinol) are important compatible solutes to respond to abiotic stress for osmotic adaptation (Krasensky and Jonak, 2012). Thus, higher levels of proline and carbohydrates (glucose, maltose, fructose, sucrose, and trehalose) in melatonin-pre-treated plants provided beneficial effects under abiotic stress conditions. In addition, higher levels of other metabolites including multiple amino acids, organic acids, and sugars in melatonin-pre-treated plants indicate the beneficial physiological processes in melatonin-pre-treated plants during abiotic stress treatment; the data further confirm the protective role of melatonin in response to abiotic stress. Notably, 18 metabolites comprising 10 amino acids, six sugars, and two sugar alcohols of the carbon metabolic pathway exhibited significantly higher levels in melatonin-pre-treated plants under abiotic stress conditions (Fig. 6A). These results indicate the widespread effects of melatonin treatment in carbon metabolism and amino acid metabolism; these metabolites might play some role in melatonin-mediated abiotic stress resistance in bermudagrass. Additionally, comparative transcriptomic analysis identified 2361 up-regulated and 1572 down-regulated transcripts as a consequence of exogenous melatonin treatment. Quantitative real-time PCR of 18 genes supported the reliability of the RNA-Seq data (Supplementary Fig. S1 at JXB online). Pathway enrichment analysis indicated that eight pathways were over-represented among differentially expressed genes between control and melatonin-treated bermudagrass plants, including N metabolism, major carbohydrate metabolism, TCA/org transformation, transport, hormone metabolism, metal handling, redox, and secondary metabolism (Table 1, group I). The enrichment of redox-related genes affected by melatonin (Table 1, group I) was consistent with the effects of exogenous melatonin on ROS detoxification in bermudagrass (Figs 4, 5). In animals, melatonin is known to be involved in circadian rhythms (Tal et al., 2011). Interestingly, the GO enrichment analysis also showed that transcripts that function in regulation of rhythms and flowering were over-represented (Fig. 7). Notably, the pathway analysis results were consistent with those of the study carried out in Arabidopsis by Weeda et al. (2014). Thus, melatonin altered many genes involved in plant defence in bermudagrass (Supplementary Fig. S2); these changes probably contributed to the enrichment of stress-related GO terms (Fig. 7). Additionally, exogenous melatonin treatment had significant effects on various signalling pathways including primary metabolism, secondary metabolism, photosynthesis, large enzyme families, receptor-like kinases, proteolysis, and autophagy pathways in bermudagrass as determined using MapMan software (Fig. 6B; Supplementary Fig. S2). Those genes modulated by exogenous melatonin treatment might also contribute to melatonin-enhanced abiotic stress resistance in bermudagrass. Asparagine accumulation shows that nitrogen re-distribution and mobilization were important features of the melatonin response (Lea et al., 2007; Maaroufi-Dguimi et al., 2011). Jia et al. (2001) also suggested that asparagine may be a signalling molecule involved in sensing the nitrogen status. In addition, asparagine is an amino group donor for the synthesis of the photorespiratory intermediate glycine. Nagy et al. (2013) and Shi et al. (2014) found that this is also a good indicator of drought stress in drought-tolerant and sensitive wheat and bermudagrass cultivars. At the same time, some carbohydrate metabolism compounds increased: these included fructose, glucose, and trehalose, but not sucrose. Such differential dynamics of carbohydrates could reflect modifications of carbon balance and carbon utilization. Moreover, asparagine synthetase genes that are involved in asparagine synthesis are regulated by the level of carbohydrates (Lam et al., 1998; Foito et al., 2009). TCA/org transformation is important for the Calvin cycle for CO2 assimilation and separation of initial carbon fixation by contact with air and secondary carbon fixation into sugars (Selwood and Jaffe, 2011). Glycolysis is an important metabolic pathway in carbohydrate metabolism, and the central role of glycolysis in the plant metabolic pathway is to provide energy such as ATP and generates precursors for anabolism such as fatty acids and amino acids (Plaxton, 1996). In accordance with the metabolic profiles, transcriptomic analysis found that many genes involved in TCA/org transformation and N metabolism were modulated by melatonin treatment (Table 1; Fig. 6B). Genes which functioned in sucrose and amino acid metabolism were also greatly changed after melatonin treatment (Fig. 6B), leading to altered sucrose and amino acid contents revealed by metabolite analysis (Figs 5, 6). These results showed that the underlying mechanisms related to melatonin may involve major reorientation of photorespiratory and carbohydrate and nitrogen metabolism. To date, various TFs have been shown to be involved in plant stress responses via activating stress-responsive gene expression, such as BASIC LEUCINE ZIPPER PROTEINS (bZIPs), CBF/DREBs, ETHYLENE-RESPONSIVE ELEMENT-BINDING FACTORS (ERFs), MYBs, WRKYs, and zinc finger proteins (ZFPs) (Shi et al., 2014a , e ). In the current study, many TFs were significantly regulated by exogenous melatonin treatment (Table 2; Supplementary Tables S4, S5 at JXB online), and these TFs might contribute to the enhanced stress tolerance of melatonin-treated plants, thus indicating that melatonin might pre-condition to be resistant to abiotic stresses. Some protein kinases [such as mitogen-activated protein kinase (MAPK)] and calcium signalling kinases [including calcium-dependent protein kinases (CDPKs), calcineurin B-like (CBL)-interacting protein kinases (CIPKs), and calcium-related protein kinases (CRKs)] were also transcriptionally regulated by exogenous melatonin treatment (Table 1; Supplementary Tables S4, S5). This suggests that kinase signalling might play critical roles in melatonin-mediated stress responses. As sessile organisms, plants cannot avoid unfavourable stress conditions by adjusting their location; thus, they have evolved complex strategies to perceive stress signals and further translate the perception into effective responses, which might largely depend on various protein kinases and TFs (Shi et al., 2014a , e ). Recently, it was found that one cysteine2/histidine2-type zinc finger TF, zinc finger of Arabidopsis thaliana 6 (ZAT6), is involved in melatonin-mediated freezing stress response, and the AtZAT6-activated CBF pathway was essential for melatonin-mediated freezing stress response in Arabidopsis (Shi and Chan, 2014a ). This study together with others in sunflower (Helianthus annuus) (Mukherjee et al., 2014) and in Arabidopsis (Shi and Chan, 2014a ; Weeda et al., 2014) indicate that melatonin is involved in long-distance signal transduction in plants. Moreover, some important genes in plant hormone signalling [RCAR/PYR/PYL, SNF1-related protein kinases 2 (SnRK2), and nine-cis-epoxycarotenoid dioxygenase (NCED) genes in ABA signalling, and jasmonate (JA)-JIM-domain proteins (JAZs) in JA signalling] that were significantly regulated by exogenous melatonin treatment (Table 1; Supplementary Tables S4, S5 at JXB online) might also have some function in melatonin-mediated cross-talk among plant hormones, as well as in stress responses. Kolář and Macháčkova (2005) also found that melatonin might function as an auxin to promote vegetative growth. These results suggested that melatonin might serve as a plant hormone that cross-talks with other plant hormones. Thus, melatonin triggered extensive transcriptional reprogramming and pre-conditioned resistance to multiple abiotic stresses. Further investigation of the in vivo roles of these genes will shed additional light on melatonin-mediated stress responses in bermudagrass. Taken together, this study provides the first evidence of the protective roles of exogenous melatonin in bermudagrass response to multiple abiotic stresses. This involved the activation of antioxidants, modulation of metabolic homeostasis, and extensive transcriptional reprogramming. Supplementary data Supplementary data are available at JXB online. Figure S1. Validation of differentially expressed genes by quantitative real-time PCR. Figure S2. Effect of exogenous melatonin treatment on stress-related pathways in bermudagrass. Table S1. The specific primers used for real-time PCR. Table S2. Concentrations of 54 metabolites in 28-day-old bermudagrass plants under control conditions and different treatments [20 μM melatonin, 300mM NaCl, drought, and cold (4 °C)] stress conditions for 14 d. Table S3. Summary of RNA-Seq data. Table S4. List of up-regulated genes in melatonin-treated bermudagrass plants. Table S5. List of down-regulated genes in melatonin-treated bermudagrass plants. Supplementary Data