Plant development and function are underpinned by redox reactions that depend on co-factors such as nicotinamide adenine dinucleotide (NAD). NAD has recently been shown to be involved in several signalling pathways that are associated with stress tolerance or defence responses. However, the mechanisms by which NAD influences plant gene regulation, metabolism and physiology still remain unclear. Here, we took advantage of Arabidopsis thaliana lines that overexpressed the nadC gene from E. coli, which encodes the NAD biosynthesis enzyme quinolinate phosphoribosyltransferase (QPT). Upon incubation with quinolinate, these lines accumulated NAD and were thus used as inducible systems to determine the consequences of an increased NAD content in leaves. Metabolic profiling showed clear changes in several metabolites such as aspartate-derived amino acids and NAD-derived nicotinic acid. Large-scale transcriptomic analyses indicated that NAD promoted the induction of various pathogen-related genes such as the salicylic acid (SA)-responsive defence marker PR1. Extensive comparison with transcriptomic databases further showed that gene expression under high NAD content was similar to that obtained under biotic stress, eliciting conditions or SA treatment. Upon inoculation with the avirulent strain of Pseudomonas syringae pv. tomato Pst-AvrRpm1, the nadC lines showed enhanced resistance to bacteria infection and exhibited an ICS1-dependent build-up of both conjugated and free SA pools. We therefore concluded that higher NAD contents are beneficial for plant immunity by stimulating SA-dependent signalling and pathogen resistance.
Pyridine nucleotides such as nicotinamide adenine dinucleotide (NAD) and its derivative NAD phosphate (NADP) are ubiquitous key redox carriers and act as coenzymes for oxidoreductases in virtually all metabolic pathways. Recently, further roles for NAD and NADP in cell signalling and gene regulation have been proposed, both in animals and plants (Hunt et al., 2004). For example, molecular mechanisms involve the conversion of NAD to cyclic ADP-ribose (cADPR), which triggers the release of Ca2+ from intracellular compartments. In addition, NAD is the substrate for mono- and poly-ADP ribosylation catalysed by poly-ADP-ribose proteins (PARPs) (Vanderauwera et al., 2007) involved in the regulation of several processes including DNA repair, transcription and apoptosis. NAD is further thought to participate in cell signalling through NAD+-dependent histone deacetylases (Silent information regulator 2 (Sir2) proteins or ‘sirtuins’) that play important roles in chromatin silencing, DNA repair, cell cycle, apoptosis, and ageing (Robyr et al., 2002; Blander and Guarente, 2004).
Recently, pyridine nucleotide pools have been found to correlate with cellular processes that are involved in stress responses (Berger et al., 2004). Similarly, the CMSII mitochondrial mutant of Nicotiana sylvestris that lacks the respiratory complex I, in which an increase in NAD is associated with high amino acid and ATP contents, improved plant ozone tolerance and resistance to virus infection, thereby suggesting that NAD may be involved in plant defence responses (Dutilleul et al., 2003, 2005; Djebbar et al., 2012). Other data also provide evidence that extracellular NAD(P) plays a role in plant immunity (Zhang and Mou, 2009). In fact, exogenous application of pyridine nucleotides induces both pathogenesis-related (PR) gene expression in Arabidopsis thaliana (Arabidopsis) and disease resistance to the bacterial pathogen Pseudomonas syringae pv. tomato (Pst), which leads to an accumulation of the defence signal molecule salicylic acid (SA) (Zhang and Mou, 2009). Induction of SA-mediated plant defence responses is also observed when plants are sprayed with neonicotinoids, which are synthetic analogues of the natural insecticide nicotine. Their chemical structure is reminiscent of that of natural NAD derivatives and allows them to mimic natural molecules that induce SA production (Ford et al., 2010). Moreover, Arabidopsis inoculated with the avirulent pathogen Pst DC3000 (AvrRpt2+) exhibited an increase in poly-ADP ribosylation mechanisms (Adams–Phillips et al., 2010), which suggested that catabolism of NAD also contributed to the protective mechanisms involved in biotic stress plant responses. Finally, NAD has also appeared as a precursor of pyridine alkaloids that are assumed to play roles in plant defence responses (Noctor et al., 2006).
In plants, five enzymatic steps are necessary to produce NAD from aspartate (Asp) (Figure 1). De novo biosynthesis starts in the chloroplast with nicotinamide adenine mononucleotide (NaMN) synthesis in Arabidopsis (Katoh et al., 2006). In brief, quinolinate (Q, pyridine 2,3-dicarboxylic acid) is formed from aspartate and dihydroxyacetone phosphate by aspartate oxidase (AO) plus quinolinate synthase (QS); quinolinate is then converted to NaMN by quinolinate phosphoribosyltransferase (QPT). The further steps occur in the cytosol and are shared between the biosynthesis and recycling of NAD. NaMN is adenylylated to nicotinic acid adenine dinucleotide (NaAD) by nicotinate mononucleotide adenyl transferase (NaMNAT). NAD synthetase (NADS) catalyses the final step of NAD biosynthesis by amidation of NaAD to NAD (Ashihara et al., 2005). However, the molecular and regulatory details of NAD synthesis from aspartate remain largely unknown. Several lines of evidence have suggested that QPT plays an important role in the control of NAD levels and its derivatives; in fact, QPT may respond to the metabolic demand for pyridines caused by increased nicotine synthesis, for instance (Mann and Byerrum, 1974; Feth et al., 1986; Wagner et al., 1986). In tobacco, QPT is believed to be critical in sustaining the biosynthesis of nicotinic acid and defensive pyridine alkaloids such as putrescine, anabasine or nicotine (Sinclair et al., 2000). The NAD biosynthetic pathway in Escherichia coli (E. coli) is similar to that in plants and QPT is encoded by the nadC gene. In other words, the bacterial enzyme catalyses the de novo formation of NaMN from quinolinate as occurs in plant cells (Griffith et al., 1975).
While the importance of NAD in cell signalling is already established, the mechanisms by which NAD influences plant gene regulation, metabolism and physiology remain unclear. Here, we took advantage of Arabidopsis lines that overexpressed the nadC gene from E. coli. Upon incubation with quinolinate, these lines accumulated NAD and were thus used as convenient inducible systems to investigate the changes due to an increased NAD content in leaves. We used a combination of genetics, transcriptomics, metabolomics and physiological analyses, and showed that the increase in NAD modulates plant responses through gene expression and mainly influences the induction of various pathogen-related genes. Higher NAD levels also lead to improved resistance to the avirulent Pst-AvrRpm1 strain and an ICS1-dependent accumulation of salicylic acid (SA). Our study therefore suggests that NAD stimulates plant resistance probably via SA-dependent signalling.
Arabidopsis plants that overexpress the nadC gene from Escherichia coli
Stable transformation with the CaMV35S construct using the PCW162 vector resulted in several independent lines that overexpressed the E. coli nadC gene (Figure 2a), and leading to the ectopic expression of the nadC gene. The control line 162.7.17 (denoted as ‘Ctrl’) was obtained from the Columbia (Col0) ecotype transformed with the empty vector PCW162. The presence of the nadC gene in Arabidopsis was examined by polymerase chain reaction (PCR) and gel blot analysis (Figure 2b), and the QPT specific activity was quantified (Figure 2c). The two lines 4.11 and 15.3 overexpressed the nadC gene in the cytosol and showed a substantial increase in QPT activity. Under typical growth conditions, the transgenic nadC lines, as well as the Ctrl, were morphologically aphenotypic.
An inducible cellular system to increase NAD levels and derivatives in Arabidopsis
We used the nadC lines that exhibited NAD accumulation when leaf discs were treated with quinolinic acid (Figure 3). Leaf discs were incubated in 3-(N-morpholino)propanesulfonic acid (MOPS) buffer that was supplemented with quinolinate during 48 h in the dark to raise the NAD content to its maximum (the time course in undetached leaves is presented below in Figure 6). Dark incubation was used to avoid the production of anthocyanins. We adopted a modified method (Queval and Noctor, 2007), which provides a highly efficient plate-reader assay to quantify pyridine nucleotides. NAD accumulated to significant levels both in the 4.11 and 15.3 lines in the presence of quinolinate (Figure 3a), and showed enrichments of three- and four-fold, respectively. In contrast, there was only a very slight increase in the NAD level in the Ctrl. Importantly, the redox status of NADH/NAD was not changed and remained near 7%. The NADP content also increased in the quinolinate-treated nadC lines (Figure 3b). Overall, line 15.3 accumulated a higher amount of pyridine nucleotides (both NAD and NADP) after quinolinate addition than did the line 4.11. We therefore expected to observe larger metabolic changes in quinolinate-treated 15.3 leaves than those of 4.11.
Furthermore, as NADPH oxidation is potentially coupled with the glutathione system (Asada, 1999), we analysed antioxidants by combining the plate-reader assay to high performance liquid chromatography (HPLC) and using fluorescence detection to determine the thiol content. No significant changes were observed in the leaf total glutathione (Figure 3c) and ascorbate (data not shown) contents in either the Ctrl or nadC lines. In the Ctrl line, however, quinolinate treatment affected the glutathione redox status as we observed the oxidation of glutathione (GSH) to glutathione disulphide (GSSG). This finding suggested that under quinolinate application, the Ctrl line had a more oxidative cellular medium compared with nadC lines. Altogether, our results indicated that the 15.3 and 4.11 lines were quinolinate-inducible systems with increased NAD synthesis.
Metabolic responses to elevated NAD contents
We did not observe a statistically significant difference in dry weight/fresh weight ratio and chlorophylls contents for Ctrl and nadC plants. To test whether the inducible enrichment of NAD in Arabidopsis influences primary C and N metabolism, we performed non-targeted metabolite profiling by gas chromatography/time-of-flight mass spectrometry (GCMS) and quantified amino acids by a fluorimetric HPLC technique (Bathellier et al., 2009; Tcherkez et al., 2010). Both methods showed altered levels of amino acids and metabolites (Table S1). However, the total amino acid content was not significantly different between the lines and did not change upon quinolinate incubation (Figure 3d). Quinolinate-treated Ctrl discs had less asparagine (Asn, Figure 3e), aspartate (Asp, Figure 3f) and Asp derivatives (Table S1). Importantly, the addition of quinolinate led to higher levels of nicotinic acid both in the control and nadC discs (Figure 3h), which strongly suggested that either NAD was recycled back to nicotinamide (NAM) and nicotinate, or that the possible degradation of NaMN (evolved by NaPT) to nicotinate was enhanced (see Figure 1). When treated with quinolinic acid, both quinolinate-treated nadC lines accumulated substantial amounts of Asn (Figure 3e), Asp (Figure 3f) and methionine (Met, Figure 3g). Under the same conditions (nadC with quinolinate), the levels of several amino acids derived from Asp also increased (isoleucine, leucine and valine; see Table S1), and so did the xylose content (Figure 3). Intermediate levels of serine, phenylalanine and glutamine were also observed.
Transcriptomic effect of elevated NAD contents
A transcriptomic analysis was carried out, given the essential role of pyridine nucleotides in various metabolisms. We adopted a global study based on the Complete Arabidopsis Transcriptome MicroArray (CATMA) (Hilson et al., 2004; Sclep et al., 2007). Because of its high NAD enrichment when treated with quinolinic acid, the nadC-overexpression line 15.3 was chosen for such a large-scale analysis. We used a ‘horseshoe’ hybridisation scheme (summarised in Figure 4a) to distinguish the effect of quinolinate treatment per se from that of the build-up of pyridine nucleotide. This approach allowed us to identify the regulation of gene expression caused by a specific increase in the NAD content. To quantify the relative abundance of each mRNA, independent experiments were performed with total RNA. We compared the following treatments: A, nadC+Q versus nadC; B, nadC+Q versus Ctrl+Q and C, Ctrl+Q versus Ctrl. The missing comparison (D, nadC versus Ctrl) was calculated simply with other ratios given by the CATMA analysis as follows: D =C + B − A (the computation is additive as it is logarithm based). As described in Gagnot et al. (2008), we considered genes to be differentially expressed if they gave a Bonferroni P-value <0.05 and showed a statistically significant change in the same direction in both biological replicates. Means were analysed by the hierarchical clustering (HCL) function of the MultiExperimentViewer (MeV) software. The HCL gave a graphical representation of 333 statistically significant regulated genes (Figure 4b and Table S2). To further characterise the transcriptomic regulations, we classified genes using the Cluster Affinity Search Technique (CAST) that eventually subdivided our data into four classes (Figure 4b). Class 1 and 2 represented NAD-repressed and NAD-induced genes, respectively; however, class 3 referred to the effect of quinolinate treatment, and class 4 denoted the effect of overexpression of the nadC gene (comparing nadC versus Ctrl, calculated condition D). The hierarchical analysis clearly showed a highly NAD-dependent transcriptome. Indeed, many more genes (92%) were significantly regulated by elevated NAD levels (97 down-regulated and 209 up-regulated genes corresponding to class 1 and 2, Figure 4b) than by quinolinate treatment (19 modulated genes) or nadC overexpression itself (eight genes).
In fact, the induction of mRNA abundance by quinolinate (logarithmic values, Table S2) was relatively weak, except for some genes that encoded a functionally unknown aspartic-type endopeptidase (At5g48430), redox proteins such as peroxidase (At5g64100) and oxidoreductase (At2g36690), or ATPase AHA2 (At4g30190). Three cadmium-responsive genes (At1g14870, At1g14880 and At5g35525) were also activated, as well as a zinc transporter-relative gene (At1g05300). Fewer genes were repressed substantially by quinolinic acid. Two plant defensins (At5g44420 and At1g75830), and an acetylCoA N-acetyltransferase related gene (At2g39030) were transcriptionally down-regulated. Conversely, the latter gene was induced in conditions A and B.
Comparing conditions A and B (classes 1 and 2) revealed NAD-regulated genes. Most genes whose expression was modified under condition A had a similar expression pattern under condition B. We determined functional categories for genes involved in the NAD-regulated transcriptome responses using the MapMan® software from the HCL results (Data S1). NAD influenced various biological processes such as amino acids, hormone and cell wall metabolisms, photosynthesis, protein, RNA, transport, signalling, redox and responses to environmental conditions (stress). Nevertheless, it should be noted that within the latter, 34 genes (i.e. 10 %) were associated with PR proteins or transcription factors (TFs). Notably, RIN4 encoding RPM1-interacting protein 4 (At3g48450), PR1 (At2g14610) and GLIP1 (At5g40990) were highly induced, as well as other biotic stress-related genes encoding chitinases (At2g43570 and At3g47540), PR proteins (At1g52900, At5g43570, At1g72910 and At1g27180), TFs ATPEP1 (At5g64900), WRKY75 (At5g13080), or CYP81F (At5g57220), a gene involved in glucosinolate metabolism. In addition, the defence-related gene involved in nucleotide metabolism (At4g30530) as well as the AvrRPT2-induced AIG2 gene (At3g28930) and α-DOX1 (induced in response to SA and oxidative stress; De León et al., 2002) appeared to be induced in our microarray analysis. Besides biotic stress responses, genes encoding various calcium-binding proteins (At3g29000, At5g54490, At5g57010 and At5g39670) were also mildly activated by the NAD build-up, which is consistent with the role of pyridine nucleotides in cADPR signalling. Moreover, transcripts of SAG20 (Senescence Associated Gene, At3g10985), SAG13 (At2g29350) and SAG24 (At1g66580) accumulated in conditions A and B. These genes were also up-regulated in the old5 mutant (Schippers et al., 2008), which exhibits early developmental senescence and increased steady state levels of NAD.
As expected, and given the metabolic responses determined by metabolomics, several genes involved in primary and secondary metabolism were also regulated. In particular, the two β-glucosidase genes BGLU46 (involved in lignification, At1g61820) and BGLU30 (also named SRG2, senescence related gene 2 At3g60140) were drastically activated; however, BGLU27 (At3g60120) was moderately induced. Moreover, genes encoding anthranilate synthase beta subunits (At5g57890 and At1g24909), amino acid related l-asparaginase (At3g16150), indole-3-glycerol phosphate synthase (IGPS, At2g04400), ATATP-PRT1 (ATP phosphoribosyltransferase1), the cell wall-related EXTENSIN3 (At1g21310) or AtCSLG2 (cellulose synthase-like G2, At4g24000) were substantially enhanced. In addition, several genes that encoded redox proteins such as glutathione S-transferases (GST22 At2g29460, GST19 At2g29490, GSTU10 At1g74590, GST21 At2g29470), and ATMP2 (At3g48890) were induced. A few abiotic-stress responsive genes were also highly induced, especially At5g15970 (cold-responsive 6), At4g28088 (drought/salt-responsive) and nitrate-responsive NOI (nitrate induced) protein-related gene (At3g48450).
Conversely, several genes were substantially down-regulated as a consequence of elevated NAD, notably several photosynthesis-relative genes that encoded chlorophyll-binding proteins (At1g29910, At2g34430, At3g47470 and At1g29930), or cell wall metabolism-associated genes CESA6 and CESA1 (cellulose synthase isomer, At5g64740 and At4g32410). In addition, BGLU38 and BGL37, which are both involved in glucosinolate biosynthesis, the SA priming gene AZI1 (azelaic acid induced 1, At4g12470) and phenylalanine ammonia lyase PAL1 (At2g37040) were repressed.
To validate the transcriptional regulation patterns, we confirmed our CATMA results by quantitative RT-PCR analysis on a few genes (see Figure 4c–f) corresponding to a high or moderate level of transcripts combined with proteins that were identified (PR1, GLIP1) and a high level of transcripts combined with proteins that were not identified (At4g28088 and At5g46950). As shown above, PR1 and GLIP1, which were induced in conditions A and B from the CATMA data, as well as At4g28088 and At5g46950 were also up-regulated in the RT-qPCR assay.
Comparison of gene expression changes with other transcriptomic data
We took advantage of the entire CATMA database to compare our transcriptomic patterns to those obtained under other experimental conditions (Data S2 and S3). We used a hierarchical clustering approach and focused the analysis on conditions in which the NAD content was increased. Our conditions A (nadC+Q versus nadC) and B (nadC+Q versus Ctrl+Q) closely clustered with expression patterns associated with biotic stress (the project numbers cited thereafter are CATdb references, http://urgv.evry.inra.fr/CATdb/) such as treatments with harpin and salicylic acid (AU07-01_PLD-SA), or elicitor flagellin (RS06-06_miRNA) reported to interact with Avr-gene defence response (Navarro et al., 2004). Also, global expression profiles were highly similar to those obtained when plants were infected with pathogens (Erwinia amylovora: RA05-02_Erwinia; Botrytis cinerea: AF13_PLP2), or when plants exhibited LAR (Localized Acquired Resistance: TRI38-LAR). In the latter case, experiments dealt with the hypersensitive response (HR) after inoculation of the avirulent strain of Pst DC3000 carrying the gene AvrRpm1; these experimental conditions matched ours. With the quinolinate-modulated transcriptome (comparison of the database with condition C, Ctrl+Q versus Ctrl), the HCL indicated quite similar expression patterns with phosphatidic acid (PA) and diacyglycerol pyrophosphate (DGPP) treatments (second messengers in abscissic acid signalling: RS05-06_DGPP). Therefore the expression profiles we observed showed no similarity to biotic stress patterns.
Higher levels of NAD induce resistance to Pst-AvrRpm1
Given the transcriptomic induction of PR1 and other pathogen-related genes in the nadC line as a consequence of increased NAD levels, we investigated the response to infection with virulent and avirulent strains of Pseudomonas syringae pv. tomato (Pst). We infiltrated Ctrl and nadC 15.3 leaves with quinolinate. After 48 h, the same leaves were inoculated with either mock buffer (MgCl2) or the different strains of Pst, and the bacterial growth kinetics were monitored. No significant difference was observed in the nadC line subjected to the virulent strain Pst DC3000 (see Data S4). Both in Ctrl and nadC 15.3 lines, bacterial growth increased from 24 to 72 h post-inoculation (hpi), which indicated that basal defences are not modulated by the increase in NAD content. Nevertheless, when the avirulent strain Pst-AvrRpm1 was inoculated in nadC 15.3 leaves in the presence of quinolinate, bacterial growth was significantly lower 48 hpi and remained lower after 72 h (Figure 5a). To determine whether quinolinate had a possible toxic effect on Pst, we investigated both the bacterial growth of leaves treated with and without quinolinate. Figure 5(b) highlights reduced Pst growth in quinolinate-treated nadC 4.11 and 15.3 leaves, but no statistically significant difference was seen in Ctrl leaves with or without quinolinate. Therefore, the observed resistance to Pst-AvrRpm1 in both nadC lines was not due to quinolinate toxicity. Hence, overexpression of nadC promoted the resistance to the avirulent strain of Pst when the levels of NAD are increased while the basal defence response was not substantially affected.
To test whether the observed resistance was accompanied with elevated NAD pools, we monitored the pyridine nucleotide content in undetached Ctrl and nadC leaves in a time course following quinolinate addition and infection with Pst-AvrRpm1 (Figure 6). After quinolinic acid infiltration (from −48 to 0 hpi), no major variations in NAD+ and NADP+ leaf contents were observed in Ctrl as well as in non-infiltrated Ctrl and nadC leaves, except for a slight increase of NAD+ in quinolinate-infiltrated Ctrl (Figure 6a,c). However, NAD(P) contents rose in nadC leaves under quinolinic acid application at −24 hpi and reached a maximum (two-fold more than Ctrl+Q) at 0 hpi, corresponding to 48 h of quinolinate treatment (Figure 6a,c). After mock and bacteria application (24 hpi and 48 hpi), a downward trend was observed for NAD+ and, to a lesser extent, for NADP+ pools (Figure 6a,c). However, quinolinate-treated nadC leaves displayed elevated NAD+ levels at 24 and 48 hpi than those of Ctrl and non-infiltrated nadC. The NADP+ content only really plummeted in the nadC line from 0 to 48 hpi under mock or bacterial inoculation, which suggested that accumulated oxidised forms of NAD(P) are consumed in the time course (Figure 6c). In addition, reduced forms NADH and NADPH tended to increase in the time course and appeared higher in quinolinate-treated nadC leaves (Figure 6b,d). Remarkably, challenging Arabidopsis with Pst-AvrRpm1 caused elevated pools of NADH except for quinolinate-infiltrated nadC line, which presented stable contents following the infection (Figure 6b). Furthermore, NADPH pools accumulated upon bacterial inoculation and were drastically increased in quinolinate-treated nadC line at 48 hpi (Figure 6d). The observed alterations in the pyridine nucleotide content led to elevated NAD+/NADH and NADP+/NADPH ratios in nadC leaves when infiltrated with quinolinate (Data S5). In addition, bacterial infection diminished those redox ratios in the different conditions.
Adenylates are key nucleotides involved in various biosynthetic processes that could be affected by perturbations of NAD cellular pools, notably upon abiotic stress (De Block et al., 2005; Djebbar et al., 2012). Given the changes in NAD/P(H) levels previously observed, the ATP content was measured in a time course after quinolinate application and bacterial infection. Nonetheless, ATP quantifications did not display substantial modifications (Table S3).
Induction of PR1 and ICS1-dependent accumulation of the defence signal molecule SA
Defence responses including pathogen-related gene expression and SA levels in leaves inoculated with Pst-AvrRpm1 were further investigated. We also carried out targeted metabolite profiling focused on NAD derivatives at 48 hpi (the time at which we observed the higher resistance) using liquid-chromatography time-of-flight mass spectrometry (LCMS), as recently developed by Guérard et al. (2011). As already seen above with the metabolomic analyses, quinolinate addition caused an accumulation of molecules derived from NAD, such as NAM and nicotinic acid (Figure 7a,b). Consistently, RT-qPCR analysis showed that infection with Pst-AvrRpm1 led to increased transcript levels of AO, which is the first enzyme of NAD de novo synthesis from aspartate (Figure 1). As shown in Figure 7(d,e), we also observed the induction of PR1 and ICS1 (isochorismate synthase 1) during bacterial infection; this is in agreement with previous studies that showed induction of PR and SA-biosynthesis ICS1 genes during pathogen interaction (Glazebrook et al., 2003). Remarkably, there was a strong increase in the PR1 and ICS1 transcripts levels in quinolinate-treated nadC leaves compared with Ctrl leaves. The higher NAD content thus induced the expression of defence markers. In addition, total and free pools of SA rose after bacterial infection (Figure 8a,b), as expected in the Pst-Arabidopsis interaction (Glazebrook et al., 2003; Heck et al., 2003). In nadC leaves treated with quinolinate, the free SA content increased significantly (more than 2-fold) compared with that of quinolinate-treated Ctrl leaves (Figure 8b); this corresponded to the condition under which PR1 and ICS1 genes were much induced (Figure 7d,e). Glycosylated SA (G-SA) levels (calculated as Total SA minus free SA, Data S6) followed the same trend. We conclude that elevated NAD levels do influence SA turnover and biosynthesis via downstream ICS1 gene expression and induce defence markers like PR1 as a result.
The inducible plant system based on the overproduction of QPT from E. coli was used to investigate the effects of an increased intracellular NAD content. This study provides strong lines of evidence for the pivotal role of NAD in the defence response by stimulating SA synthesis and inducing defence markers such as PR1, resulting in enhanced disease resistance to incompatible bacterial infection (Pst-AvrRpm1).
Specific effects of NAD
Here, technical difficulties to manipulate the NAD content within leaf cells were overcome by the use of Arabidopsis thaliana lines overexpressing nadC, the gene coding for QPT. While we showed that NAD was higher upon quinolinate addition, other metabolites, such as nicotinate or several amino acids, also increased significantly. In fact, nicotinoid compounds have been shown to improve SA-related defence responses (Ford et al., 2010). However, we argue that the specific effects investigated here are likely due to NAD itself rather than other metabolic alterations. First, the nicotinate content also increased in Ctrl leaves, where no improved infection resistance was observed. Second, the metabolic alterations (increase in the Asp, Asn or Met contents in nadC leaf discs with nicotinate) are well explained by classical regulation effects of enzymes involved in amino acid metabolism (see below). We nevertheless recognize that Ctrl discs treated with quinolinate exhibited a visible decrease in Asp, Asn and Met pools. This effect may result from quinolinate toxicity (Hsu and Fahien, 1976), a process that was apparently accompanied by glutathione oxidation (Figure 3). Third, growth of Pst-AvrRpm1 was not affected by quinolinate at all (Figure 5), and only a few genes were transcriptionally affected by quinolinate itself (Figure 4 and Table S2). Fourth, our transcriptomic data based on the CATMA microarray technique used differential signals (comparisons such as nadC versus Ctrl, with or without quinolinate); thus, any specific quinolinate or nicotinate impact was taken into account. Furthermore, there was a very clear correlation between resistance to bacterial infection, NAD content and SA (Figure 8c). Hence, we believe that the effects of NAD discussed in the following sections are probably due to NAD build-up per se.
NAD-mediated orchestration of primary metabolism
Engineering NAD metabolism did alter primary C metabolism. The metabolomic analysis revealed elevated amino acid levels upon NAD increase, in particular Asp and its derivative Asn and Met (Figure 3). These results agree with previous studies of an NAD(H)-enriched mitochondrial mutant (Dutilleul et al., 2005) that showed substantial changes in metabolite profiles. Other authors also concluded that the chloroplastic NAD kinase stimulates carbon and nitrogen assimilation in Arabidopsis (Takahara et al., 2010). As stated before, Asp levels decreased when Ctrl leaves were treated with quinolinic acid, but it increased in nadC plants under the same conditions. Presumably, quinolinate addition decreased the consumption of Asp usually required for NAD synthesis. Asp could then be consumed by other metabolic pathways or be stored as Asn. The present situation differs from the CMSII mutant where a drastic reduction in the Asp level was observed and accumulated levels of Asn are noticed (Dutilleul et al., 2005). Under our conditions, the Asn content is probably under homeostatic control because the asparaginase 2 (At3g16150) gene was induced, and Asn overaccumulation in nadC leaves was avoided. In other words, the Asp pool was undoubtedly very dynamic in nadC leaves; we observed a larger consumption by both NAD and Asn synthesis (Asn synthetase) but also an increased resynthesis from Asn (asparaginase). The alteration of the Asp and Asn contents in the CMSII mutant might also reflect a pleiotropic effect of the mutation on asparaginase activity, which may in turn explain that discrepancy.
Furthermore, Asp may be redirected towards Asp-derived amino acids such as Met, Thr and Ile, which indeed accumulated in response to higher NAD. A similar pattern was observed for valine (Val) and leucine (Leu), the biosyntheses of which share the same enzymatic machinery and are associated with Ile. We noted that these metabolic changes were more significant in the nadC line with the highest NAD content (nadC 15.3). Levels of Asp-derived amino acids may be increased for two major reasons. First, Asp was less committed to NAD biosynthesis because of quinolinate application and was thus available for other amino acid synthesis. Second, higher levels of NAD(H) and NADP(H), which act as co-factors of bifunctional Aspartate Kinase-Homoserine Dehydrogenases, may have stimulated the synthesis of homoserine and its derivatives Met, Thr and Ile. In fact, stimulating Thr production in plants results in a simultaneous increase of the Ile pool (Gakière et al., 2000; Wang et al., 2007). The Met that was overproduced may be catabolised and this represents the source of Ile accumulation (Ravanel et al., 1998; Rébeilléet al., 2006). Nevertheless, levels of lysine (Lys), another Asp-derived amino acid, were unchanged under our conditions. This reflected the tight control exerted by Lys on its own synthesis and catabolism (Blicking and Knäblein, 1997; Curien et al., 2007). Interestingly, higher amounts of glutamate (Glu), the final product of the lysine catabolism pathway, are observed in NAD-enriched leaves. It should be noted that Glu can also be synthesised with Asp as an amino donor under dark conditions, via aspartate-aminotransferase activity (Miyashita and Good, 2008).
The abovementioned amino acids probably interacted with reactive oxygen species (ROS). Met is highly susceptible to oxidative damage, resulting in changes in the structure or the activity of various proteins (reviewed in Davies, 2005). The oxidation of Met to methionine sulfoxide can be reversed by methionine sulfoxide reductases (MSR, Rouhier et al., 2006), enzymes that reduce specific Met-R-sulfoxide. In the present work, microarray analysis showed that two MSR genes were activated (MSRB3, At4g04800 and MSRB8, At4g21840) while the Met content increased (Figure 3). Recent studies (Oh et al., 2005, 2010) have reported that treatments with ABA, NaCl and cold increased the transcript abundance of the secreted ATMSRB3. It is plausible that oxidative reactions did occur under our conditions, as suggested by the decrease of the GSH/GSSG ratio in Ctrl leaves incubated with quinolinate. In nadC lines in which NAD accumulated, such an oxidative stimulation was accompanied by the induction of redox-responsive genes (peroxidase, oxidoreductases, MSR), with no net redox effect.
We also found a build-up of the xylose content in the presence of quinolinate, mostly in nadC leaves (Figure 3i). Xylose is a pentose that may have been involved in providing the phosphoribosyl moiety for NAD biosynthesis. We hypothesise that, as a residue of hemicellulose, xylose could have interacted with cell wall metabolism, which is NAD(P)H dependent. Consistently, there was an induction of the cellulose synthase-like gene AtCSLG2 and NADP(H) levels. Defence-related cell wall modifications could also be involved and require precursors such as xylose to rearrange parietal structures (Liepman et al., 2010). Finally, increased xylose levels may also reflect the stimulation of 2,3-DHBA (dihydroxybenzoic acid) and SA metabolisms, both of which are involved in the establishment of systemic acquired resistance (Bartsch et al., 2010).
NAD+ has long been considered as the common electron carrier in energy flows including photosynthesis and cellular respiration (De Block and Van Lijsebettens, 2011), and various studies have reported that the ATP content can decrease in response to enhanced biosynthetic processes (De Block et al., 2005). For instance, in the CMSII mutant, increased levels of NAD(H) are associated to higher ATP contents (Djebbar et al., 2012). This mitochondrial tobacco mutant also harbours enhanced tolerance to water stress, which correlates with a depletion of ATP. In our study, though, the higher pool of pyridine nucleotides did not link to accumulated ATP, and bacterial infection did not cause any decrease in ATP levels.
NAD-mediated stimulation of Pst-AvrRpm1 resistance is correlated with SA accumulation
Systemic acquired resistance is a well documented systemic immune response and involves an oxidative burst, accumulation of the defence signal molecule SA and expression of PR genes such as PR1, PR2, and PR5 in systemic tissues (Malamy et al., 1990; Ryals et al., 1996). SA can act together with nitric oxide (NO) and ROS to induce cell death, thus mediating the HR (Delledonne et al., 1998). Our transcriptomic data indeed highlight the involvement of redox-related genes usually expressed upon HR conditions: α-DOX1 (induced in both incompatible and compatible bacterial infections; De León et al., 2002) and several GSTs (GST22, GST19, GSTU10, and GST21), which may participate in the enzyme-catalysed peroxidation of GSH (Durrant and Dong, 2004). However, no substantial changes were observed in GSH/GSSG ratios under our conditions (see above).
In the present study, we established that engineering Arabidopsis with increased NAD content caused SA-correlated plant defence responses. This conclusion is supported by different lines of evidence. First, transcriptomic analysis revealed a highly pathogen-related NAD effect similar to that of SA (HCL in the CATMA database) and a considerable induction of PR1 during Pst infection (Figure 7d). In fact, PR gene expression is known to be associated with enhanced disease resistance (Ryals et al., 1996; Wagner et al., 2002). Second, we observed increased resistance to the avirulent pathogen Pst-AvrRpm1 in the two quinolinate-treated nadC lines (Figure 5) but not to the virulent strain Pst DC3000 (Data S4). Third, endogenous biosynthesis of SA was enhanced via higher expression of the SA-biosynthesis gene ICS1 (Figure 7e) (Wildermuth et al., 2001), resulting in the accumulation of total and free pools of SA 48 hpi (Figure 8).
The NAD(P) pools we measured in the time course following quinolinate treatment on undetached leaves were similar to those obtained with dark-incubated foliar discs. In fact, 48 h after quinolinate application, the NAD+ and NADP+ contents reached a maximum in nadC lines. However, we noticed that redox balance was not affected in dark conditions (discs, Figure 3) whereas it appeared slightly disturbed in illuminated leaves (Figure 6, Data S5). Our findings are consistent with other studies, which have reported that light induces the reduction of pyridine nucleotides pools in different plant species (Heber and Santarius, 1965; Hunt et al., 2004). In addition, we observed an accumulation of reduced forms NADH and NADPH following the infection with Pst-AvrRpm1 (Figure 6). It is worth mentioning that mutations in NADH hydrolase activity (i.e. atnudt7, atnudx6 mutants) cause perturbations of NADH homeostasis (Ge et al., 2007; Ge and Xia, 2008; Ishikawa et al., 2010; Jambunathan et al., 2010). Higher pools of NADH are linked to the stimulation of defence responses, such as NPR1-dependent and -independent genes induction (i.e. PR1 and ICS1) and SA accumulation, and to enhanced resistance to inciting agents including pathogenic and nonpathogenic micro-organisms. In the present study, the increased bacterial resistance (48 hpi) is correlated with elevated SA contents, induction of PR1 and ICS1, and accumulated levels of NAD(P)H. While NADH and NADPH levels accumulated upon infection (Figure 6), quinolinate-treated nadC leaves presented elevated pools of NADPH and stable but high contents of NADH, which in turn may reflect the consumption of NADH required for resistance to Pst-AvrRpm1. Accumulated NADPH levels may act as substrate to ROS production by NADPH oxidases, for instance (Torres et al., 2006). Alternatively, the NAD+/NADH ratio was always higher in quinolinate-treated nadC leaves compared with those of other conditions of the infection time course (0 to 48 hpi, Data S5). This particular result further suggests the importance of redox balance upon biotic stress (Ge et al., 2007; Ge and Xia, 2008).
Exogenous application of pyridine nucleotides in Arabidopsis causes Ca2 + -mediated induction of PR genes, accumulation of SA and resistance to Pseudomonas syringae pv. maculicola ES4326 (Zhang and Mou, 2009). Other studies suggested that NAD-derived cADPR participates in cell signalling (including genes regulation) in response to abscissic acid (ABA) (Wu et al., 1997; Sanchez et al., 2004), or in the induction of phenylalanine ammonia lyase (PAL) and PR1 (Durner et al., 1998). The transcriptomic pattern we observed here revealed the expression of several genes implicated in calcium signalling in response to NAD; however, the expression of PAL1 was substantially down-regulated. When Arabidopsis rosettes are sprayed with neonicotinoids (i.e. synthetic analogues of the natural insecticide nicotine), plant defence responses are induced; the corresponding transcriptomic profiles are similar to those obtained with SA and including genes implicated in (a)biotic stress and SAR because they result in reduced growth of the powdery mildew pathogen (Ford et al., 2010). Alternatively, a neonicotinoid metabolite could function as a NAM analogue in the NAD recycling pathway and increase NAD(P) to promote stress tolerance against drought or pathogens as previously reported (Thielert et al., 2009). Here, the nadC 15.3 line had a higher NAD content during bacterial infection. This situation gave rise to SA-associated defence responses and elevated contents of both NAM and nicotinic acid. Quinolinate-treated Ctrl leaves also showed enriched NAD derivatives but had neither significant NAD accumulation, SA-enrichment nor reduced growth of Pst. Therefore, the accumulated NAD derivatives were not responsible for disease resistance. Instead, we believe that NAD itself probably triggered the plant defence responses.
Besides participating in SA-dependent signal transduction, elevated NAD levels could interact with other defence responses such as ethylene (ET) signalling. Indeed, it is worth mentioning that GLIP1 is greatly induced when NAD levels are high. In plants, GLIP1 is thought to belong in SA-independent defence responses, by eliciting the ET pathway and increasing resistance to several pathogens (Kwon et al., 2009). For instance, CEJ1 (cooperatively regulated by ethylene and jasmonate, At3g50260) and the JA responsive gene JAZ8 (At1g30135) are induced in our microarray data.
As redox dynamics in NAD-enriched leaves (both oxidation and MSR enhancement) are probably complex, it would be pertinent to study GSH and other antioxidants further upon Pst infection. In addition, given the crucial role of ROS in plant defence responses (Torres et al., 2006), ROS analysis should be carried out to characterise further the oxidative response that underpins resistance during the incompatible infection of Arabidopsis with Pst-AvrRpm1. Similarly, the occurrence of cell death could be measured by electrolyte leakage experiments.
Nevertheless, our results show that NAD is involved in biotic stress and correlates with SA-mediated defence responses leading to resistance to avirulent pathogen. It is plausible that high pools of NAD trigger defence markers that elicit resistance during incompatible infection and/or promote intracellular cADPR/Ca2+-dependent signalling. Here, we used transformed plant lines with an increased QPT capacity. We recognize that other important steps may control NAD levels in plant cells. For instance, AO (Figure 1), which catalyses the first irreversible step of NAD synthesis, seems to be finely transcriptionally regulated, as reported by Genevestigator and our study (Figure 7). We did not focus on this committed step of NAD biosynthesis here, but we will address this aspect in a subsequent paper.
In the present study, the answer to whether NAD itself triggers SA pathway remains unclear. However, our NAD overproduction lines provide an efficient system for the plant pathogenic research community to further investigate how NAD levels are switching on defence mechanisms. Crossing the nadC lines with SA signalling (i.e. npr1 mutant) or modulating SA levels (sid2-NahG) mutants is required further to provide a causative role between the NAD build-up and a possible SA-associated resistance.
All reagents were purchased from Sigma Aldrich. Quantitative RT-PCR analysis reagents, glutathione reductase and glucose-6-phosphate dehydrogenase were obtained from Roche Diagnostics.
Plant materials and growth conditions
Arabidopsis thaliana transgenic lines used throughout this study were derived from Arabidopsis plants of the Columbia (Col-0) ecotype. Seeds were sown either on agar or in soil and stratified at 4°C for 2 days. Seedlings were germinated and grown in a controlled-environment growth room under a day/night regime of 8 h/16 h (SD) or 16 h/8 h (LD), an irradiance of 200 μmol quanta m−2 s−1 at leaf level, at 18–20°C and 65% humidity. Nutrient solution was given twice a week. Leaf samples were taken in the middle of the photoperiod, rapidly frozen in liquid nitrogen and stored at −80°C until further analysis. For metabolomic and transcriptomic analyses, plants were analysed and sampled at the age of 5 weeks (SD), after 1 week (LD) and at the age of 5 weeks (SD) for pathogen tests and derived analysis. Unless otherwise stated, data are the means and SE of three to five independent samples from different plants, and significant differences are expressed using a Student’s t-test at P <0.05. All experiments were repeated at least three times and yielded similar results, except for the microarray analysis that was repeated twice.
Generation of transgenic plants
The full-length cDNA of nadC from Escherichia coli was subcloned in frame into the Pst1 and BamH1-sites of the binary vector PCW162, which was kindly provided by Marie-Pascale Doutriaux, under the control of the double CaMV35S promoter and providing kanamycin resistance. The resulting plasmid was used for stable transformation of Arabidopsis plants to overexpress the nadC gene, using the Agrobacterium tumefaciens strain GV3101. Primary transformants were selected on Murashige & Skoog medium that contained 50 mg L−1 kanamycin monosulfate. Figure S1 seeds from transformed plants were surface sterilized and germinated on the medium. The number of putative transgenic resistant plants was recorded. For each selection regime, a total of 30 explants were plated in three replicate experiments. All experiments were carried out in a 25°C growth room under a constant irradiance of 200 μmol quanta m−2 s−1 at the leaf level. The expression of the bacterial gene was examined by DNA gel blots. We used genomic DNA isolated from Arabidopsis homozygous transgenics and the following primers sequences: NadCfwd (ATG TTT CTA CCT TAT GAT TCG T) and NadCrev (TTA GCG AAA ACG CAT TGA AAG).
QPT activity measurements
To quantify quinolinate PRTase activity, we used the method developed in Mainguet et al. (2009) with the following modifications: the QPTase assay was based on the formation of NaMN (see Figure 1). The QPT activity was measured in a volume of 100 μl containing 100 mm Tris–HCl (pH 7.5), 20 mm MgCl2, 2 mm quinolinic acid and 2 mm 5-phospho-d-ribose-1-diphosphate (PRPP).
Inducible cellular system to accumulate NAD contents
Leaves of the homozygous vector control and nadC 15.3 and 4.11 lines were used at the age of 5 weeks (SD) plus 1 week (LD) to obtain foliar discs of 7 mm diameter. Incubation was carried out in 50 mm MOPS (pH 6) with or without 100 μm quinolinic acid in the dark for 48 h to reach NAD concentrations that increased by three- to four-fold. The discs were then taken out of the buffer and rinsed with distilled water. The fresh weight (FW) of 10 discs corresponding to about 100 mg FW was noted, and samples were directly frozen in liquid nitrogen and stored at −80°C until further analysis.
RNA isolation and real-time qPCR
Total RNA was isolated using the NucleoSpin RNAII kit (Macherey-Nagel) according to the supplier’s instructions. One μg of total RNA was used as the template for first-strand cDNA synthesis and reverse transcribed with the SuperScript III First-Strand Synthesis System (Invitrogen). We performed quantitative RT-PCR either to confirm CATMA gene expression modulation or to quantify targeted gene transcripts. All reactions were performed in quadruplicate on separate plates. Relative expression levels were normalized using Ct values obtained for the housekeeping gene ACT2 (At3g18780). The following primer sequences were used for RT-qPCR analyses: ACT2 (LP-CTGTACGGTAACATTGTGCTCAG, RP-CCGATCCAGACACTGTACTTCC); AO At5g14760 (LP-GATCGTTCACGCTGCTGATA, RP-TGTGTTCAAGCCATCCTGAG); PR1 At2g14610 (LP-AATCGTCTTTGTAGCTCTTGTAGGT, RP-ACCCTTAGATAATCTTGTGGGCTAT); ICS1 At1g74710 (LP-GCTCTATCATTTCCATCTCTCGTAG, RP-CATTCATAGACATCGAACATGACTC); GLIP1 At5g40990 (LP-ATTCAAATACGCCCTTCACG, RP-GGCCTCCACACGTATTGATT); At4g28088 (LP-CCCTCCTCTAGGGGTTTGTC, RP-TCACGATCCACGTACACGAT); At5g46950 (LP-AAGACCCGCAATCGTCATAC, LR-GAGTAGTTGCGGCTTTGCTT).
Microarray analysis was performed using the CATMA database version 5 arrays that contained 34 648 probes corresponding to 17 622 annotated genes from Arabidopsis (Sclep et al., 2007) plus 1259 probes corresponding to 521 microRNA genes or siRNA genes (information available at http://urgv.evry.inra.fr/projects/FLAGdb++). This array resource has been employed in 40 publications over the last 5 years (Achard et al., 2008; Besson–Bard et al., 2009; Krinke et al., 2009). A regular protocol based on technical repeats of two independent biological replicates produced from pooled material from independent plants was adopted in this study. For each biological repeat and each time point, RNA samples were extracted from 100 mg of fresh material collected from two foliar discs of two different plants, using NucleoSpin RNAII kits (Macherey-Nagel) according to the manufacturer’s protocol. For each comparison, one technical replication with fluorochrome reversal was performed for each biological replicate (i.e. four hybridisations per comparison). The labeling of complementary RNAs with Cy3-dUTP or Cy5-dUTP (Perkin-Elmer-NEN Life Science Products), the hybridisation to the slides and the scanning were performed as described previously (Lurin et al., 2004).
Statistical analysis of microarray data
Experiments were designed together with the statistics group of the Unité de Recherche en Génomique Végétale. The methods are available in the R package ‘Anapuce’ (http://cran.r-project.org/web/packages/anapuce/index.html). The normalization and statistical analysis were based on two dye swaps (i.e. four arrays) per comparison (Gagnot et al., 2008). First, one normalization without background substraction per array is performed to remove systematic biases. Then, a global intensity-dependent normalization is performed using the lowess procedures (Yang and Thorne, 2003) to correct the dye bias. Finally, for each block, the log-ratio median calculated over the values for the entire block is subtracted from each individual log-ratio value to correct effects on each block, as well as print-tip, washing and/or drying effects. To determine differentially expressed genes from a dye-swap, a paired t-test is performed on the log2 ratios, with a common variance for all the genes (H homoscedasticity), leading to a robust estimation of the variance and a high power of the test. Spots with an extreme variance or genes for which only one observation is available are excluded. Then, the raw P-values are adjusted by the Bonferroni method, which controls the Family Wise Error Rate (Ge et al., 2003). Genes with a Bonferroni P-value ≤0.05 were considered differentially expressed, as described in (Gagnot et al., 2008). Clustering was performed using MultiExperimentViewer; functional categories and Venn diagram were generated with MapMan® software.
Antioxidants were analysed as previously described (Queval and Noctor, 2007) with the following modifications: the final pH for the NAD/NADP sample was between 6–7 and between 5–6 for GSH/GSSG and ascorbate samples; PMS was used at a concentration of 10 mm for the colorimetric assay. Gas chromatography coupled to time-of-flight mass spectrometry (GC-TOF-MS) and quantitative detection of amino acids by HPLC, were performed as reported previously (Bathellier et al., 2009; Tcherkez et al., 2010). Targeted nucleotides were detected using a LCMS method developed by Guérard et al. (2011). Total and free pools of SA were analysed at 48 hpi as previously described (Simon et al., 2010). ATP contents were quantified as described in Djebbar et al. (2012) using the ENLITEN® ATP Assay System Bioluminescence Detection Kit (Promega).
The virulent Pseudomonas syringae pv. tomato DC3000 strain and the avirulent strain carrying the avirulence gene AvrRpm1 (Pst-AvrRpm1 obtained from Jane Glazebrook, University of Minnesota, St Paul, USA) were used for resistance tests in a medium titer of 105 colony-forming units ml−1 (A600 = 0.0001). Whole leaves of 5-week-old plants grown under an 8 h (SD) regime were infiltrated with 5 mm quinolinate (pH 6), using a 1 mL syringe with no needle; 48 h later, the same leaves were infiltrated with 10 mm MgCl2 or a bacterial suspension. Leaf discs of 0.5 cm2 were harvested from inoculated leaves at the appropriate time points. For each time point, four to six samples were collected by pooling two leaf discs from different treated plants. Bacterial growth was assessed by homogenizing the leaf discs in 400 ml of sterile water, plating appropriate dilutions on solid lysogeny broth (LB) agar medium containing rifampicin and kanamycin, and recording the number of colonies after 2 days. All experiments were repeated at least three times and gave similar results.
Pierre Pétriacq and Linda de Bont are funded by the University of Paris-Sud 11 (Doctoral school Sciences du Végétal ED145, Orsay, France). Jutta Hager was funded by ANR Redoxome project, under Graham Noctor’s direction. The authors thank Marie-Pascale Doutriaux (Institut de Biologie des Plantes, Orsay, France) for kindly providing the overexpressor vector PCW162 and Marie Garmier (Institut de Biologie des Plantes, Orsay, France) for helpful discussions on plant immunity. Rajsree Mungur (Berkeley, USA) and Michael Mysliwieck (Detroit, USA) are also gratefully acknowledged for correcting the manuscript.
BG conceived and produced the transgenic lines and experiments, and supervised the study. GN emitted the idea to overproduce NAD in a plant system. JH obtained the inducible NAD overproduction system. CM and FG were involved in metabolomics. SP and JPR performed the microarray analysis experiments. PP analysed the CATMA data with JPR. LD and PP conceived the pathogen tests. PP performed all the experiments on the transgenic lines and wrote the manuscript. LdB contributed to metabolic measurements and pathogen tests. GT was involved in discussing the results and correcting the manuscript.