Wild accessions of sugarcane (Saccharum spp. interspecific hybrids) progenitor Saccharum spontaneum L. were used for this study. They were sourced from the Chinese S. spontaneum collection maintained in the germplasm garden of Sugarcane Research Institute, Guangxi Academy of Agricultural Sciences (SRI), Nanning, Guangxi, China. Nanning has a hot humid subtropical climate with annual temperatures ranging between 2°C and 35°C. It receives, on average, 1300 mm rainfall yearly with an annual mean humidity of 79%. For this study, thirty-three S. spontaneum accessions collected from very diverse tropical and sub-tropical agroclimatic conditions of Southern China, where sugarcane is grown commercially, were selected. All the selected accessions were free of pests and diseases, never fertilised, and grew well in the garden.

Ten-month-old S. spontaneum plants were used for population screening experiment. Plants selected for the screening experiment were grown individually in large pots with soil collected from SRI germplasm garden in a naturally-lit glasshouse. They were not fertilised or sprayed with any chemicals prior to and during the screening experiment. The soil was irrigated as and when needed. The plants of all 33 accessions were divided equally into three blocks (six to eight plants of each accession in one replicate block), and in each block, they were arranged randomly. These accessions were screened for BNF activity using nitrogenase assay, and plant height (from soil surface to the dewlap of the youngest fully expanded leaf), stalk number and brix (using juice expressed from 10 cm of the basal part of stem) were also determined. Brix of expressed juice was measured by refractometry using ATR-P Refractometer (Schmidt and Haensch, Germany). For each accession there were six independent measurements from plants randomly selected from 3 replicate blocks.

Six accessions with high BNF activity identified in the BNF screening experiment, namely G03, G152, G177, G720, G824 and G1926, were selected for studying the sensitivity of N fixation to high N condition. Single node cuttings from 12-month-old healthy plants were planted in unfertilized moist garden soil in plastic trays with perforated bottom and kept in a naturally-lit glasshouse for sprouting and plantlet development. One month after planting the cuttings, plantlets of uniform size were transplanted into 30 lit pots with unfertilized soil. All pots were watered regularly to avoid moisture stress. Two weeks after transplanting, potted plants were divided into two equal groups; one for high nitrogen (HN) and the other for low nitrogen (LN) treatments. All HN plants were supplied with Murashige and Skoog (MS) mineral nutrients (2 lit pot-1) (Murashige and Skoog, 1962), which contained 18.8 mM KNO₃ and 20.6 mM NH₄NO₃, while those in LN received MS mineral nutrients without N (2 lit pot-1) once every three weeks. This N treatment was continued during the 5-month experimental period. The experiment followed a completely randomized block design with three replicated blocks. For both treatments, there were 5 plants of each accession in a single block (replicate). At the end of the experiment, middle portion of the youngest fully-expanded leaf was sampled and nitrogenase activity was determined as described above. For each accession, three biological replicates from each block were used for the enzyme assay.

Two high-BNF S. spontaneum accessions with high brix, good stalk development, high stalk number, high propensity for flowering and a relatively low impact of applied inorganic N on their N fixation activity, namely G152 and G3, were selected for further physiological and molecular characterization. Plants were raised and the experiment was conducted as described in section 2.2. except for the following conditions. Experiment was continued for 7 months. The experiment followed a completely randomized block design with three replicated blocks for each treatment. Each block had 6-8 plants of each accession for HN and LN treatments.

At the end of the experiment, leaf tissue from the youngest fully expanded leaf and root samples were harvested for measuring nitrate reductase (NR) and glutamine synthetase (GS) activity, and total N, NH₄-N, soluble protein and soluble sugars content. Leaf tissue was also used for BNF assay and transcriptome analysis. For NR and GS activity, lamina (1 gm) from the youngest fully-expanded leaf and root tips (10 mm from the tip) were collected and immediately snap-frozen in liquid nitrogen. The tissue samples were collected between 09.30 and 13.00 hours to minimise the effect of diurnal variation.

At the end of seven months of growth, shoot height and stalk number of experimental plants were measured. For chemical analyses, oven dried (80 ⁰C for 5 days) leaf and root tissues were finely powdered and used for total N, NH₄-N and soluble sugars content measurement. Total N was determined by acid digestion of samples following Kjeldahl method. NH₄-N and soluble sugars content were measured using Plant Ammonium Nitrogen Activity Assay Kit (# BC1520, Solarbio, China), and Plant Soluble Sugar Content Assay Kit (# BC0035, Solarbio, China), respectively, following manufacture’s instruction. For each treatment, eight independent plants (measurements) randomly selected from different replicated blocks were used for analysis.

All data presented here are analysed using analysis of variance (ANOVA) in Genstat statistical system, 19th Edition (VSNI, 2017).

Analysis of nitrogenase activity in S. spontaneum accessions showed remarkable variation for BNF capacity among the clones tested ( Figure 1 ). For example, accession G152, the clone with the highest BNF activity, recorded 72-fold greater enzyme activity than the poorest performer, G103. Among the 33 clones tested, 15 had a relatively high BNF activity (>50 nmol C2H4 mg-1 protein h-1). A similar trend was also evident for plant height, stalk number and brix ( Figures 2 , 3 ) although the range of variation for these traits was much smaller than that of BNF ( Figure 2 ). On average, a six-fold variation was evident for plant height, stalk number and brix at the end of the experiment.

Analysis of nitrogenase activity in six high BNF accessions grown under externally supplied inorganic N showed remarkable genotypic variation for BNF activity ( Figure 3 ). External application of inorganic N had a relatively low impact on the nitrogenase activity (20-25% reduction) of accessions G152 and G3, whereas the activity was reduced by 42-53% in other four accessions tested. The N-induced reduction in nitrogenase activity was significant for all accessions (p<0.01)

External application of inorganic N significantly (p<0.01) increased shoot length and stalk number in both accessions ( Table 1 ). Under external N supply, stalk length was increased by 19% in G3 and 22% in G152, whereas a more remarkable rise in stalk number, 46 – 50%, was observed in both accessions. Similar to shoot growth, total N and NH4-N content of leaf and root tissues of G152 and G3 accessions also showed a substantial increase when plants received external N supply ( Figures 4A, B ). Total N content variation was more pronounced in roots than in leaf tissue for both accessions, but that was not the case for NH4-N; there was very little variation for root NH4-N in G152 and G3plants. Compared with total N content, N-induced increase in soluble sugar was markedly lower in leaf and root tissues except for G3 leaf tissue ( Figure 4C ).

From the six high-BNF accessions tested for their BNF-response to externally applied inorganic N ( Figure 3 ), two accessions with BNF least affected by external N and desirable agronomic features as well as flowering propensity, G152 and G3, were selected for more detailed physiological and molecular analyses. Consistent with the results of the previous experiment ( Figure 3 ), external application of N significantly reduced (19-23%; P<0.001) the nitrogenase activity in both accessions ( Figure 5 ).

Activity of NR in both leaf and root tissues of G152 and G3 accessions was substantially increased with the application of inorganic N ( Figure 6A ; p<0.001). The percentage increase in roots was much greater (2.4 to 3.5-fold) than that of leaf, and in both accessions rate of enzyme induction in leaf tissue was somewhat similar (~1.5 fold) ( Figure 6A ). As with NR, GS activity also showed an increasing trend in leaf and root tissues though the N-induced enzyme induction was not as pronounced as in NR, except for the G3 leaf tissue ( Figure 6B ).

Inorganic N application had a remarkable effect on endogenous hormones in both G152 and G3 accessions ( Table 2 ). The endogenous level of growth promoting hormones IAA, GA3 and ZR in both accessions increased by 60-80%, except for a doubling of ZR in G152. No consistent and significant change in ABA content was noticed with N application.

In the RNA-Seq experiment, the number of clean reads obtained from each library after trimming and filtering ranged between 44725188 and 481766948, with average 6.83 (range 6.09-7.07) gigabases, 93.6% Q30 (range 91.82-94.4%) and 53.5% GC (range 50.48-55.07%) content. Out of 266228 transcripts obtained, 110947 unigenes with a mean length of 1087 bp (range 201- 17867 bp) were identified ( Table S1 ).

The function of all unigenes obtained were annotated using seven databases. ( Table S2–S4 ). In this analysis, transcriptome assembled from RNA-Seq data was used as reference. Out of a total of 110947 unigenes identified, 64.9% (72056), the largest proportion of unigenes, were annotated from NT, followed by 67404 (60.8%) in NR. The lowest number of genes annotated, 8435 (7.6%), was in KOG. It is important to note that 76.6% (84986) of all the unigenes identified were annotated at least in one of the seven databases used in this study.

In the gene expression study, the impact of external application of N on the expression of genes associated with carbon, N, amino acids and hormones in leaf tissues of two high-BNF S. spontaneum accessions, G152 and G3, was analysed. Four different pair-wise comparisons, i.e., same accession with contrasting N treatments, high N and low N, comparison (G152 HN vs G152 LN; G3HN vs G3 LN) and different accessions with the same N treatment comparison (G152 HN vs G3 HN; G152 LN vs G3 LN), were performed to identify DEGs ( Figure S1 ). In the initial global analysis of differential gene expression using the threshold of an adjusted p value of <0.05 and log2FoldChange >1 based on DESeq2 method, 864 genes were found to be differentially expressed between G152HN and G152LN with 360 genes downregulated and 504 upregulated ( Figure S1 ). For the same comparison (HN vs LN) in G3, the number of DEGs was less than half of what was found in G152 (HN vs LN), and more genes were downregulated (155) than upregulated (87). In contrast to treatment comparison (HN vs LN for the same genotype), the genotype comparison (G152HN vs G3HN, G152LN vs G3LN) showed a substantially large number of DEGs, >2000 in each comparison, irrespective of the treatment ( Figure S2 ). However, the pattern of gene expression in genotype comparison under high N treatment (G152HN vs G3HN) was just opposite of low N treatment (G152 LN vs G3 LN) with more downregulated DEGs under HN while the opposite was true for LN.

In order to identify common and unique genes that expressed differentially in response to N application in G152 and G3 accessions, Venn diagrams of DEGs were prepared ( Figure S2 ). There were 28 common DEGs among G152 and G3 for treatment (HN vs LN) comparisons ( Figure S2A ), In this comparison (HN vs LN), G152 accession showed four-fold greater DEGs than that of G3 ( Figure S2A ). However, when G152 and G3 accessions were grown under similar soil N condition ( Figure S2B ), 1101 common DEGs were found. And, a similar number of DEGs was also found for genotype comparison under HN and LN conditions ( Figure S2B ). Interestingly, there were no common DEGs across all pair-wise comparisons ( Figure S2C ).

To determine the potential functions of DEGs identified in this study, we performed gene ontology (GO) enrichment analysis and they were grouped into two main GO functional categories- biological processes and molecular functions. There was large variation for functional classes between all four pair-wise comparisons ( Figure 7 ). For instance, DEGs identified in G152 HN vs LN comparison were mostly mapped to regulation of proteolysis, peptidases, catalytic activity, amino acid, protein and carboxylic acid metabolism and membrane transport GO terms in biological process and molecular functions categories combined ( Figure 7A ). Most of the DEGs in this comparison were upregulated with transmembrane transport and amino acid biosynthesis being the most enriched ones. In contrast, DEGs found in G3HN vs LN comparison were enriched for single-organism process, fatty acid, lipid and carboxylic acid metabolism, and anion homeostasis for biological processes, and fatty acid and iron binding for molecular functions ( Figure 7B ). Another interesting observation noted in this analysis was that when both accessions grown under the same N supply conditions (HN or LN) were compared, a substantially greater number of DEGs was highly enriched in samples from HN than those from LN ( Figures 7C, D ). Also, in the genotype comparison under HN condition, DEGs were mostly up-regulated whereas an opposite trend was true for LN plants ( Figures 7C, D ). Under LN condition, highly enriched DEGs were attributed to phosphorylation and metabolism of phosphorous-containing compounds GO terms in biological processes category, while protein kinase, transferases form the most enriched GO terms in molecular functions category. In the comparison of genotypes grown under HN, highly enriched DEGs identified were mapped to phosphorylation and DNA metabolic processes for biological processes, and nucleotide binding, kinases, carbohydrate derivative-binding, phosphotransferases, and anion binding, for molecular processes ( Figures 7C, D ). Also, it is interesting to note that under LN growth condition, though only minimally enriched, cytokinin biosynthesis was up-regulated ( Figure 7C ).

In order to gain more insights into the potential metabolic roles of genes differentially expressed in G152 and G3 accessions, they were mapped to various metabolic pathways in KEGG database ( Figure 8 ). The genes differentially expressed in G152 in response to external N supply were mostly mapped to metabolism category with amino acid metabolism and carbohydrate metabolism being the most enriched pathways followed by lipid, energy, other amino acids, terpenoids and polyketides metabolism and biosynthesis of secondary metabolites ( Figure 8A ). Transport and catabolism, and signal transduction in cellular processes and environmental information processing categories, respectively, were the other significant pathways identified in G152 HN vs G152 LN comparison ( Figure 8A ). As with G152, DEGs identified in G3 HN vs LN comparison were mostly over-represented in metabolism category ( Figure 8B ). However, in G3, DEGs were most enriched for lipid metabolism pathway followed by biosynthesis of secondary metabolites and carbohydrate and amino acid metabolism pathways. Transport and catabolism, membrane transport and signal transduction pathways were also significantly enriched.

Unlike the large variation in representation of DEGs in various KEGG metabolic pathways seen between G152 and G3 grown under contrasting N supply (HN vs LN comparison) ( Figures 8A, B ), there were remarkable similarity for enrichment of DEGs in G152 and G3 grown under same N level (HN or LN) ( Figures 8C, D ). Carbohydrate metabolism was the most over-represented category in this genotype comparison, followed by amino acid, energy, lipid and nucleotide metabolism, and biosynthesis of secondary metabolites. Significantly, DEGs mapped to component pathways of genetic information processing category and that of transport and catabolism, and environmental adaptation, showed greater representation in genotype comparison ( Figures 8C, D ) than in treatment comparison ( Figures 8A, B ).

Expression of DEGs involved in amino acid, carbon, N and hormones were further studied to gain more insights into metabolic responses of G152 and G3 accessions to inorganic N fertiliser application ( Figure 9 ). A large number of amino acid metabolism-related genes were differentially expressed in response to external N supply in both accessions ( Figure 9 ; Table S5 ). Also, there was substantial genotypic variation for this gene expression response. For instance, many amino acid metabolism-related genes that are up-regulated in G152 under high N condition were either down-regulated or remained unchanged under low N, and vice versa ( Figure 9 ). This include glutamine synthetase, asparagine synthase, alanine transaminase, glutamate synthase, tryptophane synthase, etc. A similar result was also found for G3 grown under HN and LN conditions ( Figure 9 ; Table S5 ). Several genes up-regulated in G3 under HN, such as alanine transaminase, glutamate dehydrogenase, glutamate synthase, asparagine synthase and serine acetyltransferase were either down-regulated or remained unchanged under LN condition. A similar notable result was the substantial variation in amino acid gene expression observed between G152 and G3 grown under same N condition (HN or LN). For instance, caffeoyl-CoA O-methyltransferase gene activity did not change in G152HN but it was down-regulated in G3HN. Proline dehydrogenase gene was up-regulated in G152 LN but its activity hardly changed in G3 LN ( Figure 9 ). There was little change in the expression of serine O-acetyltransferase in G152HN but was over-expressed in G3HN, and it was down-regulated in G152LN with little change in G3LN. Collectively, these results show large N-induced and genotype-dependent variation for amino acid metabolism in the S. spontaneum accessions tested.

As with amino acid metabolism, expression of DEGs involved in carbon metabolism changed significantly in G152 and G3 accessions in response to inorganic N fertiliser application ( Figure 9 , Table S6 ). For example, in both G152 and G3, genes encoding fructose-bisphosphate aldolase, phosphoglycerate kinase and pyruvate dehydrogenase were down-regulated under HN condition whereas, under LN condition, they were over-expressed in G152 but remained unchanged in G3 ( Figure 9 , Table S6 ). In contrast, phosphenolpyruvate carboxylase, 6-phosphofructokinase 1 and ribulose phosphate 3-epimerase were over-expressed in G152HN but its expression did not change under LN condition. However, all these three enzymes were down-regulated in G3 irrespective of its soil N condition. Similar to amino acid metabolism, remarkably opposite expression patterns of DEGs were observed between G152 and G3 grown under HN or LN condition. This further attests the fact that genotypic variation is remarkably greater than treatment effect for carbon metabolism in the test accessions.

Consistent with the activity of amino acid and carbohydrate metabolism genes, expression of DEGs involved in N metabolism also showed remarkable variation under contrasting N supply in G152 and G3 accessions ( Figure 9 , Table S7 ). As an example, expression of nitrate and nitrite transporters and nitrate reductase genes was mostly inhibited in G152 and G3 grown under high N condition whereas their activity under low N was strongly up-regulated in G152 but remained mostly unaffected in G3. Contrarily, glutamate synthase expression was not affected in G152HN but was down-regulated in G3HN.The gene activity was either down-regulated or unaffected under LN condition in both accessions. In general, N metabolism gene expression showed remarkable variation in different genetic background and contrasting soil N conditions.

Genes involved in plant hormones auxin, gibberellin, cytokinin, abscisic acid and brassinosteroid metabolism were substantially altered by externally applied inorganic N in high-BNF S. spontaneum accessions G152 and G3 ( Figure 9 , Table S8 ). Under high N supply condition, (+)-abscisic acid 8’-hydroxylase gene expression was up-regulated conspicuously in G152 but it remained unchanged or down-regulated under low N supply in G152, and in G3 in both soil N conditions. A similar observation was also evident for GA2-oxidase except for G152LN plants where the gene activity was down-regulated. With regard to auxin metabolism, auxin-responsive GH3 genes involved in auxin homeostasis were slightly down-regulated in G152HN, G152 LN and G3HN plants but were up-regulated in G3LN ( Figure 9 , Table S8 ). A somewhat similar trend was also observed for transcripts of cytokinin dehydrogenase, a gene regulating cytokinin homeostasis. In contrast, a cytochrome P450 gene involved in brassinosteroid biosynthesis was strongly up-regulated in G152HN plants but not in plants from the other three treatments. In brief, as with external N treatment, genotype also had a pronounced effect on hormone metabolism in the test plants.

The reliability of RNA-Seq data was confirmed by qRT-PCR analysis of 8 randomly selected unigenes involved in amino acid (c43527_g3_i1; c45156_g2_i0), carbon (c47126_g2_i0; c59722_g4_i0; c66987_g0_i0) and nitrogen metabolism (c51974_g0_i0; c55068_g0_i0; c56112_g2_i0) ( Figure 10 ). The expression data from RNA-Seq was consistent with qRT-PCR except for a few minor quantitative variations.