Multi-omics analyses revealed key factors involved in fluorescent carbon-dots-regulated secondary metabolism in Tetrastigma hemsleyanum

Background Luminescent nanomaterials (LNMs), especially newly-exploited fluorescent carbon-dots (CDs), have demonstrated promising candidates for sunlight harvesting and enhanced photosynthesis efficiency of crops. However, most of the studies focus on the design and synthesis of LNMs and primary metabolism in biomass acceleration, secondary metabolism that closely associated with the quality ingredients of plants is rarely mentioned. Results UV-absorptive and water-soluble NIR-CDs were harvested via a facile microwave-assisted carbonization method. The effect and regulatory mechanism of NIR-CDs on the secondary metabolism and bioactive ingredients accumulation in Tetrastigma hemsleyanum were explored. A total of 191 differential secondary metabolites and 6874 differentially expressed genes were identified when the NIR-CDs were adopted for enhancing growth of T. hemsleyanum. The phenolic acids were generally improved, but the flavonoids were more likely to decrease. The pivotal differentially expressed genes were involved in biosynthesis of secondary metabolites, flavonoid biosynthesis, porphyrin and chlorophyll metabolism, etc. The gene-metabolite association was constructed and 44 hub genes highly related to quality ingredients accumulation and growth were identified, among which and the top 5 genes of the PPI network might be the key regulators. Conclusion This research provided key regulatory genes and differentially accumulating quality ingredients under NIR-CDs-treatment, which could provide a theoretical basis for expanding the applications of nanomaterial in industrial crop agriculture. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s12951-022-01271-6.


Background
At present, the whole world has reached a consensus that an effective and sustainable agricultural system is essential to the existence and development of humanity. Aiming to sustainably meet future demands for food, new agri-tech revolutions including information technology, new materials, and methodologies have largely emerged [1][2][3][4]. Among them, engineered nanomaterials (ENMs)-based nanotechnology positively leads this variation, which enhances the e ciency of cropping systems associated with all inputs via more e cient delivery mechanisms, and thereby improves the e ciency of nutrient utilization, and better management of disease and crop loss [5].
Photosynthesis is the most important energy conversion process on the earth due to its massive and green utilization of sunlight. ENMs reveals great potential to elevate plant photosynthesis e ciency through the following strategies i.e. avoiding abiotic stress [6], increasing chlorophyll content [7], enhancing sunlight harvesting and utilization [8], scavenging reactive oxygen species [9] and promoting electron transfer [10]. In virtue of photoluminescence character, luminescent nanomaterials (LNMs) are ideal candidates for the sunlight absorption and photon use [11]. Because of their small size and environmentally responsive release property, these LNMs possess the characteristics of good penetrability and diffusion in the plant vasculature after root or foliar applications [12]. Many types of LNMs, such as representative semiconductor-type quantum dots (QDs), uorescent carbon dots (CDs), metal nanoclusters, lanthanide-doped up-conversion phosphors (UCPs), have been demonstrated great potentials in enhancing photon absorption in ultraviolet, red or near-infrared (NIR) region of the sunlight and elevated photosynthesis e ciency and crop yield [11]. Positive progresses have been achieved especially on the design or hybridization of these LNMs. For example, toxicity-reduced MPA-capped CdSe QDs, Si-QDs, PEI-coated UCPs, UCPs@CDs nanocomposites and diverse CDs (blue CDs-chloroplasts hybrid, bule/red dual-emissive CDs and NIR or far-red CDs) have been gradually developed and constructed for the enhanced sunlight absorption and photosynthesis e ciency [13][14][15][16][17][18][19]. Among the forementioned LNMs, as a newly-emerged zero-dimension nanophosphors, uorescent CDs have attracted more attentions due to their fascinating merits e.g. facile and low-cost preparation, low toxicity, good physicochemical stability, excellent optical properties [20]. In our recently-reported work [21], we have produced the rst comprehensive analysis of the light-harvesting effects and related mechanisms of NIR-CDs (as a preferential-selected LNMs) on primary metabolism of plant at gene level, which provides a method to research the mechanism of LNMs-caused increased sunlight absorption and photosynthesis e ciency. However, previous researches mainly focused on the design and fabrication of light-absorption nanomaterials, and the effect of these LNMs on growth of plant. There often exists an antagonistic effect between primary and secondary metabolism in plants, so more rapid photosynthesis may lead to the lower accumulation of nutritional or bioactive ingredients in plants. ENMs may be an effective approach to ensure vigorous primary metabolism that could remain a high level of biomass growth rate, but it is also essential and urgent to monitor the responses of secondary metabolism that would signi cantly in uence the quality ingredients in plants [5,11].
In recent years, technical advancements in metabolomics integrated with transcriptomics analyses have provided effective ways to explore the complex regulatory network in plants [22]. Weighted gene co-expression network analysis (WGCNA), a recently appeared approach in systems biology has been gradually applied to describe gene-related patterns under different treatment conditions, and to explore the key candidate genes involved in multiple metabolic pathways. Particularly, it has become an important tool in the identi cation of hub genes and the gene-metabolite association, then rebuild the gene regulatory sub-networks.
Herein, as a continued research work, the effect of NIR-CDs on the secondary metabolism and bioactive ingredients of Tetrastigma hemsleyanum (T. hemsleyanum, a class of medicinal plant) was determined, and the possible mechanism was also further explored by an integrated transcriptome and metabolome research. The gene-metabolite association were constructed by WGCNA, then the candidate hub genes highly related to quality ingredients accumulation and growth were identi ed. The results would lay the foundation for the application of nanotechnology in industrial crop agriculture.

Results
Characterizations of the NIR-CDs.
The NIR-CDs were prepared with a facile microwave-assisted carbonization method using GSH and formamide as the raw materials and solvent, respectively. To characterize the morphology, size, chemical composition and surface state of the harvested NIR-CDs, a series of measurements e.g. TEM, XRD, FT-IR, XPS, Raman and Zeta potential were severally performed. As seen from the TEM image (Fig. 1A), the NIR-CDs present very good mono-dispersity, uniform and spherical morphologies, and a narrow particle distribution with a mean size of 3.9 nm (Fig. 1B). Nonetheless, there are none lattice fringes are observed in the HR-TEM image (Fig. 1C), indicating that the NIR-CDs are mostly noncrystalline. In the XRD pattern (Additional les Fig. S1a), a typical peak at 26° is assigned to the (002) plane of graphite, which further veri es the noncrystalline graphite structure of the NIR-CDs [23]. Furthermore, the graphite structure of NIR-CDs also con rmed by the Raman spectrum (Additional les Fig. S1b). As shown therein, two distinct peaks centered at 1550 and 1342 cm −1 represent the typical G-band and D-band, respectively. Simultaneously, a low ratio of D to G clearly authenticates the dominate pristine carbon in the NIR-CDs [24]. In the FT-IR spectrum of NIR-CDs (1000-1100 cm −1 ) is assigned to C=S and oxidized S bonds [25]. These FT-IR assignments are further clearly con rmed by XPS analysis (Fig. 1E) Fig. S2a) reveals four typical peaks at 284.8, 286.3, 288.0 and 290.1, which is attributed to C=C/C−C, C−N/C−O, C=N/C=O and N−C=O, respectively [27]. Three obvious peaks of pyridine-like N, graphitic N, and pyrrole-like N at 398.5, 400.0, and 402.6 eV are severally observed in high resolution N 1s spectrum (Additional les Fig. S2b) [28]. O 1s XPS spectrum shows two unique peaks of C−OH and C=O at 531.3 and 533.6 eV [29], respectively (Additional les Fig. S2c). Besides, in the high solution S 2p spectrum, four binding energies at 162.2, 163.4, 164.8 and 169.7 were discriminated, corresponding to thiolate, 2p 3/2 and 2p 1/2 of thiophene S, and oxidized S, respectively [30]. Finally, zeta potential measurement reveals that the NIR-CDs are negatively charged (ζ = − 19.2 mV, Additional les Fig. S3), which brings strong electrostatic exclusion and good colloid stability.
Thereafter, the photophysical properties including Uv-vis absorption, uorescence spectrum and uorescence lifetime of the NIR-CDs were inspected, respectively. As shown in Fig. 2A, three distinct absorption bands i.e., 240-300 nm, 350-450 nm, and 550-750 nm are observed, which are generally ascribed to the typical π → π* transition of the aromatic C=C bond, π → π* and n → π* transitions of the aromatic π system containing C=O, C=N, and C=S bonds [25,31,32], respectively. In Fig. 2b, the NIR-CDs reveal strongly deep-red emission from 625 to 725 nm with a sharp peak centered at 680 nm under different excitation wavelengths. Such an excitation-independent uorescence emission property is usually attributed to the surface states/defects induced emission [33]. In addition, the average uorescence lifetime of the NIR-CDs is measured and calculated to be 3.2 ns with bi-exponential decays (Additional les Fig. S4), and the absolute uorescence quantum yield is measured to be 18.4% under an optimal excitation of 420 nm. In our previous reports [21,23], the NIR-CDs have been demonstrated excellent tolerance to photobleaching under the excitation of ultraviolet light, good storage stability at ambient environment and competitive agent in light-harvesting and electron transfer from photosystem II (PS II) to photosystem I (PS I) in chloroplasts [18,34].
Confocal images were captured using Laser-scanning confocal uorescence microscope after incubation with 0.05 mg/mL NIR-CDs. As shown in Fig. 3A, the red uorescence signals from NIR-CDs observed under 514 nm excitation were widely distributed in root, stem, and leaf of T. hemsleyanum, which con rmed that NIR-CDs could penetrate cell wall into vascular bundle system, and were transported to the aerial parts.
Metabolome pro ling and KEGG enrichment analysis of differential metabolites We pro led the widely-targeted LC-MS/MS based metabolome of the samples from the control (denoted as CK) and the NIR-CDs-treatment groups (denoted as TH), which represented the impact of the NIR-CDs on metabolite accumulation of T. hemsleyanum. We detected 493 secondary metabolites grouped into 8 classes (Additional les  Table S1), including 182 avonoids, 167 phenolic acids, 20 lignans and coumarins, 64 alkaloids, 15 terpenoids, 26 tannins and others. The differentially accumulated metabolites (DAM) between the control and the experimental groups (CK_vs_TH) were screened using the variable importance in projection (VIP)≥ 1 from the OPLS-DA model and fold change ≥ 1.5 (upregulated) or ≤ 0.667 (downregulated). A total of 191 DAM were identi ed in CK_vs_TH (Additional les Table S2). These secondary metabolites can be mainly categorized into the classes of avonoids and phenolic acids, including 106 avonoids, 43 phenolic acids, 12 lignans and coumarins, 9 alkaloids, and others. Overall, avonoids were more inclined to accumulate in CK than in TH. The content of phenolic acids, lignans and coumarins were signi cantly improved by CDs-treatment (Fig. 4A).
We focused on the two classes of secondary metabolites ( avonoids and phenolic acids) likely to be major contributors to biological activity (Fig. 4B). Flavonols and avonoid carbonoside were identi ed with a series of glycoside derivatives of kaempferol, quercetin, and apigenin, which made up the majority of the DAMs detected for CK _vs_TH compared samples. Based on fold changes and VIP values, 77 out of 106 avonoids were identi ed as downaccumulated signi cantly by CDs-treatment, the concentrations of Kaempferol- , and Epigallocatechin were signi cantly greater in CK than in TH (Student's t test, P <0.05), and the concentrations of the remaining 29 avonoids were signi cantly greater in TH than in CK. 43 out of 167 phenolic acids were identi ed as differentially accumulated by CDs-treatment, of these 13 phenolic acids were downregulated and 30 were upregulated in TH compared with CK, among which Quillaic acid, Furanofructosyl-α-D-(6-mustard acyl)glucoside, Feruloylcaffeoyltartaric acid were found only in TH, CDs-treatment also led to a 10-fold enhancement in Syringin, Neochlorogenic acid, Homogentisic acid, and 5-fold increase in 1-Caffeoylquinic acid and 2-Caffeoylquinic acid.

Transcriptome analysis and DEGs Identi cation
Differentially expressed genes (DEGs) in CK vs TH comparative group were identi ed by a transcriptomic comparison.
A total of 40.08-52.20 million clean reads were obtained, and the Q30 of the raw reads ranged from 91.08-92.21%, indicating the high quality of the transcriptome data. As shown in Figure 3A, the red dots indicate the upregulated expressed genes (Log2FC ≥ 2), the black dots indicate the non-signi cantly differentially expressed genes (0.5 < Log2FC < 2), and the blue dots indicate the downregulated expressed genes (fold change, Log2FC≤0.5). The horizontal axis indicates the fold change of differential expression, and the vertical axis indicates the signi cance level of the gene expression differences. According to the results of transcriptome data, 43340 unigenes were identi ed. A total of 6,874 DEGs were identi ed between CK vs TH (Additional les Table S3), including 2962 upregulated genes and 3912 downregulated genes (Fig. 5A).
The DEGs between CK and TH were subjected to KEGG, KOG, and GO functional pathway analyses. The top enriched KEGG terms contributed by these DEGs were ko01110 (Biosynthesis of secondary metabolites), ko04016 (MAPK signaling pathway -plant), ko00941 (Flavonoid biosynthesis), ko00860(Porphyrin and chlorophyll metabolism), and ko00909 (Sesquiterpenoid and triterpenoid biosynthesis) (Fig. 5B). The top enriched KOG terms included Posttranslational modi cation, protein turnover, chaperones(O), Carbohydrate transport and metabolism(G), Energy production and conversion(C), and Secondary metabolites biosynthesis, transport and catabolism(Q) (Fig. 5C). DEGs annotated in GO were classi ed into 52 functional groups, including 25 groups in biological process, 16 in cellular components, and 11 in molecular function. 'Cell', 'Cell part', and 'organelle' were the terms that dominated in the cellular component category. In the 'molecular function' category, the GO terms 'catalytic activity' and 'binding' predominated. 'Cellular process', 'metabolic process', 'biological regulation', and 'regulation of biological process' were the most represented GO terms in the biological process category (Fig. 5D). The above functional pathway analyses indicated that most of the identi ed DEGs acted on metabolic processes related to secondary metabolite metabolism and carbohydrate metabolism.

Co-expression network analysis of DEGs
To identify the candidate genes regulating secondary metabolite metabolism and carbohydrate metabolism, an effective system biology method called weighted gene co-expression network analysis (WGCNA), was performed to nd the modules of highly correlated genes and relate these modules to traits. Modules are clusters of genes with high correlation, and genes of a same module are co-expressed. A total of 8 gene modules were established on the clustering and signature analysis of the genes with similar expression patterns in DEGs (Additional les: Fig. S5). They were then used to correlate with the traits of avonoid metabolites content (Additional les: Fig. S6), phenolic acids metabolites content (Additional les: Fig. S7), and photosynthetic e ciency (Additional les: Fig. S8), respectively.
The module-trait relationship analysis also identi ed module "darkturquoise" as most highly related to photosynthetic e ciency. Photosynthetic rate (Pn) (r = 0.97, P < 0.01), stomatal conductance (Cond) (r = 0.96, P < 0.01), transpiration rate (Tr) (r = 0.99, P < 0.01), chlorophyll uorescence parameters (Fv/Fm) (r = 0.98, P < 0.01). Therefore, the "darkturquoise" and "green" modules were selected as the key gene modules for subsequent analysis. Key gene module expression and functional analysis Gene annotation and expression of all the gene members in the two key gene modules were performed.
"darkturquoise" and "green"gene modules were composed of 6596 and 940 unigenes, respectively. When the differential genes of CK vs TH were separately intersected with the gene members of the two key gene modules, we found that the intersection was observed between the two key gene modules and the differentially expressed genes (Fig. 6A). Downregulated expressed genes in "darkturquoise" and "green" gene modules were 985 and 83, while upregulated expressed genes in the two modules were 1621 and 44, respectively. Further expression analysis of gene members of the two key gene modules revealed distinct expression patterns in CK and TH ( Fig. 6B and 6C).
The KEGG gene-set enrichment analysis of these DEGs in "green" and "darkturquoise" gene modules were severally shown as Fig. 7 and Fig. 8, respectively. For each screened gene module, the correlation scatter plot between some phenotypes and modules were drawn ( Fig. 7A and Fig. 8A). The enriched KEGG terms of upregulated expressed genes in "green" gene module contained MAPK signaling pathway, porphyrin and chlorophyll metabolism, and galactose metabolism (Fig. 7B). The enriched KEGG terms of downregulated expressed genes in "green" gene module contained avonoid biosynthesis, other glycan degradation, glycosaminoglycan degradation (Fig. 7C). The top enriched KEGG terms of upregulated expressed genes in "darkturquoise" gene module were benzoxazinoid biosynthesis, MAPK signaling pathway, porphyrin and chlorophyll metabolism, and anthocyanin biosynthesis (Fig. 8B). The top enriched KEGG terms of downregulated expressed genes in "darkturquoise" gene module were avonoid biosynthesis, biosynthesis of secondary metabolites (Fig. 8C).
Identi cation of hub genes within network modules Two modules, "darkturquoise" and "green", were considered to take a decisive factor in the regulation of CDs on plant growth and secondary metabolites accumulation. Hub genes were de ned as these genes which were highly associated with other genes in each module network, and played a central role within the network clusters. Hub genes within the two modules were discovered in Fig. 9. Protein-protein interaction (PPI) network, which took the gene members in the two key modules as its object, were obtained using string database, and a total of 44 nodes, 44 edges were identi ed (Fig. 9A). The MCC algorithm of CytoHubba was used to score and sort the hub nodes in the PPI network, and select the top 5 genes (triosephosphate isomerase, mitochondrial carrier protein, thymidylate kinase, dehydrogenase E1 component and lyase) as the hub genes of the PPI network (Fig. 9B). Gene expression of all PPI network gene members in the two key gene modules were performed and shown in Fig. 9C. Among them, 17 upregulated genes and 27 down-regulated genes were identi ed. These interactions among key genes, pathways, metabolic type and gene regulation are shown in Fig. 9D. As for secondary metabolism, transferase family genes in anthocyanin, phenylpropanoid, and avonoid biosynthesis pathways tended to be down-regulated after CDs-treatment, and the expression of NAD dependent epimerase/dehydratase family genes in phenylpropanoid, avonoid, and iso avonoid biosynthesis pathways were restrained as well. As for primary metabolism, some gene families of carbohydrate metabolism, such as pfkB family carbohydrate kinase, glycosyl hydrolase family, triosephosphate isomerase, and carbohydrate phosphorylase, tended to be down-regulated after CDs-treatment. However, some gene families involved in carbohydrate synthesis, such as sucrose synthase and aldehyde dehydrogenase families, were upregulated. Ferrochelatase and chlorophyll A-B binding protein, which were considered to be vital components of photosynthesis and chloroplast synthesis, were both up-regulated in response to CDs-treatment.

Validation of hub genes expression by qRT-PCR
We selected the top 5 hub genes (triosephosphate isomerase, mitochondrial carrier protein, thymidylate kinase, dehydrogenase E1 component and lyase) of the PPI network to conduct a validation experiment by qRT-PCR analysis.
The patterns of RNA-Seq expressions on all the 5 hub genes were highly consistent with the qRT-PCR data, and a correlation coe cient (R2) of 95.33% was obtained (Fig. 10). Both qRT-PCR and RNA-seq analyses showed that three hub genes (triosephosphate isomerase, mitochondrial carrier protein, and dehydrogenase E1 component) showed lower level of expression after CDs-treatment. However, other two genes, thymidylate kinase and lyase showed a higher level of expression in the CDs-treatment group.

Discussion
As an emerging research hotspot in nanotechnology, carbon dots (CDs) have attracted much attention. Due to the uorescence property of CDs, it is available to trace the uptake and accumulation of CDs in plants. When the mung bean was treated with nitrogen-doped CDs by solid-state method, the blue uorescence was observed to migrate from the beans to the root ends by ultraviolet light during the sprouting process [35]. When T. hemsleyanum was treated with red CDs, the uorescence signal of CDs was mainly detected in the vascular system of the root, stem, and leaf by confocal laser scanning microscope in our study, which con rmed that CDs could easily cross the biological barrier and be widely distributed in the plants. A number of studies have demonstrated that CDs can promote the growth of various plants by enhancing the light absorption [11], stimulating biosynthase activity [36], converting ultraviolet light into blue and red light [17,18], and we rstly demonstrated that potential molecular mechanisms behind the growthstimulating effect might be related to up-regulated expression of the primary metabolism related genes, among which PsbP and PsiK genes were the hub genes [21].
Plants synthesize a myriad of secondary metabolites that are derived from central or primary metabolism. Flavonoids, alkaloids, terpenoids and phenolic acids are the common and important kinds of secondary metabolites, and they are closely related to multiple bioactive functions of medicinal plant. It is generally assumed that the level of primary metabolism would have a major positive impact on the growth rate of plant and the level of secondary metabolism would be stimulated under environmental stress [37]. Hence, it is of great signi cance to investigate the potential impact of CDs on the metabolic transition between primary and secondary metabolism.
T.hemsleyanum is a herbal plant distributed in tropical to subtropical regions, mainly in mainland China. Flavonoids and phenolic acids were the main active secondary metabolic ingredients of T. hemsleyanum [38]. Total favonoids from T. hemsleyanumon were reported to exert anti-infammatory effects on Con A-induced hepatitis in mice [39]. T. hemsleyanum avonoids could inhibit the migration and promote the apoptosis of A549 cells both in vitro and in vivo [40]. Phenolics in T. hemsleyanum, including 1-Caffeoylquinic acid, 2-Caffeoylquinic acid, 5-p-coumaroylquinic acid, isoorientin etc., were related to the antioxidant and antiproliferative activities of T. hemsleyanum [41]. A total of 13 bioactive compounds, including kaempferol, caffeic acid, apigenin, quercetin etc., were considered to be related to alleviating lung infection in Pseudomonas aeruginosa-induced mice [42]. Apigenin and luteolin glycosides of T. hemsleyanum were reported to induce apoptosis in HepG2 cells as well as inhibit the tumor growth in H22 tumorbearing mice [43].
In this study, CDs-treatment had the opposite effect on the content of phenolic acids and avonoids. The contents of avonoids decreased and phenolic acids substance increased. The ingredients associated with anti-in ammatory and antioxidant activities of T. hemsleyanum, such as syringin, neochlorogenic acid, homogentisic acid, 1-Caffeoylquinic acid and 2-Caffeoylquinic acid, were generally improved in response to CDs-treatment. But the ingredients associated with anti-tumor activity, such as glycoside derivatives of kaempferol, quercetin, and apigenin, were more likely to decrease in general terms. Thereafter, we carried out an integrated transcriptome and metabolome research on the regulatory networks of primary and secondary metabolites biosynthesis, including starch and sucrose metabolism, photosynthesis, avonoids biosynthesis etc. of T. hemsleyanum under the treatment of NIR-CDs. The pattern of association between differentially expressed genes (DEGs) and metabolite components were explored by WGCNA, then 44 candidate hub genes highly related to quality ingredients accumulation and growth were identi ed.
Maintenance of deoxyribonucleotide levels is crucial for ensuring e ciency and genome stability of DNA replication and DNA synthesis. The increase in the deoxyribonucleotide pools would lead to subtle changes in the function of DNA Pol III leading to more elongation and less proof-reading mode, thus resulting in the activation of translesion synthesis. Thymidylate kinase, referred to as TMK or TMPK in different organisms, is an important enzyme in DNA biosynthesis and catalyses the conversion of dTMP to dTDP. In the present study, the activity of thymidylate kinase was signi cantly stimulated by NIR-CDs-treatment, serving as one of the top 5 hub genes. It suggested that thymidylate kinase gene was highly associated with other genes in DNA synthesis and cell division gene regulatory network, and played a central role in promoting primary metabolism of T. hemsleyanum. Thymidylate kinase has also been studied in the photosynthetic nitrogen-xing cyanobacterium Nostoc. Recombinant Nostoc strains overexpressing AnTMK exhibited higher growth rate measured in terms of chlorophyll a content under normal growth conditions, which indicated that the TMK is likely to have a signi cant role in photosynthetic organisms [44].
Mitochondrial carrier family (MCF) proteins of plant play a major role in transporting TCA-cycle intermediates across the inner mitochondrial membrane [45]. The present study also suggested MCF served as a protagonist in the regulation of mitochondrial metabolism in response to CDs-treatment. Chlorophyll and chlorophyll A-B binding protein are indispensable for the assembly of a functional photosystem II, which is responsible for light absorption, excitation energy transfer, and charge separation. Ferrochelatase in plants possesses a conserved transmembrane chlorophyll A-B binding domain that is somewhat analogous to the rst and the third helix of light-harvesting complexes, including a chlorophyll-binding motif [46]. In the present study, CDs-treatment signi cantly stimulated photosynthesis of T. hemsleyanum by simultaneously upregulating the expression of chlorophyll A-B binding protein and ferrochelatase.
Glycosyl hydrolase family genes participated in many aspects of plant physiological processes, in particular biotic and abiotic stresses through the regulation of phytohormones and defensive components. Four genes of the glycoside hydrolase family 1 in the model legume plant Medicago truncatula, were dramatically activated by NaCl, PEG, IAA, ABA, SA and GA3 treatments [47]. Both triosephosphate isomerase and pfkB family carbohydrate kinase are key enzymes in glucolysis [48]. Sucrose synthase plays an important role in tricarboxylic acid cycle and polysaccharides synthesis. In the present study, CDs-treatment signi cantly affected glucose metabolism by simultaneously inhibiting triosephosphate isomerase and pfkB family carbohydrate kinase, while activating sucrose synthase, which suggested that CDs could switch the glycometabolism ow from the glucolysis pathway to the pyruvate carboxylation, and hence the generation of polysaccharide would be improved.
Transferases are a superfamily of abundant enzymes that play vital roles in plant growth and development and avonoid metabolism. For example, UDP-glycosyltransferases were essential for avonoid biosynthesis in Glycyrrhiza uralensis [49]. A glutathione S-transferase gene was demonstrated to be a candidate gene for manipulating anthocyanin accumulation and pigmentation in cotton tissues [50]. Transferases are likely to be responsive to exogenous chemical ingredients or environmental stress, and thus the accumulation of plant-derived avonoids was affected. The glutathione transferases had members that are selectively induced by chemical stress treatments, which were con rmed to have roles in redox homeostasis and maintaining the avonoid pool under stress conditions [51].

Conclusion
In summary, UV-absorptive and water-soluble NIR-CDs were harvested via a facile microwave-assisted carbonization method. The NIR-CDs were successfully adopted for enhancing photosynthesis and growth of T. hemsleyanum, and the in uences on the accumulation of bioactive ingredients was illustrated as well. A total of 191 differential secondary metabolites and 6874 differentially expressed genes were identi ed. The phenolic acids were generally improved, but the avonoids were more likely to decrease in general terms. The pivotal differentially expressed genes were involved in multiple metabolic processes related to biosynthesis of secondary metabolites, avonoid biosynthesis, porphyrin and chlorophyll metabolism, etc. Subsequently, DEGs were divided into 9 modules by WGCNA.
Two modules positively correlated with the avonoids content, phenolic acids content and photosynthetic e ciency were identi ed, in which 44 hub genes and the top 5 genes of the PPI network were identi ed. This research provided key regulatory genes and differentially accumulating quality ingredients under CDs-treatment, which could provide a theoretical basis for expanding the applications of nanomaterial in industrial crop agriculture.

Materials and apparatus
Reduced glutathione (GSH), Na 2 HPO 4 , KH 2 PO 4 and KCl were purchased from Aladdin Chemistry Co., Ltd (Shanghai, China). Formamide was obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). All of the chemicals were used as received without further puri cation. Aqueous solutions were prepared using deionized (DI) water.
A microwave oven (Galanz, P70F20CL) was adopted for the fabrication of NIR-CDs. The sizes and morphologies of the obtained NIR-CDs were characterized by high resolution transmission electron microscopy (HR-TEM, Tecnai F20) using an acceleration voltage of 200 kV. Fourier transform infrared (FT-IR) spectrum was carried out on a Nicolet 6700 FT-IR spectrometer through the KBr pellet technology. Raman spectrum was recorded on a Renishaw inVia Raman spectrophotometer via a 532 nm laser as the excitation resource. The crystal phase of NIR-CDs was measured on a Bruker D8 Discover X-ray diffractometer (XRD) with 2θ range from 10° to 50° at a scanning rate of 4°/min, with Cu Ka irradiation (k = 1.5406 Å). Fluorescence spectra were measured on a Perkin Elmer spectrophotometer (LS-55). UV-vis absorption spectra were recorded on an Agilent Cary 300 spectrophotometer. Fluorescence delay curve was monitored by Fluorolog 3-11 (HORIBA Jobin Yvon). Absolute uorescence quantum yield (QY) was directly measured by a Fluoromax-4 measurement system (HORIBA, JobinYvon. Inc).

Synthesis of NIR-CDs
NIR-CDs were synthesized in accordance with our previously-reported one-pot microwave-assisted carbonization manner [21,52]. Brie y, reduced GSH (2.0 g) was dissolved in 20 mL formamide under a 10 mins treatment of ultrasonic. Thereafter, the mixture was transferred into a domestic microwave oven for further carbonization reaction (700 W, 3 mins). Then, the beaker was cooled down to room temperature naturally, and a dark green mixture was obtained. The mixture was centrifuged at 10000 rpm for 5 min to remove large-sized nanoparticles, and puri ed through dialysis (cut-off molecular weight, 3500) for 5 days. Then, the NIR-CDs solution was concentrated and dried using a rotary evaporation. Finally, the dark green NIR-CDs powder ca. 350 mg was harvested.

Plant cultivation and NIRCDs treatment
Healthy cutting seedlings of T. hemsleyanum were cultured in a greenhouse under a 16/8 h photoperiod (day/night), 23 ± 2°C temperature, and 50% relative humidity. Then, seedlings showing consistent growth state were selected and uniformly divided into two groups, CK_and_TH. NIR-CDs solutions at 0.05 mg/mL were sprayed on seedlings with a dosage of 10 mL/pot every day (TH), and the same volume of ultrapure water was sprayed as the control (CK).
Phenotypic changes were observed and recorded every 15 days, respectively. The photosynthetic parameters were determined according to our previous report [53]. Each treatment was repeated three times with ten plants. The young leaves were collected on 30 nd day, frozen in liquid nitrogen immediately, and stored at -80℃ for RNA-seq and metabolomics analysis. Three biological replicates were performed, and each biological replicate consisted of a pool of samples from 10 seedlings.
RNA-seq data analysis RNA extraction, quanti cation and transcriptome sequencing were performed according to previous studies [54,55].
The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumia). The library preparations were sequenced on an Illumina Hiseq platform and 125 bp/150 bp paired-end reads were generated. Transcriptome De novo assembly was performed using Trinity software. DIAMOND BLASTX software was used to compare the assembled transcript sequence with KEGG, NR, Swiss-Prot, GO, KOG, Trembl databases [56]. After predicting the amino acid sequence of the transcript, use HMMER software to compare with Pfam database to obtain the annotation information of the transcript. The data from RNA-seq were sequentially processed, after which gene expression quantity calculations were performed by bowtie2 [57], resulting in an expression matrix. Fragments per kilobase of transcript per million mapped reads (FPKM) of each gene based on the gene length were used to estimate the transcription or gene expression level.
The differential expression between the two groups was performed using edgeR. The corrected P value and |log2foldchange| are used as the threshold for signi cant difference expression (log2|FoldChange| > 2 and P value < 0.05 were adopted as the screening threshold). The enrichment analysis is performed based on the hypergeometric test. Kyoto Encyclopedia of Genes and Genomes (KEGG) were used as pathway enrichment. At least three genes were enriched in pathways, and a p value less than 0.05 was considered to be signi cantly enriched.
Linear ion trap (LIT) and triple quadrupole (QQQ) scans were acquired on an AB4500 Q TRAP UPLC/MS/MS System, equipped with an ESI Turbo Ion-Spray interface, and controlled by Analyst 1.6.3 software (AB Sciex). The ESI source operation parameters were as follows: ion source, turbo spray; ion spray voltage (IS) 5500 V (positive ion mode)/-4500 V (negative ion mode); source temperature 550°C; ion source gas I (GSI), gas II(GSII) and curtain gas (CUR) were xed at 50, 60, and 25.0 psi, respectively. Instrument tuning and mass calibration were performed with 10 and 100 µmol/L polypropylene glycol solutions in QQQ and LIT modes, respectively. Weighted gene co-expression network analysis (WGCNA) Weighted gene co-expression network analysis (WGCNA) is a systems biology method used to describe the gene association patterns between different samples. It can be used to identify highly synergistic gene sets, and identify functional gene sets that affect or participate in phenotypes according to the interconnection of gene sets and the association between gene sets and phenotypes [58]. R software WGCNA was used to construct an undirected cocorrelation network, de ning a soft threshold β = 1 for the screening condition. The minimum number of module genes was 30 and the module merging threshold was 0.25, the cluster tree was constructed according to the correlation of gene expression and divided into modules. In order to identify gene modules and gene module members related to avonoids and phenolic acid metabolism, the contents of net photosynthetic rate (Pn), stomatal conductance (Cond), transpiration rate (TR), intercellular carbon dioxide concentration (CI), chlorophyll uorescence parameters (Fv/FM), the trait matrix of related phenotypes were used to calculate the correlation between gene module and phenotype. The expression patterns of hub module gene members were quantitatively described by intramodular gene expression heatmap with gene signi cance analysis. The phenotype and gene signi cance of the screened gene modules were analyzed, and the Venn plot was used to describe the up-down relationship of differential genes and the correlation between phenotypic related module gene members, so as to determine the signal pathways and metabolic pathways that are inhibited or stimulated in KEGG enrichment analysis and identi cation of gene modules.

Identi cation of hub genes associated with secondary metabolite synthesis
Using the functionally annotated differential transcript data, the screened modules related to phenotype are combined to generate node information for constructing protein-protein network. Then, the near source species of T. hemsleyanum was selected for protein-protein interaction (PPI) analysis through STRING database (https://www.string-db.org/). Use Cytoscape software to draw the PPI network, then use the MCC algorithm of CytoHubba plug-in to score and sort the key nodes in the PPI network, and select the top 5 genes as the hub genes of the PPI network. Then, through the cluster analysis of the expression pattern of node genes in PPI network, the expression heatmap of these network member genes was drawn. Finally, the gene information in this network was described by sanky plot.
Validation of the DEGs data using qRT-PCR

Competing interests
There are no con icts to declare.

Authors' contributions
Experiments were designed by MJL, JPZ, and YHW, and conducted by XP, ZMX, XHW, and ZYZ. Data was analysed by YXZ and CYY. Manuscript was prepared by XP and edited by YHW. All authors read and approved the nal manuscript.