We selected 16 Ginkgo biloba trees in the Hupingshan Nature Reserve, which were rarely affected by human activities, as the research object (Table 1). Ginkgo’s leaves were chosen as our object owing to their abundant flavonoids. The average annual temperature in the nature reserve is 9.2°C. The extreme maximum temperature in the sample sites is 38.2°C. The extreme minimum temperature is −15°C. The average annual sunshine in the sample sites is 1,509.9 h, and the annual precipitation is 1,898.5 mm. All trees grew in the same general area and experienced similar growing conditions. The elevation and the diameter at breast height (DBH) were determined by the Global Positioning System (GPS) instrument and a tape measure. All trees were healthy, alone, and had no shade. The leaves of the ginkgo canopy are divided into two parts longitudinally with the trunk as the axis. The sampling height is 5-8 m. The two parts are evenly sampled. Multiple leaves are mixed into one sample. About 200 g of healthy and mature leaves were collected as samples in each part. Leaves in different parts were considered as different samples. All samples were collected aseptically wearing bioclean gloves on a sunny day with the same weather condition. All samples were divided into two parts. One was frozen in liquid nitrogen immediately and then stored at −80°C for metabolome detection and flavonoid extraction. Another was surface disinfected promptly for the microbiological analysis. Leaves were sterilized with 75% ethanol for 5 min, then washed with sterile water, soaked in 8% NaClO for 2 min and washed with sterile water again (Sahu et al., 2022). After surface sterilizing, dried the leaves surface with sterile filter paper and sterile air on a sterile workbench. All leaves were triturated in liquid nitrogen and stored at −80°C until DNA extraction.

Frozen leaves were dried in an oven at 60°C until the weight was constant (Zhou et al., 1999). Afterward, these leaves were ground into powder smaller than 40 mesh and extracted in a Soxhlet extractor based on Chinese pharmacopeia (2015). 1.000 g dried powder was homogenized in 100 mL chloroform followed by two h reflux at 80°C. The residue was resolved in 50 mL methanol followed by two-hour reflux at 90°C. Then, 20 mL hydrochloride was added to the supernatant, refluxing for 30 min at 90°C and evaporating to dryness. The residue was dissolved in methanol and filtrated through a 0.22 μm membrane for HPLC (Daigle and Conkerton, 1983).

Chromatographic separation was implemented on a Shimadzu LC-20AD Series HPLC system (Shimadzu, Duisburg, Germany) equipped with SIL-20A autosampler, SPD-20A UV-VIS detector. ACQUITY UPLC™ BEH C18 column (217 mm × 2.1 mm, 1.7 μm, Waters, Milford, United States) was connected to the whole detection at 25°C. The mobile phase consisted of 10% 0.05 M sodium acetate (A) and 90% acetonitrile (B). A flow rate of 1.0 mL/min was used. Standard curves were established by a series concentration of the corresponding standard (ChemFaces, Wuhan, China) at 360 nm (Mattila et al., 2000). And the content of total flavonoids was calculated by the formula (Hasler et al., 1992):

The data were provided in the (Supplementary Table 1).

The leaf fragments were suspended in TE buffer and homogenized in a sterilized mortar and pestle with liquid nitrogen (Zhang et al., 2016). DNA was extracted from the homogenized leaf material using the DNeasyR Plant Mini Kit (Product of Germany). To get rid of the distractions of DNA of Ginkgo leaves, the DNA was polymerase chain reaction (PCR) amplified using a primer pair 799F (5′-AACMGGATTAGATAC CCKG-3′) and 1115R (5′-AGGGTTGCGCTCGTTG-3′) (Leff et al., 2015). The DNA was cleaned by DNA Clean-Up Kit (OMEGA bio-teak, Norcross, Georgia, United States). High-throughput sequencing of the PCR products was conducted on the Illumina Miseq platform (Miseq PE250).

By comparing the NCBI and Greengene databases, we removed the sequence of the chloroplasts. Raw 16S rRNA gene sequences were processed using the Galaxy | Denglab¹. These steps included: quality filtering, RDP clustering, sequence alignments and community dissimilarities analysis (Wu et al., 2018). Taxonomic assignment of 16S representative sequences was executed with the RDP (Ribosomal Database Project) classifier according to the Greengene database (Desantis et al., 2006). Resampled 16S OTU subsets (1000 sequences per sample) were to compute alpha diversity and beta diversity. The data presented in the study are deposited in the NCBI repository, accession number PRJNA852885.

Alpha-diversity indexes, including Shannon index, Simpson evenness, and Pielou evenness, were calculated at Institute for Environmental Genomics (IEG)². Non-metric multidimensional scaling (NMDS) was calculated to assess the beta diversity. The dissimilarity test identified the difference in bacterial communities. Correlation between bacterial community and flavonoids based on the Mantel test, correlation test, and redundancy analysis (RDA) were completed using the Wekemo Bioincloud³. OTUs with an average abundance of more than 0.05% was chosen to construct correlation networks by calculating Spearman’s rank correlations with Spearman’s correlation coefficient (r > 0.6, P < 0.01) (Steinhauser et al., 2008). We used Gephi to visualize the correlation networks (Bastian et al., 2009).

PICRUSt was used as a bioinformatics tool to predict microflora’s abundance of function genes (Langille et al., 2013). In this study, PICRUSt was employed to predict the functions genes of each sample based on 16S rRNA sequencing data. Comparing the sequence data of the 16S Greengene database (Greengenes 13.5), we predicted the community functional gene relative abundance of endophytic microorganisms with reference to the KEGG database.

HPLC results showed that quercetin, kaempferol and isorhamnetin were the primary flavonoid metabolites of Ginkgo leaves. According to the regression equation of the standard curve (Supplementary Table 2), the content of total flavonoids was calculated (Figure 1). The results showed that the total flavonoids increased by 178.38%, from 1,645.36 mg/kg at S1 to 4580.43 mg/kg at S4 (p < 0.05). The three flavonols quercetin, kaempferol and isorhamnetin increase steadily with elevation. The content of quercetin, kaempferol and isorhamnetin in leaves increased by 2.58, 1.11, and 1.71 times from S1 to S4, respectively. The results suggested that Ginkgo flavonoids’ content was increased along the altitude gradient.

After quality filtering, RDP clustering, sequence alignments and other measures, we obtained 312123 high-quality paired 16S rRNA sequences. At the level of 97% similarity, all sequences were assigned to 3,038 OTUs based on the 16S Greengene database. OTU table was provided in the Supplementary Document. Alpha diversity represented by the Shannon index (Figure 2A) and observed_otus (Figure 2B) revealed that alpha diversity decreased significantly with the increase of flavonoids and the elevation. Samples at lower altitudes had higher species richness and diversity. NMDS and PCA results indicated that the bacterial communities in leaves were dynamic. Samples from the same site huddled together, and samples from different altitudes showed a great distance (Figures 2C,D).

All reads were affiliated with seven bacterial phyla except 5%∼10% sequences unclassified. The dominant phylum was Proteobacteria, accounting for 90%∼95% of the reads (Figure 3A). The samples at lower altitudes had fewer unclassified reads at the phylum level. Bacteria of Actinobacteria and Firmicutes were significantly reduced at higher altitudes. But at the family level, Enterobacteriaceae dominated the bacterial community composition in leaves and remained sequences belonged to 6 families and unclassified family (Figure 3B). It was found that samples at higher altitudes have less species richness at the family level.

The Spearman correlation test and RDA analyzed the relationship between endophytic bacterial populations and factors. RDA results showed that the altitude, quercetin, kaempferol, isorhamnetin, and T_flavonoids significantly affected the endophytic bacterial community in ginkgo leaves, and the endophytic bacterial community was slightly affected by the age of Ginkgo (represented by DBH) (Figure 4A). VPA revealed that these factors constrained 19.59% of the total community distribution. The flavonoids individually constrained 12.83% of the community distribution, while the altitude and DBH constrained 1.50 and 2.28%, respectively (Figure 4B). Besides, flavonoids combined with altitude constrained 5.01% of the community distribution. The heatmap between factors and bacterial abundance was constructed by Spearman correlation values at both the family and genus levels. At the family level, we observed Ruminococcaceae and Porphyromonadaceae were positively correlated with three flavonols and T_flavonoid, while Rhodospirillaceae, Bifidobacteriaceae and Burkholderiaceae were positively correlated with T_flavonoids and one or two flavonols. There was a significant negative correlation between Bacillaceae and flavonoids (Figure 5A). It was found that endophytes of four genera were significantly positively correlated with flavonoids, including Oscillibacter, Moraxella, Clostridium and Parabacteroides. Only Staphylococcus showed a significant negative correlation with T_flavonoid and kaempferol (Figure 5B).

Using PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) as a predictive exploratory tool, we found that seven orthology groups at the level I (KOs in KEGG (Kyoto Encyclopedia of Genes and Genomes)) were observed in the Ginkgo endophyte community. Relative abundance of metabolism, genetics information processing and cellular processes tended to be constant in the four samples. Almost half of the significant function genes were classified into multiple metabolism groups. Environmental information processing accounted for nearly 20% of all genes. The difference was the relative abundance of environmental information processing gradually increased from 16.2% in S1 to 17.5% in S4 (Figure 6A). Besides, there were more pronounced changes in function genes at level III. Photosynthetic pathways (including photosynthesis - antenna proteins, photosynthesis proteins, carbon fixation in photosynthetic organisms and photosynthesis) and flavonoids biosynthesis (consisting of phenylalanine, tyrosine and tryptophan biosynthesis, phenylpropanoid biosynthesis and flavonoid biosynthesis) gradually reduced (p < 0.05). The relative abundance of function genes of photosynthesis in S1 decreased by two-thirds compared with that in S1 (Figure 6B).

We further performed network analysis to assess the impact of altitude on ginkgo leaf endophyte. Our results showed that microbial interkingdom network patterns shifted clearly along with the increased altitude (Figure 7). Specifically, endophytic networks at lower altitudes had higher network connectivity (i.e., network degree) than networks at higher altitudes. Although there was a near number of nodes contributing to the network construction, the edges of networks gradually increased with the altitude increase. Compared with the network of S1, the number of edges increased by two times at S4. The network’s average degree and weighting degree increased from 11.757 and 5.705 to 28.249 and 18.002, respectively (Figure 7E).

In this study, total flavonoids increased significantly with altitude (Figures 1A,B). There was not much difference in temperature among the locations of the samples. Meanwhile, it implied that the excess light might be the main reason for the accumulation of flavonoids. Increasing light intensity within a certain range could enhance photosynthesis (Longstreth and Strain, 1977). Another study showed that the strengthening of light intensity could motivate photosynthesis and enhance the oxidative activity in leaves (Ghasemzadeh et al., 2010). As important antioxidants in plants, flavonoids were produced to resist these oxidative active substances like (O₂^–) (Pietta, 2000; Tattini et al., 2010). Thus, we speculated that the accumulation of flavonoids could be an important strategy for plants to protect against cellular damage. With the rise of elevation, the illumination was enhanced, which promoted photosynthesis and increased oxidation activity. Ginkgo trees produced more flavonoids to weaken oxidative damage. Our recent study observed that flavonoids and related genes increased with the altitude, attributing it to the enhanced illumination (Zou et al., 2019). The fact that the improvement of illumination intensity could promote the production of flavonoids and accelerate the accumulation of oxidative activity had also been reported in the previous study (Karimi et al., 2013). As an antioxidant, flavonoids protect the plant from light damage (Kumar et al., 2018). Besides, the light was proved to induce the oxidation and degradation of lipids (Sun et al., 2006; Wiswedel et al., 2009). All the above evidence showed that Ginkgo biloba increased flavonoid output to resist the accumulation of oxidative activity caused by the elevation and enhanced light.

Many secondary metabolites of plants can affect microbial communities, and flavonoids are no exception. With the increase of flavonoids, alpha diversity and observed_otus of the endophytic community decreased (Figure 2). Besides, we observed gram-negative bacteria accounted for the majority of total microorganisms, and the abundance of gram-negative bacteria increased from S1 to S4 (Figure 3). As the flavonoids increased, some gram-positive bacteria (including Firmicutes and Actinobacteria) decreased. We further analyzed the influencing factors of the microbial community in leaves by analyzing the correlation between the microbial community and factors such as flavonoid content and altitude (Figure 4). Flavonoids and altitude were proven to regulate the endophytic bacterial community in Ginkgo biloba leaves. Flavonoids are known to be antibacterial agents. Studies showed that the accumulation of flavonoids affected the symbiosis of endophytic bacteria and plants (Abdel-Lateif et al., 2012). Thus, flavonoids may be a key factor in the loss of diversity.

Furthermore, we found some microbiomes positively correlated with flavonoids, such as Oscillibacter, Moraxella, Clostridium and Parabacteroides. Microbes of Clostridium were proved to be flavonoid-degrading bacterium (Gupta et al., 2019). Parabacteroides were also found to increase with the addition of flavonoids in the gut microbiome (Zhang et al., 2022). In contrast, Oscillibacter and Moraxella decreased with the addition of flavonoids in the gut microbiome (Goulart et al., 2019; He et al., 2021; Cao et al., 2022). This difference might be because these microorganisms in Ginkgo leaves already live in an environment with high flavonoid content and have developed the resistance to flavonoids. Staphylococcus showed a significant negative correlation with T_flavonoid and kaempferol, the same as our previous study (Deng et al., 2021). Thus, we indicated that flavonoids could affect the community’s composition and structure in Ginkgo leaves.

Microorganisms facing stress need to develop some resistant strategies for survival. Our results showed that the network of endophytic bacteria in ginkgo leaves became more complex with the increasing altitude. Thus, it can be inferred that the adjustment of microbial interactions could be a strategy for microbes to resist enhancing environmental stresses, such as the increase of flavonoid content, altitude, and light intensity.

In general, the environmental selection of microorganisms is mainly reflected in the collection of functions (Hammesfahr et al., 2011). The environmental information processing of endophytic bacteria in leaves significantly improved along with the increase of flavonoid content (Figure 5A), which also proved that the increase of altitude and flavonoids increased the environmental stress on endophytic bacteria in Ginkgo biloba leaves. Ginkgo leaves needed more environmental information processing genes to deal with the greater pressure caused by more flavonoids and more light with increased altitude. The analogical finding deemed environmental stress to be the momentous reason for enriching environmental information processing in early inquiry (Debroas et al., 2009). Furthermore, the abundance of genes on the photosynthetic pathway decreased with the flavonoids (Figure 5B). Microorganisms of Cyanobacteria were the dominating photosynthetic organisms in the endophytic communities. So the decrease of photosynthetic genes might attribute to the decreasing Cyanobacteria. We also observed genes related to flavonoid biosynthesis dropped with higher flavonoids. In vitro, flavonoid biosynthesis was inevitably inhibited by flavonoids’ accumulation by feedback inhibition (Yin et al., 2012). So we hypothesized that flavonoids could also inhibit the accumulation of microorganisms that could produce flavonoids in leaves. In conclusion, flavonoids played an essential role in coordinating endophytic community functions in Ginkgo leaves.