Analysis of the sequencing data

A total of 1,434,573 effective tags of fungal samples were obtained after filtering low-quality and other unsuitable sequences. The number of effective tags per sample ranged from 72,677 to 84,744, and the average number of clean reads was 79,699 per sampled group. The sequence lengths of all samples are mainly concentrated in 200-300 bp and 300-400 bp, accounting for 89.9% and 10.0%, respectively. The quality of the sequencing data was evaluated primarily through the statistics of sequence number, sequence length, GC content, Q20 and Q30 quality values, effective ratio, and other parameters in each sample (Table 1).

Table 1 The statistics and quality evaluation of the sequencing data

The rarefaction curves, displaying the relationship between the number of reads and operational taxonomic units (OTUs) in each sample, exhibited a stable plateau with the increase of the sample size (Fig. 1a). The results indicated that the sequencing depth and the number of OTUs were sufficient for each sample to represent the fungal communities and continue with further analyses. The rarefaction curves also showed that the abundance of fungal community in highest in leaf samples and lowest in root samples. The Good’s coverage values for the eighteen samples ranged from 98.2% to 100% (Fig. 1b), also indicating that the sequencing data confidently reflected the structure of the endophytic fungi community of the samples.

Fig. 1

Rarefaction curves of endophytic fungi based on the ITS2 sequences from each group. a. Observed OTUs numbers. b. Good coverage. Each group was comprised of 3 biological replicates (n = 3). CDR, CDS and CDY represent root sample, stem sample and leaf sample from Chengdu respectively; QXR, QXS and QXY represent root sample, stem sample and leaf sample from Qixianhu respectively

Taxonomic Analysis of Endophytic Fungi

A total of 2,521 operational taxonomic units (OTUs) were obtained from six groups, and 1,829 and 1,288 OTUs were detected in Chengdu samples and Qixianhu samples, respectively. Moreover, 140 OTUs and 102 OTUs were common to Chengdu samples and Qixianhu samples, respectively (Fig. 2a, b). As depicted in the petal diagram, 25 common OTUs were present in each sample’s, indicating that there may be great differences in endophytic fungi between the two areas due to different ecological conditions (Fig. 2c).

Fig. 2

Distribution difference of endophytic fungi in six group from two different ecological areas. Description: The Venn diagram (a and b) and petal diagram (c) based on operational taxonomic units (OTU), which represent common or unique OTUs to a given group

The taxonomic distribution of endophytic fungi in the roots, stems, and leaves of H. serrata is displayed in Fig. 3. After screening out rare OTUs, the remaining OTUs represented 9 phyla, 40 classes, 102 orders, 228 families, and 430 genera, respectively. At the phyla level, the OTUs were assigned to 8 knowns fungi, which were Ascomycota, Basidiomycota, Zygomycota, Glomeromycota, Chytridiomycota, Olpidiomycota, Mucoromycota, and Mortierellomycota. According to the results of multiple sequence alignment of features sequences of these phyla, the evolutionary tree of feature sequences is constructed (Fig. 3a). Among them, the predominant phylum was Ascomycota (54.34%, 41.14%-62.30%), followed by Basidiomycota (41.51%, 32.42%-57.17%), Fungi_unclassified (1.83%, 0.49%-3.34%), Zygomycota (0.58%, 0.15%-3.41%) and Glomeromycota (0.62%, 0%-3.52%). Among them, Basidiomycota, Ascomycota, Zygomycota, and Olpidiomycota were found in all tested samples. Otherwise, all Glomeromycota were found in root samples and stem samples, but not found in leaf samples from two sites. Chytridiomycota and Mucoromycota were only found in roots samples from Chengdu (CD), but not found from Qixianhu (QX). In addition, Mucoromycota and Mortierellomycota were only found in root samples and leaf samples from Chengdu (CD), respectively. At the genus level, a total of 430 distinct fungal genera were identified, and the compositions and proportions of the genera were significantly different among different tissues and different ecological areas. The genus Ascomycota was the most abundant in leaf and stem samples (CDY, CDS, QXY, and QXS), with relative abundances ranging from 21.45% to 28%. Whereas, Ascomycota genus was relative low abundance in root samples, with the relatively abundance was 0.77% and 1.67% in CDR and QXR, respectively. And Piskurozyma genus showed similar features. In contrast, Cladophialophora and Mycena genera were more abundant in root samples than in leaf and stem samples. It is remarkable to mention that Sebacina was the second dominant genus in Chengdu samples (CDR, CDS, and CDY), which account for 18.54%, 15.74%, and 11.76%, respectively, but it was hardly detected in Qixianhu samples (QXR, QXS, and QXY) (Fig. 3b).

Fig. 3

The relative abundance and heatmap of endophytic fungi in the six groups at different taxonomy levels. (a): Relative abundance of fungi at the phylum level. (b): Relative abundance of fungi at the genus level with a relative abundance of more than 1%. (c) and (d) The heatmap of shows the absolute abundance of taxa for endophytic fungi at the phylum and genus level, respectively

The top 9 classes of endophytic fungi in H. serrata were selected to make a clustering heatmap, which further indicating that species distributions differed greatly across the three tissue samples and different ecological areas. The heatmap representation of the results showed that the root samples, i.e., CDR and QXR clustered together, exhibiting a relatively similar community structure (Fig. 3c). The top 30 genera (i.e., those with relative abundance > 1%) were also selected to make a clustering heatmap (Fig. 3d). At the genus level, Auricularia and Mycena were the dominant genera in root samples (CDR and QXR), while Mortierella, Pestalotiopsis, Cladophialophora Agaricomycetes, and Herpotrichiellaceae were more abundant in the QXR samples than in CDR samples. The results of fungal communities showed that CDR and QXR samples clustered together, as did CDS and QXS, CDY and QXY, exhibiting a relatively similar community structure. The results hinted that the origin of endophytic fungi in roots is different from that in leaves and stems.

Alpha diversity analyses of the endophytic fungal communities in H. serrata. from different ecological areas

Alpha diversity analyses, including Shannon, Simpson, Chao1 and ACE indices, were conducted by Wilcoxon rank-sum test, to characterize differences in fungal community abundances and diversities in different groups. The results of the alpha diversity analysis of the fungal communities indicated that the Simpson index of the six samples was not significant (Fig. 4a). Specifically, the Shannon index of CDS and CDY samples was significantly higher than CDR samples (P < 0.05) (Fig. 4b). The Chao1 index of the CDS samples was significantly higher than those of CDR and QXR samples, while the Chao1 index of the QXR samples was significantly lower than those of except five samples (P< 0.05) (Fig. 4c). The ACE index of CDR samples was significantly lower than those of CDS and CDY samples, and the ACE index of QXR samples was significantly lower than that of QXS and QXY samples (P < 0.05) (Fig. 4d).

Fig. 4

Violin of the alpha diversity indices of the endophytic fungal communities of H. serrata Simpson (a) Chao1 (b) ACE (c) and Shannon (d) indices. Each violin represents the distribution of diversity present in three replicates (n = 3). Different lower-case letters represented a significant difference (P < 0.05) was assessed by one-way ANOVA followed by the Duncan's multiple range test.

Beta diversity analyses of the endophytic fungal communities in H. serrata. from different ecological areas

Principal coordinate analysis (PCoA), used to revealed variations among different H. serrata samples, was performed based on the unweighted Unifrac distance matrix. In the PCoA result showed that the first axis and the second axis explained 21.95% and 16.8% of the data’s variability, respectively. (Fig. 5). It revealed that the structures of the endophytic fungal communities of the stem and leaf samples (CDS and CDY, QXS and QXY) were relative similar. In contrast, the endophytic fungal communities in roots (CDR and QXR) were distinctly separated from those of the stems and leaves (CDS and CDY, QXS and QXY). Interestingly, the endophytic fungal communities of same tissue samples (CDR and QXR, CDS and QXS, CDY and QXY) from different ecological areas were distinctly separated (Fig. 5a). Similar results were found in Non-metric multidimensional scaling (NMDS) analysis, hinting that plant tissues and ecological areas affected the structure and diversity of endophytic fungal communities fin H. serrata (Fig. 5b).

Fig. 5

Beta diversity analysis of endophytic fungal community based on unweighted UniFrac distance for the H. serrata samples. a. Principal coordinate analysis (PCoA) plot. b. Non-Metric Multi-Dimensional Scaling (NMDS) analysis. c. UPGMA tree of different fungal community structures at the genus level in the different samples

UPGMA tree, conducted by unweighted unifrac method based on the genus level, revealed that two main different clusters were observed. Among them, the endophytic fungal communities from leaf and stem samples (CDS and CDY, QXS and QXY) clustered together, while the root samples (CDR and QXR) clustered alone and distinctly separated from the stem and leaf samples. Otherwise, the endophytic fungal communities of the leaf samples and stem samples in same ecological areas (CDS to CDY, QXS to QXY) were more similar than that of different ecological areas (CDS to QXS, CDY to QXY) (Fig. 5c). The results suggested that the endophytic fungal communities of the root samples might have species-specific, and those of the leaf and stem samples probably have ecological specificity.

Linear discriminant analysis effect size (LEfSe) analysis was used to discover the biomarkers of endophytic fungal community among the six samples from two ecological areas. A total of 53 biomarkers were discovered and employed to discern significant differences among six samples with an LDA score greater than 3.0. The result showed that more taxa with statistically significant abundance in Chengdu samples than in Qixianhu samples at the genus level. Among them, the CDY samples contain more Tremellales_unclassified, and Tremellomycetes_unclassified, CDS samples contain more Piskurozyma, Strelitziana, Pleosporales_unclassified, Spizellomycetaceae_unclassified, Rhinocladiella, Halosphaeriaceae_unclassified, Tremella, and Veronaea. And Auricularia, Gliocladium, Ilyonectria, Cotylidia, Clavulinopsis, Olpidiaceae_unclassified, and Chaetothyriaceae_unclassified are more abundant in CDR samples (Fig. 6a). In contrast, Auricularia, Glomus, Rhizophagus, and Glomeraceae_unclassified were significantly enriched in QXR samples, while Eurotiomycetes_unclassified, Tremellales_unclassified, and Strelitziana were more abundant in QXY samples. And Septobasidium and Teichosporaceae_unclassified were more abundant in QXS samples (Fig. 6b).

Fig. 6

Linear discriminant analysis effect size (Lefse) analysis of differentially abundant taxonomic clades in fungal communities from Chengdu (a) and Qixianhu(b) with an LDA score higher than 4.0. The figure shows the taxa with an LDA score greater than 4.0. The length of the horizontal bars represents the effect size for each taxon (LDA score), and different colors represent different grouped taxa

Correlation Analysis between fungal endophytes diversity and Hup A content

The content of Hup A in different samples was quantified by high performance

liquid chromatography (HPLC) (Fig. 7). The retention time of standard Hup A was 17.989 min, and the HPLC spectrum of Hup A standard is shown in Fig. 7a. The results showed that the content of Hup A in roots (CDR and QXR) was significantly lower than that in stems and leaves. The Hup A content in Qixianhu samples was significantly higher than that in Chengdu samples, hinting that Hup A content might have variety specificity (Fig. 7b). The correlation between the top 30 OTUs of endophytic fungal community and Hup A content is depicted using Pearson heat map (Fig. 8). There were 7 genera ( Fungi_unclassified, Pestalotiopsis, Rhodotorula, Ascomycota_unclassified, Cyphellophora, Sporobolomyces, and Trichomeriaceae_unclassified) were significantly and positively correlated to Hup A content of Chengdu samples(CI ≥ 0.95). At the same time, there were 7 genera (Mortierella, Russula, Auricularia, Mycena, Tomentella, Chaetothyriales_unclassified, and Cladophialophora) were significantly and negatively correlated to Hup A content of Chengdu samples (CI ≤—0.95) (Fig. 8a). On the other hand, there were 10 genera (Carlosrosaea, Ascomycota_unclassified, Sporobolomyces, Fungi_unclassified, Trichomeriaceae_unclassified, Basidiomycota_unclassified, Chaetothyriales_unclassified, Bionectria, Phialophora, and Trechispora) were significantly and positively correlated to Hup A content of Qixianhu samples (CI ≥ 0.95).There were 7 genera (Fungi_unclassified, Pestalotiopsis, Rhodotorula, Ascomycota_unclassified, Cyphellophora, Sporobolomyces, and Trichomeriaceae_unclassified) were significantly and negatively correlated to Hup A content of Chengdu samples(CI ≤ -0.95) (Fig. 8b). Of which, there are 6 genera (Ascomycota_unclassified, Cyphellophora, Fungi_unclassified, Sporobolomyces, and Trichomeriaceae_unclassified) were significantly and positively correlated to Hup A content in all two areas, whereas, there are 6 genera (Auricularia, Cladophialophora, Cryptococcus, Mortierella, and Mycena) were significantly and negatively correlated to Hup A content in all two areas. These genera, which showed positively or negatively correlated Hup A content of in all two areas, may probably have species-specific. However, those genera which showed positively or negatively correlated Hup A content in only one area may probably have eco-environmental specificity.

Fig. 7

HPLC analysis of the Hup A standard (a) and Hup A contents in different samples (b). The retention time for Hup A standard was 17.989 min. Different letters on the bars indicate significant differences among all means in different treatments using a one-way ANOVA followed by the Duncan's multiple range test (P < 0.05)

Fig. 8

The correlation of Hup A content with endophytic fungi community from Chengdu samples (a) and Qixianhu samples (b). The correlation index (CI) was shown in figures

Pearson correlation analysis showed that the Hup A contents were significantly positively correlated with endophytic fungal ACE index, and positively correlated with Chao1 and Shannon’s diversity index of endophytic fungi in H. serrata (Fig. 9).

Fig. 9

Pearson correlation analysis between diversity of endophytic fungi and Hup A content. Note: ** indicate the differences are significant at P < 0.01 and * indicate the differences are significant at P < 0.05