Basic characteristics of all participants
The demographic and clinical data of the study population are reported in Table 1. Total 120 ESCC patients were recruited from two independent clinical centers. There were 89 males and 31 females. The median age was 61 years. The majority of the tumors originated from the middle(47.50%) and lower(43.33%) esophagus. Most of the patients were with advanced tumor(60.00%). About 70.83% had a high risk index.
Esophageal microbial diversity analysis
For alpha diversity, four characteristic metrics were evaluated. The alpha diversity in para-cancerous was significantly higher than cancerous tissue except for Pielou evenness index(Faith’s phylogenetic diversity P<0.001,Observed OTUs P<0.001, Shannon index P=0.027)(Figure 1A). The difference of alpha diversity between paired cancer and para-cancerous tissue only observed in regions(Faith’s phylogenetic diversity P<0.001, Observed ASVs P<0.001) and sampling seasons(Faith’s phylogenetic diversity P<0.001, Observed ASVs P=0.003) (Figure 1B).
In multivariate Adonis test, the beta diversity of cancerous tissue was significantly different compared with para-cancerous tissue whichever measured by Jaccard distance(P=0.004), Unweighted unifrac distance(P=0.004) or Weighted unifrac distance(P=0.028) (Figure 2A). About 1%-2% of the variance in beta diversity was explained by tissue type. What is more, the differences of beta diversity between within paired cancerous and para-cancerous tissue were associated with regions(P=0.032) and sampling seasons(P=0.042) based on Unweighted unifrac distance(Figure 2B).
Microbial composition analysis
Pair-wise PCoA results were displayed in Figure 4A. Based on Bray Curtis, Jacarrd, and Unweighted unifrac distance, microbiota from the two tissue types were separated into clusters.Total 9,453 features were found after the sequence denoised in all samples(Figure 6A). More taxa were observed in para-cancerous compared with cancerous tissue(8,133 vs 6,533) and about 5,213 taxa were detected in both tissue(Figure 3B). The composition of esophageal microbiota between cancerous and para-cancerous tissue in level phylum and genus was shown in Figure 3C and Figure S1, repectively. To elucidate the phylogenetic relationship of the cancerous and para-cancerous microbiota, the heat trees of microbiota(the relative abundance >0.1%) were plotted in Figure 3D. The relative abundance of correspondent branches of bacteria in phylum Proteobacteria, Bacteroidetes, Fusobacteria and Firmicutes was similar between cancerous and para-cancerous tissue. However, some bacteria were enriched in different tissues(Figure 3E). In class Alphaproteobacteria, the branches of bacteria of order Rhizobiales and Sphingomonadales were enriched in cancerous tissue while order Rhodospirillles was higher in para-cancerous tissue. The most bacteria of class Alphaproteobacteria and its branch were enriched in par-cancerous tissue except family Klebsiella. Moreover, the microbiota from phylum TM7 and its branches of bacteria, class Bacilli, family Helicobacteraceae and its genus Helicobacter and its species pylori were higher in para-cancerous tissue.
Differential abundance analysis
A total of 56 differential taxa were picked out by ANCOM2 algorithm(Figure 3G). As the host region exerted the strongest effect on microbiota, we divided all participants into Zhangzhou group and other regions group(Table S1).There were 32, 36, 16 differential abundance taxa between cancerous and para-cancerous tissue in all regions group, Zhangzhou city group and other regions group, respectively(Figure 3F). There were only three shared differential bacteria in cancerous and para-cancerous tissue from different regions named family Enterobacteriaceae, unclassified species from genus Sphingomonas and genus Phyllobacterium (Figure 3G). It was clearly ascertained differential bacteria in phylum Firmicutes and Proteobacteria were enriched in para-cancerous tissue from Zhangzhou city. Moreover, we observed an interesting alteration that the unclassified species in genus Mycoplane was enriched in cancerous tissue from other regions whereas enriched in para-cancerous tissue from Zhangzhou city. Sampling seasons were another powerful factor that influenced host microbiota(Figure 3G). The microbiota in phylum Cyanobacteria in cancerous tissue sampled in summer had a higher relative abundance. The relative abundance of bacteria in phylum Proteobacteria were significantly enriched in para-cancerous tissue when sampled in spring and summer from Zhangzhou city.
Hence, the dominant candidate differential taxa(Table 2) between cancerous and para-cancerous tissue were selected according to the grand means of relative abundance were exceeded 0.1% from above 56 differential taxa. They were species R.Mucilaginosa, P.Endodontalis, unclassified species in genus Leptotrichia, unclassified species in genus Phyllobacterium, and unclassified species in genus Sphingomonas, which enriched in cancerous tissue. On the other hand, class Bacilli, N.Subflava, H.Pylori, A.Parahaemolyticus, A.Rhizosphaerae, unclassified species in genus Campylobacter and unclassified species in genus Haemophilus were enriched in para-cancerous tissue. Next, to explore the confounding effect of regions, sampling seasons, tumor location and risk index on host microbiota, we added these four covariates into ANCOM2 analysis. The results showed that the relative abundance of unclassified species in genus Leptotrichia, unclassified species in genus Sphingomonas and A.Rhizosphaerae could be influenced by regions; the relative abundance of unclassified species in genus Campylobacter could be influenced by sampling seasons and tumor location. However, none of the differential taxa were significantly various in either low or high risk index with ESCC.
Microbial co-occurrence networks
To understand the interaction among esophageal microbiota in cancerous and para-cancerous tissue, we illustrated the microbial co-occurrence networks of two groups. There were 3,089 positive and 348 negative correlations in cancerous tissue, while the para-cancerous samples had 3,761 positive and 355 negative correlations. Obviously, the microbial co-occurrence networks were distinct between the cancerous and para-cancerous tissue (Figure 4A). But, widely correlations were investigated in family Lachnospiraceae, species C.aerofaciens and unclassified species in genus Blautia in both of the two tissue types.
To quantify such difference, we counted the number of node and its centrality in the microbial networks under different tissue type. As expected, the numbers of interacted microbiota in para-cancerous tissue(273 nodes) were higher than that in cancerous tissue(201 nodes)(Figure 4A and B). The centrality of co-occurrence networks was described with three dimensions which were degree, betweenness, and closeness centrality, respectively. Interestingly, the degree and closeness of shared nodes between cancerous and para-cancerous tissue were quite different. Next, the edges of the networks were evaluated (Figure 4C). Despite having a few overlapped edges, the distribution of rank of overlapped edges varies in cancerous and para-cancerous tissue.
To reveal the importance of above candidate differential taxa in the network, the heatmap was shown in Figure 4D. The most of differential taxa in the network were from phylum Proteobacteria and certain differential taxa did not emerged in the co-occurrence networks The role of differential taxa in different networks(cancer vs para-cancer) was also various. The discrepancies of microbial co-occurrence networks in two groups may be attributed to the specific metabolism in different tissue types.
The association between esophageal microbiota and predict function
Most of the differential taxa in para-cancer were negatively associated with EC 18.104.22.168 which regulated the EGFR, ERBB2, ERBB4, and FGFR1 signaling pathways. Different from this, the differential taxa in cancer were positively associated with EC22.214.171.124 and EC126.96.36.199 in the MET and PTEN signaling pathways respectively (Figure.5A).
Among the eight MetaCyc metabolic pathways which were significantly differed between cancer and para-cacner tissues (Figure 5B, Table S2), the relative abundance of PWY-3661 and PWY-7431 was increased in cancerous tissue, and other six pathways(PWY-1882, PWY-5265, PWY-6565, PWY-6731, PWY-6906, and PWY-7391) enriched in para-cancerous tissue. Notably，the differential taxa played various roles in different tissues. For example, unclassified species from genus Phyllobacterium and Sphingomonas were positively associated with PWY-6906 and PWY-1882 in cancer, whereas negatively with PWY-6731and PWY-6565 in para-cancer. Similarly, the correlation was demonstrated between taxa which were A.Parahaemolyticus and A.Rhizosphaerae and pathways including PWY-7431, PWY-7391, PWY-6731, and PWY-1882 in cancer, but PWY-6565 in para-cancer. In addition, the unclassified species in genus Haemophilus only associated with the pathways in cancer.