Study population

A total of 79 patients with PJS (average age, 25.06 ± 11.82; ratio of males to females, 41:38; BMI, 21.03 ± 1.75), patients with benign polyps (average age, 48.03 ± 11.02; ratio of males to females, 35:40; BMI, 23.71 ± 3.14), and 83 healthy controls (average age, 27.64 ± 8.41; ratio of males to females, 46:37; BMI, 20.42 ± 2.08) were included in the analysis. The demographic and clinical characteristics were shown in Table 1. A total of 39 patients (49%) had experienced at least one intussusception, and 44 of these 79 patients (56%) had at least one personal history of endoscopic surgeries. A total of 28 patients (35%) had a family history, and 2 of these 79 cases (2.5%) were diagnosed with a tumor. In addition, other clinical characteristics (age of onset, 14.00 [8.00 ~ 20.00]; frequency of endoscopic surgeries, 2.50 [1.80 ~ 4.00]; length of the biggest polyps, 10 [4 ~ 23]; frequency of endoscopic surgeries, 1 [0 ~ 1]) were analyzed in Table 1.

Table 1 The demographics of patients. BMI, body mass index; Mutation, mutation of STK11 gene; Lenth, the lenth of biggest polyps; Polyps number, number of the PJS polyps; Surgery frequency, frequency of endoscopic surgeries; SD, standard deviation; IQR, interquartile rangeDysbiosis of gut microbiota in PJS patients compared with healthy controls

To characterize the signatures of gut microbiota in PJS patients, we collected stool samples from PJS patients and healthy controls and performed High throughput sequencing analysis. We originally measured gut microbial α-diversity with Wilcoxon signed-rank sum test. Consistently, different indices, including ace, chao1, observed_otus, and Shannon, showed similar tendencies and significant differences between PJS patients and healthy controls (p < 0.01, p < 0.01, p < 0.01, p < 0.05, Wilcoxon signed-rank sum test) (Fig. 1a). To identify the difference in the microbial community (β-diversity) between the two groups, principal coordinate analysis (PCoA) was performed based on Bray-Curtis metric distance, unweighted-UniFrac distance, and weighted-UniFrac distance algorithms (p = 0.001, p = 0.001, p = 0.001, PERMANOVA test) (Fig. 1b, c and d). An apparent clustering separation between amplicon sequence variants indicated different community structures between PJS patients and healthy controls. Subsequently, we assessed the features of gut microbiota from the phylum to genus levels to further evaluate the differences in the composition of gut microbiota between patients with PJS and healthy controls. At the phylum level of gut microbiota, the dominant taxa at the phylum level included Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria (Figure. S1a). Firmicutes was the most predominant phylum, accounting for 62.99% and 57.22% of the gut microbiota in PJS patients and healthy controls, respectively (p = 0.019) (Figure S1b). Consistently, the abundance of Bacteroidetes was higher in the PJS group (p < 0.01), whereas that of Actinobacteria was higher in the control group (p < 0.01) (Figure S1c-1d). In addition, compositions of gut microbiota at the class, order, family, and genus levels were also compared between PJS patients and healthy controls (Figure S2a-S2d). At the genus level of gut microbiota, different biological compositions were observed between the two groups. In particular, the genera Bacteroides, Agathobacter, Lachnospira, Roseburia, and Ruminococcus_1 were significantly enriched in the PJS group (p < 0.001, p < 0.001, p < 0.001, p < 0.001, and p < 0.001, respectively). In contrast, the genera Subdoligranulum, Romboutsia, Blautia, Bifidobacterium, and Collinsella were significantly enriched in the HC group of healthy controls (p < 0.001, p < 0.001, p < 0.001, p < 0.05, p < 0.05) (Figure S3).

To confirm which bacteria were associated with PJS, linear discriminant analysis (LDA) effect size (LEfSe) analysis was performed. LEfSe analysis showed eight discriminative features (LDA > 3.5; p < 0.05 [Fig. 1e]) at the genus (n = 2), family (n = 1), order (n = 2), class (n = 1), and phylum (n = 1) levels. The relative abundance of phylum Bacteroidetes, class Bacteroidia, order Bacteroidales, family Bacteroidaceae, and genus Bacteroides was higher in PJS patients, whereas the phylum Firmicutes, class Clostridia, order Clostridiales, and genus Blautia were significantly enriched in healthy controls. Furthermore, based on the abundance of gut microbiota at the genus level, we performed a comparison heatmap for the analysis of gut microbiota between the two groups, and the results were consistent with those shown in Figure S3 ​(Figure 1f). Collectively, the analysis indicated differences in the composition and structure of the gut microbiota in patients with PJS and healthy controls.

Significant differences of gut microbiota between PJS patients and patients with benign polyps

Gastrointestinal hamartomatous polyps are characteristic features of patients with Peutz-Jeghers syndrome. To identify the differences in gut microbiota between patients with PJS and patients with benign polyps, sequencing analysis of 16 S rRNA was performed. As for α-diversity, ace, chao1, observed otus, and shannon indices showed significant differences between the two groups (p < 0.001, p < 0.001, p < 0.001, p < 0.001, Wilcoxon signed-rank sum test) (Fig. 2a). In addition, β-diversity, including Bray-Curtis metric distance, unweighted-UniFrac distance, and weighted-UniFrac distance, showed a significant clustering separation between the two groups (p = 0.001, p = 0.001, p = 0.001, PERMANOVA test) (Fig. 2b-d), which revealed the different structure and composition of the gut microbiota in PJS patients and patients with benign polyps. However, the age of the benign polyps group is considerably higher than that of the PJS group, in that case, we conducted covariance analysis to adjust for the possible confounding factors (age, gender, and BMI) between two groups (Table S1). The results indicated that significant differences were still observed between two groups after adjusting for these characteristics. In addition, LEfSe analysis was conducted to confirm which bacteria were markedly enriched in PJS patients, and the results showed that the phylum Bacteroidetes, class Bacteroidia, order Bacteroidales, family Bacteroidaceae, family Lachnospiraceae, family Prevotellaceae, genus Bacteroides, and genus Agathobacter were enriched in PJS patients. In contrast, kindom Bacteria, phylum Firmicutes, phylum Actinobacteria, class Clostridia, order Clostridiales, family Peptostreptococcaceae, genus Romboutsia, genus Subdoligranulum and genus Blautia were enriched in patients with polyps (Fig. 2e). The differences in gut microbiota at the genus level between the two groups are shown in a heatmap (Fig. 2f), and the detailed features of gut microbiota at the phylum, class, order, family, and genus levels were also analyzed (Figure S4-S6).

No significant difference of gut microbiota composition between STK11 positive and STK11 negative groups

The STK11/LKB1 gene is involved in cell proliferation and cell-cycle signaling pathways, and the mutation of STK11 is considered to be the cause of PJS; 50–90% of PJS cases are due to mutations in this enzyme [15,16,17]. To identify whether the mutation of STK11 influences the structure and composition of gut microbiota, we compared the composition of gut microbiota in STK11 positive [43] and STK11 negative [36] patients with PJS. The α-diversity indices (ace, chao 1 indices, observed otus, shannon) (p = 0.68, p = 0.78, p = 0.73, p = 0.77, Wilcoxon signed-rank sum test) and β-diversity index (Braycurtis metric distance, unweighted-unifrac distance, weighted-unifrac distance) (p = 0.236, p = 0.603, p = 0.591, PERMANOVA test) showed no significant differences between the two groups (Fig. 3a and d). Subsequently, to identify the differences in microbial abundance between the two groups, we analyzed the composition of the gut microbiota at the phylum, class, order, family, and genus levels. At the phylum level, Proteobacteria were riched in STK11 negative group (p < 0.05) (Figure S7d). At the genus level, Ruminococcus1 and EscherichiaShigella were riched in STK11 negative group (p < 0.05, p < 0.05), while Bifidobacterium was riched in STK11 positive group (p < 0.05) (Figure S9). The composition of the gut microbiota in each sample at the class, order, family, and genus levels are shown in Fig. 3e and Figure S8. The details of the gut microbiota composition at the genus level in each sample are shown in a heat map (Fig. 3e). In addition, we compared the expression of KEGG metabolic pathways between STK11 positive and negative patients. The results showed no significant differences in the expression of the top 20 metabolic pathways (Fig. 3f). According to these results, there was no significant difference in gut microbiota composition and abundance between the STK11 negative and positive groups, and the correlation between individual microbial community differences and genetic mutations still needs further validation.

Fig. 1

Dysbiosis of gut microbiota in PJS patients compared with healthy controls. (a) Alpha diversity boxplot (based on ACE, Chao1, observed otus and shannon). (b-d) Principal coordinate analysis (PCoA) using Bray-Curtis metric distances, Unweighted-UniFrac distance, and Weighted-UniFrac distance algorithms of beta diversity. (e) LEfSe analysis depicting taxonomic association between microbiome communities from PJS patients and healthy controls. (f) Heatmap of selected most differentially abundant features at the genus level

Fig. 2

Difference of gut microbiota in PJS patients compared with patients with benign polyps. (a) Alpha diversity boxplot (based on ACE, Chao1, observed otus and shannon). (b-d) Principal coordinate analysis (PCoA) using Bray-Curtis metric distances, Unweighted-UniFrac distance, and Weighted-UniFrac distance algorithms of beta diversity. (e) LEfSe analysis depicting taxonomic association between microbiome communities from PJS patients and patients with benign polyps. (f) Heatmap of selected most differentially abundant features at the genus level

Fig. 3

Difference of gut microbiota in STK11 positive patients compared with STK11 negative patients. (a) Alpha diversity boxplot (based on ACE, Chao1, observed otus and shannon). (b-d) Principal coordinate analysis (PCoA) using Bray-Curtis metric distances, Unweighted-UniFrac distance, and Weighted-UniFrac distance algorithms of beta diversity. (e) Heatmap of selected most differentially abundant features at the genus level. (f) Top20 KEGG Boxplot Glimpse

Synthesis of SCFAs decreased in PJS patients

Previous 16 S rRNA sequencing analysis showed that Subdoligranulum, Blautia, and Bifidobacterium in the gut microbiota of PJS patients were decreased compared to healthy controls, which is associated with the metabolism of short-chain fatty acids (SCFAs) [18,19,20]. In addition, we used the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2) (V2.3.0) for functional prediction [21], and then mapped to the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, the results indicated that expression of the fatty acid biosynthesis pathway was decreased in patients with PJS (Figure S10). In recent years, the importance of SCFAs in human health has been revealed, and SCFA are important fuels for intestinal epithelial cells (IEC) and regulate gut functions and host immunity [22]. In this case, in order to identify whether the metabolism of SCFA was altered in PJS patients, we performed a targeted metabolomics assay to detect 7 kinds of seven SCFAs in the feces of PJS patients and healthy controls, including acetic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, and caproic acid. First, the content of seven types of SCFAs in each sample is shown in a total heatmap (Fig. 4a), and the SCFAs with significant differences were evaluated using the Z-score (Fig. 4b). The results showed that acetic acid, propionic acid, and butyric acid were enriched in healthy controls (p < 0.001, p < 0.001, and p < 0.001, respectively) (Fig. 4c and e), while the other four types of SCFAs showed no significant difference between the two groups. Therefore, we found that the synthesis of SCFAs was decreased in patients with PJS, which may influence the development of PJS.

Fig. 4

The synthesis of short-chain fatty acids is reduced in patients with PJS. (a) Heatmap of 7 kinds of SCFAs between PJS patients and healthy controls. (b) Z-score of three different SCFAs with variations. (c-e) Violin plot of acetic acid, propionic acid, and butyric acid between PJS patients and healthy controls

Fig. 5

The correlation between gut microbiota and clinical features. (a) Heatmap of the correlation between selected gut microbiota at genus levels and clinical features. (b-c) Significant correlation between gut microbioat and clinical features

Fig. 6

The correlation between SCFAs and clinical features. (a) Heatmap of the correlation between three types of SCFAs with significant differences and clinical features. (b-h) Significant correlation between SCFAs and clinical features (Age: age of onset. Surgery frequency: frequency of endoscopic surgeries. Length: the length of the biggest polyps. Polyps number: number of the PJS polyps)

Fig. 7

Random forest prediction model based on PJS patients. (a) Variable importance of the selected gut microbiota in genus level. (b) The Area Under the Curve of the random forest model

Bacteroides is positively correlated with biggest polyps’ length while Agathobacter is negatively correlated with age at first gastrointestinal symptom

First, to identify the correlation between the gut microbiota and clinical characteristics in patients with PJS, a heatmap was drawn using Spearman’s correlation (Fig. 5a). The clinical features included age at first gastrointestinal symptom, frequency of endoscopic surgeries, largest polyp length, and number of PJS polyps. However, some gut microbiota with higher abundance at the genus level were selected, such as Bacteroides, Agathobacter, Fusicatenibacter, Roseburia, Escherichia_Shigella, Prevotella_9, Blautia, Faecalibacterium, Bifidobacterium, and Subdoligranulum. The analysis indicated that Bacteroides was positively correlated with the longest polyps’ length (r = 0.24, p = 0.035) (Fig. 5b), and Agathobacter was negatively correlated with age at first gastrointestinal symptoms (r = -0.29, p = 0.0085) (Fig. 5c). However, there was no significant correlation between the other gut microbiota and clinical features. These results indicate that Bacteroides may promote the growth of PJS polyps, and Agathobacter may be related to the onset of gastrointestinal symptoms in patients with PJS. However, further research is needed to demonstrate the influence of gut microbiota on the clinical characteristics of PJS patients.

Higher content of acetic acid, propionic acid, and butyric acid are correlated with the improvement of clinical symptoms in PJS patients

In addition, to explore the relationship between SCFAs and clinical characteristics of PJS patients, we conducted a Spearman correlation analysis between the three SCFAs with significant differences (Fig. 6), including acetic acid, propionic acid, butyric acid, and frequency of endoscopic surgeries, maximum polyp length, age of onset, and number of polyps. The results showed that acetic acid, propionic acid, and butyric acid levels were positively correlated with the age of onset and negatively correlated with the number of polyps. In addition, butyric acid level was negatively correlated with surgical frequency in patients with PJS. Therefore, patients with higher levels of acetic acid, propionic acid, and butyric acid tend to have a later onset of initial symptoms and fewer polyps. Patients with higher levels of butyric acid underwent fewer endoscopic surgeries. The correlation of microbial feature and patient characteristics was summarized in Table S1.

The occurrence of PJS could be predicted by the randomforest model of gut microbial signature

Recognizing that the features of gut microbiota may be a potential diagnostic biomarker for PJS, we established a random forest prediction model based on gut microbiota at the genus level to visualize specific taxa that contributed to the diagnostic potential between PJS and the control group. We used a training cohort containing 80% of patients with PJS (64/79) and a validation cohort containing 20% of patients with PJS (15/79). First, we set the seeds and drew a curve to describe the relationship between the number of decision trees and error rate. After variable importance screening, we selected the top Seventeen genus species, including Romboutsia, Ruminococcus_1, Lactobacillus, Escherichia_Shigella, Subdoligranulum, Lachnospira, Blautia, Erysipelotrichaceae_UCG_003, Agathobacter, Bacteroides, Parabacteroides, Roseburia, Eubacterium_hallii_group, Bifidobacterium, Akkermansia, Sarcina, and Ralstonia. The results of MeanDecreaseAccuracy and MeanDecreaseGini show the importance rank of each taxon (Fig. 7a). After training the model, its performance was verified in the validation cohort (AUC = 0.952; Fig. 7b).