Regional distribution of Christensenellaceae and its associations with metabolic syndrome based on a population-level analysis

Data collection and criteria for MetS

The data used in this analysis from the GGMP were described previously (He et al., 2018b). In brief, 9,172 individuals were investigated, and 6,896 people finished the survey and household sampling (2,276 missing metadata). By further filtering the metadata, some individuals with sequence number less than 10,000 were excluded from this analysis, so a total of 6,879 individuals were retained. The GGMP employs an in-person questionnaire to collect individual metadata. To evaluate how Christensenellaceae members affect the metabolism and gut microbes, we also compiled information on whether the individuals had diabetes, dyslipidemia, and pre-diabetic symptoms.

MetS was diagnosed as described by the Joint Committee for Developing Chinese Guidelines on Prevention and Treatment of Dyslipidemia in adults, based on three of the following five criteria for participants: (1) waist circumference >90 cm (male) or >85 cm (female), (2) fasting blood glucose (FBG) ≥ 6.1 mmol/L (110 mg/dl) or previously diagnosed with diabetes, (3) triglyceride (TG) ≥ 1.7 mmol/L (150 mg/dl), (4) high-density lipoprotein (HDL) <1.04 mmol/L (40 mg/dl), and (5) systolic/diastolic blood pressure (SBP/DBP) ≥ 130/85 mmHg or previously diagnosed with high blood pressure.

Sample collection and DNA extraction

Stool samplers, ice bags and ice boxes were provided to collect and store samples after the questionnaire survey. After defecation, each participant recorded their Bristol stool score and stored the sample in an ice bag. All the samples were stored in a freezer (−18 °C to −20 °C) for less than 3 days and then transported to the research laboratory (Guangdong CDC) in a cold-chain vehicle to maintain a low-temperature environment. Samples were transported and stored at the research laboratory in −80 °C freezers until further processing.

A total of 200 mg of each fecal sample was used for DNA extraction using the Fecal DNA Bead Isolation Kit (Bioeasy, Shenzhen) according to the manufacturer’s instructions. Before the specimens were submitted to laboratory analysis, we prepared external standards to control for potential batch effects because multiple technicians and machines were involved in sample processing. Briefly, fecal samples were collected from three donors. For each donor, the samples were manually homogenized to obtain an even mixture, divided into 200 tubes and stored at −80 °C. All stool samples were processed with identical protocols, including three external standards for each batch. For each DNA sample, the bacterial 16S rRNA gene was amplified with the following barcoded primers (shown from 5′ to 3′): V4F (GTGYCAGCMGCCGCGGTAA) and V4R (GGACTACNVGGGTWTCTAAT) (Walters et al., 2016). The primers contained Illumina adapters and a unique 8-nucleotide barcode. The PCR conditions included an initial denaturation at 94 °C for 5 min; 30 cycles of denaturation at 94 °C for 30 s, annealing at 52 °C for 30 s, and elongation at 72 °C for 45 s; and a final extension at 72 °C for 5 min. The products were submitted for next-generation sequencing on an Illumina HiSeq 2500 platform using 500-cycle version 2 reagent kits (Beijing Genome Institute, BGI, Beijing).

Microbiome bioinformatic analysis

Raw sequence data were managed and analyzed using Quantitative Insights Into Microbial Ecology software (QIIME, version 1.9.1) (Caporaso et al., 2010), and the sequences with Phred quality scores below 20 were then discarded. The method for processing of sequences was identical to that described in our previous reports (He et al., 2018b). PCoA based on unweighted UniFrac distances comparing bacterial community structure of samples between G1 and G2. Permutational multivariate analysis of variance (PERMANOVA) was carried out to measure effect sizes and significance differences in beta diversity. The threshold of statistical significance was set at P < 0.05. We carried out multivariate association analyses with linear modeling (MaAsLin) as described by Morgan et al. (2012) to examine the relationship between sequential taxonomic units and each of the MetS diagnostic factors and several related diseases. Age was used as a confounder, and the false discovery rate was limited to 0.05. The R package (ggtree) (Yu et al., 2018) was used to visualize the evolution tree data after multi-sequence comparison with QIIME. Data plotting and statistical analyses were performed by R (3.2.2) statistical software.

Ethics approval and consent to participate

The present study was approved by the Ethical Review Committee of the Chinese Center for Disease Control and Prevention under approval notice no. 201519-A. Written consent was obtained from all participants.

Statistical analysis

The significance of differences between two groups was resolved by the Wilcoxon rank-sum test. Spearman’s rank correlation test was applied to analyze the correlation between two variables. The chi-square test was utilized to compare the ratios of two groups. P values less than or equal to 0.05 were considered significant. The Benjamini and Hochberg method was used to modulate the P value for multiple hypotheses.

The overall gut Christensenellaceae configuration of people in the GGMP

Exploration of the effects of the gut microbiota requires data from studies performed with a regionalized study design, comprehensive sampling and standardized experimental protocols. The population included in the GGMP has been previously described (He et al., 2018b). In the GGMP project, 6896 individuals were included according to our entry criteria, which followed the guidelines of the Joint Committee for Developing Chinese Guidelines as mentioned in a previous article (He et al., 2018). In the GGMP study, totally 6896 samples were characterized by 16S rRNA gene sequencing, and more than 17,083 quality-filtered sequences were obtained through QIIME analyses.Based on our statistics, the family Christensenellaceae accounted for an average of 0.11% of human fecal bacteria in the residents. Then, we grouped the individuals based on the abundance of Christensenellaceae; in total, 3316 people had no Christensenellaceae in their intestines (Group1, G1), and we selected the upper quartile population as Group2 (G2, n = 1465) according to Christensenellaceae abundance. The abundance of Christensenellaceae in this study was bias distributed, the overall median value was 0, and the maximum value was 15.1%. Statistics found that the average abundance of Christensenellaceae in the G2 group was 0.49%.To characterize the diversity and richness of the bacterial community, the alpha indices, estimated by four different parameters, were analyzed for each sample. As shown in Fig. S1, the diversity and richness estimators in the two cohorts were significantly different. Compared with the G1 subjects, the G2 subjects had a significantly high alpha diversity indexes, such as the Chao1, observed OTU, PD_whole tree and Shannon indexes (p < 0.001 for each). To measure the degree of similarity of the fecal microbial communities, we performed a principal coordinate analysis (Fig. 1A), and we found obvious differences between the two groups (R2 = 0.07184, P = 0.001). Additionally, we applied linear discriminant analysis effect size (LEfSe) (Segata et al., 2011) for quantitative analysis of biomarkers within different cohorts. A total of 20 features had significantly different abundances between G1 and G2(LDA>3) (Fig. 1B). At the genus level, the fecal microbiota of people who lacked Christensenellaceae was enriched with the taxa Proteobacteria, Enterobacteriales, Bacteroidaceae, Lachnospiraceae and Klebsiella, whereas individuals with a high proportion of Christensenellaceae exhibited enrichment of Ruminococcaceae, Mollicutes, RF39, Akkermansia and Rikenellaceae.

To examine the Christensenellaceae components in detail, we analyzed the sub-OTUs at the family level. After chimera removal and quality filtering, a total of 134 different sub-OTUs of Christensenellaceae were identified at the 100% sequence identity level, and eight of them had read counts greater than 1,000 (Fig. 1C). An evolutionary tree was constructed to further observe the evolutionary distance of each sub-OTU and several reference genome (Fig. S1).

The distribution of Christensenellaceae in GGMP

In our survey for the GGMP study, the distribution of Christensenellaceae in Guangdong Province varied based on geographic location. According to the average abundance data, Christensenellaceae is unevenly geographically distributed, with the abundance typically reduced in or near large and economically thriving centers such as Guangzhou (0.08%) and Shenzhen (0.09%). In contrast, Shanwei had the highest abundance (0.22%), followed by Huizhou (0.19%) and Meizhou (0.18%). The abundances in other cities is shown in Fig. 2A. In addition, we collected data on per capita annual income and Christensnellaceae abundance in each region (Fig. 2B). According to the map, the higher the level of urbanization, the lower the abundance of Christensenellaceae in the intestinal tract of urban residents.

When the MetS prevalence in different cities was compared, there was a significantly negative correlation between MetS and Christensenellaceae abundance (Fig. 2C). Residents in Zhanjiang, Shanwei, Foshan and Maoming had Christensenellaceae abundances higher than 0.15% in their gut and exhibited a low prevalence of MetS. Similarly, city dwellers in Guangzhou, Shenzhen and Shaoguan had a low Christensenellaceae abundance, and a high proportion of the population exhibited MetS. However, a negative correlation was not observed in residents in Jieyang and Meizhou. Notably, the prevalence of MetS in G2 subjects was lower than that in G1subjects in most regions (Fig. 2D). Moreover, there was a significant difference in the prevalence of MetS in people with different abundances of Christensenellaceae in Qingyuan (p < 0.01) and Shanwei (p < 0.05). Although Christensenellaceae abundance varies in different regions, the negative correlation between MetS and Christensenellaceae is universal.

Christensenellaceae is associated with host metabolic index and MetS status

To explore the relationship between Christensenellaceae abundance and human health, we next assessed how variable the body parameters were in terms of Christensenellaceae richness. We selected 4,781 subjects associated with Christensenellaceae from the GGMP project, and subjects in this study cohort had a wide range of age, BMI and blood test indicators (Table 1). As previously described, we classified the subjects into two categories (G1 and G2) according to the abundance of Christensenellaceae. The differences between the two groups were mainly reflected in the following aspects: age, anthropometric parameters and biochemical criteria (Table 1). Because variations in metabolites could be related to differences in age (Dunn et al., 2015), we reanalyzed the correlation between the indices after adjusting for age. Compared to the G1 group, people rich in Christensenellaceae showed significantly low BMI, waist circumference and waist-to-height ratio (WHtR). The high-value group also showed lower levels of biochemical indices, such as TG, alanine transaminase (ALT) and uric acid (UA), than the non-Christensenellaceae group (Fig. S2). The level of HDL was significantly higher in people with Christensenellaceae, indicating the abundance of Christensenellaceae has a positive correlation with HDL level.

In addition, we calculated the correlation coefficients between the waist circumference, WHtR, BMI, TG level and other clinical parameters of all subjects to evaluate the connection intensity between the abundance of Christensenellaceae and the host metabolic index (Fig. 3A). We found that waist circumference and WHtR remained significantly decreased in people rich in Christensenellaceae compared to the people who lacked Christensenellaceae. Recently, waist circumference and WHtR were proposed as predictors of the incidence of MetS (Perona et al., 2019; Suliga et al., 2019). Therefore, we measured the prevalence of MetS in individuals with GGMP. People rich in Christensenellaceae had low prevalence of metabolic diseases, including overweight, obesity, fatty liver disease, and hypertriglyceridemia.

We also tested for connections between the sequential taxonomic units of Christensenellaceae and the indicators to observe whether the changes in sub-OTUs at the family level of Christensenellaceae are consistent. These top eight sub-OTUs, with reads number greater than 1,000, had the same correlation with each parameter (Fig. 3A). They all had negative correlations with ALT, TG, UA, waist circumference and WHtR. Among the sub-OTUs, sub-OTU3228 had a significantly negative correlation with TG, ALT, UA and BMI. Nevertheless, most of the taxonomic units were positively related to HDL and blood urea nitrogen (BUN).

In addition to morphological indicators and circulating metabolites, we also identified an association between the family Christensenellaceae and metabolic diseases such as fatty liver disease, obesity, hypertriglyceridemia and digestive system disease (Fig. 3B). Moreover, we found that sub-OTU3228, sub-OTU5291, sub-OTU12860, and sub-OTU14698 were negatively correlated with overweight, obesity and hypertriglyceridemia. Although most of the associated taxa were shared across obesity and lipid metabolites, several sequences were predominantly linked to digestive diseases rather than metabolic disorders. Notably, the abundances of sub-OTU3228 and sub-OTU14698 were negatively associated with digestive disorders, while sub-OTU3228 showed a strong positive correlation with the occurrence of diabetes mellitus (T2DM). Besides the above four members, the remaining taxonomic units with read counts greater than 1,000 were not significantly associated with these metabolism-related disorders.

Functional properties predicted by PICRUSt

We performed PICRUSt analysis to predict the KEGG functional orthologs of the fecal microbiota metagenomes based on 16S rRNA sequences. Principal component analysis (PCA) showed that the KO profile of the gut microbiota in people with high abundance of Christensenellaceae diverged from that of the gut microbiota of people lacking Christensenellaceae (Fig. S3). 71.2% of the variation within these two groups was captured by the first principal component (PC1). As shown in Fig. 5, broad potential communication pathways were identified between the individuals, including metabolism, cellular process, environmental information process, genetic information process and human disease. The LEfSe algorithm was applied to detect differences in the functional pathways of the microbiota between the two groups. In total, 6 functional orthologs were significantly different in Christensenellaceae-rich people (LDA score>3); the enriched orthologs were nucleotide metabolism; cellular process; ribosome; translation; replication and repair; and genetic information processing. In contrast, the prevalent markers among the controls included those associated with transporters, membrane transport and environmental information processing (Fig. 5). Moreover, several metabolic pathways were abundant in the non-Christensenellaceae group, including those involved in energy metabolism (methane metabolism), lipid metabolism (fatty acid biosynthesis) and carbohydrate metabolism (fructose and mannose metabolism) (Table S1). In terms of metabolic disease, people with Christensenellaceae have a relatively low risk of diabetes mellitus (Table S1). These findings suggest that Christensenellaceae may affect the way in which we metabolize and regulate ponderal growth.