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

View article


Metabolic syndrome (MetS) is a disease that brings together a variety of metabolic disorders. With economic development and the improvement of living standards, the incidence of MetS has also gradually increased, making this disease a global public health problem. MetS is characterized by abdominal obesity, dyslipidemia, hypertension, and elevated blood sugar, and it currently affects approximately 20 to 30% of adults worldwide (Grundy, 2008). There are several factors involved in the development of the disease, including host genetic factors (Kraja et al., 2011), eating habits and sedentary lifestyle (He et al., 2018), but the pathogenesis of MetS has not been fully elucidated.

With technological development, the role of the microbiome in human health has been increasingly emphasized. The gut microbiota impacts a range of human health conditions, including metabolic processes, immune-related diseases and neurological disorders (Integrative HMP Research Network Consortium, 2019; Festi et al., 2014; Fung, Olson & Hsiao, 2017). Studies on MetS and the gut microbiota in mouse models have shown that the development of MetS involves a combination of the gut microbiota, host genes and diet (Ussar et al., 2015; Zmora, Suez & Elinav, 2019). Specific gut microbiota, bacterial metabolic pathways and their interactions with human health are new focuses of microbiome research (Proctor, 2019), and understanding of the microbiome is intended to pave the way for future microbiological therapies (Douillard & De Vos, 2019). An increasing number of studies have reported that probiotics in the intestinal tract can maintain intestinal homeostasis by regulating glucose and lipid metabolism, inhibiting the inflammatory response, and improving metabolic disorders (Lau et al., 2019; Plaza-Díaz et al., 2017; Sanders et al., 2019), thus preventing MetS and related complications.

In this study, we focused on a family of Firmicutes named Christensenellaceae, where the type strain Christensenellaceae minuta was first isolated from the feces of a healthy Japanese man in 2012 and named after the Danish microbiologist Henrik Christensen (Morotomi, Nagai & Watanabe, 2012; Waters & Ley, 2019). Since this family of bacteria was recently isolated, little is known about its biological function other than its association with the host and with other microorganisms. Goodrich et al. (2014) found that Christensenellaceae accounted for 0.01% of human feces from the UK Twins population. In data from China (Kong et al., 2016; Wang et al., 2015), South Korea (Kim et al., 2019) and Italy (Biagi et al., 2016), the bacteria were found to be highly abundant in centenarians. Brooks et al. reported that American females have a higher abundance of Christensenellaceae than American males (Brooks et al., 2018). In addition, a study conducted in Amsterdam showed that the relative abundance of Christensenellaceae is ethnicity specific (Brooks et al., 2018; Deschasaux et al., 2018). These studies demonstrated that age, gender, and ethnicity are associated with the abundance of Christensenellaceae. The relative abundance of Christensenellaceae in the intestine was found to be inversely proportional to body mass index (BMI) (Fu et al., 2015; Goodrich et al., 2014; Oki et al., 2016; Peters et al., 2018b) and exhibited a negative correlation with obesity and inflammatory bowel disease (Braun et al., 2019; Gevers et al., 2014; Imhann et al., 2018).

We previously performed the Guangdong Gut Microbiome Project (GGMP) study, constructing the largest intestinal microbiota database for Eastern countries to date. In the present study, we selected the Christensenellaceae-related population from GGMP to find out the relationship between Christensenellaceae and regional distribution, metabolic index and metabolic diseases. We also focused on the connection between sequential operational taxonomic unit (sub-OTU) (Amir et al., 2017) members of Christensenellaceae and the metabolic index. We also utilized the GGMP dataset to reveal the metabolic pathway signatures of Christensenellaceae. A better understanding of the relationship between MetS and Christensenellaceae may lead to new therapeutic approaches for MetS.

Materials & Methods

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.

Bacterial community structure analysis and sequential operational taxonomic units of Christensenellaceae.

Figure 1: Bacterial community structure analysis and sequential operational taxonomic units of Christensenellaceae.

(A) 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. (B) Visualization of taxa meeting a LDA threshold > 3. LEfSe cladogram showed the differential abundant taxa between the two cohorts. Subjects without Christensenellaceae in orange; chris-rich subjects in blue. (C) Sub-OTUs at the level of Christensenellaceae were arranged according to the read number.

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.

The geographical features of Christensenellaceae in different regions in GGMP.

Figure 2: The geographical features of Christensenellaceae in different regions in GGMP.

(A) Histogram based on the average abundance of Christensenellaceae in 14 regions of GGMP. City abbreviation is on the X-axis and the abundance value in on the Y-axis. (B) Annual income and Christensenellaceae abundance in 14 regions of GGMP. (C) The prevalence of metabolic syndrome and the average abundance of Christensenellaceae in each region. The blue line represents metabolic syndrome prevalence and the orange line represents average abundance of Christensenellaceae. (D) Prevalence of metabolic syndrome in people with different Christensenellaceae abundance in each area. Light orange represents G1 (n = 3316) and light blue represents G2 (n = 1465).

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).

Table 1:
Anthropometric parameters and biochemical criterions variations stratified by abundance of Christensenellaceae.
G1-value (n = 3316) G2-value (n = 1465) P-value
Median[SD] 53.0[14.5] 55.0[15.2] <0.001
Female 1830(55.2%) 824(56.2%) 0.517
Male 1486(44.8%) 641(43.8%)
Median[SD] 23.4[3.61] 22.5[3.23] <0.001
Missing 39(1.2%) 27(1.8%)
Median[SD] 80.4[10.1] 78.3[9.22] <0.001
Missing 39(1.2%) 27(1.8%)
Median[SD] 0.512[0.0651] 0.497[0.0603] <0.001
Missing 39(1.2%) 27(1.8%)
Yes 394(11.9%) 92(6.3%) <0.001
No 2883(86.9%) 1346(91.9%)
Missing 39(1.2%) 27(1.8%)
Yes 739(22.3%) 227(15.5%) <0.001
No 2529(86.4%) 1209(82.5%)
Missing 38(1.1%) 29(2.0%)
Yes 414(12.5%) 99(6.8%) <0.001
No 2864(86.4%) 1339(91.4%)
Missing 38(1.1%) 27(1.8%)
Median[SD] 330[94.5] 315[87.3] <0.001
Missing 38(1.1%) 27(1.8%)
Median[SD] 1.15[1.56] 0.9555[1.36] <0.001
Missing 38(1.1%) 27(1.8%)
Median[SD] 16.0[16.5] 14.0[13.7] <0.001
Missing 107(3.2%) 52(3.5%)
Median[SD] 143[21.9] 141[20.3] <0.001
Missing 38(1.1%) 27(1.8%)
Median[SD] 1.22[0.520] 1.25[0.435] 0.0155
Missing 40(1.2%) 27(1.8%)
DOI: 10.7717/peerj.9591/table-1



standard deviation


indicates body mass index


waist-to-height ratio


uric acid




alanine transaminase




high density lipoprotein

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.

Specific microbial taxa associated with Christensenellaceae

We used co-occurrence network analysis to investigate the interactions among the microbes in the complex intestinal microbiota. After pairwise correlation analysis of the bacteria in all the volunteers’ stool samples, the major bacteria associated with Christensenellaceae in the individuals enrolled in the study are shown in Fig. 4A. Of all the species, the abundances of Veillonella, Ruminococcus, Fusobacterium, and Blautia were most highly negatively related to Christensenellaceae. Additionally, species such as Ralstonia and Klebsiella had a negative correlation with the abundance of Christensenellaceae. In addition, we identified a significant positive association of Christensenellaceae with Oscillospira, Ruminococcaceae, RF39, Rikenellaceae and Akkermansia. Moreover, Lachnospiraceae, Roseburia and Sediminibacterium also showed expected relationship with Christensenellaceae.

The focus was not only on the family level of the microbiota but also on the sub-OTUs of Christensenellaceae. The correlation between sequential taxonomic units of Christensenellaceae and other bacteria was universal. As shown in Fig. 4B, the eight sequences with read counts in excess of a thousand were consistent with the relationships between other bacteria. The four sequences sub-OTU14698, sub-OTU3228, sub-OTU12860, and sub-OTU5219 showed the strongest correlation, and the predominant bacteria in the positive correlation included Oscillospira, Clostridiales, RF39 and Rikenellaceae. Nevertheless, Veillonella, Fusobacterium, Blautia, Megamonas and Streptococcus were negatively associated with these five sub-OTUs. In general, the bacterial network relationship at the sequence level and the family level is basically the same.

Metabolism related index and disease status related to Christensenellaceae.

Figure 3: Metabolism related index and disease status related to Christensenellaceae.

(A) Heatmap of the association between metabolic index and Christensenellaceae as well as its sOTU. Color depth indicates the pearson correlation coefficients. ***P < 0.001, **P < 0.01, *P < 0.05. (B) The network showed the significant association between Christensenellaceae sub-OTUs and metabolic diseases. The blue arrow represents the negative correlation, and the orange represents positive. The number of significant association was summarized on the edge of network. Mets, metabolic syndrome; FLD, fatty liver disease; HTN, hypertension; USD, urinary system diseases; T2DM, type 2 diabetes.
Microbial taxa associated with Christensenellaceae.

Figure 4: Microbial taxa associated with Christensenellaceae.

(A) Co-occurrence network between Christensenellaceae and the relative microbiota. The orange line indicates positive correlation and blue negative correlation. ***P < 0.001, **P < 0.01, *P < 0.05. (B) Matrix Diagram of sub-OTUs within Christensenellaceae and related microbiota. Microbiota that are significantly associated with Christensenellaceae were plotted. The size of the plots represents the strength of the correlation.

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.

Differential pathways in individuals with different abundance of Christensenellaceae.

Figure 5: Differential pathways in individuals with different abundance of Christensenellaceae.

G1, LDA score < −3, orange color; G2, LDA > 3, blue color.


This study was based on a rigorous experimental design and strict quality control, and fecal samples and host information of 6,896 volunteers were collected in Guangdong Province. The use of 16S rRNA sequencing helped us clearly elucidate the complexity of the gut microbial ecosystem. We could focus not only on the composition and distribution of intestinal bacteria but also on the metabolic functions of these bacteria. Through the integration of these data, we could focus on the associated OTUs classified within the Christensenellaceae family. And we can target these OTUs as a keystone and link the etiology and pathogenesis of MetS to intestinal microorganisms to provide potential therapeutic targets.

First, the present study showed a variation in Christensenellaceae among residents in different areas by comparing the average abundance for each region. Based on our previous research, MetS prevalence was significantly higher in individuals with higher economic status than in those of with lower economic status (He et al., 2018a). As we found that Christensenellaceae is negatively correlated with MetS, it can be explained to some extent that people in economically developed regions have lower intestinal abundance of Christensenellaceae. Moreover, the individual differences in Christensenellaceae may simply be a result of environmental or genetic factors (Waters & Ley, 2019). And our results for large study population showed that the negative correlation between MetS and Christensenellaceae is consistent in most areas.

As previously reported, the abundance of Christensenella, Bifidobacterium and Akkermansia has been recognized as a signature of the gut ecosystem for healthy aging and longevity (Biagi et al., 2016; Derrien, Belzer & De Vos, 2017; Wang et al., 2015). We also found that people with a high abundance of Christensenellaceae were older than those in the control group. Changing functions of the intestine with age may affect the abundance of bacteria.

The microbiota inhabiting the intestinal cavity affects body health by altering the metabolome and regulating the bacterial bioavailability of nutrients in the lumen. (Johnson et al., 2015; Liu et al., 2018; Romano et al., 2015). In this study, we observed a significant association between the gut microbiota and the variation in BMI and blood lipid levels, which is independent of age. We observed that the abundance of Christensenellaceae was significantly related to the individual variance in BMI and to the blood levels of TG and HDL but has not much correlation with low-density lipoprotein (LDL) or total cholesterol (TC) levels. The analysis showed that people with higher Christensenellaceae abundance in their gut had lower BMI and lower TG levels. These results are consistent with the previous report demonstrating that Christensenellaceae abundance was negatively correlated with BMI and triglycerides and positively correlated with HDL levels in the Dutch LifeLines DEEP cohort (n = 893) and reports from other countries (Fu et al., 2015; Peters et al., 2018a; Waters & Ley, 2019). As low HDL levels are one of the criteria of metabolic syndrome (Yang & Wang, 2019). The positive relationship between Christensenellaceae and HDL suggests a potentially beneficial role in metabolism. Therefore, in the complex network of indicators associated with Christensenellaceae, these findings show the negative correlation between Christensenellaceae with the prevalence of MetS.

The microbial interaction network has been considered an important biological factor in the occurrence and progression of metabolic system diseases. We observed significant microbial community changes in people with high abundance of Christensenellaceae, in whom the richness and diversity of the microbial community increased significantly. Our results highlight the positive correlation between Christensenellaceae and Oscillospira, Ruminococcaceae, RF39, Rikenellaceae and Akkermansia (Goodrich et al., 2014), while the bacteria were negatively related to Veillonella, Fusobacterium and Klebsiella. Previous studies have shown that Oscillospira and Ruminococcaceae are positively associated with health and leanness (Konikoff & Gophna, 2016; Zietak et al., 2016). Other noteworthy taxa include Akkermansia, RF39 and Rikenellaceae, which have been experimentally confirmed to be probiotics and are considered beneficial to limit high fat-induced body weight gain. (Alard et al., 2016; Wang et al., 2017). Moreover, previous reports have shown that Fusobacterium and Klebsiella are positively correlated with the levels of cholesterol and LDL and alter lipid metabolism (Fei et al., 2020; Koren et al., 2011). How Christensenellaceae members impact the diversity and structure of the intestinal microbiota is still unclear. Nevertheless, it is plausible that changes in the intestinal niche induced by Christensenellaceae can promote lipid metabolism and contribute to the maintenance of normal body weight.

Our work showed differences in the predicted microbiota function in people with different abundances of Christensenellaceae. Previous reports indicated that obesity markers were typically positively associated with the KEGG categories of fructose metabolism (Hannou et al., 2018) and methane metabolism (Mathur et al., 2013), which are enriched in people who lack Christensenellaceae. However, it’s not a general consistency with the previous result that Methanobacteriaceae increased in G2 group. PICRUSt represents the methane metabolism of the whole gut flora, but the single increased abundance of Methanobacteriaceae does not indicate the overall function. The high abundance of Christensenellaceae in individuals showed decreased fatty acid biosynthesis. Increasing evidence has shown that fatty acid accumulation is significantly associated with metabolic diseases and obesity (Sonnenburg & Bäckhed, 2016). This confirmed the previously observed negative correlation between Christensenellaceae and metabolic indicators. To develop a deeper understanding of MetS progression, further follow-up studies are required to examine the significance of microbial functional variations in body metabolism. Furthermore, metatranscriptomic and metabolomic analyses could be used to elucidate the detailed pathways of gene and metabolite interactions, enhancing the understanding of the effects of intestinal bacteria.


In the present study, we used multivariate association analyses to explore the association between Christensenellaceae and metabolic indexes in nearly 5,000 participants in South China. Based on the results, Christensenellaceae is related to a low risk of MetS and obesity, showing similarities to results obtained in Western populations. In addition, significant correlations were found between Christensenellaceae and other intestinal bacteria through analysis of the bacterial network. The bacteria that were positively correlated with Christensenellaceae were mainly beneficial bacteria that had been identified previously, while some pathogenic bacteria were negatively correlated with Christensenellaceae. Finally, PICRUSt was applied to analyze the pathway differences caused by Christensenellaceae, which provided bioinformatics evidence for predicting the correlation between Christensenellaceae and metabolic alterations in the gut microbiota community. However, the underlying mechanisms through which Christensenellaceae members regulate host metabolism need to be explored.

Supplemental Information

Alpha-diversity between two groups and evolutionary tree of sOTU

(A-D) Alpha-diversity index of the gut microbiota in two groups. Chao1, observed_OTUs, PD_whole tree and Shannon indices were compared between G1 (n=3316) and G2 (n=1465), p < 0.001. (E) Evolutionary relationships between 8 sequential taxonomic units with reading number more than 1000 and several reference genomes of 16S rRNA V4.

DOI: 10.7717/peerj.9591/supp-1

Metabolic related parameters in two groups with different abundance of Christensenellaceae.

*** p < 0.001, ** p < 0.01, * p < 0.05

DOI: 10.7717/peerj.9591/supp-2

Principal Components Analysis (PCA) comparing the KEGG orthology profiles of two groups

DOI: 10.7717/peerj.9591/supp-3

Metabolic pathways between Q1 and Q4

DOI: 10.7717/peerj.9591/supp-4

Metadata of the this manuscript

DOI: 10.7717/peerj.9591/supp-5

16S rRNA V4 Sequences of sub-OTUs within Christensenellaceae

DOI: 10.7717/peerj.9591/supp-6
12 Citations   Views   Downloads