Diversity analysis of gut microbiota in osteoporosis and osteopenia patients

Subject recruitment and bone mineral density detection

Participants in this study were recruited from Hong Hui Hospital, Xi’an Jiaotong University, Xi’an, China. Dual X-ray absorptiometry (DXA) was performed to detect the bone mineral density of lumbar vertebrae of subjects. We further excluded all patients with any malignancy, chronic liver disease, heart disease, kidney disease, or diabetes. Finally, a total of 18 subjects including six with primary osteoporosis (OP), six with primary osteopenia (ON), and six normal controls (NC; as determined by physical examination) were selected for further research (Table 1). None of the 18 participants ingested yogurt, prebiotics, or probiotics during the fecal collection period, nor had they used medication (e.g., antibiotics) within one month of sample collection. The study was approved by Hong Hui Hospital, Xi’an Jiaotong University, Biomedical research ethics committee. Each participant provided his or her written informed consent.

Fecal sample collection and DNA extraction

Fresh stool samples were collected in sterile boxes, then frozen and stored at −80 °C for further use. The microbial genome was extracted using QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Sample DNA purity and concentration were tested using a Nanodrop 2000 Spectrophotometer.

16S rRNA PCR and Illumina sequencing

We amplified the bacterial 16S ribosomal RNA gene V3-V4 region using the TransGen AP221-02 Kit (TransGen, Beijing, China). The following PCR primers were used: 338F 5′-ACTCCTACGGGAGGCAGCAG-3′ and 806R 5′-GGACTACHVGGGTWTCTAAT-3′. The reaction volume (20 μl) comprised 5 × FastPfu Buffer (4 μl), 2.5 mM dNTPs (2 μl), forward primer (0.8 μl), 5 μM reverse primer (0.8 μl), FastPfu Polymerase (0.4 μl), and template DNA (10 ng). Cycling proceeded as follows: 3 min at 95 °C twenty-seven cycles(30 sat 95 °C, 30 sat 55 °C, 45 sat 72 °C); 10 min at 72 °C. After amplicons extraction, samples were purified and quantified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, CA, USA) and QuantiFluor™-ST (Promega, Madison, WI, USA), respectively. Purified amplicons were pooled in equimolar proportions and paired-end sequenced (2 × 250 bp) on the Illumina MiSeq platform with TruSeqTM DNA Sample Prep Kit (Illumina, San Diego, CA, USA).

16S rRNA gene sequencing analysis

Raw fastq files were demultiplexed, quality-filtered by Trimmomatic and merged by FLASH with the following criteria: (i) The reads were truncated at any site receiving an average quality score <20 over a 50 bp sliding window; (ii) Primers were exactly matched allowing 2 nucleotide mismatching, and reads containing ambiguous bases were removed; (iii) Sequences whose overlap longer than 10 bp were merged according to their overlap sequence. Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff (Edgar, 2013) using UPARSE (version 7.1; http://drive5.com/uparse/) and chimeric sequences were identified and removed using UCHIME. The taxonomy of each 16S rRNA gene sequence was assigned by QIIME (version 1.7; http://qiime.org/home_static/dataFiles.html) (Caporaso et al., 2010) using RDP Classifier algorithm (http://rdp.cme.msu.edu/) (Wang et al., 2007) against the Silva (SSU123) 16S rRNA database (Quast et al., 2013) using a confidence threshold of 70%. Alpha diversity at the OTU level (e.g., Ace, Chao, Shannon and Simpson index) were calculated in QIIME following previously described methods (Jiang & Takacs-Vesbach, 2017; Lauber et al., 2009; Van Horn et al., 2016).

Statistical analysis

Results analysis and figure generation based on clinicopathological information, alpha estimators and relative bacterial abundance were performed using SPSS 21.0 and GraphPad Prism 5.01 software. Student’s t-test and the Mann–Whitney U-test were performed, with p < 0.05 indicating a significant difference between groups. Rarefaction curves were generated based on the alpha diversity estimators. The unweighted UniFrac algorithm was applied for hierarchical clustering and principal coordinates analysis at the OTU level to analyze beta diversity. We applied “Vennerable” package in R software (version 3.3.3) for the generation of venn diagram based at the OTU level. The Circos software (http://circos.ca/software/download/circos/) was performed for the generation of collonearity diagram to visualize the corresponding abundance relationship between samples and bacterial communities at the phylum and genus levels. The enriched and significant bacteria in each group were identified by linear discriminant analysis (LDA) combined with effect-size measurements (LEfSe) (Segata et al., 2011), with p < 0.05. For the Kruskal–Wallis test, LDA values >2 were considered significant (Szafranski et al., 2015).

Illumina sequencing data characteristics

The clinicopathological information for each of the three groups included in the study is presented in Table 1. There were no significant differences in terms of age or gender, while BMD, T-score and Z-score differed significantly among groups. Illumina sequencing captured a total of 694,232 high-quality sequences, with an average of 38,568 sequences/sample. Detailed information on the sequence results obtained for each sample are presented in Table S1.

Inverse correlation between the number of bacterial taxa and the value of BMD

Based on the sequencing data, the gut microbiota of all samples were classified to 507 OTUs, 367 species, 235 genera, 99 families, 63 orders, 38 classes, 25 phyla. The number of bacterial taxa tended to increase at each level in accordance with the reduction in BMD, as shown in Table 2 and Fig. S1. Figure 1 presents a Venn diagram for the OP, ON and NC groups (at the OTU level). There were 455, 378, and 282 OTUs present in the OP, ON, and NC group, respectively. In addition, 208 OTUs (41%) were shared by all samples; 154 OTUs (30.4%) were shared between the OP and ON groups. For the remaining components (28.6%), the OP group (13.6%) accounted for nearly half of all OTUs.

Diversity analysis of gut microbiota in osteoporosis and osteopenia patients

To determine alpha diversity, we calculated the mean ace index, chao index, shannon index, and simpson reciprocal index. This process allowed us to fully characterize the bacterial community diversity in samples. Detailed information on the estimators in each sample is presented in Table S2. The OTU level rarefaction curves of diversity estimators reached plateau phase (Fig. S2), indicating that most bacterial species had been captured by sequencing in all samples. Higher numbers of the estimators represent greater diversity, which suggests that alpha diversity index was inversely correlated with BMD, although there were no significant differences between the OP and ON groups, as shown in Fig. 2.

With regard to beta diversity, unweighted UniFrac analysis indicated that hierarchical clustering and principal coordinate analysis (PCoA) could discriminate the NC samples from OP as well as ON samples. However, there was substantial overlap between the OP and ON groups, and most ON samples were positioned in the middle of the OP and NC samples, as Fig. 3 illustrates.

Significance analysis of gut bacterial community abundance in osteoporosis and osteopenia patients

At the phylum level illustrated in Fig. 4, Firmicutes, Bacteroidetes, Proteobacteria and Actinobacteria constituted the four dominant phyla in all samples. The average ratios of Firmicutes/Bacteroidetes were 3.326, 1.755 and 1.290 in the OP, ON, and NC groups, respectively. Furthermore, we calculated the significance of the 10 most dominant phyla of microbial community structure among the OP, ON, and NC groups. Differences among the four dominant phyla (Firmicutes, Bacteroidetes, Proteobacteria, and Actinobacteria) were not statistically significant for comparisons between the OP and ON group or the ON and NC group (p > 0.05). Proportion of Firmicutes was significantly higher and Bacteroidetes proportion was significantly lower in OP samples than that in the NC group (p < 0.05) (Fig. 4B). As for other bacterial communities with small proportions, most of them were rare in the NC group but increased in the OP and ON groups. Gemmatimonadetes and Chloroflexi were significantly different between the OP and NC groups (p < 0.01) as well as between the ON and NC groups (p < 0.01).

At the genus level, a total of 21 genera with proportions above 1% were detected, as visualized in Fig. 5. Bacteroides accounted for the largest proportion in all samples. In the NC group, three genera (Bacteroides, Faecalibacterium and Prevotella) contributed more than half of the bacterial community. In the ON and OP groups, five and 11 genera, respectively, accounted for 50% of the bacterial community. Differentiation analysis of the 21 genera is presented in Fig. 5B. The Blautia, Parabacteroides and Ruminococcaceae genera differed significantly between the OP and NC groups. Fig. S3 depicts the collinearity diagram for the bacterial community and samples from all three groups.

We further applied linear discriminant analysis (LDA) combined effect size measurements (LEfSe) to explore the significant changes and relative richness of the bacterial community in the OP, ON, and NC groups, at phylum and genus levels. Fig. 6 summarizes the enrichment and variations in bacterial community for all three groups. At the phylum level, one phylum and seven phylum communities were enriched in the ON and OP group, respectively, while no community in the NC group was enriched. At the genus level, 35 genus, five genus and two genus communities were enriched in the OP, ON and NC groups, respectively. The significance and variance of bacterial communities, as determined by sequencing analysis, may help discriminate OP or ON patients from NC subjects.