Human participants
Ethical approvalwas granted by the Ethics Committee of Zhongshan Hospital of Xiamen University (No. 201808) before recruitment. Postmenopausal women were recruited from Xiamen, Fujian Province, China, between December 1, 2018, and February 1, 2019. Postmenopausal status was defined as at least one year since the last menstruation. Written informed consent was obtained from all the participants. We adopted the following exclusion criteria before fecal sample collection: (1) use of antibiotics or hormones within 3 months; (2) consumption of pro/pre/synbiotic products within two weeks; (3) use of medications (e.g., calcium, VD, calcitriol, alpha calcitriol, estrogen, glucocorticoids, diphosphonate, denosumab or teriparatide); and (4) presence of bone disease, hyperthyroidism, hypothyroidism, gastrointestinal disease, cancer, kidney disease, or mental illness or evidence of recent infections.
Determination of clinical parameters
The basic information of all the subjects, such as age, height and weight, was collected, and the body mass index (BMI) was calculated according to the data of the latter two items. Blood samples were collected from all the subjects in a fasting state and at similar time points in the morning. The levels of serum 25(OH)D, estradiol (E2), osteocalcin (OC), C-terminal telopeptide of type I collagen (CTX-I), procollagen type 1 n-terminal propeptide (P1NP), and parathyroid hormone (PTH) were measured with an automated Roche Osteoporosis Int electrochemiluminescence system (Roche Diagnostics GmbH, Germany). The inter- and intra-assay coefficients of variation (CVs) were 8.0% and 5.6% for 25(OH)D, 2.9% and 2.3% for E2, 4.0% and 2.9% for OC, 3.5% and 2.5% for CTX-1, 2.8% and 2.3% for P1NP, and 2.9% and 1.7% for PTH, respectively. The bone mineral densities (BMDs) of the lumbar spine (LS) (L1-4) and total hip joint (femoral neck (FN), trochanter, and intertrochanteric region) were measured with a daily calibrated Hologic 4500 A dual-energy X-ray absorptiometry scanner (Lunar Expert 1313, Lunar Corp, USA).
Sequencing and bioinformatics
Fecal samples were collected in sterile plastic cups, frozen, and stored at -80°C within 1 hour until further processing[17]. Fecal microbial DNA was extracted using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany). PCR amplification was carried out using an ABI 2720 Thermal Cycler (Thermo Fisher Scientific, USA). We used Multiskan™ GO spectrophotometry (Thermo Fisher Scientific, USA) to quantify bacterial genomic DNA as the template for amplification of the V3-V4 hypervariable region of the 16S rRNA gene in three replicate reactions with forward (Illumina adapter sequence 5’-CCTACGGGNBGCASCAG-3’) and reverse (Illumina adapter sequence 5’-GGACTACNVGGGTWTCTAAT-3’) primers. Replicate PCR products were pooled and purified with Agencourt AMPure XP magnetic beads (Beckman Coulter, USA). A TopTaq DNA Polymerase kit (Transgen, China) was used. The purity and concentration of sample DNA were assessed using a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, USA). Paired-end sequencing was performed by Treatgut Biotech Co., Ltd. with a HiSeq 2500 (Illumina, San Diego, CA, USA) with PE 250 bp reagents.
After sequencing, raw paired-end reads were assembled using FLASH[18] with the default parameters. Primers were removed using cutadapt, and clean tags were obtained by removing the lower reads using cutadapt[19]. To assign de novo operational taxonomic units (OTUs), we removed chimeric sequences and clustered sequences with 97% similarity and used Usearch (V10.0.240)[20] for the study. The representative sequences of OTUs were aligned to the SILVA132 database for taxonomic classification by RDP Classifier[21] and aggregated to various taxonomic levels.
Fecal metabolite extraction
At least fifty milligrams of sample was placed in an EP tube, and then 1000 µL of extraction liquid containing an internal target (V methanol:V acetonitrile:V water = 2:2:1, which was kept at -20°C before extraction) was added. The samples were homogenized in a bead mill for 4 minutes at 45 Hz and ultrasonicated for 5 minutes (incubated in ice water). After homogenization 3 times, the samples were incubated for 1 hour at -20°C to precipitate proteins. The samples were centrifuged at 12,000 rpm for 15 minutes at 4°C. The supernatant (750 µL) was transferred to fresh EP tubes, and the extracts were dried in a vacuum concentrator without heating. Then, 100 µL of extraction liquid (V acetonitrile:V water = 1:1) was added for reconstitution. The samples were vortexed for 30 s, sonicated for 10 minutes (4°C water bath), and centrifuged for 15 minutes at 12,000 rpm and 4°C. The supernatant (60 µL) was transferred to a fresh 2 mL LC/MS glass vial, and 10 µL was collected from each sample and pooled as quality control (QC) samples. Sixty microliters of supernatant was used for ultra-high-performance liquid chromatography combined with quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) analysis.
LC–MS/MS analysis and annotation
LC–MS/MS analyses were performed using a UHPLC system (1290, Agilent Technologies) with a UPLC BEH Amide column (1.7 µm 2.1×100 mm, Waters) coupled to a TripleTOF 6600 (Q-TOF, AB Sciex) & QTOF 6550 (Agilent). The mobile phase consisted of 25 mM NH4OAc and 25 mM NH4OH in water (pH = 9.75) (A) and acetonitrile (B), which was applied in an elution gradient as follows: 0 minutes, 95% B; 7 minutes, 65% B; 9 minutes, 40% B; 9.1 minutes, 95% B; and 12 minutes, 95% B, which was delivered at 0.5 mL/min. The injection volume was 2 µL. A TripleTOF mass spectrometer was used due to its ability to acquire MS/MS spectra on an information-dependent basis (IDA) during an LC/MS experiment. In this mode, the acquisition software (Analyst TF 1.7, AB Sciex) continuously evaluates the full-scan survey MS data as it collects and triggers the acquisition of MS/MS spectra depending on preselected criteria. In each cycle, 12 precursor ions with intensities greater than 100 were chosen for fragmentation at a collision energy of 30 V (15 MS/MS events with a product ion accumulation time of 50 msec each). The ESI source conditions were set as follows: ion source gas 1 at 60 psi, ion source gas 2 at 60 psi, curtain gas at 35 psi, source temperature at 650°C, and ion spray voltage floating at 5000 V or -4000 V in positive or negative modes, respectively.
MS raw data files were converted to mzXML format using ProteoWizard[22] and processed with the R package XCMS (version 3.2). The preprocessing results generated a data matrix that consisted of the retention time (RT), mass-to-charge ratio (m/z) values, and peak intensity. The R package CAMERA was used for peak annotation after XCMS data processing[23].
Statistical analyses and visualization
The rarefaction curves constructed from the sequenced data has been basically stable, indicating that the sequenced data has benn basically stable at this sequencing depth (Supplementary Fig. 1). The alpha diversity indexes, bacterial richness (observed OTUs), Shannon, Simpson, ACE, Chao1 index and evenness (J) were calculated based on OTU tables of the study. Significance tests between the HVD and LVD groups were conducted with the Wilcoxon test method. Differences in community structure across samples (beta diversity) were visualized by principal coordinates analysis (PCoA) plots based on Bray–Curtis distance. Significance tests were determined using permutational multivariate analysis of variance (PERMANOVA) with 999 permutations in vegan[24]. Linear discriminant analysis effect size (LEfSe)[25] was performed to identify taxa with differential abundance between the HVD and LVD groups. We further explored the correlation between different genera and fecal metabolites by Spearman’s rank test. To evaluate functional differences in the gut microbiomes of the HVD versus LVD groups, we performed PICRUSt[26] to calculate the microbial abundances, assign metabolic pathways to the gut microbiomes using KEGG and COG, and then test the differences between the two groups. All statistical and correlational analyses were conducted in R (v3.6.0)[27]. Figures were plotted mainly using ggplot2 (v3.0.0)[28].