The study protocol was approved by the Ethics Committee of Zhejiang Chinese Medical University, and written informed consent was obtained from each participant. A total of 26 male patients with diagnosed gout were recruited from three hospitals (Zhejiang Provincial Hospital of TCM, The Second Affiliated Hospital of Zhejiang Chinese Medical University, and Zhejiang Province People Hospital). All patients had suffered from gout for at least 12 months and did not receive any medical treatment within 1 month of study participation. The patients with comorbid disorders were excluded. The clinical diagnosis and blood examination reports of all patients were obtained from the hospitals. Twenty-six male volunteers were recruited by a routine physical examination. The healthy controls had no gastrointestinal tract disorders and did not receive antibiotics within 1 month of this study. In addition, there were no significant differences among the two groups in terms of age, smoking history, alcohol intake or dietary intake. The clinical data such as UA, body mass index (BMI), erythrocyte sedimentation rate (ESR), blood urea nitrogen (BUN) and serum creatinine (Cr) were shown in Table 1.

Total DNA was extracted from thawed fecal samples using the QIAamp^® DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocols. The extracted products were determined by agarose gel electrophoresis (1% w/v agarose). Quantification of the DNA yield was carried out using a NanoDrop2000 spectrophotometer (Thermo Scientific). The DNA was stored at -20°C for Illumina Miseq sequencing analysis.

The V3–V4 region of the 16S rRNA genes was amplified from the diluted DNA extracts with the primers 319f (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806r (5′-GGACTACH VGGGTWTCTAAT-3′). PCR amplification was then performed in a 30 μl mixture containing 0.5 μl of DMSO, 1.0 μl of forward primer (10 mM), 1.0 μl of reverse primer (10 mM), 5.0 μl of DNA sample, 7.5 μl of ddH2O and 15.0 μl of Phusion High-Fidelity PCR Master Mix with HF Buffer (NEB). The reactions were hot-started at 98°C for 30 s, followed by 30 cycles of 98°C for 15 s, 58°C for 15 s, and 72°C for 15 s, with a final extension step at 72°C for 1 min. The PCR products were purified using an agarose gel DNA purification kit (Qiagen, Chatsworth, CA, USA). The amplicon library was prepared using a TruSeq^TM DNA sample preparation kit (Illumina Inc, San Diego, CA, USA). The sequencing reaction was conducted using Illumina MiSeq sequencing (2 × 300 bp; Hangzhou Guhe Information and Technology Co., Ltd., Zhejiang, China).

Thawed fecal samples were suspended with Na⁺/K⁺ phosphate buffer (K₂HPO₄–NaH₂-PO₄, 0.1 M, pH 7.4) containing 100% D₂O and 0.005% sodium 3-trimethylsilyl [2,2,3,3-d4] propionate (TSP). The suspension was subjected to freeze-thaw treatment (3 times) with liquid nitrogen and 20 × ultrasonication cycles (20 s vortex-10 s waiting) followed by centrifugation (12000 rpm, 4°C, 15 min). The supernatants were removed, filtered through 0.2 μm membrane filters, and 550 μL of each filtrate was transferred to 5 mm NMR tubes (Norell ST50-7, USA) for NMR spectroscopic analysis.

High resolution ¹H NMR spectra were recorded on a Bruker Avance III 600 MHz spectrometer (Bruker, Biospin, Germany) using an inverse detection cryogenic probe. The sample temperature was controlled at 298 K. One-dimensional NMR spectra were recorded using a first increment of NOESY pulse sequence (recycle delay –90°–t1–90°–tm–90°–acquisition), and water suppression was achieved with weak continuous-wave irradiation during the recycle delay (2 s) and the mixing time (100 ms). The 90° pulse length was adjusted to approximately 10 μs, and 64 transients were collected into 32 k data points with a spectral width of 20 ppm.

The spectra were transformed with 1 Hz line broadening and zero filling, manually phased, and baseline corrected using the TOPSPIN 2.0 software. Metabolites were identified using information found in the literature (Saric et al., 2007; Le Gall et al., 2011) or on the web (Human Metabolome Database)¹ and by use of the 2D-NMR methods including ¹H–¹H correlation spectroscopy (COSY), ¹H–¹H total correlation spectroscopy (TOCSY), ¹H–¹³C heteronuclear multiple-bond correlation (HMBC), and ¹H–¹³C heteronuclear single quantum correlation (HSQC) spectra.

Paired-end reads of sequencing were first merged and demultiplexed into patient samples using the Quantitative Insights Into Microbial Ecology (QIIME version 1.9; Caporaso et al., 2010). Before assembly, sequence reads were first filtered to remove low-quality or ambiguous reads, including reads lacking exact matching with the primer, reads containing ambiguous character (N), and reads with an average quality score <25. Only two reads with a sequence overlap longer than 20 bp were assembled. The assembled sequence reads with <400 bp or >500 bp were discarded. High-quality sequences were binned into 16S rRNA Operational Taxonomic Units (OTUs) and defined at ≥97% sequence homology. Chimera detection and removal were assessed using the GOLD reference database and uchime (Edgar et al., 2011). The taxonomic affiliation of each OTU was performed with QIIME against the SILVA database (Quast et al., 2013). Alpha diversity (Chao1, observed species, Simpson, Shannon, singles, doubles) was analyzed based on the OTUs table rarefied to 1000 reads, and 20,000 reads were finally extracted from each sample for the other analyses.

To test whether gut microbial species could be differentiated between gout patients and healthy controls, a metric multidimensional scaling method based on projection known as principal coordinates analysis (PCoA) was used. Each sample was mapped based on the overall microbial composition and assessed for similarities. The online software LefSe (Segata et al., 2011) was utilized to select and demonstrate differentially abundant taxonomy based on Kruskal–Wallis test and LDA (linear discriminant analysis) score.

The statistical processing of ¹H NMR data was conducted as previously described (Shao et al., 2016). After manual corrections for phase and baseline distortion (Bruker BioSpin), ¹H NMR spectra were referenced to the TSP signal (δ 0.0) and spectral regions 9.50–0.6 were integrated into regions with an equal width of 0.004 ppm using the AMIX software package (V3.8, Bruker BioSpin). The region containing the water resonance (δ 5.16–4.68) was removed. Each bucket was normalized to the total sum of the spectral integrals, and then the peak areas of identified metabolites were extracted to calculate metabolite percentage concentrations. Total metabolite concentrations were performed with multivariate data analysis using the SIMCA-P⁺ software package (version 11.0, Umetrics, Sweden). In this study, partial least squares discriminant analysis (PLS-DA) was performed to attempt to maximize the separation between classified groups of observations. The validity of the model was assessed with the R²X and Q² values, reflecting the explained variables and the predictability of the model, respectively. The validity of the model was further evaluated with rigorous permutation tests (n = 200). Identified metabolite concentrations were calculated according to ¹H chemical shifts. Metabolite concentrations within groups were compared using Mann–Whitney non-parametric test in SPSS software 16.0.

The correlation network was performed with a Spearman’s rank correlation coefficient on the discrepant taxonomy and metabolites in R program, and only connections with a p-value less than 0.01 were retained. Meanwhile, to correct the results of this, R-p-adjust method in R program were used to calculate FDR values.

Fifty-two male individuals (26 healthy controls and 26 gout patients) were enrolled to study microbial profiles of stool samples. The results of Illumina Miseq sequencing shown each sample had at least 20,000 valid reads for OTU analysis. The intestinal microbiota of patients with gout was significantly different compared to healthy control subjects. The α diversity indices (Chao1, Observed species, Simpson, Shannon, Singles and Doubles) of the intestinal microbiota from the gout group were less than the healthy group (Supplementary Figure S1), suggesting that gout was associated with lower microbial diversity. Additionally, three-dimensional PCoA showed separation between the two groups, indicating that gout was the primary factor influencing the differences (Supplementary Figure S2).

Figure 1 showed the discrepant microbial species with a reduced significance threshold (LDA score >2) between the two groups. The LefSe method revealed that the phylum Bacteroidetes and its derivative (Bacteroidia, Bacteroidales, Bacteroidaceae as well as Bacteroidales S24_7 group and Porphyromonadaceae), the phylum Chloroflexi and its derivatives (Anaerolineae, Anaerolineales, and Anaerolineaceae), the order Corynebacteriales and its derivative (Nocardiaceae and Rhodococcus), the class Erysipelotrichia and its derivatives (Erysipelotrichales, Erysipelotrichaceae and Erysipelatoclostridium), and the class Negativicutes and its derivative (Selenomonadales) were all higher in the intestinal microbiota from the gout patients. Conversely, the family Vibrionaceae and its derivatives (Photobacterium and Vibrio), the genus Coprococcus 3, Lachnospiraceae NC2004 group, Lachnospiraceae UCG_005, Ruminococcaceae NK4A214 group and Ruminococcaceae UCG_011 were all lower in the intestinal microbiota from the gout patients.

Typical ¹H NMR spectra of fecal extracts of gout and healthy controls were shown in Supplementary Figure S3. The spectra of fecal extracts showed 46 targeted metabolites in this study. The detailed information was provided in Supplementary Table S1. To maximize class discrimination and discover the gout-related fecal metabolites, a comparative PLS-DA model was utilized on the targeted metabolite data from the gout patients and healthy controls. The values for R²X and Q² and the results of permutation tests indicated that the two models were of reasonable quality (Figure 2A). Figures 2A,B show the PLS-DA score plots and corresponding loading plots for human fecal extracts, respectively. Clear separations in the PLS-DA score plot based on 46 targeted metabolites were observed between the gout patients and healthy controls (Figure 2A). Differential metabolites were defined based on the variable importance for project values (VIP) value of PLS-DA. Metabolites with a VIP value greater than 1.0 were dispersed from the origin of the loading plot and were considered as the primary contributors for classification of the groups (Figure 2B).

According to the PLS-DA model analysis, 15 targeted metabolites with a VIP above 1.0 were selected and subjected to a significance test with Mann–Whitney non-parametric test. As shown in Table 2, 14 of 15 metabolites were significantly different with p < 0.05. Compared with healthy controls, gout patients had significantly higher concentrations of alanine, glycine, taurine, succinate, acetate, α-glucose, β-glucose and α-xylose, and significantly lower concentrations of valine, asparagine, aspartate, citrulline, phenylalanine and α-ketoisocaproate. The detailed concentrations of the 14 significant metabolites were shown in Supplementary Figure S4.

To further analyze the associations of alterations in fecal metabolome and microbiome, we conducted the correlation of 19 discrepant microbial taxa at the family and genera level and 15 discrepant metabolites using the Spearman’s rank correlation method. As shown in Figure 3, no associations were involved in some microbial taxa (Ruminococcaceae NK4A214 group, Rhodococcus, Porphyromonadaceae sp., Nocardiaceae sp. and Erysipelatoclostridium) and some fecal metabolites (acetate, asparagine). However, the associations were observed among most fecal signatures of metabolome and microbiome.

Some gout-enriched microbial taxa were associated with the gout-enriched metabolites. The positive associations were observed between Bacteroides and succinate, between Anaerolineaceae and alanine or α-xylose. The negative associations were observed between Erysipelotrichaceae sp. and inosine, between Bacteroidales S24_7 group and alanine or succinate or taurine. Meanwhile, some gout-enriched microbial taxa were also associated with the gout-depleted metabolites. The positive associations were observed between Bacteroidales S24_7 group and valine or phenylalanine.

The gout-depleted microbial taxa were negatively associated with the gout-enriched metabolites. For example, Photobacterium was negatively associated with alanine, glycine, taurine, α-glucose and β-glucose. However, the gout-depleted microbial taxa were positively associated with the gout-depleted metabolites. For instance, the positive associations were observed between Coprococcus 3 or Vibrio and citrulline.

Although the correlation between metabolites and microbial taxa doesn’t mean that they exhibit any biological interaction, they have the potential to be a combination of fecal signatures.