559 Discussion of a GQD tre ent is also onent of GOD.However.no gu red this p in T2D with the dose-dependent manner of of was also rep the adults nin ealthy ucces incre 出 losedepenieatmdlhationafthetmierahiote he improve T2D in humans. Moreover, the gut microbiota glucose home occurre ore sign T2D the gut GQL et nigh 6 6 4 MD HD Week o Week 12 ii am 28.p<005,p<00 are ropresented as means..m nan's and log- The color of the nan's co ween the OTU GThe family and gonus of th The ISME Journalabundance of this bacterium negatively correlated with HbA1c, FBG and 2h-PBG, and positively correlated with HOMA-b by Spearman’s correlation coefficient (Supplementary Table 7). Discussion To our knowledge, this study is the first registered clinical trial to evaluate the dose-dependent efficacy and safety of a Chinese herbal decoction following a randomized, double-blinded and placebo-controlled design. In our study, GQD treatment provided clinically meaningful and dose-dependent reduc￾tions in FBG and HbA1c compared with placebo. HOMA-b was also significantly improved by HD GQD treatment. Compared with those in T2D patients treated with berberine in a previously published clinical trial, the HD group in our study showed similar reductions of FBG and HbA1c; however, we did not observe an improvement of dyslipidemia, which had been significantly amelio￾rated by berberine in the other study (Zhang et al., 2008). In addition, we found a slight alleviation of inflammation, indicated by the reduction of plasma orosomucoid in the HD group. This finding was consistent with a previous nuclear magnetic reso￾nance-based plasma metabonomic study in T2D rats treated with GQD (Tian et al., 2013). These data show that the Chinese herbal formula is effective for glycemic control, making it a promising candidate for diabetes management. Concomitant with the improved glucose home￾ostasis, we observed an altered microbial composi￾tion induced by GQD. In other studies, diet (Cani et al., 2008; Turnbaugh et al., 2008; Xiao et al., 2013), drugs (Zhang et al., 2012b; Shin et al., 2013), prebiotics (Cani et al., 2007b, 2009) and probiotics (Park et al., 2013) have been used to modulate the gut microbiota in metabolic diseases, for example, the increase of Akkermansia spp. by metformin in the improvement of glucose homeostasis in high-fat￾diet-induced obese mice (Shin et al., 2013). Berber￾ine, the major pharmacologic component of a Chinese herb C. chinensis (Huang-Lian), has been shown to enrich short-chain fatty acid producers in parallel with the prevention of obesity and insulin resistance in rats (Zhang et al., 2012b). C. chinensis is also a major component of GQD. However, no gut microbiota modulation by berberine or Chinese herbs has been reported in humans. To our knowl￾edge, this study is the first direct evidence in humans that TCMs can modulate the structure of the gut microbiota. Consistent with the dose-dependent manner of T2D amelioration, GQD also exerted a dose-depen￾dent modulation on the gut microbiota, suggesting a strong association between the modulation of gut microbiota and T2D alleviation. Dose-dependent modulation of gut microbiota was also reported in prebiotic studies. For example, an stepwise increase of Bifidobacteria was observed by q-PCR when nine healthy adults received successively increased dosage levels of galactooligosaccharides (Davis et al., 2010). Our study is the first example that dose-dependent modulation of the gut microbiota was concomitant with a dose-dependent alleviation of T2D in humans. Moreover, the gut microbiota alterations occurred before significant improvement of T2D symptoms, supporting the hypothesis that changes in the gut microbiota induced by GQD might contribute to improved glucose homeostasis Figure 6 Relative abundance of F. prausnitzii as quantified by q-PCR. (a) The impact of different treatments on the relative abundance of F. prausnitzii before and after treatment. The sample number (n) for each group: placebo ¼ 21, LD ¼ 19, MD ¼ 19 and HD ¼ 20. (b) The differences of F. prausnitzii among four groups at week 12. The sample number (n) for each group at week 0: placebo ¼ 31, LD ¼ 23, MD ¼ 26 and HD ¼ 29. The sample number for each group at week 12: placebo ¼ 26, LD ¼ 25, MD ¼ 24 and HD ¼ 28. D Po0.05, DDPo0.05 vs before treatment using Wilcoxon’s signed rank test. *Po0.05 and **Po0.01. Placebo, LD, or MD vs HD using Mann–Whitney test. Data are represented as mean±s.e.m. Figure 5 Heat map of redundancy analysis (RDA)-identified key OTUs responding to GQD treatment and Spearman’s correlation between identified OTUs and HbA1c or FBG. The color of the spots in the left panel represents the mean relative abundance (normalized and log-transformed) of the OTU in each group. The OTUs are organized according to their phylogenetic positions. The color of the spots in the right panel represents R-value of Spearman’s correlation between the OTU and HbA1c or FBG. The family and genus names of the OTUs are shown on the right. þ Po0.05. Microbiota shift in alleviation of type 2 diabetes J Xu et al 559 The ISME Journal