Akkermansia muciniphila is Negatively Correlated with Hemoglobin A1c in Refractory Diabetes

Patients with refractory diabetes are defined as type 2 diabetes (T2D) patients; they cannot achieve optimal glycemic control and exhibit persistent elevations of hemoglobin A1c (HbA1c) ≥8% while on appropriate therapy. Hyperglycemia can lead to severe microvascular/macrovascular complications. However, in contrast to T2D, few studies have focused specifically on the gut microbiota in refractory diabetes. To examine this issue, we recruited 79 subjects with T2D and refractory diabetes (RT2D), and all subjects received standard therapy with Metformin or other hypoglycemic agents with or without insulin for at least one year. The α-diversity displayed no significant difference, whereas the β-diversity showed a marginal significance (p = 0.054) between T2D and RT2D. The evaluation of taxonomic indices revealed reductions in both Akkermansia muciniphila and Fusobacterium and a corresponding enrichment of Bacteroides vulgatus, Veillonella denticariosi among those with RT2D. These microbial markers distinguished RT2D from T2D with an acceptable degree of discrimination (area under the curve (AUC) = 0.719, p < 0.01) and were involved in several glucose-related functional pathways. Furthermore, the relative abundance of Akkermansia muciniphila was negatively correlated with HbA1c. Our combined results reveal unique features of the gut microbiota in RT2D and suggest that the evaluation of the gut microbiota could provide insights into the mechanisms underlying glycemic control and the impact of therapeutic modalities in patients with RT2D.


Introduction
According to recent reports from the World Health Organization, 422 million people worldwide carried a diagnosis of type 2 diabetes (T2D) in 2014, and this number is expected to reach~552 million in 2030, which will mean that this condition will rank as the seventh leading cause of death worldwide [1,2]. T2D is a metabolic disease associated with dysregulated glucose metabolism; patients typically present with insulin resistance or reduced insulin secretion, meaning that blood glucose levels are not maintained at appropriate levels. Patients with T2D typically experience both microvascular and macrovascular

Bioinformatics Analysis
The raw paired-end reads were trimmed and passed through quality filters (quality trimming, discarding short read length and removing chimeras) and were assigned to operational taxonomic units (OTUs) which shared ≥97% similarity with the Greengene database. The raw paired-end reads were also analyzed with the basespace Ribosomal Database Project (RDP) classifier. Operational taxonomic units (relative abundance, heatmap, Krona and differential abundance analysis), α-diversity (Shannon index), and β-diversity (PCoA-Unweighted UniFrac) were determined with basespace (Illumina, San Diego, CA, USA), CLC Microbial Genomics Module (Qiagen, Germantown, MD, USA) and Graphpad Prism 7 (GraphPad Software, La Jolla, CA, USA). The OTU table was generated by CLC Microbial Genomics Module to be further analyzed with the linear discriminant analysis effect size (LEfSe) and for Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) analysis. LEfSe was conducted by Galaxy/HutLab to identify specific microbial markers between groups, with an alpha value for the factorial Kruskal-Wallis test/pairwise Wilcoxon test of 0.05 and an LDA score cut-off of 2.0. PICRUSt prediction was conducted by Galaxy/HutLab according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) functional pathways database and analyzed with the Statistical Analysis of Metagenomic Profiles (STAMP) software. The STAMP criteria were set up, removing unclassified reads, with p < 0.01 and an effect size of 1. The results identified functional pathways with a significantly different abundance at level 3 between groups [20].

Statistical Analysis
A comparison of different groups was performed by the two-tailed t-test. Values of p less than p < 0.05, p < 0.01 and p < 0.001 were considered to be statistically significant. The specificity and sensitivity of the microbial markers were determined with the receiver operating characteristic curve (ROC curve) and the area under the curve (AUC) value. Correlations were calculated as the Pearson correlation coefficient [20].

Characteristics of the Study Subjects and Diversity between T2D and RT2D
In this study, we recruited 79 participants who were divided into two groups according to HbA1c. These groups included those patients with T2D who maintained good glycemic control (HbA1c <8%) and those patients with RT2D who were unable to maintain glycemic control, as indicated by HbA1c ≥8%. All study participants received standard oral glucose-lowering drugs (OGLDs) therapy with or without insulin for at least one year, and we collected biochemical indexes and feces for analysis. The biochemical indexes and general characteristics of the individuals in these two groups are summarized in Table 1. As shown, the values for HbA1c and glucose Ante Cibum (glucose AC) were significantly different between T2D and RT2D. The mean HbA1c for T2D patients was 7.03% vs. 8.99% for RT2D patients (p < 0.001), and glucose AC for T2D patients was 135.1 mg/dL vs. 158.9 mg/dL for RT2D patients (p = 0.03). Nearly all study participants were treated with Metformin as first-line standard therapy except for patients who presented with decreased renal function. Initially, we analyzed the α and β-diversity between T2D and RT2D patients. The α-diversity of the Shannon index indicated the diversity of the microbial community of different groups, and there were no significant differences between T2D and RT2D ( Figure 1A). However, in terms of the β-diversity, the principal coordinate analysis (PCoA) of Unweighted UniFrac was assessed to evaluate the total microbial composition between different groups. The PERMANOVA test revealed marginally significant differences (p = 0.054) in the overall microbial composition between T2D and RT2D ( Figure 1B); thus, we conclude that the gut microbial communities from participants diagnosed with T2D and RT2D displayed a similar α-diversity but slightly different β-diversity.

Relative Abundances of Verrucomicrobia and Fusobacteria were Reduced in RT2D
The evaluation of the β-diversity of the total microbial composition revealed marginally significant differences between samples from the patients with T2D vs. those with RT2D. This was evaluated further with a focus on operational taxonomic units (OTUs). At the phylum level, the majority of OTUs were Bacteroidetes and were slightly increased in RT2D. Furthermore, the relative abundance of Verrucomicrobia and Fusobacteria was reduced in the RT2D group (Figure 2A). This

Relative Abundances of Verrucomicrobia and Fusobacteria Were Reduced in RT2D
The evaluation of the β-diversity of the total microbial composition revealed marginally significant differences between samples from the patients with T2D vs. those with RT2D. This was evaluated further with a focus on operational taxonomic units (OTUs). At the phylum level, the majority of OTUs were Bacteroidetes and were slightly increased in RT2D. Furthermore, the relative abundance of Verrucomicrobia and Fusobacteria was reduced in the RT2D group ( Figure 2A). This tendency was observed more clearly in the Krona charts, where the species of Verrucomicrobia-Akkermansia muciniphila (A. muciniphila) was reduced in RT2D ( Figure 2B). The β-diversity of the principal coordinate analysis (PCoA) demonstrated marginally significant differences (p = 0.054) in the total microbial composition between T2D and RT2D groups.

Relative Abundances of Verrucomicrobia and Fusobacteria were Reduced in RT2D
The evaluation of the β-diversity of the total microbial composition revealed marginally significant differences between samples from the patients with T2D vs. those with RT2D. This was evaluated further with a focus on operational taxonomic units (OTUs). At the phylum level, the majority of OTUs were Bacteroidetes and were slightly increased in RT2D. Furthermore, the relative abundance of Verrucomicrobia and Fusobacteria was reduced in the RT2D group ( Figure 2A). This tendency was observed more clearly in the Krona charts, where the species of Verrucomicrobia--Akkermansia muciniphila (A. muciniphila) was reduced in RT2D ( Figure 2B).

Microbial Markers in T2D and RT2D Patients
To further elucidate the microbial markers associated with T2D and RT2D patients, we analyzed the OTUs with the linear discriminant analysis effect size (LEfSe) and heatmap analysis. The input criteria of LEfSe analysis included an LDA score >2 and an alpha value for the factorial Kruskal-Wallis test/pairwise Wilcoxon test of 0.01. We focused specifically on the genus/species level and, in an identical manner to the previous abundance of OTUs, Verrucomicrobia, A. muciniphila and Fusobacteria were enriched in T2D patients ( Figure 3A). We also conducted a heatmap analysis with the relative richness of the microbiome in high (red) and low (blue) between T2D and RT2D. Therefore, several potential microbial markers including Verrucomicrobia, A. muciniphila and Fusobacteria were revealed in T2D patients ( Figure 3B). an identical manner to the previous abundance of OTUs, Verrucomicrobia, A. muciniphila and Fusobacteria were enriched in T2D patients ( Figure 3A). We also conducted a heatmap analysis with the relative richness of the microbiome in high (red) and low (blue) between T2D and RT2D. Therefore, several potential microbial markers including Verrucomicrobia, A. muciniphila and Fusobacteria were revealed in T2D patients ( Figure 3B).

Akkermansia Muciniphila Was Negatively Correlated with HbA1c
To further elucidate the microbial markers, we utilized the basespace of the Ribosomal Database Project (RDP) classifier to reconfirm the OTUs at the genus/species level. Using the RDP classifier, we screened all of the potential microbial markers revealed by LEfSe and heatmap analysis at the genus/species level; we eliminated the microbiomes with low percentages or that were not present in statistically significant numbers in both T2D and RT2D patients. Eventually, we identified four bacterial taxa that were abundant and/or present in statistically significant quantities in samples from both T2D and RT2D patients. Compared with T2D patients, the relative abundances of A. muciniphila (T2D: 1.25%; RT2D: 0.09%, p = 0.04) and Fusobacterium (T2D: 1.29%; RT2D: 0.33%, p = 0.10) were reduced in patients with RT2D; in contrast, the relative abundances of Bacteroides vulgatus (T2D: 8.68%; RT2D: 14.34%, p = 0.03) and Veillonella denticariosi (T2D: 0.01%; RT2D: 0.10%, p = 0.01) were significantly increased in the RT2D patient cohort ( Figure 4A).

Akkermansia Muciniphila Was Negatively Correlated with HbA1c
To further elucidate the microbial markers, we utilized the basespace of the Ribosomal Database Project (RDP) classifier to reconfirm the OTUs at the genus/species level. Using the RDP classifier, we screened all of the potential microbial markers revealed by LEfSe and heatmap analysis at the genus/species level; we eliminated the microbiomes with low percentages or that were not present in statistically significant numbers in both T2D and RT2D patients. Eventually, we identified four bacterial taxa that were abundant and/or present in statistically significant quantities in samples from both T2D and RT2D patients. Compared with T2D patients, the relative abundances of A. muciniphila (T2D: 1.25%; RT2D: 0.09%, p = 0.04) and Fusobacterium (T2D: 1.29%; RT2D: 0.33%, p = 0.10) were reduced in patients with RT2D; in contrast, the relative abundances of Bacteroides vulgatus (T2D: 8.68%; RT2D: 14.34%, p = 0.03) and Veillonella denticariosi (T2D: 0.01%; RT2D: 0.10%, p = 0.01) were significantly increased in the RT2D patient cohort ( Figure 4A).
We further calculated a value representing the percentage of Bacteroides vulgatus minus the sum of the percentages of A. muciniphila, Fusobacterium and Veillonella denticariosi. This average value (T2D: 6.12 versus RT2D: 13.86, p < 0.01) was evaluated further using the receiver operating characteristic curve (ROC curve). The area under the curve (AUC) was 0.719 (p < 0.01), displaying acceptable discrimination for use in distinguishing patients with RT2D from those with T2D ( Figure 4B). Equally importantly, we found that the relative proportion of A. muciniphila was significantly negatively correlated with levels of HbA1c (r = −0.248, p = 0.02, Figure 4C). These results reveal that A. muciniphila is a unique glucose-related microbial marker that reflects the clinical status of both T2D and RT2D patients.
We further calculated a value representing the percentage of Bacteroides vulgatus minus the sum of the percentages of A. muciniphila, Fusobacterium and Veillonella denticariosi. This average value (T2D: 6.12 versus RT2D: 13.86, p < 0.01) was evaluated further using the receiver operating characteristic curve (ROC curve). The area under the curve (AUC) was 0.719 (p < 0.01), displaying acceptable discrimination for use in distinguishing patients with RT2D from those with T2D ( Figure 4B). Equally importantly, we found that the relative proportion of A. muciniphila was significantly negatively correlated with levels of HbA1c (r = −0.248, p = 0.02, Figure 4C). These results reveal that A. muciniphila is a unique glucose-related microbial marker that reflects the clinical status of both T2D and RT2D patients.

Microbiome-Related Functional Pathways between T2D and RT2D Patients
To further determine the function of the microbiome in T2D and RT2D patients, we used a PICRUSt prediction of different proportions of the microbiome against the KEGG level 3 pathway; the criteria included the removal of unclassified reads, a value of p <0.05 and an effect size of 1 from STAMP software. We identified 50 functional pathways that were enriched in RT2D versus T2D, and most pathways were too small to be considered further; thus, we concentrated on those with a proportion of ≥0.05%. Compared with the results from T2D patients, pathways contributing to the proportion of the microbiome against the biosynthesis of unsaturated fatty acids and arachidonic acid metabolism were reduced, and those associated with the phosphotransferase system (PTS) were

Microbiome-Related Functional Pathways between T2D and RT2D Patients
To further determine the function of the microbiome in T2D and RT2D patients, we used a PICRUSt prediction of different proportions of the microbiome against the KEGG level 3 pathway; the criteria included the removal of unclassified reads, a value of p <0.05 and an effect size of 1 from STAMP software. We identified 50 functional pathways that were enriched in RT2D versus T2D, and most pathways were too small to be considered further; thus, we concentrated on those with a proportion of ≥0.05%. Compared with the results from T2D patients, pathways contributing to the proportion of the microbiome against the biosynthesis of unsaturated fatty acids and arachidonic acid metabolism were reduced, and those associated with the phosphotransferase system (PTS) were increased in RT2D patients ( Figure 5). These results indicated that the differences observed concerning the microbiome content might have an impact on several critical functional pathways. increased in RT2D patients ( Figure 5). These results indicated that the differences observed concerning the microbiome content might have an impact on several critical functional pathways.

Discussion
Several studies that were published over the past few years have provided insight into the relationships between the gut microbiota and diabetes. Among these findings, specific groups and/or species of the gut microbiota were found to increase the endotoxemia of lipopolysaccharides (LPS), which can induce the secretion of proinflammatory cytokines and can promote insulin resistance and diabetes [21][22][23]; this information suggested a crucial link between dysbiosis and the development of T2D. Similarly, secondary bile acids [16,17] and short-chain fatty acid (SCFA) [18] also influence metabolism and homeostatic blood glucose levels [24]. The dynamic and adjustable nature of the gut microbiota may provide important information concerning the early detection and prevention of T2D as well as the current level of glycemic control.
However, only a few studies have described refractory diabetes, and studies on the relationship between gut microbiota and refractory diabetes are limited. The environment plays a large role in RT2D; this condition may occur more frequently within specialty practices than in primary care clinics, as a specialist physician may not have frequent contact with all patients [7,9]. Furthermore, poor healthcare, a lack of adherence to treatment and clinical inertia have all been suggested as reasons underlying the development of refractory disease. In 2014, one crucial study found a type of refractory hyperglycemia after Roux-en-Y gastric bypass (RYGB) surgery, indicating that the gut microbiota may be intimately involved in the development of RT2D [25,26]. In our studies, the microbial diversity in terms of the Shannon index showed no significant difference, but the total microbial composition revealed marginally significant differences (p = 0.054) between T2D and RT2D, indicating that the microbial diversity may not be a critical factor, but that specific species may be involved in the pathogenesis of RT2D disease.
To this end, our results revealed a reduced abundance of Fusobacterium (T2D: 1.29%; RT2D: 0.33%, p = 0.10) and an increased abundance of Veillonella denticariosi (T2D: 0.01%; RT2D: 0.10%, p = 0.01) in fecal samples from RT2D patients. Fusobacterium is a typical colorectal cancer-related pathogen [27,28]. Fusobacterium was also found to be more abundant in diabetes patients than in nondiabetic subjects [29,30]; as such, the role of Fusobacterium concerning the pathogenesis of RT2D needs further study. Veillonella is a gram-negative anaerobic cocci that normally resides in the gastrointestinal tract, oral cavity and vagina [31,32]. An increased abundance of Veillonella was detected in the saliva of the oral microbiome in T2D and gestational diabetes mellitus, as well as a pathogen in diabetic patients with osteomyelitis [33][34][35]. The species of Veillonella denticariosi (V.

Discussion
Several studies that were published over the past few years have provided insight into the relationships between the gut microbiota and diabetes. Among these findings, specific groups and/or species of the gut microbiota were found to increase the endotoxemia of lipopolysaccharides (LPS), which can induce the secretion of proinflammatory cytokines and can promote insulin resistance and diabetes [21][22][23]; this information suggested a crucial link between dysbiosis and the development of T2D. Similarly, secondary bile acids [16,17] and short-chain fatty acid (SCFA) [18] also influence metabolism and homeostatic blood glucose levels [24]. The dynamic and adjustable nature of the gut microbiota may provide important information concerning the early detection and prevention of T2D as well as the current level of glycemic control.
However, only a few studies have described refractory diabetes, and studies on the relationship between gut microbiota and refractory diabetes are limited. The environment plays a large role in RT2D; this condition may occur more frequently within specialty practices than in primary care clinics, as a specialist physician may not have frequent contact with all patients [7,9]. Furthermore, poor healthcare, a lack of adherence to treatment and clinical inertia have all been suggested as reasons underlying the development of refractory disease. In 2014, one crucial study found a type of refractory hyperglycemia after Roux-en-Y gastric bypass (RYGB) surgery, indicating that the gut microbiota may be intimately involved in the development of RT2D [25,26]. In our studies, the microbial diversity in terms of the Shannon index showed no significant difference, but the total microbial composition revealed marginally significant differences (p = 0.054) between T2D and RT2D, indicating that the microbial diversity may not be a critical factor, but that specific species may be involved in the pathogenesis of RT2D disease.
To this end, our results revealed a reduced abundance of Fusobacterium (T2D: 1.29%; RT2D: 0.33%, p = 0.10) and an increased abundance of Veillonella denticariosi (T2D: 0.01%; RT2D: 0.10%, p = 0.01) in fecal samples from RT2D patients. Fusobacterium is a typical colorectal cancer-related pathogen [27,28]. Fusobacterium was also found to be more abundant in diabetes patients than in non-diabetic subjects [29,30]; as such, the role of Fusobacterium concerning the pathogenesis of RT2D needs further study. Veillonella is a gram-negative anaerobic cocci that normally resides in the gastrointestinal tract, oral cavity and vagina [31,32]. An increased abundance of Veillonella was detected in the saliva of the oral microbiome in T2D and gestational diabetes mellitus, as well as a pathogen in diabetic patients with osteomyelitis [33][34][35]. The species of Veillonella denticariosi (V. denticariosi) was first isolated from human carious dentine, and the function of V. denticariosi remained uncertain [36]. Therefore, this study showed that V. denticariosi has a potential role in diabetes.
We also found that the abundance of A. muciniphila (T2D: 1.25%; RT2D: 0.09%, p = 0.04) was significantly reduced while Bacteroides vulgatus (T2D: 8.68%; RT2D: 14.34%, p = 0.03) was significantly increased in RT2D patients. Bacteroides vulgatus was identified as the major species associated with the biosynthesis of branched-chain amino acids (BCAAs) and increased insulin resistance [37]. Additionally, in this study, all subjects received the first-line standard therapy of Metformin except for patients with decreased renal function. Metformin is a biguanide that is in wide use for the treatment of T2D as a first-line OGLD. Several mechanisms have been proposed to explain the actions of Metformin, including decreased gluconeogenesis, the absorption of glucose from the gastrointestinal tract and increased insulin sensitivity and peripheral glucose uptake [38]. Interestingly, Metformin also alters the gut microbiota by way of its association with an increased abundance of mucin-degrading A. muciniphila as a means to promote glucose homeostasis [39,40]. However, in our studies, both groups of T2D and RT2D received Metformin therapy, and the abundance of A. muciniphila was still significantly reduced in refractory diabetes, indicating that the alteration of A. muciniphila was not affected by Metformin in refractory diabetes.
There is currently growing interest in the role and function of A. muciniphila, as it has been recognized as a probiotic that may be specifically effective for use in diabetes [41]. In a mouse model, the administration of either pasteurized A. muciniphila or its outer-membrane protein Amuc_1100* activated Toll-like receptor 2 (TLR2) and increased the expression of the tight-junction proteins, which in turn reversed high-fat diet (HFD)-induced obesity and reduced insulin resistance [42]. A. muciniphila may also mediate the negative effects of interferon-gamma (IFNγ) on glucose tolerance in IFNγ-deficient mice and reduce hepatic glycogen in streptozotocin-induced diabetic rats [43,44]. In humans, a decreased abundance of A. muciniphila was also observed in newly-diagnosed patients with prediabetes as well as those with T2D [41,[45][46][47]. Most importantly, we found that the relative proportion of A. muciniphila was significantly negatively correlated with HbA1c. Good glycemic control is critical for T2D patients, as reductions in HbA1c are directly associated with a reduced risk of microvascular/macrovascular complications [48,49]. Our findings were identical to those reported for human subjects in the Advento study [43]. As such, reductions in the abundance of the glucose homeostasis-related species, A. muciniphila, together with an increased abundance of the insulin resistance-related species, B. vulgatus, may have an overall impact on the therapeutic effect and the mechanisms underlying the development of RT2D.
We further found that these microbial markers were involved in several glucose-related functional pathways. Compared with T2D, the reduction of the biosynthesis of unsaturated fatty acids has an adverse impact on fasting blood glucose and insulin levels [50]. It has been reported in various studies that arachidonic acid (AA) has a positive effect on stimulating glucose uptake and that reductions in AA metabolism may have a direct impact on insulin-stimulated glucose uptake [51][52][53]. By contrast, the phosphotransferase system (PTS) utilizes phosphoenolpyruvate (PEP) as a phosphoryl donor for sugar phosphorylation and is involved in the transport of sugars into bacteria. The increase of the phosphotransferase system has been reported in prediabetes and gestational diabetes mellitus [54,55] and further enriched in refractory diabetes in our studies. These alterations of functional pathways may have contributed to the development of refractory diabetes.
The early evaluation of and aggressive interventions for glycemic control are critical in refractory diabetes. Clinical observations that are typically used to predict RT2D include early age of onset, longer duration of diabetes and the presence of microvascular/macrovascular complications, as well as insulin use and the overall complexity of the therapeutic regimen [10]. Our studies reveal potential markers within the gut microbiota that can distinguish RT2D from T2D (AUC = 0.719, p < 0.01), and the examination of the gut microbiota could evaluate the ability of glycemic control and the therapeutic results of diabetes patients. Additionally, microbiome intervention/modulation may offer another therapeutic approach for refractory diabetes patients. Probiotics including certain genera of Lactobacillus and Bifidobacterium are currently commonly used as human nutrition supplements and meet the goal of a positive effect on human health. In T2D patients, probiotic supplements have been widely studied, and several trials have demonstrated that probiotic intervention could improve fasting blood glucose, HbA1c and insulin sensitivity [56,57]. As a promising probiotic, A. muciniphila is also currently under investigation for use in various diseases, including diabetes [41,58,59].
The present study has certain limitations, as the subject numbers in this study were quite small, and larger studies involving more participants are required to solidify the value of the gut microbiota in distinguishing refractory diabetes patients from T2D. However, our findings reveal a new possibility of the gut microbiota in refractory diabetes and provide clinical implications that evaluate the ability of glycemic control and provide therapeutic strategies taking the gut microbiota into account in the future.

Conclusions
In conclusion, this work presents a novel study that elucidates the gut microbiota in refractory diabetes patients. The alteration of the gut microbiota pattern could distinguish refractory diabetes from diabetes with good glycemic control. Additionally, the abundance of Akkermansia muciniphila was significantly negatively correlated to HbA1c. These results indicated that the examination of the gut microbiota could be used to evaluate glycemic control in diabetes patients. Microbiome intervention and modulation could be used as a therapeutic strategy for improving glycemic control of refractory diabetes in the future.