Vitamin K2 supplementation improves impaired glycemic homeostasis and insulin sensitivity for type 2 diabetes through gut microbiome and fecal metabolites

T2DM subjects were recruited from 4 local communities in Harbin (Heilongjiang, China) through flyers and on-site advertisements. Inclusion was limited to subjects with a definite T2DM history and confirmation of diagnosis by physicians. The exclusion criteria were severe cardiovascular events, recurrent infections, any surgery, gastrointestinal disease, organ failure, use of antibiotics or warfarin in the past 3 months, and use of more than 2 kinds of medicines for glycemic control. In addition, participants did not take any vitamin K2 supplements or probiotics. Eligible subjects were randomly assigned to the vitamin K2 supplementation group (referred to as the “VK group” in subsequent analysis, 90 µg MK-7 was added in 100 g yogurt, 1 cup (100 g)/day) or the control group (“NC group,” 100 g yogurt without MK-7 added, 1 cup (100 g)/day). To ensure compliance in the study population and to take into account the characteristics of vitamin K2, we used yogurt instead of a capsule as a good carrier for vitamin K2 [23, 24] and MK-7 as a supplemental form of vitamin K2 with a longer half-life and high bioavailability. There was no difference of the yogurt in appearance or taste between the two groups. Yogurt products were pasteurized to ensure they were free of bacteria and other possible contaminants. Yogurt was allocated by community staff according to labels with the subjects’ names on the yogurt cups, but they had no information about our study design, nor did the subjects. The choice of dose was based on our previous research [25]. The detailed study design, sample calculation, fecal collection method, and implementation process are described in Fig. 1A and Additional file 1.

In the first part of the study, 42 male C57BL/6N mice (Beijing Vital River Laboratory Animal Technology Co., Ltd, China) were randomly divided into 3 groups at 8 weeks of age after 1 week of adaptation: control (“NC group,” N = 7), high-fat (“HF group,” N = 28, 60% high-fat diet), and HF + MK-7 (“HFVK group,” N = 7, 60% high-fat diet). In addition, mice in the HFVK group were administered 50 µg/kg weight MK-7 supplement (low-dose supplementation compared to other animal studies [26, 27]) every other day by oral gavage, and the other two groups were given corresponding amounts of solvent. Considering the single feeding environment of experimental mice and the difference in composition from the human gut microbiota [28], we chose the feces provided by the mice in the same environment as the source of FMT to objectively observe the effects of the gut microbiota after treatment with MK-7. The mice were subjected to an oral glucose tolerance test (OGTT) at week 9 and week 15 of the study.

In the second part of the study, 28 mice in the HF group were further randomly divided into 4 groups in study week 10: the HF group (N = 7) was treated as the previous “HF group” in the first part of the study, HF + mixed antibiotics (“HFABX group,” N = 7), HF + mixed antibiotics + feces from the NC group (“NCR group,” N = 7), and HF + mixed antibiotics + feces from the HFVK group (“VKR group,” N = 7). For mixed antibiotic (ABX) treatment, mice were treated with ampicillin (1 g/L, Aladdin, Cat# A105483), metronidazole (1 g/L, Aladdin, Cat# M109874), neomycin (1 g/L, Aladdin, Cat# N109017), vancomycin (500 mg/L, Aladdin, Cat# V301569) and sucralose (0.05 mg/ml, Aladdin, Cat# S107614) in the drinking water for 14 days [29] during weeks 10–12 of the study. From week 13, FMT was conducted once a day for the first 3 days in the first 2 weeks to promote microbiota engraftment and then twice per week thereafter in the last 2 weeks to maintain the effect of the transplant. FMT lasted a total of 4 weeks from week 13 to week 16 [30]. The mice were subjected to an OGTT at week 16 of the study. It should be emphasized that this part had a noninferior experimental design, as germ-free mice and the healthy gut microbiota could protect against diet-induced obesity (DIO) [31].

Fresh feces from the NC group and HFVK group were collected in the morning and evening before the day of transplantation and then immediately frozen at − 80 °C. On the day of transplantation, mixed feces from each group were suspended in PBS buffer (vortex, 5 min) at a ratio of 1 (mg): 100 (mL) before centrifugation at 600 × g for 5 min. The final supernatant was used to gavage each mouse (NCR group and VKR group) with a dose of 10 μL/g body weight, and correspondingly, mice in the HFABX group were given solvent.

All analyses were conducted by using R software version 4.1.1. Detailed statistical and bioinformatic methods are enclosed in the Additional file 1.

Downstream 16S rRNA analyses were conducted by the R package “microeco” and “Tax4fun.” Genus biomarkers between groups were tested by the rank-sum test and Lefse analysis (raw P < 0.05 or LDA score > 3). Differences in fecal metabolites were evaluated by using the Wilcoxon rank-sum test and variable importance for the projection score (VIP) (raw P < 0.05 or VIP score > 1). Differentially expressed genes (DEGs) were selected under the criteria of │log₂fold change│ > 1 and P < 0.05 to identify intergroup gene expression differences and whole transcriptomes were subjected to gene set enrichment analysis (GSEA) to identify intergroup functional differences by using the R packages “DESeq2” and “clusterProfiler.” All data are presented in figures as mean ± SEM.

All statistical analyses were conducted using two-tailed hypothesis testing.

Eighty eligible subjects were ultimately enrolled in our evaluation and randomized to either the 6-month supervised NC group or VK group, of whom 60 completed the entire study (n = 30 in each group) (Fig. 1A; Additional file 1: Fig.S1). It was shown that there was no significant difference between the two groups at baseline (Table 1).

After the 6-month intervention, a significant decrease in serum dp-ucMGP in the VK group compared to the NC group indicated that the circulating levels of vitamin K2 in the body had been significantly increased (VK_change =  − 165.10 ± 25.30 pmol/L, P < 0.001, Table 1) [32]. As our primary outcome, notable reductions of 13.4%, 28.3%, and 7.4% in fasting serum glucose (VK_change =  − 1.08 ± 0.49 mmol/L, P = 0.048), insulin (VK_change =  − 3.15 ± 0.39 μU/mL, P = 0.005), and Hb1Ac levels (VK_change =  − 0.57 ± 0.23%, P = 0.019) were observed in the VK group, together with modest improvements in HOMA index and lipid parameters (Fig. 1B–M, Table 1). However, body mass and the arteriosclerosis index showed no significant differences within or between groups (Table 1). Given the importance of the gut microbiota in glycemic homeostasis, we next investigated the possibility of its involvement in the metabolic effects of vitamin K2 intervention.

16S rRNA sequencing of fecal samples revealed that more taxa were observed at termination than at baseline (1682 OTUs vs 1373 OTUs) after the 6-month intervention (Fig. 2A). Although the evenness of the microbiota was not different among groups (Simpson index), the VK6 group (VK group at month 6) showed a protective effect on the richness when compared to the NC6 group (NC group at month 6) (observed index, P < 0.001, Fig. 2A, B). In addition, the beta diversity of the microbial community composition, evaluated by the weighted UniFrac distance, showed a significant difference between the NC6 group and the VK6 group after the 6-month MK-7 intervention (all P < 0.05, Fig. 2C, Additional file 1: Fig.S3A).

At the phylum level, compared to the baseline status, the ratio of Firmicutes and Bacteroidetes (F/B) was significantly increased in the NC6 group (NC0 = 4.95 ± 0.96, NC6 = 11.22 ± 2.18, P = 0.012), but there was no difference between the VK6 and VK0 groups (VK0 = 4.65 ± 1.10, VK6 = 7.41 ± 1.76, P = 0.194) (Additional file 1: Fig.S3B). Although we did not observe the difference between the VK group and NC groups at month 6 (P = 0.187), there was a significant positive correlation between the change in serum dp-ucMGP and the change in F/B value in all participants (r_Pearson = 0.33, P < 0.01, Additional file 1: Fig.S3B), indicating that the improvement nutritional status of circulating MK-7 is related to lower F/B value. Furthermore, the microbiota profile showed that the relative abundance of the genera Faecalibacterium, Intestinimonas, and Anaerofilum (order Clostridiales); genera Butyricimonas, Barnesiella, and Paraprevotella (order Bacteroidales); and the genera Dielma and Turicibacter (order Erysipelotrichales) was significantly increased in the VK6 group compared to the NC6 group (Fig. 2D). Although the genera Barnesiella, Paraprevotella, and Dielma, showed a decreasing trend throughout the study in both the NC and VK groups, the MK-7 intervention significantly delayed the decline of these genera (Fig. 2D). From an overall view of the family level, Ruminococcaceae, Lachnospiraceae, and Bacteroidaceae were relatively dominant families in the VK6 group compared to NC6, whereas Enterobacteriaceae was a dominant family in the NC6 group (Additional file 1: Fig.S3C). Moreover, coabundance network analysis suggested increased density and enhanced interactions in the microbial community after MK-7 intervention compared to baseline or the corresponding NC group (Fig. 2E; Additional file 1: Fig.S3D).

When comparing the fecal metabolites at baseline and at the end of the study, we observed significant differences in dynamic alterations between the two groups (OPLSDA, Additional file 1: Fig.S3E). Notably, a significant decrease in total branched-chain amino acids (BCAAs) and histidine concentrations was observed after 6 months of intervention, as well as an increase in lithocholic acid (LCA) and short-chain fatty acids (SCFAs) in fecal samples (Fig. 2F). Other changes were mainly related to energy metabolism, such as organic acids or long-chain fatty acids.

Similar to the differences in fecal metabolites, functional enrichment of the microbiota showed a relative enhancement of the capacity for BCAA biosynthesis and amino acid-related enzymes in the NC6 group, although the enhancement of the biosynthesis of aromatic amino acids was only observed in enrichment analysis. On the other hand, MK-7 intervention seemed to involve gut microbiota protein glycosylation reactions, secondary metabolite biosynthesis, and carbohydrate and amino acid metabolism, such as glycosylphosphatidylinositol anchor biosynthesis, flavone and flavonol biosynthesis, glycan biosynthesis and metabolism, and amino sugar and nucleotide sugar metabolism (Additional file 1: Fig.S3F). These results suggested there was a potential association between MK-7 and microbiota and co-metabolites.

After adjustment for the corresponding status in baseline, we found several strong associations between alterations in genera and fecal metabolites with improvements in glycemic parameters and a cluster of metabolic parameters in the VK group after a 6-month MK-7 intervention. SCFAs and genera whose relative abundance was higher in the VK6 group were associated with reductions in serum Hb1Ac, glucose, insulin and insulin resistance (Fig. 3). Some of these changes, such as Anaerofilum, Asteroleplasma, Paraprevotella, LCA, glutamic acid, and methysuccinic acid, which were found increased in the VK6 group were related to the reduction of lipid parameters such as fat mass, total cholesterol (TCHO), LDL-c, hip circumference as well as blood pressure and arteriosclerosis (Fig. 3). Taken together, changes in these genera as well as fecal metabolites with MK-7 intervention may provide a biological basis for the improvement in clinical indicators in T2DM participants.

To further explore the causal relationship between the MK-7-regulated microbiota and the changes in glucose tolerance and fat accumulation induced by MK-7 intervention, we first performed a 16-week supplementation study in mice fed a 60% high-fat diet and then performed a 4-week FMT study by transplanting feces from donor mice into antibiotic-treated mice (Fig. 4A). As we expected, after both 10 study weeks and 16 study weeks, significantly lower weight gain and better glucose tolerance were observed in NC mice (all P < 0.05) and VK mice (all P < 0.05) than in HF mice (Fig. 4B–D). Importantly, at 4 weeks after the first FMT, mice colonized with the microbiota from VK mice and NC mice showed reduced weight gain and improved OGTT results compared to HFABX mice (all P < 0.05, Fig. 4E–F). Intriguingly, although the glucose tolerance of VK mice (fed a 60% high-fat diet) appeared to be inferior to that of NC mice (fed a standard AIN-93 M diet), the effect of the MK-7-regulated microbiota was comparable to that of the microbiota from lean donors.

Upon further comparison of biochemical markers, the MK-7-regulated microbiota significantly improved fasting glucose, triglyceride (TG), adiponectin, and endotoxin levels and appeared to be able to slightly normalize the percentage of white adipose tissue (WAT%) and lean mass in mice with long-term high-fat feeding (Fig. 4G, Additional file 1: Fig.S4, Additional file 1: Table S2).

When we once again detected alterations in the gut microbiota to validate whether MK-7 could independently affect the microbiota (eliminating the potential effects of yogurt and antidiabetics on gut microbiota in the previous RCT study), we found no significant changes in the number of taxa and common species among the three groups, except for the significant difference in overall composition between the NC group and the other two groups (P < 0.001) (Additional file 1: Fig.S5A-S5D). Considering the single rearing and feeding environment of the mouse study, compared to our human trial, we observed fewer taxa, and fewer genera differences such as the higher relative abundance of Akkermansia, Bilophila, and Alloprevotella in the HFVK group than in the HF group (Additional file 1: Fig.S5E-S5F). Although the taxa Akkemansiaceae and Desulfovibrionaceae were not dominant families in our previous human trial, the functional enrichment of the MK-7-regulated microbiota in the mouse model still exhibited a similar enhancement of the capacity for energy metabolism, amino acid metabolism, and protein glycosylation reactions compared to the HF group (Additional file 1: Fig.S5G-S5H). It should be noticed that the genera Flexilinea (P = 0.031) and Escherichia-Shigella (P = 0.004) were significantly increased in the HF group, and the genera Lachnospiraceae_FCS020_group (P = 0.006) and Lachnospiraceae_AC2044_group (P = 0.017) were significantly decreased in the HF group when compared to the NC group. These four genera were restored after the MK-7 supplementation (all P < 0.05, Additional file 1: Fig.S6A). Flexilinea belongs to the methanogens, which have been reported to be positively correlated with the increase of blood glucose and blood lipid parameters [33]; Escherichia-Shigella are the potentially pathogenic bacteria which have been correlated with the dysbiosis of microbiota [34]; and Lachnospiraceae are the family subset of the phylum Firmicutes which are known to be responsible for the production of SCFA [35].

Due to the alterations in the gut microbiota after MK-7 intervention, we next detected the expression and functional changes in the transcriptomic profiles of colon, liver, and pancreas tissues in FMT mice. DEG analysis of the three tissues revealed hundreds of DEGs between NCR and VKR as well as between HFABX and VKR, of which the pancreas appeared to be the less affected tissue (Fig. 5A). Compared to HFABX mice (almost sterile, with low microbial activity), 22, 20, and 14 pathways were significantly enriched in the liver, colon, and pancreas, respectively (Additional file 1: Table S3). Focusing on the few top pathways based on the normalized enrichment score (NES) and leading-edge subset genes among the three tissues, most were associated with the upregulation of glycerolipid metabolism, lipid metabolism, steroid biosynthesis, and glycan biosynthesis as well as downregulation of the amino acid metabolism and the immunoinflammatory response (Fig. 5B, Additional file 1: Fig.S6B). Similar results were obtained when comparing between NCR and VKR (Fig. 5B). The MK-7-regulated microbiota seemed to have a stronger effect on the colon and pancreas compared to NCR mice and have a stronger effect on the colon and liver compared to HFABX mice.

Additionally, in the histological evaluation of liver tissue, we observed a decreased trend in steatosis as well as reduced inflammatory cells in VKR mice, which were comparable to those in NCR mice (Fig. 5C). However, no significant pathological changes were observed in HE staining of the pancreatic tissue (Fig. 5C). Furthermore, the reduction in colonic and ileal goblet cell density and normal goblet cells caused by the 60% high-fat diet was partially restored by the transplantation of MK-7-regulated microbiota (Fig. 5D, Additional file 1: Fig.S7A).

To further understand the gut-derived mechanism by which MK-7 ameliorates inflammation and glycemic homeostasis, we first measured SCFA, bile acid, and amino acid profiles in the feces of the donor group to reconfirm whether MK-7 could influence the effects of the microbiota on these metabolites. Similar to the previous RCT, several SBAs and SCFAs, such as lithocholic acid, 7-ketolithocholic acid, taurodeoxycholic acid, acetic acid, butyric acid, and valeric acid showed an increasing trend in the VK group compared with the HF group (Fig. 6A). We also found that although there was a decreasing trend in short-chain fatty acids compared to the NC group, there was still a significant increase in muricholic acid and deoxycholic acid (Fig. 6A). Additionally, it may be due to the heterogeneity of dietary sources that the fecal amino acids BCAA and histidine exhibited no difference between groups in our mouse model but were found to be significantly decreased in T2DM participants after 6 months of MK-7 intervention (Fig. 6B). On the other hand, consistent with the differences in functional enrichment and fecal metabolites of the VKR group, RT-qPCR showed significantly altered mRNA expression of the bile acid receptor TGR5 (Gpbar1, fold change = 2.42, P = 0.036), Vdr (fold change = 2.27, P = 0.004), and FXR (Nr1h4, fold change = 0.64, P = 0.100), and IHC staining also showed the increased expression of TGR5 in the cytoplasm and plasma membrane of colon tissue in VKR group (Fig. 6C). In addition, the anti-inflammatory interleukins IL2r (fold change = 2.63, P = 0.034), IL11 (fold change = 1.97, P = 0.031), and IL13 (fold change = 2.15, P = 0.004) were observed to be upregulated in the VKR group compared to the HFABX group in ileocecal tissue (Fig. 6D, Additional file 1: Fig.S7B). In visceral adipose tissue, we also found that the mRNA expression of the pro-inflammatory interleukins Il1b (fold change = 0.28, P = 0.017), Il6 (fold change = 0.50, P = 0.036), and anti-inflammatory interleukin Il13 (fold change = 3.39, P < 0.001) was significantly altered (Additional file 1: Fig.S7B). Furthermore, we observed an increased concentration of the circulating incretin GLP-1 in the VKR group compared to the HFABX group and this difference was not found between NCR and HFABX (Fig. 6E).

It should be noted that there are no differences in the concentrations of vitamin K2 (MK-4 and MK-7) in cecal contents and gla-osteocalcin in serum were detected between the donor groups (Fig. 6E), indicating that the improvement induced by FMT is not due to the direct effect of vitamin K2 itself but to the regulation of altered gut microbiota. Taken together, the alterations in the microbial metabolites SBAs and SCFAs, the mRNA expression of bile acid receptors and interleukins in several tissues, and the circulating GLP-1 concentrations with unchanged levels of intestinal vitamin K2 further revealed a potential ability of the MK-7-regulated microbiota to improve host inflammatory status and glycemic homeostasis.