Microbial and metabolic features in renal transplant recipients with post‐transplantation diabetes mellitus

Post‐transplantation diabetes mellitus (PTDM) is a common complication in renal transplant recipients (RTRs). Gut microbiome plays important roles in a variety of chronic metabolic diseases, but its association with the occurrence and development of PTDM is still unknown. The present study integrates the analysis of gut microbiome and metabolites to further identify the characteristics of PTDM.


INTRODUCTION
The short-term survival rates of renal transplant recipients (RTRs) have been greatly increased with the development of surgical technology and use of new immunosuppressive regimens, but the long-term survival rate has undergone little improvement in recent years. 1 The posttransplantation diabetes mellitus (PTDM) is one of the most important metabolic disorders after renal transplantation, which has significant effects on the long-term outcome of kidney graft. 2 PTDM is characterized by a decrease in insulin secretion and an increase in resistance. 3,4 The incidence rate of PTDM in RTRs is as high as 30%. 5 Compared with RTRs with normal glucose tolerance (NGT), RTRs with PTDM are at higher risk of adverse consequences, including cardiovascular disease and impaired graft function, which will lead to a decrease in long-term graft survival. 6,7 Thus, it is important to find novel diagnostic markers and therapeutic targets for PTDM, to improve the prognosis of RTRs with PTDM.
Gut microecosystem plays an important role in human health and diseases. 8 Specific changes of gut microbiome are associated with the development of diabetes, because they can interact with intestinal barrier function, immune homeostasis, and blood sugar levels through releasing metabolites. 9,10 It was reported that the changes in the proportion of Lactobacillus, Akkermansia muciniphila, and Faecalibacterium prausnitzii were significantly associated with the glycemic status and development of PTDM in RTRs. 11 A study by Chang et al. also demonstrated that the modification of gut microbiome profiles could improve insulin resistance in mice. 12 However, few studies have investigated the characteristics of gut microbiome in RTRs with PTDM.
In this study, we aim to evaluate the population and function of gut microbiome using Hiseq sequencing and analyze fecal metabolic profiles using untargeted metabolomics method, to identify the potential mechanism of PTDM occurrence through analyzing the characteristics of gut microbiome and fecal metabolites in RTRs with PTDM. These results may provide new strategies for individualized health management, prediction, diagnosis, and treatment (such as probiotics and metabolites intervention, and flora transplantation) of PTDM.

Study subjects
A total of 100 RTRs fecal samples were collected for analysis at Sichuan Provincial People's Hospital, using the following criteria: (i) ≥3 months after renal transplantation, (ii) stable graft function (serum creatinine concentration ≤2.0 mg/ dL), and (iii) immunosuppression regimen was stable in the last month. Recipients with preexisting diabetes, or used antibiotics in the last 3 months, were excluded. Our study has been approved by the Ethics Committee of Sichuan Provincial People's Hospital (Ethical Approval No. 2020228) and was conducted in accordance with the Declaration of Helsinki. With written-informed consent, all RTRs knew and approved the use of their clinical data and fecal samples for this study.

Sample collection
All RTRs fecal samples were collected during routine fasting blood tests. Then, samples were immediately delivered to the laboratory and stored at À80°C until use. Moreover, the exposure assessment was conducted using questionnaire (Appendix S1) with questions concerning several lifestyle variables, such as smoking, alcohol consumption, physical exercise, and eating habits.

DNA extraction, metagenomic sequencing, and sequence data process
The total genomic DNA in each fecal sample was extracted using the CTAB method. The quality of DNA was determined through 1% agarose gel electrophoresis. The DNA library was constructed according to the Novogene Co., Ltd manufacturer's instructions. Fifty-five samples were submitted to Hiseq sequencing. The general clinical characteristics of these recipients were shown in Appendix S2. Sequencing and processing were performed on an Illumina HiSeq platform by Novogene Co., Ltd. More specifically, sequencing library was constructed by using TruSeq® DNA PCR-Free Sample Preparation Kit, following manufacturer's recommendations and index codes were added. Then the library quality was assessed on the Qubit® 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on an Illumina NovaSeq platform and 250 bp paired-end reads were generated. In addition, the following four steps are included in the data analysis: paired-end reads assembly and quality control, Operational Taxonomic Units (OTU) cluster, and species annotation and b-Diversity.

Untargeted metabolomics analytical strategy
One hundred fecal samples of RTRs were subjected to Novogene Co., Ltd for untargeted metabolomics analysis based on the liquid chromatography-mass spectrometry (LC-MS) method. The fecal samples were ground in liquid nitrogen, and 80% methanol aqueous solution containing 0.1% formic acid was added. After vortexing, samples were incubated on ice for 5 min and centrifuged at 15 000 rpm for 10 min at 4°C. Then, the supernatant was collected and diluted with LC -MS grade water until the methanol content is 53%. After that, a second centrifugation at 15 000g for 10 min at 4°C was performed, and the supernatant was collected and used for LC-MS analysis. A mixed sample that was equally taken from each sample was used for quality control. And a 53% methanol aqueous solution containing 0.1% formic acid was used as the blank sample.

Statistical analysis
Statistical analyses were performed using the SPSS V.25.0 and GraphPad Prism V.8.0. LEfSe analyses were performed using LEfSe Software 1.0. Numerical data are presented as mean AE standard deviation. One-way ANOVA analysis was used for comparison of baseline characteristics between different groups. Wilcoxon rank-sum test was used to compare the differences between two independent groups. p value <0.05 was considered statistically significant.

Clinical characteristics of RTRs
As shown in Table 1, no statistical differences were found in gender, age, body mass index (BMI), time since transplantation, renal function parameters, and life styles between the three groups. Moreover, all RTRs received a triple immunosuppressive regimen consisting of tacrolimus FK506/cyclosporine A, mycophenolate mofetil, and glucocorticoid, and there was also no significant difference in blood concentrations of tacrolimus and cyclosporine A between these groups.

Assessment of gene number
As shown in the dilution curves of Core and Pan genes, the number of genes gradually stabilized as the number of sequenced samples increased (Figure 1a,b). Moreover, the Venn diagram showed that a total of 903534 genes were detected in all groups; notably, there are 264009 genes were only detected in RTRs with PTDM ( Figure 1c).

Gut microbial diversity
Our results showed that there was no statistically significant difference in gut microbial diversity between the three groups, although the average species richness in PTDM group was higher than that in other groups, which was estimated by Shannon index (Figure 1d) and Simpson index (Figure 1e). To evaluate the microbiome space between different groups, the b diversity was calculated by weighted UniFrac, and the results of principal coordinate analysis (PCoA) and the non-metric multidimensional scaling (NMDS) analysis showed that there were differences in the composition of gut microbiome among the three groups ( Figure 1f,g).

Phylogenetic profiles of fecal microbial communities
To evaluate the differences in flora composition, the relative abundance levels of the phylum and genus with the top 10 total abundances were analyzed in different groups. As shown in Figure 2a, three of the phyla, Firmicutes, Proteobacteria, and Bacteroidetes, accounted for more than 80% of the sequences on average, which could be regarded as the dominant species in RTRs. With the increase of FPG, the relative abundances of phyla Actinobacteria and Fusobacteria increased and that of phylum Proteobacteria decreased, while the relative abundance of genus Prevotella increased and those of genera Clostridium and Escherichia decreased (Figure 2b). In the IFG group, 3 types of bacteria were significantly enriched, including family Ruminococcaceae, genus Ruthenibacterium, and species Ruthenibacterium_lactatiformans ( Figure 2c). Cladograms were used to represent taxa enriched in different groups, and only the bacteria at the family level were labeled. The diameter of each circle's diameter is proportional to the taxon's abundance (Figure 2d,f).

Major bacteria associated with PTDM
Linear discriminant analysis effect size (LEfSe) was used to estimate the maximum difference of the microbial structures in the three groups, to determine the specific bacterial taxa and predominant bacteria which were associated with PTDM. Our LEfSe analysis based on LDA >3 found that there were 3 types of bacteria including genus Dialister, family Peptostreptocpccaceae, and species Dialister invisus were significantly enriched in the PTDM group (Figure 2e). It indicated that the gut microbial ecology of RTRs with PTDM was disordered.

Importance assessment of gut microbiome by random forests
The importance of gut microbiome in the PTDM group was evaluated by constructing a random forest classifier. We performed a tenfold cross validation on the random forest model of the PTDM and control groups. The result showed that there were 4 OTU markers to be selected as the best marker set, including species Dialister invisus. The POD index means the ratio of the number of randomly generated decision trees for predicting RTRs with PTDM and healthy controls, and it achieved an AUC value of 91.79% with 95% CI of 79.96% to 100% between the two groups ( Figure 3a). The results of mean decrease accuracy and mean decrease gini showed the importance ranking of four OTU markers (Figure 3b,c).

Crucial function related to PTDM
To further explore the biological metabolic pathways in RTRs with PTDM, the functional annotation analysis was performed through searching and comparing the genomic data generated by metagenomic sequencing in the Kyoto Encyclopedia of Genes and Genomes (KEGG) (Figure 4a). As shown in Figure 4b, the functions of amino acid biosynthesis such as lysine, phenylalanine and tryptophan, and peptidoglycan biosynthesis were significantly increased in RTRs with PTDM, while the functions of carbohydrate metabolism such as fructose, galactose, propionate and butyric acid, lipid metabolism, and amino acid metabolism such as alanine, aspartate, and glycine were significantly decreased, when compared with the control group.

Differences in metabolite profile
To identify the differential metabolites between groups, partial least squares discriminant analysis (PLS-DA) was used to build a model which could evaluate the correlation between metabolite expression level and RTR status. The models of the PTDM group and control group were obviously separated, as well as those of the IFG group and control group, which well fitted the explanatory variables (R2Y = 0.88, Q2Y = 0.01) (R2Y = 0.88, Q2Y = À0.06) (Figure 5a,d).
Based on the PLS-DA models, differentially expressed metabolites between groups were identified using the following criteria: variable importance in projection (VIP) value greater than 1.0, fold change of mean expression level between groups greater than 1.2 or <0.833, and p value <0.05. A total of 122 differentially expressed metabolites were identified between the PTDM and control groups, of which 114 metabolites were significantly decreased and 8 metabolites were significantly increased in the PTDM group (Appendix S3, Tables S1 and S2).
The results of KEGG analysis showed that the steroid hormone biosynthesis and tyrosine metabolism were the main pathways for the enrichment of differentially expressed metabolites (p < 0.05) (Figure 5e). And the pregnenolone, estriol, 2-methoxyestrone, 3,5-diiodotyrosine, rosmarinic acid, and homogentisic acid were presented as the differentially expressed metabolites, and their expression was significantly decreased in RTRs with PTDM.

Correlation of major gut microbiome and metabolites with FPG
The correlation of different gut microbiome and metabolites with FPG was evaluated by calculating Spearman's rank correlation coefficient. As shown in Figure 6a, the expressions of pregnenolone and estriol were negatively correlated with FPG (p < 0.05). The spearman correlation coefficients were À0.389 and À0.543, respectively. Moreover, the expressions of genus Dialister and species Dialister invisus were positively correlated with FPG (p < 0.05), and the Spearman correlation coefficients were 0.339 and 0.310, respectively (Figure 6b). Correlation among differentially expressed metabolites, gut microbiome, and microbial function To further explore the correlation between the diversity of gut microbiome and representative differentially expressed metabolites in RTRs with PTDM, we analyzed the abundance of the major gut microbiome and fecal metabolites. As shown in Figure 6c, we found that the species Dialister invisus was significantly correlated with the expression of pregnenolone and estriol, and the Spearman correlation coefficients were À0.358 and À0.401, respectively (p < 0.05). It indicated that gut microbiome has a certain effect on the metabolic characteristics of RTRs with PTDM. Moreover, many microbial functions were significantly covariant with the concentrations of gut microbiome and metabolites in the PTDM group, when compared with the control group (Figure 6d).

DISCUSSION
The disorder of gut microbiome has been reported in the pathogenesis of many chronic diseases, such as chronic kidney disease, 14 obesity, 15 and liver cirrhosis. 16 To date, few studies have investigated the association between PTDM and gut microbiome. In this study, the gut microbiome and metabolome in RTRs with PTDM were fully characterized through Hiseq sequencing and untargeted metabolomics analysis of fecal samples. To the best of our knowledge, the present study was the first to demonstrate that there were significant differences in gut microbiome and metabolite profiles between the RTRs with IFG or PTDM and RTRs with normal glucose tolerance, indicating that there might be alterations in the gut microbiome and metabolites in the development of PTDM. A previous study by Zaza et al. reported that there was a similar degree of alpha diversity in RTRs who received different immunosuppressant regiments 17 consistent with that; our results also demonstrated that there was no significant difference in the richness of the gut microbiome when using measures of alpha diversity between groups. However, we found that there were significant differences in the composition of gut microbiome when using measures of beta diversity based on weighted unifrac between groups. There might be one explanation. The immunosuppressive agents taken by RTRs could cause significant structural alterations in the gut microbiome, thus reducing the diversity of the gut microbiome in RTRs when compared with healthy controls, 11 which would lead to a similar degree of alpha diversity between groups. After we analyzed the gut microbiome of RTRs at different glucose tolerance stages at the phylum and genus levels, our results demonstrated that the abundance of phyla Actinobacteria, Fusobacteria, and genus Prevotella increased with the increase of FPG, indicating that these gut microbiomes might be the important pathogen leading to the increase of blood glucose. Moreover, the results of LEfSe analysis showed that a total of 8 types of bacteria with different classification were significantly enriched in the PTDM and control groups, and the significant difference in the abundance of these bacteria might lead to the alterations in the composition of gut microbiome. In addition, the genus Dialister, family Peptostreptocpccaceae, and species Dialister invisus were significantly enriched in RTRs with PTDM. Previous study reported that the relative abundance of family Peptostreptocpccaceae was decreased in patients with gestational diabetes when treated with metformin. 18 At present, only one study has demonstrated that the increase of the relative abundance of species Dialister invisus was significantly correlated with the development of pre-T1D. 19 Further studies are required.
Microbial function is also associated with the development of many diseases. In this study, we found that the biosynthetic functions of amino acids such as lysine, phenylalanine and tryptophan, and peptidoglycan, as well as b-galactosidase were significantly increased in RTRs with PTDM, which may reflect the decrease of amino acid availability. Moreover, our results demonstrated that the functions of carbohydrate metabolism such as fructose, galactose and butyric acid, lipid metabolism, amino acid metabolism such as alanine, aspartate and glycine, lysozyme and sucrose synthase were significantly decreased in RTRs with PTDM. The metabolic disorders of carbohydrates and lipids are consistent with the clinical symptoms of diabetes. As one of the short-chain fatty acids (SCFA), butyric acid is considered to be the main type of gut microbiome-dependent metabolites involved in the progression of diabetes, 20 which has been confirmed in a mouse model of the abnormal glucose metabolism induced by tacrolimus. Furthermore, we also performed non-targeted metabolomics analysis of fecal samples. The results of PLS-DA showed that the models of the PTDM group and control group were obviously separated, as well as those of the IFG group and control group, indicating that fecal metabolites could be used as molecular markers for early diagnosis of IFG and PTDM in RTRs.
In this study, we found that the expressions of pregnenolone and estriol in RTRs with PTDM were significantly decreased. It was reported that steroids are associated with gestational diabetes mellitus, and pregnenolone and estriol might be differential metabolites for gestational diabetes mellitus. 20 A study by Pagotto et al. also showed that pregnenolone synthesis is reduced in STZ-induced diabetic rats. 21 Up to now, no study has investigated the association between the expressions of pregnenolone and estriol and PTDM, but our results suggest that they may have important physiological effects in the development of PTDM, and further studies are required. Moreover, we also confirmed that the abnormal gut microbiome was significantly associated with the alterations of metabolites and related functions in RTRs with PTDM, when compared with the control group.
The use of multiomics approaches would accelerate our understanding of the contributions of the gut microbiome to human health and metabolic disease 22 and improve global understanding of the functional variations which occurred in RTRs with PTDM. The present study provides a novel strategy and method for non-invasive, real-time, and individualized health management after renal transplantation. However, there were several limitations. It was reported that the microbiota of metabolites can influence host metabolism, which leads to insulin resistance and the development of diabetes mellitus. 23 Thus, the changes in the gut microbiome might be a result of PTDM, but the mechanism by which gut microbiome affect the PTDM status of host is still not clear. Moreover, the use of calcineurin inhibitors (CNIs) might possibly impact gut microbiota composition and structure, and there were differences in the percentage of recipients using tacrolimus or cyclosporine A in three groups. Thus, we could not draw any concrete conclusions yet. Further studies are required to confirm the present findings.
In conclusion, our study identified the characteristics of gut microbiome and fecal metabolites in RTRs with PTDM, and we also found two major metabolites (Pregnenolone and Estriol) and a bacterium (species Dialister invisus) was significantly associated with PTDM, which might be used as novel targets in the research field of PTDM.