WHAT IS ALREADY KNOWN ON THIS TOPIC

Type 2 diabetes has been associated with an abnormal gut microbiota in several rodent models of diabetes. Similarly, we have shown the progression of type 2 diabetes alters the gut microbiota of the University of California Davis type 2 diabetes mellitus (UCD-T2DM) rat model, a polygenic model of diabetes that spontaneously develops diabetes. As the development of diabetes is independent of diet in this model, diabetes-associated alterations in the gut microbiota are likely due to an unknown mechanism that alters the luminal environment.

WHAT THIS STUDY ADDS

Several facultative anaerobes were increased in UCD-T2DM rats that had overt diabetes for 3 months compared with UCD-T2DM rats that were age matched and either free of overt diabetes or had diabetes for less than a month. Immunohistological assessment of colonocyte epithelia, using an intraperitoneal injection of pimonidazole, showed no differences in colonocyte oxygen levels, suggesting that the diabetes-associated microbiota in the UCD-T2DM rat model is not related to colonocyte oxygen levels.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

Maintenance of colonocyte hypoxia by microbial butyrate oxidation has recently been proposed as a mechanism to maintain synergistic obligate anaerobe populations in the gut. Our data do not support this mechanism as a contributing factor to explain the increase in facultative anaerobes in UCD-T2DM rats with advanced diabetes. More studies are required to determine host-derived factors that influence the composition of the microbiota during diabetes.

Introduction

Dysbiosis of the gut microbiota has been associated with several common metabolic diseases, including obesity,1 type 2 diabetes,2 and cardiovascular disease.3 While the pathogenesis of these diseases in humans is complex and multifaceted, several preclinical studies have suggested a mechanism in which gut dysbiosis leads to the systemic accumulation of microbial-derived metabolites that is an underlying contributor to disease progression. For example, high fat intake in rodents has been associated with elevated circulating levels of lipopolysaccharide which induces insulin resistance leading to diabetes by inducing chronic inflammation in key metabolic tissues, particularly the liver.4 5 Although this model highlights an effect of diet on the gut microbiota and gut permeability, it is increasingly recognized that diet-independent host-driven factors related to glucose homeostasis can influence gut microbial composition and gastrointestinal function. Thaiss et al reported that hyperglycemia, independent of obesity, results in increased intestinal barrier permeability, leading to the accumulation of circulating microbial products with inflammatory properties in genetically manipulated and chemically induced rodent models of diabetes.6 Furthermore, diet-independent rodent models of diabetes, like the monogenic db/db mouse and the polygenic University of California Davis type 2 diabetes mellitus (UCD-T2DM) rat model, have continually shown diet-independent alterations in the composition of the gut microbiota associated with worsening glucose regulation.7 8 As diet is unaltered in these genetically susceptible models, diabetes-associated shifts in microbial populations is likely driven by host-related factors that influence the luminal environment.

Studies investigating host–microbe crosstalk mechanisms related to colonocyte metabolism of butyrate may partially explain the effect of hyperglycemia on gut bacterial populations and gut permeability.9 10 For example, Byndloss et al described how the reduction of colonocyte butyrate oxidation in dextran sodium sulfate-treated intestinal-specific peroxisome proliferator-activated receptor γ (PPARγ) knockout mice increased luminal oxygen levels, resulting in the expansion of Salmonella enterica serovar Typhimurium.11 This line of evidence supports the role of host-mediated factors that regulate microbial populations in the gut lumen; in this case, the control of a hypoxic environment suitable for the maintenance of synergistic obligate anaerobes.12 Similarly, Zheng et al10 linked colonocyte hypoxia, via the oxidation of butyrate, to enhance gut barrier function. Using various colitis models, this group showed that colonocyte hypoxia, induced by the oxidation of butyrate, activates the hypoxia-inducible factor (HIF) protein complex, which in turn, upregulates genes involved in tight junction formation and mucus secretion, as well as several other responses that promote gut barrier function.13–15

These purported mechanisms have not been widely investigated under hyperglycemic conditions; however, it is possible that hyperglycemia and/or diabetes may dysregulate the preferential oxidation of butyrate, leading to oxygen infusion into the lumen and gut barrier dysfunction. In the absence of butyrate, germ-free mice increase lactate production via anaerobic glycolysis and increase basolateral glucose uptake,16 suggesting that a lack of butyrate increases the reliance on glucose oxidation for cellular energy production. Several lines of evidence suggests that diabetes is associated with a reduction in butyrate-producing bacteria,17–19 so a working hypothesis is that the reduction of available butyrate combined with the pressure to remove excess glucose during diabetes may promote colonocyte utilization of glucose for its energy production. Consequently, this would allow for an increase in luminal oxygen availability and inhibition of the HIF signaling pathway, both allowing for conditions that promote gut dysbiosis and barrier dysfunction.

As indicated above, we have observed diabetes-associated alterations in the composition of the microbiome in UCD-T2DM rats.7 8 The UCD-T2DM rat model exhibits obesity, insulin resistance that is polygenic in origin, as well as a likely monogenic defect in pancreatic beta-cell function, and spontaneously develops diabetes without any dietary manipulation.20 Given the lack of dietary manipulation needed to promote diabetes in this model, microbial alterations associated with the advancement of uncontrolled diabetes are highly likely to be driven by host-related factors that affect luminal microbial metabolism, host tissue metabolism, or a combination of both.7 Thus, the aim of this study was to determine whether the progression of diabetes alters colonocyte metabolism and hypoxia in the UCD-T2DM rat. We hypothesized that advanced diabetes will reduce the colonic abundance of microbial-derived short-chain fatty acids (SCFAs), reduce colonocyte hypoxia, and a downregulation of enzymes associated with fatty acid oxidation.

MethodsAnimals

Age-matched male (174±4 days; mean±SD) UCD-T2DM rats were selected for inclusion into this study. The polygenic background of the UCD-T2DM rat model does not allow a genetically matched control animal for the UCD-T2DM rat; therefore, age-matched lean male Sprague-Dawley (LSD) rats were included for comparisons with a metabolically healthy control group (LSD; 176±4 days; mean±SD; n=15). Rats were selected from the breeding colony in the Department of Nutrition at UCD and maintained under ad libitum feeding conditions (2018 Teklad Global; Harlan Laboratories). The colony is maintained at 22°C and 50% humidity with 14:10 hours of light-dark cycle. Rats were singly housed in polycarbonate cages with TEK-Fresh 7099 bedding (Inotiv, West Lafayette, Indiana, USA). Diabetes diagnosis in this model was defined as non-fasting blood glucose >300 mg/dL collected from the tail vein. Rats were selected for inclusion in the study if (1) pre-diabetic (PD, n=15), (2) had recent diagnosis of diabetes (RD, n=12, 24±6 days of diabetes), or (3) were 3 months post-diagnosis of diabetes (D3M, n=12, 92±10 days of diabetes).

Biospecimen collection

Following a 13-hour fast (19:00–08:00), pentobarbital sodium (200 mg/kg) was administered intraperitoneally and euthanasia was performed by a cardiac puncture and exsanguination. Immediately after sacrifice, the gastrointestinal tract was removed and placed on a stainless steel tray on ice. Two-centimeter sections of ileum, proximal, and distal colon were voided of contents and rinsed in phosphate-buffered saline (PBS) solution, then placed in tissue cassettes with foam inserts, and submerged in 10% formalin (Fisher SF100-4) for 24 hours at 4°C. Cassettes were then transferred to 70% ethanol at 4° until histomorphometric and pimonidazole analyses were conducted. Contents from ileum, proximal and distal colon were collected into sterile cryotubes and flash frozen in liquid nitrogen. Tissues from ileum, proximal and distal colon were voided, had the lumen cut open, rinsed with PBS, and flash frozen in liquid nitrogen. Intestinal content and tissue samples were stored at −80°C until analyzed.

Colonic hypoxia

Pimonidazole was used to detect in vivo colonic hypoxia levels due to its ability to form adducts with thiol groups on proteins that are exposed to low oxygen levels.21 One hour prior to euthanization, animals were given an intraperitoneal injection of 60 mg/mL pimonidazole HCl (Hypoxyprobe, Burlington, Massachusetts, USA) at 1 mL/kg body per the manufacture’s guideline. Briefly, paraffin-embedded sections were deparaffinized and rehydrated in xylene and graded alcohols, followed by antigen retrieval in Citra Plus (30 min in 90° water bath). Sections were cooled (20 min), rinsed, and then quenched for 30 min at room temperature with Bloxall. Following peroxidase quenching, sections were incubated for 30 min at room temperature with Animal Free Blocker (Vector Laboratories). Sections were then rinsed with Dulbecco’s PBS and incubated with a mouse-derived monoclonal anti-pimonidazole antibody (1:50 dilution, MAb-1, Hypoxyprobe) overnight at 4°C. Sections were then rinsed with PBS and incubated with donkey anti-mouse Alexa 594 (Jackson ImmunoResearch Laboratories, West Grove, Pennsylvania, USA) for 1 hour at room temperature. After rinsing with PBS for 15 min, sections were coverslipped with Prolong Glass Antifade Mountant with Nuc Blue (Invitrogen). Images were acquired with a Nikon Eclipse Ti2 inverted microscope (Nikon Instruments, Melville, New York, USA). Fluorescence intensity was measured using NIS Elements Imaging Software (V.5.21.02) along the lumen-to-serosa axis. A total of 10 measurements were acquired per slide and exported to R (V.4.1.0). Using R, the outer edge of the luminal epithelial cells was identified and data were captured for 100 µm from the luminal edge toward the serosa, following the oxygen gradient described by Kelly et al.14 Data were then divided into 10 10 µm sections to measure the oxygen gradient and then averaged across each of the 10 measurements as described by Kelly et al.14

Histology

Formalin-fixed tissues were embedded in paraffin and then processed for staining with H&E. Tissue sections were examined at 40× magnification using a Nikon Eclipse Ti2 microscope and captured using Nikon NIS Elements software (V.5.21.02). Ileal villus height, crypt depth, and colon gland length were measured using ImageScope (V.12.1.0.509; Asperio Technologies) by a board-certified veterinary pathologist (TL) blinded to treatment groups. At least 10 gland and crypts adjacent to intact villi were selected for measurement. The presence of increased numbers of lymphocytes and eosinophilic granular leukocytes was scored as 0=normal or 1=increased.

Real-time PCR

RNA from whole colon and ileum tissue samples was extracted using TRIzol reagent and isolated with RNAeasy mini-columns (Qiagen, Germantown, Maryland, USA). Total RNA (1 µg) was reverse transcribed using the iScript cDNA synthesis kit (Bio-Rad, Hercules, California, USA). Real-time PCR (rtPCR) was accomplished using an ABI Prism 7500 Fast instrument. Relative amounts of mRNA were quantified using the Pfaffl method to account for differences in primer efficiencies.22 Hprt1 was used as a reference gene for data normalization. Gene-specific primers were designed with Primer Express Software (online supplemental table 1).

Bacterial taxonomy

Community profiling of colonic and ileal bacteria was conducted with 16S rRNA gene amplicon sequencing as previously described.8 23 DNA from colon and ileal contents (30–150 mg) was extracted using the QIAmp Fast DNA Stool Kit (Qiagen) following the manufacturer’s instructions. Amplification of the V4 variable region of the 16S gene using extracted DNA (50 ng) was done with 515F/806R forward and reverse primers by PCR.23 Paired-end sequencing (2×250 bp) of the pooled and purified library was done with an Illumina MiSeq with 20% PhiX DNA. Demultiplexed sequence files were downloaded from BaseSpace as fastq files and processed using QIIME2.24 Filtering, denoising, and merging of forward and reverse reads were completed using DADA2.25 Default settings were used for DADA2 except for pseudo and pooled arguments for pooling and chimera methods. A SILVA26 naïve Bayes classifier, the silva-138-99-515-806-nb-classifier, was prepared for taxonomic assignment using RESCRIPt.27 Phylogenetic trees were constructed using the RAxML rapid bootstrap procedure.28 Raw fastq files are publicly available at the National Center for Biotechnology Information Sequencing Read Archive, accession number PRJNA984264.

Short-chain fatty acids

Isobutyric acid, butyric acid, 2-methylbutyric acid, isovaleric acid, valeric acid, 3-methylvaleric acid, isocaproic acid, caproic acid, heptanoic acid, octanoic acid, pyridine, 3-nitrophenylhydrazine hydrochloride (3-NPH), and N-(3-dimethylaminopropyl)-N’-ethylcarbodiimide hydrochloride (EDC) were purchased from Sigma Aldrich (St Louis, Missouri, USA). All isotopically labeled compounds used as internal standards (ISTD) were obtained from Cambridge Isotope Laboratories (Tewksbury, Massachusetts, USA). Acetic acid, propionic acid, and optima grade solvents were from Fisher Scientific (Pittsburgh, Pennsylvania, USA).

Colon content SCFAs were measured as previously described with minor modifications.29 All stock standards (1 mg/mL) were prepared in acetonitrile. Working calibration standards (0–100 µg/mL) were diluted in 50% aqueous acetonitrile and spiked with ISTD mix (500 ng/mL). Approximately 70 mg colon content samples were dried in a vacuum concentrator for at least 4 hours to remove water content. Dried samples (~17 mg) were homogenized in 500 µL of cold isopropanol spiked with 500 ng/mL ISTD using a Precellys 24 homogenizer at 5300 rpm for two cycles of 30 s. Samples were cooled on dry ice for 2 min and homogenization repeated. Homogenates (~34 mg/mL) precipitated on ice for at least 10 min before centrifuging at 18 000 g for 10 min at 4°C. Supernatant or calibration standard (100 µL) was aliquoted to a 2 mL amber glass vial and mixed with 50 µL of 200 mM 3-NPH in 50% acetonitrile and 50 µL of 120 mM EDC with 6% pyridine in 50% acetonitrile. Vials were incubated at 37°C for 30 min, then diluted with 0.1% formic acid in water to a final volume of 400 µL. The derivatized solution was centrifuged at 10 000×g for 5 min at 4°C prior to injection and a pool from 20 µL of each derivatized sample was made for quality control (QC).

Chromatographic separation was performed on an ACQUITY Premier UPLC (Waters Corporation, Milford, Massachusetts, USA) fitted with an HSS T3 C18 column (100×2.1 mm, 1.8 µm) kept at 40°C, while samples were kept at 4°C. A flow rate of 400 µL/min and injection volume of 5 µL were used. Mobile phases consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) with a 20 min elution gradient as follows: hold 10% B 1 min, ramp to 90% B over 15 min, hold 90% B 2 min, return to 10% B in 0.1 min, hold 10% B until 20 min.

Quantification was carried out on a Select Series Time of Flight mass spectrometer with data acquisition and analysis performed using MassLynx V.4.2 and UNIFI V.1.9.13 software, respectively. MSe data were acquired in negative electrospray ionization mode with source tuning parameters as follows: capillary: 1.50 kV; cone: 20 V; source offset: 10 V; source temperature: 120°C; desolvation temperature: 550°C; cone gas: 0 L/hour; desolvation gas: 800 L/hour; nebulizer gas: 6 bar; reference capillary: 3.00 kV. A 10-point calibration curve (0–100 000 ng/mL) was made for quantification of the SCFAs. A pooled QC sample was used in every 10 sample injections. The final calculated concentration was further normalized by dry weight and the data were expressed as ng of metabolite per mg of feces.

Statistical analysis

All statistical analyses were conducted in the R Statistical Language (V.4.1.0) and statistical significance was determined at alpha of <0.05. Physical characteristics, colon/ileum crypt depth, and ileum villi heights were assessed with analysis of variance (ANOVA) followed by Tukey’s honest significance difference test. Microbiome count data were filtered to remove low-abundant amplicon sequence variants (ASVs) by the cut-off of ASVs that had <25% of samples with ≤5 counts. Alpha diversity was estimated using total ASV counts, Faith’s phylogenetic diversity, and the Shannon index. Group differences in alpha diversity estimates were assessed with analysis of covariance with sample depth as a covariate. Differences in beta diversity were assessed with permutation multivariate ANOVA (PERMANOVA) using Aitchison distances. Sample depth was also included in PERMANOVA models. Differential ASVs were determined with the ZicoSeq pipeline from the GUniFrac R package.30 31 Pimonidazole staining intensity was analyzed by linear mixed models with experimental group and luminal distance as main effects and rat ID as a random effect. An interaction term between group and luminal distance was included in the linear mixed model. Quantitative rtPCR (rt-qPCR) genes and SCFAs were assessed for group differences using Kruskal-Wallis tests followed by Dunn’s test of multiple comparisons using rank sums. Results were adjusted for multiple comparisons using Benjamini and Hochberg false discovery rate (FDR) correction32 and considered statistically significant at FDR <0.05. Supplemental materials, data, and R code can be accessed at the figshare repository.33

ResultsPhysical characteristics

Total body weight of PD and RD UCD-T2DM rats was >260 g greater, on average, than LSD rats (online supplemental figure 1A). UCD-T2DM D3M rats had a significant reduction in total body weight relative to PD and RD UCD-T2DM rats, but were still ~145 g heavier compared with LSD rats. Per cent HbA1c did not differ between LSD and PD UCD-T2DM rats; however, RD and D3M groups had greater HbA1c percentages compared with LSD rats (online supplemental figure 1B). Per cent HbA1c was greatest in D3M rats, concurrent with the untreated advancement of diabetes.

Microbiome

LSD rats exhibit a distinct colonic microbial community compared with all UCD-T2DM animals, as visualized across principal components 2 and 3 (online supplemental figure 2A, bottom left panels). Using pairwise PERMANOVA, beta diversity estimates statistically differed between LSD rats and all UCD-T2DM groups (online supplemental table 2A). Pairwise PERMANOVA suggests that all UCD-T2DM groups statistically differ from each other even when accounting for multiple comparisons (online supplemental table 2). When assessing microbial community patterns in ileum contents, pairwise PERMANOVAs also showed statistical differences between LSD rats and all UCD-T2DM groups (online supplemental table 2). Similarly, all UCD-T2DM groups differed from one another (online supplemental table 2). Visualization of variation explaining group differences in ileum contents can be identified in the first five principal component analysis components (online supplemental figure 2A).

D3M rats had lower total ASVs and Faith’s phylogenetic diversity estimates relative to all other groups (online supplemental figure 2B). No differences were observed between any other groups for these alpha diversity estimates. Shannon’s index was also lower in D3M rats compared with RD rats, but did not differ to LSD and PD rats. No differences between groups were observed in any alpha diversity measurements from ileum contents.

A total of 28 genera from colon content were differentially abundant across at least one group when including LSD rats in the analysis (online supplemental table 3). However, only 14 genera from colon contents were different when assessing UCD-T2DM groups only (figure 1A). Of note, reads assigned to the genus Akkermansia were completely absent in D3M rats compared with all other groups. In addition, reads assigned to Enterorhabdus, an uncultured Flavobacteriaceae, and Butyricimonas genera had a reduction in D3M rats also. Phascolarctobacterium, Escherichia–Shigella, Allobaculum, Lactobacillus, and Streptococcus genera had significant higher abundances in D3M rats relative to other experimental groups.

Figure 1

Differential expression of genera in colon (A) and ileum (B) contents from age-matched lean Sprague-Dawley (LSD) rats and University of California Davis type 2 diabetes mellitus rats prior to the onset of diabetes (PD), within 1 month post-onset of diabetes (RD), and 3 months post-onset of diabetes (D3M). Data are presented as proportional abundances. Abundances without a common letter differ statistically by pairwise assessment using ZicoSeq.31 Models were adjusted for sample depth and statistical significance at an adjusted p<0.05. P values were adjusted for multiple comparisons using the false discovery rate correction methods of Benjamini and Hochberg.32

In ileum content, 22 genera statistically differed in at least one group when including LSD rats in the analysis (online supplemental table 4). When restricting the analysis to UCD-T2DM rats only, seven genera were found to be statistically different after correcting for multiple comparisons (figure 1B). Patterns of Phascolarctobacterium and Akkermansia genera in ileum contents mirrored patterns observed in colon contents. D3M rats had significant higher reads assigned to Butyricicoccus, Clostridia VadinBB60 group, and Lachnospiraceae UCG-006 compared with at least one other UCD-T2D group. PD rats also had higher read counts in the Muribaculum genus relative to RD rats.

Histology

Ileal villi height was greater in UCD-T2DM D3M rats compared with LSD and UCD-T2DM PD rats (figure 2A). UCD-T2DM D3M villi heights were also numerically greater in UCD-T2DM D3M rats compared with RD rats, but this was not quite statistically significant (p=0.063; figure 2A). Ileum crypt depth was higher in UCD-T2DM RD rats relative to PD rats only (p=0.035). There were no differences in ileal crypt depth across any other group, including comparisons with LSD rats. Colon crypt depths were greater in all UCD-T2DM rat groups compared with LSD rats (figure 2B); however, no differences were observed within UCD-T2DM groups.

Figure 2

Histomorphometric analysis of ileum (A) and colon (B) epithelia from age-matched lean Sprague-Dawley (LSD) rats and University of California Davis type 2 diabetes mellitus rats prior to the onset of diabetes (PD), within 1 month post-onset of diabetes (RD), and 3 months post-onset of diabetes (D3M). Ten measurements were assessed from a whole tissue slide and then averaged for final estimate. Slides shown in this figure are a section of the whole tissue slide from a single rat. Group differences were assessed by analysis of variance. Means without a common letter differ statistically by Tukey’s honest significance difference test. Statistical significance considered at p<0.05.

No differences in immune pathology from ileal tissue preparations were identified across all observed groups. In contrast to ileal tissue, LSD rats had an increased presence of lymphocyte or granular leukocytes within the colon crypts lamina propria compared with all UCD-T2DM rats (figure 3). Over 50% of colon tissue derived from LSD rats had evidence of increased numbers of lamina propria-derived lymphocytes or granular cells, compared with only 20% in PD rats, 10% in RD rats, and no visual evidence in D3M rats.

Figure 3

Histomorphometric assessment of colonic immune infiltration from age-matched lean Sprague-Dawley (LSD) rats and University of California Davis type 2 diabetes mellitus rats prior to the onset of diabetes (PD), within 1 month post-onset of diabetes (RD), and 3 months post-onset of diabetes (D3M). Presence of lymphocytes or eosinophilic granular leukocytes in colon lamina propia was coded as 1 and absence as 0. Data are the proportion of rats coded as 1 (ie, presence of immune cells). Proportional differences were assessed by Kruskal-Wallis test. Proportions without a common letter differ statistically by Dunn’s test of multiple comparisons using rank sums. Statistical significance considered at p<0.05.

Colonic hypoxia and energy metabolism

Similar to observations by Kelly et al,14 colonic hypoxia in LSD and UCD-T2DM rats is confined to fully differentiated epithelial cells that interact with the luminal environment (figure 4A). The intensity of pimonidazole staining was greatest in cells closest to the lumen and lessened significantly toward the serosal side (figure 4A). However, no difference in pimonidazole staining was found across any rat group. rt-qPCR assessment of genes associated with maintaining cellular hypoxia revealed higher expression of Pparg and Hif1α in all UCD-T2DM rat groups relative to LSD rats. Slc5a8, Acads, Cdh17, Slc16a1, Slc26a3, Ldha, and Muc3 were other genes in which UCD-T2DM rats had higher expression across all groups in comparison with LSD rats (online supplemental table 5). At least one UCD-T2DM rat group had higher expression levels of Nfkb1, Scd1, Cldn3, and Tjp1 relative to LSD rats, while Ckb expression in PD and RD UCD-T2DM rats was lower relative to LSD rats. When assessing differences in UCD-T2DM rats only (online supplemental table 6), differences in four genes were identified (figure 4B). Ckm expression was lower in RD rats compared with PD and D3M rats, Nfkb1 was lower in PD compared with RD and D3M rats, Ldha and Tjp1 were greater in D3M rats relative to PD rats.

Figure 4

Immunofluorescence assessment of colonic hypoxia (A) and real-time quantitative PCR analysis (rt-qPCR) of RNA (B) from age-matched lean Sprague-Dawley (LSD) rats and University of California Davis type 2 diabetes mellitus rats prior to the onset of diabetes (PD), within 1 month post-onset of diabetes (RD), and 3 months post-onset of diabetes (D3M). Pimonidazole adducts were identified by immunohistochemistry on paraffin-embedded sections. Fluorescence intensity of 100 µm sections along the lumen-to-serosa axis was captured using NIS Elements Imaging Software. Ten measurements were collected per tissue slide and averaged over 10 µm distances from the luminal side. Differences in experimental groups and 10 µm sections were assessed with linear mixed models with rat ID as a random effect. A representative slide of each experimental group is inlaid into the boxplot panel. rt-qPCR data are relative gene expression ratio, calculated using the Pfaffl method to account for differences in primer efficiencies. PD rats were considered as the control group. Median fold differences were assessed using Kruskal-Wallis tests. Median fold differences without a common letter differ statistically by Dunn’s test of multiple comparisons using rank sums. Statistical significance considered at p<0.05.

Short-chain fatty acids

Colon content concentration of butyric acid was greater in UCD-T2DM D3M rats compared with UCD-T2DM PD rats (p<0.01); however, no other differences were observed across any experimental group (figure 5). Butyric acid levels displayed a linear increase in PD, RD, and D3M group based on linear contrast analysis. No difference across any group was observed for colon content concentrations of acetate, propionate, and lactic acid (online supplemental table 7). Of the minor SCFAs, isocaproic acid concentrations were greater in UCD-T2DM D3M rats relative to PD and RD UCD-T2DM rats (figure 5). No other differences were observed for isocaproic acid or any other minor SCFA (online supplemental table 7).

Figure 5

Colon content concentrations of butyric acid and isocaproic acid in age-matched lean Sprague-Dawley (LSD) rats and University of California Davis type 2 diabetes mellitus rats prior to the onset of diabetes (PD), within 1 month post-onset of diabetes (RD), and 3 months post-onset of diabetes (D3M). Butyric acid and isocaproic acid concentrations were determined by liquid chromatography-mass spectrometry. Group differences were assessed by Kruskal-Wallis test. Median concentrations without a common letter differ statistically by Dunn’s test of multiple comparisons using rank sums. Statistical significance considered at an adjusted p<0.05. P values were adjusted for multiple comparisons using the false discovery rate correction methods of Benjamini and Hochberg.32

Discussion

The UCD-T2DM rat model offers a well-controlled model to identify gut microbial changes associated with progressive deterioration of metabolic glucose control. Indeed, UCD-T2DM rats are not currently available commercially and the only breeding colony has been exclusively maintained by a single laboratory at UCD for >15 years. In addition, the UCD-T2DM rat colony is maintained on standard rat chow, which is provided throughout the lifespan and was continued throughout the current study. Thus, variation attributed to husbandry, housing, and diet is minimized in this model, suggesting that previously observed alterations in cecal and colon microbiota populations are likely driven by host mechanisms. Similar to our previous observations, male UCD-T2DM rats in the current study showed compositional differences in the lower gut microbiota concurrent with the advancement of diabetes, demonstrating consistent and reproducible effects on the lower gut microbiota in this model. Using shotgun metagenomics sequencing, we previously reported that D3M UCD-T2DM rats had an increase in species within the Bacteroidota (Bacteroidetes) and a decrease in species within the Firmicutes compared with RD and PD rats, which was not replicated in this study and others.8 This discrepancy may be related to differences in sequencing and bioinformatics pipelines (amplicon vs whole genome profiling); however, we cannot rule out generational drift in the microbiome. Still, comparisons between amplicon-based sequencing studies indicate a consistent reduction of ASVs classified to the Akkermansia genus and an increase in ASVs classified to the Phascolarctobacterium in D3M UCD-T2DM rats compared with RD rats. Given A. muciniphila purported ability to improve gut barrier function and glucose regulation in rodent models of obesity and diabetes,34 35 the withdrawal of reads assigned to the Akkermansia genus in UCD-T2DM rats with advanced diabetes further indicates a strong relationship between host energy regulation and bacteria within the Akkermansia genus. Conversely, Phascolarctobacterium has been described as a commensal in the human gut36 and was reduced in streptozocin (5 mM)-treated Sprague-Dawley rats.37 Higher abundances of Phascolarctobacterium have also been associated with the tolerance of metformin administration (1700 mg/day) in adults with diabetes.38 Thus, the elevation of this commensal in UCD-T2DM rats with advanced diabetes is unexpected. However, abundances of Phascolarctobacterium were similar in D3M and PD, suggesting that the abundances of this genus are not specifically related to the metabolic health of the host.

Escherichia–Shigella and Streptococcus genera were elevated in RD and D3M UCD-T2DM rats compared with PD rats. Bacterial species derived from these genera are facultative anaerobes and common pathogens. Recent studies have proposed mechanisms by which colonocyte energy metabolism can influence luminal oxygen availability and, in turn, produce a luminal environment beneficial for facultative anaerobes.9 Given the increase in genera known to be facultative anaerobes in UCD-T2DM rats with diabetes, we hypothesized that colonocyte hypoxia would be decreased in D3M rats compared with UCD-T2DM rats at an earlier stage of diabetes development. However, we found no difference in colonocyte hypoxia using immunohistochemistry to measure the fluorescence intensity of pimonidazole-protein adducts in hypoxic cells,39 suggesting that the increase in reads assigned to facultative anaerobes is not the result of increased availability of luminal oxygen. Using C57BL/6 mice, Byndloss et al noted that an increase in either E. coli or S. enterica serovar Typhimurium strains, and a reduction in epithelial hypoxia required concurrent inactivation of PPARγ signaling and depletion of regulator T cells.11 We did not observe any differences in gene expression levels of PPARγ; however, we did not measure protein or activity level of this enzyme. We also did not measure immune cell populations directly in UCD-T2DM rats; however, microscopy analysis suggested a withdrawal of immune cells in rats with advanced diabetes. This concurs with several reports of reduced intestinal T helper 17 cells in ob/ob mice40 and in C57BL/6 mice with high fat-induced obesity41; however, further assessments should be conducted on UCD-T2DM rats to confirm the microscopy data. The failure to associate the advancement of diabetes with pimonidazole intensity combined with the lack of transcriptional changes in enzymes associated with PPARγ activity suggests that alternative mechanisms that influence microbial populations are likely occurring in UCD-T2DM rats with advanced diabetes.

Increasing colonic butyrate levels has been routinely shown to modulate obesity and insulin resistance through several mechanisms in high fat-induced obesity models.42–45 As the majority of literature points to a therapeutic role of butyrate for diabetic outcomes, we anticipated that colon content levels of butyrate would decrease with the progression of diabetes in UCD-T2DM rats. Contrary to our hypothesis, colon content levels of butyrate were elevated in D3M UCD-T2DM rats compared with PD and RD UCD-T2DM rats. This finding is aligned with early work in the ob/ob mouse, which found greater cecal abundances of butyrate in these mice compared with their lean genetic counterparts.46 Both the ob/ob mouse strain and the UCD-T2DM rat model have a genetic origin of their metabolic phenotype, which may explain why these diet-independent models have higher intestinal levels of butyrate compared with diet-induced obesity models. Although data in humans are inconsistent, several studies have observed higher fecal SCFA concentration in overweight and obese individuals compared with their lean counterparts.47–49 Due to the lack of robust changes in bacteria known to ferment carbohydrates to butyrate (eg, Clostridium-related bacteria), we suspect that the increase in butyrate levels in colon contents is likely related to alterations in epithelial uptake. Inflammation is a possible mechanism that could regulate SCFA uptake and trafficking. Indeed, a reduction in SCFA transportation has been noted in cases of active inflammatory bowel disease and several inflammatory cytokines reduce SCFA transporters in culture.50–53 While we did not observe differences in RNA abundances of SLC16A1 (ie, MCT1) in rats with advanced diabetes, that does not preclude changes in protein levels and/or other mechanisms associated with SCFA absorption and transportation. Overall, it is increasingly clear that hyperglycemia influences gastrointestinal physiology and diet-independent models should be considered to distinguish diabetes-specific mechanisms that influence SCFA uptake and metabolism.

The UCD-T2DM rat has several phenotypic features of diabetes that more closely align with type 2 diabetes progression in humans compared with other rodent models of diabetes, for example, polygenic origin of diabetes, intact leptin signaling, and fertility in female rats. While these features make the UCD-T2DM rat model an important model to evaluate therapeutic and preventative strategies for T2D, it has several limitations in relation to the gut microbiome. UCD-T2DM rats, like other murine models, have an enlarged cecum and morphological differences in the large intestine compared with the human system.54 Thus, human microbial community structure and function may not be exactly recapitulated in this model. Rats are also coprophagic and extract additional nutrients from their feces,55 which further highlights the behavioral and function differences between both species. Our hypothesis of reduced colonocyte hypoxia in advanced diabetes is dependent on butyrate oxidation; however, we did not directly measure butyrate oxidation. Thus, although we observed an increase in colon content levels of butyrate with advanced diabetes, we do not know whether there was a change in butyrate oxidation in vivo. We also measured gene expression levels of enzymes associated with butyrate transport and oxidation; however, this does not indicate whether the activity of these enzymes has been altered during diabetes. In addition to the physiological limitations, our investigation was observational in nature. It is still not known whether the alterations in the gut microbiota during the advancement of diabetes in this model are a secondary feature of worsening metabolic regulation or further contribute to the diabetic phenotype. Future studies are necessary to tease this apart.

In summary, several facultative anaerobic bacteria appeared to increase in UCD-T2DM rats with advanced diabetes; however, colonocyte hypoxia remained unaffected. The advancement of untreated diabetes in UCD-T2DM alters microbial populations in both ileal and colonic content, which is a consistent finding in this diet-independent and polygenic model of diabetes. In the absence of dietary modifications of the gut microbiome, yet to be identified factors associated with the degradation of host metabolic health are likely influencing gastrointestinal physiology and/or the luminal environment. Additional studies are necessary to determine the interconnectivity between host energy regulation and gastrointestinal function.

Data availability statement

Data are available in a public, open access repository. Data are available upon reasonable request. Data and coding will be made available upon reasonable request to the corresponding author.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

All animal protocols were approved by the Institutional Animal Care and Use Committee of the University of California Davis.