Vancomycin-Induced Modulation of Gram-Positive Gut Bacteria and Metabolites Remediates Insulin Resistance in iNOS Knockout Mice

The role of oxidative and nitrosative stress has been implied in both physiology and pathophysiology of metabolic disorders. Inducible nitric oxide synthase (iNOS) has emerged as a crucial regulator of host metabolism and gut microbiota activity. The present study examines the role of the gut microbiome in determining host metabolic functions in the absence of iNOS. Insulin-resistant and dyslipidemic iNOS−/− mice displayed reduced microbial diversity, with a higher relative abundance of Allobaculum and Bifidobacterium, gram-positive bacteria, and altered serum metabolites along with metabolic dysregulation. Vancomycin, which largely depletes gram-positive bacteria, reversed the insulin resistance (IR), dyslipidemia, and related metabolic anomalies in iNOS−/− mice. Such improvements in metabolic markers were accompanied by alterations in the expression of genes involved in fatty acid synthesis in the liver and adipose tissue, lipid uptake in adipose tissue, and lipid efflux in the liver and intestine tissue. The rescue of IR in vancomycin-treated iNOS−/− mice was accompanied with the changes in select serum metabolites such as 10-hydroxydecanoate, indole-3-ethanol, allantoin, hippurate, sebacic acid, aminoadipate, and ophthalmate, along with improvement in phosphatidylethanolamine to phosphatidylcholine (PE/PC) ratio. In the present study, we demonstrate that vancomycin-mediated depletion of gram-positive bacteria in iNOS−/− mice reversed the metabolic perturbations, dyslipidemia, and insulin resistance.


INTRODUCTION
Type 2 diabetes is a multitudinous metabolic disorder that arises from a complex interaction among genetic and environmental elements including dysregulated microbiota composition and function (Ahlqvist et al., 2018). Insulin resistance (IR) or dyslipidemia is a key attribute of obesity and diabetes due to metabolic disruptions (Ormazabal et al., 2018). Gut microbiota dysbiosis, specifically changes in Firmicutes to Bacteroidetes ratio, has been linked to a cluster of chronic metabolic diseases including IR, diabetes, and obesity in both humans and mice (Eid et al., 2017). Prebiotics (such as inulin), probiotics, and other microbiota-targeted bacteriotherapy including antibiotics interventions are actively being explored to prevent and manage metabolic disorders (Shapiro et al., 2017). Treatment with antibiotics reduces bacterial diversity in the host with reduced resistance to colonization of non-autochthonous microbes (Sullivan, 2001;Dethlefsen et al., 2008). Thus, antibiotics add an interesting dynamism to the host-microbiome association (Jernberg et al., 2010). Antibiotic-induced depletion of the short-chain fatty acid (SCFA)-producing gut microbes reduced the levels of microbialderived colonic SCFA in children and enhanced the likelihood to develop obesity and metabolic syndrome later in life (Korpela et al., 2017;Li et al., 2017a;Wilkins and Reimer, 2021). On the contrary, antibiotic-induced microbiota depletion (AIMD) by the antibiotic cocktail or vancomycin provides protection against diet or genetically inherited obesity and glucose dysmetabolism (Carvalho et al., 2012;Hwang et al., 2015;Fujisaka et al., 2016) by reducing inflammation and oxidative stress and improving metabolic homeostasis . Studies in humans have implied both gram-positive and gram-negative bacteria in the metabolic perturbations and obesity (Rawat et al., 2012;Pushpanathan et al., 2016;Radilla-Vaźquez et al., 2016;Liu et al., 2019). These studies thus indicate the strong association of gut microbiota with obesity and related metabolic disorders (Cox and Blaser, 2013).
Nitric oxide (NO) has a pivotal role in the regulation of cardiovascular and metabolic functions. Among the different NOS isoforms, the importance of iNOS has been predominantly investigated in inflammatory and infectious diseases (Wong and Billiar, 1995;Kröncke et al., 1998) but is now emerging as an important metabolic regulator (Perreault and Marette, 2001;Cha et al., 2011;Kanuri et al., 2017;Kakimoto et al., 2019;Pathak et al., 2019;Aggarwal et al., 2020). iNOS-derived NO controls the growth of certain microorganisms as an adaptive response of host defense, while survival of certain microbes also seems to be facilitated (Bogdan, 2015). iNOS −/− mice are protected from LPS-induced IR (Carvalho-Filho et al., 2006) and cardiovascular (House Ii et al., 2012) and endothelial dysfunction (Chauhan et al., 2003) and also displayed reduced inflammatory cytokines (Perreault and Marette, 2001;Cha et al., 2011) with enhanced survival during septicemia (Hollenberg et al., 2000). Moreover, previous studies from our lab (Kanuri et al., 2017;Pathak et al., 2019;Aggarwal et al., 2020) and others (Nakata et al., 2008;Kakimoto et al., 2019) had demonstrated IR and disrupted metabolic homeostasis in iNOS −/− mice fed with chow diet, low-fat diet (LFD), or high-fat diet (HFD) which was reversed by enhancing NO bioavailability via nitrite treatment (Aggarwal et al., 2020), suggesting the importance of redox status in host metabolism. As oxidative stress and changes in gut microbiota could play a role in IR (Cui et al., 2009), iNOS −/− mice might have dysbiosis due to the absence of iNOS-derived NO (Aggarwal et al., 2020). Therefore, our present study explored the contribution of gut microbiota in regulating the metabolic abnormality observed in mice lacking iNOS. We demonstrate that iNOS −/− mice exhibit an altered gut microbial community with enhanced enrichment of gram-positive bacteria. Vancomycin (largely depletes grampositive bacteria)-induced depletion of gut bacteria established that such enrichment of vancomycin-sensitive microbes is a vital contributor to metabolic abnormalities found in iNOS −/− mice.

Body Weight, Body Mass Index, and Food Consumption
Body weight was measured weekly from day 0 to the completion of the study at 4 weeks. The body length was also measured from the nose tip to the base of the tail of each mouse. Body mass index (BMI) was calculated using the formula body weight (g) divided by the square of nose to anus length (cm) as previously described (Aggarwal et al., 2020) at the end of 4 weeks of antibiotic treatment. The weekly food consumption was measured by housing two to three mice per cage and adding the preweighed food pellets to each cage and measuring the food left over on a sensitive weighing balance twice weekly. The average weekly food consumption and the average food intake/ day/mice during the study period were calculated.

Body Composition Analysis
Body composition (fat and lean mass) was analyzed in live, conscious mice by echo MRI (E26-226-RM Echo MRI LLC, USA) allowing limited horizontal and vertical movements by applying radiofrequency pulses at a distinct static magnetic field (Kanuri et al., 2018).

In-Vivo Gut Permeability Assay
FITC-labeled dextran (4 kDa) was used to assess in-vivo intestinal permeability (Thevaranjan et al., 2017). Mice were fasted for 4 h with free access to water. A total of 0.8 mg/ml FITC-dextran tracer was given orally in 200 µl PBS followed by removal of both food and water after the gavage. Blood was collected retro-orbitally after 4 h, the serum was separated, and fluorescence intensity was measured using an excitation wavelength of 493 nm and an emission wavelength of 518 nm.

Serum Biochemistry
Retro-orbital blood was collected from 6-h fasted mice. Estimation of lipids like total cholesterol (TC), triglycerides (TG), low-and high-density lipoproteins (LDL and HDL), and non-esterified fatty acids (NEFA) was performed in the serum using kits (Pathak et al., 2019) (Randox, UK). Insulin was measured using the kit from Crystal Chem, USA. Indices of IR (HOMA-IR) and insulin sensitivity (QUICKI) were calculated from fasting blood glucose and serum insulin as per the formulae used by other investigators (Yokoyama et al., 2003).

Total Nitrite Estimation
The animals were sacrificed to retrieve the tissues (liver, epididymal white adipose tissue, and small intestine). Tissues (liver, eWAT, and intestine, 50 mg) were homogenized in 500 µl of hypotonic TKM buffer [25 mM Tris-HCl (pH 7.4), 2 mM MgCl 2 , 5 mM KCl, and 1% NP-40] followed by sonication. The supernatant was obtained by centrifugation at 15,000g at 4°C for 20 min. Total nitrite (nitrate and nitrite) was estimated in serum (100 ml) and tissue homogenates using Griess reagent by reducing nitrate to nitrite using preactivated cadmium pellets followed by deproteinization in tissue homogenates with 3% trichloroacetic acid (Kanuri et al., 2017).

Tissue Biochemistry
Liver tissue (50 mg) was processed as described previously (Aggarwal et al., 2020) for the estimation of hepatic total cholesterol, triglycerides, and free fatty acids (FFA) using the Randox kit. Briefly, for TC estimation, samples were homogenized on ice in 1 ml hexane:isopropanol mixture (3:2 ratio) followed by centrifugation, supernatant collection, and drying in a CentriVap concentrator (Labconco, USA). Samples were homogenized in 1 ml of 1% Triton X-100 in chloroform on ice, centrifuged, and lower phase dried in CentriVap for FFA estimation. The dried products for TC and FFA estimation were redissolved in ethanol:NP-40 (9:1). For triglyceride estimation, samples were homogenized in 500 µl of 5% NP-40 on ice, boiled for 5 min in a water bath at 80°C-100°C, then cooled and reheated three times, and centrifuged, and the supernatant was collected.

Hepatic Glycogen
Insulin was administered at a dose of 0.6 IU/kg i.p. and the animals were sacrificed after 30 min to collect the liver for insulin-stimulated glycogen estimation along with unstimulated controls in both WT and iNOS −/− mice with or without vancomycin and Abx treatment. A 50-mg liver tissue was homogenized in 500 µl distilled water on ice, boiled for 10 min, and centrifuged to remove insoluble substances followed by supernatant collection and glycogen estimation using the kit from Biovision (K646-100) as per the protocol of the manufacturer.

Alcian Blue (AB) Staining
Formalin-fixed, paraffin-embedded small intestinal and colonic tissue was sectioned into 5 mm thin serial slices and stained with 1% Alcian blue solution (in 3% acetic acid, pH 2.5) for morphological examination of acid mucosubstances and acetic mucins and counterstained with 0.1% nuclear Fast Red solution (Otsuka et al., 2010). The Alcian blue-stained area (%), villi, and crypt lengths were calculated using ImageJ software.

Real-Time PCR
Quantitative gene expression analysis was performed using SYBR Green as described previously (Kanuri et al., 2018). Briefly, total RNA was extracted using the TRIzol reagent followed by cDNA synthesis using RevertAid first-strand cDNA synthesis kit using the protocol of the manufacturer. Real-time PCR was performed using LightCycler 480II Real-Time PCR system (Roche Applied Science, Indianapolis, IN, USA) with primers listed in Table S1. 18S rRNA was used as a reference gene for normalization in liver and adipose tissues and RPL10 in intestinal tissue to calculate the expression of candidate genes. Relative fold change was calculated from mean normalized gene expression between different groups as compared with WT mice.

Sample Preparation
A 100-µl serum was lyophilized and stored at −80°C until further processing. The sample was reconstituted in 200 µl methanol, 50 µl water, and 870 µl methyl tert-butyl ether (MTBE) and vortexed for 1 h to extract the metabolites. Organic and aqueous phase separation was induced by adding 250 µl water and centrifuged at 15,000g for 15 min at 4°C. The upper organic and lower aqueous layers (100 µl each) were vacuum dried in a SpeedVac concentrator and stored until further analysis at −80°C . Samples were reconstituted in 50 µl 15% methanol and kept on ice for 30 min, vortexed for another 30 min, and centrifuged at 15,000g for 15 min at 4°C, and the supernatant was collected and subjected to metabolomic analysis by using the LC-MS platform.

Metabolomics Measurement
The metabolomics data were acquired on the Orbitrap fusion mass spectrometer (Thermo Scientific, USA) equipped with a heated electrospray ionization (HESI) source. Data were acquired on positive and negative modes at 120,000 mass resolution in MS mode and 30,000 resolution in datadependent MS2 scan mode. Spray voltages of 4,000 and 35,000 V were used for the positive and negative modes, respectively. Sheath gas and auxiliary gas was set to 42 and 11, respectively. Mass scan range of 50-1,000 m/z, automatic gain control (AGC) target at 200,000 ions, and maximum injection time of 80 ms for MS and AGC target of 20,000 ions and maximum injection time of 60 ms for MSMS were used. The extracted metabolites were separated on UPLC ultimate 3000 using HSS T3 column (100 × 2.1 mm i.d., 1.7 µm, waters) maintained at 40°C temperature. Mobile phase A was water with 0.1% formic acid and mobile phase B was acetonitrile with 0.1% formic acid. The elution gradient used is as follows: 0 min, 1% B; 1 min, 15% B; 4 min, 35% B; 7 min, 95% B; 9 min, 95% B; 10 min, 1% B; and 14 min, 1% B. The flow rate was 0.3 ml/min and the sample injection volume was 5 µl. The pool quality control (QC) sample was prepared by collecting 10 µl from each sample and was run after every five samples to monitor retention time shift, signal variation, and drift in mass error (Kumar et al., 2020).

Data Processing
All acquired data were processed using the Progenesis QI software (Waters Corporation) using default setting. The untargeted metabolomics workflow of Progenesis QI was used to perform retention time alignment, feature detection, elemental composition prediction, and database search. Identification of the metabolite was done on the basis of an in-house metabolite library with accurate mass, retention time, and fragmentation pattern information match. Additionally, spectral data matching with mzCloud and MassBank for the fragmentation match for the identification of metabolites were also used. Metabolomics data were normalized by sum and Pareto scaled before multivariate analysis. Relative fold change values in metabolite expression analysis were calculated for each treated sample with respect to the untreated time-matched control (WT) for two sets of experiments, and then differential analysis was performed on the pooled fold change data. Fold change values were log transformed for a clearer representation in the heatmap analysis. Statistical analysis was performed by multiple t-tests followed by a two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli to correct for multiple comparisons by controlling the false discovery rate (<0.05).

16S rRNA Gene Sequence Analysis
Genomic DNA from approximately 200 mg mice stool samples was extracted by an optimized enzymatic, chemical, physical, and mechanical lysis method as described previously (Bag et al., 2016). The V3-V4 region of the 16S rRNA gene was amplified and gel extracted for metagenomic sequencing as described previously (Tandon et al., 2018). Metagenome libraries were prepared using V3-V4 region-specific primers. Approximately, 40 ng of DNA samples were amplified for 26 cycles of round 1 PCR using KAPA HiFi Hot-Start PCR Kit (KAPA Biosystems Inc., USA). The forward and reverse primers were used at a concentration of 5 µM each. The amplicons were analyzed on 1.2% agarose gel, and l µl of diluted round 1 PCR amplicons was used for indexing PCR (round 2). Here, the round 1 PCR amplicons were amplified for 10 cycles to add Illumina sequencing barcoded adaptors (Nextera XT v2 Index Kit, Illumina, USA). Round 2 PCR amplicons (sequencing libraries) were analyzed on 1.2% agarose gel. The Illumina adapter sequences were as follows: 5′-AATGATACGGCGAC

CACCGAGATCTACAC[i5]TCGTCGGCAGCGTC and 5′-C A A G C A G A A G A C G G C A T A C G A G A T [ i 7 ]
GTCTCGTGGGCTCGG, where [i5, i7] were unique dual index sequence to identify sample-specific sequencing data. Furthermore, the samples were loaded into an Illumina MiSeq v3 600 cycles flow cell (Illumina, CA, USA) and the sequencing run was performed according to standard Illumina protocol.

Data Analysis
The Illumina paired-end raw reads having V3-V4 primer sequence and high-quality bases (>Q30) were selected. Bcl2fastq software v2.20 was used for data demultiplexing and FastQ files were generated based on the unique dual barcode sequences. FastQC v0.11.8 software was used to assess the sequencing quality. The adapter sequences were trimmed and bases above Q30 were considered. During read preprocessing, low-quality bases were filtered off and then data were used for downstream analysis. The reads were further stitched using Fastq-join. These stitched reads were subjected to QIIME pipeline for microbiome analysis. The query sequences were clustered using the UCLUST method against a curated chimera-free 16S rRNA database, Greengenes v 13.8. The taxonomies were assigned using the RDP classifier to these clusters at ≥97% sequence similarity against the reference Greengenes database. The sequencing stats and absolute read counts are represented in Tables S2, S3, respectively. PCA and alphadiversity analysis were performed using MicrobiomeAnalyst. The raw reads data were normalized and relative abundance in % was calculated. For comparative analysis, minimal data filtering was done and features containing non-zero values in less than~10% samples were removed. Statistical analysis at the levels of phylum, family, and genus was performed by multiple t-tests with the assumption that all rows are sample from populations with the same scatter (SD) followed by a two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli to correct for multiple comparisons by controlling the false discovery rate (<0.05). Due to taxonomic resolution limit of the partial 16S rRNA gene sequencing-based microbiome study, the abundance difference of the bacterial species in WT and insulin-resistant iNOS −/− mice has not been included in the present study.

Statistical Analysis
Data were presented as mean ± SEM. Independent unpaired Student's t-test was used for comparisons as appropriate using the GraphPad Prism 8 software. More than two groups were compared by one-way analysis of variance (ANOVA) followed by post-hoc Tukey's multiple comparison test or Dunnett's test. Differences were considered statistically significant at p <0.05. For the correlation analysis, Pearson correlation coefficients were calculated and p-value was corrected according to the Benjamini-Hochberg correction for multiple comparisons, with a false discovery rate <0.05.

Availability of Data and Materials
All data used in this study are present in the main text and Supplementary Material. The raw 16S rRNA gene sequencing data were deposited in the Sequence Read Archive (SRA) of the National Center for Biotechnology Information (NCBI) under accession number PRJNA740126. However, no significant difference was observed in the Simpson diversity index ( Figures 1A-D). Multidimensional scaling analysis through principal coordinate plot represents that the bacterial communities between the WT and iNOS −/− mice varied significantly ( Figure 1E). At the phylum level, the relative abundance of Firmicutes was decreased with increased Verrucomicrobia. At the family level, Erysipelotrichaceae, Bifidobacteriaceae, and Verrucomicrobiaceae were increased, while Lactobacillaceae and Ruminococcaceae were decreased. Allobaculum, Bifidobacterium, and Akkermansia were increased significantly, while the Lactobacillus genus was reduced in insulin-resistant iNOS −/− mice as compared with WT ( Figure 1F) indicating differential gut microbiome with major changes in gram-positive bacteria. In the PCA score scatter plot from serum metabolomics, distinctly separated clusters between the WT and iNOS −/− mice in ESI-positive mode were seen ( Figure 1G) with significantly differential metabolites (p < 0.05) being visualized through a volcano plot ( Figure 1H). iNOS −/− mice displayed enhanced purine and pyridimidine metabolites, PE lipids, PE to PC ratio, 10-hydroxydecanoate, 3nitrotyrosine, cysteamine, cysteate, carbohydrate metabolites, indole-3-ethanol, diosmetin, and phosphonoacetate. PC, PA, and PS lipids; laurate; lauroyl carnitine; anthranilate; and cystathionine were decreased in iNOS −/− mice suggesting altered metabolic profile as compare with WT ( Figure 1I).

Vancomycin-Induced Modulation of Gut Microbiota Rescues iNOS −/− Mice From Systemic IR and Dyslipidemia
Subsequently, we used vancomycin to deplete the gram-positive and antibiotic cocktail to deplete both gram-positive and gramnegative gut microbiota in iNOS −/− mice to assess the correlation, if any, between gut microbiome and altered metabolic homeostasis in iNOS −/− mice. Vancomycin (Figures 2A-D) and Abx ( Figures S3A-D  were decreased in Abx-treated iNOS −/− mice with increased Brucellaceae in WT. At the genus level, Lactobacillus, Bifidobacterium, and Akkermansia were decreased by Abx treatment in WT and iNOS −/− mice with increased Klebsiella. Abx decreased Allobaculum and increased Serratia in iNOS −/− mice, while it increased Pseudomonas, Elizabethkingia, and Ochrobactrum in WT ( Figures S5H, S3F). Both vancomycin ( Figures 2G-L) and Abx ( Figures S3G-L) reversed the systemic glucose intolerance and IR and enhanced gluconeogenesis in iNOS −/− mice as seen by a decrease in AUC during GTT, ITT, and PTT, respectively. The increase in the circulating level of glucose and insulin in iNOS −/− mice was decreased by both treatments (Figures 2M, N, S3M, N). HOMA-IR was decreased by antibiotic treatment with increased QUICKI reversing the IR phenotype observed in iNOS −/− mice ( Figures 2O, P, S3O, P).
The enhanced circulating TC, TG, LDL, and NEFA in iNOS −/− mice were reversed by the antibiotic treatment, thus rescuing the dyslipidemia in iNOS −/− mice ( Figures 2Q-T, S3Q-T). The body weight ( Figures S6A, B) and food intake ( Figures S6D, E), however, remained unaltered in obese iNOS −/− mice upon vancomycin and Abx treatment. Body composition and BMI also remained unaltered in vancomycin-treated iNOS −/− mice with increased fat mass in Abx-treated iNOS −/− mice ( Figures  S6C, F). Liver, adipose tissue, and heart weight ratio were decreased in vancomycin-and Abx-treated iNOS −/− mice with increased weight and length ratio of the small intestine and cecum ( Figures S6G, H).

Improvement in Disrupted Lipid and Glucose Homeostasis in the Liver in iNOS −/− Mice Following Treatment With Vancomycin
Hepatic TC, TG, and FFA levels were reduced in the vancomycin ( Figures 4A-C) and Abx ( Figures S7A-C)-treated iNOS −/− mice. FFA levels were also decreased in vancomycin-and Abxtreated WT (Figures 4C, S7C). The enhanced expression of lipid synthesis genes (PPARg, SREBP-1c, and ACC1) in the liver of iNOS −/− mice was decreased by vancomycin ( Figure 4F) and Abx ( Figure S7F) with unchanged LXRa, LXRb, and HMGCR in the liver. FAS and SREBP-2 expression was decreased significantly by vancomycin in iNOS −/− mice, but not by Abx. The expression of LXRa, LXRb, and HMGCR was increased in the liver of vancomycin-treated WT ( Figure 4F). PGC-1b expression was decreased in the liver of vancomycin ( Figure 4G) and Abx ( Figure  in iNOS −/− mice following vancomycin ( Figure 4H) and Abx ( Figure S7H) treatment with decreased SR-1B. CD36 and LPL expression was increased in vancomycin-treated WT. LDLR expression was reduced in iNOS −/− mice and was unaltered by vancomycin and Abx. The expression of genes involved in hepatic lipid efflux-ABCG5 and ABCG8-was decreased in iNOS −/− mice and was increased significantly in Abx-treated iNOS −/− mice ( Figure S7I) and were comparable with WT in vancomycin-treated iNOS −/− mice ( Figure 4I). Hepatic PC expression was decreased by vancomycin with decreased G6PC expression by vancomycin ( Figure 4E) and Abx ( Figure S7E) in iNOS −/− mice along with unchanged FOXO1 and PEPCK. Hepatic glycogen levels also remained unaltered upon vancomycin treatment ( Figure 4D), while insulin-stimulated glycogen levels were increased significantly by Abx ( Figure S7D).

Improvement in the Disrupted Lipid and Glucose Homeostasis in Adipose Tissue and Intestine in iNOS −/− Mice Following Treatment With Vancomycin
The enhanced expressions of lipid synthesis genes (PPARg, SREBP-1c, FAS, and ACC1) in adipose tissue of iNOS −/− mice were decreased by vancomycin, while Abx treatment decreased SREBP-1c and FAS. LXRa expression was increased in the eWAT of vancomycin-( Figure 5C) and Abx-treated iNOS −/− mice ( Figure S8C). PGC-1a and UCP2 expressions in adipose tissue were increased in Abx-treated iNOS −/− mice, while PGC-1b, ACC2, and PPARa remain unchanged by vancomycin ( Figure 5D) or Abx ( Figure S8D). The enhanced expression of CD36 and LPL in adipose tissue of iNOS −/− mice was decreased by vancomycin ( Figure 5E) and Abx treatment ( Figure S8E). PEPCK and G6PC were decreased in adipose tissue by vancomycin ( Figure 5A) and Abx ( Figure S8A) in iNOS −/− mice with unchanged FOXO1 and PC. Decreased Akt2 gene expression was increased in white adipose tissue of vancomycintreated WT and iNOS −/− mice suggesting improved insulin signaling upon gram-positive bacteria depletion. The glucose transporters in adipose tissue remained unaltered upon vancomycin ( Figure 5B) or Abx treatment ( Figure S8B).
The morphometric calculations ( Figures S9A, D) displayed increased intestinal villi and colonic crypt length in iNOS −/− mice ( Figures S9C, F). The barrier functionality of the gut remained unchanged in iNOS −/− mice as assessed by in-vivo FITC-dextran gut permeability assay ( Figure S9G) and Alcian blue staining in intestinal ( Figures S9A, B) and colonic tissues ( Figures S9D, E). The gene expression of the tight junction protein Claudin-2 was increased with decreased ZO-1 and unaltered occludin in iNOS −/− mice. Mucins (Muc-2 and Muc-5AC) remained unaltered in iNOS −/− mice with reduced expression of antimicrobial peptide Reg3g ( Figure S9H). The FITC-dextran gut permeability assay showed significant reduction in the intestinal permeability following vancomycin and Abx treatment in WT and iNOS −/− mice as compared with untreated controls ( Figure S9G). The mRNA expression for tight junction proteins, mucins, and antimicrobial peptide in the small intestine largely remained unchanged in antibiotic-treated iNOS −/− mice. Claudin-2 mRNA expression was increased in Abx-treated WT and iNOS −/− mice ( Figure S9H). Furthermore, the status of NO was checked and the decreased total nitrite levels in iNOS −/− mice were not altered in the serum, adipose tissue, and intestine after treatment with antibiotics. In the liver, the total nitrite levels were further decreased in iNOS −/− mice by antibiotics ( Figures S9I-L). nNOS expression remained unchanged in the liver and intestine and was decreased in adipose tissue upon treatment with antibiotics in iNOS −/− mice. Decreased eNOS expression was rescued in the adipose tissue and intestine by antibiotics and remained unaltered in the liver ( Figures S9M-O).

Association of Serum Metabolites With Metabolic Profile of iNOS −/− Mice Upon Treatment With Antibiotics
The phenotypic, biochemical, functional, metabolic, and molecular analyses suggest that iNOS −/− mice displayed IR and dyslipidemia along with altered gut microbiome as compared with WT mice. Gut microbiota depletion/modulation by antibiotics showed marked improvement in metabolic parameters with precise alterations in the serum metabolites. Pearson's correlation was used to identify the serum metabolites which strongly correlated with the metabolic biomarkers quantified in control (WT), insulin-resistant (iNOS −/− ), and antibiotics (vancomycin or Abx)-treated WT and iNOS −/− mice. The lipid species MGDGs, PEs, PAs, PSs, PGs, ceramides, and 10-hydroxydecanoate positively correlated with bacterial diversity, dyslipidemia, and IR and negatively with cecum weight. Laurate and palmitate were negatively correlated liver lipids and also negatively correlated with bacterial diversity ( Figure S10A). Majority of amino acids positively correlated with biomarkers of IR, dyslipidemia, and a-diversity and negatively correlated with HDL levels and cecum weight ( Figure S10B). Indole-3-methyl acetate, 5-oxoproline, 2-methylhippuric acid, hippurate, indole-3-ethanol, ophthalmate, mevalonate, 6carboxyhexanoate, sebacic acid, 4-chlorophenol, d-undecalactone, and aspartame positively correlated with IR, dyslipidemia, and adiversity and negatively correlated with cecum weight. Scarboxymethylcysteine and formate were negatively correlated with dyslipidemia and a-diversity ( Figure S10C). Nucleic acid metabolites-methylthioadenosine, xanthine, allantoin, thymine, cytosine, and uracil-positively correlated with biomarkers of IR and dyslipidemia and negatively correlated with cecum weight. Glycolysis and Krebs cycle intermediates were correlated positively and glyceraldehyde negatively with liver lipids and a-diversity. 4-Pyridoxate and melatonin were positively correlated with glucose intolerance and dyslipidemia. Bile acids and bile pigments were positively correlated with IR and dyslipidemia and negatively correlated with cecum weight ( Figure S10D). These results suggest that serum metabolites exhibited strong association with specific metabolic biomarkers and were modulated by microbiota depletion by antibiotics.

Association of Gut Microbiota With Metabolic Parameters in iNOS −/− Mice Following Treatment With Antibiotics
Furthermore, the association between antibiotic-induced compositional changes in microbiota with the alterations in metabolic parameters was investigated. The phyla Bacteroidetes and Actinobacteria were positively associated with IR, dyslipidemia, and a-diversity, while Proteobacteria correlated negatively. Firmicutes and Verrucomicrobia were negatively correlated with cecum weight and positively correlated with dyslipidemia ( Figure 6A). Bifidobacteriaceae (Actinobacteria), Erysipelotrichaceae (Firmicutes), and S24-7 (Bacteroidetes) were positively correlated with IR, dyslipidemia, and a-diversity and negatively with cecum weight. The microbial families-Ruminococcaceae, Eubacteriaceae, Lachnospiraceae, Dehalobacteriaceae, Mogibacteriaceae, and Lactobacillaceae (Firmicutes); Coriobacteriaceae (Actinobacteria); and Verrucomicrobiaceae (Verrucomicrobia)-were positively correlated with dyslipidemia and a-diversity and negatively correlated with cecum weight. Weeksellaceae and Sphingobacteriaceae (Bacteroidetes); Enterococcaceae, Bacillaceae, Tissierellaceae, and Leuconostocaceae (Firmicutes); Microbacteriaceae, Micrococcaceae, Gordoniaceae, Nocardiaceae, Nocardiodaceae, Promicromonosporaceae, Beutenbergiaceae, Streptomycetaceae, Pseudonocardiaceae, Brevibacteriaceae, and Geodermatophilaceae (Actinobacteria); a n d m a n y f a m i l i e s o f P r o t e o b a c t e r i a i n c l u d i n g Enterobacteriaceae were negatively correlated with serum NEFA and positively correlated with cecum weight ( Figure S11).
These results suggest that gram-positive bacteria depletion (Allobaculum, Bifidobacterium, and Ruminococcus) following treatment with vancomycin is directly correlated with the improvement in metabolic biomarkers.

Gut Microbiota Association With Serum Metabolites in iNOS −/− Mice Upon Antibiotics Treatment
Pearson's correlation coefficients were calculated between serum metabolites and microbial taxa, which positively correlated with blood glucose and glucose intolerance in WT and iNOS −/− mice without or with antibiotics (vancomycin or Abx) treatment to gain better insights into the host metabolism. Interestingly, most of these microbial taxa linked to impaired glucose metabolism are gram-positive bacteria. The lipid species MGDG, PE, PC, PA, PS, PG, ceramide, 10-hydroxydecanoate, dodecanedioic acid, and myristate were positively correlated and palmitate was negatively correlated with taxa directly associated with glucose intolerance (Actinobacteria, Bacteroidetes, Erysipelotrichaceae, Bifidobacteriaceae, S24-7, Ruminococcaceae, Allobaculum, Bifidobacterium, Ruminococcus) ( Figure S12A). Most amino acid metabolites including anthranilate, histidine, carnosine, serine, N-formylmethionine, N-acetyl cysteine, N-acetyl norvaline, carnitine, aminoadipate, pipecolinic acid, methyl A B FIGURE 6 | Association of gut microbiota with metabolic parameters in iNOS −/− mice following treatment with antibiotics. Heatmap of Pearson's correlation coefficients between changes in different metabolic parameters and microbial taxa at the (A) phylum and (B) genus levels caused by gut microbiota modulation by vancomycin and Abx in WT and iNOS −/− mice. *p < 0.05, **p < 0.01, ***p < 0.001 represent significant correlations between metabolic biomarker and bacterial taxa. Blue color represents negative and red positive correlations. See also Figure S11. glutarate, glutamate, N-acetyl ornithine, betaine, creatine, creatinine, proline, leucyl proline, and 3-amino-4hydroxybutyric acid were positively correlated with taxa associated with glucose intolerance ( Figure S12B).

DISCUSSION
Involvement of both gram-positive and gram-negative bacteria has been reported in diabetes and metabolic syndrome; however, most of the studies on gram-negative bacteria linked LPSmediated inflammation with metabolic dysregulation (Rawat et al., 2012;Pushpanathan et al., 2016;Radilla-Vaźquez et al., 2016;Liu et al., 2019). iNOS −/− mice are, however, protected against LPS-induced inflammation and the associated metabolic perturbations (Chauhan et al., 2003;Carvalho-Filho et al., 2006;House et al., 2012). iNOS −/− mice were insulin resistant and dyslipidemic on chow, LFD, and HFD diets, which was improved substantially by nitrite supplementation (Kanuri et al., 2017;Aggarwal et al., 2019;Pathak et al., 2019;Aggarwal et al., 2020). To investigate further the role of gut bacteria in metabolic perturbations in iNOS −/− mice, we checked the microbial diversity in fecal samples. Reduced bacterial diversity in iNOS − / − mice correlated with less diverse microbial communities in obese subjects (Turnbaugh et al., 2009). Similar to that reported in ileum (Matziouridou et al., 2017), we observed increased relative abundance of Bifidobacterium and Allobaculum in the feces belonging to the phyla Actinobacteria and Firmicutes, respectively, in iNOS −/− mice. Actinobacteria are augmented in obese subjects and in patients with liver diseases and diabetes (Turnbaugh et al., 2009). Allobaculum, the efficient energy harvesters and active assimilators of glucose (Herrmann et al., 2017), was enhanced in LFD-and HFD-fed mice (Ravussin et al., 2012), during metabolic dysbiosis (Nobel et al., 2015), and in diabetic ZDF rats (Zhou et al., 2019). Allobaculum is a secondary degrader and seems to be dependent on primary degraders like Bifidobacterium for the utilization of complex polysaccharides for their growth. We and others (Kim et al., 2017;Zhou et al., 2019) did find a positive correlation between Bifidobacterium, Ruminococcus, and Allobaculum in the hyperglycemic condition. Gut barrier dysfunction and increased availability of microbial products in the systemic circulation, commonly known as metabolic endotoxemia, are considered as key drivers of obesity and metabolic syndrome (Cani et al., 2007). Surprisingly, iNOS-deficient mice did not display alterations in gut permeability, possibly due to an increase in the relative abundance of Akkermansia muciniphila, a gram-negative bacteria that is shown to maintain and restore mucosal barrier integrity in obese and diabetic mice and in inflammatory conditions (Everard et al., 2013;Bian et al., 2019) and is reported to be a beneficial microbe in maintaining metabolic health. However, despite its beneficial effects, we observed metabolic disruptions in iNOS −/− mice. Intestinal villi and colon crypt length were also increased in iNOS −/− mice, and this could be an adaptation for improved absorption of nutrients under hyperglycemic conditions (Isah and Masola, 2017).
As iNOS −/− mice exhibited a higher abundance of the grampositive bacteria Allobaculum and Bifidobacterium, we employed vancomycin, which largely depletes gram-positive bacteria to study their association with IR and dyslipidemia. The vancomycin intervention altered the gut microbiome profiles as reported earlier (Tulstrup et al., 2015;Ray et al., 2021) in both WT and iNOS −/− mice with decreased bacterial alpha-diversity and increased relative abundance of Proteobacteria (Ray et al., 2021), the signature markers of antibiotic treatment. Proteobacteria exhibit metabolic plasticity (Stecher et al., 2012) and are reservoirs of several horizontally acquired antimicrobial resistance genes (Jiang et al., 2017) leading to their expansion after treatment with antibiotics. Intriguingly, vancomycininduced reduction of Allobaculum, Bifidobacterium, and Ruminococcus was linked with improved IR and dyslipidemia in iNOS −/− mice. Antibiotics strengthen the intestinal integrity (Tulstrup et al., 2015;Fujisaka et al., 2016) with no distinguishable change in mucins and tight junction proteins (Tulstrup et al., 2015;Ray et al., 2021) as was also observed in iNOS −/− mice. The depletion of microbiota by treatment with antibiotics in WT (Rodrigues et al., 2017;Kuno et al., 2018;Zarrinpar et al., 2018) and in diet-induced obese mice Membrez et al., 2008;Carvalho et al., 2012;Hwang et al., 2015) improved glucose tolerance and IR as found by us in iNOS −/− mice. Veillonella has recently been reported to improve cardiovascular health and exercise performance by utilizing lactic acid and converting it into propionate (Scheiman et al., 2019) which might have a symbiotic relationship with the lactic acidproducing Lactobacillus. This might have been the reason for the expansion of both Veillonella and Lactobacillus in vancomycintreated iNOS −/− mice. The susceptibility of Lactobacilli to vancomycin varies as Lactobacillus acidophilus and Lact. delbreuckii are vancomycin sensitive, while Lact. rhamnosus are vancomycin resistant (Hamilton-Miller and Shah, 1998). We also used a broad-spectrum antibiotic cocktail to deplete the majority of gut gram-positive and gram-negative bacteria so as to further assess the role of gut microbiota in iNOS −/− mice. As expected, Allobaculum, Bifidobacterium, Lactobacillus, and Akkermansia were also depleted by Abx, with metabolic improvements along with reversal of IR and dyslipidemia. Lactobacillus rhamnosus and Lactobacillus casei improved glucose tolerance and IR in db/db mice , HFD-fed obese mice (Kim et al., 2013), and diabetic mice (Li et al., 2017b); however, Abx treatment in this study also reduced Lactobacillus. We therefore failed to notice any additional advantage in IR and metabolism. Captivatingly, vancomycin and the antibiotic cocktail reshaped the gut microbiome and the correlation analysis suggests the importance of gram-positive bacteria Allobaculum, Bifidobacterium, and Ruminococcus depletion in the reversal of IR and dyslipidemia in iNOS −/− mice with a seemingly less important role of gram-negative bacteria in the redox imbalanced state.
Reduced gluconeogenesis was evident by the change in PTT and a decrease in the expression of gluconeogenic enzymes in the liver and adipose tissue of iNOS −/− mice after treatment with antibiotics (Membrez et al., 2008;Zarrinpar et al., 2018). A decrease in the insulin-stimulated glycogen levels in iNOS −/− mice suggests enhanced flux of glucose into lipogenic pathways and their storage (Irimia et al., 2017), which was enhanced by Abx as reported previously in ob/ob mice (Membrez et al., 2008). Treatment with antibiotics improved insulin signaling in iNOS −/ − mice via Akt (Carvalho et al., 2012). Interestingly, treatment with antibiotics rescued dyslipidemia (Kuno et al., 2018) in iNOS −/− mice via reduction in the expression of genes involved in fatty acid synthesis (Zarrinpar et al., 2018) in the liver and adipose tissue and of lipid uptake in the adipose tissue, while the expression of genes involved in the lipid efflux was augmented in the liver and intestine.
Systemic changes in metabolism were reflected by untargeted metabolomics in the serum and revealed that purine and pyrimidine metabolites were elevated in iNOS −/− mice as supported by other studies in diabetic human subjects and rodents (Dudzinska, 2014;Urasaki et al., 2016). Bile acids and bile salts have been positively correlated with IR in humans (Cariou et al., 2011) and also in iNOS −/− mice and were decreased upon gut microbiota depletion by antibiotics as reported previously in HFD-fed mice (Fujisaka et al., 2016;Zarrinpar et al., 2018). Lipids are the major source of energy metabolism and are independent predictors for impairment of insulin actions and progression to diabetes (Zhao et al., 2017). The present study revealed that lipid metabolism seems to be the principal disordered pathway in iNOS −/− mice. Following vancomycin treatment, dyslipidemia and IR were reversed suggesting an association of lipid species with glucose intolerance. It has been reported that in NAFLD/NASH in humans, PE concentration is increased relative to the disease progression (Ma et al., 2016) and NASH patients have a decreased ratio of PC to PE (Li et al., 2006). Nozaki et al. observed higher propensity of fatty liver damage in iNOS −/− mice (Nozaki et al., 2015), and perturbed PE/PC ratio observed by us in iNOS −/− mice might provide a plausible explanation. Moreover, the PE/PC ratio was improved upon treatment with antibiotics in iNOS −/− mice similar to WT along with metabolic improvements by gram-positive bacteria depletion. Hydroxydecanoate is enhanced in obesity-associated IR and type 2 diabetes (Al-Sulaiti et al., 2019); we, similarly, observed an increase in this metabolite in iNOS −/− mice and a reduction with vancomycin intervention. Laurate and lauroyl carnitine were decreased in iNOS −/− mice, which could be due to the enhanced oxidation of fatty acids, as they are suggested to be accumulated during defects in fatty acid oxidation in diabetic states (Möder et al., 2003;Nowak et al., 2018). Interestingly, laurate and lauroyl carnitine were increased upon vancomycin treatment specifically and not by antibiotic cocktail in iNOS −/− mice suggesting the role of vancomycin-selective gram-positive bacteria in their metabolism.
Hippurate, a bacterial metabolite, is an early biomarker of IR and diabetes (Zhang, 2015). Allantoin, an oxidative stress biomarker, is the end product of uric acid oxidation, and it is enhanced in streptozotocin-induced diabetic rats, db/db mice, obese dogs, and gestational diabetic females (Hernandez-Baixauli et al., 2020). Gram-positive bacteria are positive contributors of hippurate, allantoin, glucose, 2-oxoglutarate, and glutamine as these were found to be present in the serum of mice infected with gram-positive bacteria (Hoerr et al., 2012). In the present study following vancomycin-induced depletion of gram-positive bacteria, these metabolites were reduced in iNOS −/ − mice. Formate, produced by gram-negative bacteria (Hoerr et al., 2012), was increased in vancomycin-treated iNOS −/− mice but not by antibiotic cocktail treatment, suggesting the differential metabolic response of gram-positive and gram -negative bacteria. Indole-3-ethanol, a bacterial-derived metabolite of tryptophan (Roager and Licht, 2018), was increased in iNOS −/− mice. Sebacic acid was higher in diabetic (Hameed et al., 2020) and NAFLD patients (Liu et al., 2021). Aminoadipate, a diabetogen, is reported to be involved in glycemic perturbations (Wang et al., 2013). Ophthalmate is a potential biomarker for oxidative stress and glutathione depletion (Mehta et al., 2019). These bacteria-derived metabolites were decreased upon microbiota depletion by antibiotics, suggesting their altered metabolism and possible association with IR in iNOS −/− mice. The present study demonstrates an association of the metabolic perturbations in iNOS −/− mice with select gut microbiota and serum metabolites. It might be helpful if the causative effect of these gut bacteria and metabolites could be also confirmed by implanting these bacteria in germ-free and conventionally raised iNOS −/− and WT mice. The dominance of gram-positive bacteria in insulin-resistant iNOS −/− mice and the improvement in metabolic markers with vancomycin intervention indicate that gram-positive bacteria are crucial drivers of observed metabolic phenotype in these mice. However, we do not completely rule out the role of gramnegative bacteria and other microorganisms such as archaea in regulating the metabolic functions. The association of the gramnegative bacteria, if any, could be also examined in iNOS −/− mice as this could not be investigated in our study. Understanding the interplay between nutrition, genetics, and microbial metabolism with redox imbalance is necessary to combat the scourge of IR.
Altogether, in the present study, we report that increased relative abundance of gram-positive Allobaculum and Bifidobacterium in redox imbalanced iNOS −/− mice led to metabolic perturbations with enhanced PE/PC ratio, 10hydroxydecanate, and indole-3-ethanol levels. Vancomycinmediated depletion of Allobaculum, Bifidobacterium, and Ruminococcus was associated with the reversal of IR and dyslipidemia in iNOS −/− mice. Vancomycin treatment also led to improvement in PE/PC ratio, 10-hydroxydecanoate, indole-3ethanol, allantoin, hippurate, sebacic acid, aminoadipate, and ophthalmate ( Figure 7). Overall, the results obtained in the present study demonstrate that gram-positive gut microbiota, the associated metabolites, and their crosstalk with the host drive dyslipidemia and insulin resistance in iNOS −/− mice.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

ETHICS STATEMENT
The animal study was reviewed and approved by the Institutional Animal Ethics Committee of CSIR-CDRI (IAEC/2014/43) in accordance with CPCSEA guidelines.

AUTHOR CONTRIBUTIONS
HA performed most of the experiments and wrote the manuscript. PP performed some experiments. VS provided deeper understanding and critical suggestions during the study and manuscript finalization. MS provided critical inputs and interpretation for gut microbiome studies/data. YK performed the metabolomics experiments and analyzed the data. BD provided critical inputs regarding fecal DNA isolation and interpretation for gut microbiome studies/data. KJ provided the animals and critical suggestions during the study. MD conceptualized the whole project, designed and supervised the studies, and interpreted the data as being presented in the MS; she also edited and finalized the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
The present study was supported by JC Bose National fellowship (SB/SE/JCB-017/2015) and THSTI core grant to MD and by DST SERB Project (EMR/2017/003730) to KJ (CSIR-CDRI). Research fellowships to HA from the Indian Council of Medical Research and to PP from the Council of Scientific and Industrial Research, India, are acknowledged.