SGLT2 inhibition leads to a restoration of hepatic and circulating metabolites involved in the folate cycle and pyrimidine biosynthesis

Abstract

Inhibition of sodium-glucose cotransporter 2 (SGLT2) by empagliflozin (EMPA) and other “flozins” can improve glycemic control under conditions of diabetes and kidney disease. Though they act on the kidney, they also offer cardiovascular and liver protection. Previously, we found that EMPA decreased circulating triglycerides and hepatic lipid and cholesterol esters in male TallyHo mice fed a high-milk-fat diet (HMFD). The goal of this study was to determine whether the liver protection is associated with a change in metabolic function by characterizing the hepatic and circulating metabolic and lipidomic profiles using targeted LC-MS. In both male and female mice, HMFD feeding significantly altered the circulating and hepatic metabolome compared with low-fat diet (LFD). Addition of EMPA resulted in the restoration of circulating orotate (intermediate in pyrimidine biosynthesis) and hepatic dihydrofolate (intermediate in the folate and methionine cycles) levels in males and acylcarnitines in females. These changes were partially explained by altered expression of rate-limiting enzymes in these pathways. This metabolic signature was not detected when EMPA was incorporated into an LFD, suggesting that the restoration requires the metabolic shift that accompanies the HMFD. Notably, the HMFD increased expression of 18 of 20 circulating amino acids in males and 11 of 20 in females, and this pattern was reversed by EMPA. Finally, we confirmed that SGLT2 inhibition upregulates ketone bodies including β-hydroxybutyrate. Collectively, this study highlights the metabolic changes that occur with EMPA treatment, and sheds light on the possible mechanisms by which this drug offers liver and systemic protection.

NEW & NOTEWORTHY Sodium-glucose cotransporter 2 (SGLT2) inhibitors, including empagliflozin, have emerged as a new treatment option for individuals with type 2 diabetes that have positive impacts on kidney and cardiovascular disease. However, less is known about their impact on other tissues, including the liver. Here, we report that empagliflozin reduces hepatic steatosis that is associated with restoring metabolic intermediates in the folate and pyrimidine biosynthesis pathways. These changes may lead to new approaches to treat nonalcoholic fatty liver disease.

INTRODUCTION

According to the Centers for Disease Control, diabetes is currently the seventh leading cause of death in the United States, and its growing prevalence worldwide has led to its classification as an epidemic (1). Out of all cases, >90% of diabetics are classified as having type 2 diabetes (T2D) and nearly half of all individuals with T2D are also diagnosed with nonalcoholic fatty liver disease (NAFLD) (2–6). Recently, NAFLD was retermed “metabolic dysfunction-associated fatty liver disease” or MAFLD (7). MAFLD/NAFLD is characterized as a rise in hepatic steatosis that cannot be explained by alcohol consumption, immune disorders, and/or prescription of lipid-altering medications. Although its pathogenesis is incompletely understood, it is associated with insulin resistance, dysbiosis, inflammation, cholestasis, and hyperlipidemia, all of which can promote the development of nonalcoholic steatohepatitis (NASH)/metabolic dysfunction-associated steatohepatitis (MASH), fibrosis, and cirrhosis (2–5, 8).

For patients with T2D, managing glucose levels is a crucial concern as hyperglycemia can promote the development of diabetic comorbidities. Circulating blood glucose is continually filtered by the renal glomerulus where it is rapidly returned to circulation via the sodium-glucose cotransporters (SGLT1 and SGLT2). SGLT2, found in the S1 and S2 segments of the renal proximal tubule, accounts for ∼90% of all glucose reabsorption. Recently, inhibitors of SGLT2 (SGLT2is), such as empagliflozin (EMPA), have emerged as a novel therapeutic for patients with T2D (9–11). By blocking SGLT2-mediated glucose reabsorption, glucose is excreted in the urine, which in turn lowers blood glucose levels. In addition, SGLT2is offer significant cardiovascular protection and can even prevent the worsening of chronic kidney disease in the absence of overt diabetes (12).

Diabetes, obesity, and MAFLD can be considered “multiorgan” disorders. Indeed, a multitude of studies have pointed to the interplay between organ systems that result in, or protect from, the development of these disorders. Recently, we used EMPA to investigate the effects that modulation of renal glucose handling has on other tissues, mainly the liver (13, 14). Using obese male and female TallyHo mice maintained on a high-milk-fat diet (HMFD), a model that mimics the polygenicity of the human population with diabetes, we determined that EMPA attenuates the development of steatosis and modulates hyperglycemia by decreasing circulating triglycerides, attenuating hepatic lipid accumulation, lowering hepatic cholesterol levels, and promoting healthy adipose expansion (14). However, although SGLT2is appear to have therapeutic potential for the treatment of MAFLD, a disorder in desperate need of new approaches and targets, the mechanisms behind this mitigation are still unknown. To investigate, we performed targeted metabolomics and lipidomics (LC-MS) to narrow down the list of metabolites and metabolic pathways that are restored to healthy levels, in TallyHo mice treated with EMPA. This study reveals that HMFD significantly alters the circulating and hepatic metabolite profile including increasing 18 of 20 and 11 of 20 amino acids in males and females, respectively. EMPA treatment restores the amino acid profile and the expression of orotate and dihydrofolate, metabolites that are part of the pyrimidine biosynthesis pathway and folate cycle, in males and acylcarnitines in females.

MATERIALS AND METHODS

In Vivo Mouse Studies

Male and female TallyHo mice (n = 4–10) were transitioned to a high-milk-fat diet (HMFD; Research Diets) at 8–12 wk old. Both male and female mice were maintained on the HMFD alone or a HMFD incorporated with 0.01% empagliflozin (EMPA; MedChem Express) for 24 wk. The final concentration of EMPA provided a dosing of 10 mg/kg body wt. A similar cohort was maintained on a low-fat (normal chow) diet (LFD) for the same length of time. All mice were kept on a 12:12 light-dark cycle and provided food and water ad libitum. Mice were weighed weekly. At the conclusion of the 24 wk, mice were fasted for 4 h (∼09:00 am to 01:00 pm) and blood was collected via cardiac puncture and placed into EDTA-coated tubes, spun at 2,000 g for 15 min, with plasma stored at −80°C until used for metabolomic analysis or other downstream confirmatory assays (see Targeted Metabolomics below). Lobes of the livers were collected and either placed in 10% neutral-buffered formalin for histological analysis, lysed directly into lysis buffer for Western blotting, or flash frozen in liquid nitrogen for downstream applications. In a separate cohort, male and female mice (n = 5) were maintained on the LFD for 24 wk in the continued absence or presence of 10 mg/kg EMPA. Plasma was collected as described and processed for metabolomic analysis. All animal studies were performed in accordance with the policies of Georgetown University’s Division of Comparative Medicine with approval from the Institutional Animal Care and Use Committee.

Histological Analysis

One lobe of each liver was drop-fixed in 10% buffered formalin for 24 h before being transferred to 70% ethanol. The samples were subsequently embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E) by the Georgetown University Lombardi Comprehensive Cancer Center’s Histopathology Shared Resource. Stained slides were scanned using an Aperio Imager (Leica Biosystems, Buffalo Grove, IL) at ×40. To quantify lipid accumulation, scanned sections were analyzed using QuPath software using pixel density thresholds set to 255. Total area and mean lipid size were calculated for each droplet.

Cholesterol Analysis

In a ceramic mortar and pestle, one frozen liver lobe was immersed in liquid nitrogen and ground into a fine powder. Powered liver tissue of 10 mg was mixed with 400 µL of chloroform-isopropanol-NP-40 (7:11:0.1) and allowed to mix overnight to extract lipids. The following day, 200 µL of the lysate was collected, spun at 13,000 g, and left to dry. Samples were reconstituted in 40 µL of assay buffer, and total and free cholesterol esters were measured according to the manufacturer’s instructions (Abcam ab65359).

Triglyceride Assay

To determine liver triglycerides, the lipid extractions that were obtained for the cholesterol assay were diluted 1:3 in assay buffer (Abcam ab65359). Triglyceride standard (Thermo Fisher 23-666-422), diluted tissue lysate, and a blank (buffer) were loaded into a 96-well plate in triplicate (1.5 µL of each). Next, 150 µL Infinity Triglycerides Liquid Stable Reagent (Thermo Fisher TR-22421) was added to each well, and the plate was incubated for 15 min at room temperature. After a brief, gentle shake, absorbance was read at 540 nm using a plate reader (BMG Labtech, Ortenberg, Germany). To calculate plasma triglyceride concentration, 1.5 µL of isolated plasma (see In Vivo Mouse Studies above) was added per well, combined with 150 µL of triglycerides stable reagent, incubated for 15 min, and absorbance was read at 540 nm.

Targeted Metabolomics

To identify the metabolites that are changed with EMPA treatment, targeted metabolomics were performed on both liver and plasma samples in collaboration with the mass spectrometry (MS) and Analytical Pharmacology Shared Resource using a method developed in-house. This method can quantitate >450 endogenous metabolites using the QTRAP 7500 LC-MS/MS system (Sciex, Massachusetts). Flash-frozen liver samples were dissolved in 150 µL of methanol-water 50:50 extraction buffer that contained 250 ng/mL debrisoquine (DBQ) for the positive mode standard and 250 ng/mL 4-nitrobenzoic acid for the negative mode standard. Samples were vortexed for 30 s, homogenized for 1–2 min on ice, and incubated for 20 min, followed by the addition of 150 µL of acetonitrile and incubation at −20°C for 20 min. Samples were centrifuged at 13,000 rpm for 20 min at 4°C with the supernatant transferred to MS vial for LC-MS analysis. Plasma samples of 25 µL were dissolved in 100 µL of extraction buffer containing 200 ng/mL of DBQ for the positive-mode standard and 200 ng/mL 4-nitrobenzoic acid for the negative-mode standard. The samples were vortexed for 30 s and incubated on ice for 20 min, followed by the addition of 100 µL of acetonitrile. The samples were incubated for 20 min and centrifuged at 13,000 rpm for 20 min at 4°C, and the supernatant was transferred to an MS vial for LC-MS analysis.

Five microliters of prepared sample were injected onto a Kinetex 2.6 µm FS 100A, 150 × 2.1 mm (Part No. 00 F-4723-AN; Phenomenex, California) using SIL-30 AC auto sampler (Shimadzu) connected with a high-flow LC-30AD solvent delivery unit (Shimadzu) and Exion 30AD communication bus module (Shimadzu) online with QTRAP 7500 operating in positive and negative ion mode. A binary solvent composed of water with 0.1% formic acid (solvent A) and acetonitrile with 0.1% formic acid (solvent B) was used. The extracted metabolites were resolved at 0.2 mL/min flow rate. The LC gradient conditions were as follows: initial—100% A, 0% B for 2.1 min; 14 min—5% A, 95% B; 15.1 min—100% A, 0% B. The auto sampler and oven were kept at 15°C and 30°C, respectively. Source and gas settings for the method were as follows: curtain gas = 45, CAD gas = 10, ion spray voltage = 2,000 V in positive mode and ion spray voltage = 4,500 V in negative mode, temperature = 500°C, ion source gas 1 = 45, and ion source gas 2 = 70. The data were normalized to internal standard area and processed using Sciex OS software. To ensure high quality and reproducibility of LC-MS data, the column was initially conditioned using a pooled control serum or liver QC sample. The QC sample was also injected periodically to monitor shifts in signal intensities and retention time. Furthermore, a standard plasma sample designed to represent “normal” human plasma developed by the National Institute of Standards and Technology (NIST) was run periodically (prepared in the same manner) to evaluate instrumental variance. After analysis, metabolites with pooled liver or plasma reference coefficient variation (CV) greater than 20% were removed.

Statistical analysis was performed using MetaboAnalyst (15). Raw expression data were imported as a CSV file. A three-way heat map highlighting the 50 greatest changed metabolites among all samples (LFD, HMFD, and HMFD + EMPA) was generated to identify those metabolites that changed with different treatments, whereas a volcano plot was used to assess statistical significance (HMFD vs. HMFD + EMPA; LFD vs. HMFD). The statistical threshold was set at a fold change of 2.0 and a P < 0.05.

Targeted Lipidomics

This method is designed to measure 21 classes of lipid molecules, which includes diacylglycerols (DAG), cholesterol esters (CE), sphingomyelins (SM), phosphatidylcholine (PC), triacylglycerols (TAG), free fatty acids (FFA), ceramides (CER), dihydroceramides (DCER), hexosylceramide (HCER), lactosylceramide (LCER), phosphatidylethanolamine (PE), lysophosphatidylcholine (LPC), lysophosphatidylethanolamine (LPE), phosphatidic acid (PA), lysophosphatidic acid (LPA), phosphatidylinositol (PI), lysophosphatidylinositol (LPI), phosphatidylglycerol (PG), acylcarnitines, and phosphatidylserine (PS) using QTRAP 5500 LC-MS/MS System (Sciex). For this purpose, all tissue samples were dissolved in 300 μL of chilled isopropanol containing internal standards for lipid classes. The samples were vortexed for 30 s and homogenized for 1–2 min on ice and incubated on ice for 20 min followed by incubation at −20°C for 30 min. Samples were centrifuged at 13,000 rpm for 2 h at 4°C. The supernatant was transferred to MS vial for LC-MS analysis.

Five microliters of each sample were injected onto a Xbridge amide 3.5 μm, 4.6X 100 mm (waters) using SIL-30 AC auto sampler (Shimadzu) connected with a high-flow LC-30AD solvent delivery unit (Shimadzu) and CBM-20A communication bus module (Shimadzu) online with QTRAP 5500 (Sciex, Massachusetts) operating in positive and negative ion mode. A binary solvent comprising acetonitrile-water 95/5 with 10 mM ammonium acetate as solvent A and acetonitrile-water 50/50 with 10 mM ammonium acetate as solvent B was used for the resolution. Lipids were resolved at 0.7 mL/min flow rate, initial gradient conditions started with 100% of solvent A, shifting toward 99.9% of solvent A over a time period of 3 min, 94% of solvent A over a time period of 3 min and 25% of solvent A over a period of 4 min. Finally, samples were washed with 100% of B for 6 min and equilibrated to initial conditions (100% of solvent A) over a time period of 6 min using an auto sampler temperature of 15°C and an oven temperature of 35°C. Source and gas settings were as follows: curtain gas = 30, CAD gas = medium, ion spray voltage = 5.5 kV in positive mode and −4.5 kV in negative mode, temperature = 550°C, nebulizing gas = 50, and heater gas = 60. The data were normalized to respective internal standard area for each class of lipid and processed using MultiQuant 3.0.3 (Sciex). To ensure quality and reproducibility of LC-MS data, the column was conditioned using the pooled QC samples initially, followed by periodic QC injections to monitor shifts in signal intensities and retention time.

Statistical analysis was performed using both MetaboAnalyst (to generate volcano plots and heat maps) and LipidSig (16) to classify changes in lipid properties. The statistical threshold was set to a fold change of 2.0 and P < 0.05.

β-Hydroxybutyrate Assay

Plasma was deproteinized using a 10-kDa column (Abcam, ab93349) to remove interfering substances found in the blood. Twenty-five microliters of deproteinized plasma were used to detect β-hydroxybutyrate using a colorimetric assay performed according to the manufacturer’s specifications with absorbance read at 450 nm (Abcam, ab83390). Concentrations were extrapolated from the standard curve, and total concentration was calculated using the following formula: (β-hydroxybutyrate in well/volume added) × 2 (dilution factor).

Western Blotting

At the conclusion of the study, livers were collected and ∼50 mg were homogenized on ice in 500 µL lysis buffer [250 mM sucrose and 10 mM triethanolamine at pH 7.6 with cOmplete Mini (Sigma 11836153001)], and PMSF protease inhibitors were added immediately before use. All samples were centrifuged at 2,800 g for 15 min at 4°C, and protein concentrations were determined by the Bradford assay (Bio-Rad 500-0201). Total protein of 40 μg was suspended into Laemmli sample buffer and loaded into precast stain-free gels (Bio-Rad 17000435) and transferred to a PVDF membrane. Total protein loading and normalization were determined by exposing the transferred membranes to UV light at an excitation of 302 nm according to the stain-free parameters. Primary antibodies and their blotting conditions used in the study are listed in Table 1, and HRP-conjugated secondary antibodies (Invitrogen A16035) were added at a 1:5,000 dilution for 45–60 min at room temperature. Images were obtained on the Azure Biosystems 300 imager, and quantification was performed using ImageJ (National Institutes of Health) by converting the images to 8-bit and calculating the mean pixel intensity of all bands.

Table 1.

Primary antibodies and their blotting conditions used in the study

Protein Detected	Antibody Source	Blotting Conditions
DHFR	Abcam ab124814	1:1,000 o/n
HCP1	Abcam ab25134	1:1,000 o/n
CAD	Fortis Life Sciences A301-374A	1:1,000 o/n
CPS1	Abcam ab129076	1:1,000 o/n
OTC	Abcam ab203859	1:1,000 o/n
DHODH	Proteintech 14877-1-AP	1:1,000 o/n

CAD, carbamoyl-phosphate synthase 2/aspartic acid carbamoyl transferase/dihydroorotase; CPS1, carbamoyl-phosphate synthase 1; DHFR, dihydrofolate reductase; DHODH, dihydroorotate dehydrogenase; HCP1, heme carrier protein 1; o/n, overnight; OTC, ornithine transcarbamylase.

Statistical Analysis

Values presented are means ± standard error. Two-way ANOVA followed by Tukey’s post hoc test was performed by Prism GraphPad, and significance markings are indicated on the graphs if P < 0.05 in the multiple comparison tests. A Kolmogorov–Smirnov test was performed on lipid droplet data to examine the differences in droplet size distribution across different treatment groups using RStudio.

RESULTS

We previously determined that a 12-wk HMFD feeding was sufficient to induce steatosis, but not widespread hepatic inflammation (14). Thus, we extended our treatment to 24 wk with both male and female TallyHo mice receiving a HMFD in the continued absence or presence of EMPA. Male mice fed a HMFD with or without EMPA continued to gain weight throughout the treatment period, with no statistically significant changes detected in the week-by-week analysis. However, the HMFD-fed mice did begin to plateau and even show a decline in body weight (BW) toward the end of the study, and at this timepoint, male mice treated with EMPA weighed more (P = 0.04; Fig. 1A). EMPA did not alter BW or weight gain in the female TallyHo mice, mirroring our previous findings (14) (Fig. 1A). Although liver weights (normalized to final BW) were slightly lower in both male and female EMPA-treated mice, this did not reach significance (Fig. 1B). Despite this, we observed a decrease in hepatic lipid accumulation with EMPA treatment (Fig. 2A). Compared with LFD-fed TallyHo mice, the HMFD significantly increased the total number of lipid droplets (calculated as % area) in both male and female mice (Fig. 2, A and B). EMPA did not reduce the total lipid droplet count (calculated as % area) but did significantly decrease the median lipid droplet size in both male and female TallyHo mice (Fig. 2, B and C). To examine the composition of lipids, we first measured hepatic triglycerides and cholesterol. HMFD significantly increased liver triglycerides (males and females; Fig. 2D) and total liver cholesterol (males only; Fig. 2E) compared with the LFD. In line with our previous findings (14), EMPA only partially decreased hepatic triglycerides (EMPA vs. HMFD, P = 0.16) but did reduce total liver cholesterol (EMPA vs. HMFD, P = 0.003) in the male TallyHo mice. Consumption of a HMFD in male mice sufficiently increased the ratio of cholesterol esters to free cholesterol, a marker of liver damage and potential for developing hepatocellular carcinoma (17), but this was only marginally decreased by EMPA following 24 wk of treatment (EMPA vs. HMFD, P = 0.37; Fig. 2F). Neither triglycerides nor cholesterol (total or ester/free) were changed in the female mice.

Figure 1.

Empagliflozin (EMPA) promotes sustained growth in obese TallyHo mice. TallyHo mice were maintained on a high-milk-fat diet (HMFD) in the continued absence or presence of 10 mg/kg body wt EMPA and weighed weekly for 24 wk. A: weights for males (HMFD n = 14, EMPA n = 15 each with 1 fatality) and females (HMFD n = 12, EMPA n = 12) were graphed and a two-way ANOVA followed by Sidak’s multiple comparisons comparing time and treatment was performed. Although initial body weights were similar, EMPA treatment promoted increased growth in the later weeks for male mice. No differences in body weight were detected among the female groups. B: at the conclusion of the study, livers were isolated, weighed, and normalized to body weight. A two-way ANOVA followed by Tukey’s post hoc test was performed and no significance was noted.

Figure 2.

Empagliflozin (EMPA) decreases hepatic lipid droplet size and total liver cholesterol. At the conclusion of the study, livers were stained with hematoxylin and eosin (H&E), and the total slide was scanned and analyzed by QuPath. A: representative H&E images are shown for the male and female high-milk-fat diet (HMFD) (male n = 4, female n = 5), HMFD + EMPA (male n = 4, female n = 5), and low-fat diet-fed mice (n = 3 for males and females). The scale bar indicates 500 μm. B: using a mean pixel density of 225 as a maximum to distinguish lipid droplets from tissue, the lipid droplet density relative to the total liver tissue was determined and analyzed by two-way ANOVA followed by Tukey’s post hoc test. C: the median lipid droplet size was calculated with a means ± SE graphed for each sample analyzed. Analysis was performed by a Wilcoxon test followed by a Dunn test with *P < 0.05, **P < 0.005, ****P < 0.0001. HMFD feeding greatly increased the % lipid droplets and median lipid droplet size compared with low-fat feeding. EMPA decreased lipid droplet size but not the total % lipid droplets. Liver triglycerides (D), total liver cholesterol (E), and ester:free liver cholesterol levels (F) are plotted and were analyzed by two-way ANOVA followed by Tukey’s post hoc test. P values displayed on the graph indicate the significance value as determined by the multiple comparisons.

To explore the lipid profiles more deeply, livers from low-fat-, HMFD-, and HMFD + EMPA-fed mice were analyzed by LC-MS using a targeted lipid panel. In the livers, more than 730 different lipids were detected. Visualization using a three-treatment group heat map revealed that of the 50 lipids with the greatest change, in male mice, HMFD feeding upregulated all but three compared with the low-fat control (Fig. 3A). Overall, the addition of EMPA into the HMFD only marginally decreased the lipid profile compared with the HMFD alone (Fig. 3A; note the lighter shades of red in the HMFD + EMPA group). To narrow in on the statistically changed lipids, we defined the inclusion criteria as a fold change of >2 and a P < 0.05. Using this criterion, EMPA increased five triglyceride species (red font) and decreased hydroxyceramide (blue font) in males (Fig. 3B). Similar findings were observed in the female mice. When compared with the low-fat control, the HMFD increased all lipids included in the heat map (showing the 50 lipids with the greatest change) (Fig. 3C). Specifically, EMPA increased several species of triglycerides (red font) and significantly decreased several species of phosphatidyl choline (blue font) in the females (Fig. 3D). It should also be noted that there was more mouse-to-mouse variability in the female mice as one of the EMPA-livers clustered most strongly with the HMFD cohort (Fig. 3C). Taken together, the lipidomics analysis indicates that EMPA does not drastically alter hepatic lipids following 24 wk of HMFD feeding in either male or female mice.

Figure 3.

High-milk-fat diet (HMFD) feeding significantly upregulates total lipids. Lipidomics was performed on male and female liver samples from HMFD, HMFD + empagliflozin (EMPA), and low-fat diet groups (n = 4) using a panel of 730 different lipids. A: a three-treatment heat map highlighting the 50 lipid species with the largest changes across the groups is shown for the male samples. Blue colors indicated decreased expression, whereas red colors indicate increased expression. Lipid identifiers are listed on the right, and each column represents an individual liver grouped according to treatment. B: a volcano plot was generated from the entire 730-lipid panel (male samples HMFD vs. EMPA) with circles representing a different lipid. The y-axis indicates the P value as a log10 scale, and the x-axis indicates fold change (FC) in a log2 scale. Lipids that had a fold change >2 and a P < 0.05 were deemed significant. Those that are shifted to the right (red circles) indicates lipids that were increased upon EMPA treatment and those in blue are decreased upon treatment with EMPA. C: a three-treatment heat map highlighting the 50 lipid species with the largest changes across the groups is shown for the female samples. D: a volcano plot was generated from the entire 730-lipid panel for the female samples (HMFD vs. EMPA).

To identify possible mechanisms for the altered lipid profiles, we performed targeted MS on both serum and livers from the HMFD-fed TallyHo mice in the presence or absence of EMPA. Significantly altered metabolites were defined as greater than twofold expression difference with P < 0.05. In the male liver, HMFD feeding significantly altered the metabolic profile by increasing seven metabolites (Fig. 4, A and B, red font) and decreasing 17 (Fig. 4, A and B, blue font) compared with the low-fat feeding. When compared with the HMFD alone, the addition of EMPA (HMFD + EMPA) upregulated six metabolites including urate, phenylacetylglycine, glutathione, and dihydrofolate (Fig. 4, A and B, red font). No metabolites were significantly decreased with treatment. To narrow in on potentially relevant changes to the metabolic profile, we assumed that those metabolites that are restored to, or close to, that of LFD levels with EMPA treatment may point to metabolic pathways that are physiologically important. In the male liver, only one metabolite met this criterion (dihydrofolate), the reduced form of folic acid that participates in the one-carbon metabolism cycle leading to DNA-synthesis or methylation reactions (Fig. 4C). This finding was male-specific as there were no changes in hepatic dihydrofolate levels among the female mice (Fig. 4C). Dihydrofolate is the precursor to the important folate cycle that generates one-carbon units to be used in the methionine cycle (18). Importantly, folate deficiency has been reported in cases of MAFLD making the restoration of this metabolite upon EMPA treatment a clinically relevant finding (19). To dive into the mechanism behind the restoration of this metabolite, we examined the expression of the hepatic folate transporter, heme carrier protein 1 (HCP1), and the key enzyme that is responsible for reducing folic acid to dihydrofolate and tetrahydrofolate, dihydrofolate reductase (DHFR) (20). No significant expression differences were noted for HCP1 in either male or female livers. EMPA treatment in males did upregulate hepatic DHFR relative to both LFD and HMFD conditions (Fig. 5B) but did not change folic acid levels (Fig. 5C). Although the enzymes were not changed in the female livers, both HMFD and EMPA-treated females had reduced levels of folic acid compared with low-fat controls (HMFD vs. LF, P < 0.0001; EMPA vs. LF, P < 0.0001; Fig. 5C). Collectively, in males, the increase in DHFR coupled with the lack of change in folic acid suggests that EMPA may promote the metabolism of dihydrofolate to tetrahydrofolate and that may explain the decrease in hepatic dihydrofolate levels (Fig. 5D). Given the interconnection between the folate cycle and the methionine cycle, we also examined the metabolic intermediates that are produced in these pathways that were included in the MS screen (Fig. 5E). As indicated by the heat map, there was sufficient mouse-to-mouse variability in the expression of these metabolites across all treatment groups with no clear trend (apart from dihydrofolate levels). Thus, EMPA does not appear to influence the methionine cycle.

Figure 4.

Empagliflozin (EMPA) changes the liver metabolite profile in male TallyHo mice. Targeted metabolomics was performed on the livers isolated from male TallyHo mice treated with either a high-milk-fat diet (HMFD), HMFD + EMPA, or low-fat diet (LFD) for 24-wk (n = 4 for each group). A: volcano plots were generated to compare either HMFD vs. LFD (left) or EMPA vs. HMFD (right) with circles representing different metabolites. The y-axis indicates the P value as a log10 scale, and the x-axis indicates fold change (FC) in a log2 scale. Metabolites that had a FC > 2 and a P < 0.05 were deemed significant. Those that are shifted to the right (red circles) indicate metabolites that were increased compared with the comparison group and those in blue were decreased. B: significant metabolites identified from the volcano plots were organized into Venn diagrams to determine metabolites that were reciprocally changed. The left circle lists the significantly changed metabolites when comparing HMFD vs. LFD and the right reflects those changed between EMPA vs. HMFD. Those in blue font were decreased and those in red font were increased. In the center is the lone liver metabolite (dihydrofolate) that was determined to be restored by EMPA treatment. C: the expression graph for dihydrofolate is shown as a dot plot, with the y-axis indicating expression and the x-axis indicating the different treatment groups for both male and female samples. HMFD decreased dihydrofolate expression in male mice and this was partially restored with EMPA treatment.

Figure 5.

Empagliflozin (EMPA) does not change the expression of folate enzymes and transporters. A: male and female livers from high-milk-fat diet (HMFD), EMPA, and low-fat diet-fed mice (n = 4–8) were immunoblotted for heme carrier protein 1 (HCP1), the folate transporter expressed in the liver, and dihydrofolate reductase (DHFR), the enzyme responsible for converting folic acid to dihydrofolate (DHF) and trihydrofolate (THF). B: densitometry quantification is plotted for each protein, with the y-axis reflecting the protein expression/total protein loading and the x-axis reflecting the three treatment groups for both males and females. A two-way ANOVA followed by Tukey’s post hoc test was performed with multiple comparison significance noted as indicated by P value. C: liver folic acid levels (ng/mL) for all treatment groups. D: key steps in the folate cycle and the closely associated methionine cycle pathway are shown. Arrows indicate either increased, decreased, or no change with regards to EMPA treatment. E: a three-treatment heat map is shown for each individual liver clustered into three groups (low fat, HMFD + EMPA, HMFD) for folate cycle and methionine cycle metabolites that were included in the targeted metabolomics screen. Blue colors indicated decreased expression, whereas red colors indicate increased expression.

Although dihydrofolate levels were not changed with EMPA in female livers (Fig. 4C), notable changes did occur with respect to liver carnitine levels. As for the male metabolites, volcano plots highlighting significantly altered metabolites (>2-fold expression difference with P < 0.05) were generated to compare HMFD versus LFD and HMFD versus EMPA in the female mice (Fig. 6A). Although EMPA treatment did not significantly increase any metabolites relative to HMFD alone, it did return expression of four hepatic acylcarnitines to levels found in the low-fat diet controls (Fig. 6, B–E). Hexanoylcarnitine was also decreased in the male livers upon EMPA treatment (Fig. 6C). Acylcarnitines have recently been linked to a number of metabolic disorders and mitochondrial dysfunction (21). Thus, in the female mice, it appears that EMPA properly restores their expression suggesting an improvement in metabolic function.

Figure 6.

Empagliflozin (EMPA) decreases liver acylcarnitines in female mice. Targeted metabolomics was performed on livers isolated from female TallyHo mice treated with either a high-milk-fat diet (HMFD), HMFD + EMPA, or low-fat diet (LFD) for 24-wk (n = 3–5 for each group). A: volcano plots were generated to compare either HMFD vs. LFD (left) or EMPA vs. HMFD (right) with circles representing different metabolites. The y-axis indicates the P value as a log10 scale, and the x-axis indicates fold change (FC) in a log2 scale. Metabolites that had a FC > 2 and a P < 0.05 were deemed significant. Those that are shifted to the right (red circles) indicate metabolites that were increased compared with the comparison group and those in blue were decreased. Comparison of the expression values for those changed metabolites revealed that EMPA significantly decreased four acylcarnitines. Short-chain butylcarnitine (B), medium-chain hexanoylcarnitine (C), medium-chain octanoylcarnitine (D), and short-chain valerylcarnitine (E). P values noting significance by Tukey’s multiple comparisons are noted in each graph.

SGLT2 inhibitors are known to improve physiological functioning of a number of tissues including the heart, adipose, and kidney. With this likely comes a change in metabolic function that may be detected in the expression of circulating metabolites. These metabolites have the potential to impact the liver. Thus, we examined the circulating metabolome of LFD- and HMFD-fed male and female TallyHo mice in the absence or presence of EMPA. Examination of the 50 greatest expression changes (Fig. 7A) revealed unique signatures for the metabolic profile in male mice. In particular, we noted that the HMFD increased expression of 18 of 20 circulating amino acids (detected in serum), and this pattern was reversed with the addition of EMPA (Fig. 7B). These data align with other reports indicating that patients with T2D present with higher circulating amino acids, and there is a strong association between elevated amino acids and insulin resistance, ketoacidosis, and visceral fat accumulation (22, 23). This pattern was serum specific, as we did not note global increases in hepatic amino acids in the HMFD-fed mice nor did we see a complete restoration of that profile upon EMPA treatment. Although clear differences were also noted in the female mice, there was more variation between individual mice (Fig. 8A). This is in keeping with a less-penetrant diabetic phenotype in the female TallyHo mice (24–32). HMFD feeding in female mice elevated 11 amino acids, but the EMPA’s ability to restore this profile was diminished (Fig. 8B).

Figure 7.

Empagliflozin (EMPA) alters circulating metabolites in male mice, including decreasing amino acids and increasing β-hydroxybutyrate (BHB). Targeted metabolomics was performed on serum isolated from high-milk-fat diet (HMFD), EMPA, and low-fat diet-fed (LFD) TallyHo male mice (n = 4 for each group) to screen for expression of >450 endogenous metabolites. A: the 50 most changed metabolites (EMPA vs. HMFD) are shown (listed on the right) in a heat map with each column representing a different sample grouped by treatment. Blue colors indicated decreased expression whereas red colors indicate increased expression. B: expression from the targeted metabolomics was averaged for each treatment group, and the expression profiles of all 20 amino acids are shown. HMFD feeding increased (noted by red colors) 18 of 20 circulating amino acids and this was partially restored by EMPA treatment (denoted by blue or lighter red coloring). C: confirmatory expression of plasma BHB (nmol/µL) was analyzed by ELISA for all treatment groups for both males and females (n = 4 or 5/group). A two-way ANOVA followed by Tukey’s post hoc analysis was performed, and significance is noted by P value.

Figure 8.

Empagliflozin (EMPA) alters circulating metabolites in female mice. Targeted metabolomics was performed on serum isolated from high-milk-fat diet (HMFD), EMPA, and low-fat diet-fed (LFD) TallyHo female mice (n = 5 for each group) to screen for expression of >450 endogenous metabolites. A: the 50 most changed metabolites (EMPA vs. HMFD) are shown (listed on the right) in a heat map with each column representing a different sample grouped by treatment. Blue colors indicated decreased expression, whereas red colors indicate increased expression. B: expression from the targeted metabolomics was averaged for each treatment group, and the expression profiles of all 20 amino acids are shown. HMFD feeding increased (noted by red colors) 11 of 20 circulating amino acids and this was partially restored by EMPA treatment (denoted by blue or lighter red coloring).

Of particular interest was the determination that EMPA increased the circulation of the ketone body β-hydroxybutyrate (BHB) in both male and female mice relative to HMFD-fed mice (Fig. 7A and Fig. 8A). This finding was confirmed using traditional ELISA, whereby EMPA-treated male and female mice showed a twofold increase in circulating BHB (male mice: EMPA vs. HMFD P = 0.045, EMPA vs. LFD P = 0.002; female mice: EMPA vs. HMFD P = 0.01, EMPA vs. LFD P = 0.002; Fig. 7C).

To narrow in on the most important alterations in the circulating metabolome, we once again honed in on those metabolites whose expression is restored to that of control (low-fat) levels. As for the liver, we noted only a single metabolite that was reciprocally expressed upon EMPA treatment in male mice: orotate (Fig. 9, A–C). Upon HMFD feeding, orotate levels are significantly increased in male serum, and upon treatment with EMPA, its levels return close to control levels (Fig. 9, B and C). In the female mice, HMFD feeding decreased indoxyl sulfate levels and this was partially restored with EMPA treatment (Fig. 9D).

Figure 9.

Empagliflozin (EMPA) restores circulating orotate levels. A: volcano plots were generated from the targeted metabolomics (Fig. 7) performed on male serum to compare either high-milk-fat diet (HMFD) vs. low-fat diet (LFD) (left) or EMPA vs. HMFD (right) with circles representing different metabolites. The y-axis indicates the P value as a log10 scale, and the x-axis indicates fold change (FC) in a log2 scale. Metabolites that had a FC > 2 and a P < 0.05 were deemed significant. Those that are shifted to the right (red circles) indicate metabolites that were increased compared with the comparison group and those in blue were decreased. B: significant metabolites identified from the volcano plots were organized into Venn diagrams to determine metabolites that were reciprocally changed. The left circle lists the significantly changed metabolites when comparing HMFD vs. LFD and the right reflects those changed between EMPA vs. HMFD. Those in blue font were decreased and those in red font were increased. In the center is the lone circulating metabolite (orotate) that was determined to be restored by EMPA treatment. C: the expression graph for orotate is shown as a dot plot with the y-axis indicating expression and the x-axis indicating the different treatment groups. Each dot indicates a different liver sample. HMFD increased orotate expression by >2.0× with a P < 0.05. D: volcano plots for the female circulating metabolomics (Fig. 8) is plotted with the HMFD vs. LFD shown to the left and EMPA + HMFD vs. HMFD alone shown to the right.

The circulating metabolic profile of mice fed a HMFD will be different than those fed a LFD. To determine whether EMPA exerts similar changes to the metabolome in the context of a LFD, we performed a targeted metabolomics screen in a separate cohort of mice fed a LFD with or without EMPA (Fig. 10). EMPA significantly changed a number of unique metabolites under these conditions (Fig. 10, A and B). Notably, it did not impact serum orotate levels as was the case for the HMFD feeding (Fig. 10C). Thus, EMPA’s ability to restore this metabolite and to change expression of others is dependent on the HMFD feeding and the change in metabolic profile under these pathophysiological conditions.

Figure 10.

Empagliflozin (EMPA) changes metabolites based on diet consumed. In a separate cohort of mice, targeted metabolomics was performed in serum isolated from both male and female TallyHo mice fed a low-fat diet (LFD) in the presence or absence of EMPA. A: the 50 most changed metabolites (LFD vs. LFD + EMPA) are shown for both male (left) and female (right) in a heat map, with each column representing a different sample grouped by treatment. Blue colors indicated decreased expression, whereas red colors indicate increased expression. B: volcano plots were generated from the targeted metabolomics to compare the LFD vs. LFD + EMPA in both males (left) and females (right) with circles representing different metabolites. The y-axis indicates the P value as a log10 scale, and the x-axis indicates fold change (FC) in a log2 scale. Metabolites that had a FC > 2 and a P < 0.05 were deemed significant. Those that are shifted to the right (red circles) indicate metabolites that were increased with EMPA and those that shifted to the left (blue circles) were decreased with EMPA. C: normalized serum orotate levels from the HMFD (Figs. 7 and 8) or LFD metabolomics study are graphed. In male mice, EMPA only decreased orotate following consumption of a HMFD, whereas female mice showed a nonsignificant (P = 0.07) decrease in the LFD + EMPA group.

Orotate is an intermediate in the pyrimidine biosynthesis pathway, arising from the conversion from glutamine by enzyme complex carbamoyl-phosphate synthase 2/aspartic acid carbamoyl transferase/dihydroorotase (CAD) and dihydroorotate dehydrogenase (DHODH). HMFD feeding increased the expression of CAD (P = 0.049) and DHODH (P = 0.051) in male mice (compared with the low-fat control), which aligns with increased levels of orotate (Fig. 11, A and B). EMPA treatment trended toward reducing CAD levels (n/s; P = 0.19) but DHODH expression remained elevated (Fig. 11, C and F). No changes were observed for expression levels of carbamoyl-phosphate synthase 1 (CPS1) and ornithine transcarbamylase (OTC), two enzymes that feed into the urea cycle, a closely related metabolic pathway that is highly linked with pyrimidine biosynthesis (Fig. 11, D and E). Taken together, our data suggest that a HMFD feeding is sufficient to induce a shift toward pyrimidine biosynthesis, and this is partially restored upon treatment with EMPA (Fig. 11G).

Figure 11.

High-milk-fat diet (HMFD) feeding increases expression of enzymes involved in the pyrimidine biosynthesis pathway. Male (A) and female (B) livers from HMFD, empagliflozin (EMPA), and low-fat-fed mice (n = 4–8) were immunoblotted for enzymes involved in the pyrimidine biosynthesis pathway. C–F: densitometry quantification is plotted for each protein, with the y-axis reflecting the protein expression/total protein loading and the x-axis reflecting the three treatment groups for both males and females. A two-way ANOVA followed by Tukey’s post hoc test was performed with multiple comparison significance noted by P value on the graph. G: key steps in the pyrimidine biosynthesis pathway and the closely associated urea cycle are shown. Arrows indicate either increased, decreased, or no change with regard to EMPA treatment. CAD, carbamoyl-phosphate synthase 2/Asp carbamoyl transferase/dihydroorotase; CPS1, carbamoyl-phosphate synthetase 1; DHODH, dihydroorotate dehydrogenase; OTC, ornithine transcarbamylase.

DISCUSSION

Recently, SGLT2 inhibitors, including EMPA, have seen an explosion in their therapeutic potential. Although it was initially accepted that a reduction in glucose reabsorption in the kidney would promote glycosuria and improved glycemic control, their benefits are now observed across different tissues and pathophysiological states (9–11). Indeed, several of these drugs are now approved as treatments for nondiabetic conditions including chronic kidney disease (12, 33–35). Our group and others have focused on the impacts of SGLT2 inhibition on the liver. Previously, TallyHo mice treated for 12 wk with EMPA showed a reduction in steatosis alongside an expansion in white adipose tissue (14). Similar findings were reported in other mouse strains (db/db), implying that these protective effects may be universal (36, 37). After 12-wk of HMFD feeding, we also observed a decrease in circulating, but not hepatic, triglycerides. We also observed a decrease in hepatic cholesterol esters (14). After expanding the HMFD feeding to 24 wk, we observed many of the same findings. As for the 12-wk study, a 24-wk treatment with EMPA in the presence of the HMFD also decreased total cholesterol and the ester-to-free cholesterol ratio but did not change hepatic triglycerides (circulating triglycerides were not measured in this current study). Although it is clear that 24-wk HMFD feeding induced a significant accumulation of ectopic lipids including triglycerides, the impacts of EMPA treatment were mild and resulted in a slight decrease in total hepatic triglycerides and a significant reduction in hepatic cholesterol, phosphatidyl choline (18:1/20:2 and 14:0/14:0), phosphatidyl ethanolamine (14:0/18:1 and 16:0/20:5), and hydroxyceramide (26:0) (Figs. 2 and 3). Interestingly, we also noted that EMPA significantly increased certain triglyceride species including several with the fatty acid length of 22:6 (Fig. 3). Although counter-intuitive, studies have determined that the long-chain highly unsaturated fatty acids 22:6 correlate with improved insulin sensitivity and reduced body mass index (38). It is also enriched in the Mediterranean diet and a Jiangnan diet, both of which are positively associated with improved heath (38). Moreover, other studies have concluded that it is the ratio of lipids and not simply an overexpression of certain species that is a better predictor of liver health (39). Collectively, our data confirm that a HMFD significantly increases hepatic lipid accumulation, and this is somewhat mitigated with long-term treatment of EMPA.

With data from our laboratory and others suggesting a role for EMPA in mitigating hepatic steatosis, it has remained unclear if this was simply an outcome of better glycemic control [which we reported in these same mice previously (14)] or if treatment offers direct protection through the alteration of metabolic profiles. We previously noted that EMPA resulted in an expansion of nuchal white adipose tissue with an accompanying decrease in visceral adipose size (14). This phenotype was suggestive of a “healthy” adipose expansion, whereby an increase in white fat protects against exogenous lipid accumulation in tissues, including the liver. Although we did not directly test EMPA’s impacts on adiposity in the current study, the finding that EMPA increased circulating BHB levels (Fig. 7) supports the notion that similar changes took place. BHB has been shown to enhance white adipose function and induce de novo lipogenesis, hallmarks of healthy adipose tissue expansion (40). Moreover, BHB is synthesized in the liver during β-oxidation of fatty acids and can exert protective effects on the liver and extrahepatic tissues through several identified mechanisms, including the attenuation of inflammasome signaling, direct and indirect modification of lysine residues (BYBylation and lysine acetylation), and the activation of GPR109a signaling that has been linked to a decrease in triglyceride hydrolysis (41–44). Although the action of BHB in the context of EMPA treatment needs to be carefully explored, its elevation does provide a direct link between liver lipolysis, enhanced adipose function, and improved liver function.

The goal of this current study was to identify additional metabolic pathways that are altered with EMPA treatment and to correlate these to improved liver function. To examine this, we performed targeted metabolomics on male and female TallyHo mice and compared their metabolic profiles to mice fed a normal LFD. From these data, we discovered that EMPA does indeed alter the metabolic profile of these mice. In particular, it partially restored the circulating amino acid profile in both males and females, as well as dihydrofolate and orotate expression in males and acylcarnitines in females. Thus, our data highlight several potential pathways whereby EMPA may have a direct effect on physiology, including pyrimidine biosynthesis and the urea cycle.

In a recent meta-analysis, the global prevalence of MAFLD among the overweight population was found to be 70% (45). The most common form of obesity is termed polygenic obesity, whereby a combination of factors (mainly genetic and environmental) ultimately leads to the pathophysiological state (46). Despite the prevalence of polygenic obesity, most animal models that have been developed fall into the category of monogenic obesity, where a single genetic mutation gives rise to the condition. In this study, we used obese TallyHo mice, which recapitulate the polygenicity of obesity. Similar to the human population, this mouse line spontaneously develops elevated cholesterol, diabetes, and endothelial dysfunction (47, 48). Because of the incomplete penetrance of this phenotype, the mice in this study were challenged with a HMFD, selected for its high concentration of saturated fatty acids, mainly myristate and palmitate (49–51). This diet has been linked to insulin resistance, cardiometabolic dysfunction, and, importantly for this study, liver disease. Despite the propensity toward development of obesity and diabetes, previous data from our laboratory have shown that SGLT2i is an efficient therapeutic in the TallyHo mice (13, 14). It is important to point out that this study focused on the ability of EMPA to restore metabolite levels to that of control TallyHo mice. Although our findings that orotate and dihydrofolate are returned to “normal” levels, this is relative to the nonchallenged TallyHo mice. A separate metabolomic study also showed that these changes only occurred in the context of a HMFD indicating that the HMFD promotes a change in metabolism that can be partially corrected for with the addition of EMPA (Fig. 10). Future studies should examine the levels of orotate and dihydrofolate in more healthy strains of mice (i.e., C57BL6) to determine whether these metabolic changes are universal.

In the circulating metabolome, we identified that the HMFD significantly upregulates amino acids (induces amino acidemia) in male TallyHo mice and these were mostly reduced/restored upon treatment with EMPA. The increase in amino acids aligns well with findings that individuals with diabetes often present with amino acidemia, and increased amino acids are correlated with insulin resistance and visceral fat accumulation (22, 23). Amino acids also serve as substrates for the TCA cycle and are required for protein synthesis and acid-base regulation, suggesting that the HMFD feeding may lead to an overabundance of energy generation that is mitigated upon treatment with EMPA. Most of the circulating amino acids that are upregulated by HMFD feeding are glucogenic, indicating that the HMFD has a higher capacity for producing its own glucose in the kidney and liver, and this is reduced with EMPA. Although the restoration of amino acid levels in EMPA-treated mice was confined to the plasma (we did not detect the same degree of restoration in the livers), it does match previous data reported from our group that examined the metabolome on the renal cortex (13). In this study, we found that EMPA downregulated 14 of 20 amino acids in the kidney cortex, and this was possibly due to altered brush-border membrane reabsorption leading to a reduced energy use. In addition, a recent publication on the systemic metabolic communication that is impacted by dapagliflozin treatment found that SGLT2i led to an increase in aromatic amino acids and this was likely due to the reduction of gut microbiota that are involved in the metabolism of these amino acids (52). It will be important to determine whether the restoration of the amino acid levels observed in our study correlates with a reduction in gluconeogenesis and/or a change in the gut microbiota.

Folate (vitamin B) functions as a coenzyme in single carbon transfers for DNA synthesis and amino acid metabolism. It is also involved in methylation reactions and has been associated with an improvement in autophagy (via reduction of homocysteine) and protection from hepatotoxicity. Methylation reactions are highly prevalent in the liver, and a deficiency in folate has been linked to increased lipid accumulation and an increased risk of MAFLD, diabetes, and diabetic comorbidities (53, 54). Thus, our finding that EMPA partially restores hepatic dihydrofolate levels in TallyHo male mice, the reduced form of folate, unveils a possible protective mechanism linked to a mitigation of MAFLD. Dihydrofolate and tetrahydrofolate (not included in our targeted metabolomic screen) feed directly into the folate and methionine cycle to impact methylation and DNA synthesis. Liver expression of the folate transporter, HCP1, and folic acid levels themselves (the nonreduced version) were not found to be altered in any treatment group. However, EMPA did increase the folic acid → dihydrofolate → tetrahydrofolate enzyme, HFR (Fig. 5D). Whether the enzymatic activity of DHFR is also changed should be examined. It should be noted, however, that studies have found that mice and rats only express one form of the DHFR enzyme, whereas humans and nonhuman primates express two (55). Thus, these animal models are not suitable for studying the full effects of folate deficiency. In addition, the folate cycle is interconnected with the methionine cycle, and methionine metabolism is highly active in hepatocytes. An impaired methionine pathway is highly correlated with a number of liver disorders including alcoholic liver disease, MAFLD, and hepatitis, fibrosis, and cirrhosis (56). Clearly further examination of how EMPA impacts the methionine cycle is warranted.

In addition to dihydrofolate, our findings indicate that EMPA restores (reduces) circulating levels of orotate, a precursor for pyrimidine biosynthesis that interfaces with the urea cycle. In male TallyHo mice, we found that HMFD feeding significantly elevated circulating concentrations of orotate. Elevations in orotate are associated with ammonia toxicity and dysfunction of the urea cycle. Although we did not see any changes in hepatic expression of key urea cycle enzymes (OTC, CPS1), our previous metabolomic profile of the renal cortex revealed a downregulation of this cycle in the kidney (13), the other site of urea cycle activity. Interestingly, orotic acid has long been used to induce fatty liver in animal models and cell culture (57, 58). Elevated circulating levels of orotic acid are linked to increased expression of sterol regulatory element binding protein-responsive genes in the liver via decreased activity of AMPK (57, 58). Thus, the decrease in circulating levels of orotate may protect the mice from exogenous lipid accumulation at sites including the liver. Moreover, although the urea cycle did not appear to change in the TallyHo mice, we did see a change in the multienzyme CAD that catalyzes the three rate-limiting steps of pyrimidine synthesis and is required for orotate production. This enzyme was increased in HMFD-fed mice and partially downregulated on EMPA treatment matching our metabolomic data. Similarly, the enzyme that catalyzes the conversion of dihydroorotate to orotate (DHODH) was similarly increased under conditions of HMFD feeding. Although the expression of this enzyme was not reduced on EMPA treatment, we cannot rule out changes to enzyme activity.

Although we have unveiled several metabolic pathways that are changed with EMPA treatment, there are some limitations to the current work. Most notably, in this study, we fed the mice ad libitum rather than pair-feeding to control for dietary intake. In our previous study, we observed that addition of EMPA into the HMFD actually increased food intake (14), thus, we do not anticipate that EMPA treatment reduces dietary intake. Nonetheless, a change in dietary intake (either increased or decreased) could alter hepatic lipids and other circulating sugars. In this study, we observed a decrease in circulating levels of fructose with EMPA feeding (Fig. 7). Although the HMFD does not contain fructose and is actually formulated to induce obesity due to dietary fats, we cannot rule out the possibility that addition of EMPA into any diet will alter intake of sugars leading to liver improvements. Moreover, given the role of SGLT2 in transporting glucose, it would be interesting to see whether the metabolic pathways identified in this study are similarly changed in mice fed a high-sugar diet. Finally, we opted to define significance in our metabolomic datasets as those metabolites that were changed by greater than twofold with a P < 0.05. With such large datasets, controlling for the false discovery rate may be a more accurate representation of true significance.

In 2020, the term “NAFLD” was proposed to be reclassified as “metabolic dysfunction-associated fatty liver disease” or MAFLD (7). This name change, which reflects the growing prevalence of hepatic steatosis associated with T2D, obesity, and metabolic dysfunction, underscores the need to identify new therapeutics for this growing epidemic. Here, we find that EMPA mitigates hepatic steatosis following a HMFD feeding in obese TallyHo mice through the alteration of circulating and hepatic metabolites. This study is the first step toward identifying the mechanisms involving systemic protection with SGLT2is and may ultimately translate into appropriate and well-tolerated therapeutics to mitigate the detrimental impacts of NAFLD/MAFLD.

DATA AVAILABILITY

Data will be made available upon reasonable request.

GRANTS

This work was supported by the Dekkers Endowed Chair in Human Science (to B.D.S.) and by National Institutes of Health (NIH) Grant R03-DK123546 (to B.D.S.). The Proteomics & Metabolomics Shared Resource is partially supported by NIH Grant P30 CA051008. This study was also supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Grant K01DK106400 (to B.D.S.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.M.E. and B.D.S. conceived and designed research; I.M.E., E.N.D.C., and B.D.S. performed experiments; I.M.E., E.N.D.C., C.M.E., and B.D.S. analyzed data; C.M.E. and B.D.S. interpreted results of experiments; B.D.S. prepared figures; B.D.S. drafted manuscript; I.M.E., E.N.D.C., C.M.E., and B.D.S. edited and revised manuscript; I.M.E., E.N.D.C., C.M.E., and B.D.S. approved final version of manuscript.