A role for tubular Na⁺/H⁺ exchanger NHE3 in the natriuretic effect of the SGLT2 inhibitor empagliflozin

Abstract

Inhibitors of proximal tubular Na⁺-glucose cotransporter 2 (SGLT2) are natriuretic, and they lower blood pressure. There are reports that the activities of SGLT2 and Na⁺-H⁺ exchanger 3 (NHE3) are coordinated. If so, then part of the natriuretic response to an SGLT2 inhibitor is mediated by suppressing NHE3. To examine this further, we compared the effects of an SGLT2 inhibitor, empagliflozin, on urine composition and systolic blood pressure (SBP) in nondiabetic mice with tubule-specific NHE3 knockdown (NHE3-ko) and wild-type (WT) littermates. A single dose of empagliflozin, titrated to cause minimal glucosuria, increased urinary excretion of Na⁺ and bicarbonate and raised urine pH in WT mice but not in NHE3-ko mice. Chronic empagliflozin treatment tended to lower SBP despite higher renal renin mRNA expression and lowered the ratio of SBP to renin mRNA, indicating volume loss. This effect of empagliflozin depended on tubular NHE3. In diabetic Akita mice, chronic empagliflozin enhanced phosphorylation of NHE3 (S552/S605), changes previously linked to lesser NHE3-mediated reabsorption. Chronic empagliflozin also increased expression of genes involved with renal gluconeogenesis, bicarbonate regeneration, and ammonium formation. While this could reflect compensatory responses to acidification of proximal tubular cells resulting from reduced NHE3 activity, these effects were at least in part independent of tubular NHE3 and potentially indicated metabolic adaptations to urinary glucose loss. Moreover, empagliflozin increased luminal α-ketoglutarate, which may serve to stimulate compensatory distal NaCl reabsorption, while cogenerated and excreted ammonium balances urine losses of this “potential bicarbonate.” The data implicate NHE3 as a determinant of the natriuretic effect of empagliflozin.

INTRODUCTION

Na⁺-glucose cotransporter 2 (SGLT2) is expressed in the brush border of the early proximal tubule, where it reabsorbs ~97% of the filtered glucose in normoglycemia (56). Inhibitors of SGLT2 are new antihyperglycemic drugs that have protective effects on the kidney and heart in patients with type 2 diabetes (37, 66, 75). Pleiotropic effects of SGLT2 inhibition likely contribute to these beneficial effects (62). This includes natriuretic and diuretic effects that can lower blood pressure and are, at least in part, due to direct inhibition of Na⁺ uptake via SGLT2 in the early proximal tubule as well as the secondary osmotic effect of delivering high amounts of glucose to the further downstream tubule segments (27).

In the brush border of the small intestine, glucose uptake through Na⁺-glucose cotransporter 1 (SGLT1) can promote apical Na⁺-H⁺ exchanger 3 (NHE3) translocation and activation and thereby facilitate postprandial glucose and Na⁺ absorption (31, 55, 74). In comparison, in the early proximal tubule of the kidney, NHE3 is coexpressed with SGLT2. NHE3 in the proximal tubule is responsible for the reabsorption of a large fraction of the filtered Na⁺ and, secondarily, of Cl⁻ and fluid; moreover, NHE3 is important for proximal tubular bicarbonate reabsorption and ammonia secretion and thereby contributes to acid-base regulation (3, 19, 29, 33, 39, 48, 61, 69). In analogy to an SGLT1-NHE3 interaction in the small intestine, a functional relationship may exist between SGLT2 and NHE3 in the early proximal tubule.

Evidence is accumulating that transport processes in the early proximal tubule are coordinated. An example is the proposed coordinated stimulation of SGLT2, NHE3, and urate transporter 1 by insulin (56). A potential functional link between NHE3 and SGLT2 in the proximal tubule has also been suggested by the observation that glucose induced the regulation of both NHE activity and SGLT2 expression in human embryonic kidney cells (1). The scaffolding protein membrane-associated protein 17 kDa (MAP17) strongly enhances SGLT2 activity and may link SGLT2 and NHE3 regulation in the proximal tubule (7, 8). Furthermore, luminal glucose enhanced NHE activity during stationary perfusion studies in rat proximal tubules, and the dual SGLT1/SGLT2 inhibitor phlorizin inhibited NHE activity and bicarbonate reabsorption in rat proximal tubules and enhanced urinary bicarbonate excretion (41). Finally, a preliminary study in mice have reported that the SGLT2 inhibitor empagliflozin acutely increased the urinary anion gap, potentially reflecting increased bicarbonate excretion (17).

Therefore, the goal of the present studies was to gain further insights into a potential role of NHE3 in the renal response to SGLT2 inhibition. To this end, a mouse model was used with knockdown of NHE3 along the tubular system (39). Study 1 determined whether knockdown of tubular NHE3 affects the acute natriuretic response to an SGLT2 inhibitor. SGLT2 inhibition shifts Na⁺/glucose reabsorption from SGLT2 in the early proximal tubule to SGLT1 in the downstream late proximal tubule (46). SGLT2 transports 1 Na⁺ per glucose, whereas SGLT1 transports 2 Na⁺ per glucose (70). Thus, under appropriate conditions, SGLT2 inhibition theoretically can be antinatriuretic by shifting Na⁺-glucose cotransport to downstream SGLT1, unless another natriuretic mechanism is operating. To maximize the proposed SGLT1 effect, a dose of the SGLT2 inhibitor empagliflozin was tested that just saturated glucose transport via SGLT1 and therefore induced minimal glucosuria and osmotic effects.

Study 2 determined the role of tubular NHE3 for the chronic effect of the SGLT2 inhibitor empagliflozin. This study builds on a previously published data set that determined the role of tubular NHE3 in nondiabetic mice and Akita diabetes (39). This data set can be summarized as follows: mice with tubule-specific NHE3 knockdown (NHE3-ko mice) had higher urine pH and bicarbonate excretion; this was associated with compensating upregulation of renal mRNA expression of genes involved in proximal tubule generation of ammonium, glucose, and bicarbonate as well as distal tubule H⁺ and ammonia secretion. This compensation left blood pH and bicarbonate in the normal range in nondiabetic and diabetic NHE3-ko mice. NHE3 knockdown altered urine tricarboxylic acid (TCA) cycle metabolites in nondiabetic mice consistent with a response to metabolic acidosis, and upregulated renal proinflammatory markers. Moreover, NHE3-ko was associated with lesser kidney weight and glomerular filtration rate (GFR) independent of diabetes. NHE3-ko, however, did not attenuate hyperglycemia or prevent diabetes from increasing kidney weight and GFR. Building on the above data, study 2 aimed to determine the role of NHE3 in the chronic effects of empagliflozin on volume status but also explored effects on acid-base status including renal expression of related genes and urine metabolomics. Like study 1, study 2 was performed in nondiabetic mice to exclude the potentially confounding blood glucose-lowering effect of SGLT2 inhibition, but also included Akita diabetic mice.

Third, in study 3, we determined in a set of nondiabetic and Akita diabetic mice, whether empagliflozin enhanced the renal phosphorylation of NHE3, which has previously been associated with natriuretic effects (10, 45).

METHODS

Animals

All animal experimentation was conducted in accordance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals (NIH, Bethesda, MD) and was approved by the local Institutional Animal Care and Use Committee. For study 1, nondiabetic C57BL/6 mice with NHE3 knockdown along the entire tubular system (NHE3-ko) were used (39). These mice had been generated by crossing floxed NHE3 mice (29) and mice heterozygously expressing Cre recombinase under the control of the Pax8 promoter (4). For study 2 and as previously described (39), the above line was crossed with Akita mice (Ins^2+/C96Y; Akita/+; on C57BL/6 background), a genetic murine model of insulin-dependent type I diabetes mellitus (The Jackson Laboratories, Bar Harbor, ME). For study 3, C57BL/6 Akita mice and their littermate nondiabetic control mice were used (57). Male adult mice were used for experiments and housed in the same animal room with a 12:12-h light-dark cycle and free access to standard rodent chow and tap water.

Studies

Study 1: determination of the acute natriuretic effect of the SGLT2 inhibitor empagliflozin in the absence of tubular NHE3.

The dose of the SGLT2 inhibitor empagliflozin to just saturate SGLT1-mediated glucose reabsorption has been determined in a previous metabolic cage study (46) that compared the glucosuric response to empagliflozin in wild-type (WT) and SGLT1 knockout mice (Fig. 1). This empagliflozin dose of 0.3 mg/kg was estimated to inhibit ~30–40% of SGLT2.

Fig. 1.

Absence of tubular Na⁺/H⁺ exchanger 3 (NHE3) unmasked an acute antinatriuretic effect of the Na⁺-glucose transporter 2 (SGLT2) inhibitor empagliflozin when dosed to just saturate Na⁺-glucose transporter 1 (SGLT1). A: SGLT2 inhibition shifts Na⁺/glucose reabsorption downstream to SGLT1 in the late proximal tubule. SGLT2 and SGLT1 transport 1 and 2 Na⁺ per glucose, respectively. Thus, SGLT2 inhibition could be antinatriuretic due to downstream SGLT1, unless an additional natriuretic mechanism is operating. B: to maximize such an SGLT1 effect, a dose of the SGLT2 inhibitor was picked (0.3 mg/kg) that just saturates glucose transport via SGLT1, as determined in a previous metabolic cage study comparing the glucosuric effect of empagliflozin in wild-type (WT) and SGLT1 knockout mice (46). The difference between the two curves reflects glucose reabsorption via SGLT1 in WT mice for a given empagliflozin dose. C and D: in WT mice, this low dose of empagliflozin induced a small increase in glucosuria (consistent with SGLT1 saturation; compare with maximum glucosuria in B) associated with an increase in urinary pH and flow rate and urinary excretion of bicarbonate, Na⁺, and Cl⁻. Empagliflozin induced a similar low-level increase in glucosuria in mice with tubule-specific NHE3 knockdown (NHE3-ko mice) but did not increase pH or bicarbonate excretion and reduced urinary Na⁺ and Cl⁻ excretion. Values are means ± SE; n = 8–12 mice/group. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) further identified the significant effects. #P < 0.05 vs. WT mice; *P < 0.05 vs. vehicle.

NHE3-ko and littermate WT mice (9–15 wk of age, no difference in mean age between groups) were randomized to application of empagliflozin (0.3 mg/kg by oral gavage; provided by Boehringer Ingelheim) or vehicle (water). After the bladder was emptied, mice were water loaded by oral gavage (25 µl/g body wt, including empagliflozin or vehicle) to facilitate subsequent timed quantitative urine collection in metabolic cages over 3 h without access to food or water (46). Urine was analyzed for pH and concentrations of Na⁺, Cl⁻, and bicarbonate (see details below). Blood glucose measurements before and after urine collection by tail snip revealed no significant differences between groups (data not shown).

Study 2: determination of the chronic effect of SGLT2 inhibitor on volume and acid-base status in nondiabetic and Akita diabetic NHE3-ko mice.

NHE3-ko and littermate WT, with or without Akita, were randomized to application of empagliflozin (150 mg/kg of diet) or vehicle starting at 4 wk of age for 20 wk (total of 8 groups). All groups were studied at the same time. Part of the data in groups that received vehicle were previously published in a study to determine the role of tubular NHE3 in the diabetic kidney (39). Briefly, at 17–18 wk of age, blood pressure was measured in awake mice by an automated tail cuff system (BP-2000 Blood Pressure Analysis System, Visitech-Systems, Apex, NC) for 6 consecutive days after appropriate training (39). At 18–20 wk of age, GFR was measured in awake mice by plasma elimination kinetics of FITC-sinistrin (Fresenius-Kabi, Linz, Austria) following retroorbital injection of this GFR marker (39). Three to five days later, blood glucose was determined by tail snip, spot urine was collected, body weight was determined, and the animal was anesthetized with isoflurane followed by retroorbital blood collection and subsequent kidney tissue harvesting as previously described (39). All sample collections were performed between 9 and 11 AM.

Study 3: determination of the effect of SGLT2 inhibitor on expression of total and phosphorylated NHE3 in renal membranes in nondiabetic and diabetic Akita mice.

This series included nondiabetic and Akita-diabetic C57BL/6 mice treated with empagliflozin (300 mg/kg of diet) or vehicle starting at 4 wk of age for 15 wk, when the kidneys were harvested. The general characteristics and kidney harvesting of this mouse set have been previously published (57). Renal membranes were prepared in the presence of protease and phosphatase inhibitors, as previously described (45). Immunoblot analysis was performed at 4°C overnight with the primary NHE3 antibody (Millipore, Billerica, MA), phosphorylated S552-NHE3 antibody (Novus Biologicals, Littleton, CO), and phosphorylated S605-NHE3 antibody (Santa Cruz Biotechnology, Santa Cruz, CA) diluted 1:1,000, 1:1,000, and 1:200, respectively (45). The latter two antibodies recognize NHE only when S552 or S605 is phosphorylated (25). Chemiluminescent detection was performed using a 1:5,000 dilution of enhanced chemiluminescent (ECL) donkey anti-rabbit IgG and anti-mouse IgG linked to horseradish peroxidase and ECL detection reagent (GE Healthcare, Buckinghamshire, UK). The membrane was stripped (0.2 M NaOH for 5 min) and reprobed with anti-β-actin antibody (Sigma-Aldrich) for normalization.

Blood and Urine Analysis

Blood glucose was determined using the Ascensia Elite XL glucometer (Bayer, Mishawaka, IN). If levels were “high” (>600 mg/dL), glucose concentration was determined by the hexokinase/glucose-6-phosphate dehydrogenase method (Infinity, Thermo Electron, Louisville, CO), which was also used for all urine samples. Urine albumin (Bethyl Laboratories, Montgomery, TX) and ammonium (Pointe Scientific, Canton, MI) were determined using commercial assays. Acid-base analyses were performed in freshly collected blood and urine as previously described (39); bicarbonate concentrations were calculated from pH and CO₂ concentrations measured by a blood gas analyzer (OPTI Medical Systems, Inc., Roswell, GA) according to the Henderson-Hasselbalch equation. Urine Na⁺ and Cl⁻ concentrations were measured using an electrolyte analyzer (EasyElectrolytes, Medica Corporation, Bedford, MA). Plasma and urine phosphate concentrations were measured using the QuantiChrom Phosphate Assay Kit (BioAssay Systems). Concentrations in spot urine were normalized to creatinine concentrations measured by kinetic analysis of Jaffe's reaction (ThermoFisher Scientific, Waltham, MA).

Western Blot Analysis for Renal SGLT2 and NHE3 Expression

Western blot analysis was performed using the membrane fraction extracted from whole kidneys as previously described (39, 57, 58). Protein concentration was determined using a DC Protein Assay kit (Bio-Rad Laboratories, Hercules, CA). Then, 40–60 µg/lane of extracted proteins were resolved on NuPAGE gels in MOPS buffer (Invitrogen, Carlsbad, CA). Gel proteins were transferred to nitrocellulose membranes and immunoblotted with rabbit anti-NHE3 antibody (EMD Millipore, Darmstadt, Germany) (45), rabbit anti-SGLT2 antibody (57, 59), and mouse anti-β-actin antibody (Sigma-Aldrich, St. Louis, MO) (45) followed by treatment with horseradish peroxidase-conjugated secondary antibody. Protein expression was detected autoradiographically by ECL Plus kit (Amersham Pharmacia, Piscataway, NJ). Densitometric analysis was performed using ImageJ Software (NIH). Results were normalized to β-actin expression and the control WT group. Whole kidney expression was also estimated by considering kidney weight because diabetes had a strong effect on the latter.

Quantitative RT-PCR

Quantitative RT-PCR was performed as previously described (39). Briefly, frozen kidneys were homogenized in lysis buffer provided by the RNeasy Plus Mini Kit (catalog no. 74136, Qiagen), and RNA was purified. cDNA was prepared using the SuperScript IV First-Strand Synthesis System (catalog no. 18091050, Thermo Fisher Scientific). Real-time PCR was performed in a 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA) using Taqman assays (catalog no. 4331182, ThermoFisher Scientific) as shown in Table 1. Expression levels were expressed as relative fold increases/decreases normalized to the average of two housekeeping genes, β-actin and ribosomal protein L19 (rpl19), which were not different between groups (not shown). Each experiment was performed in duplicate.

Table 1.

Real-time PCR primers used

Target Gene	Assay ID
Actb	Mm00607939_s1
Atp6v1b	Mm00460309_m1
Ccl2	Mm00441242_m1
Cd14	Mm00438094_g1
Epo	Mm01202755_m1
Glud1	Mm00492353_m1
Gls	Mm01257297_m1
Hif1a	Mm00468869_m1
Pepck	Mm01247058_m1
Pfkp	Mm00444792_m1
Renin	Mm02342889_g1
Rhcg	Mm00451199_m1
Rpl19	Mm02601633_g1
Tnfa	Mm00443258_m1

See text for abbreviations.

Urine Metabolomics

Urine metabolites were measured as previously described (23, 39, 73). Briefly, aliquots corresponding to 50 nmol creatinine were derivatized and analyzed by GC-MS/MS. A mix of 10 heavy isotope labeled internal standards, corresponding to different groups of metabolites under analysis, were added to aliquots of urine samples before derivatization. Keto acids were oximated with pentafluorobenzylhydroxylamine. Following overnight lyophilization, organic acids were isolated by liquid partition chromatography on silica (45% 2-methyl-2-butanol in chloroform). The eluted samples were evaporated under N₂, and the dry residue was silylated with 1:1 volumes of N,O-bis(trimethylsilyl)trifluoroacetamide and Tri-Sil HTP for 2 h at 60°C. One microliter of sample was applied onto a 20 m × 0.36 mm column (Agilent DB-5) at 250°C in a gas chromatogram and eluted with helium gas at a temperature gradient of 80−300°C over 20 min. Target metabolites were analyzed by electron ionization and detected on an extended dynamic range detector on a triple-quadrupole mass spectrometer (Bruker Scion). Each metabolite was quantified using an 8-point calibration standard curve (0.1−150 nmol) with R2 values above 0.98. The ratio of the metabolite peak area to the internal standard peak area was used to determine metabolite concentration.

Statistical Analysis

Data are presented as means ± SE. To analyze for statistical differences between groups, two-way ANOVA was performed to probe for a significant effect of the two specific factors (study 1 and study 2: NHE3-ko and empagliflozin, study 2: metabolomics data, and study 3: Akita and empagliflozin) and for the interaction between the two factors. If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) was used to identify the significant effects. Values that were more than two SDs above or below the average were considered outliers and excluded from analysis. P values of <0.05 were considered statistically significant.

RESULTS

Study 1: Absence of Tubular NHE3 Unmasked an Antinatriuretic Effect of the SGLT2 Inhibitor Empagliflozin When Dosed to Just Saturate Glucose Reabsorption by SGLT1

In metabolic cage experiments, the tested dose of empagliflozin (0.3 mg/kg), chosen to just saturate glucose reabsorption via SGLT1 (46) (Fig. 1, A and B), induced only a minor glucosuric effect in WT mice (Fig. 1C); this was associated with an increase in urine pH, flow rate, and excretion of bicarbonate, Na⁺, and Cl⁻ (Fig. 1D). Urine pH and bicarbonate excretion were higher in vehicle-treated NHE3-ko versus WT mice, consistent with a role of NHE3 in renal bicarbonate reabsorption. In the absence of tubular NHE3, empagliflozin induced a similarly small glucosuric effect but did not increase urinary pH or bicarbonate excretion. Furthermore, the diuretic effect of empagliflozin was blunted, and empagliflozin, in contrast to WT mice, lowered urinary Na⁺ and Cl⁻ excretion in NHE3-ko mice (Fig. 1D).

Study 2: Effect of Chronic Application of Empagliflozin on Volume and Acid-Base Status in Nondiabetic and Akita Diabetic Mice: Role of Tubular NHE3

The goal of the present studies was to compare chronic effects of SGLT2 inhibition with NHE3-ko and to probe statistically for potential interactions. The analysis was performed separately for nondiabetic and Akita diabetic mice since the confounding blood glucose-lowering effect of empagliflozin can complicate data interpretation in Akita diabetic mice.

Evidence for overall similar and at least in part independent effects of chronic empagliflozin and NHE3-ko on renal mRNA expression of genes involved in glutamine metabolism, gluconeogenesis, and acid-base handling in nondiabetic mice.

Empagliflozin did not significantly alter renal NHE3 protein expression (Fig. 2A) or blood pH or bicarbonate concentrations (Fig. 2B) in any of the groups versus vehicle application. However, in nondiabetic control mice, empagliflozin modestly lowered urine pH in both WT and NHE3-ko mice (Fig. 2B). In NHE3-ko mice, this effect of empagliflozin was associated with a modest reduction in urinary bicarbonate-to-creatinine ratios (Fig. 2B) and a trend (P = 0.065) for higher renal mRNA expression of the B1 subunit of H⁺-ATPase (ATP6V1B1; Fig. 3A), which contributes to H⁺ secretion into the lumen of the distal nephron, thereby facilitating bicarbonate reabsorption at this site (15). Empagliflozin enhanced renal mRNA expression of phosphoenolpyruvate carboxykinase (PEPCK; Fig. 3A), a principal gluconeogenic enzyme, the activity of which is regulated at the level of transcription and contributes to both glucose and bicarbonate formation (Fig. 4) (44). Empagliflozin also enhanced renal mRNA expression of glutaminase (Fig. 3A), which, together with glutamate dehydrogenase 1, catalyzes the oxidative deamination of glutamine to glutamate and α-ketoglutarate, thereby contributing to ammonium formation (ammoniagenesis), a primary mechanism to enhance net renal acid excretion (51). Empagliflozin increased urinary α-ketoglutarate levels (Fig. 4). Moreover, empagliflozin increased renal mRNA expression of the ammonia transporter Rhesus C glycoprotein (Fig. 3B), which secretes ammonia into the lumen of the distal nephron (67). This was associated with higher urine ammonium-to-creatinine ratios in response to empagliflozin (Fig. 3C). The described findings are consistent with an empagliflozin-induced upregulation of proximal tubular glutamine metabolism leading to the formation of ammonia and α-ketoglutarate. α-Ketoglutarate formed by ammoniagenesis can be further processed by the TCA cycle and thereby contributes to the generation of both bicarbonate and glucose via upregulated PEPCK (Fig. 4). Newly formed ammonia is delivered to the distal nephron for secretion. H⁺ secretion by ATP6V1B1 facilitates both distal tubule bicarbonate reabsorption and ammonia secretion.

Fig. 2.

Chronic empagliflozin did not affect renal Na⁺/H⁺ exchanger 3 (NHE3) expression but modestly reduced urine pH in nondiabetic mice. A: renal membrane NHE3 protein expression was effectively knocked down in mice with tubule-specific NHE3 knockout (NHE3-KO mice; KO) and not affected by empagliflozin in control and Akita wild-type (WT) mice. Renal NHE3 protein expression was normalized to β-actin or, in addition, multiplied by kidney weight (KW) to estimate whole kidney expression since diabetic kidneys were significantly larger (n = 8–10 mice/group for WT mice and 4 mice/group for NHE3-KO mice). B: empagliflozin did not alter blood pH or bicarbonate but lowered urine pH in nondiabetic WT and NHE3-ko mice and reduced urinary bicarbonate-to-creatinine ratios in NHE3-ko mice. It may be impossible to further reduce the very low urinary bicarbonate values in vehicle-treated nondiabetic WT mice, i.e., the ANOVA cross term may not indicate a real interaction. Values are means ± SE; n = 13–28 mice/group. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. WT mice; *P < 0.05 vs. vehicle.

Fig. 3.

Evidence for similar and at least in part independent effects of chronic empagliflozin and Na⁺/H⁺ exchanger 3 knockout (NHE3-ko) on renal mRNA expression of genes involved in acid-base handling in nondiabetic mice. Renal mRNA expression normalized to β-actin and ribosomal protein L19 (rpl19) is shown (n = 8–10 mice/group). A and B: both empagliflozin and NHE3-ko enhanced or tended to increase (P < 0.1) renal mRNA expression of genes involved in proximal tubular gluconeogenesis [phosphoenolpyruvate carboxykinase (Pepck)] and in ammonia and α-ketoglutarate formation [glutaminase (Gls) or glutamate dehydrogenase (Glud1)] and in distal tubular H⁺ secretion [H⁺-ATPase B1 subunit (Atp6v1b1)] and ammonia secretion [Rhesus C glycoprotein (Rhcg)]. The effects of empagliflozin and NHE3-ko appeared at least in part additive or independent (no significant interactions), and part of the effects were also detectable in Akita mice. C: empagliflozin increased urinary ammonium to creatinine ratios in nondiabetic wild-type (WT) mice but attenuated the increase in Akita WT mice (n = 10–20 mice/group). Treatment of NHE3-ko mice with empagliflozin enhanced urinary phosphate-to-creatinine ratios independent of diabetes; empagliflozin caused a small increase in plasma phosphate concentrations in nondiabetic mice (n = 6–15 mice/group for plasma and n = 12–15 for urine). Values are means ± SE. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. WT mice; *P < 0.05 vs. vehicle.

Fig. 4.

Distinct effects of empagliflozin (empa) on urinary excretion of tricarboxylic acid (TCA) cycle metabolites in nondiabetic and diabetic mice. In nondiabetic mice, empagliflozin consistently enhanced urinary excretion of the TCA cycle metabolites α-ketoglutarate, fumarate, and malate. The Akita diabetic kidney potentially reduced mitochondrial energy metabolism, as indicated by reduced urinary excretion of aconitate, isocitrate, and succinate. In Akita diabetic mice, empagliflozin significantly increased urine excretion of four of seven TCA cycle metabolites. Na⁺/H⁺ exchanger 3 knockout [NHE3-ko; indicated for reference by gray bars (39)] induced very different and in part opposite responses than empagliflozin. Green arrows indicate the direction of significant effects of empagliflozin. Values are means ± SE; n = 5–6 mice/group. Two-way ANOVA was performed to probe for a significant effect of Akita diabetes (P_Akita), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. control; *P < 0.05 vs. wild-type (WT) mice.

Thus, the responses in renal expression of genes related to proximal tubular ammonia, glucose, and bicarbonate formation and H⁺ and ammonia secretion were largely changed in the same direction by NHE3-ko and empagliflozin. Hence, it appears these changes aim to compensate for impaired urine acidification and bicarbonate reabsorption in NHE3-ko but contribute to the enhanced urine acidification and potentially bicarbonate retention (see effect in NHE3-ko mice) in response to empagliflozin. The empagliflozin-induced changes in nondiabetic mice indicated a response to an increased acid load or formation. This may reflect metabolic adaptations to the urinary loss of glucose and calories, which can increase the formation of acidic ketone bodies (43) or enhance the dietary acid intake, as reflected by a trend (P = 0.051) for higher food intake in response to empagliflozin (Fig. 5B).

Fig. 5.

Distinct effects of empagliflozin (empa) on urinary metabolites that were previously associated with chronic kidney disease (CKD) or linked to rewiring of fuel preference to fats. Urinary metabolite excretion is given as nmoles of metabolite per 50 nmoles of creatinine. A: the eight metabolites previously found to be reduced (ratios to creatinine) in urine of patients with diabetes (50) or without diabetes (23) with CKD versus individuals with normal kidney function. Five of the eight metabolites were also lower in urine of wild-type (WT) Akita mice. Moreover, empagliflozin, but not Na⁺/H⁺ exchanger 3 knockout (NHE3-ko), increased urine excretion of all eight metabolites in Akita mice. With regard to other urine metabolites, some were changed in similar directions in response to empagliflozin and NHE3-ko (B), whereas others showed contrasting effects (C). Empagliflozin increased urinary azelaic acid and reduced urinary stearate and palmitate in nondiabetic and diabetic mice, which may reflect rewiring of the renal fuel preference to fats. Values are means ± SE; n = 6–8 mice/group. Two-way ANOVA was performed to probe for a significant effect of Akita diabetes (P_Akita), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. control; *P < 0.05 versus WT mice. TCA cycle, tricarboxylic acid cycle.

For none of the described changes in gene expression was the interaction term for the effects of NHE3-ko and SGLT2 inhibition (P_inter) statistically significant, indicating the responses were largely independent of each other. Urinary phosphate facilitates titratable acid excretion, which can serve as a compensatory response when renal ammonium handling is impaired or acid loading is enhanced. Notably, treatment of NHE3-ko mice with empagliflozin enhanced urinary phosphate-to-creatinine ratios independent of diabetes (Fig. 3C). On the other hand, empagliflozin induced a small increase in plasma phosphate concentrations that was significant in nondiabetic mice (Fig. 3C) and is consistent with observations in humans (2, 52).

Urine metabolomics revealed different patterns in response to empagliflozin and NHE3-ko in nondiabetic mice (Figs. 4 and 5). No samples had been collected for metabolomics in empagliflozin-treated NHE3-ko mice and thus no interactions could be tested. Therefore, metabolomics data have only been statistically analyzed and presented for effects of empagliflozin and Akita, whereas data for NHE3-ko are shown for reference (statistical analyses for NHE3-ko in Ref. 39). With regard to urinary excretion of TCA cycle metabolites in nondiabetic mice, NHE3-ko mice showed significantly increased excretion of citrate, aconitate, and isocitrate, whereas α-ketoglutarate, fumarate, and malate were reduced (indicated as the gray bars in Fig. 4) (39), a pattern consistent with a response to metabolic acidosis or a strong acid challenge (39). The response to empagliflozin was different, and empagliflozin, in contrast to NHE3-ko, enhanced urinary levels of α-ketoglutarate, fumarate, and malate in nondiabetic mice (Fig. 4). The differences in α-ketoglutarate metabolism have potential implications for the paracrine stimulation of distal transport processes, as further discussed below.

While NHE3-ko had no effect on urine excretion of TCA cycle metabolites in Akita mice (39), empagliflozin enhanced the urinary excretion of four of seven TCA cycle metabolites (Fig. 4). Figure 5A shows eight metabolites that are related to mitochondrial dysfunction and were found to be reduced (ratios to creatinine) in urine of diabetic (50) and nondiabetic (23) patients with chronic kidney disease (CKD) versus without CKD. Five of the eight metabolites were reduced in urine of WT Akita versus nondiabetic mice, and urine excretion of all eight metabolites was higher in empagliflozin-treated Akita mice versus vehicle (Fig. 5A). In contrast and except for one isoleucine metabolite, NHE3-ko did not alter urinary concentrations of these metabolites in Akita mice (gray bars in Fig. 5A) (39). With regard to other urine metabolites, some were changed in similar directions in response to empagliflozin and NHE3-ko (Fig. 5B), whereas others showed contrasting patterns (Fig. 5C). See further details in the discussion.

NHE3-ko, but not empagliflozin, significantly upregulated renal mRNA expression of the proinflammatory genes chemokine (C-C motif) ligand 2 (CCL2), CD14, or TNF-α in nondiabetic controls (Fig. 6A). The proinflammatory effect of NHE3-ko may reflect an increase in proximal tubular synthesis of ammonia that cannot exit across the luminal membrane, causing toxic intracellular or interstitial levels of ammonia (39).

Fig. 6.

Chronic empagliflozin and Na⁺/H⁺ exchanger 3 knockout (NHE3-ko) have different effects on gene expression of markers of kidney inflammation in nondiabetic mice. Renal mRNA expression normalized to β-actin and ribosomal protein L19 (rpl19) is shown (n = 8–10 mice/group). A: in contrast to NHE3-ko, empagliflozin did not significantly upregulate renal mRNA expression of the proinflammatory genes chemokine (C-C motif) ligand 2 (Ccl2), Cd14, and TNF-α (Tnfa). B: both empagliflozin and NHE3-ko shift reabsorption to distal segments, where Na⁺ reabsorption requires more energy and glucose is a preferred energy substrate (56). Both empagliflozin and NHE3-ko enhanced or tended to increase (P = 0.072) renal mRNA expression of phosphofructokinase platelet isoform (Pfkp), a key glycolytic enzyme in distal tubules, as well as hypoxia-inducible factor-1α (Hif1a). This was associated with elevated mRNA expression of erythropoietin (Epo) in response to empagliflozin but not NHE3-ko. Values are means ± SE. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. wild-type (WT) mice; *P < 0.05 vs. vehicle.

Shifting transport to distal segments.

Based on mathematical modeling (27, 28) and experimental data (38, 39), both SGLT2 inhibition and NHE3-ko are expected to shift Na⁺ and/or glucose reabsorption downstream of the early proximal tubule and enhance oxygen consumption in distal tubular segments, which prefer glucose as an energy substrate (56). Both empagliflozin and NHE3-ko independently (no significant interaction) enhanced renal mRNA expression of phosphofructokinase platelet isoform (Fig. 6B), which is a rate-limiting enzyme of glycolysis, primarily expressed in distal tubules and the collecting duct (30). Moreover, empagliflozin enhanced renal mRNA expression of hypoxia-inducible factor-1α and erythropoietin in nondiabetic mice (Fig. 6B). A modeling study has proposed that such changes in response to SGLT2 inhibition reflect enhanced transport work in distal segments (27). NHE3-ko also tended (P = 0.072) to increase renal hypoxia-inducible factor-1 mRNA expression but did not enhance erythropoietin mRNA expression. Possible mechanism include the enhanced renal inflammatory processes in NHE3-ko, which can lower erythropoietin production (13, 72), including an inhibitory effect of TNF-α (Fig. 6A) (20).

SGLT2 expression and glucosuria.

Empagliflozin enhanced renal SGLT2 protein expression in nondiabetic and diabetic WT mice (Fig. 7A), confirming a previous study (57). This also occurred in the absence of tubular NHE3 and may reflect negative feedback regulation of SGLT2 expression by apical glucose uptake and the resulting reduction in cytosolic glucose concentrations (56). In contrast, NHE3-ko is associated with reduced renal SGLT2 expression (Fig. 7A) (39), which may reflect the increase in gluconeogenesis and thereby enhanced cytosolic glucose concentrations. This effect of NHE3-ko was also evident during empagliflozin treatment as well as in nondiabetic and diabetic mice and was associated with higher PEPCK expression in all groups (Fig. 3A).

Fig. 7.

Opposite effects of empagliflozin and Na⁺/H⁺ exchanger 3 knockout (NHE3-ko) on renal Na⁺-glucose cotransporter 2 (SGLT2) protein expression. A: empagliflozin increased and NHE3-ko reduced renal SGLT2 protein expression in nondiabetic and Akita diabetic mice (with no significant interaction) when normalized to β-actin or, in addition, multiplied by kidney weight (KW) to estimate whole kidney expression (n = 7–10 mice/group). To illustrate the observed expression range, * shows a sample with particularly high SGLT2 expression; while this mouse fits the observation that empagliflozin-treated wild-type (WT) Akita mice had the highest SGLT2 expression, the statistical analysis identified it as an outlier that was not used for the analysis. Figures 2A and 7A are from the same experiment and have the same loading control. B: in nondiabetic mice, empagliflozin increased glucosuria, associated with higher levels of food and fluid intake, while blood glucose was unchanged. These effects of empagliflozin were similar in NHE3-ko mice. In WT Akita mice, empagliflozin strongly reduced blood glucose associated with lesser glucosuria and fluid intake versus vehicle, and similar effects were observed in NHE3-ko mice. See text for a further discussion. n values for nondiabetic and Akita diabetes were as follows: for glucosuria, n = 7–13 and 8–12 mice/group, respectively; for blood glucose, n = 8–21 and 16–24 mice/group, respectively; for food and fluid intake, n = 12–24 mice/group for nondiabetic and Akita diabetes. Values are means ± SE. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. WT mice; *P < 0.05 vs. vehicle.

Empagliflozin enhanced glucosuria to the same extent in nondiabetic WT and NHE3-ko mice (Fig. 7B). This occurred without changing blood glucose (Fig. 7B) and GFR (Fig. 8A) and thus filtered glucose, consistent with similar inhibition of renal glucose reabsorption by empagliflozin in nondiabetic WT and NHE3-ko mice. Empagliflozin also increased fluid intake and tended to increase food intake similarly in nondiabetic WT and NHE3-ko mice (Fig. 7B), consistent with similar glucosuria-induced osmotic diuresis and urine calorie loss.

Fig. 8.

Evidence for an interaction between empagliflozin and Na⁺/H⁺ exchanger 3 knockout (NHE3-ko) with regard to the effect on volume status in nondiabetic mice. A: effects on kidney weight, glomerular filtration rate (GFR), and albuminuria; see text for details. B: systolic blood pressure (BP), renal renin mRNA expression, and their ratio were used as indicators of volume status. In nondiabetic mice, empagliflozin tended to lower systolic BP despite higher levels of renal mRNA expression of renin, which is expected to increase BP, and both effects seemed to be restricted to wild-type (WT) mice. The ratio of systolic BP to renal renin mRNA expression revealed a significant interaction (P_inter = 0.006) between empagliflozin and NHE3-ko, such that empagliflozin reduced this ratio, but only in WT mice, i.e., in the presence of tubular NHE3. An overall similar pattern was observed in Akita mice for systolic BP and renin expression, but the interaction term did not achieve statistical significance, potentially due to the confounding effect on blood glucose. n = 12–28 mice/group for body and kidney weight, n = 10–26 mice/group for GFR, n = 9–16 mice/group for albuminuria, n = 14–26 mice/group for systolic BP, and n = 8–10 mice/group for renin mRNA and the ratio. Values are means ± SE. Two-way ANOVA was performed to probe for a significant effect of NHE3-ko (P_NHE3), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. WT mice; *P < 0.05 vs. vehicle.

In Akita mice, empagliflozin reduced blood glucose levels to the same extent in WT and NHE3-ko mice (Fig. 7B). NHE3-ko enhanced glucosuria in vehicle-treated Akita mice. This has been proposed to reflect downregulation of SGLT2 expression and potentially enhanced renal gluconeogenesis with glucose spilling across the luminal membrane (39, 56). The effect of NHE3-ko to increase glucosuria in Akita mice was numerically attenuated during empagliflozin treatment (P_inter = 0.075), which is consistent with a role of SGLT2 downregulation.

Effects on kidney weight and GFR.

In nondiabetic mice, empagliflozin did not affect body weight but increased kidney weight, particularly in NHE3-ko mice (Fig. 8A). An increase in kidney weight by empagliflozin or SGLT2 knockout in nondiabetic mice has been previously reported (57, 59). Empagliflozin did not affect GFR or albuminuria in nondiabetic WT or NHE3-ko mice, whereas NHE3-ko reduced GFR and increased albuminuria in the absence and presence of empagliflozin with no significant interaction (Fig. 8A). The reduction in GFR in NHE3-ko is thought to reflect activation of tubuloglomerular feedback, whereas increased albuminuria may reflect a role of NHE3 in proximal tubule albumin reuptake (18).

In Akita NHE3-ko mice, the empagliflozin-induced improvement in glycemic control and the reduction in renally filtered and excreted glucose were associated with an increase in body weight (Fig. 8A). Whereas NHE3-ko lowered kidney weight, particularly in vehicle-treated Akita mice (P_inter = 0.055), the effect of empagliflozin did not reach significance. NHE3-ko also reduced GFR in Akita mice, independent of empagliflozin treatment. Empagliflozin did not significantly alter GFR when expressed in absolute terms (Fig. 8A) and lowered GFR when related to body weight (data not shown). Both empagliflozin and NHE3-ko lowered or normalized albuminuria in Akita mice (Fig. 8A).

Are effects of empagliflozin on volume status dependent on tubular NHE3?

Systolic blood pressure and renal renin mRNA were used as indicators of volume status. In nondiabetic mice, empagliflozin overall tended to lower systolic blood pressure (P = 0.098) despite significantly higher renal mRNA expression of renin, and both effects seemed largely due to effects in WT mice rather than in NHE3-ko mice, but the two-way ANOVA interaction terms did not reach statistical significance (P = 0.194 and 0.209; Fig. 8B). Since an increase in renal renin expression is expected to attenuate any volume- and blood pressure-lowering effect, the ratio of systolic blood pressure to renal renin mRNA expression was calculated to take this compensation into account. Doing so revealed a highly significant interaction (P_inter = 0.006) between empagliflozin and NHE3-ko, such that empagliflozin reduced this ratio, but only in WT mice, i.e., in the presence of tubular NHE3 (Fig. 8B). The data indicate that effects of empagliflozin on volume status were at least in part dependent on tubular NHE3. An overall similar pattern was observed in Akita mice for systolic blood pressure and renin expression, but the interaction term did not achieve statistical significance, which may in part reflect the confounding effect of lowering blood glucose levels on volume status.

Study 3: Chronic Application of Empagliflozin Enhances the Expression of Phosphorylated NHE3 Protein Expression in Akita Diabetic Mice

The renal phenotype of this mouse set has previously been published (57) and can be summarized as follows: empagliflozin, which was applied in a higher dose than in study 2 (300 vs. 150 mg/kg diet), also reduced blood glucose to lower levels in Akita mice (200 vs. 300 mg/dL). Moreover, high dose empagliflozin significantly reduced kidney weight, GFR, and systolic blood pressure (57). For study 3 and since the kidneys had been homogenized with phosphatase inhibitor, the renal protein expression and phosphorylation of NHE3 were determined. Empagliflozin did not alter total renal NHE3 protein expression in nondiabetic control or Akita mice (Fig. 9, A and B) and increased the expression of NHE3 phosphorylated at S605 (Fig. 9, C and D) and S552 (Fig. 9, E and F) only in Akita mice.

Fig. 9.

Chronic empagliflozin enhanced Na⁺/H⁺ exchanger 3 (NHE3) phosphorylation in Akita mice. This series of mice was treated with empagliflozin (300 mg/kg of diet) or vehicle starting at 4 wk of age for 15 wk. The general characteristics of this mouse set has been previously published (57). Here, it was observed that empagliflozin did not alter total renal NHE3 protein expression (A and B) but increased expression of NHE3 phosphorylated at S605 (C and D) and S552 (E and F) in Akita mice, an effect not observed in nondiabetic mice. Values are means ± SE; n = 8–9 mice/group. Two-way ANOVA was performed to probe for a significant effect of Akita diabetes (P_Akita), empagliflozin (P_empa), or the interaction between the two factors (P_inter). If the interaction was statistically significant, then a pairwise multiple-comparison procedure (Holm-Sidak method) identified the significant effects. #P < 0.05 vs. wild-type (WT) mice; *P < 0.05 vs. vehicle.

DISCUSSION

The present studies provide support in murine models for a functional role of tubular NHE3 in the acute natriuretic and chronic volume effect of the SGLT2 inhibitor empagliflozin. Study 1 indicated a role for tubular NHE3 in the acute natriuretic effect of empagliflozin. Study 2 provided evidence that also the chronic effect of empagliflozin on a marker of volume status was affected by the presence of tubular NHE3. Both of these observations were made in nondiabetic mice, where the potential confounding effect of SGLT2 inhibition on blood glucose is null or minimal. Study 3 showed in a genetic model of type 1 diabetes that chronic empagliflozin enhanced the renal expression of NHE3 phosphorylated at S552 and S605, changes previously associated with natriuresis (10, 45).

The finding of study 1 that the SGLT2 inhibitor empagliflozin can induce a natriuretic response and increase in bicarbonate excretion, which both depend on tubular NHE3, is not unprecedented. The dual SGLT1/SGLT2 inhibitor phlorizin has been previously shown to inhibit NHE activity and bicarbonate reabsorption in rat proximal tubules and enhance urinary bicarbonate excretion (41). SGLT2 inhibition by empagliflozin, titrated to just saturate glucose reabsorption via downstream SGLT1, increased urinary pH and the excretion of fluid, Na⁺, and bicarbonate in WT animals. These effects were absent in NHE3-ko animals, which actually showed an antinatriuretic response, and the statistical analysis indicated a significant interaction between empagliflozin and NHE3-ko. It is proposed that shifting glucose reabsorption from SGLT2 to SGLT1 may have induced an antinatriuretic effect due to their Na⁺-to-glucose transport ratios of 1:1 versus 2:1, respectively, and that this effect was unmasked in the absence of the natriuretic effect via NHE3 inhibition. A smaller natriuretic effect could, in part, be due to the lesser renal SGLT2 expression observed in mice lacking tubular NHE3, but this in itself is not expected to cause an antinatriuretic effect of SGLT2 inhibition. Moreover, the picked empagliflozin dose was estimated to have inhibited only ~1/3 of SGLT2 in WT mice (46). Furthermore, the glucosuric response was similar in response to acute and chronic empagliflozin in nondiabetic NHE3-ko versus WT mice, indicating a similar net inhibition of SGLT2 transport proteins between genotypes.

Study 2 indicated that chronic empagliflozin and NHE3-ko induced a similar natriuretic influence in nondiabetic mice as indicated by similar compensatory reductions in the ratio of systolic blood pressure to renal renin mRNA expression. The latter serves to limit urinary losses of glucose, NaCl, and bicarbonate when tubular function is impaired. Moreover, the effects of these two maneuvers were not additive, and the statistical analysis indicated a significant interaction between empagliflozin and NHE3-ko, consistent with the notion that part of the effect of empagliflozin on this volume marker depended on intact NHE3 (Fig. 10).

Fig. 10.

Proposed effects of Na⁺-glucose cotransporter 2 (SGLT2) inhibitor empagliflozin in the nondiabetic kidney and its interaction with Na⁺/H⁺ exchanger 3 (NHE3) on Na⁺ reabsorption. Empagliflozin reaches its target, SGLT2, in the lumenal brush border of the early proximal tubule (S1/S2 segment) by glomerular filtration and tubular secretion (16). By inhibiting SGLT2, empagliflozin induces a glucosuric effect. Empagliflozin acutely induces a diuresis and natriuresis, associated with a small increase in bicarbonate excretion and urine pH. Chronic empagliflozin induces a diuretic/natriuretic tone, as indicated by lower systolic blood pressure (SBP) despite a compensatory upregulation of renal renin expression. The full acute natriuretic and chronic volume effect of empagliflozin depends on intact tubular NHE3 and therefore may in part involve NHE3 inhibition. The underlying mechanism is unclear but may include a scaffolding role of membrane-associated protein 17 kDa (MAP17)/NHE-regulatory cofactor 3(NHE-RF3) and/or effects on NHE3 phosphorylation, which affects NHE3 localization. Metabolic adaptations to empagliflozin include responses to urinary losses of glucose, calories, and Na⁺. This involves enhanced expression of genes related to proximal tubular glucose, bicarbonate, and ammonium production [glutaminase (GLS) and phosphoenolpyruvate carboxykinase (PEPCK)] and ammonium secretion in type A intercalated cells [Rhesus C glycoprotein (RHCG)]. These effects were at least in part independent of tubular NHE3. Intracellular glucose may feedback inhibit on PEPCK expression, as previously described in hepatocytes (49), an effect potentially reduced by inhibition of SGLT2-mediated glucose uptake. PEPCK may also be disinhibited by the compensatory lowering of insulin levels. SGLT2 inhibition enhances load-dependent glucose and/or NaCl reabsorption in downstream S2/3 segments and the medullary thick ascending limb (mTAL) and further distal segments. Compensatory hyperreabsorption of NaCl may be facilitated by proximal tubular secretion of α-ketoglutarate (α-KG). The latter has been shown to stimulate electroneutral NaCl reabsorption through combined pendrin/Na⁺-dependent Cl⁻-HCO₃⁻ exchanger NDCBE in non-A, non-B intercalated cells (by activation of its receptor OXGR1) and inhibits Na⁺ reabsorption via the epithelial Na⁺ channel (ENaC) and thus Na⁺/K⁺ exchange in principal cells (independent of OXGR1), causing NaCl retention while limiting K⁺ loss (22, 42, 54). The increase in urinary ammonium (same molar increase vs. α-ketoglutarate) may balance for the loss of “potential bicarbonate” in the form of α-ketoglutarate. The enhanced workload in distal segments is associated with an increase in glucose metabolism [phosphofructokinase platelet isoform (PFKP)] and may lower oxygen availability in the outer medulla, thereby triggering hypoxia-inducible factor-1 (HIF-1) in these tubular segments and erythropoietin (EPO) expression in fibroblast-like interstitial cells (FLIC). Glucose leaves proximal tubules through basolateral glucose transporter 2 (GLUT2) and is taken up into distal tubular segments by basolateral glucose transporter 1 (GLUT1). NBC1, Na⁺-bicarbonate cotransporter 1; OAT1, organic anion transporter 1.

In study 3, the phosphorylation status of NHE3 was studied in response to chronic empagliflozin application. S552 and S605 of NHE3 are two consensus sites for protein kinase A that are physiologically regulated in vitro (26) and in vivo (25, 26). It has been proposed that phosphorylation of these sites may not directly alter NHE activity (25) but that NHE3 phosphorylation is a potential marker for NHE3 redistribution to the coated pit region where NHE3 is inactive (9, 25, 26) or to the microvillar base (64) where its activity is depressed by local pH change caused by NHE3 clustering (5). The dephosphorylation of NHE3 at S552 has been implicated as a key event in the stimulation of renal proximal tubule Na⁺ reabsorption by angiotensin II (11). Moreover, NHE3 phosphorylation at S552 and S605 has been linked to functional NHE3 inhibition and lesser Na⁺ and bicarbonate reabsorption in response to activation of the glucagon-like protein-1 receptor (10, 45, 47). The present studies found that chronic empagliflozin enhanced the phosphorylation of NHE3 at S552 and S605 in Akita mice. These changes were not observed in nondiabetic mice, indicating that the diabetic setting facilitated this interaction. These results are consistent with a recent study in rats in which the SGLT2 inhibitor ipragliflozin, given for 8 wk, did not affect total renal NHE3 expression but increased NHE3 phosphorylated at S605 in spontaneously diabetic Torii rats, a nonobese type 2 diabetic model; ipragliflozin, however, had no significant effect on NHE3 phosphorylation in nondiabetic Sprague-Dawley rats (35). The finding of greater levels of phosphorylated NHE3 in diabetic rodents in response to empagliflozin and ipragliflozin (Fig. 10) support a potential functional link between SGLT2 inhibition and NHE3. Further studies are needed to determine the quantitative contribution of this proposed, phosphorylation-dependent natriuretic interaction.

NHE3 is a major pathway for bicarbonate reabsorption in the kidney, although other Na⁺-dependent (6) and Na⁺-independent [H⁺-ATPase (65)] transport systems also contribute to proximal tubular bicarbonate reabsorption and can compensate (Fig. 10), which likely explains the overall modest renal acid-base phenotype of tubular NHE3-ko mice (39). Since empagliflozin acutely induced a small increase in urinary bicarbonate excretion, it seemed possible that chronic empagliflozin could also have a sustained effect on acid-base handling. Along these lines, in nondiabetic mice, chronic empagliflozin changed renal expression of genes that are related to proximal tubular ammonia, glucose, and bicarbonate formation and distal tubular H⁺ and ammonia secretion in the same direction as NHE3-ko mice (Fig. 10). However, while mice with a lack of tubular NHE3 made less acidic urine, chronic empagliflozin made the urine more acidic. Therefore, it appears that these changes in gene expression compensated for impaired urine acidification and bicarbonate reabsorption in NHE3-ko but contributed to the enhanced urine acidification and ammonium excretion and potentially bicarbonate retention and generation in response to SGLT2 inhibition. This response may indicate a modestly increased acid load in nondiabetic animals given empagliflozin, which could reflect metabolic adaptations to the urinary loss of glucose and calories [e.g., increased formation of acidic ketone bodies (43)]. Moreover, the effects of NHE3-ko and chronic empagliflozin on urine pH and renal gene expression were at least in part independent of each other, indicating that any functional inhibition of NHE3 by chronic empagliflozin may not have been strong enough to significantly impair the role of NHE3 in renal acid-base handling, which, as pointed out above, is not the only mechanism of renal bicarbonate reabsorption. This may have clinical relevance since a mild chronic metabolic acidosis could have a negative effect on the progression of CKD.

Consistent with dissimilar effects on the kidney and its metabolism, empagliflozin and NHE3-ko had different effects on the urine metabolite pattern. This included opposite effects of empagliflozin and NHE3-ko on TCA cycle metabolites in the urine of nondiabetic mice, including α-ketoglutarate. Downregulation of urinary α-ketoglutarate in NHE3-ko is consistent with reduced renal levels of α-ketoglutarate due to acidosis-induced activation of α-ketoglutarate dehydrogenase, which converts α-ketoglutarate to succinyl-CoA (34), thereby accelerating the conversion of glutamate to α-ketoglutarate and the formation of ammonium. Moreover, the resulting lesser proximal tubular secretion of α-ketoglutarate reduces the stimulation of its receptor OXGR1 in type B intercalated cells and thereby pendrin-mediated bicarbonate secretion (54), which would otherwise further aggravate the acid-base disturbance in NHE3-ko. In comparison, the fact that empagliflozin did not suppress urinary α-ketoglutarate, like NHE3-ko, indicates that it did not induce a large acid challenge. But why did empagliflozin even increase urinary α-ketoglutarate? Notably, activation of OXGR1 by luminal α-ketoglutarate also stimulates electroneutral reabsorption of NaCl in non-A, non-B intercalated cells, and luminal α-ketoglutarate inhibits the epithelial Na⁺ channel independently of OXGR1 (54). In other words, the hypothesis is proposed that an increase in luminal α-ketoglutarate in response to empagliflozin serves to enhance NaCl reabsorption in distal segments, which receive an increased NaCl load in response to SGLT2 inhibition, while limiting epithelial Na⁺ channel-dependent K⁺ loss (Fig. 10). Urinary excretion of α-ketoglutarate has been proposed to represent the loss of “potential bicarbonate,” which provides the advantage of minimizing bicarbonaturia under alkali load (40). This allows excretion of base at lower urinary pH, thereby lessening the risk of calcium-phosphate precipitates and nephrolithiasis. Notably, in nondiabetic WT mice, empagliflozin increased urinary α-ketoglutarate-to-creatinine ratios and urinary ammonium-to-creatinine ratios by the same amount of ~1 mol/mol, i.e., the urinary loss of “potential bicarbonate” was matched by acid excretion in the form of ammonium. In other words, a significant portion of the observed renal metabolic adjustments in response to empagliflozin may relate to using α-ketoglutarate as a paracrine signal to induce compensatory distal NaCl reabsorption and generating and excreting more ammonium to establish this in an acid-base neutral way. The body can engage this mechanism in response to empagliflozin because chronic empagliflozin, in contrast to NHE3-ko, may not significantly impair the renal ability for bicarbonate conservation. Notably, this very same compensatory mechanism, associated with increased urinary α-ketoglutarate and ammonium, has recently been established in a mouse model with impaired activity of the renal Na⁺-Cl⁻ cotransporter (21, 22). In contrast to empagliflozin, however, this model did not upregulate renal PEPCK expression, which may point to a role of urinary Na⁺ loss and altered glucose handling in the renal metabolic adaptations to empagliflozin (Fig. 10). Further studies are needed to follow up on this hypothesis.

In Akita mice, empagliflozin, but not NHE3-ko (39), improved a potentially deleterious urine metabolite profile, which consisted of metabolites that are mainly related to the TCA cycle and mitochondrial function and that have previously been associated with CKD progression in individuals with or without diabetes (23, 50). These findings are consistent with observations in individuals with type 2 diabetes in which the SGLT2 inhibitor dapagliflozin increased a panel of urinary metabolites linked to mitochondrial metabolism (36).

We discovered that empagliflozin increases urinary azelaic acid in nondiabetic and diabetic mice, a compound that has recently been linked to mitochondrial biogenesis and autophagy by activation of olfactory receptor 544 (53). Azelaic acid is a C9 α,ω-dicarboxylic acid (nonanedioic acid) that is found in grain foods but is also endogenously produced by the peroxisomal ω-oxidation pathway as an end product of linoleic acid (32). Azelaic acid is a ligand for the mouse olfactory receptor Olfr544 and oral administration of azelaic acid in mice reduces adiposity by rewiring fuel preference to fats (71). Empagliflozin reduced urinary stearate and palmitate in nondiabetic and diabetic mice, which may reflect this rewiring of the renal fuel preference to fats. NHE3-ko mice showed even higher urinary azelaic acid levels than induced by chronic empagliflozin, but, in contrast, this was associated with elevated levels of urinary stearate and palmitate. Further studies are needed to better understand these distinct metabolic effects and their implications.

Both tubular NHE3 and SGLT2 activity have been shown to increase GFR in Akita diabetic mice (39, 57) consistent with the tubular hypothesis of diabetic glomerular hyperfiltration and a positive link between proximal tubular reabsorption and GFR through tubuloglomerular feedback (63). If empagliflozin in part inhibited NHE3, then part of the GFR-lowering effect of an SGLT2 inhibitor could be due to NHE3 inhibition. Empagliflozin, however, did not significantly lower absolute GFR levels in WT Akita mice in the present studies. According to mathematical modeling, a potential explanation is that the severe levels of hyperglycemia in Akita mice of >600 mg/dL in the present studies had elevated tubular glucose to high enough levels such that the osmotic effect of luminal glucose enhanced the tubular flow and NaCl delivery to the macula densa and thereby attenuated glomerular hyperfiltration (68). This is in accordance with experimental evidence showing that at very high levels of glucose, the net influence of tubuloglomerular feedback on GFR is reversed such that the tubuloglomerular feedback mechanism actually lowers hyperfiltration (14, 60, 63). Since empagliflozin did not induce a robust GFR-lowering effect in Akita mice, the present studies could not truly address the role of NHE3 in its GFR effects. Further studies are needed at lower levels of blood glucose to probe this potential interaction.

In summary, empagliflozin induced an acute small increase in urine pH and bicarbonate excretion in nondiabetic mice in a NHE3-dependent way. In comparison, chronic empagliflozin application lowered urine pH and altered renal gene expression consistent with an increase in renal glucose, bicarbonate, and ammonium formation, and this occurred at least in part independent of NHE3-ko. These metabolic changes in response to empagliflozin reflect adaptations to urinary glucose and Na⁺ loss. The latter may involve proximal tubular α-ketoglutarate generation and secretion as a paracrine luminal signal for compensatory distal NaCl reabsorption, while generating and excreting more ammonium to balance urine losses of α-ketoglutarate or “potential bicarbonate.” Upregulation of renal gluconeogenesis by empagliflozin may serve to enhance glucose and bicarbonate formation to compensate for urinary glucose loss and for metabolic adaptations that enhance acid formation (e.g., ketone bodies). Consistent with dissimilar effects on the kidneys and their metabolism, chronic empagliflozin and NHE3-ko had different effects on the urine metabolite pattern, including TCA cycle metabolites (like α-ketoglutarate), metabolites previously associated with CKD progression, and fatty acid fuels stearate and palmitate. In this regard, empagliflozin may enhance azelaic acid formation to rewire the renal fuel preference to fats. Most importantly, the presented data support the notion that NHE3 is a determinant of the acute natriuretic and chronic volume effect of SGLT2 inhibition. A coordinated regulation of proximal tubular transporters and reabsorption has been proposed in response to changes in luminal shear stress (12) but also in response to hormonal signals like insulin (56), indicating that molecular mechanisms must be in place to link these transporters. SGLT2 inhibitors can change the phosphorylation status of NHE3 to a natriuretic profile, but this may be restricted to the diabetic setting. The scaffolding protein MAP17 is expressed in proximal tubules, where it binds to SGLT2 and is necessary for its activation; MAP17 is known to interact with NHE-regulatory cofactor 3 (PDZK1), a scaffolding protein linked to other transporters, including NHE3 (7). Thus, MAP17 may contribute to the molecular basis for coordinating the actions of SGLT2 and NHE3 and their functional dependency in the early proximal tubule. Further studies are needed to better define the underlying mechanisms and molecular interactions.

GRANTS

This work was supported by National Institutes of Health (NIH) Grants R01DK112042 (to V.V. and S.C.T.), R01DK106102, R01HL142814, RF1AG061296 (to V.V.), and University of Alabama at Birmingham/University of California-San Diego O’Brien Center of Acute Kidney Injury NIH Grant P30DK079337 (to V.V. and S.C.T.), National Natural Science Foundation of China Grant 81800649 and Natural Science Foundation of Hunan Province Grant 2018JJ3727 (to P.S.), the Department of Veterans Affairs (to V.V. and S.C.T.), and an investigator-initiated research project by Boehringer Ingelheim (to V.V.).

DISCLOSURES

Over the past 36 mo, V. Vallon has served as a consultant and received honoraria from Bayer, Boehringer Ingelheim, Eli Lilly, Janssen Pharmaceutical, Merck, and Retrophin and received grant support for investigator-initiated research from Astra-Zeneca, Bayer, Boehringer Ingelheim, Janssen Pharmaceutical, and Novo-Nordisk. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

A.O., M.D., M.S., K.S., S.C.T., and V.V. conceived and designed research; A.O., Y.F., R.P., M.D., M.C.-M., W.H., P.S., B.F., Y.C.K, and V.V. performed experiments; A.O., R.P., M.D., M.C.-M., W.H., Y.C.K., and V.V. analyzed data; A.O., M.D., Y.C.K., K.S., S.C.T., and V.V. interpreted results of experiments; A.O. and V.V. prepared figures; V.V. drafted manuscript; A.O., Y.F., R.P., M.D., M.C.-M., W.H., P.S., B.F., Y.C.K., M.S., K.S., S.C.T., and V.V. edited and revised manuscript; A.O., Y.F., R.P., M.D., M.C.-M., W.H., P.S., B.F., Y.C.K., M.S., K.S., S.C.T., and V.V. approved final version of manuscript.