Empagliflozin in Heart Failure: Regional Nephron Sodium Handling Effects

Visual Abstract

Abstract

Significance Statement

The effect of sodium–glucose cotransporter-2 inhibitors (SGLT2i) on regional tubular sodium handling is poorly understood in humans. In this study, empagliflozin substantially decreased lithium reabsorption in the proximal tubule (PT) (a marker of proximal tubular sodium reabsorption), a magnitude out of proportion to that expected with only inhibition of sodium–glucose cotransporter-2. This finding was not driven by an “osmotic diuretic” effect; however, several parameters changed in a manner consistent with inhibition of the sodium–hydrogen exchanger 3. The large changes in proximal tubular handling were acutely buffered by increased reabsorption in both the loop of Henle and the distal nephron, resulting in the observed modest acute natriuresis with these agents. After 14 days of empagliflozin, natriuresis waned due to increased reabsorption in the PT and/or loop of Henle. These findings confirm in humans that SGLT2i have complex and important effects on renal tubular solute handling.

Background

The effect of SGLT2i on regional tubular sodium handling is poorly understood in humans but may be important for the cardiorenal benefits.

Methods

This study used a previously reported randomized, placebo-controlled crossover study of empagliflozin 10 mg daily in patients with diabetes and heart failure. Sodium handling in the PT, loop of Henle (loop), and distal nephron was assessed at baseline and day 14 using fractional excretion of lithium (FELi), capturing PT/loop sodium reabsorption. Assessments were made with and without antagonism of sodium reabsorption through the loop using bumetanide.

Results

Empagliflozin resulted in a large decrease in sodium reabsorption in the PT (increase in FELi=7.5%±10.6%, P = 0.001), with several observations suggesting inhibition of PT sodium hydrogen exchanger 3. In the absence of renal compensation, this would be expected to result in approximately 40 g of sodium excretion/24 hours with normal kidney function. However, rapid tubular compensation occurred with increased sodium reabsorption both in the loop (P < 0.001) and distal nephron (P < 0.001). Inhibition of sodium–glucose cotransporter-2 did not attenuate over 14 days of empagliflozin (P = 0.14). However, there were significant reductions in FELi (P = 0.009), fractional excretion of sodium (P = 0.004), and absolute fractional distal sodium reabsorption (P = 0.036), indicating that chronic adaptation to SGLT2i results primarily from increased reabsorption in the loop and/or PT.

Conclusions

Empagliflozin caused substantial redistribution of intrarenal sodium delivery and reabsorption, providing mechanistic substrate to explain some of the benefits of this class. Importantly, the large increase in sodium exit from the PT was balanced by distal compensation, consistent with SGLT2i excellent safety profile.

Clinical Trial registry name and registration number

ClinicalTrials.gov (NCT03027960).

Introduction

Sodium–glucose cotransporter-2 inhibitors (SGLT2i) represent an advance for cardiorenal medicine with improvement in both kidney and heart failure outcomes across diverse clinical settings.^1–4 Given the remarkable consistency in outcomes between SGLT2i and the fact that sodium–glucose cotransporter-2 (SGLT2) is expressed almost exclusively in the proximal tubule (PT) of the kidney, it is likely that, at least in part, the benefit to cardiorenal outcomes stems from the proximal tubular effects of these drugs.^5,6 The direct effect of SGLT2i is to block sodium and glucose cotransport through SGLT2.⁶ Importantly, the beneficial effects of this class do not seem to be driven by antihyperglycemic effects since outcomes are improved independent of the presence or severity of diabetes.^7–9 Although the net natriuretic effects of SGLT2i have been described in multiple populations,^10–13 redistribution of reabsorption from proximal to distal nephron segments could have important downstream effects, such as activation of tubular glomerular feedback, reduction in neurohormonal activation, increased erythropoietin production, and overall increased sodium chloride delivery to renal salt sensing elements. Modulation of these pathways could help to explain some of the powerful effects on renal and heart failure outcomes with SGLT2i.

Despite the potential importance, specific mechanisms of SGLT2i on renal tubular sodium handing in humans are poorly understood. Antagonism of SGLT2 will obligatorily reduce sodium transport directly through SGLT2. However, this is stoichiometrically a minor sodium reabsorptive pathway and compensation through SGLT1, which absorbs twice the sodium per glucose as SGLT2, and could actually lead to antinatriuresis in euglycemic patients with low tubular glucose delivery.^14–17 In contrast to the predicted limited direct effects of SGLT2i on sodium reabsorption, studies in rodents have demonstrated profound effects of SGLT2 inhibition on proximal tubular solute handling.^16,18,19 This seems to occur primarily via sodium–glucose cotransporter (SGLT)–mediated regulation of sodium hydrogen exchanger 3 (NHE3), the transport pathway responsible for the largest fraction of renal sodium reabsorption.^16,18 Of note, in micropuncture studies, direct antagonism of NHE3 with a specific inhibitor leads to an approximately 30% reduction in proximal tubular reabsorption, a value similar to that reported with SGLT2 inhibition.^19,20

To investigate the cardiorenal effects of SGLT2i, we conducted a mechanistic study in patients with heart failure and diabetes with dual primary aims: (1) quantify the diuretic and cardiorenal effects and (2) describe the regional tubular sodium handling. We have previously described the diuretic and cardiorenal effects.¹⁰ The objective of the current analysis was to investigate the magnitude and mechanisms of SGLT2 inhibition on proximal tubular sodium handling and better understand how the kidney adapts to these changes.

Methods

Study Oversight

This study was an investigator-initiated trial that was conceived and designed by the investigators in collaboration with the sponsor, Boehringer Ingelheim. Approval was obtained by the Yale University Institutional Review Board, and written informed consent was obtained from all patients. This study was registered in ClinicalTrials.gov (NCT03027960).

Study Population

The detailed methods of this study and results of empagliflozin's diuretic and cardiorenal effects have been published previously.¹⁰ In brief, we included patients with stable heart failure (diagnosed by an advanced heart failure cardiologist) as defined by no hospitalizations during the preceding 60 days, stable heart failure medications for at least 2 weeks, and stable diuretics (if prescribed) for 4 weeks, and the patient was at optimal volume status. Other key inclusion criteria were diagnosis of type 2 diabetes, eGFR ≥45 ml/min per 1.73 m², and age 18 years or older. Key exclusion criteria included the use of a nonloop diuretic except an aldosterone antagonist (≤25 mg spironolactone or ≤50 mg eplerenone); history of diabetic ketoacidosis, brittle diabetes, and/or frequent hypoglycemia or severe hypoglycemic episodes requiring emergent intervention in the past 6 months; history of bladder dysfunction or incontinence; or use of another SGLT2i. To improve enrollment rate and on the basis of the safety experience in SGLT2i trials, the eGFR inclusion criterion was modified to ≥20 ml/min per 1.73 m² in July 2018.

Study Design

This study was a randomized, double-blind, placebo-controlled crossover study consisting of treatment with either 10 mg empagliflozin or matched placebo daily for 14 days followed by a 2-week washout period and crossover to 14 days of treatment with the alternate therapy, as previously described.¹⁰ Randomization was performed in permuted blocks by the Yale New Haven Hospital Investigational Drug Service.

Study Procedures

Detailed study procedures have been previously described.¹⁰ In brief, on days 1 and 14 of each study arm, patients underwent an intensive biospecimen collection. Patients performed an overnight fast before the study visit. At the beginning of each study visit, baseline body weight, vital signs, and blood and urine samples were collected. Next, empagliflozin and matched placebo were administered orally. Patients then received a 500 ml bolus of 5% dextrose in water administered intravenously over 30 minutes followed by a continuous infusion of 100 ml/h to optimize urine flow for serial collection of timed specimens. Urine produced for the first 1.5 hours after water loading is not included in any analysis given the absorption kinetics of empagliflozin and any possible effect of medullary/dead space washout after water loading. In 12 patients, because of a national shortage of 5% dextrose in water, the 500 ml bolus was replaced with consumption of 500 ml of an oral sports beverage (Gatorade) over 30 minutes followed by 100 ml/h. Each patient received the same hydration route during crossover treatment.

Blood and urine samples were obtained at 1.5, 3, 4.5, and 6 hours after empagliflozin or placebo administration. During this time, all urine produced was collected in 1.5-hour cumulative collections, ending with each specified time point. Three hours after empagliflozin or placebo, intravenous bumetanide, a specific sodium–potassium–chloride (NKCC2) inhibitor, was administered in a dose equivalent to the patient's home loop diuretic dose (maximum of 4 mg). Patients not on a long-term loop diuretic (n=1) received 0.5 mg intravenous bumetanide. After a 2-week washout, participants crossed over to the opposite therapy, and the above protocol was repeated.

Outcomes

For the current analysis, the primary outcome was the effect of empagliflozin on regional sodium handling in the PT, loop of Henle, and distal nephron for the overall study period and on individual days 1 and 14. Exploratory outcomes included analysis of urinary extracellular vesicles (uEV) and acid–base parameters to better understand potential effect on NHE3-based transport.

Assays and Calculations

Fractional excretion of sodium (FENa) quantified natriuresis. Fractional excretion of endogenous lithium (FELi), absolute fractional distal reabsorption (FELi–FENa), and relative fractional distal reabsorption ([FELi–FENa]/FELi) were used to describe sodium handling in the PT/loop of Henle and distal nephron, respectively (Figure 1). The use of endogenous lithium as a measure of regional nephron sodium handling has been extensively investigated, and the interested reader is directed to reviews on this literature.^21–23 Although there is debate in this literature regarding the relative contribution of PT, loop of Henle, and distal reabsorption via the epithelial sodium channel across different species and different experimental conditions, relative consensus has been reached that in humans with endogenous plasma lithium levels, (1) lithium reabsorption occurs almost exclusively in the PT and loop of Henle, (2) tracks paracellular sodium reabsorption in those segments, (3) and is the best in vivo technique available to query this physiology in humans (Figure 1).^21–25 Fractional excretions were chosen because it captures instantaneous assessment of sodium handling, which is required given the various experimental conditions encountered across each study visit. We defined and analyzed within study time points as follows: empagliflozin monotherapy (hour 3), empagliflozin–bumetanide combination therapy (hours 4.5 and 6), and “on-drug” (combination of hours 3, 4.5, and 6). To better understand the potential effect of SGLT2 inhibition on NHE3, a multi-ion gap was calculated that incorporated most stoichiometrically relevant anions and cations that are found in urine using the formula: ([Na+K+2Ca+2Mg+Ammonium]−[Cl+1.8×P+0.5 uric acid]). In the setting of inhibition of NHE3, the “gap” in this formula should relate to urine bicarbonate and “bicarbonate equivalent” organic ions, such as citrate. Detailed methods on the measurements of lithium, uEV, and all other analyses are presented in the Supplemental Analytical Methods.

Figure 1.

Methods to quantify regional sodium handling. In the PTs and loop of Henle, both sodium and lithium are reabsorbed in parallel. However, in the distal tubules and collecting duct, sodium is reabsorbed but lithium is not. FELi, fractional excretion of lithium; FENa, fractional excretion of sodium; Na⁺, sodium; Li⁺, lithium; SGLT2i, sodium glucose cotransporter-2 inhibitor; PT, proximal tubule.

Statistical Analysis

Data that were approximately normally distributed are presented as means±SDs, and data with skewed distribution are shown as medians (quartile 1 to quartile 3). Categorical values are presented as frequencies and percentages. Data are presented as box plots and duplicated as violin plots with individual data points in the Supplemental Results (Supplemental Figures 2–4). Skewed variables were log-transformed to approximate normal distribution. We examined the difference on the biologic outcomes between two interventions using linear mixed models accounting for correlations within subjects and within time points. Details on biostatistical methods used are described in the Supplemental Statistical Methods. Given the lack of independence of these end points from each other, we did not include an adjustment for multiple testing reduction in alpha. The results are reported as point estimates with 95% confidence intervals and P values, which have not been adjusted for multiplicity. Statistical analyses were performed using Stata version 17 (Statacorp; College Station, TX).

Results

Global Effects on Renal Sodium Handling

Patient characteristics at baseline have been previously described.¹⁰ Patients had a mean age of 60±12 years and eGFR of 69±19 ml/min per 1.73 m². On average across “on-drug” time points (hours 3, 4.5, and 6), empagliflozin significantly increased sodium excretion compared with placebo (FENa 4.1%±3.7% versus 2.9%±2.7%, P < 0.001; Figure 2 top right panel). This effect was observed both during empagliflozin monotherapy and in combination with bumetanide (Figure 2, left and middle panels). Absolute values for urine and solute clearance across the study visit time points can be found in Supplemental Figure 5.

Figure 2.

(Central illustration): effect of empagliflozin monotherapy, combination therapy with loop diuretic, and the “on-drug” effect on FENa, FELi, and absolute fractional distal sodium reabsorption. Empa, empagliflozin; Reabsorption, absolute fractional distal sodium reabsorption (FELi–FENa); “on-drug” includes hours 3, 4.5, and 6 after administration of empagliflozin. Other abbreviations as in previous figures.

Empagliflozin induced significant changes in proximal tubular sodium handling compared with placebo, as determined by change in endogenous lithium clearance (FELi 26.3%±10.3% versus 20.6%±9.8%, P < 0.001; Figure 2, middle row, right panel) during “on-drug” time points. The increase in FELi was evident both on empagliflozin monotherapy (Figure 2, left panel) and in combination with bumetanide (Figure 2, middle panel). The peak placebo-corrected increase in FELi from hour 0 to hour 4.5 (during peak inhibition of loop of Henle sodium transport with bumetanide) was 7.5%±10.6% (P = 0.001). FELi was an additional 3.9%±8.3% (P < 0.001, Figure 3) higher during bumetanide+empagliflozin versus bumetanide+placebo, indicative of compensatory sodium reabsorption through NKCC2. However, this likely represents an underestimation of reabsorption through NKCC2 since the 1 mg (1–3.5 mg) bumetanide dose administered during the placebo period only increased FELi 7.7%±8.6% (P < 0.001 versus FELi increase >20% expected with full antagonism).^26,27

Figure 3.

Synergistic effect of combination therapy with empagliflozin and loop diuretic on FENa and FELi. Delta FENa and Delta FELi are calculated as the augmentation on empagliflozin and bumetanide (hour 4.5) versus empagliflozin monotherapy (hour 3). Differences between blue and red bars represent the synergistic effect of empagliflozin+bumetanide.

Association between Metrics of Renal Glucose, Sodium, and Water Handling

As previously reported, empagliflozin dramatically reduced the renal glucose threshold linearizing (r=0.79, P < 0.0001) the relationship between glucose filtration and urinary glucose excretion compared with placebo (Figure 4, top panel). A dominant tubular osmotic effect of glucose reducing proximal tubular paracellular sodium reabsorption was not apparent because there was no relationship with any metric of glucose excretion and sodium handling; FELi was unrelated to glucose filtration, glucose reabsorption, glucose excretion, and fractional excretion of glucose (FEGlu) (Figure 4). FENa was similarly unrelated to these parameters (Supplemental Figure 6).

Figure 4.

Association between renal glucose and sodium water handling. The association between renal glucose filtration and urinary glucose excretion (top center panel) is presented for patients on placebo (blue dots) and empagliflozin (red dots). The association between FELi is presented with urinary glucose excretion (middle left panel), renal glucose filtration (middle right panel), renal glucose reabsorption (bottom left panel), and FEGlu (bottom right panel) in patients on empagliflozin only (red dots). All data are “on bumetanide+empagliflozin/placebo,” which includes hours 4.5 and 6. FEGlu, fractional excretion of glucose.

Tubular Compensation

Significant compensatory sodium reabsorption occurred distal to the loop of Henle as the absolute fractional distal sodium reabsorption (FELi–FENa) increased with empagliflozin versus placebo during “on-drug” periods (absolute fractional reabsorption 22.2%±8.2% versus 17.6%±8.7%, P < 0.001; Figure 2, bottom right panel). This increased distal sodium reabsorption occurred to a similar degree (P = 0.31 between conditions) with both empagliflozin monotherapy (Figure 2, left panel) and with empagliflozin+bumetanide (Figure 2, middle panel).

Adaptation of Renal Sodium Handing over 14 Days of Empagliflozin

A statistically significant attenuation of empagliflozin's natriuresis emerged when all three “on-drug” time points were analyzed together. The empagliflozin increased ΔFENa 1.7%±2.5% at day 1, but this attenuated to 0.6%±2.0% at day 14 (mean difference 1.0%; 95% CI, 0.1% to 1.9%; P = 0.004) (Figure 5).

Figure 5.

Effect of empagliflozin therapy on glucose and sodium handling with the first dose and after 14 days of therapy. Deltas represent the difference in fractional excretions between empagliflozin and placebo at day 1 (hashed bar) and day 14 (solid bar) “on-drug,” including hours 3, 4.5, and 6. DSR, absolute fractional distal sodium reabsorption. Other abbreviations are in previous figures.

Despite stable SGLT2 inhibition on empagliflozin (day 1 FEGlu 36%±15% versus day 14 FEGlu 33%±14%; P = 0.14) and by inference stable antagonism of sodium transport through SGLT2, sodium exiting the loop of Henle decreased substantially from day 1 (ΔFELi 8.1%±9.5%) to day 14 (ΔFELi 3.6%±10.7%, P = 0.009) during “on-drug” time points (Figure 5). Notably, the synergistic effect in ΔFELi with bumetanide plus empagliflozin did not change from day 1 (3.8%±3.6%) to day 14 (3.6%±2.1%; P = 0.91). However, this was in the setting of the previously noted incomplete antagonism of NKCC2.

Importantly, augmentation of distal nephron sodium reabsorption was not the primary mechanism for chronic adaptation to SGLT2i. Absolute fractional distal sodium reabsorption decreased significantly from day 1 to day 14 (placebo-corrected distal reabsorption 6.3%±9.1% to 2.8%±9.9%, P = 0.036) consistent with the decrease in distal delivery. Notably, relative fractional distal sodium reabsorption did not change from day 1 to day 14 (P = 0.83), indicating a stable proportion of sodium delivered to the distal nephron was reabsorbed.

Effects on Potassium and Magnesium Handling

There was a small but significant increase in serum potassium from day 1 to day 14 with empagliflozin (4.2±0.5 to 4.3±0.5 mmol/L, P = 0.04) but no change with placebo (4.2±0.5 to 4.1±0.5, P = 0.07, time by treatment P = 0.008). There was no significant difference in potassium excretion with 7.4±3.7 mmol/1.5-hour clearance period excreted on drug with empagliflozin versus 6.8±4.0 mmol/1.5 hours excreted with placebo (P = 0.08). This did not differ between days 1 and 14 (time by treatment P = 0.74) or on versus off bumetanide (P = 0.1). However, natriuresis (and thus distal sodium delivery) was substantially higher on empagliflozin resulting in a significant increase in the urinary Na/K ratio (empagliflozin 5.6±3.5 versus placebo 5.1±3.7, P = 0.005).

Serum magnesium significantly increased from baseline to 14 days with empagliflozin (2.17±0.31 to 2.37±0.33 mEq/L, P < 0.001) but did not change with placebo (2.14±0.32 to 2.15±0.34 mEq/L, P = 0.65), which was highly significant (time by treatment interaction P < 0.001). Fractional excretion of magnesium (FEMg) with the first exposure to empagliflozin was not different with empagliflozin monotherapy (placebo 2.6%±2.0% versus empagliflozin 2.5%±1.8%, P = 0.67). However, FEMg was significantly lower when combined with bumetanide (placebo 9.5%±4.3% versus empagliflozin 8.2%±3.9%, P = 0.037). Consistent with a new magnesium steady state being established by day 14, the difference in FEMg on bumetanide at day 14 had attenuated and was no longer significant (placebo 8.8%±4.4% versus empagliflozin 8.4%±4.3%, P = 0.32), but the time by treatment for this attenuation was not significant (P = 0.74).

Exploratory Analysis on Possible NHE3 Inhibition

Given the large magnitude decrease in proximal tubular sodium reabsorption observed with empagliflozin and evidence that SGLT2i inhibit NHE3 in rodents, we next looked for signals consistent with an NHE3 inhibiting effect in this study. We first evaluated a multi-ion urine anion gap incorporating Na, Cl, K, Mg, Ca, uric acid, phosphate, and ammonium. With this metric, the largest fraction of unmeasured urinary ions should be bicarbonate and bicarbonate equivalent organic ions, such as citrate. There was a highly significant increase in the muti-ion urine anion gap “on-drug” with empagliflozin versus placebo (P $\le$ 0.0001). This relationship was largest at day 1 (P < 0.0001, Figure 6) but had attenuated by day 14 (time by treatment interaction, P = 0.03). Importantly, the substantial increase in FELi observed with empagliflozin was attenuated by nearly 70% after adjustment for the empagliflozin-induced change in multi-ion gap (Figure 6), indicating that the processes acutely increasing multi-ion gap were driving much of the acute increase in FELi.

Figure 6.

Empagliflozin and urinary multi-ion gap. Top panel represents changes in urine multi-ion gap with empagliflozin. Multi-ion gap calculated as ([Na+K+2Ca+2Mg+Ammonium]−[Cl+1.8×P+0.5 uric acid]), and thus, changes should be influenced by urine bicarbonate and bicarbonate equivalent organic ions. Bottom panel is the change in delta FELi observed with empagliflozin relative to placebo during NKCC2 antagonism with bumetanide (hours 4.5 and 6) and after adjustment for the multi-ion gap. The substantial attenuation suggests that the same processes (e.g., inhibition of NHE3) that are increasing multi-ion gap with empagliflozin are also driving the increase in FELi. Data are presented as means for observed data and adjusted means for adjusted FELi with error bars representing 95% confidence intervals. NHE3, sodium–hydrogen exchanger 3; NKCC2, sodium–potassium–chloride.

Several other parameters consistent with a potential inhibition of NHE3 tended to change with empagliflozin. Total ammonium excretion did not differ between empagliflozin and placebo (P > 0.05 for various acute and chronic changes). This was likely due to high visit to visit variability in ammonia excretion independent of the study intervention. For example, ammonium excretion only correlated r²=0.23 between baseline and day 14 during placebo. However, accounting for the hour zero urine ammonia excretion, there was a significant increase with empagliflozin “on-drug” (P = 0.021, Supplemental Figure 7, top left panel). There was also a small but significant decrease in serum bicarbonate levels “on-drug” (empagliflozin 23.5±4.1 mmol/L versus placebo 24.4±3.9 mmol/L, P = 0.01, Supplemental Figure 7, top right panel). Venous pCO₂ decreased by −2.1±4.1 mm Hg with empagliflozin (P = 0.02) but did not change with placebo (0.2±6.9 mm Hg, P = 0.86). However, the time by treatment interaction did not reach statistical significance (P = 0.16). Venous pH was not different between empagliflozin and placebo (P = 0.9). In addition, we analyzed uEVs from participants with adequate baseline urine samples allowing isolation (N=65 time points from N=20 patients). uEV NHE3 substantially deceased from baseline to day 14 (P = 0.009, Supplemental Figure 7, bottom panel). We additionally performed a sensitivity analysis comparing the observed changes in FELi with the modeled change in proximal tubular sodium reabsorption on the basis of observed glucose filtered/reabsorbed and the known SGLT1/2 sodium to glucose stoichiometry. Here, we find that the observed increase in sodium exit from the PT is substantially greater than expected if empagliflozin only inhibited sodium transport through SGLT2 (Supplemental Results Text).

Discussion

The main findings of this study investigating the mechanisms of SGLT2i sodium handling across nephron segments in patients with heart failure are (1) empagliflozin substantially increased sodium delivery to the loop of Henle and distal nephron; (2) the magnitude of proximal tubular sodium reabsorption was larger than expected from blocking SGLT2 alone and several parameters suggestive of NHE3 inhibition changed on empagliflozin; (3) plasma glucose levels did not correlate with the changes in proximal tubular sodium reabsorption, indicating that “osmotic” diuresis is not a dominant mechanism for the observed changes in sodium handling; (4) acutely, sodium exiting the PT is compensated by increased reabsorption in both the loop of Henle and the distal nephron, resulting in modest acute natriuresis; (5) despite stable antagonism of renal glucose reabsorption after 14 days of empagliflozin, natriuresis waned due to the attenuation of sodium exiting the PT and/or loop of Henle. In addition to our prior observations on natriuresis and cardiorenal adaptations in this cohort,¹⁰ these findings confirm in humans that SGLT2i have complex and important effects on intrarenal tubular solute handling.

The most important observation in this study is the translation of the observation, from animals to humans, that SGLT2i cause a large change in intrarenal solute handling, with a shift from reabsorption at proximal to more distal segments. Prior publications have described the modest natriuresis associated with SGLT2i in various settings.^10–13 Beyond SGLT2i′s direct natriuresis, the redistribution of sodium reabsorption observed in this study has critical mechanistic implications that may partially explain many of the cardiorenal benefits of this class. Reduced proximal tubular salt reabsorption leads to increased sodium chloride delivery to the macula densa, which would be expected to activate tubuloglomerular feedback and reduce neurohormonal activation. Tubuloglomerular feedback activation likely plays a role in SGLT2's renal protective effects via improvement in intraglomerular hemodynamics and proximal tubular workload.²⁸ SGLT2i′s direct effect on the macula dense could potentially balance the expected worsening neurohormonal activation from natriuresis. Notably, we have previously published from this cohort that plasma volume decreased with empagliflozin, without the associated increase in neurohormonal activation.¹⁰ Increased sodium reabsorption in the loop of Henle may also drive important beneficial effects. Sodium reabsorption in the loop of Henle cause electrochemical gradients that facilitate paracellular transport of magnesium.^29,30 Thus, increased serum magnesium concentrations from SGLT2i may stem from the increased sodium transport in the loop.^10,31 Increased sodium reabsorption in the loop in response to SGLT2i also reduces oxygen tension in the outer medulla, which is a stimulus of erythropoietin secretion.³² Increases in erythropoietin and hematocrit have been reported in SGLT2i trials.^33–35 Increased hematocrit is a key statistical “mediator” of SGLT2i′s benefits on renal and cardiovascular outcomes, meaning that adjusting for the change in hematocrit attenuates the statistical benefit associated with randomization to SGLT2i in clinical trials.^36–38 A rise in hematocrit occurs via changes in plasma volume and/or red blood cell mass, both of which are affected by intrarenal changes from SGLT2i.^10,33 Although the rise in hematocrit itself is unlikely to truly mediate the benefit of this class, the fact that processes identified in these statistical mediation analyses are traceable to SGLT2i′s direct physiologic effects in the PT and are strongly associated with cardiorenal outcomes underlies the importance of these mechanisms.

The mechanism(s) underlying the acute and chronic renal adaptation to empagliflozin, dampening what would otherwise be a dangerous acute and chronic natriuresis, is critical to the safety of these agents. In micropuncture studies, acute SGLT2 inhibition reduces total proximal reabsorption by nearly 30%.¹⁹ The PT reabsorbs approximately 70% of filtered sodium and water, translating into >100 L of water and nearly 1 kg of salt reabsorbed daily with normal GFR.³⁹ If this large decrease in proximal reabsorption was not compensated for, SGLT2i would result in lethal hypovolemia. However, hypovolemia and AKI have not been observed in clinical trials.^2,3,7 Even in heart failure trials, diuretic down titration and hypovolemia were uncommon.^40,41 We found rapid tubular compensation to the increased distal sodium delivery, which occurred by augmentation of sodium reabsorption in the loop of Henle and distal nephron. These observations provide a mechanism to explain the seemingly contradictory findings of dramatic local tubular effects on solute handling seen in animals with the observed mild net diuretic effects and excellent safety profile in SGLT2i clinical trials.

Sodium reabsorption in the proximal convoluted tubule is mediated by multiple transporters, including SGLT1, SGLT2, and NHE3.^30,39 Under normal physiologic conditions, SGLT2 leverages the energetics of transmembrane sodium gradients to drive the reabsorption of most filtered glucose with a stoichiometry of 1:1 sodium:glucose.¹⁷ SGLT1 scavenges glucose escaping SGLT2 consisting of just 3% of filtered glucose during euglycemia.^5,42–44 SGLT1, with 2:1 sodium:glucose stoichiometry, is located distal to SGLT2 and at idle during euglycemia, allowing SGLT2 to more efficiently reabsorb most filtered glucose.^15,30 Despite the dominant role of SGLT2, genetic absence or pharmacologic inhibition of SGLT2 only produces a FEGlu of 30%–50%.^5,30,44 This is due to compensatory increase in glucose reabsorption through SGLT1, accompanied by twice the sodium reabsorption.⁴⁴ Since euglycemic plasma glucose levels do not meaningfully engage SGLT1, SGLT2 inhibition should be antinatriuretic.

However, SGLT2i induced an acute natriuresis, and we found empagliflozin reduced proximal tubular sodium reabsorption by a magnitude that substantially exceeds the expected quantity on the basis of filtered and transported solute through SGLT2. Glucose transport through SGLT is a key regulator of sodium–hydrogen exchangers in the gut, and SGLT2 modulation inhibits renal NHE3 transport in animals.^16,18,45,46 Importantly, in NHE3 knockout mice, SGLT2 inhibition is indeed antinatriuretic, while SGLT2 inhibition induces natriuresis in littermate wild-type mice.¹⁶ There are several observations in this study that suggest NHE3 may in fact be modulated by SGLT2i in humans as well, such as an acute increase in urine multi-ion gap (likely indicating increased secretion of bicarbonate and bicarbonate equivalents), a decrease in serum bicarbonate, and signals for increased urine ammonium excretion. A key finding is that the large increase in FELi produced by SGLT2i was attenuated by nearly 70% after adjustment for the empagliflozin induced increase in multi-ion gap. This indicates that the same processes driving the increase in multi-ion gap are also likely responsible for the majority of the increase in FELi, with NHE3 inhibition being the most likely explanation.

The Acetazolamide in Acute Decompensated Heart Failure with Volume Overload trial, which excluded patients on SGLT2i, found improved rates of decongestion with acetazolamide and bumetanide versus loop monotherapy.⁴⁷ Acetazolamide's diuretic effect occurs largely by inhibiting renal NHE3-mediated sodium transport.^39,48,49 Given SGLT2i inhibit NHE3 in animal models and likely also in humans, acetazolamide's diuretic effect in the presence of SGLT2i is unknown.⁵⁰ The change in FELi with acetazolamide in healthy volunteers is similar to the change in FELi reported in this study with empagliflozin.⁵¹ Although the direct effects of SGLT2 inhibition and carbonic anhydrase inhibition on NHE3 would be expected to be subadditive, the myriad of additional proximal tubular and compensatory effects differentially induced by the strategies (e.g., opposite effects on renal ammoniagenesis) could potentially lead to synergistic effects in the setting of chronic SGLT2i use. Additional research is needed to understand this biology.

Limitations

This was a small, intensive, mechanistic study from a single center, limiting generalizability. To facilitate the crossover design, only stable, euvolemic patients with heart failure were enrolled. We can thus not conclude that the results apply to patients without heart failure. Similarly, we only studied patients with type 2 diabetes mellitus. We used endogenous lithium to estimate sodium handling in the PT and loop of Henle. Although data in human subjects without exogenous lithium administration consistently support utility of this metric, there is a conflicting literature particularly regarding animal models in extreme physiologic conditions. As a result, the results should be interpreted in light of these inconsistencies in the literature. Treatment and washout periods were short to ensure stable volume status and medical therapies for cross over. Therefore, long-term renal adaptations to empagliflozin cannot be determined from this study. The bumetanide dose was the patients home loop diuretic dose and likely did not saturate NKCC2, underestimating empagliflozin's full effect on proximal tubular and loop of Henle sodium handling. We focused on major transporters driving paracellular solute transport, without modeling theoretical paracellular transport from unabsorbed glucose. Because there was no relationship between FELi and plasma glucose (thus filtered glucose load), osmotic forces driving paracellular transport are likely not significant. The exploratory NHE3 uEV analyses should be considered hypothesis generating only. These reported observations are in the context of water loading. Although water loading is critical from a practical standpoint to allow sufficient urine flow for accurate clearance measurements, all reported physiologic observations are on the backdrop of a water loaded state. As a result, findings may differ in the setting of antidiuresis.

In conclusion, empagliflozin leads to a large decrease in proximal tubular sodium reabsorption, significantly more than would be expected with SGLT inhibition alone. This could be explained by the known effects of SGLT2 inhibition on NHE3 reported in animal models, a hypothesis which is supported by several observations in this study, but further definitive research in humans is needed. These large changes in intrarenal solute handling may contribute to several of the clinical benefits noted with SGLT2i. Despite the large decrease in sodium reabsorption in the PT, subsequent nephron segments rapidly adapt to reabsorb the increased delivery of sodium, preventing over diuresis. This observed biology of powerful effects in the PT but restrained total natriuresis is congruent with the significant improvement in cardiorenal outcomes and excellent safety profile observed with SGLT2i.

Supplementary Material

jasn-35-189-s001.pdf ^{(1.2MB, pdf)}

Disclosures

J. Butler reports Consultancy: Consulting fees from 3ivelabs, Abbott, American Regent, Amgen, Applied Therapeutics, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Cardiac Dimension, Cardior, CVRx, Cytokinetics, Edwards, Element, Faraday, G3 Pharmaceutical, Imbria, Impulse Dynamics, Innolife, Inventiva, Ionis, Janssen, LivaNova, Lexicon, Medtronic, Merck, Novartis, Novo Nordisk, Otsuka, Occlutech, PharmaCosmos, Roche, Sanofi, Secretome, Sequana, Tricog, and Vifor; Honoraria: As above consultancy; Advisory or Leadership Role: As above consultancy; and Speakers Bureau: AstraZeneca, Boehringer Ingelheim-Lilly, Impulse Dynamics, and Novartis. L. Bellumkonda reports Consultancy: CareDx; Research Funding: Natera; Honoraria: CareDx; and Advisory or Leadership Role: CareDx. S.P. Collins reports Consultancy: Abbott, Boehringer Ingelheim, and Reprieve Cardiovascular; and Research Funding: Beckman Coulter. Z.L. Cox reports grants from AstraZeneca; and Consultancy with Roche and Translational Catalyst. D.H. Ellison reports Consultancy: Boehringer Ingelheim; Honoraria: Boston University School of Medicine and Renaissance School of Medicine; Patents or Royalties: Author, UpToDate; and Advisory or Leadership Role: Consulting Editor, Hypertension and Editorial Board, American Journal of Physiology-Renal Physiology and JASN. M.P. Field reports Employer: Elemental Scientific. S.E. Inzucchi has served on clinical trial committees or as a consultant to Abbott, AstraZeneca, Bayer, Boehringer Ingelheim, Esperion, Lexicon, Merck, Novo Nordisk, Pfizer, and vTv Therapeutics; and has delivered lectures sponsored by AstraZeneca and Boehringer Ingelheim. S.E. Inzucchi reports Consultancy: AstraZeneca, Bayer, Boehringer Ingelheim, Merck, Novo Nordisk, and Pfizer; Honoraria: AstraZeneca, Bayer, Boehringer Ingelheim, Merck, Novo Nordisk, and Pfizer; Advisory or Leadership Role: Editor-in-Chief, Diabetes Channel, PracticeUpdate website (honoraria received from Elsevier); and Speakers Bureau: Not on any speaker's bureaus/rosters but gave lectures supported by AstraZeneca and Boehringer Ingelheim in the past. J.B. Ivey-Miranda reports personal fees from AstraZeneca, Boehringer Inglheim, Merck, Moksha8, Novartis, and Translational Catalyst. J.B. Ivey-Miranda reports Research Funding: AstraZeneca and Sequana; and Speakers Bureau: AstraZeneca, Boehringer Ingelheim, Merck, Moksha8, and Novartis. C. Maulion reports Consultancy: Sequana Medical. V.S. Rao has a patent for Treatment of diuretic resistance US20200079846A1 issued to Corvidia Therapeutics Inc. and Yale with royalties paid to Yale University, V.S. Rao and J.M. Testani and a patent Methods for measuring renalase WO2019133665A2 issued to Yale. V.S. Rao reports personal fees from Translational Catalyst. V.S. Rao reports Consultancy: Translational Catalyst; and Patents or Royalties: Corvidia Therapeutics. J.M. Testani reports Consultancy: 3ive labs, AstraZeneca, Bayer, BD, Boehringer Ingelheim, Bristol Myers Squibb, Cardionomic, Cytokinetics, Edwards Life Sciences, FIRE1, Lexicon Pharmaceuticals, Lilly, MagentaMed, Merck, Novartis, Otsuka, Precardia, Regeneron, Relypsa, Reprieve, Inc., Sanofi, Sequana Medical, Windtree Therapeutics, and W.L. Gore; Ownership Interest: Reprive, Inc. and Sequana Medical; Research Funding: 3ive labs, Abbott, Boehringer Ingelheim, Bristol Myers Squibb, FIRE1, Lexicon Pharmaceuticals, Merck, Otsuka, Reprieve, Inc., Sanofi, and Sequana Medical; Honoraria: 3ive labs, AstraZeneca, Bayer, BD, Boehringer Ingelheim, Bristol Myers Squibb, Cardionomic, Cytokineitcs, Edwards Life Sciences, FIRE1, Lexicon Pharmaceuticals, Lilly, MagentaMed, Merck, Novartis, Otsuka, Precardia, Regeneron, Relypsa, Reprieve, Inc., Sanofi, Sequana Medical, Windtree Therapeutics, and W.L. Gore; and Patents or Royalties: Corvidia, Reprieve, Inc., Sequana Medical, and Yale University. J.M. Turner reports grants and/or personal fees from 3ive labs, Abbott, AstraZeneca, Bayer, BD, Bristol Myers Squibb, Cardionomic, Edwards Life Sciences, FIRE1, Lexicon Pharmaceuticals, Lilly, MagentaMed, Merck, Novartis, Otsuka, Precardia, Regeneron, Relypsa, Reprieve, Inc., Sanofi, Sequana Medical, Windtree Therapeutics, and W.L. Gore. In addition, J.M. Turner has a patent for Treatment of diuretic resistance issued to Corvidia Therapeutics Inc. and Yale, a patent methods for measuring renalase issued to Yale, and a patent Treatment of diuretic resistance pending with Reprieve, Inc. J.M. Turner reports Consultancy: Sequana Medical; and Research Funding: Bayer, Boehringer Ingelheim, and Sequana Medical. D.R. Wiederin is the President of Elemental Scientific. D.R. Wiederin reports Employer: Elemental Scientific; Consultancy: Elemental Scientific; Ownership Interest: Elemental Scientific; Research Funding: Elemental Scientific; Patents or Royalties: Elemental Scientific; Advisory or Leadership Role: Elemental Scientific; and Speakers Bureau: Elemental Scientific. C.S. Wilcox reports Consultancy: AstraZeneca, Bayer, Bristol Myers Squibb, Relypsa, Sarfez Inc., Sequana, and Vifor; Ownership interest: Sarfez Inc.; Honoraria: Sequana; and Advisory or Leadership Role: Sarfez Inc., scientific advisor and Sequane, Data monitoring and safety committee for SAHARA and MOJAVE trials. F.P. Wilson reports Ownership Interest: Owner of Efference, LLC; Research Funding: Amgen, Boeringher-Ingelheim, Vifor, and Whoop; Advisory or Leadership Role: Editorial Board—American Journal of Kidney Disease and CJASN; and Other Interests or Relationships: Medical Columnist—Medscape. All remaining authors have nothing to disclose.

Funding

National Institutes of Health (NIH) K23HL114868, L30HL115790, R01HL139629, R21HL143092, R01HL128973, R01HL148354 (to JMT), R01DK113191, and P30DK079210 (to F.P. Wilson) Grants. This original clinical trial was supported by Boehringer Ingelheim Pharmaceuticals Inc. (BIPI). The funding source had no role in the original study design, data collection, analysis, or interpretation. The current biochemical and statistical analysis in this manuscript was funded by internal Yale University research funds, without sponsorship or support from BIPI.

Author Contributions

Conceptualization: Zachary L. Cox, Juan B. Ivey-Miranda, Veena S. Rao, Jeffrey M. Testani.

Formal analysis: Juan B. Ivey-Miranda, Julieta Moreno-Villagomez, Venna S. Roa.

Funding acquisition: Jeffrey M. Testani.

Investigation: Jeffrey M. Testani.

Methodology: David H. Ellison, M. Paul Field, Juan B. Ivey-Miranda, Veena S. Rao, Jeffrey M. Testani, Christopher S. Wilcox, Daniel R. Wiederin.

Supervision: Jeffrey M. Testani.

Writing – original draft: Zachary L. Cox, Juan B. Ivey-Miranda, Veena S. Rao, Jeffrey M. Testani.

Writing – review & editing: Lavanya Bellumkonda, Javed Butler, John Chang, Sean P. Collins, David H. Ellison, M. Paul Field, Silvio E. Inzucchi, Christopher Maulion, Juan B. Ivey-Miranda, Veena S. Rao, Jeffrey M. Testani, Jeffrey M. Turner, Daniel R. Wiederin, Christopher S. Wilcox, F. Perry Wilson.

Data Sharing Statement

All data is included in the manuscript and/or supporting information.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/JSN/E559.

Supplemental Analytical Methods

Supplemental Figure 1. Lithium concentration determination.

Supplemental Statistical Methods

Supplemental Results Figures

Supplemental Figure 2. Effect of empagliflozin monotherapy, combination therapy with loop diuretic, and the “on-drug” effect on FENa, FELi, and absolute distal sodium reabsorption.

Supplemental Figure 3. Synergistic effect of combination therapy with empagliflozin and loop diuretic on FENa and FELi.

Supplemental Figure 4. Effect of empagliflozin therapy on glucose and sodium handling with the first dose and after 14 days of therapy.

Supplemental Figure 5. Absolute values for urine and solute clearance.

Supplemental Figure 6. Association between renal glucose and FENa.

Supplemental Figure 7. Exploratory analysis on possible NHE3 inhibition.

Supplemental Results Text. Predicted and observed sodium excretion on the basis of SGLT2 stoichiometry with different assumptions of percentage SGLT2 reabsorptive capacity (text).

Supplemental Figure 8. Predicted and observed sodium excretion on the basis of SGLT2 stoichiometry with different assumptions of percentage SGLT2 reabsorptive capacity.

References for Supplemental Methods and Results