SGLT2 inhibition versus sulfonylurea treatment effects on electrolyte and acid-base balance: secondary analysis of a clinical trial reaching glycemic equipoise Tubular effects of SGLT2 inhibition in type 2 diabetes

SGLT2 inhibitors increase plasma magnesium and plasma phosphate and may cause ketoacidosis, but the contribution of improved glycemic control to these observations as well as effects on other electrolytes and acid-base parameters, remain unknown. Therefore, our objective was to compare the effects of SGLT2 inhibitors dapagliflozin and sulfonylurea gliclazide on plasma electrolytes, urinary electrolyte excretion, and acid-base balance in people with type 2 diabetes (T2D). We assessed the effects of dapagliflozin and gliclazide treatment on plasma electrolytes and bicarbonate, 24-hour urinary pH and excretions of electrolytes, ammonium, citrate, and sulfate in 44 metformin-treated people with T2D and preserved kidney function. Compared to gliclazide, dapagliflozin increased plasma chloride by 1.4 mmol/l (95% CI 0.4 to 2.4), plasma magnesium by 0.03 mmol/l (95% CI 0.01 to 0.06), and plasma sulfate by 0.02 mmol/l (95% CI 0.01 to 0.04). Compared to baseline, dapagliflozin also significantly increased plasma phosphate, but the same trend was observed with gliclazide. From baseline to week 12, dapagliflozin increased the urinary excretion of citrate by 0.93 ± 1.72 mmol/day, acetoacetate by 48 µmol/day (IQR -17 to 138), and β-hydroxybutyrate by 59 µmol/day (IQR 0 to 336), without disturbing acid-base balance. Dapagliflozin increases plasma magnesium, chloride, and sulfate compared with gliclazide, while reaching similar glucose-lowering in people with T2D. Dapagliflozin increases urinary ketone excretion without changing acid-base balance. Therefore, the increase in urinary citrate excretion by dapagliflozin may reflect an effect on cellular metabolism including the tricarboxylic acid cycle. This potentially contributes


Introduction
Sodium-glucose transporter (SGLT)2 inhibitors was the first glucose lowering drug-class to improve cardiovascular and kidney outcomes in people with type 2 diabetes (T2D) at high risk for cardiovascular disease [1]. SGLT2 inhibitors lower plasma glucose levels by reducing sodium-coupled glucose reuptake in the proximal tubule [2]. Although glycosuria is the main change induced by SGLT2 inhibition, it also alters the tubular handling of other electrolytes including sodium. SGLT2 inhibitors have been found to increase certain plasma electrolytes, including magnesium and phosphate [3]. SGLT2 inhibition also predisposes people with diabetes to euglycemic diabetic ketoacidosis (DKA) [1]. However, an integrative assessment of the effects of SGLT2 inhibition on electrolyte and acid-base balance in people with T2D is lacking. Plasma glucose concentrations also influence tubular transport of different solutes and water [4]. Therefore, the aim of this study was to analyze the effects of SGLT2 inhibition on electrolyte and acid-base balance compared to sulfonylurea treatment with equal glucose lowering. This allowed us to dissect the direct tubular effects of SGLT2 inhibition from those related to glycemia.

Trial design
This is a pre-specified secondary analysis of the RED (Renoprotective Effects of Dapagliflozin in Type 2 Diabetes) trial; a phase-4, monocenter, randomized, double-blind, comparator-controlled, parallelgroup, intervention trial [5]. The trial was originally designed to determine treatment-induced changes in gold-standard measured renal hemodynamics after 12-week dapagliflozin and gliclazide therapy in people with T2D. After a 4-week run-in period, participants were randomized to dapagliflozin 10 mg or gliclazide 30 mg (block-size of 4, performed by an independent trial pharmacist using computer-generated numbers), and treated for 12 weeks. Before and after treatment, participants collected urine during a 24 hour period that ended on the night before a visit to the clinical research unit. Participants were given oral and written instructions on how to collect a 24-h urine-sample, and were instructed to postpone collection in case of fever or urinary-tract infection, and to refrain from strenuous exercise during the collection-period. After an overnight fast, blood samples were obtained at 07:30 AM. The week before the test visits, participants adhered to normal sodium (9 -12 g/day) and protein (1.5 -2.0 g/kg/day) diets, in order to minimize variation.
Written informed consent was obtained from all participants before any trial-related activities. The study complied with the Declaration of Helsinki and Good Clinical Practice guidelines and was registered at ClinicalTrials.gov (ID: NCT02682563) [5].

Outcome measures
The endpoints of this pre-specified secondary analysis of the RED trial [5] were dapagliflozin and gliclazide induced changes in plasma concentrations and urinary excretion of electrolytes and ketones, and acid-base balance.

Measurements and calculations
GFR was measured by plasma inulin/iohexol clearances, as described [5]. Fractional excretions were calculated using fasting plasma and 24h urine creatinine concentrations. Plasma and urinary sodium, potassium, chloride, magnesium, calcium, phosphate, urea, creatinine, bicarbonate, glucose, and uric acid were all measured using routine chemistry methods on a Cobas 8000 routine analyser (Roche diagnostics, Mannheim Germany). Creatinine was measured using the enzymatic method. Urine pH was measured by with the Edge benchtop pH-meter HI2020-02 with CAL Check. Plasma pH and bicarbonate were measured in heparinized venous blood by the ABL800 FLEX blood gas analyzer (type ABL825; Radiometer, Zoetermeer, the Netherlands). Plasma and urinary sulfate were measured by means of a validated ion-exchange chromatography assay with conductivity detection (Metrohm, Herisau, Switzerland) [6]. Citrate in urine was determined by an enzymatic method (Instruchemie, Delfzijl, The Netherlands) on a ABX Pentra 400 clinical chemistry analyzer (Horiba, Montpellier, France). Urinary ammonium was measured in triplicate using the Berthelot method [7,8]. Plasma and urine acetoacetate was measured in plasma and urine samples by an automated spectrophotometric enzymatic method on a Beckman UniCel® DXC600 Synchron® Analyzer (Fullerton, CA), and β-hydroxybutyrate (BHB) concentrations were measured on a Synchron system CX4 (Beckman Coulter, Fullerton, CA) [9]. The miliequivalents of each measured anion and cation in the electrogram ( Figure 4) were calculated by multiplying the charge of the ion, if needed calculated from pK of the ion and urine pH, with the absolute excretion in the 24 hour urine collections.

Statistical analysis
Statistical analyses were performed in the per protocol population using SPSS 24.0 (IBM SPSS, Chicago, IL). Multivariable linear regression models were used to analyze and compare the effects of dapagliflozin and gliclazide. Corresponding baseline values were added as independent variables to correct for potential between-group baseline differences. Within-group comparisons were performed using paired samples t-tests (normally distributed data) or Wilcoxon signed-rank tests (non-normally distributed data). Significance was set at a 2-sided α-level of < 0.05. Statistically nonsignificant p-values are not shown in figures. Data are presented as mean ± SD for normally distributed variables, median (IQR) for non-normally distributed variables, or as baseline corrected mean difference with a 2-sided 95% confidence interval (CI) unless otherwise specified. Since we aimed to provide an inclusive overview of the treatment effects, we did not correct for multiple testing to minimize the risk of type 2 error. Downloaded from http://portlandpress.com/clinsci/article-pdf/doi/10.1042/CS20201274/897612/cs-2020-1274.pdf by guest on 19 November 2020

Clinical trial results
As previously described [5], 44 people were randomized to 12-week treatment with dapagliflozin (n = 24) or gliclazide (n = 20). Baseline characteristics were well balanced between treatment groups ( Table 1) and medication remained unchanged during the treatment period. No (serious) adverse events, including hypoglycemic events, episodes of acute kidney injury or electrolyte disturbances occurred in either group and no participants dropped out after treatment commenced. There were 5 genital fungal infections in the dapagliflozin group versus none in the gliclazide group.
Gliclazide had no significant effects on renal glucose handling and GFR.

Discussion
In this pre-specified secondary analysis of the RED trial, we analyzed the effects of dapagliflozin on electrolyte and acid-base balance in people with T2D and preserved renal function. Because the effects of dapagliflozin were compared with the sulfonylurea gliclazide, we were able to study these effects while plasma glucose levels were equally lowered. We confirm that dapagliflozin causes small but significant increases in plasma magnesium and show for the first time that this is also the case for plasma chloride and sulfate. Of interest, the dapagliflozin-induced increase in plasma phosphate appeared to depend on glycemia, because the same trend was observed with gliclazide. Increases in plasma ketone concentrations were likely prevented by an increase in urinary ketone excretion with dapagliflozin. Accordingly, there were no effects on systemic acid-base balance, which was also supported by the absence of changes in urinary ammonium excretion. We did observe an increase in urinary citrate excretion, which is a novel finding. Below we will discuss possible explanations for these findings, including their potential clinical relevance. SGLT2 inhibition has been consistently found to increase plasma magnesium concentrations in people with T2D [10, 11], although the mechanism remains unclear. Hypomagnesemia and urinary magnesium wasting are associated with T2D, possibly due to reduced activity of TRPM6, a transient receptor potential cation channel in the distal convoluted tubule, and may be secondary to insulin resistance [12,13]. We recently reported that clamp-measured insulin resistance was unaffected by dapagliflozin or gliclazide, making improved insulin resistance a less likely mediator [14]. However, since we did not measure glucose disposal and endogenous glucose production separately, and others have found SGLT2 inhibitor-induced improvements in insulin sensitivity [15], we cannot fully discard the contribution of improved insulin resistance. Glucagon reduces magnesium reabsorption in the proximal tubule, but strongly increases magnesium reabsorption in the thick ascending limb and distal convoluted tubule [16]. Furthermore, plasma insulin may cause a redistribution of magnesium from the extracellular to the intracellular compartment [17]. The increased plasma Downloaded from http://portlandpress.com/clinsci/article-pdf/doi/10.1042/CS20201274/897612/cs-2020-1274.pdf by guest on 19 November 2020 magnesium concentration upon SGLT2 inhibition could thus be caused by the increase in glucagon/insulin ratios [18]. Another contributing factor to the increased magnesium reabsorption is an increased chloride reabsorption leading to an increased lumen-positive potential and increase in paracellular reabsorption of magnesium [19]. The increase in plasma magnesium following SGLT2 inhibition is modest. Yet, the drug class is currently under investigation as a potential treatment for refractory hypomagnesemia [20].
Dapagliflozin increased plasma sulfate and chloride. The majority of filtered sulfate is reabsorbed in the proximal tubule, where secondary active sulfate uptake takes place via sodium-coupled sulfate transport, driven by the electro-chemical gradient for sodium across the luminal membrane [21].
Because SGLT2 inhibition increases the luminal sodium gradient, this may explain the increase in plasma sulfate concentrations. In patients with heart failure sulfate clearance is related to diuretic use, creatinine clearance, and sodium excretion [22]. Thus, dapagliflozin may increase plasma sulfate, because of its diuretic properties and its effects on GFR and sodium excretion. As shown by the urine electrogram, the increase in plasma chloride concentrations is due to increased chloride reabsorption coupled to other ions than sodium. In the proximal tubule chloride is mainly reabsorbed paracellularly in the S3 segment. Therefore, a shift of solute transport from the S1/S2 segments to the S3 segment could explain the increase in plasma chloride [23]. Effects of dapagliflozin on transcellular chloride transport seem less likely, as these are coupled to bicarbonate, and we found no effects on plasma bicarbonate [24]. Although disturbances in plasma chloride levels have been implicated in worsened outcome in heart failure (hypochloremia) [25] or renal disease (hyperchloremia) [26], chloride levels remained within the normal range in our study and the observed increase is probably too small to affect clinical outcome.
Plasma phosphate levels were increased by dapagliflozin after 12 weeks of treatment. It is believed that sodium-coupled phosphate reabsorption (via NaPi2) is increased in the proximal tubule by the Downloaded from http://portlandpress.com/clinsci/article-pdf/doi/10.1042/CS20201274/897612/cs-2020-1274.pdf by guest on 19 November 2020 higher availability of sodium following SGLT2 inhibition [27]. Increases in phosphate reabsorption and serum phosphate levels, together with increases in plasma parathyroid hormone (PTH), fibroblast growth factor 23 (FGF23) and reductions in 1,25-dihydroxyvitamin D levels, have been previously found, both in healthy volunteers treated with canagliflozin [28] and in people with T2D and CKD stage 2 to 4 treated with dapagliflozin [29]. Because plasma phosphate also increased with gliclazide, we found no between group differences, which might indicate a glucose-dependent mechanism. It is however also possible that gliclazide increases plasma phosphate levels via increasing insulin levels [30].
It has been hypothesized that increased PTH and FGF23 secretion following increased phosphate reabsorption upon SGLT2 inhibition may be detrimental for bone health [27]. However, a dedicated analysis found no effect of dapagliflozin versus placebo on markers of bone formation/resorption, bone mineral density, or fractures, while PTH and FGF23 also remained unchanged despite a slight increase in plasma phosphate [31]. Although bone fractures have been added to the label of SGLT2 inhibitor drug-class as an adverse effect, data remain conflicting.
Dapagliflozin increased fracture risk versus placebo in a phase III trial including 252 people with T2D and stage 3 CKD [32], as did canagliflozin in the post-marketing CANVAS program [33]. In contrast, fracture risk was not increased in other outcome trials with SGLT2 inhibitors, including the renal CREDENCE trial using canagliflozin [34][35][36].
SGLT2 inhibitors increase the risk for diabetic ketoacidosis, which is a major problem in people with T1D, where absolute risks up to 6% have been found [37]. Although absolute risks are extremely low in people with T2D (0.005 to 0.6%), especially those with sufficient endogenous insulin production, relative risks are significantly increased [33,35,36]. In addition to increased ketogenesis [18], it has been hypothesized that SGLT2 inhibition might reduce urinary ketone excretion. The increased luminal sodium concentration could theoretically drive the reabsorption of filtered ketone bodies via the proximal tubule transporters SLC5A8 (S3 segment) and SLC5A12 (S1, S2, and S3 segments) [38]. In Downloaded from http://portlandpress.com/clinsci/article-pdf/doi/10.1042/CS20201274/897612/cs-2020-1274.pdf by guest on  contrast to this hypothesis, we observed an increase, rather than a decrease, in urinary ketone excretion. These results are in agreement with a previous study, although this study did not report the effects on acetoacetate [9]. The increased delivery of ketone bodies to the proximal tubule seemingly contributes to attenuated diabetes associated kidney damage, potentially via correcting mTORC1 hyperactivation [39]. Furthermore, we observed no changes in plasma ketones following dapagliflozin, despite increased urinary glucose excretion and decreased plasma insulin concentrations following dapagliflozin. Accordingly, venous and urinary pH, venous bicarbonate levels, and urinary ammonium excretion were unaffected by both treatments. Dapagliflozin did increase urinary citrate excretion. The proximal tubule regulates citrate reabsorption in response to changes in systemic acid-base balance [40], but because we did not observe these changes, other explanations must apply. Since citrate can be released from bone, increased bone resorption, as discussed above, might be an explanation [41]. Of interest, citrate accumulates in the kidney cortex of diabetic mice, and this effect is nullified by SGLT2 inhibition [42]. The normalization of accumulated tricarboxylic acid cycle intermediates such as citrate by SGLT2 inhibitors may reduce oxidative stress and thereby contribute to kidney protection. Indeed, patients with diabetes and low urinary citrate experience more rapid progression of CKD [43,44]. The use of urine metabolomics in future research may be helpful to further explore this pathway.
The strength of this study is the randomized and double-blind design of the trial with the incorporation of an active comparator group providing the possibility to compare the effects of dapagliflozin with a control-group that reached similar glucose lowering. However, this study also has a number of limitations. First, we excluded patients with heart failure or CKD, and those using diuretics. The results can therefore not be translated to patients with these comorbidities in which electrolyte and acid-base disorders are common. Second, we did not measure plasma citrate concentrations. Third, it is to be expected that a new electrolyte balance is reached after 12 weeks and that urinary excretion mainly depends on dietary intake. Although we standardized sodium and protein intake, diet was not fully controlled. Fourth, the study was not primarily powered for the Downloaded from http://portlandpress.com/clinsci/article-pdf/doi/10.1042/CS20201274/897612/cs-2020-1274.pdf by guest on 19 November 2020 analysis presented here. Therefore, we provide an overview of the observed effects, with statistical testing within groups and between groups, without rigorous correction for multiple testing. This increased the risk of type 1 error. Finally, we were unable to pinpoint which tubular mechanisms were responsible for the changes in plasma electrolyte concentrations, and propose that experimental studies are required to resolve this. Of note, dapagliflozin reduced blood pressure, GFR, body weight and increased hematocrit, while gliclazide did not. We have previously described the mechanisms explaining these effects [2,5], and cannot exclude that these effects might have influenced our results. Lastly, despite the fact that randomization rendered well balanced groups, blood pressure and plasma insulin levels differed somewhat at baseline. Although we statistically corrected for baseline values, this potentially affected the physiological effects of both drugs.
In conclusion, dapagliflozin caused small increases in plasma chloride, magnesium, and sulfate, while gliclazide did not despite similar glucose lowering. Dapagliflozin also increased urinary ketone and citrate excretion. Especially the increase in urinary citrate excretion by dapagliflozin is a novel finding that may reflect an effect on cellular metabolism including the tricarboxylic acid cycle. This could potentially contribute to kidney protection by this drug-class.

Data Availability Statement:
We are unable to share our data due to patient privacy.

Acknowledgments
We are very grateful for all the time and commitment of our study participants without whom we could have never completed this trial. We also acknowledge the help of the assistants who were indispensable in the process of data collection: Ingrid Knufman, Renée de Meijer, and especially Jeanette Boerop. Finally, we thank Nel Willekes-Koolschijn (Department of Nephrology and Hypertension, University Medical Center, Utrecht, The Netherlands), for her much appreciated technical laboratory assistance.

Funding
This trial was funded by AstraZeneca as an investigator-initiated study. The funder had no role in the study design, the analyses or interpretation of the data, or drafting the manuscript. The funder had no role in the decision to submit this manuscript for publication.  Table 1 -Baseline characteristics. Data are represented as mean ± SD, frequency, or median (IQR). Abbreviations: ASCVD, atherosclerotic cardiovascular disease; eGFR, estimated glomerular filtration rate; CKD-EPI, chronic kidney disease epidemiology collaboration; UACR, Urinary albumin-creatine ratio; RAS, renin-angiotensin system.    -The effects of dapagliflozin and gliclazide on 24 hour urinary excretion of electrolytes, ammonium, and citrate, and urinary pH. Data were analyzed with paired ttests or wilcoxon signed-rank test within-group and by baseline corrected linear regression analysis between-group. Data are represented as mean, or geometric mean (pH) and 95% confidence intervals. Only statistically significant p-values are shown. Abbreviations: DAPA, dapagliflozin; GLIC, gliclazide.