Effects of the SGLT2 Inhibitor Dapagliflozin on Energy

Diabetes Care Volume 44, June 2021 1

Effects of the SGLT2 Inhibitor Yvo J.M. Op den Kamp,1 Marlies de Ligt,1 Bas Dautzenberg,1 Esther Kornips,1 Dapagliflozin on Energy Russell Esterline,2 Matthijs K.C. Hesselink,1 Joris Hoeks,1 Vera B. Schrauwen-Hinderling,1,3 Metabolism in Patients With Bas Havekes,4 Jan Oscarsson,5 Esther Phielix,1 and Patrick Schrauwen1,a Type 2 Diabetes: A Randomized, Double-Blind Crossover Trial Diabetes Care 2021;44:1–10 | https://doi.org/10.2337dc20-2887

OBJECTIVE SGTL2 inhibitors increase urinary glucose excretion and have beneﬁcial effects on cardiovascular and renal outcomes. The underlying mechanism may involve caloric restriction-like metabolic effects due to urinary glucose loss. We investigated the effects of dapagliﬂozin on 24-h energy metabolism and insulin sensitivity in patients with type 2 diabetes.

RESEARCH DESIGN AND METHODS 1Department of Nutrition and Movement There were 26 patients with type 2 diabetes randomized to a 5-week double- Sciences, Maastricht University Medical Centre, REGIMENS AND DRUGS THERAPIES: EMERGING blind, crossover study with a 6- to 8-week washout. Indirect calorimetry was Maastricht, the Netherlands 2BioPharmaceuticals R&D, AstraZeneca, used to measure 24-h energy metabolism and the respiratory exchange ratio Gaithersburg, MD (RER), both by whole-room calorimetry and by ventilated hood during a two-step 3Radiology and Nuclear Medicine, Maastricht euglycemic-hyperinsulinemic clamp. Results are presented as the differences in University Medical Centre, Maastricht, the least squares mean (95% CI) between treatments. Netherlands 4Internal Medicine, NUTRIM School for Nutrition and Translational Research in Metabolism, RESULTS Maastricht University Medical Centre, Maastricht, Evaluable patients (n = 24) had a mean (SD) age of 64.2 (4.6) years, BMI of 28.1 the Netherlands 2 5BioPharmaceuticals R&D, Late-Stage Development, (2.4) kg/m , and HbA1c of 6.9% (0.7) (51.7 [6.8] mmol/mol). Rate of glucose dis- Cardiovascular, Renal and Metabolism, AstraZeneca, appearance was unaffected by dapagliflozin, whereas fasting endogenous glucose Gothenburg, Sweden production (EGP) increased by dapagliflozin (12.27 [1.39, 3.14] lmol/kg/min, Corresponding author: Patrick Schrauwen, p. P < 0.0001). Insulin-induced suppression of EGP (–1.71 [–2.75, –0.63] lmol/kg/ [email protected] P – – – P min, = 0.0036) and plasma free fatty acids ( 21.93% [ 39.31, 4.54], = 0.016) Received 26 November 2020 and accepted 17 was greater with dapagliflozin. Twenty-four-hour energy expenditure (–0.11 February 2021 [–0.24, 0.03] MJ/day) remained unaffected by dapagliflozin, but dapagliflozin re- Clinical trial reg. no. NCT03338855, clinicaltrials. duced the RER during daytime and nighttime, resulting in an increased day-to- gov. nighttime difference in the RER (–0.010 [–0.017, –0.002], P = 0.016). Dapagliflozin This article contains supplementary material online treatment resulted in a negative 24-h energy and fat balance (–20.51 [–27.90, at https://doi.org/10.2337/figshare.14058038. –13.12] g/day). Y.J.M.O.d.K., M.d.L., and B.D. share equal authorship. © 2021 by the American Diabetes Association. CONCLUSIONS Readers may use this article as long as the Dapagliflozin treatment for 5 weeks resulted in major adjustments of metabolism work is properly cited, the use is educational fi mimicking caloric restriction, increased fat oxidation, improved hepatic and adi- and not for pro t, and the work is not altered. More information is available at https://www. pose insulin sensitivity, and improved 24-h energy metabolism. diabetesjournals.org/content/license. Diabetes Care Publish Ahead of Print, published online April 15, 2021 2 Dapagliflozin Affects Energy Metabolism Diabetes Care Volume 44, June 2021

Sodium–glucose cotransporter 2 inhibitors Therefore, the aim of the current (EHC), when the timing of medication (SGLT2is) inhibit glucose and sodium study was to investigate whether SGLT2i was at 6:30 A.M., before start of tracer reabsorption in the proximal renal tu- treatment for 5 weeks affects 24-h infusion. During the end-of-treatment bules and have been proven a valuable energy balance and substrate metabol- visit, blood was drawn at 1:30 P.M.for additional treatment for type 2 dia- ism in patients with type 2 diabetes. To assessment of HbA1c, uric acid, and hs- betes to improve glycemic control and this end, whole-body 24-h energy ex- CRP. Hereafter, participants stayed in a also moderately reduce body weight penditure and substrate oxidation, whole-room calorimeter from 6:00 P.M. and blood pressure (1). Importantly, whole-body and tissue-specificinsulin onward to measure O2 consumption SGLT2is have been shown to reduce sensitivity, and body composition, in- and CO2 production (16). The last 24 h cardiovascular (CV) risk, including re- cluding ectopic fat accumulation, were starting at 5:30 A.M. were used for ana- duced hospitalization for heart failure investigated in large detail in patients lysis. In the respiration chamber, partici- and reduced risk of kidney disease pro- with type 2 diabetes after 5 weeks of pants adhered to an activity protocol gression (2,3). How SGLT2is exert their SGLT2i or placebo treatment. consisting of multiple times standing remarkable effects on CV and renal upright and including twice a 15-min outcomes is not fully understood. For RESEARCH DESIGN AND METHODS stepping exercise at 30 steps/min to in- example, only 12% of the relative risk crease physical activity levels. Physical Study Design and Participants reduction for all-cause death caused by activity was measured by Actigraphy A double-blind, randomized, placebo- SGLT2is could be explained by the (ActiGraph, Pensacola, FL). Participants controlled, crossover study was con- combined changes in conventional CV received three standardized meals at ducted at the Metabolic Research Unit risk factors (4). 0830, 1300, and at 1800. Twenty-four- Maastricht (MRUM) of Maastricht Uni- The primary action of SGLT2i on glu- hour energy intake was determined versity between 5 March 2018 and 4 cose and sodium reabsorption has sev- from the sleeping metabolic rate (SMR) November 2019. Ethics Committee of eral metabolic consequences that may during the first night multiplied by an Maastricht University Medical Centre help to explain the effects on CV and re- activity factor of 1.5, and all meals were (Maastricht, the Netherlands) approved nal outcome (1). Such metabolic effects completely consumed by the volun- the study, which was conducted to con- may reflect the adaptive response to teers. -Twenty-four hour energy expend- form to the Declaration of Helsinki (15). the loss of 50–100 g glucose/day in iture and RER were calculated using Upon written informed consent, pa- the urine, resulting in a form of mild Brouwer et al. (17). Daytime RER was tients were randomized to double-blind caloric restriction and caloric loss over- calculated from 0800 until 2200 and night. Consistent with this line of treatment for two treatment periods, nighttime RER from 0000 till 0530. SMR thinking, SGLT2i treatment has been re- each 5 weeks or a maximum duration was defined as the two consecutive – ported to result in increased glucagon of 40 days, separated by a 6 8weeks hours with lowest energy expenditure levels (5) and decreased insulin secre- washout (see Supplementary Material using the Weir formula (18) and was tion (6), most likely mediated by de- for details). End points were assessed at averaged for two nights. Twenty-four- creased plasma glucose levels (7). Other the end of each 5-week period. Male hour diet-induced thermogenesis (DIT) observed metabolic effects include in- and female patients with type 2 dia- was determined by plotting energy ex- creased endogenous glucose production betes, diagnosed for at least 6 months, penditure against physical activity level, (EGP) (8), increased fasting fatty acid with HbA1c levels between 6 and 9% and the intercept of the regression line oxidation (9), and decreased intrahe- (42 and 75 mmol/mol), and on a stable at the lowest physical activity repre- > patic lipid (IHL) content (10). In add- dose of metformin for 3monthsor sents SMR1DIT (19). Glucose and nitro- ition, several (8,9,11) but not all (10) drug naïve, were eligible. Detailed eligi- genweremeasuredin24-hurine studies have reported improvements in bility criteria are in Supplementary collection (in 6-h aliquots), the latter to whole-body/peripheral insulin sensitiv- Table 1. calculate daytime and nighttime protein ity. All these separate effects resemble oxidation. Blood samples were drawn at effects that are also observed after cal- Procedures seven time points (see Supplementary oric restriction, which is also accompa- The study comprised seven visits, in- Table 2). nied by enhanced fat oxidation (12), cluding screening (visit 1). Visit 2 (or 5), After leaving the respiration chamber improved metabolic flexibility (13), re- the first visit of the treatment periods, after an overnight fast, proton MRS was ductions in plasma glucose, and im- included randomization and measure- used to quantify IHL content (see provements in insulin sensitivity (13). ment of safety markers. A safety visit Supplementary Materials for details). Despite these separate finding, so far (visit 3 or 6) was scheduled after 2 Subsequently, an EHC with coinfusion of no studies have investigated the hy- weeks, and an end-of-treatment visit D-glucose [6,6-D2] tracer (0.04 mg/kg Á pothesis that SGTL2i treatment of pa- (visit 4 or 7) after 5 weeks. All outcome min) was performed (20). After 3 h, an tients with type 2 diabetes can indeed measures were performed at visit 4/7 insulin infusion was started with 10 restore the fed-to-fasting cycle in pa- and were spread over a period of 6–8 mU/m2/min for 3 h to assess hepatic in- tients with type 2 diabetes and whether days, and visit 4 was followed by a 6–8 sulin sensitivity and subsequently in- the treatment has caloric restriction-like weeks wash-out period. Timing of study creased to 40 mU/m2/min for 2.5 h to effect on 24-h energy and substrate me- medication during end of treatment assess whole-body insulin sensitivity. tabolism, as was hypothesized previous- was 8:30 A.M., except on the day of Glucose (20%) was coinfused to keep ly (14). the euglycemic-hyperinsulinemic clamp glucose levels at 5 mmol/L. care.diabetesjournals.org Op den Kamp and Associates" Do for all odd pages 3

2 Arterialized blood (hotbox) was drawn and no important protocol deviations, of 28.1 (2.4) kg/m and HbA1c of 6.9% every 5–10 min to monitor glucose con- was used for the statistical analyses, us- (0.7) (51.7 [6.8] mmol/mol). Patients centration. During the last 30 min of ing SAS version 9.04 software and per- had a mean estimated glomerular filtra- every steady-state period, blood sam- formed by IQVIA Biostatistics, Reading, tion rate of 141 (13.0) mL/min. There ples were collected, and indirect calor- U.K. The expected difference between were 17 patients on metformin treat- imetry (Omnical; Maastricht treatment groups was estimated using a ment, and 7 patients did not use any Instruments, Maastricht, the Nether- linear mixed-effects model. This model antidiabetic drugs. No other antidiabetic lands) was performed to assess sub- had treatment group, treatment se- drugs were used. strate utilization according to Peronnet quence, and period as fixed effects, as Compliance was >95% during both et al. (21). Nonoxidative glucose dis- well as a random intercept for each sub- treatment periods. Body weight was sig- fi fl posal (NOGD) was defined as the Rd ject. If deviations from normality were ni cantly reduced by dapagli ozin treat- corrected for urinary glucose loss minus detected, a nonparametric test of treatment compared with placebo (–1.26 carbohydrate oxidation. Metabolic flexi- ment difference against zero was per- (–1.85, –0.66) kg, P = 0.0003) (Fig. 1A). bility was defined as the change in the formed (Wilcoxon signed rank test) Body weight was measured also at the RER from basal to high-insulin stimu- using all of the data and ignoring the se- beginning of each period, and no carry- lated state. Isotopic enrichment of plas- quence. The least-squares means (LSM) over effects as a result of the treatment ma glucose was determined using and the corresponding 95% CIs for sequence were noticed. Assessment of electron ionization gas chromatography- treatment effect in the respective treat- body composition by DEXA showed sig- fi mass spectrometry (22). Calculation of ment periods are presented. The differ- ni cantly reduced lean mass by dapagli- fl glucose R and R was performed ac- ence in the LSM between the two ozin treatment compared with placebo a d – – – cording to Steele’s single pool non- treatments was computed, with the cor- ( 0.67 [ 1.29, 0.04] kg, P =0.038)(Fig. steady state equations (23). Basal R responding 95% CI and P value tabu- 1A), whereas whole-body fat mass was d fi fl values were corrected for fasting urin- lated. A two-sided 0.05 level was not signi cantly affected by dapagli o- zin treatment (Fig. 1A). Trunk fat mass ary glucose excretion during the night, considered statistically significant. We (–0.48 [–0.89, –0.07] kg, P =0.023)(Fig. and R values during insulin infusion used the Benjamini-Hochberg procedure d 1B) as well as IHL content, as measured were corrected for urinary glucose ex- for multiple testing and found that ad- by 1H-MRS, were lower after dapagliflo- cretion during the clamp. In two partici- justed P values were in agreement with zin treatment in 18 of 22 patients (P = pants, glucose excretion during the unadjusted P values. For power calcula- 0.036) (Fig. 1C). clampcouldnotbemeasuredinoneof tion, see the Supplementary Material. Systolic blood pressure after 2 weeks the periods, and fasted urinary glucose The trial is registered with Clinical- of treatment was significantly lower excretion was used, multiplied by the Trials.gov, number NCT03338855. after dapagliflozin treatment (–6.77 percentage average difference on urin- [–12.09, –1.45] mmHg, P = 0.015) (Sup- Data and Resource Availability ary glucose excretion in the fasted plementary Table 4), but diastolic blood l Data underlying the findings described (2.209 and 0.008 mol/kg/min on dapa- pressure was not significantly affected fl in this manuscript may be available gli ozin and placebo) and insulin-stimu- by dapagliflozin (–1.85 [–5.68, 1.98] l upon request in accordance with Astra- lated state (2.656 and 0.006 mol/kg/ mmHg, P = 0.33) (Supplementary Table fl Zeneca’s data sharing policy described min on dapagli ozin and placebo). Ex- 4). Systolic blood pressure (–3.58 at https://astrazenecagroup-dt.pharmacm. clusion of these two subjects did not al- [–10.33, 3.16] mmHg, P = 0.28) (Supple- com/DT/Home. ter the EHC results. mentary Table 4) or diastolic blood On day 6, 7, or 8 during the end-of- pressure (–0.65 [–4.61, 3.31] mmHg, P treatment visit, a DEXA scan (Hologic, RESULTS = 0.74) (Supplementary Table 4) after 5 Marlborough, MA) was used to deter- The study enrolled 38 patients, and 26 weeks of treatment was not significant- mine body composition. patients were randomized (Supplemen- ly changed after dapagliflozin treat- Changes in vital signs and safety la- tary Fig. 1). Twelve patients were ineli- ment. Levels of plasma hsCRP (0.22 boratory values were monitored during gible for participation. Two patients [–0.45, 0.90] mg/L, P = 0.50) (Supple- the study. Adverse events leading to dis- were excluded from the evaluable data mentary Table 5) and HbA1c (–0.07% continuation from the study as well as set because of important protocol devi- [–0.22, 0.08], P = 0.33) (Supplementary serious adverse events and cases of sus- ations. One of the excluded participants Table 5), measured during the end-of- pected diabetic ketoacidosis were discontinued the study after completing treatment visits, were not significantly collected. one period as a result of a change in altered with dapagliflozin treatment. medical treatment affecting insulin sen- HbA1c was measured also at the begin- Biochemical Analysis sitivity. The other patient was excluded ning of each period, and no carryover Details of the biochemical analysis are because safety and efficacy assessments effects as a result of treatment se- outlined in Supplementary Material. were not performed as defined in the quence were noticed with respect to protocol. HbA1c levels. Urate levels were lower Statistics Baseline characteristics of the 24 after dapagliflozin treatment (–56.8 The evaluable analysis set, consisting of evaluable patients are reported in Sup- [–75.4, –38.1] lmol/L, P < 0.0001) patients with at least one dose of the plementary Table 3. The patients had a (Supplementary Table 5). Furthermore, investigational product (per protocol) mean (SD) age of 64.2 (4.6) years, BMI levels of hemoglobin were higher with 4 Dapagliflozin Affects Energy Metabolism Diabetes Care Volume 44, June 2021

Figure 1—Body composition (n = 24) (A), fat mass (n = 24) (B), and IHL (C) content upon placebo (P) and dapagliﬂozin (D) treatment (n = 22). BW, body weight; FM, fat mass; LBM, lean body mass. Results are in LSM and 95% CI, obtained through a linear mixed model, with the exception of IHL, where a Wilcoxon paired rank sum test was used. *P < 0.05 is considered signiﬁcantly different.

dapagliﬂozin treatment (0.19 [0.02, Fig. 2B); an increase that was similar in min, P = 0.012) (Supplementary Table 6 0.35] mmol/L, P = 0.03) (Supplementary magnitude to the increase in urinary and Fig. 2E). Fasting nonesteriﬁed fatty

Table 5), while there was no significant glucose excretion rate. Basal Rd,cor- acid (NEFA) levels (208.88 [53.88, change in erythrocyte volume fraction rected for urinary glucose loss, was 363.89] mmol /L, P =0.011)(Supple- (0.004 [–0.006, 0.014] L/L, P =0.43) similar between the dapagliflozin and mentary Table 6, Fig. 2F) and fasting (Supplementary Table 5). placebo treatment periods (0.11 [–1.12, glycerol levels (0.022 [0.005, 0.039] A two-step EHC with indirect 1.25] mmol/kg/min, P = 0.85) (Supple- mmol/L, P = 0.013) (Supplementary calorimetry was performed after an mentary Table 6). Fasting fat oxidation Table 6 and Fig. 2G) were all significant- overnight fast when the patients had was higher (0.53 [0.22, 0.85] mmol/kg/ ly higher after dapagliflozin treatment been in the respiration chamber for 36 min, P = 0.0022) (Supplementary Table compared with placebo. Fasting insulin h. Urinary glucose excretion during the 6andFig.2C), while fasting carbohy- levels were lower after dapagliflozin clamp increased after dapagliflozin drate oxidation was lower after dapagli- treatment compared with placebo treatment (2.46 [2.06, 2.86] mmol/kg/ flozin treatment compared with placebo (–18.18 [–23.07, –13.28] pmol/L, P < min, P < 0.0001) (Supplementary Table (–1.73 [–2.72, –0.74], mmol/kg/min, P = 0.0001) (Supplementary Table 6 and 6 and Fig. 2A). Basal or fasting EGP was 0.0016) (Supplementary Table 6 and Fig. 2H). higher after dapagliflozin treatment Fig. 2D). NOGD in the fasted state was Next, a low dose of insulin (10 mU/ (2.27 [1.39, 3.14] mmol/kg/min, P < higher with dapagliflozin treatment ver- m2/min) was infused for 3 h to assess 0.0001) (Supplementary Table 6 and sus placebo (1.85 [0.45, 3.24] mmol/kg/ hepatic insulin sensitivity. The insulin- care.diabetesjournals.org Op den Kamp and Associates" Do for all odd pages 5

R G d

Figure 2—Urinary glucose excretion (A), EGP (B), fatty acid (FA) oxidation (C), carbohydrate oxidation (D), NOGD (E), plasma NEFA levels (F), plasma glycerol levels (G), plasma insulin levels (H), DEGP (I), DRd (J), percentage of NEFA suppression (K), Dsuppression of glycerol plasma levels (L), and DRER (M) measured during a two-step EHC upon placebo (P) and dapagliﬂozin (D) treatment. Results (n = 22) are in LSM and 95% CI, obtained through a linear mixed model. *P < 0.05 is considered signiﬁcantly different.

induced suppression (DEGPlow-basal)was insulin state, fat oxidation was signiﬁ- Insulin sensitivity expressed as the larger with dapagliﬂozin treatment cantly higher (0.50 [0.11, 0.89] mmol/ change in Rd from basal to the high-in- compared with placebo (–1.71 [–2.78, kg/min, P = 0.015) (Supplementary Ta- sulin state and corrected for urinary glu-

–0.63] mmol/kg/min, P = 0.0036) ble 6 and Fig. 2C), whereas carbohy- cose loss (DRdhigh-basal) was not (Supplementary Table 6 and Fig. 2I), drate oxidation was significantly lower significantly affected by dapagliflozin suggestive of enhanced hepatic insulin after dapagliflozin treatment compared (–1.07 [–3.18, 1.05] mmol/kg/min, P = sensitivity. Improvements in insulin sup- with placebo (–2.03 [–3.85, –0.21] 0.33) (Supplementary Table 6). Plasma pression of EGP were observed despite mmol/kg/min, P = 0.030) (Supplemen- insulin levels during the high-insulin in- lower plasma insulin levels after dapa- tary Table 6 and Fig. 2D). fusion were not affected by dapagliflo- gliflozin treatment (–20.73 [–37.62, To assess peripheral insulin sensitivity, zin (–9.73 [–36.54, 17.09] pmol/L, P = –3.83] pmol /L, P = 0.019) (Supplemen- a high dose (40 mU/m2/min) of insulin 0.46) (Supplementary Table 6 and Fig. tary Table 6 and Fig. 2H). In the low- was subsequently infused for 2.5 h. 2H), and DRdhigh-basal corrected for 6 Dapagliflozin Affects Energy Metabolism Diabetes Care Volume 44, June 2021

steady-state insulin concentrations was affected by dapagliflozin treatment (P = both during daytime (–53.35 [–75.68, not different between the treatment 0.36) (Supplementary Table 7). Food –31.03] g/day, P < 0.0001) (Supplemen- arms (–0.0029 [–0.0083, 0.0024], P = intake was similar during both treat- tary Table 7) and nighttime (–47.69 0.27) (Supplementary Table 6 and Fig. ment periods, and activity levels did not [–67.64, –27.74] g/day, P < 0.0001) 2J). Plasma NEFA was more suppressed differ between the two periods (6.33 (Supplementary Table 7). Twenty-four- upon the high-insulin infusion after da- [–9.65, 22.32] counts/min, P =0.42). hour protein oxidation was not signifi- pagliflozin treatment compared with Diet-induced thermogenesis was not cantly affected by dapagliflozin treat- placebo (–21.93% [–39.31, –4.54], P = significantly affected by dapagliflozin ment versus placebo (–0.91 [–7.30, 0.016) (Supplementary Table 6 and Fig. treatment (P = 0.64) (Supplementary 5.47] g/day, P = 0.77) (Supplementary 2K), resulting in lower NEFA levels dur- Table 7). Table 7 and Fig. 4D) and did not differ ing the high-insulin state after dapagli- Twenty-four-hour urinary glucose loss between treatment periods neither dur- flozin treatment versus placebo (–53.21 was significantly higher by dapagliflozin ing daytime (–1.99 [–8.86, 4.87] g/day, [–101.2, –5.23] mmol/L, P =0.031)(Sup- treatment compared with placebo (3.53 P = 0.55) (Supplementary Table 7) nor plementary Table 6 and Fig. 2F). Similar- [3.04, 4.00] g/h, P < 0.0001) (Supple- during nighttime (–0.192 [–8.40, 8.04] ly, plasma glycerol levels were more mentary Table 7). Interestingly, urinary g/day, P = 0.96) (Supplementary Table suppressed by the high-insulin infusion glucose loss during dapagliflozin treat- 7). after dapagliflozin treatment compared ment was approximately twofold higher While in the respiration chamber, with placebo (–0.029 [–0.050, –0.008] at daytime versus nighttime (4.3 [3.7, blood was sampled seven times during mmol/L, P = 0.0085) (Supplementary 5.0] g/h vs. 2.0 [1.6, 2.4] g/h) (Fig. 3B the day. Fasting glucose levels were Table 6 and Fig. 2L). Differences in fatty and C). After correction for energy lost lower after dapagliflozin (P < 0.0001) acid oxidation (0.30 [–0.01, 0.60] mmol/ as urinary glucose, the energy balance (Supplementary Table 5), and the area kg/min, P = 0.055) (Supplementary Ta- was negative in the dapagliflozin period under the curve (AUC) for the plasma ble 6 and Fig. 2C) or carbohydrate oxi- (Fig. 3D). glucose profile during the day was – – m dation ( 1.11 [ 2.33, 0.11] mol/kg/ Twenty-four-hour RER was lower significantly lower with dapagliflozin fl – min, P = 0.071) (Supplementary Table 6 after dapagli ozin treatment ( 0.02 treatment (P = 0.0006) (Supplementary fl – – and Fig. 2D) between dapagli ozin [ 0.03, 0.01], P = 0.0001) (Supplemen- Table 5 and Fig. 4E). Fasting NEFA levels and placebo during the high-insulin tary Table 7 and Fig. 3E). This effect was were unaffected after dapagliflozin fi state did not reach statistical signi - observed both during daytime (P = treatment (P = 0.22) (Supplementary cance. NOGD during the high-insulin in- 0.0001) (Supplementary Table 5 and Table 5); however, the AUC for the plas- < fusion rate was not changed by Fig. 3F) and nighttime (P 0.0001) ma NEFA profile during the day was fl – m dapagli ozin (0.16 [ 1.89, 2.22] mol/ (Supplementary Table 7 and Fig. 3G). higher after dapagliflozin (P =0.0026) kg/min, P = 0.87) (Supplementary Table RER was lower at night compared with (Supplementary Table 5 and Fig. 4F). 6 and Fig. 2E ). The insulin-induced daytime, and the decrease in RER from Fasting (P = 0.045) (Supplementary Ta- increase in carbohydrate oxidation day to nighttime, which can be consid- ble 5) and AUC of plasma levels of b-hy- (DCHOox ,0.77[–0.37, 1.92] ered as a marker of metabolic flexibility, high-basal droxybutyrate during the day were mmol/kg/min, P = 0.17) and reduction was larger after dapagliflozin compared higher after dapagliflozin treatment (P = in fatty acid oxidation (DFAox , with placebo (–0.010 [–0.017, –0.002], high-basal 0.047) (Supplementary Table 5 and Fig. –0.28 [–0.65, 0.09] mmol/kg/min, P = P = 0.016) (Supplementary Table 7 and 4G). The AUC for plasma FGF21 was not 0.13) were more pronounced after da- Fig. 3H). Twenty-four-hour fatty acid oxi- different with dapagliflozin (P =0.16) pagliflozin treatment but did not reach dation was higher after dapagliflozin (Supplementary Table 5), while the AUC statistical significance (Supplementary treatment (19.70 [12.48, 26.92] g/day, P for plasma glucagon was higher with Table 6). Changes in RER between basal < 0.0001) (Supplementary Table 7 and dapagliflozin treatment (P =0.039)(Sup- and high-insulin state (as a measure or Fig. 4A), resulting in a negative fat bal- plementary Table 5). metabolic flexibility) were numerically ance after dapagliflozin treatment No serious adverse events, adverse larger after dapagliflozin treatment but ((À20.51 [À27.90, À13.12] g/day, Fig. events leading to discontinuation, or did not reach statistical significance 4B). The higher fat oxidation was ob- events of diabetic ketoacidosis were re- (0.01 [–0.01, 0.03], P = 0.18) (Supple- served both during daytime (20.19 ported. One suspected diabetic ketoaci- mentary Table 6 and Fig. 2M). [12.57, 27.81] g/day, P < 0.0001) (Sup- dosis was diagnosed as a gastroenteritis To investigate the effect of dapagliflo- plementary Table 7) and nighttime that was not regarded as due to study zin on whole-body substrate metabol- (22.04 [13.83, 30.26] g/day, P < 0.0001) treatment. ism in detail, patients stayed in a (Supplementary Table 7). Consequently, respiration chamber for 36 h, and meas- 24-h carbohydrate oxidation was lower urements were performed during the after dapagliflozin treatment (–49.39 CONCLUSIONS last 24 h of the stay. Twenty-four-hour [–69.02, –29.77] g/day, P < 0.0001) SGLT2is are indicated for glucose con- total energy expenditure was not signifi- (Supplementary Table 7 and Fig. 4C), trol in type 2 diabetes and have recently cantly affected by dapagliflozin treat- and carbohydrate balance, corrected for been shown to have unanticipated ment compared with placebo (–0.11 urinary glucose loss, remained positive favorable effects on CV and renal [–0.25, 0.03] MJ/day, P = 0.11) (Supple- during both treatment periods (Fig. 4B). outcomes (2,3,24–26). The primary mentary Table 7 and Fig. 3A). The sleep- A lower carbohydrate oxidation after pharmacological action of SGLT2is cause ing metabolic rate was not significantly dapagliflozin treatment was observed glucose and energy loss in urine, which care.diabetesjournals.org Op den Kamp and Associates" Do for all odd pages 7

Figure 3—Twenty-four-hour energy expenditure (n = 24) (A), glucose excretion daytime (n = 23) (B), glucose excretion nighttime (n = 23) (C), energy balance (n = 24) (D), 24-h RER plot (n = 24) (E), daytime RER (n = 24) (F), nighttime RER (n = 24) (G), DRER between day and night (n = 24) (H). Re- sults are in LSM and 95% CI, obtained through a linear mixed model. *P < 0.05 is considered significantly different. would mimic daily mild caloric restric- levels or GFR during daytime may con- sleeping metabolic rate, or diet-induced tion but also result in extra nighttime tribute to larger urinary glucose loss. thermogenesis. Because food intake loss of energy. Here, we demonstrate We also observed higher glucose excre- and activity was kept constant while in that dapagliflozin has beneficial effects tion rates during the insulin phase com- the respiration chamber, the overall 24- on 24-h energy and substrate metabol- pared with the basal phase of the h energy balance was negative after da- ism that mimic the effects previously glucose clamp, suggesting that higher pagliflozin treatment, inducing a state observed with mild caloric restriction or daytime insulin levels may also be of caloric restriction. By design, we time-restricted feeding (12,27,28). We involved. provided a diet relatively high in carbo- confirm in a controlled setting the A glucose loss of 90 g/day would re- hydrate, and the overall 24-h carbohy- pharmacological action of SGLT2is, lead- sult in an energy deficit of 1,500 kJ/ drate balance did not become negative ing to a loss of urinary glucose of 90 day and an expected decrease in body by dapagliflozin treatment. This suggests g/day. Interestingly, the rate of urinary weight of 1.5–2.0 kg during the treat- that the increase in 24-h fat oxidation glucose loss appeared to be twice as ment period, which is close to what we cannot solely be attributed to a lack of large during daytime compared with report here. However, there was no sig- substrate availability for glucose oxida- nighttime, which is a novel finding in nificant effect of dapagliflozin treatment tion, because it is well known that the this study. A plausible explanation for on body fat, but a small but significant intake of carbohydrates stimulates their the differences in daytime and night- decrease in lean body mass, which own oxidation (31). time glucose excretion is the diurnal most likely reflects a loss of extracellular However, in the postabsorptive change in dapagliflozin concentration water, as recently reported after 14 state—specifically during the night and, therefore, the lower nighttime con- days of dapagliflozin treatment (30). when the energy and carbohydrate bal- centration of dapagliflozin (29). Further- Interestingly, we found that the SGLT2i ance is per definition negative—dapagli- more, higher (dietary) carbohydrate did not result in compensatory reduc- flozin may have resulted in a more availability and higher plasma glucose tions in 24-h energy expenditure, pronounced negative energy balance. 8 Dapagliflozin Affects Energy Metabolism Diabetes Care Volume 44, June 2021

Figure 4—A: Twenty-four-hour fatty acid oxidation measured during the stay in the respiration chamber after placebo (P) or dapagliﬂozin (D) treatment. Substrate balance (n = 24) (B), 24-h carbohydrate oxidation (n = 24) (C), 24-h protein oxidation (n = 24) (D), plasma glucose (n = 21) (E), NEFA (n = 23) (F), and b-hydroxybutyrate levels (G) as measured in the respiration chamber (n = 23). Blood draws at 8:30 A.M., 1:00 P.M., and 6:00 P.M. were taken before meals. Results are in LSM and 95% CI, obtained through a linear mixed model. *P < 0.05 is considered signiﬁcantly different

Indeed, a novel finding is that fat oxida- since reduced glycogen levels trigger an are needed to further investigate this tion was more strongly increased at increased fat oxidation. Moreover, a finding. nighttime, leading to a distinct fed-to- typical fed-fasting cycle, with alternating In the current study, we did not find fasting transition with dapagliflozin periods of energy excess and restriction, abeneficial effect of SGLT2is on periph- treatment. These findings are interest- triggers a cascade of molecular path- eral insulin sensitivity. Lack of improve- ing, because we recently showed that ways that are involved in the mainten- ment in peripheral insulin sensitivity the 24-h variability in substrate oxida- ance of optimal cellular functioning and was also observed by Latva-Rasku et al. tion—as reflected by measurement of thereby affect whole-body metabolic (10), whereas other studies with 24-h RER—is markedly blunted in volun- health. Indeed, inducing alternating pe- SGLT2is reported improvements in per- teers with prediabetes compared with riods of transient energy deficit also ipheral insulin sensitivity (8,9,34). Pos- lean, healthy volunteers. This suggests characterize interventions such as exer- sible explanations for the discrepant that in prediabetes the diurnal oscilla- cise training, caloric restriction, or time- findings include differences in baseline tions in the fed-fasting cycle are blunted restricted feeding, which all have been glucose control of the patients and dif- (32), but similar measurements have shown to improve cardiometabolic ferent rates of insulin infusion. In the fi not been done in patients with type 2 health (27,28,33). Thus, the bene cial current study, the patients’ baseline health effects of dapagliflozin in type 2 diabetes. However, it has been shown HbA1c level was 6.9% (51.7 mmol/mol), in type 2 diabetes that 24-h fluctuations diabetes may be due to the restoration which is lower than reported in studies of a blunted fed-to-fasting transition. in glycogen content are blunted (31), that did show an improved peripheral Future investigations including molecu- and it is tempting to speculate that da- insulin sensitivity, in which baseline lar analysis of underlying mechanisms pagliflozin helps to restore the 24-h HbA1c levels averaged 7.5% (58 mmol/ rhythmicity in glycogen metabolism mol) (34), 8.4–8.8% (68–73 mmol/mol) care.diabetesjournals.org Op den Kamp and Associates" Do for all odd pages 9

(8), and 8.5% (69 mmol/mol) (9). Worse contributor of carbon for gluconeogene- insulin sensitivity toward suppression of baseline glucose control would result in sis (31). Interestingly, dapagliflozin im- lipolysis. These results may provide new larger improvement in glucose control proved adipose tissue insulin sensitivity insights in the mechanisms underlying by the intervention and therefore re- toward suppression of lipolysis reflected the beneficial health effects of SGTL2is, duced glucotoxicity that could help to in larger suppression of NEFA and gly- although further studies are needed to explain improved insulin sensitivity. cerol during insulin stimulation, which provide support for this hypothesis. Furthermore, the insulin infusion rate was paralleled by enhanced suppression during the high-insulin phase of the glu- of EGP. These novel findings indicate 2 cose clamp in our study (40 mU/m / regulation of Acknowledgments. The authors would like min) was lower compared with the infu- gluconeogenesis by dapagliflozin via to thank all study participants and staff in the sion rate in those studies that did show changed adipose tissue insulin sensitivity Maastricht University Medical Centre. The au- 2 thors also thank staff at AstraZeneca and IQ- improved insulin sensitivity: 80 mU/m / and suggest that the enhanced hepatic VIA for their valuable support. 2 min (8,9) and 120 mU/m /min (34). Al- insulin sensitivity reported here and pre- Duality of Interest. This clinical trial was though the insulin infusion in our study viously (36) can be explained by im- funded by AstraZeneca. R.E. and J.O. are em- tended to suppress fat oxidation more proved adipose tissue insulin sensitivity. ployed by AstraZeneca and are AstraZeneca fl during dapagliflozin treatment com- In addition, improved adipose tissue in- shareholders. No other potential con icts of interest relevant to this article were reported. pared with placebo, the lower insulin insulin sensitivity toward suppression of The study funder was involved in the design 2 fusion rate of 40 mU/m /min did not lipolysis, together with enhanced fat oxi- of the study, the interpretation of data, and completely suppress fat oxidation, and dation, suggests an increased fatty acid writing the report and did not impose any re- absolute values for fat oxidation during turnover, which may also explain the strictions regarding the publication of the the clamp remained higher with dapa- lower IHL content found in this and pre- report. fl Author Contributions. Y.J.M.O.d.K., M.d.L., gli ozin treatment compared with vious studies (10,36), despite higher fast- and B.D. share equal authorship. Y.J.M.O.d.K., placebo; this elevated fat oxidation ing and daytime free fatty acid levels. M.d.L., B.D., M.K.C.H., J.H., V.B.S.-H., E.P., and could impair skeletal muscle glucose up- Because IHL content is known to correl- P.S. designed and performed the experiments. take via substrate competition. Similar ate with hepatic insulin sensitivity (37), Y.J.M.O.d.K., M.d.L., B.D., V.B.S.-H., E.P., and changes have been observed in athletes enhanced fatty acid turnover may be an- P.S. analyzed the data. Y.J.M.O.d.K., J.O., and P.S. drafted the manuscript. R.E., M.K.C.H., in the morning after running a mara- other factor contributing to improved J.H., V.B.S.-H., B.H., J.O., E.P., and P.S. de- thon when fat oxidation and plasma hepatic insulin sensitivity. signed and conceived the study. All authors NEFA levels were also elevated— prob- A limitation of the study is the short reviewed and approved the final version of ably due to the negative energy bal- duration of treatment. Patients receiv- the manuscript. P.S. is the guarantor of this ance—and insulin sensitivity was ing SGLT2 inhibition for 6 months have work and, as such, had full access to all the data in the study and takes responsibility for paradoxically reduced (35). been shown to increase food intake to the integrity of the data and the accuracy of The higher fat oxidation and lower compensate for the glucose losses (38). the data analysis. carbohydrate oxidation in the fasting It has been suggested that the in- state as well as the unchanged fasting creased compensatory food intake re- References € Rd corrected for urinary glucose loss sulting in stabilization of weight occurs 1. Bolinder J, Ljunggren O, Johansson L, et al. also led to the unexpected observation 2–3 months after initiation of treat- Dapagliflozin maintains glycaemic control while reducing weight and body fat mass over 2 years of increased NOGD in the fasted state. ment with SGLT2is (39). Therefore, the in patients with type 2 diabetes mellitus In line with previous studies investi- metabolic adaptations described in this inadequately controlled on metformin. Diabetes gating the effect of SGLT2is on EGP study may look differently after 3 Obes Metab 2014;16:159–169 (8,9), an increased EGP in the overnight months of treatment. Furthermore, pa- 2. Neal B, Perkovic V, Mahaffey KW, et al.; fasted state of similar magnitude as the tients included in this study were well CANVAS Program Collaborative Group. Canag- liflozin and cardiovascular and renal events in urinary glucose excretion rate was ob- controlled and had relatively low HbA1c, type 2 diabetes. N Engl J Med 2017;377:644– served. The liver is the main organ re- so effects of SGLT2 inhibition on HbA1c 657 sponsible for EGP and is dependent on and glucose control are limited in this 3. Zinman B, Wanner C, Lachin JM, et al.; EMPA- the combined effects of glycogenolysis study. In addition, the better glucose REG OUTCOME Investigators. Empagliflozin, and gluconeogenesis. We did not deter- control of volunteers in this study may cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med 2015;373:2117–2128 mine glycogenolysis, but enhanced pro- have affected the results, such as ef- 4. Coleman RL, Gray AM, Broedl Md UC, et al. tein oxidation upon SGTL2 inhibition fects of treatment on peripheral insulin Can the cardiovascular risk reductions observed has been suggested to contribute to the sensitivity, and limits direct comparison with empagliflozin in the EMPA-REG OUTCOME elevated gluconeogenesis and, thereby, with studies that included less well-con- trial be explained by concomitant changes seen elevated EGP (14). However, in the cur- trolled patients. in conventional cardiovascular risk factor levels? – fl Diabetes Obes Metab 2020;22:1151 1156 rent study, protein oxidation based on To summarize, dapagli ozin treat- 5. Alatrach M, Laichuthai N, Martinez R, et al. urinary nitrogen excretion was not af- ment for 5 weeks resulted in caloric re- Evidence against an important role of plasma fected by dapagliflozin treatment, and striction-like effects on 24-h substrate insulin and glucagon concentrations in the therefore, a major contribution of flexibility as a result of increased urinary increase in EGP caused by SGLT2 inhibitors. glycogenolysis to EGP cannot be ex- glucose loss. Consistent with a caloric Diabetes 2020;69:681–688 fl 6. Ferrannini E, Muscelli E, Frascerra S, et al. cluded. Alternatively, gluconeogenesis restriction-like effect, dapagli ozin also Metabolic response to sodium-glucose cotran- can be fueled by adipose tissue lipolysis, reduced hepatic lipid content and im- sporter 2 inhibition in type 2 diabetic patients. J and glycerol could become a major proved hepatic and adipose tissue Clin Invest 2014;124:499–508 10 Dapagliflozin Affects Energy Metabolism Diabetes Care Volume 44, June 2021

7. Lundkvist P, Pereira MJ, Kamble PG, et al. output) and urine-N. Acta Physiol Pharmacol induces dose-dependent glucosuria in healthy Glucagon levels during short-term SGLT2 Neerl 1957;6:795–802 subjects. Clin Pharmacol Ther 2009;85:520–526 inhibition are largely regulated by glucose 18. Weir JB. New methods for calculating 30. Scholtes RA, Muskiet MHA, van Baar MJB, changes in patients with type 2 diabetes. J Clin metabolic rate with special reference to protein et al. Natriuretic effect of two weeks of Endocrinol Metab 2019;104:193–201 metabolism. J Physiol 1949;109:1–9 dapagliflozin treatment in patients with type 2 8. Merovci A, Solis-Herrera C, Daniele G, et al. 19. Westerterp KR, Wilson SA, Rolland V. Diet diabetes and preserved kidney function during Dapagliflozin improves muscle insulin sensitivity induced thermogenesis measured over 24h in a standardized sodium intake: results of the but enhances endogenous glucose production. J respiration chamber: effect of diet composition. DAPASALT trial. Diabetes Care 2021;44:440–447 Clin Invest 2014;124:509–514 Int J Obes Relat Metab Disord 1999;23:287–292 31. Petersen MC, Vatner DF, Shulman GI. 9. Daniele G, Xiong J, Solis-Herrera C, et al. 20. DeFronzo RA, Tobin JD, Andres R. Glucose Regulation of hepatic glucose metabolism in Dapagliflozin enhances fat oxidation and ketone clamp technique: a method for quantifying health and disease. Nat Rev Endocrinol 2017;13: production in patients with type 2 diabetes. insulin secretion and resistance. Am J Physiol 572–587 Diabetes Care 2016;39:2036–2041 1979;237:E214–E223 32. Wefers J, Connell NJ, Fealy CE, et al. Day- 10. Latva-Rasku A, Honka MJ, Kullberg J, et al. 21. Peronnet F, Massicotte D. Table of night rhythm of skeletal muscle metabolism is The SGLT2 inhibitor dapagliflozin reduces liver fat nonprotein respiratory quotient: an update. Can disturbed in older, metabolically compromised – but does not affect tissue insulin sensitivity: a J Sport Sci 1991;16:23 29 individuals. Mol Metab 2020;41:101050 randomized, double-blind, placebo-controlled 22. Ackermans MT, Pereira Arias AM, Bisschop 33. Lin X, Zhang X, Guo J, et al. Effects of study with 8-week treatment in type 2 diabetes PH, Endert E, Sauerwein HP, Romijn JA. The exercise training on cardiorespiratory fitness and fi patients. Diabetes Care 2019;42:931–937 quanti cation of gluconeogenesis in healthy men biomarkers of cardiometabolic health: a 11. Merovci A, Abdul-Ghani M, Mari A, et al. by (2)H2O and [2-(13)C]glycerol yields different systematic review and meta-analysis of results: rates of gluconeogenesis in healthy Effect of dapagliflozin with and without acipimox randomized controlled trials. J Am Heart Assoc men measured with (2)H2O are higher than on insulin sensitivity and insulin secretion in 2015;4:e002014 those measured with [2-(13)C]glycerol. J Clin T2DM males. J Clin Endocrinol Metab 2016;101: 34. Mudaliar S, Henry RR, Boden G, et al. Endocrinol Metab 2001;86:2220–2226 1249–1256 Changes in insulin sensitivity and insulin 23. Steele R. Influences of glucose loading and 12. Most J, Redman LM. Impact of calorie secretion with the sodium glucose cotransporter of injected insulin on hepatic glucose output. Ann restrictiononenergymetabolisminhumans.Exp 2 inhibitor dapagliflozin. Diabetes Technol Ther N Y Acad Sci 1959;82:420–430 Gerontol 2020;133:110875 2014;16:137–144 24. McMurray JJV, Solomon SD, Inzucchi SE, 13. Johnson ML, Distelmaier K, Lanza IR, et al. 35. Tuominen JA, Ebeling P, Bourey R, et al. et al.; DAPA-HF Trial Committees and Investi- Mechanism by which caloric restriction improves gators. Dapagliflozin in patients with heart failure Postmarathon paradox: insulin resistance in the insulin sensitivity in sedentary obese adults. and reduced ejection fraction. N Engl J Med face of glycogen depletion. Am J Physiol 1996; – – Diabetes 2016;65:74 84 2019;381:1995–2008 270:E336 E343 fl 14. Esterline RL, Vaag A, Oscarsson J, Vora J. 25. Perkovic V, Jardine MJ, Neal B, et al.; 36. Cusi K, Bril F, Barb D, et al. Effect of canagli ozin MECHANISMS IN ENDOCRINOLOGY: SGLT2 CREDENCE Trial Investigators. Canagliflozin and treatment on hepatic triglyceride content and fi inhibitors: clinical bene ts by restoration of renal outcomes in type 2 diabetes and glucose metabolism in patients with type 2 diabetes. – normal diurnal metabolism? Eur J Endocrinol nephropathy. N Engl J Med 2019;380:2295–2306 Diabetes Obes Metab 2019;21:812 821 2018;178:R113–R125 26. Wiviott SD, Raz I, Bonaca MP, et al.; 37. Targher G, Byrne CD. Clinical Review: 15. World Medical Association. World Medical DECLARE–TIMI 58 Investigators. Dapagliflozin nonalcoholic fatty liver disease: a novel cardio- Association Declaration of Helsinki: ethical and cardiovascular outcomes in type 2 diabetes. metabolic risk factor for type 2 diabetes and its principles for medical research involving human N Engl J Med 2019;380:347–357 complications. J Clin Endocrinol Metab 2013;98: subjects. JAMA 2013;310:2191–2194 27. Cioffi I, Evangelista A, Ponzo V, et al. 483–495 16. Schoffelen PF, Westerterp KR, Saris WH, Ten Intermittent versus continuous energy restriction on 38. Matsuba I, Kanamori A, Takihata M, et al. Hoor F. A dual-respiration chamber system with weight loss and cardiometabolic outcomes: a Canagliflozin increases calorie intake in type 2 automated calibration. J Appl Physiol (1985) systematic review and meta-analysis of randomized diabetes without changing the energy ratio 1997;83:2064–2072 controlled trials. J Transl Med 2018;16:371 of the three macronutrients: CANA-K study. 17. Brouwer E. On simple formulae for 28. Patterson RE, Sears DD. Metabolic effects of Diabetes Technol Ther 2020;22:228–234 calculating the heat expenditure and the intermittent fasting. Annu Rev Nutr 2017;37: 39. Ferrannini G, Hach T, Crowe S, Sanghvi A, quantities of carbohydrate and fat oxidized in 371–393 Hall KD, Ferrannini E. Energy balance after metabolism of men and animals, from gaseous 29.KomoroskiB,VachharajaniN,BoultonD, sodium-glucose cotransporter 2 inhibition. exchange (Oxygen intake and carbonic acid et al. Dapagliflozin, a novel SGLT2 inhibitor, Diabetes Care 2015;38:1730–1735