Empagliflozin modulates renal sympathetic and heart rate baroreflexes in a rabbit model of diabetes

We determined whether empagliflozin altered renal sympathetic nerve activity (RSNA) and baroreflexes in a diabetes model in conscious rabbits. Diabetes was induced by alloxan, and RSNA, mean arterial pressure (MAP) and heart rate were measured before and after 1 week of treatment with empagliflozin, insulin, the diuretic acetazolamide or the ACE inhibitor perindopril, or no treatment, in conscious rabbits. Four weeks after alloxan administration, blood glucose was threefold and MAP 9% higher than non-diabetic controls (p < 0.05). One week of treatment with empagliflozin produced a stable fall in blood glucose (−43%) and increased water intake (+49%) but did not change RSNA, MAP or heart rate compared with untreated diabetic rabbits. The maximum RSNA to hypotension was augmented by 75% (p < 0.01) in diabetic rabbits but the heart rate baroreflex was unaltered. Empagliflozin and acetazolamide reduced the augmentation of the RSNA baroreflex (p < 0.05) to be similar to the non-diabetic group. Noradrenaline (norepinephrine) spillover was similar in untreated diabetic and non-diabetic rabbits but twofold greater in empagliflozin- and acetazolamide-treated rabbits (p < 0.05). As empagliflozin can restore diabetes-induced augmented sympathetic reflexes, this may be beneficial in diabetic patients. A similar action of the diuretic acetazolamide suggests that the mechanism may involve increased sodium and water excretion. Graphical abstract Graphical abstract


Introduction
Individuals with diabetes have an increased risk of developing cardiovascular disease which increases their morbidity and mortality. A new generation of glucose-lowering drugs have been developed that increase sodium and glucose excretion in urine by inhibiting the sodium-glucose cotransporter 2 (SGLT2). In patients with type 2 diabetes, treatment with SGLT2 inhibitors such as empagliflozin has been reported to reduce cardiovascular mortality [1,2]. SGLT2 inhibitors block glucose and sodium reabsorption in the close proximal tubules of the renal nephron where SGLT2 is primarily expressed. Excessive resorption of sodium/glucose in the proximal tubules leads to activation of renal afferent nerves and results in central sympathetic overactivity in individuals with poorly controlled type 2 diabetes in whom sympathetic nervous system (SNS) activation is also associated with cardiovascular outcomes [3]. We hypothesise that through their actions in the proximal tubule, SGLT2 inhibitors also modulate the renal SNS. Selective SGLT2 inhibition has been shown to reduce the power of low-frequency fluctuations in systolic BP in a rat model of the metabolic syndrome displaying hypertension, diabetes and hyperlipidaemia, which has been interpreted as a reduction in sympathetic vasomotor tone [4].
Currently, the effect of an SGLT2 inhibitor on direct recording of renal sympathetic nerve activity (RSNA) is unknown and thus we assessed the changes in RSNA following alloxan-induced diabetes in conscious rabbits, a model in which RSNA can be compared between groups [5]. We explored the effects of acute and 1 week treatment with SGLT2 inhibitor empagliflozin on blood biochemical, cardiovascular and RSNA baroreflex variables and compared these responses with the effects of insulin, the natriuretic acetazolamide and the ACE inhibitor perindopril.

Animals
Experiments were conducted in 48 male New Zealand white rabbits weighing 2.6-3.3 kg. Experiments were approved by the Alfred Medical Research Education Precinct Animal Ethics Committee and conducted according to the Australian code for the care and use of animals for scientific purposes (2013). Rabbits were housed individually with free access to water and food under controlled humidity (40-70%) and temperature (22 ± 2°C) and a 12/12 light cycle.

Induction of diabetes
Alloxan [6] (60 mg/kg i.v.; Sigma-Aldrich, Saint-Louis, MO, USA) was injected under brief propofol sedation (6 mg/kg i.v.; Fresofol 1%, Fresenius Kabi, Pymble, NSW, Australia). Blood glucose was monitored 4 hourly during the first 48 h and animals were supplemented with 25% glucose when necessary. From day 2, rabbits received long-acting glargine insulin s.c. daily (Lantus, Sanofi, Macquarie Park, NSW, Australia) to maintain their blood glucose in the target range of 15-25 mmol/l [7]. One rabbit became irreversibly hypoglycaemic after alloxan and one other became unwell during the study, and these were euthanised.
Daily body weight, blood glucose, and water and food intake measurements Body weight and blood glucose measurements at day 0 and water and food intake, measured over 3 days before alloxan administration, were used as baseline. All variables were then measured daily before water and food replenishment. To evaluate the effect over 1 week of each treatment drug, the 3 days before treatment/non-treatment were used as that control period. Blood glucose was measured by glucometer (Accu-Chek Performa, Roche, Bella Vista, NSW, Australia) from a drop of blood taken from a superficial ear vein.

Blood biochemistry measurements
Blood glucose, sodium, chloride, potassium, ionised calcium (iCa), total carbon dioxide (TCO 2 ), creatinine, blood urea nitrogen/urea (urea), haemoglobin, haematocrit and anion gap (AnGap) were measured from a blood sample (0.2 ml) using the i-stat system (Chem8 + cartridge, Abbott, Melbourne, VIC, Australia) before alloxan administration, before the first treatment drug and at the end of the 1 week treatment period.

Conscious experimental measurements
One week after electrode surgery, rabbits were placed in single holding boxes in the laboratory and the central ear artery was catheterised to measure mean arterial pressure (MAP) and heart rate [5]. RSNA was rectified and integrated [9]. See the ESM Methods for details. RSNA was normalised by the nasopharyngeal response expressed as 100 normalised units (nu). After 30 min recovery from handling, control measurements were recorded for 60 min and blood glucose was measured. Rabbits were then given a single dose of the treatment drug and data were recorded over the following 3 h, with blood glucose measured each hour.
For the 1 week treatment, rabbits received the same drug but at a chronic dose (ESM Table 1). After 1 week of treatment, cardiovascular measurements were recorded for 3 h, with hourly blood glucose. RSNA and heart rate baroreflexes were derived from slow ramp rises and falls in MAP by i.v. infusions of 25 mg/kg phenylephrine (Sigma-Aldrich) and 30 mg/kg nitroprusside (Fluka AG, Buchs, Switzerland), respectively. Noradrenaline (norepinephrine) spillover was determined using [ 3 H]-noradrenaline (4070 Bq kg −1 min −1 i.v.; Perkin Elmer Life Sciences, Boston, MA, USA; see ESM Methods). Rabbits were killed with sodium pentobarbitone (160 mg/kg i.v.; Lethobarb, Virbac, Milperra, NSW, Australia) and organs removed and weighed.
Baroreflex curves and spectral analysis methods MAP, heart rate and RSNA were averaged over 2 s intervals and fitted to a five-variable sigmoid logistic function to produce MAP-RSNA and MAP-heart rate baroreflex curves as previously described [10]. Variables included the lower plateau (minimum RSNA or heart rate), the range between the lower and upper plateau and the median blood pressure at half the reflex range (BP50). The beat-to-beat instantaneous MAP over the 3 h of chronic treatment was analysed for spectral analysis [11]. See the ESM Methods for further details.

Confirmation of ACE inhibition
Rabbits in the ACE inhibitor perindopril treatment group were given a bolus injection of angiotensin I (50-200 pmol, Auspep, Tullamarine, VIC, Australia) before and after treatment with perindopril, both acute and chronic, and the changes in MAP were compared to determine whether the doses of perindopril were adequate to block the renin-angiotensin system.

Statistical analysis
Data were expressed as mean ± SEM or mean difference ± standard error of the difference (SED) and analysed by a mixed model multivariate ANOVA followed by Student's t tests to determine the effect of diabetes (p diab ), differences at baseline (p base ), and treatment effects compared with control (p treat ) and between-group treatment effects (p delta ) with a pooled variance [12]. Family-wise error was controlled by the Bonferroni procedure and multisample asphericity by the Greenhouse-Geisser correction [13]. A probability of p < 0.05 was considered significant.

Effects of diabetes on metabolic and cardiovascular variables
Alloxan induced a stable level of hyperglycaemia with blood glucose 20.5 ± 0.4 mmol/l compared with 6.5 ± 0.4 mmol/l in non-alloxan rabbits (p diab < 0.001). Diabetic rabbits gained weight at a slower rate than controls (+3% vs +7%, respectively, p diab < 0.001) even though water intake was greatly elevated (+55% vs −17%, p diab < 0.001) and food consumption was greater (p diab = 0.045; ESM Fig. 1). The MAP of diabetic rabbits was 9% greater than non-diabetic (p diab = 0.016; Fig. 1). There was no effect of diabetes on heart rate, total RSNA, burst amplitude, burst incidence or burst frequency ( Fig. 1). Plasma sodium and chloride were reduced by 5-6% compared with non-diabetic animals (p diab < 0.001; ESM Table 2) but potassium and other blood biochemistry measures were similar (ESM Table 2).

Acute effects of treatments on blood glucose, cardiovascular variables and RSNA
After establishment of diabetes, rabbits underwent acute treatment with empagliflozin, insulin, acetazolamide, perindopril or vehicle and the effects were followed for 3 h. In the vehicletreated group, blood glucose and MAP remained stable over 3 h after vehicle injection and not different from the control period (p treat > 0.05; Fig. 2), but heart rate and total RSNA increased (+13% and + 26%, respectively, p treat < 0.05; Fig Table 3).
Empagliflozin induced a fall in blood glucose of −5.5 ± 1.2 mmol/l over 3 h (−31%, p treat < 0.001, n = 7; Fig. 2). There was little change in MAP but we observed linear  increases in heart rate and RSNA (total RSNA, burst incidence and frequency, all p linearity < 0.001), similar to the effect of vehicle (ESM Table 3). Treatment with insulin lowered blood glucose over 3 h (−12.4 ± 2.2 mmol/l, p treat < 0.001, n = 8). There was a linear fall in MAP of 6% accompanied by increases in heart rate (17%) and total RSNA (41%, p treat < 0.001), which was similar to vehicle treatment (Fig. 2). However, MAP in the insulin-treated rabbits was lower after 3 h than the MAP in non-diabetic controls (68 ± 2 mmHg vs 73 ± 2 mmHg, p = 0.022) despite similar blood glucose. We observed a linear relationship between change in blood glucose and fall in MAP in the insulin group (r = 0.57, p < 0.001), but not in the empagliflozin group (ESM Fig. 3). Acetazolamide and perindopril administration had no observable effects on blood glucose or MAP (p treat > 0.05) but heart rate increased by 10% and 13%, respectively (both p treat < 0.001), and RSNA in response to acetazolamide increased by 39% (p treat < 0.01), all similar to the responses to vehicle (Fig. 2, ESM Table 3). In the non-diabetic group, MAP, heart rate and total RSNA remained stable over the 3 h recording period (all p treat > 0.05; ESM Fig. 2).

Differences between treatment groups in control period for acute treatment
In diabetic rabbits, there were no differences in blood glucose, MAP and heart rate between treatment groups and the vehicletreated group during the control period (ESM Table 4). Whilst control total RSNA was 52% lower in the empagliflozin group compared with the vehicle-treated group (ESM Table 4), it remained 50% lower throughout the treatment period (mean over 3 h: 3.1 ± 0.5 vs 6.1 ± 1.2 nu, respectively, p = 0.005).
There were no differences in control total RSNA between the untreated diabetic group and the other treatment groups (ESM Table 4).

One week effects of empagliflozin, insulin, acetazolamide and perindopril
Cardiovascular variables and RSNA Diabetic rabbits were treated for 1 week with empagliflozin, insulin, acetazolamide or perindopril, or were untreated for that period. Between the acute and chronic recording days for the non-diabetic group, there was an increase in MAP of 6% (p treat = 0.014), but in this group neither heart rate nor total RSNA changed between the two experiments (ESM Fig. 4). However, all diabetic animals decreased MAP and heart rate over the week between the two recording days and there was no difference among groups for MAP or RSNA. In untreated diabetic rabbits, mean MAP and heart rate decreased by 8% and 7%, respectively (both p treat < 0.001; Fig. 2). Total RSNA was also 13% lower (p treat = 0.022) due to a reduction in burst amplitude (−32%, p treat < 0.001; ESM Table 3). The fall in MAP and the stable heart rate and total RSNA after 1 week empagliflozin treatment were not different from those in untreated diabetic rabbits (Fig. 2), although there was an increase in RSNA amplitude (ESM Table 3). Similarly, we observed no difference between insulin-treated and untreated groups in MAP and total RSNA changes but there was a 10% increase in heart rate compared with the untreated group in which heart rate decreased (p delta = 0.003; Fig. 2). Acetazolamide did not alter the pattern observed in untreated rabbits, but after perindopril treatment MAP was 9% lower than in the untreated group (p = 0.045; Fig. 2).
Blood glucose, body weight, and water and food intake In the untreated diabetic group, blood glucose, body weight, and water and food intake remained stable and similar to the control levels over the 1 week period (ESM Fig. 5). Empagliflozin treatment over 1 week produced a stable decrease in blood glucose (43%, p treat < 0.001) without requiring insulin to prevent hyperglycaemia. Water consumption increased by 49% (p treat < 0.001), markedly greater than changes in water intake in untreated diabetic rabbits (p delta < 0.001; ESM Fig. 5). There was a small but significant 2% loss of body weight during empagliflozin treatment (p treat = 0.001) but the drug did not alter food intake (ESM Fig. 5). In the insulin-treated group, blood glucose gradually declined so that by day 7 it was 32% lower than control (p linearity < 0.001) but similar to the empagliflozin-treated group (ESM Fig. 6). Weight gain in insulin-treated rabbits (+2%, p treat < 0.001) was greater than in untreated rabbits (p delta < 0.001) despite food intake not changing, but water intake was reduced by 26% (p treat < 0.001, p delta = 0.024; ESM Fig. 5). In the acetazolamide group, the ongoing treatment with insulin maintained blood glucose at a stable level over 1 week (ESM Fig. 6), but water and food intake both fell markedly over the course of 7 days (−44% and −35%, p treat < 0.001). This was associated with a 3% loss of body weight (p treat < 0.001; ESM Fig. 5). One week treatment with perindopril had no observable effects on blood glucose, water or food consumption and body weight (all p treat > 0.05; ESM Figs. 5, 6). In non-diabetic rabbits, there was a small but significant 1% weight gain over 1 week (p treat < 0.001).
Blood biochemistry In the untreated diabetic group, we observed no change in blood biochemistry after the 1 week period. Empagliflozin, insulin or perindopril treatment was not associated with any differences in biochemical measurements compared with untreated diabetic rabbits (ESM Table 5). Acetazolamide lowered potassium levels by −17%, a greater effect than in untreated diabetic rabbits (p delta = 0.012), and increased blood urea concentrations (+29%, pdelta = 0.042 vs untreated; ESM Table 5), but did not affect any other blood variables. Blood biochemical variables remained stable over 1 week in the non-diabetic group (ESM Table 6).
Organ weights Kidney weight, normalised to body weight, was 20% greater in the untreated diabetic group compared with the non-diabetic group. We observed no additional effects of any of the 1 week treatment drugs (ESM Table 7). There were no detectable effects on either whole heart or left ventricular weight of diabetes or any of the treatment drugs (ESM Table 7).

Effects of diabetes and 1 week treatments on RSNA and heart rate baroreflexes
Effects of diabetes The main effect of diabetes on RSNA baroreflex curves was to markedly increase the upper plateau by 75%, range (+69%) and gain (+87%, p diab < 0.01; ESM Fig. 7). A similar effect was seen in the upper plateau and range of the RSNA amplitude curves but the upper plateau and range of the RSNA frequency curves were inhibited. Heart rate baroreflex curves were similar in the diabetic and non-diabetic groups (ESM Fig. 7).
Effects of 1 week treatment Empagliflozin reduced both the RSNA upper plateau and range by 34% (p diab = 0.016) such that they were similar to non-diabetic rabbits (Fig. 3). A similar effect was observed after acetazolamide treatment, with 30% reductions (p diab < 0.05; Fig. 3). Empagliflozin also reduced the heart rate baroreflex upper plateau (p diab = 0.046; Fig. 3). There was little effect of these drugs on RSNA burst frequency curves (ESM Table 8). There were no effects of insulin or perindopril on either the RSNA or heart rate baroreflex. Effect on short-term cardiovascular variability Low, mid, high and total power of MAP and heart rate were similar in non-diabetic and diabetic rabbits (ESM Table 9). In empagliflozin-treated rabbits, MAP power in all bands and low and high heart rate power were similar to diabetic controls. By contrast, mid-frequency heart rate power was elevated compared with controls (ESM Table 9). Insulintreated rabbits exhibited reduced mid-and high-frequency MAP power and reduced high-frequency heart rate power compared with untreated diabetic controls (ESM Table 9). Acetazolamide treatment did not affect any MAP power but low-and mid-frequency heart rate power was elevated compared with controls (ESM Table 9). Perindopril-treated rabbits had lesser low-, mid-and high-frequency and total MAP power and reduced high-frequency heart rate power compared with untreated diabetic rabbits. Baroreflex gain estimated by cross-spectral analysis was similar in all groups (ESM Table 9). Thus, the pattern of effects observed in the empagliflozin group more closely aligns with the effects of acetazolamide than with insulin or perindopril.

Effects of angiotensin I
To confirm the inhibition of ACE by perindopril, angiotensin I was administered after cardiovascular control period recording and at the end of the 3 h recording. Before perindopril treatment, angiotensin I induced a 60% increase in MAP (ESM Fig. 8). After perindopril treatment, MAP was only increased by 25% after angiotensin I administration, whether after the acute or the chronic treatment doses of perindopril (ESM Fig. 8), confirming ACE inhibition in both cases.

Principal findings and mechanisms
The present study examined the SNS effects of acute and 1 week treatment effects of SGLT2 inhibition, in alloxaninduced diabetic rabbits with hyperglycaemia and mild hypertension, compared with non-diabetic controls. One week treatment with empagliflozin produced a stable and marked reduction in blood glucose (−43%) and an increase in total noradrenaline spillover, but did not change RSNA, MAP or heart rate. The main effect on RSNA was to reduce the exaggerated response to hypotension (baroreflex upper plateau) to a level similar to that seen in non-diabetic controls. These changes were also observed with acetazolamide but not insulin, which indicates that empagliflozin modulation of the SNS is related to its natriuretic rather than its glucose-lowering action. The main effect of empagliflozin on the cardiac baroreflex was a reduction in the upper plateau, which most likely reflects a reduction in cardiac sympathetic nerve activity (SNA) since vagal activity is inhibited and baro-sensitive cardiac SNA is at its greatest at this point on the curve. This change to baroreflex responsiveness is similar to that induced during mild-intensity exercise [14] and is thought to buffer the BP against vasodilation associated with exercise. Presumably, in early diabetes these same changes would allow BP and single-nephron GFR to be maintained at a higher level. Enhanced muscle SNA responses to exercise and the cold pressor test have been reported in diabetic individuals, and occurred without changes to the gain of the baroreflex [15]. However, it has also been reported that muscle SNA responses to an oral glucose tolerance test are blunted in type 2 diabetic patients [16]. In our model there was greater RSNA baroreflex gain in diabetic compared with non-diabetic rabbits. Thus, for small alterations in BP near the resting level, the responsiveness of RSNA was enhanced. The augmentation of the total integrated RSNA baroreflex in diabetic rabbits could be attributed to an increase in the upper plateau and range of the RSNA burst amplitude baroreflex. The mechanism underlying this may be additional discharges from previously silent presynaptic neurons in the central nervous system, as modulation of burst amplitude reflects recruitment of neurons [17]. With the restoration of the RSNA baroreflex involving reduced sympathetic responsivity (gain), we might have expected an increase in short-term BP variability, but surprisingly there was no effect of empagliflozin or acetazolamide on BP variability. By contrast, the treatments that did not affect the RSNA baroreflex, insulin and perindopril, did reduce short-term BP variability.
The mechanism whereby empagliflozin was able to normalise the RSNA baroreflex remains to be established. SGLT2 plays a key role in sodium balance, reabsorbing from the lumen of the early proximal tubule one molecule of sodium for every glucose molecule. Following SGLT2 inhibition, increased delivery of sodium and glucose also influences tubular function including the absorption of other solutes and regulation of renal blood flow. We speculate that SGLT2 inhibition also modulates renal afferent activity to inhibit sympathetic neurons to not only prevent tachycardia in response to volume lowering, but also to suppress maximal RSNA in response to BP lowering. This is most likely due to changes in renal sodium balance as we observed a similar effect with acetazolamide, a carbonic anhydrase inhibitor that also acts as a proximal tubular natriuretic, partly by . Data are mean ± SEM, indicating between-animal variance. *p < 0.05, **p < 0.01 for treated diabetic vs untreated diabetic rabbits; † p < 0.05 for treated diabetic vs untreated non-diabetic rabbits modulating the activity of the sodium/proton exchanger isoform 3 [18]. Notably, the effect of empagliflozin was not observed with a single dose, suggesting that it represents a realignment of renal function rather than a direct (instantaneous) effect of the drug on RSNA. Treatment with insulin did not alter the RSNA response, despite a similar lowering of blood glucose as for the SGLT2 treatment. ACE inhibition did not change the RSNA in this model, despite its known actions on renal and systemic haemodynamics and renoprotective actions in vivo. Noradrenaline spillover to plasma was greater in diabetic empagliflozin-treated rabbits than in untreated rabbits. The higher spillover likely accounted for greater arterial concentrations as noradrenaline clearance did not change. Higher noradrenaline release was also observed in animals treated with acetazolamide, suggesting it may be related to its comparable actions on plasma volume, BP and natriuresis. Taken together with the measurement of RSNA and heart rate baroreflexes, the higher spillover is likely to be in other vascular beds (not renal or cardiac) and less critical for long-term cardiovascular outcomes.

Strengths and weaknesses
A key strength of the current study is the use of alloxaninduced diabetic rabbits, a model of stable chronic hyperglycaemia, and the ability to have RSNA and cardiovascular variables recorded with these animals over several weeks without anaesthesia. Further, RSNA can be calibrated to allow comparison among different groups. We used comparison treatments that allowed us to compare the independent roles of hypoglycaemia, natriuresis and hypotension. We confirmed that acetazolamide had a diuretic effect because levels of urea increased but serum potassium was reduced, consistent with its known diuretic action [19]. We also confirmed that perindopril blocked angiotensin I conversion.
Limitations were that we used only one dose of each agent, that treatment was for only 1 week and that we did not use relevant combination treatments such as SGLT2 together with ACE inhibition. Chronic treatment with SGLT2 inhibitors such as empagliflozin in humans is most commonly associated with a small reduction in BP [20,21]. The effect size in this study was actually similar but was not statistically significant. In a rat model of diabetes, treatment for 7 weeks with empagliflozin had no effect on BP [22] but did increase urine volume and water intake without affecting sodium balance [22]. Clearly, increased water intake is consistent with glycosuria-mediated diuresis [23], but any reduction in blood volume was not enough to lower BP. Also, the effect of empagliflozin in rats was maximal after 3 days and stable for 7 weeks, suggesting that our study lasting 1 week may well reflect longer-term chronic treatment [22].

Meaning of the study
The lower levels of sympathetic excitation occurring when BP fluctuates may better protect single-nephron GFR by preventing excessive afferent vasoconstriction when the GFR is already reduced. This could also partly explain the reduced risk of acute kidney injury observed in the EMPA-REG OUTCOME trial in patients receiving empagliflozin [24]. Our findings may be particularly important for patients with comorbidities including established cardiovascular disease, chronic heart failure and chronic kidney disease, in whom RSNA is often augmented. The reduction in the upper plateau of the cardiac baroreflex may also indicate a beneficial effect of SGLT2 inhibition on reducing cardiac SNA during hypotension, and may be one explanation for the better long-term cardiovascular outcome [25].

Strengths and weaknesses in relation to other studies
Many studies have indicated that muscle SNA is increased in diabetes [3,26,27], and this has been associated with poor clinical outcomes and increased cardiovascular risk and end organ damage [26,28]. We hypothesised that SGLT2 inhibition would modulate resting RSNA but this was not the case. We did not observe any differences in plasma noradrenaline, total noradrenaline spillover or clearance in the diabetic compared with non-diabetic rabbits, suggesting that there was no increase in global SNA associated with diabetes. In humans with type 2 diabetes there were no changes in muscle SNA after several days of empagliflozin treatment, despite an increase in urine volume, weight loss and a reduction in BP [29]. In the present study, once RSNA was increased (by hypotension), the normalising effects of empagliflozin were clearly observed, with lower maximal RSNA and heart rate in diabetic rabbits treated with an SGLT2 inhibitor.

Future research
Clearly, a longer treatment would be an important next step to determine whether these effects on the SNS are sustained. Furthermore, it would be of interest to use this model to explore the end organ protective effects of combination treatments such as SGLT2 and ACE inhibition.