Unravelling the effect of renal denervation on glucose homeostasis: more questions than answers?

The sympathetic nerves are a part of the autonomic nervous system, regulating many functions of the human physiology. Central sympathetic neurons are located in the rostral ventrolateral medulla (RVLM), which is a key area for the regulation of BP and metabolism [12, 13]. The RVLM neurons signal directly the sympathetic preganglionic neurons, located in the spinal cord, and innervate several organs, while feedback is conveyed by a number of afferent inputs from carotid and organ receptors (i.e., mechanoreceptors, chemoreceptors), as well as hormones [14]. SNS plays an important role in regulation of daily energy expenditure by controlling metabolic rate, food intake and temperature. It has been generally recognized that increased sympathetic neural activity (SNA) results in catabolic effects on glucose and lipid metabolism, whereas increased parasympathetic neural activity results in anabolic effects. A number of afferent nerves from peripheral organs convey metabolic information that modulate activation of RVLM. Circulating molecules such as insulin and angiotensin, which are able to cross the blood–brain barrier can also influence central sympathetic outflow and thereby modulate peripheral lipid and glucose metabolism [15]. Moreover, sympathetic neurotransmitter norepinephrine and epinephrine are known to affect glucose transport and metabolism in the liver, pancreas, adipose tissue, and skeletal muscle. The liver, which plays a key role in glucose metabolism, is richly innervated by the autonomic components from the splanchnic sympathetic nerves and vagal parasympathetic nerves. Additionally, the part of sympathetic nerve fibers which innervate the liver arise directly from the hypothalamus, which is responsible for food intake and appetite regulation [16]. SNA and catecholamines increase glucose by activation of α1 and β2 receptors in the liver, leading to glycogenolysis and gluconeogenesis.

Furthermore, sympathetic nerves innervating skeletal muscle can modulate glucose uptake and glycogenolysis independent of concomitant increase in plasma insulin levels, via activation of β2 adrenergic receptors [17, 18]. Of note, administering a medical β2 agonist appears to improve glucose tolerance due to increased glucose uptake in skeletal muscle [19]. Conversely, neuronal stimulation of α-adrenergic receptors in arterioles elicits vasoconstriction, thus decreasing glucose utilization and leading to peripheral insulin resistance (IR). All these processes increase BP and glucose concentration in the blood, actions that are expected after activation of the SNS. However, SNS overactivation, due to chronic increase of stimulating factors or decreased activation of the parasympathetic system, contributes to the development of resistant hypertension, obesity, metabolic syndrome and IR.

The kidneys also play a role in glucose homeostasis by ensuring that glucose is not lost in urine through three membrane protein-sodium-glucose cotransporters (SGLT1, SGLT2, and GLUT2), which are responsible for glucose reabsorption [20]. Additionally, multiple studies have shown that the human kidney, through renal gluconeogenesis, releases significant amounts of glucose into circulation in the postprandial state under both physiological and pathological conditions and seems to be regulated by postprandial increases in SNS activity [21, 22].

Finally, intrarenal sympathetic hyperactivity may cause increased renal afferent nerve signaling, leading to further stimulation of the central SNS, thus initiating a vicious cycle (Fig. 1). RDN breaks this vicious cycle resulting in suppression of the SNS, and thereby improved glucose metabolism and insulin sensitivity. Therefore, there is a pathophysiologic interaction among insulin resistance, obesity and the central nervous system, which is potentially mediated by ameliorating renal sympathetic afferent and efferent signals. Interestingly, in the clinical use of a centrally acting sympatholytics drug, which are known to reduce central sympathetic activity [23], they reduce efferent sympathetic nerve traffic to the kidney, suppress glucose intolerance and increase skeletal muscle blood flow with less glycogenolysis. Therefore, an interventional procedure mimicking sympatholytics and reducing the renal-SNS crosstalk, such as RDN, may also lead to a benefit in the homeostasis of glucose and insulin, resulting in better disease control.

One of the first animal trials that has studied the effect of denervation on glucose metabolism was in 1977, where Szalay et al. studied the effect of glucose per os loading on dogs that have been previously unilaterally surgically denervated [24]. They demonstrated that glucose tubular reabsorption in the kidney was significantly decreased in denervated dogs, compared to healthy controls.

Schaan et al. [25] evaluated the efficacy of bilateral denervation in rats with diabetic nephropathy; they studied 26 rats—13 of them with diabetic nephropathy and 13 healthy—that were randomized to either surgical RDN or no procedure, demonstrating a significant decrease in the levels of cortical GLUT1 protein toward normal values, without regression of diabetic-induced albuminuria. The same research team demonstrated, in an animal model of diabetic and non-diabetic rats, that there is direct association between renal sympathetic activity and GLUT2 levels, as denervation reduced GLUT2 levels in both non-diabetic and diabetic rats (-21% and -15%, respectively) and concluded that denervation modulates GLUT2 expression in kidney cells [26].

The multiple pleiotropic effects of surgical RDN on glucose metabolism were also demonstrated in an animal setting of uninephrectomised diabetic rats [27], in which denervation was associated with reduction in BP and renal tissue noradrenaline levels. RDN was also associated with improvement in glucose metabolism and insulin sensitivity, reflected by an increase in peripheral in-cell glucose uptake, as well as significant decrease of SGLT2 levels in renal cells, followed by attenuation of diabetic glycosuria.

As forewritten, insulin sensitivity plays a pivotal role in glucose metabolism and RDN could lead to attenuation of insulin sensitivity. Iyer et al. [28] carried out an animal study, showing that RDN could lead to regression of high hepatic insulin resistance. Specifically, they evaluated insulin sensitivity in a nonhypertensive obese canine model, before and after a 6-week period of high-fat diet, and after either surgical RDN or sham procedure. They reported a significant decrease of hepatic insulin sensitivity after RDN, suggesting a direct effect of RDN, via downregulation of hepatic gluconeogenesis through the natriuretic peptide pathway. Similar findings were demonstrated at another animal study by Chen et al. resulting that RDN, when applied unilaterally in high-fat-fed rats, led to 18.2% decrease of glucogenolysis and 16.3% decrease of gluconeogenesis; on the other hand, when applied bilaterally, it led to 31.9% decrease of glucogenolysis and 42.8% decrease of gluconeogenesis [29].

Another possible mechanism that denervation could lead to improved glucose tolerance and metabolism in wild-type and diabetic mice may be increased glycosuria, via the decreased expression of GLUT2 in renal tissue. These findings are supported by an animal study, in which improved glucose levels and increased glycosuria are met in mice undergoing renal denervation and in hypothalamic arcuate nucleus-specific pro-opiomelanocortin deficient (ArcPomc − / −) mice, demonstrating diminished renal sympathetic activity in both animal populations [30].

Although surgical denervation leads to glucose metabolism improvement, there were few data regarding the effect of catheter-based RDN. Pan et al. carried a study at which they divided 33 diabetic dogs into three arms: bilateral RDN arm, left RDN arm, and sham procedure arm [31]. They demonstrated that fasting plasma glucose (9.64 ± 1.57 mmol/L vs. 5.12 ± 1.08 mmol/L; P < 0.0001), fasting insulin (16.19 ± 1.43 mIU/mL vs. 5.07 ± 1.13 mIU/mL; p < 0.0001), and insulin resistance in the bilateral RDN arm, had statistically significant reduction, compared with the sham arm, supporting their hypothesis that multi-electrode catheter-based leads to significant reduction of gluconeogenesis and glycogenolysis 3 months post-intervention, without any procedure-related adverse events. Another animal study that evaluated the effect of increased sympathetic activity on hypertension and insulin sensitivity had similar findings [32]. The researchers included glucose-fed rats or fructose-fed rats that underwent bilateral RDN using cryoablation or sham denervation and demonstrated that RDN led to significant decrease of renal noradrenaline spillover and plasma renin activity, with an imminent effect not only in BP reduction but also to an improvement in glucose-to-insulin ratio and insulin sensitivity.

De Oiveira et al. also studied the effect of catheter RDN on diabetic mice [33]. For the purpose of the study, they divided mice into three groups: control, diabetic and diabetic that underwent RDN. They demonstrated that renal sympathetic activity was higher in diabetic rats, while RDN led to significant reduction of both glycemia and glycosuria, as well as reduction of SGLT2 expression on the kidney cells’ membrane and normalization of renal sympathetic activity. Regarding SGLT2 expression, newfangled evidence supports that SGLT2 expression is increased in the proximal tubule of the kidney, especially on the presence of HF [34]. In specific, the researchers performed RDN in rats with HF and demonstrated that sodium and glucose excretion increased, as a response to dapagliflozin. Moreover, in vitro high levels of noradrenaline induced translocation of SGLT2 to the cell membrane, highlighting the pleiotropic effects that RDN could have in human patients with HF, by not only reducing BP levels, but also improving glucose metabolism and altering the expression of SGLT transporters at renal tissue.

Following the results of preclinical studies, a number of clinical trials investigating the use of RDN in hypertension also reported outcomes regarding glycemic control (Table 1). Bhatt et al. [35], in a randomized trial (SYMPLICITY HTN 3) assessing the safety and efficacy of RDN in patients with resistant hypertension, investigated the effect of RDN on the levels of Hba1C, along with outcomes regarding BP controls. The trial, which included 364 patients in the RDN arm and 171 in the sham procedure arm, reported no benefit of RDN in the levels of glycated hemoglobin in all patients, as well as in patients with diabetes. Similarly, Rosa et al. [36], in the Prague-15 study, which compared RDN to intensified pharmacotherapy with spironolactone in resistant hypertensives, showed no statistically significant reduction of fasting glucose levels in patients that received RDN treatment, compared to the pharmacotherapy group.

Following studies, however, also investigating the role of RDN in BP control, showed benefit after RDN in several parameters of glucose control. In specific, Bohm et al. [37] evaluated the RDN-associated BP reduction in patients with hypertension, as well as the HbA1c and fasting glucose levels. The study, which enrolled a total of 998 patients, showed no difference between baseline and 6 months levels of HbA1c in both diabetics and non-diabetics, while also reporting significant reduction of fasting glucose levels in both aforementioned groups (p < 0.05). Similar reduction in fasting glucose levels were observed in a study by Pourmoghaddas et al. [38], which included 30 patients with resistant hypertension undergoing RDN. The investigators reported a reduction in fasting glucose levels of −4.50 mg/dL at 1-year follow-up, as well as significant reduction in BMI and waist circumference, parameters associated with metabolic syndrome (p = 0.008 and p = 0.003, respectively).

Moreover, Hopper et al. [39], in the SYMPLICITY HF Feasibility study, which included 39 patients with HF and renal impairment, showed significant reduction of the 120-min oral glucose tolerance test (OGTT) at 12 months, compared to baseline. Also, similar studies have shown possible alterations of insulin sensitivity mediated by RDN. More specifically, Aripov et al. [40], in a study assessing RDN’s effect in 63 patients with resistant hypertension, noted a significant reduction of the homeostasis model assessment-insulin resistance (HOMA-IR) index, which was significantly reduced by a mean of 0.5 at 12 months.

Mahfoud et al. [41], in a pilot study, aimed to assess the impact of RDN in several aspects regarding glucose and insulin homeostasis. They enrolled 50 patients, out of which 37 underwent RDN and 17 served as control, and they evaluated them 1- and 3-month post-intervention. Besides significant reductions in BP in both follow-up periods, the study also revealed significant decrease in fasting glucose levels, insulin and C-peptide levels in the RDN cohort. Furthermore, patients undergoing RDN had significantly decreased HOMA-IR index and 2 h OGTT was reduced by 27 mg/dL (p = 0.012). Interestingly, similar changed were not observed in the control group, implying the positive effect of RDN in glycemic control markers.

Despite earlier positive results, Matous et al. [42], in a subsequent study, enrolled 51 resistant hypertensives, which were grouped regarding the presence of diabetes and the number of radiofrequency applications to each renal artery (greater or lower than 4). Regardless of the group on which each patient was assigned, the investigators reported significantly increased fasting levels of glucose, higher levels of HbA1c and no difference in the levels of C-peptide.

Similarly, Verloop et al. [43], in the DREAMS randomized controlled trial, evaluated the effect of RDN in patients with metabolic syndrome. In particular, they enrolled 29 patients with metabolic syndrome undergoing RDN and assessed, besides BP post-intervention, a number of metabolic markers and insulin sensitivity. The median insulin sensitivity, as assessed by the simple index assessing insulin sensitivity OGTT (SIiSOGTT) was not different at 6 months and 12 months after RDN. Furthermore, fasting glucose levels, fasting insulin levels and C-peptide levels were not significantly changed post-RDN.

Tsioufis et al. [44] also investigated the effect of RDN in the sympathetic control of glycemic status. In specific, the enrolled 17 hypertensive patients which fulfilled 4 or more criteria for metabolic syndrome, and assessed muscle sympathetic nerve activity (MSNA) both at rest and during 75 g OGTT. At 3-month follow-up, the investigators reported a reduction of MSNA bursts at rest and improved sympathetic response to OGTT in the RDN group. However, no significant improvement was noticed in the HOMA-IR index (p = NS).

In another study, Miroslawska et al. [45] investigated insulin resistance in 23 patients with resistant hypertension, undergoing RDN, using a hyperinsulinemic-euglycemic step clamp (HEC). Out of them, 18 met the criteria for metabolic syndrome. At 4 months after RDN, mean fasting plasma glucose, median insulin and C-peptide concentration remained unchanged, compared to baseline. Furthermore, during HEC, the endogenous glucose production remained unchanged, while at high-dose insulin infusion, the measured glucose disposal remained unchanged to baseline. These results, thus, were indicative of no effect of RDN in peripheral and hepatic insulin sensitivity in patients with resistant hypertension.

Kampmann et al. [46], in their respective trial, aimed to study the effect of RDN in patients with resistant hypertension without diabetes. They included, therefore, 8 patients undergoing RDN, which were re-examined regarding BP metabolic parameters at 6 months. Regarding insulin sensitivity, as assessed by HEC and expressed as M-value, it was improved non-significantly, however investigators reported a significant correlation between greater insulin improvement and lower BMI (R = 0.75, p = 0.03). Moreover, the endogenous glucose production, as assessed by a 3-³H glucose tracer, was also non-significantly decreased during both the basal and clamp period of the test. Finally, no difference was reported in insulin, C-peptide, IGF-1 and glucagon levels.

Eikelis et al. [47], in a trial investigating how RDN affects the adipokine profile of patients with resistant hypertension, also investigated the response of insulin levels to sympathetic denervation. The study, which included 57 patients, showed at 3-month follow-up, besides a significant increase in adiponectin concentration, a significantly higher fasting insulin concentration after RDN, suggesting a possible positive effect in both lipid and glucose homeostasis.

Witkowksi et al. [48], in a trial assessing RDN in patients with resistant hypertension and sleep apnea, enrolled 10 patients, which were also followed up for glycemic control markers. The study showed that, in resistant hypertensive patients with sleep apnea, there was a significant decrease of plasma glucose concentration 2 h post-glucose administration, as well as in HbA1c levels.

Similarly, Daniels et al. [49] also investigated the effect of RDN in 20 patients with RDN and sleep apnea. At 6-month follow-up, the investigators reported a non-statistically significant reduction of fasting glucose levels and a significant reduction of glucose levels following a 120 min OGTT. Other parameters, such as insulin sensitivity, HbA1c and percentage of pancreatic beta cell function, did not alter significantly from baseline at the time of the follow-up. Moreover, another study by Warchol-Celinska et al. [50] involving 60 patients with resistant hypertension and sleep apnea, also examined the effect of RDN on glycemic control variables. The total cohort of patients was randomly assigned to either RDN arm (n = 30) or control arm (n = 30). Regarding glucose metabolism, at 3 months the study did not show any change from baseline at the RDN arm in fasting glucose levels, fasting insulin levels or HbA1c levels.

A following study by Manukyan et al. [51], which aimed to evaluate RDN and its relation to renal resistance and function, also evaluated several parameters in regards to glycemic control. The trial included 59 patients with resistant hypertension and type 2 diabetes. At 12-month follow-up, the investigators reported no significant changes in fasting plasma glucose and HbA1c, in both patients with a renal resistance index (RRI) ≥ 0.7 and < 0.7.

Finally, an observational study by Miroslawska et al. [52] evaluated long-term metabolic effects of RDN in patients with resistant hypertension. In specific, 20 resistant hypertensives without diabetes underwent RDN and were followed up at 6 and 24 months. The study reported no significant change in several metabolic parameters, such as HbA1c, fasting glucose levels, fasting insulin levels, fasting C-peptide levels and HOMA-IR.