Slow but Steady—The Responsiveness of Sympathoadrenal System to a Hypoglycemic Challenge in Ketogenic Diet-Fed Rats

The sympathoadrenal counterregulatory response to hypoglycemia is critical for individuals with type 1 diabetes due to impaired ability to produce glucagon. Ketogenic diets (KD) are an increasingly popular diabetes management tool; however, the effects of KD on the sympathoadrenal response are largely unknown. Here, we determined the effects of KD-induced ketosis on the sympathoadrenal response to a single insulin-induced hypoglycemic challenge. We investigated how a 3 week KD feeding regimen affected the main components of the sympathoadrenal counterregulatory response: adrenal sympathetic nerve activity (ASNA), adrenal gland activity, plasma epinephrine, and brainstem glucose-responsive C1 neuronal activation in anesthetized, nondiabetic male Sprague-Dawley rats. Rats on KD had similar blood glucose (BG) levels and elevated ketone body β-hydroxybutyrate (BHB) levels compared to the control Chow diet group. All KD rats responded to hypoglycemia with a robust increase in ASNA, which was initiated at significantly lower BG levels compared to Chow-fed rats. The delay in hypoglycemia-induced ASNA increase was concurrent with rapid disappearance of BHB from cerebral and peripheral circulation. Adrenal gland activity paralleled epinephrine and ASNA response. Overall, KD-induced ketosis was associated with initiation of the sympathoadrenal response at lower blood glucose levels; however, the magnitude of the response was not diminished.


Introduction
Insulin therapy is the gold standard of treatment for type 1 diabetes and patients with advanced type 2 diabetes, but maintaining consistent blood glucose levels is challenging. Hypoglycemia frequently results from inadvertent excessive administration of insulin [1] and, if untreated, can lead to serious neurological consequences, such as mental confusion, seizures, and coma. Severe hypoglycemia is a major obstacle to optimal management of type 1 diabetes with insulin and it is not just debilitating, it can be fatal, accounting for up to 10% of deaths in this population [2].
In individuals without diabetes, hypoglycemia is initially counteracted by increased glucagon secretion and suppression of endogenous insulin release [3]. If blood glucose levels continue to fall (<3.8 mmol/L), a highly coordinated neural counterregulatory reflex is triggered which activates the sympathoadrenal system [4]. This reflex response is initiated by falling extracellular glucose, which activates a subset of glucose-sensing neurons in the hypothalamus and brainstem, termed glucose-inhibited neurons. Activated glucose-inhibited neurons convey signals via sympathetic preganglionic neurons in the spinal cord, through the adrenal sympathetic nerve to adrenal chromaffin cells. Activation of epinephrine-producing chromaffin cells results in rapid release of epinephrine into acutely recorded in anesthetized rats [26]. As ASNA positively correlates with epinephrine release [27], and the actions of epinephrine on target tissues produces physical symptoms of hypoglycemia, in vivo ASNA recordings in anesthetized rats can provide a sensitive, real-time measure of hypoglycemia progression without confounders such as handling stress, multiple blood draws for plasma epinephrine quantification, and difficulties in registering mild physical symptoms in conscious animals. We hypothesize that the CRR is likely to be reduced, and the time course of its initiation may be shifted, as the KD may provide additional substrate and spare glucose for neuronal metabolism in the face of diminishing glucose availability, thereby delaying the central sympathoadrenal response.

Anesthesia
Sodium pentobarbital (Virbac, Milperra, NSW, Australia) has minimal effects on blood glucose levels and Fos expression [28,29] and, therefore, was used for anesthesia in overnight fasted rats (induction: 65 mg/kg ip; maintenance: 13 mg/h iv). The depth of anesthesia was monitored by checking blood pressure responses to tail or paw pinches and was adjusted if changes exceeded 10 mmHg. Arterial PO 2 , PCO 2 , and pH were kept within the physiological range by monitoring end-tidal CO 2 throughout the experiment and adjusting ventilatory rate and volume when necessary. Core body temperature was maintained at 36.5-37.5 • C.

Cannulations and Nerve Recordings
The right carotid artery and jugular vein were cannulated (for the recording of arterial blood pressure and administration of drugs, respectively), and a tracheostomy was per-formed for mechanical ventilation and recording of end-tidal CO 2 . Animals were ventilated with room air and supplemental 100% O 2 . Heart rate was derived from a three-lead ECG. The left adrenal sympathetic nerve was isolated and carefully placed on silver bipolar electrode as previously described [26]. Neurograms were sampled at 2000 Hz, bandpass filtered (10 Hz-3 kHz), and amplified 5000 times (BMA-931 Bioamplifier, CWE, Ardmore, PA, USA). As the adrenal sympathetic nerve can contain a mixture of preganglionic (activated by glucoprivation and synapsing on catecholaminergic chromaffin cells) and postganglionic nerve fibers (innervating the adrenal cortex), upon completion of the 2 h recording period, a ganglion blocker hexamethonium (40 mg/kg, Sigma-Aldrich, St. Louis, MO, USA) was administered intravenously. Experiments (n = 2, KD + Ins group) were excluded from analysis of adrenal sympathetic nerve activity (ASNA) and all related analyses as the ratio of pre-to-postganglionic activity was less than 50% [27]. As such, these recorded nerves were not predominantly preganglionic and did not reflect the specific hypoglycemia-induced changes in activity; therefore, they were deemed suboptimal. ASNA recordings were rectified, smoothed (τ = 2 s), and normalized to baseline (100% activity) by subtracting residual activity (0%) after crushing the nerve with forceps Nutrients 2021, 13, 2627 4 of 16 at the conclusion of experiment. Mean ASNA, expressed as a percentage change from baseline, was analyzed in 3 min intervals taken immediately before insulin administration (0 min) and 15,30,45,60,90, and 120 min post insulin administration.

Insulin Injections
In unpublished pilot studies conducted in our laboratory, we established that a single intravenous injection of 5 U/kg of insulin in anesthetized, overnight fasted rats produced a consistent and highly reproducible level of hypoglycemia in both Chow-and KD-fed groups. Such a consistent degree of hypoglycemia allowed tracking the changes in ASNA with changes in blood glucose levels over time without the need for a hypoglycemichyperinsulinemic clamp. In the current study, hypoglycemia was induced in the Chowfed (n = 9) and KD-fed (n = 7) rats by a single intravenous injection of insulin (human recombinant, 5 U/kg, I2643, Sigma-Aldrich, North Ryde, NSW, Australia). The control Chow (n = 3) and KD-fed (n = 3) rats were injected the same volume of PBS. Peripheral blood glucose and BHB were measured from a drop of blood, obtained via a tail nick, at 0 (pre insulin administration), 15,30,45,60,90, and 120 min. Baseline (0 min, pre insulin administration) blood glucose and BHB were consistently measured between 11:00 a.m. and 12:00 p.m. An Accu-check Performa glucose meter (Roche Diabetes Care, North Ryde, NSW, Australia) and corresponding glucose strips were used for glucose determination, and an Abbott Optium Neo glucose and ketone meter with Freestyle Optium ketone strips (Abbott Diabetes Care, Doncaster, VIC, Australia) were used for BHB measurements.

Measurement of Cerebral Venous BHB Levels
To confirm the presence and utilization of BHB in the brain, cerebral venous blood BHB was sampled in a small separate group of animals (KD + Ins n = 2 and Chow + Ins n = 3) at baseline (0 min) and 15,30,45,60,90, and 120 min post insulin injection. The skull was exposed, and a 4-5 mm diameter hole was drilled above the confluence of sinuses posterior and central to Lambda (Supplementary Figure S1). The confluence of sinuses was opened by puncturing the dura with a needle. Venous blood was sampled at specified times with the ketone strips as described above, by reopening the hole in the confluence of sinuses and letting it bleed for~7 s before occluding it with gauze.

Epinephrine ELISA
At the conclusion of the recording period, blood was collected via the carotid artery catheter. Plasma was separated, aliquoted, and stored at −80 • C. Ultra-sensitive epinephrine ELISA (KA3837, Abnova, Taipei, Taiwan) was performed according to the manufacturer's instructions.

Image Acquisition and Analysis
Adrenal medullae (randomly selected five sections per treatment group) were imaged (Zeiss Axio Imager Z2; 20×) and analyzed in Fiji ImageJ software. The thresholded Fos-positive area, measured in pixels, was normalized to the PNMT-positive area and ex-pressed as a percentage of PNMT-positive area. Fos-positive staining outside of the PNMT-immunoreactive area was excluded. Four 40 µm brainstem sections containing the C1 catecholaminergic nucleus (bregma level −12.48 to −12.12) per animal were used to count the number of TH immunoreactive cells (TH + ), Fos immunoreactive cells (Fos + ), and double-labeled (TH + /Fos + ) cells bilaterally within the region. A rectangular region of interest (ROI, 1680 × 810 µm) was superimposed on the images taken with Zeiss Axio Imager Z2, 10× objective, and cells were counted manually within the ROI by an operator blinded to the experimental conditions.

Statistical Analysis
All statistical analyses were performed using GraphPad Prism software (version 9.0.0). Comparisons between groups were made with unpaired t-test or Mann-Whitney test for non-normally distributed residuals, and one or two-way ANOVA with Holm-Šidák's post-tests for multiple comparisons where appropriate; for data with unequal SDs, Brown-Forsythe ANOVA test was used. Results are reported as means ± SEM unless stated otherwise, with statistical significance set at p < 0.05. The smallest effect size was derived for ASNA, epinephrine, PNMT + /Fos + , and TH + /Fos + measurements from the data generated from the Chow + Ins group which was considered a normal physiological response to insulin injection. To detect a difference of >20%, a minimum of three animals per group was required to achieve the power >80% for all measured variables (ASNA, epinephrine, blood glucose, BHB and Fos). Sample sizes for all treatment groups and measured variables are outlined in Figure 1. Sample sizes for dietary groups, treatments, and measured variables. MAP-mean arterial pressure; HR-heart rate; BHB-β-hydroxybutyrate; ASNA-adrenal sympathetic nerve activity, PNMT -phenylethanolamine N-methyltransferase; TH-tyrosine hydroxylase; Fos-protein product of immediate early gene c-fos.

Figure 2.
Ketogenic diet increases BHB but does not change blood glucose levels. Sprague-Dawley rats were fed ketogenic diet (KD, n = 10, red circles) or control diet (Chow, n = 12, black circles) for 3 weeks. KD-fed rats had comparable (a) blood glucose levels and significantly higher (b) blood BHB levels. (c), KD-fed rats had significantly elevated baseline mean arterial pressure (MAP), but not (d) heart rate. Crosses indicate animals that were excluded from further analysis due to hypoglycemia-induced convulsions (black) or suboptimal nerve recordings (red). Data are means ± SEM, Mann-Whitney test (a) and unpaired t-test (b-d); *** p < 0.001, **** p < 0.0001.

Figure 2.
Ketogenic diet increases BHB but does not change blood glucose levels. Sprague-Dawley rats were fed ketogenic diet (KD, n = 10, red circles) or control diet (Chow, n = 12, black circles) for 3 weeks. KD-fed rats had comparable (a) blood glucose levels and significantly higher (b) blood BHB levels. (c), KD-fed rats had significantly elevated baseline mean arterial pressure (MAP), but not (d) heart rate. Crosses indicate animals that were excluded from further analysis due to hypoglycemia-induced convulsions (black) or suboptimal nerve recordings (red). Data are means ± SEM, Mann-Whitney test (a) and unpaired t-test (b-d); *** p < 0.001, **** p < 0.0001.

Insulin Administration Lowers Blood Glucose and BHB
Insulin (5 U/kg) consistently induced a progressive decrease in blood glucose levels in both Chow and KD groups (Figure 3a). In the KD-fed rats, BHB fell significantly within the first 30 min of insulin injection but remained considerably higher than the Chow-fed rats at every timepoint ( Figure 3b). Within 120 min, BHB decreased from 3.2 ± 0.3 mmol/L to 1.5 ± 0.3 mmol/L, levels similar to the pre-insulin levels in the Chowfed rats. The BHB levels following insulin in Chow-fed rats were physiological for fasting conditions [34]. These data indicate that both glucose and ketones are consumed following administration of insulin. Within 60 min of insulin administration, MAP decreased significantly in both Chow-fed and KD-fed rats and was no longer different between the groups (Supplementary Figure S2).
To assess the presence and utilization of BHB in the brain, a preliminary subset of animals (KD n = 2 and Chow n = 3) was used to measure BHB levels in the cerebral venous blood obtained from the confluence of sinuses (Supplementary Figure S1), concurrently with the peripheral blood BHB measurements. In this subgroup, there was a rapid decrease in cerebral BHB levels in KD-fed animals in the first 30 min (from 3.0 ± 0.2 mmol/L to 1.0 ± 0.2 mmol/L, Figure 4a); this change was larger than the peripheral blood BHB decrease within the same time period (2.3 ± 0.0 mmol/L to 1.2 ± 0.0 mmol/L, Figure 4a). Baseline BHB levels in the Chow-fed rats were substantially lower in both cerebral and peripheral blood (1.6 ± 0.4 mmol/L and 1.4 ± 0.3 mmol/L, respectively) compared to the KD-fed rats, and the hypoglycemia-induced decrease was modest, reaching the lowest of 0.5 mmol/L at 60 min post insulin injection (Figure 4b). In comparison, BHB nadir in the KD-fed animals was 1 mmol/L (Figure 4a).
Nutrients 2021, 13, 2627 7 of 16 to 1.5 ± 0.3 mmol/L, levels similar to the pre-insulin levels in the Chow-fed rats. The levels following insulin in Chow-fed rats were physiological for fasting condition These data indicate that both glucose and ketones are consumed following administ of insulin. Within 60 min of insulin administration, MAP decreased significantly in Chow-fed and KD-fed rats and was no longer different between the groups (Supple tary Figure S2). To assess the presence and utilization of BHB in the brain, a preliminary sub animals (KD n = 2 and Chow n = 3) was used to measure BHB levels in the cerebral v blood obtained from the confluence of sinuses (Supplementary Figure S1), concur with the peripheral blood BHB measurements. In this subgroup, there was a rap crease in cerebral BHB levels in KD-fed animals in the first 30 min (from 3.0 ± 0.2 m to 1.0 ± 0.2 mmol/L, Figure 4a); this change was larger than the peripheral blood decrease within the same time period (2.3 ± 0.0 mmol/L to 1.2 ± 0.0 mmol/L, Figur Baseline BHB levels in the Chow-fed rats were substantially lower in both cerebra peripheral blood (1.6 ± 0.4 mmol/L and 1.4 ± 0.3 mmol/L, respectively) compared KD-fed rats, and the hypoglycemia-induced decrease was modest, reaching the low 0.5 mmol/L at 60 min post insulin injection (Figure 4b). In comparison, BHB nadir KD-fed animals was 1 mmol/L (Figure 4a).

KD-Induced Ketosis Shifts the Initiation of ASNA to Lower Blood Glucose Levels but Preserves the Maximal Response
Insulin-induced hypoglycemia initiated a robust increase in ASNA (Figure 5a-sample traces and Figure 5b-grouped data) in all animals compared to the PBS-treated controls, indicative of a functional sympathetic CRR. Four out of nine insulin treated Chowfed rats were not included in further analysis due to the development of clonic convulsions (presumably neuroglycopenic seizures), terminatable by glucose administration, within 2 h of recording (grouped data shown in Supplementary Table S1). None of the KD-fed rats showed any signs of clonic convulsions. The blood glucose level required to induce a doubling in ASNA (Figure 5c dashed blue line) was much lower (~2.8 mmol/L) in KD-fed than in Chow-fed rats (~3.7 mmol/L; Figure 5c). Following an initial delay, the ASNA response in KD-fed rats exponentially increased reaching a similar (if not higher) maximum to Chow-fed rats (KD 284.2 ± 38% (n = 5); Chow 285.6 ± 34% (n = 5); Figure 5b).

KD-Induced Ketosis Shifts the Initiation of ASNA to Lower Blood Glucose Levels but Preserves the Maximal Response
Insulin-induced hypoglycemia initiated a robust increase in ASNA (Figure 5a-sample traces and Figure 5b-grouped data) in all animals compared to the PBS-treated controls, indicative of a functional sympathetic CRR. Four out of nine insulin treated Chow-fed rats were not included in further analysis due to the development of clonic convulsions (presumably neuroglycopenic seizures), terminatable by glucose administration, within 2 h of recording (grouped data shown in Supplementary Table S1). None of the KD-fed rats showed any signs of clonic convulsions. The blood glucose level required to induce a doubling in ASNA (Figure 5c dashed blue line) was much lower (~2.8 mmol/L) in KD-fed than in Chow-fed rats (~3.7 mmol/L; Figure 5c). Following an initial delay, the ASNA response in KD-fed rats exponentially increased reaching a similar (if not higher) maximum to Chow-fed rats (KD 284.2 ± 38% (n = 5); Chow 285.6 ± 34% (n = 5); Figure 5b).

Insulin-Induced Hypoglycemia Increases the Activation of Medullary C1 Neurons
The activation of tyrosine hydroxylase-expressing (TH) neurons in the ventrolateral medulla (C1 area) is necessary for stimulation of epinephrine secretion from the adrenal gland [35]. The number of TH-immunoreactive neurons in the RVLM did not differ between treatments (Figure 6 a-d), whereas the proportion of activated (TH + /Fos + ) neurons significantly increased in the Chow + Ins group (Figure 6a,e) compared to the corresponding PBS control (Figure 6a vs. Figure 6c,e). Although there was an increase in TH + /Fos + in the KD + Ins group compared to KD + PBS (30.3% ± 6.6% vs. 11.7% ± 1.7%, respectively, Figure 6b,e), the difference was not statistically significant. These results demonstrate the involvement of the brainstem glucoregulatory nuclei in the CRR under both Chow and KD conditions.

Insulin-Induced Hypoglycemia Increases the Activation of Medullary C1 Neurons
The activation of tyrosine hydroxylase-expressing (TH) neurons in the ventrolateral medulla (C1 area) is necessary for stimulation of epinephrine secretion from the adrenal gland [35]. The number of TH-immunoreactive neurons in the RVLM did not differ between treatments (Figure 6a-d), whereas the proportion of activated (TH + /Fos + ) neurons significantly increased in the Chow + Ins group (Figure 6a,e) compared to the corresponding PBS control (Figure 6a vs. Figure 6c,e). Although there was an increase in TH + /Fos + in the KD + Ins group compared to KD + PBS (30.3% ± 6.6% vs. 11.7% ± 1.7%, respectively, Figure 6b,e), the difference was not statistically significant. These results demonstrate the involvement of the brainstem glucoregulatory nuclei in the CRR under both Chow and KD conditions.

Discussion
It was previously reported that acute administration of ketones leads to attenuation of the counterregulatory response to insulin-induced hypoglycemia, but studies on the effects of chronic diet-induced ketosis are lacking. Here, we report that a 3 week KD in (g) Plasma epinephrine levels were significantly higher in Chow + Ins (n = 5, black solid circles) and KD + Ins (n = 5, red solid triangles) rats compared to Chow + PBS (n = 3, black empty circles) and KD + PBS (n = 3, red empty triangles) rats, respectively. Data are means ± SEM; * p < 0.05, ** p < 0.01; Brown-Forsythe ANOVA with Dunnett's test for multiple comparisons. Chow + Ins-chow-fed rats injected with insulin; KD + Ins-ketogenic diet-fed rats injected with insulin; Chow + PBS-chow-fed rats injected with vehicle PBS; KD + PBS-ketogenic diet fed rats injected with vehicle PBS.

Discussion
It was previously reported that acute administration of ketones leads to attenuation of the counterregulatory response to insulin-induced hypoglycemia, but studies on the effects of chronic diet-induced ketosis are lacking. Here, we report that a 3 week KD in rats (1) induces stable ketosis, (2) does not impair the counterregulatory epinephrine release in response to severe insulin-induced hypoglycemia, (3) does not diminish the sympathoadrenal response as assessed by adrenal sympathetic nerve activity and adrenal gland function, and (4) shifts the onset of sympathoadrenal counterregulatory response to lower blood glucose levels. The delay in the sympathoadrenal response is concurrent with rapid cerebral and peripheral ketone consumption and diminished activation of low-glucose-sensitive C1 neurons in the brainstem. Our findings support the notion that ketones may be an effective alternative (or additional) energy source at times of decreased glucose availability; however, in contrast to other studies, nutritional ketosis does not decrease the magnitude of counterregulatory sympathoadrenal response.
Elevated levels of plasma ketone bodies are associated with increased cerebral uptake and utilization of ketones [13,36,37], as well as a parallel decrease in brain glucose metabolic rates [34], indicating a glucose-sparing effect and a metabolic switch toward ketone metabolism, possibly explaining the delayed ASNA response in KD-fed rats in the present study. The degree of cerebral ketone uptake and the glucose-sparing effect of ketones depend on the duration and magnitude of ketosis [34,38]. During prolonged fasting ketones can provide up to 60% of the brain's energy requirements [14]; additionally, with every 1 mmol/L increase in plasma ketones, brain glucose utilization rate decreases by 9-10% [15,34,39]. In our study, a 3 week KD regimen was enough to raise blood BHB levels to 4.1 mmol/L, a level close to the K m for brain endothelial ketone transport by MCT1 (monocarboxylate transporter 1) [39], suggesting that the concentration of BHB was sufficient to exert effects on the brain. In contrast, overnight fasting-induced elevation of BHB to~1.6 mmol/L in Chow-fed rats was insufficient to optimally contribute to neuronal metabolism and probably did not play a significant role in the CRR to hypoglycemia. A sharp reduction in cerebral and peripheral BHB levels in the KD-fed rats was observed within 30 min of insulin administration, which coincided with a slow ASNA increase during this period. Once the fall in BHB plateaued at 30 min, ASNA increased exponentially. Given that insulin rapidly inhibits ketogenesis [40], the decreasing levels of BHB in both peripheral and cerebral blood are indicative of net BHB consumption by extrahepatic tissues, including the brain.
While we did not directly measure brain tissue consumption of BHB, we can speculate on the basis of BHB measurements in the cerebral venous blood that central glucoseinhibited neurons (i.e., neurons that increase their firing rate in the presence of low glucose) were not sufficiently activated until a substantial amount of BHB disappeared from the circulation, which manifested as a delayed increase in ASNA and epinephrine release. One of the important brain nuclei necessary for the CRR and specifically for the sympathetically mediated epinephrine secretion is the C1 area in the rostral ventrolateral medulla [24,41]. In our study, a smaller proportion of activated C1 neurons in KD-fed rats may indicate a delayed response to hypoglycemia due to the time-course of Fos expression, which is reported to peak at 2 h post stimulus in the medulla [42]; however, in the absence of a longer time-course (e.g., 3 h post insulin), lower (but not delayed) activation of C1 neurons cannot be excluded.
In addition to the preserved ASNA, adrenal gland activity was not adversely affected in the KD-fed rats, as evidenced by high level of chromaffin cell activation and plasma epinephrine concentration measured at the end of the experiment. This suggests a good agreement between the function of the adrenal gland and ASNA, which is not always the case. Previous studies in mice [43] and rats [25] revealed that, under conditions of recurrent hypoglycemia, ASNA does not correlate with epinephrine release, suggesting impairment of the CRR at the level of the adrenal gland.
The mechanism of action of ketones in various physiological and pathological states has been studied extensively, mostly by analyzing the effects of acutely administered exogenous ketones [44][45][46][47], which produce similar level of ketosis as the KD. However, Poff et al. cautioned against the extrapolation of findings from acute studies to chronic KD-induced conditions, as the metabolic state resulting from KD consumption, although similar, is not identical to the one resulting from exogenous ketone administration [48]. This indeed appears to be the case in studies investigating the CRR to hypoglycemia; previous human studies showed that acute infusion of exogenous ketones causes a reduction in epinephrine levels [11,49], which contrasts with our findings in KD-fed rats. This discrepancy might be explained by the stimulatory effects of acute ketone administration on cerebral blood flow [50], which is associated with significantly reduced perception of hypoglycemia symptoms and blunted sympathoadrenal response in healthy humans [51]. These findings suggest that attenuated CRR to hypoglycemia in the settings of acute administration of ketones, as opposed to KD-induced sustained ketosis, might be due to the direct effects of ketones on cerebral blood flow, the elevation of which allows a faster delivery of all energy substrates, including glucose, to the brain rather than preferential neuronal metabolism of ketones. In fact, it was shown that chronic diet-induced ketosis does not alter cerebral blood flow in rats but increases capillary density in the brain [52]. In order to maximize cerebral uptake and efficiency of ketone utilization as an energy substrate, a period of adaptation is required, as occurs with KDs [53,54] and extended fasting [55][56][57][58]. Adaptation enables the upregulation of MCTs and ketolytic enzyme expression in the brain [59,60], whereas acute administration of exogenous sources of ketones does not [37,61]. In the present study, 3 weeks of KD was sufficient to induce a stable level of ketosis (>2.5 mmol/L), which was not further augmented by an overnight fast (fed level of ketones 3.2 ± 0.2 mmol/L, unpublished pilot studies) and, therefore, may be considered a maximum nutritional level; these characteristics were previously described and defined as keto-adaptation in KD-fed Sprague-Dawley rats [33]. The observed glucose-sparing effect may provide a metabolic energy buffer for the glucose-sensing neurons in the hypothalamus and brainstem during the initial stages of hypoglycemia, thereby delaying autonomic neurogenic symptoms such as shakiness, hunger, palpitations, and sweating (or evidence of increased ASNA in anesthetized animals); once the buffering capacity of ketones is exceeded, glucoseinhibited neurons are activated and the CRR is rapidly initiated. Further studies are warranted to investigate whether there is sufficient time for the sympathoadrenal response to increase endogenous glucose production before the development of neuroglycopenia under conditions of nutritional ketosis.
The effect of the KD on blood pressure was not a focus of the current study; however, it is worth noting the significantly elevated baseline level of mean arterial pressure in the KD-fed rats. Although the exact mechanisms of the development of this observed increase in blood pressure is unknown, one possibility is the BHB-induced increase in sympathetic nerve activity. Indeed, one previous study did report a stimulatory effect of BHB added to the rat chow for 4 days on sympathetic nerve activity in rats [62]; however, the existing literature in this area is limited, particularly in animals [63]. It is well established that chronically elevated sympathetic nerve activity can lead to hypertension [64]; taken together, we can hypothesize that in the present study the elevated baseline blood pressure may have resulted from sympathoexcitatory effects of KD-induced chronic ketosis. However, without further investigation, other mechanisms cannot be ruled out.

Technical Considerations
This study has technical and interpretive limitations that must be considered. First, the invasive ASNA recordings reported in this study were only achievable in anesthetized animals and, as such, limited us to recording only severe symptoms of hypoglycemia, such as clonic convulsions in some Chow-fed animals. However, the measurements of physical symptoms of hypoglycemia were not our primary objective. One of the advantages of using an anesthetized preparation is that it eliminates handling stress-induced changes in epinephrine, one of the key measures of the CRR in our study. Second, it can be argued the measurements of epinephrine release in response to hypoglycemia are sufficient to assess the function of the sympathoadrenal system as it positively correlates with ASNA [27]; however, it is not always the case. As such, Sivitz et al. demonstrated that repeated exposure to insulin-induced hypoglycemia significantly reduces plasma epinephrine without reducing the level of ASNA [25]. Further studies by Ma et al. confirmed that the dysregulation of sympathoadrenal counterregulatory response originates at the level of adrenal gland, leading to impaired epinephrine release [43]. Therefore, it was prudent in the current study to assess multiple components of the sympathoadrenal response to pinpoint the possible source of impairment. Advantageously, ASNA recordings give a direct real-time measure of the sympathetic output following a hypoglycemic stimulus. This is important because, in our study, the precipitous decline in ASNA after an initial hypoglycemia-induced rise was predictive of the occurrence of clonic convulsions in Chow-fed rats (Table S1), a finding not possible with epinephrine measurement after termination of the experiment. Although not explored in the current study, this observation deserves further examination. Third, this study was focused primarily on the effects of circulating ketones on the CRR; however, the KD used here contained 34% of medium-chain fatty acids (from Copha-hydrogenated coconut oil), which are not only readily metabolized into ketone bodies in the liver but can also cross the blood-brain barrier (unlike long-chain fatty acids) and directly affect brain energy metabolism through various mechanisms [65]. Our experiments were not designed to assess the relative contribution of medium-chain fatty acids on the CRR, but their contribution to the demonstrated response cannot be excluded. Lastly, the study was conducted in healthy, nondiabetic rodents and, although it does provide novel insights into the normal hypoglycemia-induced CRR under the conditions of chronic nutritional ketosis, it would be optimal, albeit challenging, to study this response in animal models of diabetes.

Conclusions
The current study provides a novel insight into the sympathoadrenal CRR to insulininduced hypoglycemia under conditions of nutritional ketosis in healthy, nondiabetic rats. We demonstrated that the sympathoadrenal response in KD-fed rats is as strong as in Chowfed rats; however, it occurs at lower blood glucose levels. Were similar effects to occur in type 1 diabetes, KD might be a useful strategy to mitigate symptoms and consequences of hypoglycemia while providing the brain with an additional metabolic substrate.