Evidence That the Sympathetic Nervous System Elicits Rapid, Coordinated, and Reciprocal Adjustments of Insulin Secretion and Insulin Sensitivity During Cold Exposure

The capacity to adjust insulin secretion to compensate for changes of systemic insulin sensitivity is an essential but poorly understood aspect of glucose homeostasis (1–3). In insulin-resistant conditions such as obesity (3), pregnancy (4), or adolescence (5), for example, insulin secretion must increase to maintain euglycemia, and the reverse applies to conditions of increased insulin sensitivity (1–3,6). Failure to mount an appropriate pancreatic β-cell response to worsening insulin resistance is a hallmark of the progression from normal glucose tolerance to type 2 diabetes (T2D) (7).

The concept that insulin sensitivity and insulin secretion are reciprocally regulated stems from human studies in which plasma glucose and insulin levels measured during a frequently sampled intravenous glucose tolerance test (FSIGT) were analyzed by the minimal model method. This approach generates reliable estimates of key determinants of glucose tolerance: insulin secretion (typically measured as the acute insulin response to glucose [AIR_G]), insulin sensitivity (based on the insulin sensitivity index [S_I]), and glucose effectiveness (S_G), a measure of glucose disposal at basal insulin (independent of an increment in insulin secretion), which can be further separated into the basal insulin effect (BIE) and glucose effectiveness at zero insulin (GEZI). The relationship between S_I and AIR_G is hyperbolic in humans, such that the product of the two, referred to as the “disposition index” or DI, a measure of the net contribution made by insulin on glucose levels during the FSIGT, tends to remain constant irrespective of the prevailing level of insulin sensitivity (3), which provides for stable glucose homeostasis in the face of diverse internal and external challenges to tissue glucose utilization (6).

The coupling of insulin secretion to insulin sensitivity cannot be explained by changes of plasma glucose levels (8). Although rising nocturnal plasma levels of nonesterified fatty acids (NEFA) were identified as a potential mechanism linking increased insulin secretion to obesity-induced insulin resistance (9), the extent to which this mechanism contributes to the hyperbolic relationship between insulin secretion and insulin sensitivity remains unknown. To investigate the role of the brain in this coupling mechanism, we sought to interrogate its contribution to the adaptive metabolic response to a physiological challenge. To this end, we used cold exposure as a provocative intervention, based on the following considerations: 1) maintenance of core body temperature in cold environments depends on temperature-sensitive hypothalamic neurocircuits that increase sympathetic nervous system (SNS) outflow to thermogenic tissues (10,11); 2) cold exposure increases SNS outflow to the pancreas (12), which likely explains the associated reduction of insulin secretion (via a mechanism involving activation of α-adrenergic receptors on pancreatic β-cells [13–15]); and 3) glucose tolerance does not change or is improved despite the effect of cold exposure to decrease insulin secretion (16,17). Although the latter observation can be accounted for by the known effect of cold exposure to increase insulin sensitivity in thermogenic tissues (17–19), the goal of the current work was to determine whether the brain serves to ensure that insulin secretion declines in a manner that preserves glucose homeostasis and averts hypoglycemia.

To this end, we measured the effect of cold exposure on determinants of glucose tolerance in rats using the FSIGT/minimal model approach in the presence and absence of an intravenous (i.v.) infusion of the α-adrenergic receptor (α-AR) blocker phentolamine. Because the proper interpretation of our data hinges on whether the relationship between insulin secretion and insulin sensitivity in rats comports with the hyperbolic relationship characteristic of humans, we also sought to create a normative rat database with which to interrogate this relationship. This goal was achieved by analyzing pooled data from FSIGT studies conducted in either of two strains of rat (Wistar or Long-Evans) fed a standard chow diet or a high-fat diet (HFD), some of which were included as normal control data in a previous publication (20).

We report that, as expected, the relationship between insulin sensitivity and insulin secretion in rats conforms to the hyperbolic function characteristic of humans. We further demonstrate that the potent effect of cold exposure to the increase of insulin sensitivity is perfectly offset by a proportionate decline of glucose-induced insulin secretion such that glucose tolerance remains unchanged. Moreover, the effects of cold exposure on insulin secretion and insulin sensitivity are rapidly reversed by systemic α-AR blockade. Together, these findings offer direct evidence of a key role for the brain in the highly coordinated, potent, and reciprocal adjustments of insulin sensitivity and insulin secretion that maintain stable glucose homeostasis during cold exposure.

All procedures were performed in accordance with National Institutes of Health Guidelines for the Care and Use of Animals and were approved by the University of Washington Animal Care Committee.

Adult male Wistar and Long Evans rats (Harlan Laboratories, Indianapolis, IN) were individually housed under specific pathogen-free conditions in a temperature-controlled room with a 12:12 h light/dark cycle and provided ad libitum access to water and standard laboratory chow (PMI Nutrition, St. Louis, MO) or an HFD containing 60% kcal fat (D12492; Research Diets, Inc., New Brunswick, NJ), unless otherwise stated.

Rats underwent implantation of catheters into the carotid artery and jugular vein and received buprenorphine hydrochloride (Reckitt Colman Pharmaceuticals, Richmond, VA) at the completion of the surgery as previously described (20–22).

After an overnight fast, multiple blood samples were obtained from unrestrained, conscious animals via an arterial catheter as previously described (20–23). Plasma insulin and blood glucose levels generated from the FSIGTs were analyzed using MinMod software to quantify S_G and S_I as previously described (20,23,24). AIR_G, the incremental area under the insulin curve during the FSIGT (AUC_insulin), the BIE, GEZI, and the glucose disappearance rate constant (K_G) were calculated as previously described (20,23,24). A detailed description of the FSIGT and minimal model analysis is provided in the Supplementary Data.

To independently assess the effect of phentolamine on insulin sensitivity in cold-exposed rats, a euglycemic-hyperinsulinemic clamp was performed as previously described (21,22). A detailed description of the clamp procedure is provided in the Supplementary Data.

Insulin clearance was estimated by first calculating the insulin secretion rate using deconvolution of the measured C-peptide concentrations, followed by calculation of the rate of insulin clearance as AUC(insulin secretion)/AUC(plasma insulin). The parameters for rat C-peptide kinetics were obtained from a kinetic study using 5 nmol/kg injections of rat C-peptide 2 in Sprague-Dawley rats. The C-peptide concentration data after the injection were well described by a one-compartment model with clearance = 29.7 mL/min/kg and volume = 656 mL/kg (25).

To create a normative database with which to analyze relationships between the determinants of glucose tolerance in rats, we assembled data from studies previously conducted in adult male rats in our laboratory that used the FSIGT and minimal model analysis protocol described above that included two different rat strains (Wistar [n = 54] or Long-Evans [n = 25]), provided ad libitum access to chow (n = 62) or an HFD (n = 17) for 5 days or 3 months before study, vehicle-treated animals from a previous study (20), and animals from the current studies.

A key goal was to determine whether the relationship between insulin sensitivity and insulin secretion conforms to a rectangular hyperbola in rats as it does in humans (2,3,26,27). To test this hypothesis, we used reduced major axis regression (RMA) with bootstrapped estimates of the SEs (28) (http://www.exetersoftware.com/cat/biomstat/book_biometry.html) to fit linear models for the relationship between a measure of insulin secretion (AIR_G) or an integrated measure of the circulating insulin level during the FSIGT (AUC_insulin), which reflects insulin secretion and its clearance from plasma, and the inverse of insulin sensitivity (i.e., as a function of [1/S_I]). For completeness, we also modeled insulin secretion equaling a regression slope times (1/sensitivity) without an intercept term using ordinary least squares regression (29), such that the product of insulin secretion and insulin sensitivity equals a constant. Comparison of the RMA versus intercept-excluded hyperbolic fits enabled a comprehensive qualitative and quantitative assessment of the extent to which the “hyperbolic law” applies to rats.

Animals matched by body weight were housed at room temperature (22°C, n = 13) or in a cold environment (5°C, n = 13) for 28 h without access to food before a FSIGT conducted at these same temperatures. Experiments were performed using a crossover design, such that each animal was studied at each temperature separated by at least 10 days. To assess the durability of cold-induced changes in determinants of glucose tolerance, separate cohorts of rats (n = 9/group) were housed for 24 h at room temperature or at 5°C, and then housed at room temperature for an additional 4 h before and during an FSIGT.

To determine whether α-AR signaling is required for the effects of cold exposure on AIR_G and S_I, rats with chronic indwelling catheters were exposed to the cold (5°C) for 28 h without access to food. At 30 min before the FSIGT, a continuous i.v. infusion of the nonselective α-AR blocker phentolamine at a dose of 8 μg/kg/min (identified in pilot studies as a dose that has minimal effects on mean arterial pressure and heart rate) (Supplementary Fig. 1) or vehicle (n = 9/group) was commenced, with infusion continuing throughout the FSIGT.

After implantation of catheters, adult male Wistar rats were exposed to the cold (5°C) for 28 h without access to food, and a euglycemic-hyperinsulinemic clamp was performed (21,22). Although tracer dilution analysis using tritiated glucose was included in the study design, the data are not included owing to technical problems that confound data interpretation. Animals received a continuous i.v. infusion beginning at t = −30 min of phentolamine (8 μg/kg/min) or its vehicle. A primed continuous infusion of regular human insulin (16 mU/kg bolus, followed by 2 mU/kg/min HumulinR; Eli Lilly, Indianapolis, IN) was initiated at t = 0 min. A 50% dextrose solution was infused as required to maintain euglycemia, and serial blood samples were obtained for determination of plasma glucose and insulin levels.

Whole blood samples for plasma hormonal measures were collected in appropriately treated tubes (22) and centrifuged, and the plasma was removed and stored at −80°C for subsequent assay. Plasma glucose was measured using a GM9D glucose direct analyzer (Analox Instruments, Stourbridge, U.K.), plasma insulin (Crystal Chem, Downers Grove, IL), glucagon (Mercodia, Uppsala, Sweden), and corticosterone (Alpco, Salem, NH) by ELISA and NEFAs by colorimetry (Wako Chemicals, Richmond, VA).

Results are expressed as mean ± SEM. Significance was established at P < 0.05 (two-tailed). Statistical analyses of effects of cold exposure or phentolamine on metabolic parameters were performed using Statistica 7.1 software (StatSoft, Inc., Tulsa, OK). A one-way ANOVA with a least significant difference post hoc test was used to compare mean values among multiple groups, a two-sample unpaired Student t test was used for two-group comparisons, and a paired Student t test was used for within-group comparisons. RMA regression was performed via constrained nonlinear regression in SPSS 23 software (IBM Corp., Armonk, NY) using sequential quadratic programming and the loss function specified in the IBM technical note available at http://www-01.ibm.com/support/docview.wss?uid=swg21476031. The general linear model in SPSS was used to model the relationships among measures of glucose tolerance and measures of insulin- and noninsulin-mediated determinants of glucose homeostasis.

Analysis of data compiled from a cohort of 79 normal rats revealed the predicted hyperbolic relationship between insulin sensitivity (S_I) and the insulin response to glucose, whether the latter was measured as AIR_G or as AUC_insulin (Fig. 1A and B). Although RMA regression identified asymptotic values (y-intercepts) with bootstrap estimated 95% CIs that did not include zero (383, 689 for AIR_G; 889, 2673 for AUC_insulin), indicating that the intercepts are statistically significant, inspection of Fig. 1 suggests that the relationship between S_I and either measure of the insulin response is fundamentally hyperbolic in nature. We suspect that because AUC_insulin reflects the exposure of body tissues to insulin during the FSIGT more effectively than does AIR_G, the relationship between S_I and AUC_insulin conforms to a hyperbola more closely than does that between S_I and AIR_G (Fig. 1A and B). Moreover, the insulin sensitivity of rats fed the HFD was reduced relative to that of chow-fed controls (mean S_I: 0.80 ± 0.39 vs. 3.80 ± 2.84 [0⁻⁴ × min⁻¹]/µU/mL; P < 0.0001), whereas insulin secretion was increased (mean AIR_G: 2,157 ± 700 vs. 1,003 ± 480 µU/mL × min; P < 0.0001) (Fig. 1C).

To assess the relative contributions of insulin-dependent and insulin-independent mechanisms to glucose disposal during the FSIGT, we performed multiple regression analyses in which the dependent variables were either of two measures of glucose tolerance (K_G or AUC_glucose) during the FSIGT. These models included as predictor variables the interaction of the insulin response (measured as AIR_G or AUC_insulin) and insulin sensitivity, along with S_G (Table 1 and Fig. 1D). As indicated in Table 1, S_G makes a substantial contribution to glucose tolerance whether the latter is measured as K_G (models 1 and 2) or AUC_glucose (models 3 and 4), as in humans (30). Relatedly, the contribution made by overall insulin action (represented by the DI or by the product of S_I × AUC_insulin) to glucose tolerance appears to be larger when AUC_glucose is used instead of K_G.

To analyze the relationship between insulin action and glucose tolerance among rats with relatively high or low values of S_G, regression lines were fit by ANCOVA for each cohort (Fig. 1D). Although the slopes of the lines fit to each of the two groups did not differ significantly (P = 0.54), for any given value of insulin action, the rats in the upper S_G split had significantly better (P < 0.0001) mean overall glucose tolerance than did rats in the lower S_G split. These observations suggest that the effect of insulin action on glucose tolerance in rats is influenced by variation in S_G, a conclusion consistent with data obtained from three other models (data not shown). Similar outcomes were obtained using the models that replaced S_G with GEZI (data not shown), consistent with human data (30). Indeed, GEZI (but not BIE) appears to be a substantial determinant of glucose tolerance in rats, given that the Pearson correlation (r) between K_G and GEZI was 0.56 (P < 0.0001).

Overall, our results indicate that either AIR_G or AUC_insulin is a useful index of the insulin response to glucose and that together they support the conclusion that variation in insulin sensitivity across animals is effectively compensated by a proportionate change of insulin secretion.

To investigate mechanisms underlying the coordinate regulation of insulin sensitivity and insulin secretion, we sought to validate previous evidence that cold exposure inhibits insulin secretion in rats (16,19). As expected (16,26), fasting levels of blood glucose (112.6 ± 3.5 vs. 97.5 ± 2.3 mg/dL; P < 0.05) and plasma insulin (22.3 ± 2.9 vs. 10.0 ± 1.1 μU/mL; P < 0.05) were lower in fasted rats housed in a cold environment (5°C for 28 h) than at room temperature (22°C) (Fig. 2A and B). Moreover, in rats that underwent a FSIGT, followed by minimal model analysis (20,23,24,31), as expected, AIR_G was potently inhibited by cold exposure by ∼50% (AIR_G: 667 ± 41 for room temperature vs. 372 ± 33 μU/mL × min for cold exposure; P < 0.001), despite glucose tolerance remaining virtually unchanged (AUC_glucose: 5,869 ± 233 for room temperature vs. 5,848 ± 442 for cold exposure; P = NS) (Fig. 2A and B) (16). To explain this outcome, we hypothesized that a compensatory increase of insulin sensitivity must have occurred in the cold-exposed group. As predicted, we found that the potent effect of cold exposure to inhibit insulin secretion was offset by a near doubling of insulin sensitivity (S_I) (Fig. 2C and D). Consequently, neither the DI nor glucose tolerance was substantially altered (Fig. 2E). In addition, cold exposure had no significant effect on S_G, the BIE, or GEZI (Fig. 2F–H).

To assess the durability of the potent and highly coordinated (and offsetting) changes of insulin secretion and insulin sensitivity elicited by cold exposure, we performed an FSIGT on separate cohorts of fasted rats that were maintained at room temperature or were housed in the same cold environment (5°C) for 24 h and then returned to room temperature (22°C) for 4 h before study. Although fasting blood glucose and plasma insulin levels remained slightly lower in cold-exposed animals that were then returned to room temperature (Fig. 3A and B), the effect of cold exposure to inhibit glucose-induced insulin secretion (AIR_G) (Fig. 3B) and/or increase S_I (Fig. 2D) was no longer evident 4 h after returning to room temperature (Fig. 3D). As before, neither DI nor S_G (Fig. 3E and F), GEZI, or BIE (data not shown) differed between groups (Fig. 3E). Thus, the potent and highly coordinated effects of cold exposure on insulin secretion and insulin sensitivity are reversed within 4 h after the return to room temperature.

To determine whether a change in SNS output contributes to the rapidly reversible effect of cold exposure on insulin secretion and insulin sensitivity, fasted rats housed at 5°C for 28 h received a continuous i.v. infusion of the α-AR blocker phentolamine or its vehicle for 30 min before and then during an FSIGT. As predicted, the effect of cold exposure to suppress basal- and glucose-induced insulin secretion was reversed by i.v. phentolamine, such that the low plasma insulin levels typical of animals housed in the cold were restored to values observed in animals housed at room temperature (13.9 ± 2.4 vs. 28.4 ± 1.9 μU/mL; P < 0.05). Despite this marked increase of insulin secretion, glucose tolerance was once again largely unaffected (AUC_glucose: 5,151 ± 351 for vehicle vs. 5,078 ± 378 for phentolamine; P = NS) (Fig. 4B and C). In addition, the potent effect of cold exposure to increase S_I was also rapidly reversed by phentolamine-induced α-AR blockade, such that the DI did not change (Fig. 4D and E). Interestingly, S_G was also increased in phentolamine-treated animals (Fig. 4F), an effect due largely to an increase of GEZI with no change in BIE (data not shown). Collectively, these data show that the potent, highly coordinated, and reciprocal adjustments of insulin secretion and insulin sensitivity induced by cold exposure are both reversible within 30 min and dependent on intact α-AR signaling.

To confirm our finding that α-AR blockade reverses the effect of cold exposure to increase insulin sensitivity, we measured the effect of i.v. phentolamine infusion on the glucose infusion rate (GIR), a direct measure of whole-body insulin sensitivity, in cold-exposed rats during a euglycemic-hyperinsulinemic clamp. By design, arterial blood glucose levels were similar between groups during the clamp period (Fig. 5A), although plasma insulin levels during the clamp were significantly increased in cold-exposed animals receiving phentolamine (Fig. 5B), raising the possibility that phentolamine infusion affected insulin clearance as well as insulin secretion. Yet despite higher plasma insulin levels, the GIR required to maintain euglycemia was decreased by ∼25% (P < 0.05) in cold-exposed rats that received i.v. phentolamine (Fig. 5C and D). This finding offers direct, independent confirmation of the effect of systemic α-AR blockade to reduce insulin sensitivity in cold-exposed rats.

To determine whether either cold exposure or phentolamine administration affected insulin clearance from plasma as well as insulin secretion, we measured C-peptide levels in plasma samples obtained during the FSIGT studies reported above and subjected these data to a model-based deconvolution analysis (25). We found that changes in the plasma level of C-peptide closely paralleled those of insulin from both studies (Fig. 6A and B). Further analysis also revealed no difference in rates of insulin clearance between animals housed at room temperature versus the cold (0.261 vs. 0.279; P = 0.59) or between cold exposed animals treated with i.v. vehicle versus phentolamine (0.184 vs. 0.199; P = 0.31). Both the effect of cold exposure to reduce the plasma insulin response to glucose and the effect of phentolamine to reverse this effect, therefore, resulted from changes of insulin secretion rather than insulin clearance.

An increase in nocturnal NEFA levels has been reported to contribute to the compensatory increase of insulin secretion in a dog model of insulin resistance induced by HFD feeding (8,9). However, we found no difference of fasting NEFA levels between rats housed at room temperature and cold-exposed animals. Further, NEFA levels tended to increase in cold-exposed animals during the FSIGT, whereas insulin secretion was reduced (Fig. 6C). Combined with the observation that plasma NEFA levels dropped comparably in the first 10 min after administration of glucose in cold-exposed animals, irrespective of whether they received i.v. saline or phentolamine (P = NS) (Fig. 6D), our data suggest that the effect of phentolamine to acutely increase insulin secretion in cold-exposed rats was not a consequence of increased plasma NEFA levels.

In addition, we found that plasma glucagon levels were elevated in cold-exposed animals relative to those housed at room temperature (Fig. 6E) and were similarly suppressed after an i.v. glucose bolus during the FSIGT, irrespective of the temperature at which they were housed (Fig. 6F). Similarly, plasma corticosterone levels were elevated in cold-exposed rats, and phentolamine treatment, if anything, lowered plasma corticosterone levels (Fig. 6G and H), despite reversing the effect of cold to improve insulin sensitivity (Figs. 4 and 5). Taken together, these data suggest that neither changes in circulating NEFA levels nor the neuroendocrine response to cold is likely to explain the robust and highly coordinated changes of insulin secretion and insulin sensitivity.

That cold exposure elicits increased whole-body glucose utilization (to support the increased demand for heat production) (19) and decreased insulin secretion (16,32) with little or no change of glucose tolerance points to a highly coordinated metabolic response that enables thermogenic needs to be met while preserving glucose homeostasis. Our data suggest that these adaptive changes of insulin secretion and sensitivity to cold exposure occur quickly via a mechanism involving the SNS. Specifically, we found that the effect of cold exposure to inhibit the insulin response to glucose (whether measured as AIR_G or AUC_insulin) in rats was precisely offset by a proportionate increase of insulin sensitivity (measured as S_I), such that glucose tolerance was unchanged. Moreover, this reduced insulin response to glucose was due entirely to reduced insulin secretion with no change of plasma insulin clearance, was fully established within 28 h of exposure to a cold environment, and returned to baseline within just 4 h after the return to room temperature. Our finding that the effect of cold exposure on these parameters was fully reversed by α-AR blockade implicates the SNS in the coordinate regulation of insulin secretion and sensitivity in this setting. That this reversal was achieved within just 30 min of the onset of phentolamine infusion attests to the remarkable rapidity with which this coordinate regulation takes place.

The proper interpretation of these findings partly hinges on the extent to which the well-documented hyperbolic relationship between insulin secretion and insulin sensitivity in humans (3) applies to rats as well. To address this question, we analyzed the relationship between S_I and two different measures of the insulin response to a glucose challenge (AIR_G and AUC_insulin) across a large cohort of normal rats housed at room temperature (22°C). To ensure a sufficiently broad distribution of these parameters, we included data from two different rat strains (Wistar and Long-Evans) fed either of two diets (standard chow or a 60% kcal HFD). As predicted, the relationship between S_I and each of the two measures of insulin response was largely hyperbolic. Although we also found that S_G makes a major contribution to glucose tolerance, which is consistent with prior evidence (33) and may help explain why the RMA regression hyperbolic fits included significant nonzero asymptotic values (intercept terms), our data establish that the relationship between insulin sensitivity and insulin secretion in rats is quite similar to that in humans. By recapitulating the relationship observed in humans, our data also support the validity of the FSIGT/minimal model method for metabolic studies in rats. Extending this validation is our findings that 1) consuming an HFD reduces insulin sensitivity in otherwise normal rats and that this effect is offset by a proportionate increase of insulin secretion (Fig. 1C), and 2) in cold-exposed rats, the effect of i.v. phentolamine to rapidly reduce insulin sensitivity when measured as S_I was replicated using the euglycemic-hyperinsulinemic clamp technique.

By investigating the role played by the brain (acting via the SNS) in the adaptive response to a physiological challenge, our approach is a departure from more conventional strategies that rely on administration of peptides or drugs into the brain to investigate its role in glucose homeostasis. Indeed, metabolic studies are usually conducted under constant environmental conditions to avoid the potentially confounding effect of commonly encountered environmental challenges. This strategy reduces variability in measured end points but also precludes investigation into how animals adapt to such challenges. Should the brain’s role in glucose homeostasis take on greater importance when animals are confronted with a real-world environmental challenge, therefore, housing them in an unchanging (and unchallenging) environment by definition places major limits on the insights that can be gained. In this context, we note that the current study paradigm involved exposure to temperatures comparable to what rodents throughout much of the world experience daily.

To understand how cold exposure effects these highly coordinated changes of insulin secretion and insulin sensitivity, we consider three possibilities: 1) adaptive changes of insulin sensitivity are secondary to the change of insulin secretion, 2) adaptive changes of insulin secretion are secondary to the change of insulin sensitivity, or 3) adaptive changes of both insulin sensitivity and insulin secretion are part of an integrative regulatory process involving the SNS. With respect to the first possibility, we note previous evidence demonstrating that insulin sensitivity does not increase when insulin secretion is reduced, whether by a pancreatic β-cell toxin (e.g., streptozotocin) (34), inhibitory hormones (e.g., somatostatin) (35), partial pancreatectomy (36), or disease (e.g., T2D) (37). The alternative possibility that reduced insulin secretion is a consequence of a cold-induced increase of insulin sensitivity is consistent with the known effect of cold exposure to increase insulin-mediated glucose uptake into thermogenic tissues, including skeletal muscle, heart, and white and brown adipose tissue (17–19). One potential mechanism to explain how this effect is coupled to a proportionate decrease of insulin secretion is through reduced nutrient stimulation of β-cells via a mechanism involving, for example, reduced circulating glucose or NEFA levels (8–10). Such a mechanism is inconsistent with both the modest effect of cold exposure on plasma levels of these nutrients and the lack of any effect of phentolamine on these plasma values (despite its ability to fully reverse cold-induced inhibition of insulin secretion). Similarly, although cold exposure increased circulating levels of glucagon and corticosterone via a mechanism that was partially reversed by phentolamine, such changes do not offer a viable explanation for the associated changes of insulin secretion and sensitivity.

Could cold-induced changes of insulin secretion and insulin sensitivity be mediated via the SNS? That the effect of cold exposure to increase insulin sensitivity was reversed by α-AR blockade offers direct evidence that increased SNS outflow is required for this effect, consistent with an extensive literature implicating the brain in the control of insulin sensitivity (38) via mechanisms involving the SNS (39). The hypothesis that the cold-induced decline of insulin secretion was also mediated via the SNS is supported by pioneering work documenting autonomic control of pancreatic islet function some 40 years ago (15) and by evidence that cold exposure increases sympathetic tone to the pancreas (12). The mechanism underlying this effect was subsequently shown to involve α-AR signaling in pancreatic islets (13–15), a possibility reinforced by our finding that the effect of cold exposure to reduce insulin secretion was rapidly reversed (within 30 min) by phentolamine administration.

These observations collectively support a model in which SNS outflow to both pancreatic islets and thermogenic tissues underlies the highly coordinated and proportionate changes of insulin sensitivity and insulin secretion that effectively redirect substrate to thermogenic tissues without altering glucose tolerance (Fig. 7). Future studies are warranted to clarify the cellular basis for these effects and to further delineate the brain’s role in these metabolic responses. For example, the specific role played by islet sympathetic nerves could be investigated by sympathetic denervation of the pancreas or by targeted deletion of α-AR in pancreatic β-cells. An additional priority is to determine whether reciprocal adjustment of insulin secretion and insulin sensitivity is triggered by activation of hypothalamic cold-sensitive neurons involved in thermoregulation.

In conclusion, we report that the hyperbolic relationship between insulin secretion and insulin sensitivity characteristic of humans is operational in rats as well. We demonstrate further that in rats, the rapid, potent, and highly coordinated changes of insulin sensitivity and insulin secretion elicited by cold exposure are critically dependent on the SNS. We speculate that these metabolic adjustments are components of a larger and highly integrated set of responses elicited by the brain to enable effective thermogenesis while preserving metabolic homeostasis and thereby promote survival during cold exposure.

Acknowledgments. The authors greatly acknowledge valuable guidance provided by Dr. David Wasserman, Vanderbilt University, and technical assistance provided by Amy Martinson, from the laboratory of Charles E. Murry, University of Washington, for measures of heart rate and mean arterial pressure.

Funding. This work was supported by National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases grants DK-089056 (G.J.M.), DK-27619 and DK-29867 (R.N.B.), DK-50154 (G.J.T.), DK-083042, DK-090320, and DK-101997 (M.W.S.); the National Institute of Diabetes and Digestive Kidney Diseases–funded Nutrition Obesity Research Center (DK-035816) and Diabetes Research Center (DK-017047); the Nutrition, Obesity and Atherosclerosis Training Grant from the National Heart, Lung, and Blood Institute (T32-HL-007028) and the Diabetes and Metabolism Training Grant (T32-DK-0007247) at the University of Washington; and the Department of Veterans Affairs grant BX001060 (S.E.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.J.M., K.M., K.J.K., F.P., D.S., R.N.B., G.J.T., S.E.K., and M.W.S. analyzed data and edited the manuscript. G.J.M., K.M., J.M.R., J.M.S., M.E.M., J.T.N., and N.K.A. conducted experiments and acquired data. G.J.M., K.M., and M.W.S. designed research studies and wrote the manuscript. M.W.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.