Metabolic Sensors Mediate Hypoglycemic Detection at the Portal Vein

Male Wistar rats (241 ± 7 g, n = 36) in the conscious relaxed state were used in all experimental procedures. All animals were housed in individual cages, fed ad libitum, and subjected to standard 12-h light-dark cycle. The University of Southern California Institutional Animal Care and Use Committee approved all surgical and experimental procedures. One week before the study, rats were anesthetized (3:3:1 ketamine HCl:xylazine:acepromizine malate; 0.1 ml/0.1 kg body wt i.m.), and indwelling catheters were inserted into the portal vein (silastic tubing, 0.051 cm internal diameter [ID] with 0.03-cm-ID tip) for infusion of studied metabolites, into the carotid artery (polyethylene tubing, 0.058 cm ID) for arterial blood sampling, and into the jugular vein (dual silastic cannula, 0.03 cm ID) for peripheral insulin, glucose, and metabolite infusions. All cannulas were tunneled subcutaneously, exteriorized on the back of the neck, and incased in an infusion harness (Instech Laboratories, Plymouth Meeting, PA). All animals were allowed at least 6 days to recover from surgery and to regain their original body weight. All animals were fasted overnight before initiation of the experiment.

All rats were exposed to a general protocol for the induction of hypoglycemia. On the day of the experiment, animals were placed in a metabolic chamber and allowed to acclimatize to the new environment for at least 60 min (−90 to −30 min). At min 0, whole-body hypoglycemia was induced via jugular vein insulin infusion (25 mU · kg⁻¹ · min⁻¹) and variable jugular glucose infusions (20% dextrose). Basal glucose concentrations were measured at −30 and 0 min, and subsequent serial sampling for glucose was performed at 10-min intervals throughout the duration of the clamp (0–105 min). Plasma concentrations of epinephrine, norepinephrine, and infused metabolites (lactate, pyruvate, or BHB) were determined at basal (−30 and 0 min) and during sustained hypoglycemia (60, 75, 90, and 105 min). Plasma samples for the insulin assay were drawn during the basal period (−30 min) and at the end of hypoglycemia (105 min).

Experiments were subdivided according to the infused metabolite, i.e., portal lactate clamp (n = 13), portal pyruvate clamp (n = 11), and portal BHB clamp (n = 12). Animals were randomly selected to take part in one of six separate protocols distinguished by the site of infusion, portal vein versus jugular vein, and metabolite infused, lactate, pyruvate, and BHB.

For portal-lactate animals (n = 7), whole-body hypoglycemia was induced as described above with a lactate (2.5 mol/l lactic acid in phosphate buffer, pH 4.5; Sigma, St. Louis, MO) infused via the portal vein at variable rates starting at 40 min to attain local portal vein hyperlactatemia (∼5 mmol/l) in the midst of systemic hypoglycemia. This lactate infusate has previously been shown to elevate lactate levels without impacting blood pH (25). Jugular-lactate control animals (n = 6) were exposed to an identical hyperinsulinemic-hypoglycemic clamp with lactate infused (variable rate) via the jugular vein beginning at 40 min to match arterial lactate concentrations in the portal lactate group. To isolate the increase in lactate levels to the portal vein, dichloroacetic acid (DCA; 300 mg/kg) was infused in both groups via the jugular vein to prevent any elevation in arterial lactate concentration above basal. Dichloroacetate selectively inhibits tissue lactate release via activation of pyruvate dehydrogenase and has been used in a number of animal and human studies without any adverse affects (26,27).

Portal-pyruvate (n = 6) and jugular-pyruvate (n = 5) rats were exposed to a protocol identical to that above for the lactate clamp with the exception that animals received pyruvate infusions (1 mol/l sodium pyruvate in sterile 0.9 mol/l NaCl, pH 7.4; Sigma), without DCA.

Portal-BHB (n = 5) and jugular-BHB (n = 7) animals were exposed to a protocol identical to that above for the lactate clamp with the exception that animals received BHB infusions (1 mol/l sodium BHB in sterile 0.9 mol/l NaCl, pH 7.4; Sigma), without DCA.

Glucose and lactate were assayed by the glucose oxidase method (YSI, Yellow Springs, OH). Arterial BHB (28) and pyruvate (29) concentrations were assayed spectrophotometrically. Catecholamines were analyzed using a single-isotope radioenzymatic method (30). Insulin samples were assayed using a radioimmunoassay kit (Linco Research, St. Charles, MO).

The estimated local portal vein metabolite concentrations (M_PV) were calculated as M_PV = M_A+ (MINF_POR/PVPF), where M_A is arterial metabolite concentration (in micromoles per milliliter), MINF_POR is the portal vein metabolite infusion rate (in micromoles per minute), and PVPF is portal vein plasma flow rate in (milliliters per minute). The portal vein plasma flow rate was assumed to be 80% of hepatic plasma flow rate, where hepatic plasma flow rate was estimated to be 1.3 ml · g liver wt⁻¹ · min⁻¹ (31). Mean steady-state metabolite concentrations were determined as the average of at least four serial samples that differ nondirectionally by <0.5% per minute. Experimental results were expressed as means ± SE. A one-way ANOVA was used to compare characteristics between independent groups. Repeated measures ANOVAs were used to determine group differences over time with Tukey’s test for post hoc analysis. Significance was set at P < 0.05.

Arterial plasma insulin increased from an average basal value of 98 ± 10.41 pmol/l (P = 0.76, between groups) to a hyperinsulinemic value of 5,246.34 ± 310.59 pmol/l (P = 0.52, between groups). Basal glucose concentrations for portal-lactate versus jugular-lactate (6.76 ± 0.18 vs. 7.10 ± 0.16 mmol/l, P = 0.12) and glucose concentrations at the hypoglycemic nadir (2.54 ± 0.14 vs. 2.51 ± 0.14 mmol/l, P = 0.81) were not significantly different between the groups (Fig. 1A). Furthermore, no significant differences were found between portal-lactate versus jugular-lactate protocols with respect to basal arterial lactate (0.88 ± 0.08 vs. 0.9 ± 0.09 mmol/l; P = 0.81; Fig. 1B) and arterial lactate during sustained (60–105 min) hypoglycemia (1.01 ± 0.1 vs. 0.91 ± 0.09 mmol/l; P = 0.36; Fig. 1B). As expected, portal vein lactate infusions significantly elevated mean portal vein plasma lactate concentrations in the portal-lactate group, whereas portal vein lactate concentrations in the jugular-lactate group remained at basal levels (5.03 ± 0.20 vs. 0.91 ± 0.09 mmol/l, P < 0.0001; Fig. 1C)

In response to insulin-induced hypoglycemia, jugular-lactate animals demonstrated a robust sympathoadrenal response where plasma epinephrine concentrations increased from 1.68 ± 0.51 nmol/l at basal to 14.44 ± 2.70 nmol/l by 105 min (Fig. 2A). In contrast, selective elevation of portal vein lactate resulted in a 64% suppression of the epinephrine response (14.44 ± 2.70 vs. 5.26 ± 0.77 nmol/l, P < 0.01; Fig. 2A) at 105 min of hypoglycemia. Additionally, statistically significant suppressions in the epinephrine response to hypoglycemia were observed in the portal-lactate group at min 60, 75, and 90 of hypoglycemia. Correspondingly, the norepinephrine response above basal was also suppressed by 67% in the portal-lactate versus the jugular-lactate group (1.66 ± 0.57 vs. 5.05 ± 1.21 nmol/l, P < 0.01; Fig. 2B).

Mean arterial glucose concentrations were not significantly different between the portal-pyruvate versus jugular-pyruvate groups at basal (5.77 ± 0.2 mmol/l, P = 0.69) nor were they statistically different during the hypoglycemic plateau (2.55 ± 0.12 mmol/l, P = 0.84; Fig. 3A). Furthermore, at basal (P = 0.76) and during sustained hypoglycemia (P = 0.17), no significant differences between the two groups were observed with respect to mean arterial pyruvate concentrations (Fig. 3B). On the other hand, as a result of the selective portal vein pyruvate infusion, mean portal vein pyruvate levels were significantly elevated (1.81 ± 0.21 vs. 0.42 ± 0.05 mmol/l, P < 0.0001; Fig. 3B) in portal-pyruvate versus jugular-pyruvate animals, respectively. In jugular-pyruvate animals, hypoglycemia elicited ∼16-fold increase in the epinephrine response from basal 1.00 ± 0.19 to 16.15 ± 3.21 nmol/l at min 105 of the hyperinsulinemic-hypoglycemic clamp. On the other hand, as a result of selective elevation of portal vein pyruvate concentrations, epinephrine response to whole-body hypoglycemia in the portal-pyruvate group was suppressed by ∼64% at min 105 compared with the jugular-pyruvate group (5.84 ± 0.71 vs. 16.15 ± 3.21 nmol/l, P < 0.01; Fig. 3C). In jugular-pyruvate animals, norepinephrine concentrations increased approximately threefold from the basal value of 2.25 ± 0.38 to 6.13 ± 0.77 nmol/l at the end point of hypoglycemia. Similar to the epinephrine response, elevation of portal vein pyruvate resulted in a ∼80% suppression of the norepinephrine response above basal at min 105 of hypoglycemia (0.78 ± 0.41 vs. 3.88 ± 0.69 nmol/l, P < 0.01; Fig. 3D).

By design, there were no significant differences between the portal-BHB and jugular-BHB groups with respect to arterial glucose concentrations at basal (5.86 ± 0.17 vs. 5.61 ± 0.16 mmol/l; P = 0.45; Fig. 4A), during the fall in glycemia, and during sustained hypoglycemia (2.53 ± 0.14 vs. 2.52 ± 0.14 mmol/l; P = 0.91; Fig. 4A). Also, there were no significant differences between portal-BHB and jugular-BHB treatments with respect to mean arterial BHB levels at basal (0.55 ± 0.23 vs. 0.67 ± 0.10 mmol/l; P = 0.59, data not shown) and during hypoglycemia (0.12 ± 0.03 vs. 0.16 ± 0.03 mmol/l; P = 0.32; Fig. 4B). Portal vein BHB infusions elevated BHB concentrations in the portal vein ∼12-fold in the portal-BHB versus jugular-BHB groups (2.02 ± 0.03 vs. 0.163 ± 0.03 mmol/l; P < 0.0001; Fig. 4B). In contrast to selective elevation of portal vein lactate or pyruvate, no significant differences were observed between the portal-BHB versus jugular-BHB groups with to respect epinephrine (5.42 ± 0.92 vs. 6.84 ± 1.59 nmol/l; P = 0.51 at min 105; Fig. 4C) and norepinephrine (2.67 ± 0.49 vs. 3.45 ± 0.81 nmol/l; P = 0.47 at min 105; Fig. 4D) responses to hypoglycemia.

In the current study, we investigated the effects of selective portal vein elevation of lactate, pyruvate, or BHB on the sympathoadrenal response to insulin-induced systemic hypoglycemia. Clamping portal vein lactate or pyruvate concentrations at approximately fivefold above basal, while maintaining arterial concentrations during systemic hypoglycemia, resulted in a substantial suppression of the epinephrine and norepinephrine responses, 64 and 75%, respectively. In contrast, elevating portal vein BHB to levels 12-fold above basal failed to impact the sympathoadrenal response to hypoglycemia. The ability to detect alternative metabolic substrates known to enter the glycolytic pathway distal to the uptake of glucose is indicative of a portal vein metabolic sensor rather than a glucoreceptor specific for the glycosyl moiety. The metabolic nature of hypoglycemic detection at the portal vein is further underscored by the fact that portal elevation of either lactate or pyruvate led to a quantitative blunting of the sympathoadrenal response comparable with that achieved via portal glucose normalization (Fig. 5).

Whole-body infusions of lactate have been shown to blunt the sympathoadrenal response to insulin-induced hypoglycemia in humans (16,17). It is generally assumed that such lactate-induced suppression of the sympathoadrenal response is mediated via its action on brain metabolic sensors. Lactate can be readily used by nervous tissue (32) and has been shown, both in vitro and in vivo, to activate glucose-sensitive neurons in the brain (19–21,33,34). The ability of lactate to alter the firing rate of glucose-responsive neurons has been demonstrated in the hypothalamus (19,20) and in the hindbrain (35). However, in vitro studies have not been uniform in their observations of the effect of lactate on glucose-inhibited hypothalamic neurons. Yang et al. (21), reported that the addition of lactate, similar to glucose, decreased the firing rate of hypothalamic glucose-inhibited neurons, whereas Song and Routh (19) observed an increased discharge from such neurons. Local microdialysis of the ventromedial hypothalamus with 100 mmol/l lactate has been shown to substantially suppress the sympathoadrenal response to systemic hypoglycemia in rats in vivo (33). Exogenous lactate delivery (50 μmol/h) into the caudal fourth ventricle has also been observed to prolong recovery from insulin-induced hypoglycemia (34).

Although the abovementioned findings clearly demonstrate that elevated brain lactate levels can impact the counterregulatory response to systemic hypoglycemia, it is not clear that the central nervous system (CNS) is the locus for those metabolic sensors responding to plasma lactate elevations in vivo as previously reported (16,17). The current findings clearly demonstrate that the selective elevation of portal vein plasma lactate levels, similar to that used in studies of whole-body plasma lactate elevations, substantially suppresses the sympathoadrenal response to systemic hypoglycemia. The suppression in the epinephrine response to hypoglycemia with local portal vein lactate elevation in the current study, a 64% suppression, is comparable with that reported for whole-body lactate elevation during hypoglycemia, i.e., 60–70% (16,17). As noted above, normalizing glycemia or elevating lactate in the portal vein results in a comparable suppression in the sympathoadrenal response to hypoglycemia, which is similar in magnitude to that reported for whole-body lactate infusions. In contrast, elevating brain glucose levels during systemic hypoglycemia is purported to result in a substantially greater suppression of the epinephrine response to hypoglycemia (∼90%) (7). Thus, the response to elevating whole-body plasma lactate levels during hypoglycemia in quantitative terms appears more consistent with the blunting of portal vein metabolic sensors, as opposed to CNS metabolic sensors.

Nutrient sensing by hypothalamic glucose-responsive neurons has been attributed to intracellular changes in the ATP-to-ADP ratio and the subsequent binding of ATP to K_ATP, leading to changes in cell potential (19,24). Exposure of hypothalamic glucose-responsive neurons to glucose or lactate results in closure of K_ATP channels, which can be reversed by the application of diazoxide (19). The intracerebroventricular application of the K_ATP channel closer, glibenclamide, blunts the sympathoadrenal response to hypoglycemia, suggestive of an ATP-dependent mechanism (36). The current observation that elevating portal vein pyruvate concentration during hypoglycemia results in a suppression of the sympathoadrenal response comparable with that of lactate or glucose is consistent with such an ATP-dependent mechanism. However, Yang et al. (20,21) reported that glucose-responsive and glucose-inhibited neurons in the hypothalamus do not respond to elevated pyruvate despite the fact that pyruvate is apparently metabolized by such neurons. These same investigators observed that early glycolytic intermediates, e.g., glyceraldehyde and lactate, were able to substitute for glucose and alter the firing rate in both glucose-responsive and glucose-inhibited hypothalamic neurons. Based on the inability of pyruvate to stimulate hypothalamic glucose sensors, Yang et al. concluded that ATP production was not the primary metabolic event responsible for “glycemic detection” in these cells. They postulated that hypothalamic glucose sensing must be mediated by products of glucose metabolism proximal to pyruvate oxidation, e.g., cytosolic NADH. This hypothesis is supported by a recent report demonstrating a glucose-induced increase in cytosolic NADH, but not ATP levels, in hypothalamic glucose-responsive neurons (37).

Although it remains unclear as to whether glucose-sensing neurons in the hypothalamus and portal vein use a common metabolic event(s), our findings are consistent with a postulated mechanism for pancreatic β-cell nutrient sensing (38). It is proposed that the intracellular rise in malonyl CoA signals nutrient abundance in the β-cell, and thus only those compounds capable of both substrate oxidation and citric acid cycle anaplerosis should be effective (39,40). Elevating portal vein pyruvate, lactate, or glucose but not BHB proved comparable in their efficacy for suppressing the sympathoadrenal response to hypoglycemia (Fig. 5). Of these substrates, only BHB is incapable of supplying citric acid cycle intermediates via anaplerosis. That the metabolic sensing mechanisms for the β-cell and portal vein glucose sensors might be similar is supported by reports that GLUT2 (41) and the pancreatic form of glucokinase (42), critical components of the β-cell glucose-sensing “machinery,” are both expressed in the portal region.

As noted above, elevating portal vein BHB levels fails to impact either epinephrine or norepinephrine secretion during hypoglycemia (Fig. 4). This is in contrast to whole-body ketone infusions, which have previously been shown to suppress the counterregulatory responses to insulin-induced hypoglycemia (16). Clearly the impact of whole-body hyperketonemia on the sympathoadrenal response to hypoglycemia must be manifest at some locus distal from the portal vein, most likely the CNS. Iontophoretic application of BHB has been observed to modify neuronal discharge of hypothalamic glucose-responsive and glucose-inhibited neurons in vitro in a dose dependent manner (18). Although it is possible that the portal vein BHB level achieved in our current study (2.02 ± 0.03 mmol/l) was insufficient to mask portal vein metabolic sensing or that BHB is not transported into portal glucose sensors, neither proposal appears plausible. Previous studies showed that elevating ketone levels ∼10-fold above basal to arterial concentrations of ∼1–2 mmol/l significantly diminishes the sympathoadrenal response to hypoglycemia in humans (16), dogs (43), and rats (44). Additionally, transport of lactate, pyruvate, and BHB across cellular plasma membranes uses the same family of monocarboxylate transporter proteins (45). Therefore, transport across plasma membrane would not account for the differences observed between lactate or pyruvate and BHB. Assuming that the impact of BHB is effected at the level of the CNS, the inability of portal vein metabolic sensors to respond to ketones may be further evidence distinguishing that metabolic sensing at the portal vein is distinct from that which governs sensing at the brain.

Metabolic sensors have long been known to reside in the hepatoportal area and are understood to be involved in the regulation of feeding (46). Early studies by Niijima (47,48), using an in situ guinea pig liver perfusion preparation, demonstrated that the rates of hepatic vagal afferent discharge are inversely proportional to glucose availability in the perfusion medium. Subsequent studies showed that afferent discharge was also augmented by intraportal administrations of 2-deoxy-d-glucose (glucose oxidation inhibitor) (47) or mercaptoacetate (fatty acid oxidation inhibitor) (49), suggesting that the glucose sensory mechanism is dependent on the intracellular fuel oxidation and the subsequent ATP production. In support of this, intraperitoneal administration of lactate, pyruvate, or BHB has been shown to suppress feeding; however, no change in food intake in response to these metabolites was observed after hepatic branch vagatomy (50). It was therefore proposed that glucose-sensitive hepatic vagal afferents regulate feeding as a result of intracellular fuel oxidation and subsequent changes in mitochondrial ATP production (50).

Despite the fact that some investigators have suggested that glucose-sensitive vagal afferents are also involved in the regulation of the sympathoadrenal response (47), studies have failed to observe this relationship in vivo (15,51). On the other hand, we have shown that sectioning the celiac ganglion, through which spinal afferents originating from the portohepatis ascend, substantially impairs the ability of the animal to mount a full sympathoadrenal response to hypoglycemia (15). Thus, it appears that feeding in the periphery is mediated by glucose-sensitive hepatic vagal afferents, where hypoglycemic detection is mediated by spinal afferents transcending the celiac ganglion. Furthermore, this study points out yet another revealing distinction between the two sets of hepatoportal sensors, at least in respect to their responsiveness to ketone bodies. Clearly, further studies are warranted to elucidate exact molecular and cellular mechanisms of portal vein metabolic sensors.

In summary, our current findings indicate that the portal vein glucose sensor responds to alternative metabolic substrates such as lactate and pyruvate and therefore behaves as a metabolic sensor rather than a glucose-specific receptor. Portal vein elevations of glucose, lactate, and pyruvate but not BHB equally suppressed (60–80%) the sympathoadrenal response to hypoglycemia. These findings suggest a mechanism of cellular glycemic detection mediated by events distal to pyruvate oxidation, requiring anaplerotic input into the citric acid cycle. In this regard, the mechanism for glycemic detection at the portal vein appears similar to that of the β-cell. That pyruvate and BHB demonstrate differential effects at the portal vein and CNS indicates distinct mechanisms for hypoglycemic detection in these two loci. Elucidation of the cellular mechanism(s) involved in portal vein metabolic sensing and how that information is integrated with glucose sensors in the CNS remains a significant challenge to our understanding of hypoglycemic counterregulation in diabetes.

C.M.D. has received National Institutes of Health Grant DK-55257.

We thank Satoshi Fujita, Maziyar Saberi, and MaryAnn Bohland for extensive contributions to the study.