Intense Exercise Has Unique Effects on Both Insulin Release and Its Roles in Glucoregulation: Implications for Diabetes

Moderate postabsorptive exercise is an extraordinary example of euglycemic homeostasis (Fig. 1A). There is a precise match between the increases in GU and GP. This is considered to be feedback-regulated by signals associated with the increased demand by the exercising muscles, causing responses that increase GP to match GU. Insulin secretion is inhibited to below fasting levels (Fig. 1B) by α-adrenergic receptor activation (2,3), both via the sympathetic innervation of the islets and by circulating catecholamines (Figs. 1C,D). This decrease of insulin secretion is important because it increases GP by the liver by sensitizing it to glucagon (4–6). It is established that the decreases in insulin and unchanged or increased glucagon account for the corresponding increases in GP (7,8) (Figs. 1B and 2A). Thus, the ratio of glucagon to insulin is the main regulator of GP during moderate exercise (5,9) (Fig. 2B). Catecholamines play a role in increasing GP, presumably through gluconeogenesis, only during prolonged exercise >2 h duration (10–12). Their increases during 40 min of moderate exercise (Fig. 1C and D) are modest and predicted to have small GP effects based on results in animal studies. The increment of GU (Fig. 2D) is precisely adapted to the increment of GP (Fig. 2C). The signal(s) coordinating these liver and muscle responses, and whether insulin plays any role, remain unknown. What has been known for many years is that muscle GU increases despite decreased insulin, because exercise causes translocation of GLUT4 glucose transporters from a different pool than insulin (13,14), and the exercise-induced signaling of GU is independent of insulin signaling (15). Also, increased peripheral blood flow augments total insulin delivery to muscle and thus compensates at least in part for the decreased plasma insulin concentrations. This also explains why the insulin and exercise effects are additive (16). The effects of both insulin and exercise counteract the effect of catecholamines, which under most conditions inhibit muscle GU.

The importance of insulin in moderate exercise can be best illustrated in type 1 diabetes (4,17,18). GP and GU are well balanced in experiments in which basal insulin is infused intravenously at a constant rate appropriate to maintain euglycemia at rest. Under these conditions, plasma insulin does not decrease during exercise, but its relationship to portal glucagon levels is such as to cause increased GP by the liver. However, in severe insulin deficiency the increments in GP with exercise may exceed those in GU and lead to further hyperglycemia. In contrast, in the usual treatment of type 1 diabetes, hyperinsulinemia may ensue because exercise can increase insulin mobilization from its subcutaneous depot, particularly if the injection site is in an exercised region. This results in falling glucose levels that may reach the hypoglycemic range, depending on the pre-exercise concentration, because the rise in insulin blocks the effect of glucagon on liver GP and concurrently amplifies GU even more than that required for the exercise. In contrast, we will describe how the insulin concentration perfusing the liver during intense exercise is of minor importance. We will present data to show that this is because catecholamines, rather than the glucagon/insulin interaction, are the main regulators of GP in this special setting. Unless otherwise indicated, “exercise” hereafter refers to intense exercise.

Metabolic responses to intense exercise can be divided temporally into those occurring during the muscular work itself, those occurring immediately after exhaustion as the acute and massive changes reverse, and those that occur later in recovery (19). Typically, even very fit persons can sustain exercise at ≥80% Vo_2max for periods of only 10–15 min, although elite athletes may do so for longer. Exercise experiment paradigms, as well as certain sports that require this level of exertion, often use exercise that generates maximal or supramaximal Vo₂ for up to 3–5 min. These bouts may be repeated with variable intervening time intervals. If the rest interval is short (e.g., several 5-min bouts) the metabolic responses, especially hyperglycemia, are additive (20). If the intervals are as long as 1 h, most of the components of a single response are repeated (21). These responses to multiple, repeated bouts will not be considered in detail here, in part because existing techniques for studying glucose turnover in this setting would be difficult to apply and interpret. Indeed, even for single bouts, it was necessary for us to devise a modified approach to the conventional primed-continuous tritiated glucose infusion method in order to define the rapidity and magnitude of changes in glucose turnover. We therefore used stepwise increments in the 3-[³H]glucose infusion rates by up to 4- to 7.5-fold (21–23) during the exercise, with corresponding steps downward in recovery. Blood samples were taken at 2-min intervals. This approach attenuated the magnitude of changes in circulating glucose specific activity and total radioactivity that otherwise would have rendered estimates of GP and GU less precise. Thus, we were able to define the extremely rapid changes that occurred, in contrast to the more typical 10-min sampling intervals that are satisfactory in lower intensity exercise of longer duration (18). Our studies have been conducted in fit individuals undergoing exercise at >80%, with Vo_2max having been previously determined by a test using a progressive increment in resistance (ramp) on a cycle ergometer. The workload selected was subject to minor adjustments that resulted in an initial rapid increase in Vo₂ to a near steady state, followed by a minor, progressive rise by exhaustion, which in our studies occurred at 14–15 min of exercise.

In intense exercise, a signal for rapid hepatic glycogenolysis has been invoked, which anticipates the need for the marked increase in GP. In contrast to the “feedback” signal at lower exercise intensities, this has been suggested to be “feedforward” (24,25). Irrespective of the site of origin of the signal, we and others (26) have proposed that it is the marked catecholamine response during intense exercise that is the prime regulator of GP. In moderate exercise, plasma catecholamine concentrations increase only two- to fourfold (Fig. 1C and D) (27,28). In contrast, in intense exercise there is a marked 14- to 18-fold increase of both EPI and NE (Fig. 1C and D) (e.g., 19,21–23,25,26) to levels seen in pheochromocytoma. This induces a seven- to eightfold increase in GP (Fig. 2C), the largest increases seen under any physiological or pathophysiological condition. In contrast to moderate exercise, the increase in GU, though very large, is less than the increase in GP. This restraint in GU increment (Fig. 2D) is due to the largely catecholamine-mediated stimulation of muscle glycogenolysis. Despite the increase of plasma glucose that this GP-GU difference causes, plasma insulin concentration does not change or may even increase slightly. There are two contributing factors. The first is that the high concentrations of catecholamines (acting through a dominant α-adrenergic effect) can prevent glucose stimulation of insulin secretion. The second may be a decrease in insulin disposal (see below). These glucoregulatory responses and their regulation during intense exercise are shown in schematic fashion in Fig. 3.

Further marked changes ensue at exhaustion. Notable is the marked increase of plasma insulin that lasts up to 60 min during recovery (Fig. 1B) (e.g., 19,21,22), starting with the immediate and rapid decrease of NE and EPI concentrations. This reflects a rapid waning of the α-receptor–mediated inhibition of the β-cell response to hyperglycemia (with a probable return toward resting rates of insulin disposal). The simultaneous increases of glucose and insulin concentrations create a milieu that should favor repletion of at least part of the muscle glycogen mobilized during the exercise. Such rapid replenishment of muscle glycogen is important because in many such activities there is frequent repetition of short bouts of intense exercise. That the recovery hyperinsulinemia is a physiologically essential early recovery response is illustrated by results in diabetic subjects, as described below.

To more precisely define the regulatory contributions of insulin, glucagon, and growth hormone to the metabolic responses to intense exercise, we blocked the endogenous secretion of these hormones with intravenous octreotide and replaced their endogenous secretion by basal-rate infusions (Fig. 4) (22). The roles of the catecholamines could thereby be inferred more robustly, because their responses were not altered (not shown). GP increased normally ( Fig. 4A), strongly suggesting that it’s main regulators are catecholamines and not insulin and glucagon. GU likewise increased normally (Fig. 4B). Interestingly, despite continuing the insulin infusion at a constant rate, its plasma concentration increased threefold (Fig. 4C), indicating that intense exercise decreases insulin disposal. Furthermore, decreasing its infusion rate by half was associated with a lesser increase in plasma levels, and again the catecholamines and GP responses were the same (22). That GP was not attenuated by this hyperinsulinemia indicates that the stimulus for GP increase was able to overcome any suppressive effects of insulin and a decreasing glucagon-to-insulin ratio. As noted above, these responses of insulin suggest that the failure of intense exercise to suppress insulin without the islet clamp is not only due to the hyperglycemia overcoming its α-adrenergic suppression, but also to the effect of intense exercise to decrease insulin degradation. Finally, there appears to be little role for growth hormone in these metabolic responses, because they were the same with both the normal >10-fold increase as well as with the maintenance of its levels during infusion of octreotide (22).

Previous studies at rest and during moderate exercise to define the roles of hepatic α- and β-adrenergic receptor activation in stimulation of GP have yielded inconsistent results and species differences (29–31). Many different experimental approaches have been used. Factors in interpretation of in vivo studies include concurrent effects on insulin and glucagon secretion, and NE delivery to hepatocytes both by local release from sympathetic nerve terminals and via the circulation. The intensity, or perhaps specificity, of the stimulus to increased sympathetic tone and/or circulating catecholamines could also lead to differing interpretations, particularly in human studies (e.g., 32–37). Furthermore, the physiological status of the subjects plays a role—level of fitness and presence of disease determine the absolute as well as relative workloads during exercise. Examples of differing results include the following: α-receptors have been found to be most important in stimulating hepatic GP in rats (38), whereas in dog hepatocytes, β-receptors have been reported to be dominant, and both play a role in cats (reviewed in 30). Propranolol and phentolamine infused together into the portal vein were unable to prevent the GP increment in dogs exercised to 85% of their maximum heart rate, which caused a fourfold GP increment (31); however, they were able to block the GP response to portal infusion of NE and EPI, possibly related to the difference between neural and “hormonal” effects.

In an attempt to define the contributions of α- and β-receptors in intense exercise, we infused the receptor-blocking agents phentolamine (Fig. 5) or propranolol (Fig. 6) intravenously, starting 30 min before the bout and ending at 60 min of recovery (35,36). With phentolamine (36), the importance of the usual islet α-receptor effect of inhibiting insulin secretion during exercise (2,3) is clearly illustrated by the threefold increase in insulin (Fig. 5B), despite the fact that the increment of glycemia was substantially decreased (Fig. 5A). Phentolamine thus unmasked the β-receptor–mediated stimulation. Accordingly, because the glucagon response was little altered, the glucagon-to-insulin ratio decreased. The hyperglycemia was less marked because α-blockade decreased the peak increment of hepatic GP (Fig. 5C). This was associated with unaltered GU and with markedly enhanced elevations of both NE and EPI, the latter possibly due to decreases in their clearance (see below). These results are consistent with phentolamine blocking part of the GP response to the large rises in catecholamines, and indicate the presence of GP stimulation mediated by α-receptors. That more restraint of the GP increment did not occur could likewise have been due to some β-receptor–mediated stimulation of GP. Though we cannot exclude a partial attenuation of GP related to the effect on the glucagon-to-insulin ratio, the failure of similar changes in other experiments to do so makes this less likely.

Most human studies have suggested a predominance of β-receptor stimulation of GP (30,39). Were this is the case in intense exercise, β-blockade should decrease GP. However, we found the opposite with the nonspecific β-blocker propranolol (35); at matched exercise intensities, a twofold higher peak GP occurred versus control responses (Fig. 6C). Interpretation of this response must take account of the following. During exercise with propranolol infusion, there was also a greater increase in NE and EPI levels (not shown), along with an early decrease in plasma insulin (Fig. 6B) and an increase in plasma glucagon, resulting in an increased glucagon-to-insulin ratio (not shown). Furthermore, while β-blockade was clearly present because of attenuated increases of heart rates and lipolysis (not shown), it cannot be argued that the blockade was complete, and some β-adrenergic contribution to the increase in GP remains a possibility. In similar studies with β-blockade in type 1 diabetic patients kept euglycemic before exercise by insulin infusion, a similarly enhanced increase in GP was found, which occurred despite increase in plasma free insulin during exercise (40). The increased plasma catecholamine response was also observed. Thus, because the same GP increment occurred with propranolol in the face of both decreased and increased insulin, it seems unlikely that the glucagon-to-insulin responses played a role. Previous studies in humans at rest demonstrated that the increase in NE levels with propranolol infusion (41,42) is due to decreased clearance, rather than increased release, related primarily to increased vasoconstriction (42). This is the likely explanation for the markedly greater response of the plasma catecholamines in our studies.

On the basis of these three studies, we have suggested that during intense exercise both α- and β-receptor effects of catecholamines can increase GP (35,36,40). Because with blockade of either of the receptors the plasma concentrations of catecholamines rose considerably more, this suggests that one system can compensate for the attenuated activity of the other, i.e., there is redundancy in the receptor mechanisms in this specific situation.

The net glycemic effect of propranolol in both diabetic and control subjects was to lessen the magnitude and duration of the exercise hyperglycemia (Fig. 6A) (35,40). This was due to greater enhancement of GU (Fig. 6D) (35,40), likely to occur by inhibition of the well-documented β-adrenergic stimulation of muscle glycogenolysis, thereby allowing for greater uptake of blood-borne glucose. One cannot exclude a contributing role for the concurrent inhibition of lipolysis (leading to a greater decline in plasma free fatty acid) that is normally stimulated by increased β-adrenergic activity at the adipocyte. In the diabetic subjects, this greater GU occurred in the absence of the normal sustained postexercise hyperinsulinemia and returned the glycemia to pre-exercise levels. If it were not for other less desirable effects of β-blockade, this might be part of the strategy for controlling postexercise hyperglycemia in diabetes (see below).

We hypothesized that if the proposed catecholamine mechanism is necessary for the GP response to intense exercise independently of a glucose feedback mechanism, such a response should occur even if pre-exercise endogenous GP was completely suppressed by continuous intravenous glucose infusion and its accompanying hyperinsulinemia. Indeed, it should occur even if an amount of glucose equivalent to the normal GP increment were infused exogenously during exercise in addition to the constant infusion begun before exercise. This would be a further contrast to the feedback mechanism for glucoregulation in lower intensity exercise in which attenuation or complete prevention of the endogenous GP increment has been shown (43–48), due to increase of insulin and decrease of glucagon in experiments in which the hormone responses were endogenous, as well as in somatostatin studies in which levels were adjusted by exogenous hormone replacement. We therefore determined the effects of stepwise increment of glucose infusion during exercise and decrement in the early postexercise period that mimicked the endogenous GP responses of control subjects (Fig. 7) (23). To mimic the fed state, glucose was infused before, during, and after exercise; endogenous GP can thereby be measured accurately by subtracting the known glucose infusion rate from total GP. As shown in Fig. 7A, plasma glucose concentration was elevated before exercise and for the first 45 min of recovery, compared with values in uninfused subjects. Despite the total suppression of GP at rest, its endogenous increment was only modestly attenuated during intense exercise (Fig. 7B). This is in contrast to moderate exercise, in which exogenous glucose infusion abolishes endogenous GP (43–48). The endogenous elevation of plasma insulin during glucose infusion appears to inhibit GP only marginally because the huge catecholamine response (which was unaffected, not shown) overcomes the insulin-induced suppression (Fig. 7C). Alternatively, this could be interpreted as catecholamine-induced hepatic insulin resistance. The markedly greater postexercise increment in plasma insulin (Fig. 7C) was parallel to that of the fasted state (albeit even higher) and was associated with increased GU (not shown), probably mostly into muscle.

We subsequently studied the response 3 h following a mixed meal (49). Before exercise, there was elevated glucose appearance into the circulation associated with hyperinsulinemia (and therefore probable suppression of endogenous GP), yet the increase in GP was comparable with that in fasted subjects, and catecholamine responses were the same. It is highly improbable that there would be increase in glucose absorption from the gut during intense exercise. Thus, these results are compatible with those of the glucose infusion, with the same conclusion: that prior suppression of GP by insulin can be overcome by the catecholamine response. In both studies, the GU responses were greater during exercise and persisted during recovery, compatible with the higher insulin levels enhancing glucose uptake into insulin-sensitive tissues, both as primary energy source and substrate for restoration of glycogen depleted during exercise. Accordingly, in the postprandial study, the hyperglycemic response was less than in the fasted state.

Convincing evidence for catecholamines being the primary mediators of the huge GP response to intense exercise would be if mimicking their plasma levels by exogenous infusion reproduced the response. Such infusions at rest would probably produce unacceptable symptoms. Given the exertion required at 50% Vo_2max and that glucoregulation at lower intensities is not primarily catecholamine-mediated, we tested whether infusion of NE, EPI, or both could “convert” the response of moderate to that of intense exercise. Therefore, fit normal subjects were exercised for 40 min at 50% Vo_2max, and either the individual catecholamines were infused for the last 10 min or both together were infused during the last 14 min. Figure 8 illustrates the effects of EPI infusion at a constant rate that produced circulating concentrations that peaked in the range found in intense exercise (27). This infusion induced rapid increases in GP from the typical twofold increment over resting rates to over half the peak rate of intense exercise (Fig. 8C). Though GU also increased to greater than the twofold rise of moderate exercise, it peaked at less than the GP increment (Fig. 8D), so both during the infusion and in early recovery, hyperglycemia occurred (Fig. 8A). Unlike after moderate exercise, these effects were accompanied by recovery hyperinsulinemia (Fig. 8B). Thus, most qualitative aspects of the intense exercise response were reproduced, but were quantitatively smaller. Similar observations were also made with infusions of NE that reproduced the plasma levels of intense exercise (28), though the magnitudes of all of the responses were somewhat less than those to EPI. The sum of GP increments achieved separately with EPI and NE (not shown) would approximate that of intense exercise. This would only be a valid calculation if the effects were additive. Therefore, with the knowledge of these effects at the plasma catecholamine concentrations achieved, we performed combined infusions of EPI and NE at incremental rates designed to fully mimic the patterns and magnitudes of intense exercise (50). The latter was achieved, and the increase in GP was indeed predicted by the sum of the two and closely matched that in intense exercise (not shown). Most other hormonal and metabolic responses did likewise. In contrast to both the individual and combined infusions, the catecholamines augmented rather than restrained the increased GU (not shown), although some hyperglycemia did occur because these increments were of smaller magnitude than the increase in GP. We conclude, therefore, that the effects of catecholamines on GU during exercise are different from those during the resting state, in which they are typically reduced. In these experiments, there was a postexercise hyperinsulinemia that corresponded to the magnitude and duration of the hyperglycemia (Fig. 8B). Once again, given the absence of “appropriate” changes in insulin and glucagon during the infusions (27,28,50), virtually the whole of the glucoregulatory responses can be attributed to the levels of catecholamines achieved. These results clearly indicate the capacity of NE and EPI to explain the huge increase in GP of intense exercise. That they can do so by peripheral venous infusion implies that they can act as hormones rather than requiring local release of NE to generate the intense exercise response (a contribution of the latter is nonetheless probable). Furthermore, EPI appears to be a more potent stimulator than NE at identical infusion rates and at the circulating levels achieved both endogenously and by infusion. Finally, this infusion model confirms the need for the postexercise hyperinsulinemic response to return plasma glucose to resting concentrations.

Very little research has been done in persons with type 2 diabetes, as most are unlikely to perform strenuous exercise. Given the importance of the increased secretion of insulin in the postexercise period and the relatively small roles of insulin and glucagon during the bout, it would be predicted that persons with type 1 diabetes might have normal GP and GU responses during and immediately following exercise, but abnormal responses for the rest of the recovery period. Our studies in such subjects allow for some compelling inferences with respect to these predictions (40,51,52) (summarized in Fig. 9). Fit, young subjects with no residual subcutaneous insulin received continuous intravenous insulin overnight that maintained either normal or elevated plasma glucose concentrations (Fig. 9A). In all three studies, the rates of infusion producing these levels were kept constant during exercise. In one euglycemic study and in the hyperglycemic study these rates were maintained during all of recovery, and in one euglycemic study the rate was doubled at the end of exercise. The catecholamine responses and those of GP and GU during exercise and the first minutes of recovery were normal (not shown). Notably, in both studies with constant insulin infusion, the hyperglycemia of exercise continued for the 2-h recovery period (Fig. 9A). However, doubling the infusion rate returned glycemia to pre-exercise values, albeit from higher levels and more slowly than occurred in the control subjects. The plasma insulin levels approximated those in control subjects by this simple maneuver (Fig. 9B). From this we conclude that in type 1 diabetes the response that is catecholamine-mediated is preserved, but that without extra postexercise insulin, intense exercise can cause sustained hyperglycemia. With repeated bouts, it is probable that this would progressively worsen. We tested whether inducing an even greater immediate postexercise hyperinsulinemia in the normal subjects could accelerate the decrease in GP—it could not (51). Thus, the decline is due to the drop in catecholamines and therefore a strategy to mimic the normal pattern and magnitude of postexercise hyperinsulinemia may be feasible with available insulins given subcutaneously.

As noted, the increments and early recovery decline of GU were normal (not shown). Noteworthy is that metabolic clearance of glucose was also normal when euglycemia was present at rest. Although in all studies in diabetic subjects it increased somewhat more slowly during exercise, it was much lower postexercise in the hyperglycemic than in the control subjects (Fig. 9C). GU was normal in the latter because increased glucose mass compensates for decreased glucose metabolic clearance. We have speculated that its suppression represents a protective mechanism against hyperglycemia, and that this is the reason that muscle may not have the diabetic complications seen in organs that do not have this mechanism. Most recently, we have shown in the perfused rat hindquarter that this mechanism applies to both hyper- and hypoglycemia, and that it is due to altered translocation and synthesis of glucose transporters (53). Thus, with hypoglycemic and hyperglycemic infusions, the number of glucose transporters in plasma membrane increases and decreases, respectively. This is proportional to changes in GU (53).

As noted, β-blockade before, during, and after exercise normalized the postexercise decrease of plasma glucose (40). This effect is similar to the increment of insulin given in the postexercise period (Fig. 9A) and is due to a marked increase of glucose metabolic clearance. We made a similar observation in diabetic dogs, where stress (intracerebroventricular injection of carbachol) caused hyperglycemia in normal but not in diabetic dogs; this response was prevented by infusion of propranolol (54). The potential role of β-blockade in preventing hyperglycemia following stress or intense exercise remains to be explored.

What are the practical implications of the present findings in type 1 diabetic subjects? One must be circumspect in extrapolating from a small number of fit postabsorptive, complication-free, young diabetic men to the larger number of individuals who would more usually perform exercise in the nonfasting state, be less fit, and not necessarily young, male, and complication-free. There is need to develop strategies for diagnosing and managing exercise-related hyperglycemia (55). From our data, it seems probable that brief intense exercise in the postabsorptive state before insulin injection may lead to prolonged hyperglycemia. Second, if plasma glucose is elevated, it could become more elevated after intense exercise. This would depend on the types of insulin used and the timing of exercise after injection, as well as the type of exercise and its duration. Third, increasing the plasma insulin concentration very soon after intense exercise, even if given in a less-than-physiological fashion, can restore glucose clearance and glycemia toward normal. For the moment, we can only suggest an individualized approach based on blood glucose monitoring before and for several hours following exercise. For those individuals who regularly and predictably become hyperglycemic after intense exercise, it may be appropriate to recommend a small additional insulin dose immediately after exercise. The optimal agent, dose, and timing to achieve this goal remain to be defined, but one promising approach may be the use of rapidly absorbed monomeric insulin analogs. We would also predict that the approach might need to be different in women than in men with type 1 diabetes. We have shown that for comparable relative exercise intensity based on Vo_2max, normal young women have higher postexercise hyperglycemia and correspondingly greater hyperinsulinemia (56). Finally, it should be reassuring to competitive athletes that their principal fuel-homeostatic responses during the performance of these types of strenuous exercise is, in fact, intact.

There is need for much more study in type 2 diabetes. In obese subjects without diabetes there are abnormal responses, probably related to insulin resistance (57). The hyperglycemic response is prolonged and associated with a much greater and longer hyperinsulinemic period, and while NE responses are normal, EPI responses are half those of lean subjects. The same general pattern of response was found in obese persons during low-energy diets, very-low-energy ketogenic diets, and total fasting, though at different resting levels of all substrates and of insulin (58), perhaps reflecting in part the diets’ effects on insulin secretion and action. The combined effects of hypoenergetic diets and diabetes have not been defined. One recent study in type 2 diabetic subjects on diet only addressed postprandial metabolic responses to four bouts of stepped exercise with the highest intensity at 98% Vo_2max for 4 min (59). Glucose and insulin excursions were lower with exercise because of increased GU and clearance but only after breakfast, not after lunch. Because this absence of a hyperglycemic effect is unusual, it needs confirmation in other protocols before being generalizable to all type 2 diabetes and intense exercise. As with other studies in unfit persons, the absolute workload may have been lower than that which generates the endogenous GP responses described herein, despite being high-intensity relative to these subjects’ individual values for Vo_2max.

In conclusion, in intense exercise, insulin secretion does not decrease, or it does so minimally; hyperglycemia occurs because GU increases less than GP. We postulate that the α-adrenergic inhibition of insulin secretion is counterbalanced by hyperglycemia and decreased degradation of insulin. Results of studies with α- or β-blockers suggest that both receptors can stimulate GP and that the lack of one can be compensated for by the other. Compelling evidence that catecholamines are the main regulators of hepatic GP derives from studies in moderate exercise with catecholamine infusions. These mimic all the main parameters of intense exercise, including the responses of insulin, GP, and GU. Hyperglycemia occurs and persists in insulin-infused diabetic subjects in the postexercise period, indicating the need for additional exogenous insulin during this period. Thus, those patients who become hyperglycemic after intense exercise may need an injection of ultra-rapidly acting insulin at a time determined empirically. More study of strenuous exercise in type 2 diabetes is required.

This work was supported by Canadian Institutes of Health Research (CIHR) grants to E.B.M. and M.V.

We appreciate the help of D. Bilinski, L. Vranic, Marie Lamarche, and Madeleine Giroux in preparing this manuscript. The studies were the result of an invaluable collaboration with Dr. Jeffrey B. Halter (Ann Arbor, MI) and a series of superb postdoctoral fellows, graduate students, summer medical students (all co-authors on our cited papers), and could not have been performed without the assistance of our dedicated technicians and research nurse.

We dedicate this paper to our colleague and friend over many decades, Dr. Daniel Porte, Jr., on the occasion of his recent retirement from the University of Washington, Seattle, WA.