Pharmacological Antilipolysis Restores Insulin Sensitivity During Growth Hormone Exposure

Seven adult GH-deficient patients (six men and one woman) were included in the study (Table 1). One patient had childhood-onset GH deficiency (no. 5), whereas the other patients had adult-onset GH deficiency secondary to juxtasellar pathology and/or its treatment. The mean (± SE) fasting blood glucose level was 5.0 ± 0.2 mmol/l (range: 4.3–5.8 mmol/l), and the mean (± SE) HbA_1c level was 5.3 ± 0.2% (range: 4.8–6.0%). The diagnosis of GH deficiency was ultimately based on a GH response <3 μg/l after an arginine stimulation test (Table 1). Patients 3 and 6 also underwent an insulin tolerance test. All pituitary replacement therapy, including GH treatment, was administered in an unchanged dosage for at least 6 months before the study. Informed consent was obtained from all participants, and the study was approved by the local ethics committee.

All seven patients were studied four times in random order: during continuation of GH replacement (A); after discontinuation of GH replacement for 2 days (B), during continuation of GH replacement plus administration of acipimox (C), and after discontinuation of GH, but with administration of acipimox (D). GH was administered as subcutaneous self-injections at 2200; for study conditions B and D, the last two GH injections were discontinued. In study conditions C and D, the patients received four doses of acipimox 250 mg, p.o., with two doses administered at 2000 and 2300 the evening before and two administered at 0600 and 1000 on the day of the metabolic study. All metabolic studies were performed between 0800 and 1400 (0−360 min.). Intravenous catheters were placed in an antecubital vein and in a heated dorsal hand vein on the contralateral arm for the measurement of forearm arterial and deep venous metabolite balances. The subjects were studied in the basal postabsorptive state for 180 min (0800–1100), followed by a hyperinsulinemic/euglycemic clamp for 180 min (1100–1400). At 0800, a primed (20 μCi) continuous (10.4 μCi/h) infusion of [3-³H]glucose was started. After allowing 3 h for tracer equilibration, glucose metabolism in the basal state was assessed. At 1100, a constant infusion of insulin (0.6 mU · kg⁻¹ · min⁻¹; Actrapid, Novo Nordisk, Gentofte, Denmark) was begun. During insulin infusion, plasma glucose was clamped at 5.0 mmol/l by adjusting the rate of infusion of 20% glucose according to plasma glucose measurements every 10 min. Indirect calorimetry was performed at the end of the basal period (1030−1100) and at the end of the hyperinsulinemic clamp period (1330−1400).

Plasma glucose was determined in duplicate immediately after the samples were drawn using a Beckman glucose analyzer (Beckman Instruments, Palo Alto, CA). Whole-blood glycerol, 3-hydroxybutyrate, lactate, and alanine were analyzed by fluorometric enzymatic methods (12). FFAs were measured by a colorimetric method. GH and cortisol were assayed by a time-resolved fluoroimmunoassay (Delfia, Wallac, Turku, Finland). The intra- and interassay coefficients of variation of the GH assay were 1.7–2.4 and 1.9–3.0%, respectively; the lower detection limit was 0.01 μg/l. Serum IGF-I was determined by an in-house, time-resolved fluoroimmunoassay with average intra- and interassay coefficients of variation <5 and <10%, respectively (13). The mean level and range of serum IGF-I concentrations in healthy adults (age 27–59 years) with this assay are 177 and 75–275 μg/l, respectively. Insulin and C-peptide were measured by commercially available immunoassays (DAKO, Glostrup, Denmark). The specific activity of tritiated glucose was determined as previously described by Møller et al. (14). Rates of appearance (R_a) and disappearance (R_d) of glucose were calculated using Steele’s equation for non−steady state, and a pool fraction of 0.65 was used. Endogenous glucose production during the clamp was calculated by subtracting the rate of glucose infusion (M value) from glucose R_a, as determined isotopically. Indirect calorimetry using a computerized open circuit system was performed for the measurement of gas exchange across a canopy (Deltatrac; Datex Instruments, Helsinki, Finland). From the measurement of gas exchange, the energy expenditure (EE) and respiratory exchange rate were calculated. Rates of lipid and glucose oxidation were derived from indirect calorimetry after correction for protein oxidation, which was estimated from the urinary excretion of urea. Urine for the assessment of urea excretion was collected at 0800−1400 (15). Nonoxidative glucose disposal was determined by subtracting oxidative glucose disposal from total glucose R_d, as determined by the infusion of tritiated glucose. Arterialized and deep venous blood samples were drawn simultaneously, and forearm blood flow was determined by venous occlusion plethysmography. Forearm blood flow was not affected by administration of GH or acipimox; consequently, forearm exchange of metabolites are presented as arteriovenous differences in plasma concentrations of these metabolites. Data for circulating hormones and metabolites were derived from triplicate measurements during the final 30 min of the basal and clamp periods.

All results are expressed as means ± SE. Statistical calculations were done by analysis of variance (ANOVA) for repeated measures. For time series (i.e., circulating hormones), the area under the curve (AUC) was calculated by the trapezoidal method, and comparisons were made by ANOVA. Where appropriate, post hoc comparisons of the different study days were made by means of a paired t test. All calculations were done using the computer program SPSS version 10.0 (SPSS, Chicago, IL). P < 0.05 was considered statistically significant.

During the basal period, serum levels of GH were higher when the patients continued GH treatment (A and C). As expected, serum GH levels gradually declined with time in the studies where GH was administered (A and C). When GH was discontinued, administration of acipimox tended to elevate GH levels (P = 0.05) (Fig. 1).

AUC_insulin during the basal period was not different on the 4 study days. During the hyperinsulinemic clamp, AUC_insulin was higher when GH and acipimox were administered together than when acipimox was administered alone (AUC_insulin in pmol · l⁻¹ · min⁻¹: 48,568 ± 6,982 [A] vs. 44,006 ± 6,140 [B] vs. 46,979 ± 5,984 [C] vs. 40,653 ± 4,425 [D]; P = 0.039) (Fig. 2). However, insulin levels during the final 30 min of the clamp did not differ among the four studies (data not shown). AUC_C-peptide during the basal period was higher when GH was given (A and C) than when patients were off GH (B and D). Furthermore, AUC_C-peptide during the hyperinsulinemic clamp was higher with GH than after discontinuation of GH (A vs. B; P = 0.004) and with GH than with acipimox alone (A vs. D; P = 0.003). Finally, AUC_C-peptide during the hyperinsulinemic clamp was higher with GH plus acipimox than with acipimox alone (C versus D; P = 0.005) (Fig. 2; Table 2).

IGF-I levels were lower in study conditions B (no treatment) and D (acipimox alone) than when GH was continued, whereas IGF-I was uninfluenced by acipimox (Table 2).

On all 4 study days, glucagon increased significantly from the beginning of the study (0 min) to the end of the basal period (180 min). During the clamp, a small but insignificant decrement in glucagon was found. Glucagon was not different at the beginning of the study (0 min) or at the end of the clamp (360 min). However, at 180 min, glucagon was lower in study condition A than C (P = 0.014) or D (P = 0.031). Cortisol was identical on all 4 study days at 0, 180, and 360 min (Table 2).

As expected, FFA levels were higher during GH replacement than after GH was discontinued. Basal circulating FFA levels were markedly suppressed by acipimox irrespective of GH status. During the clamp, FFA levels were uniformly suppressed in all four studies (Fig. 3). Whole-blood concentrations of glycerol and 3-hydroxybutyrate were suppressed by acipimox, and the levels were highest during GH replacement alone. Insulin infusion suppressed the levels of both glycerol and 3-hydroxybutyrate in all studies. Circulating levels of lactate and alanine did not exhibit significant intra- or interstudy fluctuations (Table 3).

EE was similar in all four studies, and no significant difference was recorded between basal and insulin-stimulated EE in any study (Table 4). Discontinuation of GH induced a 14% decline in lipid oxidation, which did not reach statistical significance. An ∼50% reduction in lipid oxidation was observed after acipimox administration alone (B versus D; P = 0.008). Moreover, lipid oxidation during GH treatment was significantly reduced by acipimox (A versus C; P = 0.04). By contrast, the antilipolytic effect of acipimox was not significantly influenced by GH (C versus D; P = 0.379) (Fig. 4). During the clamp, the reduction in lipid oxidation was more pronounced during GH discontinuation (28%) and lowest during acipimox administration alone (3.5%), but these differences were not statistically significant.

Fasting plasma glucose during the basal period tended to be higher when GH was administered alone. The basal glucose R_d did not differ significantly among the four studies (P = 0.104). Basal glucose oxidation was lower with GH alone than with GH plus acipimox (P = 0.037). Moreover, acipimox alone resulted in higher basal glucose oxidation rates than in the two studies without acipimox (A and B). Nonoxidative glucose disposal was low in all four experiments. Likewise, forearm glucose uptake in the basal state was low in all studies (arteriovenous differences of glucose in mmol/l: GH alone, 0.00±.0.03; no treatment, 0.06 ± 0.02; GH plus acipimox, 0.06 ± 0.04; acipimox alone, 0.14 ± 0.08) (Table 4).

The rate of total glucose turnover (R_d) during the clamp was significantly lower during GH administration alone as compared with the three other studies (P = 0.004). The oxidative glucose R_d did not significantly differ among the studies. Nonoxidative glucose disposal was elevated during hyperinsulinemia compared with the basal state, and GH tended to suppress nonoxidative glucose disposal; however, overall no statistical significant differences were found (P = 0.08).

Insulin sensitivity, as assessed by the M value, was reduced during GH replacement alone as compared with all other studies (in mg · kg⁻¹ · min⁻¹: GH alone, 2.55 ± 0.64; no treatment, 4.01 ± 0.70; GH plus acipimox, 3.96 ± 1.34; and acipimox alone, 4.96 ± 0.91). Administration of GH plus acipimox increased insulin sensitivity to a level not significantly different from the two studies in which the patients did not receive GH (no treatment and acipimox alone). By contrast, GH did not significantly influence the effects of acipimox in the sense that the M value during GH plus acipimox did not significantly differ from the M value during acipimox alone (P = 0.19) (Fig. 5).

In all studies, the arteriovenous difference of glucose (mmol/l) over the forearm was elevated as compared with the basal state (GH alone, 0.17 ± 0.06; no treatment, 0.40 ± 0.13; GH plus acipimox, 0.36 ± 0.10; and acipimox alone, 0.83 ± 0.24). Moreover, administration of acipimox alone tended to increase glucose uptake in the forearm, and compared to administration of GH alone, this nearly reached statistical significance (P = 0.05).

Administration of supraphysiological GH doses in normal subjects induced both peripheral and hepatic insulin resistance (7,14,16). Moreover, during GH exposure, circulating levels of FFAs and lipid oxidation rates were elevated, whereas glucose oxidation rates were suppressed. Randle et al. (5) proposed the glucose fatty acid cycle in 1963, and the concept of substrate competition has been supported by numerous studies. Elevation of the circulating lipid levels causes insulin resistance (17), whereas suppression of circulating levels of FFAs, induced by pharmacological antilipolysis, increases insulin sensitivity (9). The purpose of the present study was to test the hypothesis that the short-term effect of GH on insulin sensitivity significantly depends on its lipolytic action.

Our study demonstrated that insulin resistance induced by GH administration is nearly abolished by concomitant suppression of FFAs. We studied a group of adult GH-deficient patients receiving their usual GH replacement doses, which made it possible to evaluate the effects of both constant physiological GH levels and low to nonexistent GH concentrations.

The levels of circulating FFAs were suppressed by acipimox, which is a nicotinic acid derivative known to suppress lipolysis in rat as well as human adipose tissue (18,19) by depression of intracellular cAMP levels, leading to decreased activity of a protein kinase, and thereby inhibiting the activity of the hormone-sensitive lipase (18). An intrinsic effect of acipimox on glucose metabolism has also been suggested (20); however, this effect was only assumed to increase the M value during a hyperinsulinemic clamp by 13% at insulin levels much higher than in the present study. Consequently, it seems reasonable to assume that acipimox in the present study primarily exerted its effects by reducing FFA release.

In the basal postabsorptive state, lipid oxidation rates were suppressed by acipimox, irrespective of GH status concomitantly with reciprocal changes in glucose oxidation rates. Thus the changes in lipid and glucose oxidation rates were found to be predominantly related to differences in FFAs, a finding that is in line with a previous report that lipid oxidation is regulated by FFAs and insulin levels (21). In accordance with previous studies in healthy humans (22) and in GH-deficient patients (23), we found that GH does not affect basal glucose production rates. Moreover, it has recently been reported that neither GH nor FFA levels influence basal glucose production (24). On the other hand it has been shown that pharmacological antilipolysis by acipimox in healthy subjects stimulates basal hepatic glucose production at the expense of increased protein breakdown (25). It is noteworthy, however, that the GH levels in the latter study increased to 37 μg/l at the end of the study period, which made the data difficult to interpret in a physiological context. The same concern might be raised about other studies with acipimox administration in which the rebound increase in GH secretion was not accounted for. In this regard, GH deficiency provides a unique model for studying the isolated effects of pharmacological antilipolysis. Notably, we recorded a minor increase in serum GH levels when acipimox was administered alone, which implied some residual pituitary GH secretion. Our patients did, however, fulfill the criteria of severe GH deficiency, and we did not consider this minimal GH secretion to significantly impact the overall findings.

Insulin secretion tended to be higher during GH exposure. Elevated insulin secretion could be a compensatory phenomenon related to insulin resistance; however, because C-peptide levels were also elevated during concomitant GH and acipimox administration when insulin sensitivity was restored, it could be attributed to the direct insulinotropic effect of GH (26,27). It cannot be ruled out that small differences in endogenous insulin secretion may have contributed to total insulin exposure during the clamp; on the other hand, there was a subtle rise in both insulin and C-peptide concentrations after GH administration, regardless of whether acipimox was given, meaning that any possible GH-induced stimulation of pancreatic insulin secretion was identical whether in the presence or absence of acipimox. In addition, there was no difference between circulating insulin levels during the final 30 min of the clamp.

As expected, insulin sensitivity, as assessed by the M value, was distinctly decreased by GH replacement. More importantly, the effect of GH on insulin sensitivity was markedly reduced by concomitant lowering of FFA levels, a finding that strongly supports the hypothesis that the insulin antagonistic properties of GH are causally linked to its lipolytic action. Additional data exist to support the idea that the effect of GH on glucose disposal is mediated by elevated FFA levels: 1) glucose disposal during an intravenous glucose tolerance test was inhibited by GH, but restored by administration of nicotinic acid (28); and 2) GH-induced insulin resistance during a hyperinsulinemic clamp in healthy subjects was restored by coadministration of acipimox (24). Somewhat surprisingly, Piatti et al. (24) also reported that administration of heparin together with GH and acipimox increased (normalized) serum FFA levels and lipid oxidation without impairing insulin sensitivity. The two latter studies (28,24) differed from ours by using supraphysiological GH dosages, but findings from all three studies support the concept that the insulin antagonistic effects of GH on glucose disposal can be opposed by concomitant, experimental antilipolysis. The present study design also allowed us to assess whether GH affects insulin sensitivity when lipolysis is suppressed. We observed that insulin sensitivity during acipimox administration was not significantly affected by GH. Still, the M value was slightly lower during administration of acipimox plus GH as compared with acipimox alone (P = 0.19), which suggests that GH may also cause insulin resistance through non–FFA-dependent mechanisms.

Forearm glucose uptake tended to be decreased in our study during GH exposure, and previous studies (4,29,30) have shown an acute (i.e., within minutes) decrement in forearm glucose uptake after administration of GH. Because the lipolytic action of GH is seen only after 2–3 h (4), this immediate inhibition of forearm glucose uptake is not readily explained by changes in circulating FFA levels.

Several lines of evidence support an inverse relationship between FFA levels and insulin sensitivity. First, elevation of circulating FFA levels by infusion of lipids or stimulation of intravascular lipolysis leads to insulin resistance (17,31,32). Second, circulating FFA levels are often elevated in insulin-resistant type 2 diabetic patients (33), and the level of plasma FFAs is correlated to insulin sensitivity in healthy subjects (34). Third, lowering of FFA levels by acipimox administration has led to increased insulin sensitivity in short-term studies (9,10). Moreover, intravenous infusion of FFAs seems to affect glucose metabolism in a specific time sequence: 1) glucose oxidation decreases as a consequence of increased lipid oxidation after ∼2 h (31,32), 2) glucose transport or glucose phosphorylation decreases after ∼4 h (32), and finally 3) glycogen synthesis is inhibited after ∼6 h. In fact, intracellular concentrations of acetyl-CoA in muscle increase and glycogen synthesis decreases during lipid infusion (31).

The specific mechanisms by which FFAs influence glucose metabolism and insulin sensitivity have not been resolved yet (6). In the original hypothesis, Randle et al. (5) proposed the existence of the glucose fatty acid cycle based on in vitro studies of substrate oxidation in rat heart and rat diaphragm. It was suggested that an increased delivery of FFAs to muscle tissue would lead to increased lipid oxidation and decreased glucose oxidation through increased intracellular levels of NADH/NAD and acetyl-CoA/CoA ratios together with increased levels of citrate, which would lead to allosteric inhibition of pyruvate dehydrogenase. Moreover, accumulation of citrate would inhibit phosphofructokinase, which would lead to accumulation of glucose-6-phosphate and decreased cellular glucose uptake. On the other hand, recent data conflict with the Randle hypothesis, as the FFA-induced decrement in glycogen synthesis has been shown to be preceded by a lowering of intracellular glucose-6-phosphate (32), and FFAs have been demonstrated to lower intracellular glucose concentrations (35). Taken together, these recent observations suggest that FFAs lower glucose disposal primarily by inhibition of insulin-dependent glucose transport (6).

From a clinical point of view, our study results indicate that even physiological GH replacement therapy impairs insulin sensitivity, although we did not include a comparison group of untreated healthy control subjects. Endogenous GH release is characterized by small and short-lived GH surges 2−3 h after each meal and a more pronounced secretion shortly after onset of sleep. Moreover, fasting is associated with sustained elevations in GH levels. This implies that endogenous GH release is low in the prandial and immediate postprandial period, when insulin secretion and action are stimulated. This inverse relation between GH and insulin is disturbed during constant GH exposure, such as in active acromegaly and during subcutaneous GH administration. It has previously been shown that evening as opposed to morning GH administration is associated with lower daytime insulin levels and is a closer imitation of the normal circadian pattern of circulating lipid intermediates (36). Still, in that study, evening GH administration was accompanied by prolonged, albeit small, elevations in daytime GH levels (36). We favor the explanation that the insulin antagonistic effects of GH relate to the nonphysiological occurrence of sustained GH elevations during periods of β-cell challenge. It is therefore important to strive for the lowest effective GH replacement dosage and to rigidly monitor treatment, including regular assessment of serum IGF-I levels and glucose homeostasis.

In conclusion, our study in GH-deficient subjects provides new and strong evidence that the acute effects of GH on glucose metabolism and insulin sensitivity are tightly connected to the concomitant effects on lipid metabolism. The molecular mechanisms underlying these effects remain to be investigated.

This study was financially supported by Novo Nordisk and the Århus University –Novo Nordisk Center for Research in Growth and Regeneration. J.O.J. has received consulting fees and research support from Novo Nordisk.