Central and Opposing Effects of IGF-I and IGF-Binding Protein-3 on Systemic Insulin Action

Young (3 months old, n = 6 in each group), male Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA) were used for this study. Rats were housed in individual cages and were subjected to a standard light (6:00 a.m. to 6:00 p.m.)/dark (6:00 p.m. to 6:00 a.m.) cycle. All rats were fed ad libitum using regular rat diet, which consisted of 64% carbohydrate, 30% protein, and 6% fat with a physiological fuel value of 3.3 kcal/g. The rats that received intracerebroventricular (ICV) infusions, 2–4 weeks before the in vivo study, were anesthetized by inhalation of isoflurane and ICV cannula was placed in the third ventricle for infusions of IGF-I, IGFBP-3, or IGFBP-3 mutant or artificial cerebrospinal fluid (aCSF), as previously described (26) (Fig. 1). Briefly, a 26-gauge stainless steel guide cannula (Plastics One, Roanoke, VA) was chronically implanted into the third ventricles using the following coordinates from bregma: anterior-posterior, +0.2 mm dorsal-ventral, −9.0 mm medial-lateral, 0.0 directly on the midsagittal suture, followed by a 28-gauge dummy cannula inserted to prevent clogging of the guide cannula. The implant was secured to the skull with caulk grip dental cement, and the skin was closed over the implant using wound clips. The recovery of rats from the surgical stress was monitored by daily weight, food intake, and movement.

Upon recovery of body weight, usually a week, indwelling catheters were placed in the right internal jugular vein and in the left carotid artery (26). The venous catheter was extended to the level of the right atrium, and the arterial catheter was advanced to the level of the aortic arch. Recovery was continued until body weight was within 3% of the preoperative weight (∼4–6 days). In the rats that received peripheral IGFBP-3, only the vascular catheters were placed.

Studies were performed in unrestrained rats using the insulin clamp technique, in combination with high-performance liquid chromatography–purified [3-³H]glucose and [U-¹⁴C]-lactate infusions, as described previously (27). Food was removed for ∼5 h before the in vivo protocol. All studies lasted 360 min and included a 120-min equilibration period, a 120-min basal period for assessment of the basal glucose turnover, and a 120-min hyperinsulinemic clamp period. All rats received ICV infusions (bolus followed by continuous ICV infusions of IGF-I, IGFBP-3, IGFBP-3 mutant, or aCSF, depending on the group) over the entire 6 h of the study. At the beginning of the basal period and 120 min before starting the glucose/insulin infusions, a primed-continuous infusion of high-performance liquid chromatography–purified [3-³H]glucose (20 μCi bolus, 0.2 μCi/min; NEN Life Science Products, Boston, MA) was initiated and maintained throughout the remaining 4 h of the study. In the final 2 h of the study, the rats were subjected to hyperinsulinemic clamp. The protocol followed during the insulin clamp study was similar to that previously described (27). Briefly, a primed-continuous infusion of regular insulin (3 mU · kg⁻¹ · min⁻¹) was administered, and a variable infusion of a 25% glucose solution was started and periodically adjusted to clamp the plasma glucose concentration at 7–8 mmol/l. To prevent endogenous insulin secretion and in order to control for possible effects of ICV infusions on the endocrine pancreas, somatostatin (1.5 μg · kg⁻¹ · min⁻¹) was also infused in all the groups. [U-¹⁴C] lactate (5 μCi bolus, 0.25 μCi/min) was infused during the last 10 min of the study.

A bolus of human IGF-I (0.3 μg; a gift from Tercica, South San Francisco, CA) was followed by a continuous infusion over 6 h (total dose of 1 μg, 0.06 μg · kg⁻¹ · h⁻¹). No human IGF-I was detected in the rat periphery, confirming that there was no leak from ICV infusions. (Human IGF-I does not cross-react with rat’s endogenous IGF-I in our specific enzyme-linked immunosorbent assay [ELISA]).

A bolus of nonglycosylated human IGFBP-3 (1.25 μg; generously provided by Celtrix, Mountain View, CA) was followed by a continuous infusion over 6 h (total dose of 5.25 μg, 0.26 μg · kg⁻¹ · h⁻¹). No human IGFBP-3 was detected in the rat periphery, confirming that there was no leak from ICV infusions. To further characterize the effects on glucose metabolism, we infused a nonglycosylated NLS mutant IGFBP-3 (Celtrix). This NLS mutant has two amino acid substitutions (K228E and R230G) and has been characterized (28).

To study the acute effects of an infusion of IGFBP-3, two groups of awake, unstressed, chronically catheterized Sprague-Dawley rats (∼300 g) were studied for 300 min (n = 6 in each group) under hyperinsulinemic clamp (as previously described). From 120 min, the rats received a primed continuous infusion of IGFBP-3 (0.06 mg · kg⁻¹ · h⁻¹) or saline (control) for an additional 3 h.

Plasma samples for determination of [³H]glucose and [³H]water specific activities (SAs) were obtained at 10-min intervals during the basal and clamp periods. Steady-state conditions for the plasma glucose concentration and SA were achieved within 90 min in all the studies. Plasma samples for determination of plasma IGF-I (rat and human), IGFBP-3 (human IGFBP-3), growth hormone, insulin, leptin, and free fatty acid (FFA) concentrations were collected at 30-min intervals throughout the study. All determinations were also performed on portal vein blood obtained at the end of the experiments. The total amount of blood drawn during the entire study for various assays is ∼3 ml. After separation of plasma (subsequently used for analysis), the erythrocytes were reconstituted in saline and infused back into the animal.

At the end of the clamp study, rats were killed using 100 mg pentobarbital sodium/kg body wt i.v. Epididymal, mesenteric, and perinephric fat pads (visceral fat) were dissected and weighed at the end of each experiment. The study protocol was reviewed and approved by the animal care and use committee of the Albert Einstein College of Medicine.

Plasma glucose (sample volume 10 μl) was measured by the glucose oxidase method (Glucose Analyzer II; Beckman Instruments, Palo Alto, CA). Plasma [³H]glucose radioactivity was measured in duplicates in the supernatants of Ba(OH)₂ and ZnSO₄ precipitates of plasma samples (20 μl) after evaporation to dryness to eliminate tritiated water. Uridine diphosphate glucose (UDPG) and phosphoenolpyruvate (PEP) concentrations and SAs in the liver were obtained through two sequential chromatographic separations, as previously reported (27). Plasma insulin (10 μl) was measured by radioimmunoassay using rat insulin standard for basal studies and human insulin standard for insulin clamp studies. Plasma leptin (10 μl) was assayed using the Linco leptin assay kit (Linco Research, St. Charles, MO). Plasma nonesterified fatty acid concentrations (5 μl) were determined by an enzymatic method with an automated kit according to the manufacturer’s specification (Waco Pure Chemical Industries, Osaka, Japan). Rat IGF-I (50 μl) was measured with an rIGF-I ELISA kit (DSL, Webster, TX). Human IGF-I (50 μl) was measured with an in-house kit developed by the University of California Los Angeles using a double-monoclonal antibody ELISA with a sensitivity of 0.1 ng/ml and intra- and interassay coefficients of variation of <6 and <8%, respectively, in the range 1–6 ng/ml. Human IGFBP-3 (30 μl) was measured with an ELISA kit (DSL). This kit reads both IGFBP-3 and NLS IGFBP-3. We could not demonstrate any human IGF-I, hIGFBP-3, or NLS IGFBP-3 in the rat periphery in the groups that received these ICV peptides, indicating that no leaks occurred during the infusions and that the effects observed in our studies are strictly central in nature. Serum levels of rat growth hormone were determined using an enzyme immunoassay kit (ALPCO, Windham, NH) according to the manufacturer’s instructions. The detection limit of this assay is 0.5 ng/ml, and the intra- and interassay variations are <10 and <14%, respectively.

Under steady-state conditions for plasma glucose concentrations, the rate of glucose disappearance (R_d) equals the rate of glucose appearance (R_a). The latter was calculated as the ratio of the rate of infusion of [3-³H]glucose (disintegrations per minute per minute) and the steady-state plasma [³H]glucose SA (disintegrations per minute per milligram). The rate of endogenous glucose production was calculated as the difference between R_a and the infusion rate of glucose. The rates of glycolysis were estimated as described previously (29). Glycogen synthesis was estimated by subtracting the rate of glycolysis from the R_d.

The direct contribution of plasma glucose to the hepatic glucose-6-phosphate (G-6-P) pool was calculated from the ratio of SAs of hepatic [³H]UDPG and plasma glucose after [3-³H]glucose infusion. This represents the percentage of the hepatic G-6-P pool, which is derived from plasma glucose. The indirect contribution of plasma glucose to hepatic G-6-P was derived from the ratio of SAs of hepatic [¹⁴C]UDPG and [¹⁴C]PEP × 2 after [¹⁴C]lactate infusion. This represents the percentage of the hepatic G-6-P pool, which is derived from PEP gluconeogenesis. Total glucose output (TGO) is the sum of the HGP + glucose cycling ([³H]UDPG SA/plasma [3-³H]glucose SA × TGO). Therefore, TGO = HGP/(1 − [³H]UDPG SA/plasma [3-³H]glucose SA). Gluconeogenesis was estimated from the SAs of ¹⁴C-labeled hepatic UDPG (assumed to reflect the SA of hepatic G-6-P) and hepatic PEP after the infusion of [U-¹⁴C]lactate and [3-³H]glucose (27). Therefore, gluconeogenesis = TGO × [¹⁴C]UDPG SA/[¹⁴C]PEP SA × 2. Glycogenolysis was calculated as the difference between HGP and gluconeogenesis.

Total RNA was extracted from the mediobasal wedge of rat (control, aged 3 months) hypothalamus following Clontech’s protocol. First-stranded cDNA was synthesized from 1 μg total RNA using Superscript III (Gibco, Gaithersburg, MD). Expression of IGF-I, IGF-I receptor, and IGFBP-3 was demonstrated by RT-PCR. For IGF-I, the sequences used in the reaction were: sense 5′-TCTGAGGAGGCTGGAGATGT-3′ and antisense 5′-GTTCCGATGTTTTGCAGGTT-3′, for IGF-I receptor: sense 5′-GCGTCTTCACCACTCATTCC-3′ and antisense 5′-GCGCATAAGTTCAAACAGCA-3′, and for IGFBP-3: sense 5′-GGCCCAGCAGAAATATCAAA-3′ and antisense 5′-TACCAGGGTCTCCAACAAGG-3′. The conditions for PCR were 94°C for 10 min, cycles of 94°C for 30 s, then annealing temperature of 60°C for 30 s and then 72°C for 1.5 min. The product sizes were 240 bp, 179 bp, and 194 bp for IGF-I, IGF-I R, and IGFBP-3, respectively.

Total RNA was extracted following Clontech’s protocol. First-stranded cDNA was synthesized from 1 μg total RNA using Superscript III (Gibco). Hepatic glucose-6-phosphatase (G6Pase) and PEPCK mRNA abundance were assessed by quantitative RT-PCR as described previously (30). Quantification of these peptides and their copy number were normalized for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and 18S to correct for loading irregularities. The data presented are those normalized by GAPDH.

All values shown are expressed as means ± SE. Statistical analyses were performed using ANOVA in multiple comparisons and unpaired, nonparametric Student’s t test. When the main effect was significant, a two-tailed post hoc test (Tukey’s) was applied to determine individual differences between means. A P value <0.05 was considered to be statistically significant. All statistical analyses were performed using SPSS for Windows.

Figure 2 demonstrates the gene expression of IGF-I, IGF-I receptor, and IGFBP-3 in the mediobasal segment of the hypothalamus. To examine the effect of the acute ICV administration of IGF-I (n = 6), IGFBP-3 (n = 6), and the NLS IGFBP-3 mutant (n = 6) on peripheral and hepatic insulin action, rats that received a primed constant infusion of these ICV peptides were compared with control rats (n = 6 for each group) receiving vehicle infusion. There were no differences in the body weight and average food intake among the different groups of rats compared with controls. The amount of visceral fat was comparable in all the groups at the time of killing. Plasma insulin, glucose, and FFA concentrations were similar in the rats assigned to the different experimental groups (Tables 1 and 2). Plasma growth hormone and IGF-I levels were similar between the groups at the end of the clamp study, demonstrating that ICV infusions did not alter the peripheral growth hormone and IGF-I levels (Tables 3 and 4). Consistent with our findings, the lack of an effect of acute infusion of ICV IGF-I on the peak, trough, interpeak, or mean 6-h peripheral growth hormone levels have been demonstrated previously (31).

Under basal conditions, there was no difference in the endogenous glucose production between the groups. During hyperinsulinemic clamp, similar insulin levels were achieved, and plasma glucose was clamped at 7–8 mmol/l in all the groups. Hepatic glucose production was significantly decreased in all the groups under hyperinsulinemic clamp. In the presence of ICV IGF-I, the HGP was significantly lower (Fig. 3A, Table 3) compared with controls in spite of no changes in the peripheral IGF-I levels between the groups. In contrast, in the presence of ICV IGFBP-3 (Fig. 3B, Table 4), HGP was significantly increased compared with controls. While ICV IGF-I resulted in a 50% lower HGP than controls, IGFBP-3 resulted in a 45% increase in HGP compared with controls. In the presence of ICV NLS mutant infusion, HGP during both basal and hyperinsulinemic conditions were similar to controls (Fig. 3B, Table 4).

With infusion of IGF-I or IGFBP-3 ICV, there were marked changes in HGP (Fig. 3C). We examined whether ICV IGF-I and IGFBP-3 modified the relative contributions of plasma glucose, gluconeogenesis, and glycogenolysis to the hepatic pool. A marked decrease in the contribution of glycogenolysis to HGP was demonstrated in rats receiving ICV IGF-I (P < 0.01), while gluconeogenesis showed no change. IGFBP-3 on the other hand, resulted in a significant increase in glycogenolysis (P < 0.01 compared with controls) with no changes in gluconeogenesis (0.3 ± 0.1, 0.54 ± 0.2, and 0.37 ± 0.1 mg · kg⁻¹ · min⁻¹ in aCSF, IGF-I, and IGFBP-3, respectively). There were no changes in glucose fluxes compared with controls in the NLS mutant group.

At the end of the clamp, the relative abundance of key gluconeogenic enzyme PEPCK and G6Pase mRNA in the liver was calculated by quantitative RT-PCR and compared as a ratio to the concentration of a reference gene, GAPDH. PEPCK mRNA in the liver of rats receiving ICV IGF-I or IGFBP-3 for 6 h was similar to vehicle-infused rats (2.5 ± 0.5, 3.4 ± 0.6, and 3.2 ± 0.3 in saline, IGF-I, and IGFBP-3, respectively), reflecting the lack of change in gluconeogenesis. In parallel, with the significant reduction in HGP and glycogenolysis by tracer methodology, the expression of G6Pase was significantly threefold lower in animals that received IGF-I (7.4 ± 0.5 vs. 2.6 ± 0.6 in aCSF and IGF-I, respectively, P < 0.001). In spite of an increase in HGP, there were no changes in expression levels of G6Pase in the group that received ICV IGFBP-3. This effect of ICV IGF-I and IGFBP-3 may explain the redistribution of hepatic glucose fluxes between gluconeogenesis and glycogenolysis.

The effect of increase in the circulating insulin concentrations in the presence of ICV IGF-I on the rates of tissue glucose uptake (R_d), glycolysis, and glycogen synthesis are displayed in Fig. 4A and C. All measurements were performed during the final 60 min of the clamp study, a time when steady-state conditions were achieved for plasma glucose and insulin concentrations, glucose SA, and rates of glucose infusion. ICV IGF-I tended to augment the effect of physiologic increments in the plasma insulin concentration on whole-body glucose disposal, though this increase was not statistically significant (Table 3, Fig. 4A). In contrast, IGFBP-3 significantly decreased peripheral glucose uptake (Table 4, Fig. 4B). The infusion of NLS mutant IGFBP-3 did not result in significant changes in R_d, but the values obtained were intermediate between IGFBP-3 and aCSF. Notably, in separate experiments, IGFBP-3 and NLS IGFBP-3 bound equally to 125-labeled IGF-I in Western ligand blots (data not shown), but NLS IGFBP-3 fails to internalize into cells due to its inability to bind to transferrin and importin (25,28). We next examined whether ICV IGF-I, IGFBP-3, or IGFBP-3 NLS mutant exerted any effect on the partitioning of glucose disposal into glycogen synthesis and glycolysis. Concomitant with the trend in increase in R_d, both glycolysis and glycogen synthesis was increased in IGF-I groups compared with controls. There was a significant decrease in both glycolysis and glycogen synthesis in IGFBP-3–infused animals (Fig. 4C), while the NLS mutant showed no change compared with controls.

Hyperinsulinemia during the clamp suppressed FFA levels in the control groups as expected. In the rats that received ICV IGF-I, the suppression of FFA was significantly higher (compared with controls), reflecting an overall improvement in insulin action. On the contrary, with ICV infusion of IGFBP-3, insulin-induced suppression of serum FFA levels was significantly reduced; this occurring along with a decrease in R_d and increased HGP reflects an overall decline in insulin action.

To determine if peripheral infusion of IGFBP-3 affects insulin action, we infused saline or IGFBP-3 during hyperinsulinemic clamp to two groups of rats that were matched for body weight (∼260 g), visceral fat (∼ 3.5 g), basal glucose (∼135 mg/dl or 7 mmol/l), FFA (∼0.55 mEq/l), and insulin levels (∼1.7 ng/ml). The resulting human IGFBP-3 levels during the IGFBP-3 infusion were 1,960 ± 360 ng/ml. The levels are similar to levels seen in physiological insulin resistance states such as puberty (13) and old age (32). IGFBP-3 induced an ∼17% reduction in insulin-stimulated glucose uptake and 30% higher HGP (Fig. 5A and B) compared with controls under similar hyperinsulinemic clamp conditions. We tested rat CSF for entry of peripherally infused IGFBP-3 through the blood-brain barrier and detected levels with a mean of 7 ng/ml, indicating some uptake into the central nervous system. These studies demonstrate that IGFBP-3 directly affects peripheral glucose metabolism, probably through the central nervous system.

We demonstrate that IGF-I and IGFBP-3 affect peripheral insulin action through central mechanisms. Infusion of IGF-I into the third ventricle was able to affect peripheral hepatic insulin action even in the absence of demonstrable leak of human IGF-I into the periphery or change in peripheral growth hormone levels. These studies may explain the apparent puzzle of significant metabolic effects of IGF-I at the level of the liver in the absence of substantial amounts of hepatic IGF-I receptors (9). Considering the metabolic effects of peptides such as leptin and insulin, which affect hepatic insulin action through the hypothalamus (33), and the abundance of IGF-I and its receptors in the brain, a central action of IGF-I is not surprising (34,35). Similar to leptin and insulin, IGF-I has been demonstrated to be transported across the blood-brain barrier and is present in the CSF (36). In addition, the tissue levels of IGF-I in the hypothalamic area may be higher than the levels in the CSF because of local production of IGF-I. Because a recent study (37) showed a rapid clearance of CSF IGF-I following a lateral ventricle infusion, we designed the study to achieve physiological metabolic effects through a continuous infusion. The doses of IGF-I we chose in these studies are small and are not significantly higher (in molar quantities) than the doses of ICV insulin that demonstrate a physiological effect (38). Thus, we believe that the central effects of IGF-I demonstrated here on peripheral glucose metabolism are physiologically relevant.

Insulin resistance observed during clamps with IGFBP-3 (in both peripheral and ICV infusions of IGFBP-3) provides the first conclusive evidence of an acute effect in vivo of this molecule. We demonstrate that IGFBP-3 influences overall insulin action through decreases in insulin-induced suppression of HGP and decrease in R_d and FFA. Under physiologic hyperinsulinemic clamp conditions, the effects of central IGFBP-3 appears to be on overall systemic insulin action, while the effects of central IGF-I are predominantly at the level of liver. As is often observed in other systems, it is possible that some of the in vivo effect of IGFBP-3 is secondary to IGF-I binding (and thus decreased availability of IGF-I) in addition to IGF-independent effects (39). Liver-specific IGF-I knockout (LID) mice demonstrate significant insulin resistance at the level of muscle. In LID mice, the circulating IGF-I levels are low, at 10–20% of control (40), but IGFBP-3 levels are only 50% reduced (P.C. and D.H., unpublished data) and growth hormone levels are very high. Inhibition of growth hormone action in LID mice leads to improved insulin sensitivity; however, this is also associated with a reduction in growth hormone–dependent proteins such as IGFBP-3, which can contribute toward the improved sensitivity, further demonstrating the complex role of growth hormone/IGF-I/IGFBP-3 axis in glucose metabolism (41).

IGFBP-3 mutants with altered interactions with its protein partners serve as useful tools to evaluate the mechanisms of specific effects of this pleiotropic molecule. The NLS mutant of IGFBP-3 is the only mutant that is currently available in sufficient quantities for in vivo use. The NLS mutant IGFBP-3 protein binds IGF-I normally but fails to bind to other IGFBP-3 ligands, including cell-surface proteins (preventing cell binding and internalization [28,42]), importin-β (preventing nuclear transport [25]), various extracellular matrix proteins, such as transferrin and type-1 collagen (43,44), and various soluble, intracellular, extracellular, and circulating proteins (45). We have previously characterized the mutant and showed that it fails to internalize into cells and does not mediate nuclear actions of IGFBP-3 in spite of normal IGF binding and equal inhibition of IGF-dependent effects. Here, in vivo, this mutant minimally inhibited hepatic and peripheral insulin action. The decreased activity of the NLS mutant IGFBP-3 demonstrates that IGF-I binding alone is insufficient to mediate the central effects of IGFBP-3 that we observed. Therefore, it can be concluded that the action of IGFBP-3 on the hypothalamus involves activity mediated by interfacing with additional molecules in addition to IGFs. The partial, but not significant effects, of the NLS mutant is compatible with some inhibition of local IGF action as a component of the effect of IGFBP-3 in the central nervous system. IGF-independent effects of IGFBP-3 on Akt have been described in adipocytes (19) and endothelial cells (46). Once inside the cell, IGFBP-3 can rapidly internalize into the nucleus, where it has the ability to bind RXR and modulate its signaling (21) and, as noted, in 3T3-L1 adipocytes in response to tumor necrosis factor-α (47). This binding of IGFBP-3 to RXR-α can impair peroxisome proliferator–activated receptor γ signaling (48). Peroxisome proliferator–activated receptor has been located in various regions of the brain and have been implicated in glucose metabolism in the brain (49). Therefore, we hypothesize that IGFBP-3 induced impairment of nuclear signaling or some other intracellular action of IGFBP-3 leads to insulin resistance.

It is interesting to compare the effects of central and peripheral infusions of IGFBP-3 on insulin action. It appears that IGFBP-3 antagonizes insulin action through distinct mechanisms in the central nervous system involving both IGF-dependent and -independent pathways. The effects of peripheral infusion of IGFBP-3 on glucose metabolism could also be centrally mediated, as the levels of human IGFBP-3 detected in the CSF during a peripheral infusion of this protein cannot attest to its availability near the hypothalamus, its internalization, or local production in the brain. IGFBP-3 is recognized to be expressed in brain tissue, and its levels are within the physiologically active range in the CSF (24). Furthermore, IGFBP-3 is elevated in the brains of patients with Alzheimer’s disease (26), and it is intriguing to speculate that this is involved in the well-recognized insulin resistance in this condition. The extent of the individual contribution of IGF-I–dependent versus non–IGF-I–dependent effects of a peripheral IGFBP-3 infusion can only be discerned by studying the effects of peripheral infusions with a non–IGF-I–binding mutant IGFBP-3 in vivo; however, these studies must await large-scale production of these proteins.

Thus, IGF-I and IGFBP-3 appear to have opposing effects on glucose metabolism, and their balance may play a prominent role in glucose homeostasis. As with other peripheral peptides such as insulin and leptin, much of these effects seem to be mediated through the hypothalamus. An independent effect of IGFBP-3 on insulin action is particularly significant as IGFBP-3 is about to undergo phase II clinical trials as an anticancer drug. Dissecting the differences in effects of these peptides is extremely important to develop treatment modalities without unwanted side effects.

This work was supported by Serono Clinical Scholar award from the Lawson Wilkins Pediatric Endocrine Society (to R.M.), grants from the National Institutes of Health (AG21654 and AG18381 to N.B. and K08HD042172 to P.V.), and by the core laboratories of the Albert Einstein Diabetes Research and Training Center (DK 20541).

We thank Stanislaw Gaweda for his assistance in estimating hepatic glucose fluxes.