Dipeptidyl Peptidase IV-Resistant [d-Ala²]Glucose-Dependent Insulinotropic Polypeptide (GIP) Improves Glucose Tolerance in Normal and Obese Diabetic Rats

Synthetic human GIP_1–42, [Ala¹-Tyr²]GIP_1–42, [d-Ala²]GIP_1–42, and GIP_3–42 were purchased from Hukabel (Montreal, Canada) and synthetic porcine GIP_1–42 from Bachem (Torrance, CA). Batches of GIP_1–42 (Mass+H⁺: measured, 4,984.7 g/mol; expected, 4,983.6 g/mol) and [d-Ala²]GIP_1–42 (Mass+Na⁺: measured, 4,998.6 g/mol; expected, 5,002.7 g/mol) were also synthesized (Probiodrug, Halle, Germany), using methods described in detail elsewhere (22).

Male Wistar rats were purchased from the University of British Columbia Animal Care Facility (Vancouver, Canada); male VDF Zucker rats (23) were obtained from the colony maintained in the Department of Physiology, University of British Columbia. Animals were held in group housing with free access to rat food and water, with a 12-h light:dark cycle. Animals were fasted overnight (15–18 h) before experimentation. Anesthesia, where indicated, was achieved with intraperitoneal Somnotol (65 mg/kg sodium pentobarbital; MTC Pharmaceuticals, Cambridge, Canada). Animal experiments conformed to the guidelines set forth by the University of British Columbia Committee on Animal Care and the Canadian Council on Animal Care.

Monocomponent [¹²⁵I-Tyr¹⁰]GIP_1–42 (∼300 μCi/mg) was prepared according to established protocols (12), and iodination of [d-Ala²]GIP_1–42 was based on these techniques. Briefly, 1 mCi of carrier-free iodine-125 (PerkinElmer Life Sciences, Boston, MA) was added to 1 nmol peptide dissolved in 100 μl of 0.4 mol/l phosphate buffer (pH 7.5), iodination was initiated by addition of 10 μl chloramine-T (14.2 mmol/l in 0.4 mol/l PO₄ buffer [BDH, Toronto, Canada]) for 15 s, and the reaction was quenched with 20 μl sodium metabisulfite (66.3 mmol/l in 0.4 mol/l PO₄ buffer [Fisher Scientific, Fair Lawn, NJ]). Radiolabeled peptide was separated from free iodine by gel filtration (Sephadex G-15 [Pharmacia, Uppsala, Sweden]) in 0.2 mol/l acetic acid with 0.5% BSA (RIA fraction V; Sigma Chemical, St. Louis, MO) and 2% Trasylol (aprotinin; Bayer, Etobicoke, Canada). Labeled peptides were then purified by high-performance liquid chromatography (HPLC) to yield an apparent single molecular species, as described below. Iodinated products of [d-Ala²]GIP_1–42 have not been previously characterized; five major peaks were identified on the chromatogram, and peaks 2 and 4 showed the greatest specific binding at the GIP receptor (data not shown; protocol described below). Peak 2 eluted with a similar retention time as that of [¹²⁵I-Tyr¹⁰]GIP_1–42 and appeared to be the predominant iodinated product, and thus was used in subsequent experiments.

Purified pork kidney DP IV was prepared as previously described (24) to a specific activity of 31.2 units/mg. The hydrolysis of iodinated peptides by purified DP IV was studied as previously described (12). In brief, radiolabeled peptides (750,000 cpm/150 μl) were incubated in HEPES-buffered saline (HBS) (40 mmol/l HEPES, 154 mmol/l NaCl, pH 7.6) with or without DP IV (10 mU) for 16 h. Samples were resolved by HPLC using a CH₃CN (Fisher Scientific)/trifluoroacetic acid (TFA) (0.1% vol/vol; Pierce Chemical, Rockford, IL) and H₂O/0.1% TFA solvent system at a flow rate of 1 ml/min, according to published methods (12). Iodinated peptides (∼50,000 cpm/20 μl injection) were separated by a protocol consisting of 14 min at 32% CH₃CN/0.1% TFA and a linear gradient to 38% CH₃CN/0.1% TFA over 10 min, followed by a further 5 min at 38% CH₃CN/0.1% TFA. Fractions were collected at 15-s intervals between minutes 5 and 27 of this protocol. Data were normalized to the total radioactivity recovered in all of these fractions. Between each sample injection, the C-18 μ-Bondapak HPLC column (Waters, Milford, MA) was rinsed for 20 min in 100% CH₃CN/0.1% TFA, and any residual radioactivity was removed with three consecutive injections of 300 μl DMSO (Sigma); the sample port was rinsed with 400 μl H₂O, and the column was allowed to re-equilibrate at 32% CH₃CN/0.1% TFA for 5 min before the next injection. Under these conditions, a second peak eluting earlier than intact [¹²⁵I]GIP_1–42 has been confirmed to be [¹²⁵I]GIP_3–42 by Edman degradation analysis (12).

Receptor studies were performed in a heterologous expression system: CHO-K1 cells transfected with the rat pancreatic islet GIP receptor (wtGIPR cells) (25,26). Cells were cultured in 1:1 Dulbecco’s modified Eagle’s medium and Ham’s F12 nutrient mixture (DMEM/F12) (Gibco BRL, Burlington, Canada) with 10% newborn bovine calf serum (Cansera, Rexdale, Canada), antibiotics (Gibco), and 0.8 mg/ml Geneticin (G418; Gibco). At 80–90% confluence, cells were harvested with trypsin/EDTA (Gibco) and seeded at 5 × 10⁴ cells per well into 24-well plates (Falcon; Becton Dickinson, Mississauga, Canada). Experiments were carried out 48 h later.

Wild-type GIPR cells (3–5 × 10⁵ cells per well) were washed twice with 37°C 15 mmol/l HEPES (Sigma)-buffered (pH 7.4) DMEM/F12 containing 0.1% BSA and allowed to preincubate for 1 h at 37°C. A 30-min stimulation of cells followed in the same buffer, additionally supplemented with 0.5 mmol/l isobutylmethyl xanthine (Research Biochemicals International, Natick, MA), at the peptide concentrations indicated in the figures (0–1 μmol/l, except where noted). Cells were then lysed in ice-cold ethanol (70%), cellular debris was removed by centrifugation, and cell contents were dried in vacuo (Speed Vac; Savant, Farmingdale, NY). cAMP levels were determined using a radioimmunoassay kit (Biomedical Technologies, Stoughton, MA) and expressed in terms of fmol/1,000 cells or % maximal GIP_1–42-stimulated cAMP production. Antagonism experiments were carried out according to previously published protocols (25,27,28). In these studies, cells were additionally preincubated for 15 min with varying concentrations of antagonist, followed by addition of 1 nmol/l GIP_1–42; after 30 min, cells were treated according to the above protocol.

Receptor binding studies using [¹²⁵I]GIP were completed essentially as described previously (26,28). A separate series of experiments was also carried out using [¹²⁵I]-[d-Ala²]GIP_1–42 as the radioligand, according to the same methodology. wtGIPR cells were washed twice in ice-cold HEPES-buffered DMEM/F12 with BSA (defined above) and incubated at 4°C for 12–16 h with 50,000 cpm [¹²⁵I]GIP or [¹²⁵I]-[d-Ala²]GIP in the presence of GIP or analogs (0–1 μmol/l, except where noted) in the same buffer, additionally supplemented with 1% Trasylol. After incubation, cells were washed twice with ice-cold buffer, solubilized with 0.1 mol/l NaOH (1 ml), and transferred to test tubes for counting of cell-associated radioactivity. Nonspecific binding was defined as radioactivity measured in the presence of 1 μmol/l GIP_1–42.

The insulinotropic activity of [d-Ala²]GIP_1–42 relative to native GIP was assessed in the perfused rat pancreas as previously described (28,29). In brief, overnight-fasted Wistar rats were anesthetized, and the pancreas plus associated duodenum were isolated. Perfusate consisted of a modified Krebs-Ringer buffer containing 3% dextran (Sigma) and 0.2% BSA, gassed with 95% O₂/5% CO₂ to achieve a pH of 7.4. The peptides were delivered as a linear gradient (0–1 nmol/l) over a 45-min period in the presence of 16.7 mmol/l glucose. Effluent samples were collected, and insulin content was measured by radioimmunoassay (RIA) (described below).

To examine the effectiveness of GIP or analogs in vivo, a bioassay was developed for the conscious unrestrained rat. Male Wistar rats (∼290 g) were fasted overnight and then given an oral glucose tolerance test (OGTT) (1 g glucose/kg body wt) with concurrent subcutaneous saline or peptide (8 nmol/kg) injection (500 μl). Blood samples were taken at indicated times from the tail vein using heparinized capillary tubes, and the plasma was separated by centrifugation (6,000 rpm, 12 min, 4°C) for RIA. Blood glucose was measured every 10 min using a SureStep glucose analyzer (LifeScan Canada, Burnaby, Canada). Similar experiments were subsequently carried out using age-matched 16- to 20-week-old lean (Fa/?) and obese (fa/fa) VDF Zucker rats (∼335 g and 575 g, respectively).

Insulin measurement by RIA was done using GP01 insulin antisera (1:1,000,000 final dilution) under disequilibrium conditions with rat insulin (Linco Research, St. Charles, MO) as a standard. After 24 h at 4°C, 2000 cpm of chloramine-T iodinated porcine insulin (Sigma) was added to all tubes; subsequent to another 24 h 4°C incubation, antibody-bound and free radioactive insulin were separated by centrifugation with dextran-coated charcoal. This protocol has been previously reported (28).

Measurement of GIP by RIA was performed as previously described using a COOH-terminally directed antibody (30). Plasma samples were appropriately diluted in GIP RIA buffer (5% charcoal-extracted human donor plasma, 2% Trasylol, 40 mmol/l phosphate buffer, pH 6.5) and incubated with GIP antiserum RK343F (1:300,000 final dilution) at 4°C. [¹²⁵I]GIP (5,000 cpm) was added 24 h later and, after a second 24-h incubation, antibody-bound and free radiolabeled GIP were separated by PEG-8000 precipitation (12.5% wt/vol final concentration; BDH) and centrifugation.

Data are presented as means ± SE. Receptor binding and cAMP data were analyzed using the nonlinear regression analysis program Prism (version 3.02; Graphpad, San Diego, CA). Student’s t test or ANOVA followed by the Newman-Keuls multiple comparison test was performed where appropriate. P < 0.05 was considered significant; F tests confirmed that variances were equal. Integrated glucose and insulin responses were calculated using the trapezoidal method, using the algorithm included in the Prism software package and basal levels as a baseline.

Initial studies were targeted at determining the importance of the NH₂-terminal amino acids, Tyr¹ and Ala², for GIP receptor binding and activation of adenylyl cyclase. In competitive binding studies with wtGIPR cells, removal of the NH₂-terminal dipeptide Tyr¹-Ala² from GIP resulted in a peptide (GIP_3–42) that displaced binding of [¹²⁵I]GIP, but with a significant shift to the right in the displacement curve: half-maximal inhibitory concentration (IC₅₀) for GIP_3–42 was 58.4 ± 18.8 nmol/l; for GIP_1–42, the IC₅₀ was 3.56 ± 0.81 nmol/l (Fig. 1A). Despite high affinity for the GIP receptor, synthetic GIP_3–42 was found to have no effect on cAMP production in wtGIPR cells at concentrations as high as 10 μmol/l (Fig. 1B). Reversing the order of the first two amino acids of GIP ([Ala¹-Tyr²]GIP_1–42) resulted in a molecule with a similar binding affinity to GIP_3–42 (IC₅₀ 67.0 ± 20.3 nmol/l) that also failed to stimulate cAMP production, emphasizing the importance of the integrity of the first two amino acid residues for high-affinity binding (Fig. 1). Both GIP_3–42 and [Ala¹-Tyr²]GIP_1–42 significantly inhibited cAMP production in response to 1 nmol/l GIP_1–42, in a concentration-dependent manner. cAMP production was inhibited by 45.6 ± 5.2% (GIP_3–42) and 40.5 ± 4.6% ([Ala¹-Tyr²]GIP_1–42) and by 91.0 ± 0.9% (GIP_3–42), and 90.9 ± 0.4% ([Ala¹-Tyr²]GIP_1–42) at peptide concentrations of 1 and 10 μmol/l, respectively (Fig. 2A). This antagonism appears to be a reversible competitive blockade of receptor activation, since 10 μmol/l GIP_3–42 right-shifted the half-maximal cAMP production of GIP_1–42 by ninefold when added to a concentration-response curve, but did not significantly diminish the maximal response (Fig. 2B).

For a peptide to be a suitable substrate for DP IV proteolysis, amino acid residues in the first two positions must be l-isomers and the linkage between them in the trans conformation. Therefore the l-amino acid in position 2 of GIP was substituted with the d-isomer. Resistance of peptides to hydrolysis by purified pork kidney DP IV was previously monitored by MALDI-TOF (matrix-assisted laser/desorption ionisation, time of flight analyzer) mass spectrometry; GIP_1–42 was shown to be degraded rapidly (t_1/2 = 2.68 ± 0.16 min) by DP IV, whereas P₂ substitution with d-Ala resulted in a peptide completely resistant to DP IV degradation (22). When incubated with purified porcine DP IV, [¹²⁵I]GIP_1–42 (retention time [RT] = 14.3 ± 0.3 min) was completely degraded to [¹²⁵I]GIP_3–42 (RT = 10.3 ± 0.1 min), as resolved by HPLC (Fig. 3A). Similar studies using [¹²⁵I]-[d-Ala²]GIP_1–42 (RT = 16.4 ± 0.2 min) indicated that it was not a substrate of DP IV (Fig. 3B), since no peak corresponding to [¹²⁵I]GIP_3–42 eluted.

Receptor binding and biological activity of GIP_1–42 and [d-Ala²]GIP_1–42 in vitro were not significantly different. In binding competition assays, regardless of whether the radioligand used was [¹²⁵I]GIP_1–42 or [¹²⁵I]-[d-Ala²]GIP_1–42, IC₅₀ values were equivalent (Fig. 4). [d-Ala²]GIP_1–42 showed nearly equal cAMP-stimulating potency to that of native GIP on wtGIPR cells (half-maximal stimulatory concentration [EC₅₀] values: GIP_1–42, 183 ± 18 pmol/l; [d-Ala²]GIP_1–42, 630 ± 119 pmol/l; P < 0.05) (Fig. 5A). In view of the strong resistance to DP IV degradation and the minimal change in receptor affinity and cAMP potency with [d-Ala²]GIP_1–42 (Figs. 3–5), its effect on insulin secretion in the isolated perfused rat pancreas was examined (Fig. 5B). In agreement with the cAMP data, integrated insulin responses to [d-Ala²]GIP_1–42 (0.927 ± 0.029 μmol, n = 5) were only slightly smaller than those with GIP_1–42 (1.130 ± 0.037 μmol, n = 5; P < 0.05) (Fig. 5B); when we perfused the pancreases of these animals with 16.7 mmol/l glucose alone, the integrated insulin response was 0.103 ± 0.056 μmol (n = 4) (28). Given the in vitro data, we hypothesized that [d-Ala²]GIP_1–42 may have enhanced bioactivity in vivo relative to native GIP, resulting from its DP IV resistance.

Initial experiments were performed with conscious Wistar rats (287 ± 6.5 g; fasting glycemia, 4.7 ± 0.1 mmol/l; fasting insulin, 87.6 ± 11.6 pmol/l; fasting GIP, 606 ± 82 pg/ml; n ≥ 20). Subcutaneous injection of 8 nmol/kg GIP_1–42 with a concurrent OGTT significantly reduced the glycemic profile relative to the saline control, and this was associated with increased circulating insulin levels (Fig. 6). Measurement of GIP levels by RIA indicated that this dose of GIP resulted in a 10-fold greater peak GIP level (control, 1.68 ± 0.17 ng/ml; treated, 16.6 ± 3.0 ng/ml) during the OGTT. The exogenous GIP appeared to be rapidly absorbed and to follow a similar elimination profile to that of endogenous GIP in control animals (data not shown). In contrast, the same dosage of GIP_3–42 had no effect on postprandial glycemia or insulin release (Fig. 7). Subcutaneous injection of [d-Ala²]GIP_1–42 resulted in a more pronounced reduction in the glycemic profile and an enhanced insulin time course than that of native GIP during an OGTT (Fig. 6). Notably, both GIP and d-Ala²-modified GIP appeared to exert their glucose-lowering effects by significant enhancement of the early phase of insulin release, whereas the latter peptide displayed more protracted bioactivity. Integrated glucose and insulin profiles can be found in Table 1.

Comparison of GIP and [d-Ala²]GIP was subsequently studied in the VDF Zucker animal model of type 2 diabetes (23). Age-matched obese animals (fa/fa; 576.1 ± 9.1 g) displayed significantly higher fasting glycemia than their lean (Fa/?; 335.6 ± 14.9 g) littermates (7.3 ± 0.3 mmol/l vs. 4.8 ± 0.1 mmol/l) and, similarly, a fasting hyperinsulinemia (979 ± 109 pmol/l vs. 8.5 ± 2.4 pmol/l), typical of this animal model (P < 0.05, n ≥ 20). Fasting GIP levels in these animals were not significantly different (lean, 954 ± 72 pg/ml; obese, 926 ± 110 pg/ml; n ≥ 13), whereas after an OGTT in control animals, postprandial GIP levels (the mean of samples taken at 10, 20, and 30 min) in obese rats (1,730 ± 170 pg/ml, n = 15) were significantly greater than levels in lean animals (1,150 ± 120 pg/ml, n = 19; P < 0.05). Injection of 8 nmol/kg GIP_1–42 resulted in 15.4-fold (lean) and 9.6-fold (obese) greater peak GIP levels in vivo after the OGTT, relative to peak values in saline control animals. In lean animals, GIP injection produced moderate reductions in postprandial glycemic levels (16.8% reduction compared with saline control at 40 min using fold-basal values), whereas [d-Ala²]GIP_1–42 was more potent (46.8% reduction at the same time point) (Fig. 8A). Similarly, in obese animals, comparison at the 40-min time point indicated that GIP reduced glycemia by 18.7% and [d-Ala²]GIP reduced glycemia by 41.5%, relative to the saline control (Fig. 9A). In lean VDF Zucker rats, both peptides appeared to similarly augment insulin release, with [d-Ala²]GIP resulting in more elevated insulin levels at the first time point (3.5 min) (Fig. 8B). However, in obese rats, differences in the potencies of GIP and DP IV-resistant GIP were more evident, with insulin levels remaining at near-peak values at the 60-min time point for the [d-Ala²]GIP-treated group, whereas insulin levels approached control values for the GIP-treated group after 1 h (Fig. 9B). Integrated glucose and insulin profiles for bioassay data can be found in Table 1.

Recognition of the importance of the structure of the NH₂-terminus for biological activity of peptides in the secretin-glucagon superfamily has resulted in the development of numerous analogs with reduced in vivo catabolism and increased biological activity. Substitution of l-Tyr¹ with the d-isomer in growth hormone-releasing hormone (GRH) (31–33), l-His¹ with d-His¹ in glucagon (34) and GLP-1 (35), or d-amino acids in P₂ of glucagon (27) or GLP-1 (36,37) results in peptides with increased in vivo potency. Although conformational changes in the molecules may play a role in increasing biological activity (34), a more prolonged biological half-life as a result of their resistance to enzymatic degradation is probably the more important factor. Frohman et al. (31,38) first showed the importance of DP IV in the physiological degradation of members of the secretin-glucagon superfamily, by demonstrating that GRH is metabolized to biologically inactive GRH_3–44 by DP IV both in vitro and in vivo. Additionally, it was shown that amino-terminal substituted analogs, including des-amino-Tyr¹-, d-Tyr¹-, and d-Ala²-GRH, were resistant to DP IV cleavage (31). More recently, studies on GLP-1 have shown a similar resistance to DP IV degradation with analogs containing d-amino acids in the P₂ NH₂-terminal position (36,37). The current study was targeted at developing long-acting analogs of the second important incretin, GIP. During the preparation of this article, Flatt and colleagues (39–42) reported on [Tyr¹-Glucitol]GIP, a peptide analog modified postsynthesis, displaying both DP IV resistance and improvement of glucose tolerance in a diabetic mouse model.

Experiments were first carried out to establish the minimal structural requirements for the NH₂-terminus of GIP for receptor binding and stimulation of adenylyl cyclase. Removal of the first two amino-terminal residues (GIP_3–42), or inversion of these amino acids ([Ala¹-Tyr²]GIP_1–42), resulted in peptides that displayed reduced receptor affinity in competition binding studies and that were devoid of the ability to stimulate cAMP production in wtGIPR cells at concentrations as high as 10 μmol/l (Fig. 1). This supports the early claim that GIP_3–42 isolated from porcine intestinal extracts lacked insulinotropic activity in the perfused rat pancreas (43). Both GIP_3–42 and [Ala¹-Tyr²]GIP_1–42 were antagonists of GIP_1–42-induced cAMP production in the micromolar range, inhibiting cAMP production by >90% (Fig. 2). Schmidt et al. (44) tested GIP_3–42 for antagonism of GIP_1–42 action on isolated rat islets and found that it was unable to reduce GIP-stimulated insulin release when administered in equal or 10-fold greater concentrations; Fig. 2 indicates that at least a 1,000-fold greater concentration of GIP_3–42 is necessary to reduce native GIP action on the cloned receptor. The claim that GIP_3–42 lacks insulinotropic activity is further supported by its lack of effect on glucose excursions and insulin profile when bioassayed in conscious Wistar rats (Fig. 7). In this bioassay, excess exogenous GIP_3–42, injected subcutaneously, was unable to block or even change the insulinotropic effect of endogenously released GIP_1–42 resulting from the oral glucose load; thus, it is extremely unlikely that GIP_3–42 plays an antagonistic role in vivo.

Given the sensitivity of the NH₂-terminus of GIP to inactivation by DP IV, we sought to generate peptide analogs resistant to this enzyme for use in vivo. Initial in vitro studies looked at modification or substitution of amino acids in positions 1–4 of GIP, using peptides based on the shorter 30-amino acid bioactive core of the hormone (1–30) (22). From this data, substitution with d-Ala in position 2 was shown to have the greatest potential for further development; preliminary trials found that this molecule was completely resistant to DP IV degradation for over 24 h and had minimal changes in receptor activity (22). When testing full-length [d-Ala²]GIP_1–42 in the current report, these results were corroborated and studies continued in animal models. Although DP IV resistance did little to the effectiveness of the analog in vitro (consistent with the lack of DP IV activity in CHO cells; S.A.H., J.A. Pospisilik, R.A.P., C.H.S.M., unpublished data), when tested in vivo, [d-Ala²]GIP reduced glycemic excursions in all animal models to a greater extent than native GIP. This was associated with enhanced early-phase insulin release in lean animals (Figs. 6 and 8; Table 1); in diabetic rats, where the first phase of insulin release is compromised, an augmentation of the entire insulin time course was observed (Fig. 9). The latter finding is of particular interest, as GIP’s effectiveness in type 2 diabetes and animal models of the disease has been questioned and remains controversial. Lean Zucker rats showed significant differences in glycemic profiles between GIP and [d-Ala²]GIP, whereas both peptides appeared equally insulinotropic except at the first time point (2 min) (Fig. 8). Although the importance of the first phase of insulin release is well established, these data are also consistent with either an increased insulin sensitivity in these animals, as noted by Pederson et al. (18), and/or enhanced ability of this compound to stimulate glucose uptake in peripheral tissues (42).

Although GIP_1–42 has been shown to exhibit insulinotropic activity equivalent to that of GLP-1 (45), there have been no previous reports targeted at developing long-acting analogs with therapeutic potential in type 2 diabetes until very recently, and these have been limited. The reason for the lack of such studies probably originates in the report of Nauck et al. (46), who observed that human GIP was almost devoid of insulinotropic activity in type 2 diabetic patients. It had been shown earlier that there was a reduced incretin response in type 2 diabetic patients, characterized by a reduction in the component of β-cell secretion resulting from oral glucose relative to that obtained with intravenous glucose (47). Additionally, the responsiveness of type 2 diabetic patients to exogenous porcine GIP, at concentrations resulting in physiological (48) or near-physiological (49) circulating levels, was blunted. It was also observed that porcine GIP_1–42 stimulated insulin secretion under fasting conditions (50,51), whereas normal control subjects did not respond, presumably because circulating fasting glucose levels in type 2 diabetic patients reach the required threshold for the insulinotropic activity of GIP. Therefore, there appears to be little doubt that insulin responses to exogenous GIP are reduced in type 2 diabetic patients. Nevertheless, the pancreas still retains some GIP sensitivity. The reason for the almost complete lack of response to human GIP observed in the study of Nauck et al. (46) may lie elsewhere. In their study, responses of normal controls to the human peptide were also extremely weak, unlike those described by the same group in an earlier study with GIP from a different commercial source (52), suggesting that the synthetic human peptide used in the type 2 diabetes study exhibited only weak biological activity. It has recently been shown by us that some (45), but not all (26), commercial preparations of human GIP exhibit very low biological activity, and it is therefore critical that further clinical trials with human GIP and GIP analogs are performed with peptides of established biological activity.

It is also important to define the origin of the resistance to GIP. One possibility is increased receptor desensitization/downregulation or altered signal-transduction pathways (6,53,54). Alternative explanations include antagonism of GIP_1–42 action by GIP_3–42, as suggested for GLP-1 (21), or a genetic defect resulting in reduced receptor expression (55). Antagonism by GIP_3–42 appears unlikely in view of its low binding affinity for the receptor and strength as an antagonist, with at least 1,000-fold higher concentrations being required to show a significant effect. However, some support for the latter hypothesis has come recently, with the finding that obese VDF Zucker rats have compromised GIP receptor expression, at both the mRNA and protein levels, in isolated islets of Langerhans (23). During an intraperitoneal glucose tolerance test, when GIP was infused at minimum threshold levels necessary to obtain a biological response in lean animals (4 pmol · min⁻¹ · kg⁻¹), obese animals were unable to respond to the same dose. This was linked to a defect in the ability to stimulate cAMP in isolated islets, and quantitative PCR and immunoblotting suggested a reduction in GIP receptor expression (23). The etiology of the diminished expression, however, has not been elucidated and may result from elevated postprandial GIP levels and subsequent desensitization/internalization/downregulation of the receptor. In the current study, fasting GIP levels in lean and obese VDF Zucker rats were not significantly different, confirming the result of Lynn et al. (23); however, elevated levels of GIP were detected in obese animals after an OGTT, lending support for the downregulation hypothesis.

The significance of GIP as a physiological incretin has been emphasized in studies using specific GIP antagonists in rat (56) and inhibition of DP IV activity in GLP-1 receptor knockout mice (57). With the important findings that sulfonylureas improve β-cell sensitivity to GIP (58), that smaller fragments of GIP are bioactive (28,29), and in the present report, that even in animals with compromised GIP receptor expression, supraphysiological concentrations of GIP and analogs with improved plasma stability are still capable of improving glucose tolerance, the pharmacological potential of GIP in treatment of human diabetic states is preserved. To truly recognize the potential of GIP and degradation-resistant analogs, the extent of the reduced sensitivity to GIP in diabetic patients needs to be quantitatively assessed using a wide range of peptide concentrations, as has been done for GLP-1 (59), rather than single low-dosage protocols, biased to show a lack of effectiveness. No matter what underlying cause is ultimately determined to be responsible for the reduced responsiveness to GIP in human type 2 diabetes, analogs based on the DP IV-resistant form described here may be useful in stimulating insulin secretion through the residual islet capacity to respond to this incretin.

The study was funded by grants from the Medical Research Council of Canada/Canadian Institute of Health Research (590007; R.A.P./C.H.S.M.), Canadian Diabetes Association (C.H.S.M./R.A.P.), and the Department of Science and Technology of Sachsen Anhalt (grant no. 9704/00116; H.-U.D.). S.A.H. is funded by the Medical Research Council of Canada/Canadian Institute of Health Research and the Killam Trusts.