Glucose-Dependent Insulinotropic Polypeptide Stimulates Osteopontin Expression in the Vasculature via Endothelin-1 and CREB

Glucose-dependent insulinotropic polypeptide (GIP) and glucagon-like peptide 1 (GLP-1) are the main incretin hormones secreted by the intestine after a meal to stimulate insulin secretion (1). Both hormones are rapidly degraded by the enzyme dipeptidyl peptidase IV (DPP-IV), inhibition of which is a novel approach to enhance incretin concentrations in the treatment of type 2 diabetes (2). In addition to their insulinotropic activity, a plethora of actions in other tissues have been described, including effects on the cardiovascular system (1,3). While a cardioprotective role for GLP-1 has been suggested (3), less is known about GIP in this context. Early studies of cats (4) and dogs (5) showed that GIP infusion increased blood flow into the superior mesenteric artery and portal vein, while decreasing it in the pancreatic and hepatic arteries, optimizing nutrient delivery to the liver after a meal. GIP has also been suggested to promote the redistribution of blood from the periphery to the gut after a meal (6). Vasoconstriction and vasodilation seem to be achieved by differential production of endothelin-1 (ET-1) and nitric oxide (NO), as suggested from studies using cultured endothelial cells (ECs) (7).

ET-1 is not only a potent regulator of cardiovascular homeostasis but also a stimulator of vascular smooth muscle cell (VSMC) proliferation, migration, and synthesis of extracellular matrix (8), all features of a transition from a differentiated, contractile state to a more dedifferentiated, proliferative, and synthetic phenotype. Circulating ET-1 concentrations are increased in most cardiovascular diseases (CVDs), playing a critical role in the structural and functional alterations associated with the development of diabetic vascular complications and hypertension (9). ET-1 is also elevated in atherosclerotic plaques and promotes the expression of inflammatory genes in VSMCs (8).

One emerging key player in the context of vascular disease is the matrix cytokine osteopontin (OPN), which is not only a marker of inflammation but also an active player in disease progression. Several growth factors, hormones, and vasoactive agonists, including ET-1, have been shown to increase OPN expression in the vasculature (10,11). While OPN deficiency results in reduced atherosclerotic lesions, OPN overexpression leads to enhanced lesion size (12,13). This cytokine regulates the proliferation and migration of VSMCs and ECs during vascular repair and remodeling and promotes leukocyte recruitment to the vessel wall and macrophage retention (12,14). Clinically, plasma OPN concentrations are associated with the presence and extent of coronary artery disease, independent of traditional risk factors (15), and with restenosis after balloon angioplasty (16). Hyperglycemia is another potent stimulus for the induction of OPN in the vascular wall (17), and increased plasma concentrations and arterial expression of OPN have been demonstrated in patients and mice with diabetes (17,18), suggesting a role for OPN in the development of diabetic vascular complications.

We recently demonstrated that GIP increases OPN expression in pancreatic β-cells and hence also has a proliferative and antiapoptotic role in this tissue (19). Similarly, GIP stimulates OPN expression in adipocytes, which was associated with insulin resistance (20,21). We also showed that a variant (rs10423928) in the GIP receptor (GIPR) gene is associated with impaired glucose- and GIP-stimulated insulin secretion and decreased BMI (19). Interestingly, another single nucleotide polymorphism (SNP) (rs1800437) in the GIPR gene has been associated with features of the metabolic syndrome and CVD (22). Here we explore the impact of GIP signaling in the vasculature and its potential link to OPN.

Primary human coronary artery smooth muscle cells (HCASMCs; Cascade Biologics), human microvascular ECs (HMECs; Centers for Disease Control and Prevention/Emory University), mouse aortic VSMCs and human myometrial resistance arteries (as described in ref. 23), carotid arteries from Wistar Kyoto rats, coronary arteries from healthy domestic pigs, and mouse aortas and carotid arteries were used. The following mouse strains were used: NMRI (Taconic Europe), FVBN nuclear factor of activated T-cells (NFAT) luciferase (NFAT-luc) transgenic mice (24), NFATc3^−/− and control NFATc3^+/+ littermates (25), and Akita^+/− LDLr^−/− (B6.Cg-Ins2^AkitaLDLr^tm1Her/J) and control LDLr^−/− littermates (The Jackson Laboratory).

Plasma from patients with a confirmed diagnosis of critical limb ischemia and healthy control subjects was analyzed for OPN, ET-1, and GIP concentrations. Human carotid plaques and plasma were collected during carotid endarterectomies. Clinical materials have been described in refs. 26 and 27. Characteristics of the individuals included are described in Supplementary Tables 1 and 2. Cross-sectional 1-mm fragments from the most stenotic region of the carotid plaques were taken for histology and adjacent fragments for RNA isolation. The remaining portion of the plaques was homogenized for protein and cytokine analyses (27). GIP concentrations were measured in plasma from patients with CVD or type 2 diabetes and healthy control subjects. The effect of GIP on plasma OPN concentrations was explored in individuals subjected to hyperglycemic clamp with infusion of GIP, performed as previously described (28). All participants gave informed consent. The study protocols conformed to the Declaration of Helsinki and were approved by the local human ethics committee. Experiments involving animals were approved by the Animal Ethical Committee in Lund/Malmö. More detailed protocols and description of the cohorts investigated are available in the Supplementary Data.

GIPR and OPN were detected in HCASMCs, HMECs, and/or arterial sections, as previously described (17), using primary antibodies for GIPR and OPN. von Willebrand factor and α-smooth muscle actin were used to identify endothelial and smooth muscle cells, respectively. Expression was detected and quantified in sections and cells using a Zeiss LSM 5 Pascal laser scanning confocal microscope.

OPN, GIPR, phospho-CREB, and total CREB were detected in cells, intact aortas, and pancreas homogenates, as previously described (17), using β-actin as the loading control. Band intensity was measured using Quantity One 1-D Analysis software.

Aortas from NFAT-luc transgenic mice (24) were stimulated ex vivo, and luciferase activity was measured as previously described (17). Optical density was normalized to protein content.

Cell death was measured by quantification of lactate dehydrogenase activity in the culture medium using a Cytotoxicity Detection Kit (Roche Applied Science) according to the manufacturer’s instructions.

Plasma OPN (Human Osteopontin Quantikine ELISA kit; R&D Systems, Abingdon, U.K.), plasma and culture medium ET-1 (Human Endothelin-1 QuantiGlo ELISA Kit; R&D Systems), and serum GIP (Human GIP [total] ELISA kit, no. EZHGIP-54K; Millipore, St. Charles, MO) were analyzed according to the manufacturers’ instructions. Direct cAMP ELISA (Enzo Life Sciences) was used to measure cAMP in cells.

DNA synthesis was measured by thymidine incorporation. Cells were pulsed with 1 mCi [methyl-3H]thymidine (Amersham Biosciences, Uppsala, Sweden) during the last 24 h.

Cytokine concentrations were measured in aliquots of human carotid plaque homogenates and plasma (Milliplex Human Cytokine/Chemokine Immunoassay) and analyzed with Luminex 100 IS 2.3, as previously described (29).

Immunohistochemistry was performed on sections of carotid atherosclerotic plaques, as previously described (29), using primary antibodies detecting GIPR, OPN, CD68, and α-actin. Briefly, Oil Red O was used to detect lipids and Masson’s trichrome with Ponceau-acid fuchsin and aniline blue was used to detect plaque collagen content. Slides were scanned with the ScanScope Console version 8.2 and photographed with Aperio ImageScope version 8.0. The percentage area of the plaque constituted by the different components was quantified blindly using Biopix iQ 2.1.8.

Total RNA was isolated from cultured cells, mouse arteries, and human carotid plaques as previously described (29). It then was reverse transcribed, and thereafter GIPR and OPN mRNA levels were quantified by real-time PCR using TaqMan Gene Expression assays and normalized to the expression of housekeeping genes.

Associations of the GIPR rs10423928 SNP with cardiovascular events were explored in the following studies: Prevalence, Prediction and Prevention of Diabetes-Botnia (PPP-Botnia [30]); the Malmö Diet and Cancer Study (31); Metabolic Syndrome in Men (32); Scania Diabetes Registry (33); Diabetes Genetics Initiative (30); Steno type 1 and type 2 diabetes studies (34,35); Malmö Preventive Project (36); and the Finnish Diabetic Nephropathy study (37). Genotyping of rs10423928 was performed as previously described (19). A detailed description of the cohorts investigated is available in the Supplementary Data.

Results of in vitro studies are expressed as means ± SEM, and GraphPad Prism 5.0 was used for the analyses. Significance was determined using the two-tailed Student t test or one-way ANOVA followed by Bonferroni post-tests for normally distributed variables, and Mann-Whitney U or Kruskal-Wallis tests for non–normally distributed variables. SPSS version 17.0 (SPSS Inc., Chicago, IL) was used to analyze nonparametric, bivariate correlations in human carotid plaques. Associations between genotype and CVD outcomes were investigated using logistic regressions, and associations between genotype and blood pressure variables were tested using linear regression. Fixed-effect meta-analyses were performed with the metan command in STATA/SE version 12.1 (StataCorp LP; College Station, TX). Non–normally distributed variables were logarithmically transformed before analysis.

Using confocal immunofluorescence microscopy, we showed that in normal healthy mouse aorta, GIPR is predominantly expressed in the endothelium and to a small extent in smooth muscle cells of the media (Fig. 1A–C). This was also true in intact arteries from different vascular beds and species (Supplementary Fig. 1A). Using RT-PCR, GIPR mRNA was detected by two different primer pairs in intact mouse aorta (Fig. 1D). Western blotting revealed a band at the expected molecular weight (∼65 kDa) in ECs (Fig. 1E, left panel) and in pancreas homogenate from GIPR-competent mice but not from GIPR knockout mice (Fig. 1E, right panel).

Culture of intact mouse aortas with various concentrations of GIP for 3 days resulted in upregulation of OPN in the media of the arteries, as determined by confocal immunofluorescence microscopy (Fig. 2A and B) and Western blotting (Fig. 2C), with a significant effect at physiological nanomolar concentrations. Even though a tendency toward increased OPN expression was observed after 24 h of GIP stimulation (Supplementary Fig. 2A), the effects became significant only after 2 (Fig. 3C and D) and 3 days (Fig. 2).

In agreement with previous studies showing that GIP stimulates ET-1 release from cultured ECs (38), stimulation of intact mouse aortas with GIP resulted in dose-dependent ET-1 release into the culture media (Fig. 3A). ET-1 stimulation of intact mouse aorta in turn resulted in increased OPN protein (Fig. 3B). Culture of mouse aorta with GIP in the presence of the ET-1 receptor blockers BQ788 and BQ123 completely abolished the induction of OPN (Figs. 3C–E), whereas the blockers alone had no significant effect on OPN expression (Fig. 3E). To elucidate better the contribution of ECs and VSMCs to the GIP responses observed in intact arteries, experiments using isolated cells were performed (Fig. 4, upper panels A–C for ECs; panel D for intact aorta, lower panels E–H for VSMCs). Data show that GIP stimulates ET-1 release from ECs (Fig. 4A), but not from VSMCs (Fig. 4E). Further, ET-1 dose-dependently increases OPN expression in VSMCs (Fig. 4F) but has no effect on OPN in ECs (Supplementary Fig. 3A). No direct effects of GIP on OPN expression in either ECs or VSMCs were detected (Supplementary Fig. 3B and C), suggesting that both cell types must be present for arterial GIP-induced OPN expression to occur. Figure 4 also includes an illustration summarizing the proposed mechanism underlying GIP-induced OPN expression in intact arteries.

We recently reported that the calcium-dependent transcription factor NFAT regulates the expression of arterial OPN in response to hyperglycemia (17). Moreover, we recently demonstrated that GIP-induced OPN expression in adipocytes is mediated by NFAT (20). Therefore, we speculated whether the effect of GIP on arterial OPN could be mediated via NFAT activation. We found that stimulation of aortas from NFAT-luc reporter mice for 12 h with various concentrations of GIP had no effect on NFAT-dependent transcriptional activity (Supplementary Fig. 4A). Also, incubation of aortas with the NFAT inhibitor A-285222 (17), or lack of NFATc3 protein in aortas from NFATc3-deficient mice, did not prevent GIP-induced OPN expression (Supplementary Fig. 4B and C), ruling out a role for NFAT in GIP-induced OPN expression in intact arteries.

In VSMCs, several transcription factors in addition to NFAT—including CREB, NF-κB, AP-1, and upstream stimulatory factors 1 and 2—have been shown to participate in the regulation of OPN expression (17). In both β-INS-1 cells and adipocytes, GIP stimulation was shown to increase CREB phosphorylation (39,40). Here we found that stimulation with GIP increases CREB phosphorylation in ECs but not in VSMCs, as assessed by Western blotting (Fig. 4B and Supplementary Fig. 3D). Moreover, GIP-induced ET-1 release from ECs was prevented by the small-molecule CREB antagonist KG-501 (2-naphthol-AS-E-phosphate; Fig. 4C). CREB phosphorylation was also increased in VSMCs in response to ET-1 stimulation (Fig. 4G), and ET-1-induced OPN expression was prevented by KG-501 (Fig. 4H), whereas the blocker alone had no effect (Supplementary Fig. 3E). No toxic effects of KG-501 were observed at the concentrations used here (≤10 µmol/L), as assessed by measurements of lactate dehydrogenase activity in ECs and VSMCs (Supplementary Fig. 3F and G). Taken together, these data suggest that CREB plays a role at two levels, being involved in 1) the regulation of GIP-stimulated ET-1 production in ECs and 2) the regulation of ET-1-stimulated OPN expression in VSMCs. This is in agreement with previous studies showing involvement of CREB in the regulation of both ET-1 (41) and OPN (42,43). The involvement of CREB at these two levels may explain the dramatic effect of KG-501 on OPN expression that is observed in experiments using intact arteries when both cell types are present (Fig. 4D).

In line with previous work demonstrating a lack of effect of GIP on cAMP concentrations in ECs isolated from a hepatic artery or portal vein (7), here we found no effect of GIP on cAMP concentrations in cultured ECs at a GIP dose that effectively increases ET-1 levels (Supplementary Fig. 5A). Moreover, GIP was able to induce ET-1 release, even in the presence of the protein kinase A (PKA) inhibitors Rp-cAMPS and H-89 (Supplementary Fig. 5B). Thus, data suggest that GIP does not engage the traditional cAMP-PKA pathway in ECs. Intriguingly, the two PKA inhibitors had opposite effects on basal ET-1 release (Supplementary Fig. 5B): While Rp-cAMPS increased ET-1 release, suggesting an inhibitory effect of constitutively active PKA under basal nonstimulated conditions, H-89 reduced ET-1 release (Supplementary Fig. 5B). A relatively large number of PKA-independent effects described for H-89, including the inhibition of at least 8 other kinases, could account for the discrepancies between the effects of Rp-cAMPS and H-89 (44).

Having established a link between GIP, ET-1, and OPN in the vasculature, we next examined the level of these in individuals suffering from vascular disease. Significantly higher plasma ET-1 and OPN concentrations were measured in patients with critical limb ischemia when compared with control individuals (Fig. 5A and B). Several clinical characteristics of the patients with critical limb ischemia (e.g., age, smoking) could contribute to these increased concentrations of ET-1 (Supplementary Table 1). A positive correlation was found between plasma ET-1 and OPN (r = 0.424; P < 0.0001; Fig. 5D). Nonfasted plasma concentrations of GIP were not different between ischemic patients and control subjects (Fig. 5C), and they were consistent with previously reported concentrations (45).

Expression of GIPR and OPN were demonstrated in human carotid endarterectomy sections by immunohistochemistry (Fig. 5H and Supplementary Fig. 6). Significantly higher GIPR mRNA levels were found in plaques from patients with symptoms (stroke, transient ischemic attacks, amaurosis fugax) when compared with those from patients without symptoms (P = 0.0177; Fig. 5E). Plaque OPN mRNA and plasma OPN concentrations also were higher in symptomatic patients compared with asymptomatic patients (P = 0.0022 and P = 0.044, respectively; Fig. 5F and G), and there was a significant positive correlation between plaque GIPR and OPN mRNA levels (r = 0.566; P < 0.0001; Fig. 5I). Moreover, both GIPR and OPN mRNA levels correlated with the number of clinical events (Table 1). As opposed to what we found in patients with critical limb ischemia when blood samples were collected under nonfasting conditions, fasting GIP concentrations were significantly higher in individuals with a history of CVD (myocardial infarction or stroke) when compared with individuals with no history of CVD in the PPP-Botnia study (P = 0.002; Fig. 5J).

Plaque histology revealed that GIPR and OPN mRNA were positively correlated to the extent of lipid accumulation (Oil Red O staining), macrophage infiltration (CD68 staining), and elastin contents and negatively correlated to α-actin contents (Table 1). Positive correlations between GIPR and OPN mRNA levels and plaque interleukin (IL)-10, IL-1β, IL-6, MCP-1, macrophage inflammatory protein-1β, and RANTES were found. Also, GIPR mRNA was positively correlated to plaque platelet-derived growth factor AB/BB, whereas OPN mRNA correlated with plaque eotaxin, fractalkine, IL-12p70, and vascular endothelial growth factor. A complete list of the parameters analyzed, including the above-mentioned histological and plaque cytokines, is shown in Table 1. Taken together, these data provide evidence of a link between the expression of GIPR and OPN and parameters that are characteristic of more unstable and inflammatory plaques.

The elevated GIPR mRNA expression in plaques from symptomatic patients suggested that plasticity of GIPR expression is of potential relevance for GIP signaling in the context of vascular disease. Considering also the positive correlation observed between the expression of GIPR and OPN mRNA in the plaque (Fig. 5I), we wanted to explore what could drive GIPR expression in vascular cells. GIPR was detected primarily in the endothelium of freshly dissected healthy arteries (Fig. 1A and C), but also outside the endothelium in atherosclerotic vessels (Fig. 5H). Interestingly, GIPR protein was detected in arterial smooth muscle cells when cultured under growth-stimulating conditions (Fig. 6A), which is known to result in a proliferative VSMC phenotype, as shown by numerous previous publications (reviewed in ref. 46) and confirmed here by significantly increased thymidine incorporation (Fig. 6B, right panel). It is widely accepted that while VSMCs in the healthy vessel wall are contractile, dispersed cells in culture rapidly modulate from a contractile differentiated phenotype to a synthetic phenotype (47). This phenotypic switch has been shown to take place and contribute to vascular disease states (e.g., atherosclerosis, hypertensive microvessels, restenosis) (48). A time-dependent decrease in GIPR expression was found instead when smooth muscle cells were cultured in differentiating medium (Fig. 6B, left panel). Corresponding experiments also were performed using ECs, showing upregulation of GIPR mRNA when ECs are stressed by the removal of serum from the culture media (Fig. 6C). Interestingly, dramatic upregulation of GIPR mRNA was observed when intact aorta was cultured (Fig. 6D), a procedure known to result in rapid phenotypic switch caused by the loss of tensile stress (49,50).

GIPR expression also was modulated by high glucose and insulin. As shown in Fig. 6E, stimulation of arterial smooth muscle cells with 20 mmol/L glucose increased GIPR expression, and this was prevented by the addition of insulin to the culture medium. By contrast, glucose and insulin had no effect on GIPR mRNA expression in ECs (Fig. 6F). Along these lines, a trend toward elevated GIPR expression was found in carotid arteries from Akita^+/−LDLr^−/− mice when compared with nondiabetic LDLr^−/− controls (P = 0.0503; Fig. 6G). In these mice, GIPR mRNA expression correlated significantly to blood glucose concentrations (r = 0.470; P = 0.027; n = 22). Fasting GIP concentrations were significantly higher in patients with diabetes than in normal glucose-tolerant individuals (57.3 ± 51.2 vs. 37.2 ± 25.6 pg/mL; P = 4.08e-11; n = 310 vs. 4009 from the PPP-Botnia study; Fig. 6H).

Combining data from several studies, we analyzed whether the SNP rs10423928 in the GIPR gene would influence the risk of vascular disease. Patients with type 2 diabetes who carry the A allele of this SNP had an increased risk of stroke (odds ratio 1.22; P_meta = 0.00799; Fig. 7A and Supplementary Tables 3 and 4), but this association was not observed in patients without diabetes or in patients with type 1 diabetes (Fig. 7B and Supplementary Tables 3, 4, and 7). Genotype had no effect on the risk of myocardial infarction or of retinopathy, nor did it affect blood pressure (Supplementary Tables 3 and 5–8).

To explore whether GIP stimulation could induce OPN in vivo, we performed GIP infusions in 47 healthy individuals and measured the concentrations of OPN in plasma before and 105 min after the GIP infusion. As shown in Fig. 7C, GIP infusion increased plasma OPN concentrations in a genotype-dependent fashion, since only carriers of TA/AA responded with increased OPN, whereas carriers of TT did not.

The impact of the SNP rs10423928 in the GIPR gene also was examined in the patients undergoing carotid endarterectomy. Significant associations were observed between mRNA levels of GIPR or OPN and the number of events in TA/AA carriers, but not in TT genotype carriers (Supplementary Table 9). Other parameters showing an association with GIPR and/or OPN mRNA in TA/AA genotype carriers included plaque CD68, eotaxin, IL-10, and vascular endothelial growth factor, as well as plasma IL-1β (Supplementary Table 9). Also of interest, we found a significant association between GIPR mRNA and HbA_1c in TA/AA carriers only.

The major findings in this study were that 1) GIP stimulation increased the expression of OPN in mouse native arteries ex vivo by a mechanism involving the release of ET-1 and activation of CREB; 2) infusion of GIP increased plasma concentrations of OPN, and plasma ET-1 and OPN concentrations were positively correlated in patients with critical limb ischemia; 3) fasting GIP concentrations were significantly higher in individuals with a history of CVD (myocardial infarction or stroke) when compared with control subjects; 4) patients with symptoms (stroke, transient ischemic attacks, amaurosis fugax) exhibited higher plaque GIPR and OPN mRNA levels and higher plasma OPN than asymptomatic patients, and mRNA expression levels associated with parameters that are characteristic of more unstable and inflammatory plaques; 5) plasticity in GIPR expression was observed in vascular cells, with increased expression when cells were cultured under conditions leading to a more “vascular disease–like” phenotype, or upon changes in glucose and insulin that mimic the diabetic phenotype, or in arteries from diabetic mice; and 6) a common variant (rs10423928) in the GIPR gene associated with increased plasma OPN after GIP infusions and with increased risk of stroke in patients with type 2 diabetes.

Here we extend our previous findings that GIP can stimulate OPN expression in pancreatic islets (19) and adipose tissue (20) to the vascular system. Plenty of evidence supports a role for OPN in the initiation and progression of vascular disease and diabetic vascular complications, which has led to the view that OPN could serve as a potential biomarker for CVD (51). In human atherosclerotic plaques, OPN mRNA correlates with the stage of the disease (52), and OPN protein expression in carotid lesions has been ascribed prognostic value for future cardiovascular events (53). Here we found significant upregulation of OPN expression in intact arteries stimulated with physiological GIP concentrations that are comparable to concentrations reached after a mixed meal (45). The in vivo half-life of GIP in serum is approximately 3–5 min because of rapid degradation by DPP-IV (54). In our in vitro studies, some endothelial DPP-IV activity can be predicted since arteries were cultured intact; therefore, GIP was supplemented every 24 h to mimic the concentrations observed in humans (55). The functional outcome of the effect of GIP on OPN expression seems to be very different depending on the tissue. While OPN seems to be protective of pancreatic β-cells by increasing cell proliferation and reducing cytokine-induced apoptosis (19), it promotes lipogenesis, inflammation, and insulin resistance in adipose tissue (20). Therefore, the inhibition of GIP-induced OPN expression might be desirable in vessels and fat but not in islets.

GIP-induced OPN expression in intact vessels was dependent on local release of ET-1 since it was inhibited by blockers of ET-1 receptors. This is in line with previous work showing dose-dependent GIP-induced ET-1 release from cultured human ECs (38). The effect was GIPR specific and limited to cells from certain vascular beds, which was thought to be due to the differential expression of GIPR splice variants (7). ET-1 is thought to bind to receptors close to the site of release, acting in a paracrine or autocrine fashion (56). To our knowledge, no dynamic ET-1 biosensor has yet been developed, so a limitation of this study is that we could not determine the concentrations of ET-1 that vascular cells face in situ upon stimulation with GIP. Nevertheless, GIP doses capable of inducing OPN in intact arteries (i.e., 0.1 nmol/L; Fig. 2C) resulted in a ∼2.5 pg/mL change in ET-1 measured in the culture media, which would equate to ∼0.1 nmol/L ET-1 but presumably higher concentrations in the vicinity of the site of release. Long-lasting activation of the GIP–ET-1 axis could potentially lead to vasoconstriction, increased mitogenic and proinflammatory status, formation of free radicals, and platelet activation—all of which are known actions of ET-1 at nanomolar concentrations and are determinants of CVD.

Under normal physiological conditions, the effects of GIP on the vascular wall may be limited by readily available degrading enzymes (DPP-IV and neutral endopeptidases), by the sparse and restricted expression of GIPR to the endothelium (Fig. 1A), and by the many well-functioning systems of damage control (i.e., endothelium-derived relaxing factors). Under pathological conditions as well as during inhibition of degrading enzymes, however, the scenario may be different. Our in vitro data demonstrate an effect of GIP on vascular OPN; hence, elevated circulating GIP concentrations in vivo, reported here in patients with CVD and in patients with type 2 diabetes, may be anticipated to exacerbate this effect. GIP concentrations have been shown to be increased, decreased, or unaffected in adults with type 2 diabetes or impaired glucose tolerance when compared with normoglycemic individuals. Many of the early discrepancies can be explained by differences in the assays used (57) but also by indistinctly referring to incremental (i.e., secreted after a standard oral glucose tolerance test or meal test), nonfasting unstimulated, and fasting GIP concentrations. In a recent meta-analysis of 22 studies including 688 patients, the authors concluded that patients with type 2 diabetes (n = 363), in general, exhibit normal GIP secretion in response to oral glucose tolerance tests or meal tests but have elevated fasting plasma GIP concentrations compared with healthy control subjects (58). The latter is in agreement with our data (Fig. 6H). Altogether, this underscores the potential relevance of monitoring fasting—and not only secreted—GIP concentrations. Along these lines, a three times higher serum GIP concentration was reported in patients with the metabolic syndrome when compared with patients with premetabolic syndrome (19.0 ± 45.7 vs. 6.5 ± 10.2 pg/mL; P = 0.034) (59).

Changes in vascular GIPR expression, such as those reported in this study, may also contribute to the aggravation of GIP effects. Under pathological conditions, expression of GIPR was not restricted to the inner lining of the vessels; it also was detected in atherosclerotic lesions and in the media of the arteries (Fig. 5H), with higher mRNA expression in samples from symptomatic patients. It is difficult to speculate about the functional impact of this increased mRNA expression, and differences between symptomatic and asymptomatic patients may seem small, yet they were discernible despite both groups of patients having severe atherosclerosis with a degree of stenosis >80%. Carotid endarterectomies from symptomatic patients show increased levels of inflammatory markers and increased macrophages and lipid infiltration (60) as well as increased OPN expression (53,61) when compared with plaques from asymptomatic patients. Our data show that GIPR and OPN mRNA levels in plaque were positively correlated to each other and associated with parameters that are characteristic of a more unstable and inflammatory plaque. This is in line with the in vitro data, which clearly demonstrate increased GIPR expression under conditions leading to EC stress and de-differentiation of smooth muscle cells toward a more proliferative state or when cultured in the presence of large amounts of glucose, whereas insulin prevented the effects of glucose. GIPR plasticity has been previously demonstrated in other tissues. Higher GIPR (and GIP) expression was reported in the retinas of streptozotocin-induced diabetic rats (62). In pancreatic islets, on the other hand, GIPR expression is downregulated by hyperglycemia in rats and humans (19,63,64). In type 2 diabetes, a considerable loss of GIP efficacy has been demonstrated (reviewed by Meier et al. [65]), and this refers to the loss of incretin activity or pancreatic effects of GIP on insulin secretion. The exact mechanisms behind this reduced insulinotropic effect are not completely clear but are inherent to the ability of pancreatic β-cells to respond to GIP. Our data do not support a loss of GIP efficacy in arteries of patients with type 2 diabetes, highlighting the need for a distinction between pancreatic and extrapancreatic effects of GIP.

Finally, because the SNP rs10423928 in the GIPR gene had been associated with increased glycemia, we explored whether this variant in the GIPR gene would influence risk of vascular complications in patients with diabetes. Patients with type 2 diabetes carrying the less common A allele had an increased risk of stroke, but we did not observe any increased risk of myocardial infarction, nor of retinopathy, in patients with type 1 and type 2 diabetes. There was a clear effect of this genotype on the OPN response to GIP stimulation, which was restricted to carriers of AA/AT genotypes. Only these carriers showed significant correlations between plaque GIPR and OPN expression and the number of clinical events, further supporting a functional role of the SNP. Given the stronger stimulatory effect of GIP on OPN in AA/AT carriers, one could envision a stronger stimulatory effect on ET-1 and thereby increased blood pressure in these carriers, but we did not detect significant effects on blood or pulse pressure. One should keep in mind that blood pressure measurements were not sufficiently standardized in these studies.

At this point, we do not know how the variant influences the molecular function of GIPR in the vasculature—whether it influences expression of a specific isoform or whether it is associated with a gain or reduction of function. Based on information from the 1000 Genomes Project, there are five SNPs in the GIPR gene that are in perfect linkage disequilibrium (D′ = 1) with the rs10423928 SNP. Four of these five SNPs are intronic, but one (rs1800437) results in nonsynonymous coding, resulting in a residue substitution (E[Glu] → Q[Gln]). Using PolyPhen-2, a tool to predict the possible impact of an amino acid substitution on the structure and function of a human protein (http://genetics.bwh.harvard.edu/pph2/), this amino acid substitution is predicted to be “probably damaging,” with a score of 1.000. This variant, also designated E354Q, was recently shown to cause GIP-induced desensitization by increasing internalization of the GIPR (66). However, based on the impact of the variant studied here (rs10423928) on OPN concentrations after GIP infusion, it seems to be associated with a gain of function, but final proof will require additional studies to elucidate the effects of GIP on different human vessels in relation to genotype. Several splice isoforms with potentially divergent functions have been reported in cultured ECs (6), and we have observed a large number of splice isoforms in human adipose tissue, some of which seem to be regulated by SNPs in the GIPR gene (21). Additional support for a vascular role for GIP comes from the Coronary ARtery DIsease Genome wide Replication and Meta-analysis (CARDIoGRAM) Consortium, where a variant in the GIP gene was associated with myocardial infarction (67).

Adding complexity to the vascular effects of GIP, recent studies of dyslipidemic apolipoprotein E knockout mice suggested antiatherogenic effects of GIP or DPP-IV inhibition, apparently via decreased CD36 expression in macrophages and decreased foam cell formation (68–70). In this mouse model, treatment with GIP significantly reduced plasma nonesterified fatty acid concentrations, which could in part explain the reduced CD36 and macrophage foam cell formation (71). Several other discrepancies were observed when compared with studies using other experimental animals or with clinical data, such as a decreased body weight and remarkable reduction in non-HDL cholesterol. Apolipoprotein E knockout mice have high levels of VLDL and contain apolipoprotein B-48 in their lipoproteins, not fully recapitulating human lipoproteins. These marked differences between mice and humans emphasize the need to study GIP effects in humans.

Despite promising results in a recently published meta-analysis of DPP-IV inhibitors and cardiovascular risk (72), the short observation period of the trials included here (mean follow-up of 44.1 weeks), the extremely low cardiovascular event rate in the DPP-IV inhibitor arms, and the fact that studies were not controlled for the use of cardioprotective drugs question the value of the results. Based on lessons from the UK Prospective Diabetes Study, longer follow-up periods (i.e., 10 years) may be required to demonstrate potential long-term risks or benefits of therapy. Data from the more recent and superiorly designed studies Saxagliptin Assessment of Vascular Outcomes Recorded in Patients with Diabetes Mellitus (SAVOR) (73) and EXamination of cArdiovascular outcoMes with alogliptIN versus standard of carE in patients with type 2 diabetes mellitus and acute coronary syndrome (EXAMINE) (74) showed neutral effects of saxagliptin and alogliptin, respectively, on a composite of death from cardiovascular causes, nonfatal myocardial infarction, or nonfatal stroke. Similar neutral results for sitagliptin were also reported more recently (75). All these results were reported in spite of a predicted cardioprotective benefit (based on GLP-1-related experimental data and pooled data from phase IIb/III studies of saxagliptin) (76). Also, data from two independent clinical trials recently demonstrated that DPP-IV inhibition attenuated endothelial function (assessed by measurements of flow-mediated dilation) in patients with type 2 diabetes (77). These findings may not be definitive but merit further investigation.

In conclusion, we demonstrate an unprecedented link between the incretin hormone GIP and the inflammatory cytokine OPN that is known to promote atherosclerotic disease in humans, an effect that seems partially influenced by variants in the GIPR gene (Supplementary Fig. 7). In conditions with increased GIP concentrations or GIPR expression, these untoward extrapancreatic effects of GIP should be taken into account. This study also highlights the need for hard end points from trials designed to evaluate long-term CVD benefits and side effects of therapies using DPP-IV inhibitors. Considering the results presented here, randomization or closer analysis of already collected data based on GIPR genotype may add an important dimension to future studies. In the context of DPP-IV therapy, it is also possible that the effects of GIP may be counteracted by the beneficial effects of GLP-1 described to date, which have been shown in numerous publications to increase NO availability in a wide range of vascular beds (recently reviewed in ref. 78) and to inhibit ET-1 production (79). GIP and GLP-1 may behave as the “yin and yang” for systemic ET-1 and NO production. Simultaneous infusion of GIP and GLP-1 should be tested to determine whether one or the other incretin will dominate, and whether the dominance would be vascular bed specific or affected by age, given that the balance between vasoconstriction/vasodilation is normally shifted with age due to reduced NO bioavailability and increased oxidative stress.

Acknowledgments. The authors thank Drs. D. Drucker and Y. Seino from the Department of Medicine, University of Toronto, Toronto, Canada, and Kansai Electric Power Hospital, Osaka, Japan, respectively, for provision of tissues from GIPR knockout mice for use in antibody characterization. The authors also thank Drs. Eva Bengtsson and Daniel Engelbertsen for sharing tissue from Ins2^+/Akita:LDLr^−/− mice, as well as Ana Persson, Marie M. N. Nilsson, and Lena Sundius, all from the Department of Clinical Sciences, Lund University, Malmö, Sweden, for assistance with the carotid plaque experiments.

Funding. A-285222 was kindly provided by Abbott Laboratories (Abbott Park, IL). This study was supported by the Swedish Heart and Lung Foundation (HLF 20130700, HLF20100532, HLF20080843 to M.F.G.; HLF20090419 to I.G.; HLF20090704 to L.G.), Swedish Research Council (2009-4120, 2011-3900, 2014-3352 to M.F.G.; 2010-2932 to I.G.), European Foundation for the Study of Diabetes (EFSD), a European Research Council Advanced Research Grant (GA269045 to L.G.), European 7th Framework Programme HEALTH 2007-201413 (ENGAGE to L.G.), HEALTH-F2-2009-241544 (PREDICTIONS), and QLG2-CT-2001-01669 (EURAGEDIC). Support was also provided by the Swedish Medical Society; the Swedish Diabetes Association (Diabetesfonden); the Magnus Bergvall, Crafoord, Albert Påhlsson, Lars Hierta Memorial, Åke Wiberg, Thelma Zoéga, Ernhold Lundström, Lundgren, Tore Nilsson, Segerfalk, Hulda Almroth, Marianne and Marcus Wallenberg, and Knut and Alice Wallenberg (KAW 2009-0243) Foundations; the Royal Physiographic Society in Lund; the Malmö and Skåne Hospital research funds; regional research funds; the Vascular Wall Programme; and the Lund University Diabetes Centre. L.M.B. and O.K. received support from the Swedish Society for Medical Research.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.M.B. designed the study, performed in vitro the experiments, analyzed data, and wrote the article. V.L. designed the study, performed the Diabetes Genetics Initiative (DGI) genome-wide association study (GWAS), and analyzed genetic data. C.L. analyzed genetic data. O.K. and A.V.Z. performed in vitro experiments and analyzed data. A.E., M.N., and P.D. phenotyped carotid plaque endarterectomies. K.P. and C.B. performed the incretin clamp. S.A. performed genotyping in the DGI-GWAS study. C.F. performed phenotyping in the Finnish Diabetic Nephropathy study. A.J. performed genotyping and analyzed data. C.R.M. performed confocal imaging and analyzed data. A.S. and J.K. performed phenotyping and analyzed data in the Metabolic Syndrome in Men study. E.A. analyzed the DGI-GWAS. M.Laj., L.T., and P.R. performed phenotyping and analyzed data in the Steno studies. S.M. and A.V. were the principal investigators of the Steno studies. T.J.K. provided GIPR antibodies and expertise regarding the use of the antibodies. O.M., M.O.-M., and P.N. performed phenotyping in the Malmö study. P.-H.G. was the principal investigator of the Finnish Diabetic Nephropathy study. B.L. and A.G. were the principal investigators of the critical limb ischemia study. M.Laa. was the principal investigator of the Metabolic Syndrome in Men study. I.G. was the principal investigator of the carotid endarterectomy study. L.G. designed and supervised all parts of the study and drafted the article. M.F.G. designed and supervised all parts of the study, performed in vitro and confocal experiments, analyzed data, and wrote the article. All authors critically revised and approved the final version of this article. M.F.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.