Low Circulating Levels of Insulin-Like Growth Factor Binding Protein-1 (IGFBP-1) Are Closely Associated With the Presence of Macrovascular Disease and Hypertension in Type 2 Diabetes

One hundred sixty subjects were enrolled from two diabetes clinics (Hope Hospital and Manchester Royal Infirmary) in Manchester, U.K. The mean age of the participants was 58.1 ± 9.1 years, and subjects were matched for duration of diabetes (mean 8.2 ± 1.2 years). Previously, individuals had been randomly assigned to type of treatment according to U.K. Prospective Diabetes Study (UKPDS) protocols. Their current treatments were diet alone (n = 20), sulfonylurea (chlorpropamide or glipizide) treatment (n = 71), insulin alone (n = 29), and insulin (once or twice a day) plus sulfonylurea treatment (n = 40). They belonged to three different ethnic groups: European (n = 121), Pakistani (n = 23), and African-Caribbean (n = 16). There was no significant difference between the ethnic groups for age or duration of diabetes. All subjects gave informed consent, and the protocol was approved by local ethics committees.

All blood samples were obtained after an overnight fast, and the patients’ usual medications were withheld until after venesection. Plasma was separated from the remaining blood by centrifugation at 4°C. Plasma samples were stored at −70°C until analysis. Blood samples were taken at baseline and 12 months later.

Blood pressure was the mean of three readings taken with the subject in the supine position. A survey of medical records was carried out to obtain a detailed medical history.

Macrovascular disease was defined as a history of any of the following: ischemic heart disease including previous myocardial infarction and congestive heart failure, intermittent claudication, cerebrovascular accident, transient ischemic attack, peripheral vascular disease (defined by unilateral or bilateral absence of foot pulses), or systolic blood pressure ≥170 mmHg.

Levels of IGF-I (19), IGF-II (20), total IGFBP-I (tIGFBP-1) (using monoclonal antibody [mAb] 6303, which detects all IGFBP-1 phosphoforms) (21), and lesser-phosphorylated/nonphosphorylated IGFBP-I (lpIGFBP-1) (using mAb 6305, which detects only the lesser- and nonphosphorylated forms of IGFBP-1) (21) were determined by previously reported antibody-based assays. The assays have respective detection limits of 28 ng/ml, 30 ng/ml, and 3 μg/l and within- and between-assay coefficients of variation (CVs) of <10%. Where necessary, highly phosphorylated IGFBP-1 (hpIGFBP-1) was calculated by subtraction of lpIGFBP-1 from tIGFBP-1 for each subject.

Insulin and insulin C-peptide were measured using the Mercodia (Uppsala, Sweden) enzyme-linked immunosorbent assay (ELISA) for intact insulin and the Dako (Ely, U.K.) ELISA for intact insulin C-peptide, with respective detection limits of 7 and 70 pmol/l and within- and between-assay CVs of <8%. Cross-reactivity of the insulin assay for insulin C-peptide and the C-peptide assay for insulin is <0.1%.

Fasting cholesterol and triglycerides were measured by enzymatic methods using the Cobasintegra autoanalyzer (Roche Diagnostics, Welwyn Garden City, U.K.), with HDL cholesterol measured by a direct method (Cobasintegra autoanalyzer). LDL cholesterol was calculated according to the Friedewald formula (22).

Fasting nonesterified fatty acid (NEFA) concentrations were measured using the Wako (Neuss, Germany) ELISA for total NEFAs. Sensitivity of the assay is 0.05 mEq/l and within- and between-assay CVs <10%. Glucose was assayed using the hexokinase method (Cobasintegra autoanalyzer). HbA_1c was measured using the Cobasmira autoanalyzer (Roche Diagnostics).

To confirm the validity of antibody-based measurements of IGFBP-1 phosphoforms, we also separated and visualized the IGFBP-1 phosphoforms using nondenaturing PAGE (23). Briefly, for a subset of subjects with no manifest macrovascular disease, IGFBP-1 in 250 μl plasma was immunoprecipitated with mAb 6303 (1:1,000; volume 250 μl) (provided by Medix Biochemica, Kauniainen, Finland), which immunoprecipates all phosphoforms of IGFBP-1 (21), and incubated overnight at 4°C. IGFBP-1 in 500 μl plasma from subjects with known macrovascular disease was required for immunoprecipitation to be able to detect lower concentrations. Two controls were used, HepG2 conditioned medium containing all the phosphoforms of IGFBP-1 and amniotic fluid, which contains all the lesser-phosphorylated isoforms but not hpIGFBP-1. After incubation, the antigen-antibody complexes were immunoprecipitated using Saccel rabbit anti-mouse antibody (IDS, Tyne & Wear, U.K.). Immunoprecipitated IGFBP-1 was applied to a 10% native polyacrylamide gel (23). After overnight transfer to nitrocellulose membranes, nonspecific binding was blocked by incubation with 1% BSA in 0.15 mol/l sodium chloride. The proteins were revealed by incubation with 150,000 cpm/ml [¹²⁵I]IGF-I for 3 h at 25°C in accordance with the method of Hossenlopp et al. (24), followed by phosphor-image analysis (Fuji bio-imaging analyzer) and autoradiography.

Multiple logistic regression was used to assess the risk of macrovascular disease in relation to changes in IGFBP-1 concentration. Spearman correlations, both simple and partial, were calculated to investigate the relation between IGFBP-1 and other metabolic variables, BMI, and blood pressure.

Differences in metabolic variables between treatment groups and ethnic groups were assessed using ANCOVA. Non–normally distributed data were logarithmically transformed before analysis. Post hoc testing used Scheffe’s test. Significant predictors for IGFBP-1 were examined using multiple regression analysis. Comparison of IGFBP-1 at initial assessment and follow-up was carried out using linear regression analysis. All analyses were carried out using the statistical package SPSS, version 9.0.

tIGFBP-1 was markedly reduced in subjects (n = 40) with known macrovascular disease (48.7 μg/l [33.7–63.6], geometric mean [95% CI]) compared with patients (n = 120) with no manifest macrovascular pathology (80.0 μg/l [52.2–107]; F = 5.4, P = 0.01) (Table 1). A similar relationship was also found for hpIGFBP-1 concentration (known macrovascular disease, 45.1 μg/l [35.1–55.2]; no macrovascular disease, 75.8 μg/l [56.2–95.3]; F = 4.8, P = 0.01) (Table 1). Logistic regression showed that for every unit decrease in the log tIGFBP-1 (increase of 2.73 μ g/l in the tIGFBP-1 concentration), there was a 43% increase in the odds of the subject having macrovascular disease (odds ratio 0.57 [95% CI 0.40–0.83]; P = 0.001) (Fig. 1).

There was no difference in lpIGFBP-1 by macrovascular disease status. For the whole group, lpIGFBP-1 (12.2 μg/l [10.9–13.5]) was significantly lower than tIGFBP-1 (76.1 μg/l [56.8–95.5]; P < 0.001).

To confirm the assay results, native PAGE was performed on a subset of subjects with high (n = 11) versus low (n = 12) tIGFBP-1. Representative results are shown in Fig. 2. The concentration of hpIGFBP-1 was less in those with macrovascular disease than in those without manifest macrovascular pathology. There was no corresponding difference in lpIGFBP-1 by macrovascular disease status.

tIGFBP-1 correlated negatively with mean arterial pressure (MAP) (Spearman’s ρ = −0.26, P = 0.002). The latter relationship persisted even when corrected for age, sex, BMI, and insulin C-peptide concentration (partial ρ = −0.30, P < 0.001). hpIGFBP-1 correlated negatively with systolic blood pressure (ρ = −0.30, P < 0.01), diastolic blood pressure (ρ = −0.45, P < 0.001), and MAP (ρ = − 0.41, P < 0.001). In contrast to tIGFBP-1 and hpIGFBP-1, no consistent relationships were found with lpIGFBP-1 and macrovascular disease, hypertension, or cardiovascular risk markers.

The relationship between tIGFBP-1 and hypertension was robust. Using multiple regression modeling, 40% of the variance in tIGFBP-1 was accounted for by three variables, MAP (unstandardized coefficient B = −3.4 ± 0.8; P < 0.001), triglycerides (B = −23.1 ± 7.4, P = 0.003), and NEFAs (B = −0.6 ± 0.03, P = 0.04) (Figs. 3A–C) in a model that included IGF-I, IGF-II, insulin C-peptide, cholesterol, triglycerides, NEFAs, fasting glucose, HbA_1c, age, sex, BMI, and MAP. When this analysis was repeated by treatment group, the relationship with MAP was greatest in the group treated with insulin alone, where 69% of the variance in IGFBP-1 was accounted for by MAP alone (P < 0.001) (Fig. 3D).

As anticipated, there was a strong negative correlation of tIGFBP-1 concentration with insulin C-peptide concentration and also with peripheral insulin concentration and triglyceride level (Table 2). However, there was no relationship between tIGFBP-1 and fasting glucose or HbA_1c. This result was in complete contrast to lpIGFBP-1, which correlated positively with fasting glucose (ρ = 0.350, P < 0.01).

In agreement with our previously reported findings (25), tIGFBP-1 concentration was higher in subjects treated with insulin alone than in any other group (Table 3; Fig. 4). This difference persisted even when tIGFBP-1 was corrected for BMI and C-peptide in addition to age and sex (F = 6.5, P < 0.001). In contrast, there was no variation in lpIGFBP-1 concentration by treatment group (Table 3). Peripheral insulin levels were higher in those treated with insulin (with or without additional sulfonylureas) than in subjects not receiving insulin, whereas C-peptide levels were lower in subjects treated with insulin (with or without additional sulfonylureas). There was no significant difference in fasting glucose level or HbA_1c by treatment group (Table 3).

There were no differences in IGF-I or IGF-II levels between subjects with and without evidence of macrovascular disease. IGF-I correlated positively with IGF-II and negatively with NEFAs. Interestingly, IGF-II concentration was strongly positively correlated with elevated total, LDL, and HDL cholesterol levels (Table 2).

IGF-I concentration was not affected by treatment category, but IGF-II levels were lower in subjects treated with insulin alone than in any other group (Table 3). There was a trend for total cholesterol to be lower in subjects treated with insulin alone (F = 2.2, P = 0.06), but otherwise there were no detectable differences in lipid profile by treatment group.

Because hepatic IGFBP-1 synthesis is acutely regulated by glucoregulatory hormones, we wished to establish whether assessment of fasting IGFBP-1 levels could robustly predict association with macrovascular disease even after a gap of 12 months. Therefore, tIGFBP-1 concentration was assessed at initial visits and 12-month follow-ups as shown in Fig. 5. Linear regression analysis demonstrated that tIGFBP-1 on follow-up was closely related to tIGFBP-1 at initial visit by the equation tIGFBP-1_follow-up = 3.2 + (tIGFBP-1_{initial visit}) × 0.92 (r² = 0.90, P < 0.001). This suggests that fasting tIGFBP-1 concentration is a consistent measure with little variation over a 12-month period.

There were no differences in IGFBP-1 levels between the ethnic groups (Table 4). IGF-I concentration was higher in African-Caribbeans than in Europeans or Pakistanis, and IGF-II was higher in Europeans than in African-Caribbeans or Pakistanis. The IGF-I results are in accord with previous findings in nondiabetic subjects (26) and indicate that the onset of type 2 diabetes does not mask underlying differences in circulating IGF concentrations related to ethnicity. Also in keeping with previous studies (2), we found a close correlation between IGF-II and total, LDL, and HDL cholesterol. The significance of the IGF-II/cholesterol relationship is unknown, although similarities in clearance pathways between the β-very low density lipoprotein (β-VLDL) receptor and the mannose-6-phosphate receptor may suggest a possible mechanism (27). Insulin levels were similar in all three ethnic groups, with C-peptide being lower in African-Caribbeans than in Europeans or Pakistanis. There was no significant difference in total cholesterol, LDL cholesterol, or triglyceride levels by ethnic group. There was no difference in age, duration of diabetes, BMI, or fasting glucose between the three ethnic groups.

This is the first study to demonstrate close links between fasting hpIGFBP-1 and the presence of macrovascular disease and hypertension in type 2 diabetes. In addition to the negative correlation between IGFBP-1 and macrovascular disease, we found significant negative relationships between tIGFBP-1 and other known cardiovascular risk factors, namely BMI, triglyceride concentrations, circulating insulin C-peptide, and insulin levels. These relationships were robust and persisted over a 12-month period.

IGFBPs have been clearly shown to modulate the actions of IGFs. IGFBP-1 in particular is accepted as an acute modulator of IGF bioavailability (13). We and others have shown a close relationship between low IGFBP-1 levels and surrogate markers of cardiovascular risk in a nondiabetic population (3,15). We showed that low circulating IGFBP-1 concentrations significantly correlated with impaired glucose tolerance (IGT). Because the subjects in our previous study (15) did not have any evident macrovascular pathology, and because IGT precedes the development of type 2 diabetes, it is probable that a low circulating IGFBP-1 concentration predates the development of macrovascular disease. As proposed previously (15), we believe that the development of peripheral insulin resistance, with maintenance of hepatic insulin sensitivity, explains the decrease in circulating IGFBP-1 (and thus the increase in IGF bioavailability) long before the development of type 2 diabetes. Our current data suggest that this pattern persists in a type 2 diabetic population and further predisposes affected individuals to premature macrovascular disease.

The majority of circulating IGFBP-1 is in the highly phosphorylated state, implying that it is this isoform of IGFBP-1 that is most relevant to the development of hypertension and macrovascular disease (2). hpIGFBP-1 has the highest affinity for IGF-I relative to other isoforms (16,17), and its affinity for ligand is significantly greater than that of the type 1 IGF receptor (28). Thus hpIGFBP-1 has the greatest potential to inhibit IGF-I/IGF receptor interactions.

In nondiabetic hypertensive humans, circulating IGF-I levels independently predict 16% of the variability in left ventricular mass (29). In another study, a reduction in serum IGF-I level in relation to bezafibrate treatment in post–myocardial infarction patients was independently associated with angiographic measures of coronary artery disease regression (30). Furthermore, in acromegaly, a human model of IGF-I excess (where World Health Organization criteria indicate that 37.5% patients have hypertension), blood pressure positively correlates with IGF-I concentration (31). These findings strongly support an endocrine effect of IGF-I in vivo. It is important, however, to recognize that IGFBP-1 is capable of inhibiting endocrine, paracrine, and autocrine IGF effects because of its ability to access the extracellular space and to associate with specific cell-surface integrins (32). Thus a high circulating level of hpIGFBP-1 will inhibit IGF effects whatever the cellular source of the IGF-I.

An elevated concentration of hpIGFBP-1 would therefore serve to protect vascular tissues from the mitogenic potential of circulating IGFs, whereas a consistently reduced level—induced by the diabetic state per se, hyperinsulinemia, or both—would increase IGF availability to susceptible tissues. This situation could be further worsened in type 2 diabetes, where reports of increased IGFBP-3 proteolysis (33) would further contribute to the exposure of the vascular compartment to the mitogenic actions of IGFs. It is interesting in this respect to note that African-Caribbean subjects in this and other studies tend to have higher IGF-I levels and are known to be more susceptible to resistant hypertension than other ethnic groups (34).

There are no studies examining the ability of liver to differentially phosphorylate IGFBP-1 in humans. Peripheral dephosphorylation seems a more likely explanation for the presence of lpIGFBP-1. Evidence for the ability of ectophosphatases (present in many cell types) to dephosphorylate IGFBP-1 has been reported (35). Cell-surface dephosphorylation would significantly affect the bioavailability of IGF-I at the cellular level. Thus IGFBP-1 in the highly phosphorylated state would significantly reduce interaction of IGF-I with its receptors, but with subsequent dephosphorylation could enhance local IGF bioavailability. Thus changes in tissue phosphatase/protease activity would further influence IGFBP-1/IGF-I binding kinetics and so IGF bioavailability at the tissue level (35). Further studies will be required to determine how the regulation of these factors influences cell-specific IGF effects.

In this study, we found that lesser-/nonphosphorylated IGFBP-1 correlated positively with fasting glucose. This finding was unexpected and has not been described previously. It implies that IGFBP-1 phosphorylation status may play an important part in mediating the effects of IGFBP-1 on glucose homeostasis. The finding of a positive correlation between fasting glucose and lpIGFBP-1 may suggest a mechanism to increase insulin-like activity at times of high ambient glucose concentration and, given the lower affinity of lpIGFBP-I for IGF-I, parallels reports of increased IGFBP-3 proteolysis associated with hyperglycemia in type 2 diabetes (35).

This study confirms our earlier findings of higher IGFBP-1 levels in subjects with type 2 diabetes who were treated with insulin alone (25). A simple explanation for this finding would be that a reduced portal insulin concentration in these subjects would remove the tonic inhibition of hepatic IGFBP-1 synthesis, leading to increased circulating levels of IGFBP-1. To address this issue in this study, we measured C-peptide levels, which undergo little hepatic extraction and therefore better represent hepatic insulin exposure via the portal system. There were no significant differences in insulin and insulin C-peptide concentrations between the groups treated with insulin alone or insulin plus sulfonylureas, although IGFBP-1 levels were significantly different, suggesting that it is less likely to be pancreatic β-cell insulin output and more likely to be hepatic insulin resistance that determines the differences in tIGFBP-1 level between treatment groups. This result is in keeping with the multiple linear regression analysis showing that factors other than insulin were important in determining tIGFBP-1 concentration, particularly MAP, NEFAs, and triglycerides.

The close relationship between fasting tIGFBP-1 levels at inception of the study and 1-year follow-up confirms the robustness of fasting tIGFBP-1 measurement over time and points to the utility of tIGFBP-1 as a predictive biomarker for macrovascular disease in type 2 diabetes.

In summary, these studies demonstrate the novel and important finding of close links between low circulating levels of the highly phosphorylated isoform of IGFBP-1 and the presence of macrovascular disease and hypertension in type 2 diabetes. In addition, we have shown that lesser-phosphorylated IGFBP-1 levels are associated with ambient glucose concentration, suggesting that the phosphorylation status of IGFBP-1 may be important in glucose homeostasis. Because IGFBPs largely determine the mitogenic and metabolic potential of IGFs, our findings suggest an etiopathogenic role for IGFBP-1 in modulating the development of large vessel disease in type 2 diabetes.

We thank the National Health Service (NHS) North-West Research and Development Directorate for funding the work of A.H.H. in this project and Salford Royal Hospitals NHS Trust for the funding of technical support.