Circulating and Urinary Transforming Growth Factor β1, Amadori Albumin, and Complications of Type 1 Diabetes: The EURODIAB Prospective Complications Study

The EURODIAB Prospective Complications Study is a follow-up of the EURODIAB IDDM Complications Study (9). Baseline investigations (1988–1991) were performed on 3,250 men and women with type 1 diabetes drawn from 31 European centers. Sample selection was stratified by sex, age-group, and duration of diabetes to ensure sufficient representation in all categories. Type 1 diabetes was clinically defined as a diagnosis made before the age of 36 years, with a continuous need for insulin therapy within 1 year of diagnosis.

The follow-up (EURODIAB Prospective Complications Study) was performed on average 7–9 years later. Of the 3,250 patients, 1,880 (57.8%) returned for examination. Of these, full data on cardiovascular disease (CVD), microvascular complications, and neuropathy were available on 1,296 patients. Of those who were not reexamined, 437 had been recruited from centers not taking part in the follow-up, 101 had died, 465 provided morbidity data only, and 367 were untraced or otherwise unavailable.

All complications were again measured to a standard protocol (10). Two 24-h urine collections were performed (only one was performed at baseline) to measure the albumin excretion rate (AER). Aliquots were frozen and sent to London for analysis of urinary albumin with an immunoturbidimetric method using goat anti-human albumin antisera (Sanofi Diagnostics Pasteur, Minneapolis, MN) and human serum albumin (ORHA 20/21 grade human serum albumin [HSA]; Behring Diagnostics, Hoechst, U.K.) standards (11). Retinopathy was assessed from retinal photographs (two fields per eye) according to the EURODIAB protocol (12). CVD was defined as a positive medical history, bypass graft surgery, and/or major Q waves on a centrally coded electrocardiogram.

Blood samples were taken, fasting if possible, for measurement of lipids and glycemic control. At a central laboratory, cholesterol and triglyceride levels were measured by enzymatic colorimetric tests (13), and HDL cholesterol was measured directly (14). HbA_1c was measured by a latex enhanced turbidimetric immunoassay (Roche, Welwyn Garden City, U.K.). The reference range for this assay was 4.2–6.2%.

Total TGF-β1 was measured by an enzyme-linked immunosorbent assay (ELISA) development system (R&D Systems, Oxon, U.K.). To activate the latent plasma TGF-β1 to the immunoreactive form, 0.1 ml plasma was acid-activated by 0.1 ml 2.5 N acetic acid/10 mol/l urea for 10 min at room temperature and neutralized by 0.1 ml 2.7N NAOH/1 mol/l HEPES. For urine, 1 ml was activated by 0.2 ml 1 N HCl and neutralized by 0.2 ml 1.2 N NAOH/0.5 mol/l HEPES. Intra- and interassay variations were 5.0 and 7.0%, respectively. Urinary TGF-β1 was indexed to urinary creatinine excretion. Amadori albumin was determined in a competitive ELISA as recently described (15).

Briefly, each well was coated with 100 μl of 0.1 mg/ml glycated HSA as prepared by the incubation of HSA with 0.5 mol/l glucose for 4 weeks. This preparation was also used as standard in the competitive ELISA. Then, 50 μl of the antibody conjugated with biotin (1:2,000) and 50 μl of standard or a sample to be tested, diluted in PBS including 30 mg/ml HSA, was added to each well and incubated for 2 h. After three washes with PBS-Tween, the wells were incubated with streptavidin-peroxidase (CLB, Amsterdam, the Netherlands) for 1 h, followed by substrate development with 100 μl tetramethylbenzidine. The reaction was stopped with 100 μl 0.5 mol/l H₂SO₄. The extinction at 450 nm was measured with a multichannel spectrophotometer (SLT Microplate Reader; Wilten Bioteknika, Etten-Leur, the Netherlands). Plasma levels were expressed as Amadori albumin units, and 1 unit was defined as the antibody-reactive material equivalent to 1 μg glycated HSA standard.

We used a nested case-control approach to maximize efficiency. Case subjects who had the greatest complication burden were selected to provide sufficient numbers for subgroup analyses. Control subjects who were completely free of complications were selected. Thus, case subjects were individuals with CVD or proliferative retinopathy or macroalbuminuria at follow-up and individuals with microalbuminuria and some degree of retinopathy (n = 356). Control subjects were individuals who had no evidence of CVD, retinopathy, or neuropathy and were normoalbuminuric at follow-up (n = 185). This selection allowed us to compare people with and without single or multiple complications. Case and control subjects were unmatched, so that the impact of key variables, such as age, could still be assessed, and any adjustments were taken care of at the analysis stage. Of these 541 individuals, sufficient sampling material was available for 535 for Amadori albumin, 534 for circulating TGF-β1, and 532 for urinary TGF-β1. These numbers allow us to detect a difference between groups of at worst one-third of an SD, with 90% power and 5% significance.

Key risk factors in case and control subjects were compared, reporting crude means and SDs or geometric means and 25th and 75th percentiles for skewed variables. Urinary and circulating TGF-β1 were log-transformed before analysis. Pearson correlation coefficients were first calculated for circulating and urinary TGF-β1 and Amadori albumin and other risk factors for complications. Then mean values of circulating and urinary TGF-β1 and Amadori albumin were estimated using ANOVA for each complication. Finally, standardized regression effects (SREs) were compared. These SREs were calculated by multiplying the β coefficient of each variable in a multivariate logistic regression model by its SD. This calculation allowed us to compare the impact of each factor directly with the other in standardized units. All SREs were first adjusted for age; adjustment for diabetes duration alone produced similar findings. There were no significant sex differences in these relationships; therefore, both sexes were combined, and no adjustment was made for sex.

Case subjects were more likely to have an adverse risk factor profile and raised levels of urinary and circulating TGF-β1 and Amadori albumin than control subjects (Table 1).

Urinary and circulating TGF-β1 was not correlated (coefficient 0.05, P = 0.3, Table 2). Urinary TGF-β1 was correlated with diabetes duration, HbA_1c, systolic pressure, AER, Amadori albumin, and lipids and was correlated negatively with measures of body size. Circulating TGF-β1 was only correlated with HbA_1c and Amadori albumin. Amadori albumin, as anticipated, was correlated with HbA_1c, AER, and lipids and was correlated negatively with estimates of body size.

Mean urinary TGF-β1 and Amadori albumin were elevated in microalbuminuric and macroalbuminuric patients (Table 3). However, circulating TGF-β1 was only significantly raised in macroalbuminuric patients compared with normoalbuminuric patients.

The strength of the association between TGF-β1 and macroalbuminuria was second only to systolic blood pressure in the univariate analyses (Table 4, SRE 2.45 vs. 3.18). All univariate associations were attenuated in bivariate models. For example, the SRE for TGF-β1 was attenuated from 2.45 to 2.02 with the addition of HbA_1c, and the univariate association between HbA_1c and macroalbuminuria was attenuated from 2.27 to 1.91 in the same model. In the multivariate model, the association between TGF-β1 and macroalbuminuria was severely attenuated, but by no means abolished, by the inclusion of HbA_1c and Amadori albumin; HbA_1c was the least affected by these inclusions. The impact of systolic blood pressure was less strong than that of HbA_1c and was almost equivalent to that of Amadori albumin on TGF-β1 in the bivariate models. But in the multivariate model, blood pressure had an almost unchanged impact on macroalbuminuria compared with the univariate data, indicating that the impact of blood pressure on albuminuria is likely to be independent of the consequences of hyperglycemia.

In contrast to the findings on nephropathy, there was a striking difference in circulating TGF-β1 and Amadori albumin between patients with proliferative retinopathy and no retinopathy, which remained on adjustment for AER (Table 3). The association between urinary TGF-β1 and retinopathy could be accounted for by adjustment for AER. For proliferative retinopathy, the univariate association between circulating TGF-β1 was attenuated by the addition of the other covariates (Table 5). The SRE for Amadori albumin was barely altered with the addition of TGF-β1. However, the association between Amadori albumin and proliferative retinopathy was almost abolished with the addition of HbA_1c. In the multivariate model, associations between HbA_1c, and systolic blood pressure and retinopathy, differed little from their univariate associations, but the relationship for circulating TGF-β1 and Amadori albumin became statistically nonsignificant.

There was no association between Amadori albumin and either urinary or circulating TGF-β1 and CVD (data not shown).

Circulating TGF-β1 was higher in individuals on ACE inhibitors than in individuals not on any antihypertensive therapy (3.67 vs. 2.92 ng/ml, P = 0.001). A nonsignificant difference was observed for urinary TGF-β1 (5.99 vs. 5.88 pg/mmol, P = 0.9). These associations may be confounded by nephropathy status. Strikingly, elevated levels of circulating TGF-β1 were observed in individuals on ACE inhibitors in the normoalbuminuric and microalbuminuric range (5.87 vs. 5.58 and 7.46 vs. 6.05 ng/ml, respectively), whereas levels were lower in individuals on ACE inhibitors who had macroalbuminuria (5.58 vs. 7.03 ng/ml, respectively; P = 0.04 for interaction). Similar differences were observed for urinary TGF-β1. Mean systolic blood pressure was 13 mmHg higher in normoalbuminuric patients on ACE inhibitors than in individuals not on treatment; this difference was 8 mmHg in microalbuminuric patients and 5 mmHg in macroalbuminuric patients. Adjustment for these blood pressure differences largely attenuated differences in both circulating and urinary TGF-β1 in normoalbuminuric and microalbuminuric patients but not in macroalbuminuric patients.

We show that both urinary TGF-β1 and Amadori albumin are elevated in microalbuminuric and macroalbuminuric patients with type 1 diabetes. Separately, Amadori albumin and HbA_1c accounted for a similar proportion of the association between TGF-β1 and albuminuria, but their joint impact attenuated this association still further. This result indicates that HbA_1c effects TGF-β1 via mechanisms in addition to its impact on Amadori albumin and on albuminuria via means that are not exclusive to TGF-β1. This is not unexpected because it is likely that several other mechanisms account for the impact of hyperglycemia on diabetic renal disease.

Cell culture studies show that a high glucose medium can enhance autocrine production and activation of TGF-β1, and TGF-β1 in turn can enhance glucose uptake through fibroblasts (16), indicating an amplification of the effect of hyperglycemia. In human studies of type 2 diabetes, both urinary and circulating TGF-β1 have been found to be raised in individuals with proteinuria but not in individuals with microalbuminuria compared with control subjects (17). However, these previous studies were relatively small and may not have been powered to detect a difference for earlier stages of disease. Further, these studies have not generally accounted for the confounding effects of other risk factors and complications. Few data exist for type 1 diabetes and have so far not indicated elevations of urinary TGF-β1 with albuminuria as we do here (18).

We also show that the association between Amadori albumin and macroalbuminuria could largely be explained by glycemic control because the univariate SRE was attenuated from 1.67 to 1.21 in the bivariate model with HbA_1c, with little attenuation to 1.15 in the multivariate model, indicating that much of the mechanisms of HbA_1c and Amadori albumin on diabetic nephropathy are shared and that at least part of that effect is mediated via TGF-β1. There is now good mechanistic evidence demonstrating the role of Amadori albumin in the etiology of diabetic nephropathy. Mesangial cells can recognize and bind Amadori albumin in a dose-response fashion, and animal studies show that infusion of Amadori albumin produces renal glomerular changes that are similar to those observed in diabetes (19,20). That this nephropathic effect is mediated via TGF-β1 is shown by cell culture studies, in which incubation with glycated albumin resulted in overexpression of TGF-β1 mRNA levels (21). Administration of anti-glycated albumin antibody in animal models reduces plasma levels of glycated albumin, reduces AER, and reverses glomerular pathology (22). Human studies have been conflicting, perhaps in part because of small numbers and confounding by other risk factors, but one of the largest previous studies clearly indicated that Amadori albumin levels were elevated in individuals with nephropathy compared with individuals without nephropathy (15,23–25).

Interestingly, the inclusion of both Amadori albumin and HbA_1c in our models could only account for around one-quarter of the association between TGF-β1 and albuminuria, indicating that other mechanisms shape the association between TGF-β1 and albuminuria. A clear candidate is hypertension, which alone was the strongest correlate of macroalbuminuria. Certainly the inclusion of blood pressure with TGF-β1 had the same attenuating effect on the risk estimate of TGF-β1 as Amadori albumin did, suggesting that raised blood pressure, an important risk factor for albuminuria, was in part working through TGF-β1 to mediate its effects; however, clearly from the multivariate analyses, this was not the sole, or even the most important, mechanism. This finding can be accounted for by the observation of a stimulatory effect of TGF-β1 on the renin angiotensin system (6) and the beneficial effect of angiotensin II receptor blockers on TGF-β1 levels and renal pathology (26). Other mechanisms not studied here include activation of protein kinase C, other growth factors such as vascular endothelial growth factor and insulin-like growth factor, oxidative stress, and intraglomerular hemodynamic factors (27–29).

Interestingly, urinary TGF-β1 was only lower in macroalbuminuric patients on ACE inhibitors compared with individuals on no treatment. In the normoalbuminuric and microalbuminuric categories, there appeared to be little difference or even a tendency to higher levels in individuals on ACE inhibitors. This was in part accounted for by the greater blood pressure differentials at lower albuminuric levels between individuals on treatment and those not on treatment. A separate explanation is that ACE inhibition suppresses TGF-β1 receptor expression, which could reduce TGF-β1 signaling and thus TGF-β1–mediated effects (30).

In contrast, circulating TGF-β1 was only elevated in the presence of macroalbuminuria, and not microalbuminuria, compared with normoalbuminuria, indicating that it may be less sensitive to renal changes. However, clear differences were observed between proliferative retinopathy and no retinopathy in circulating TGF-β1, which unlike differences observed for urinary TGF-β1, could not be accounted for by confounding with AER. The impact of TGF-β1 on proliferative retinopathy remained largely unexplained by elevated blood pressure but could be largely accounted for by HbA_1c. Others (8) have indicated that circulating TGF-β1 is elevated in nonproliferative retinopathy in type 2 diabetes. By inhibiting replacement of damaged cells, TGF-β1 may contribute to capillary and pericyte loss in diabetic retinopathy (31). In contrast though, whereas total vitreous TGF-β1 did not differ between individuals with proliferative retinopathy compared with control subjects, latent TGF-β1 was substantially decreased in the former group, indicating that TGF-β1 may also have an anti-mitogenic effect (32).

Animal studies show that overexpression of TGF-β1 occurs within as little as 3 days of streptozotocin-induced diabetes and that this can be inhibited by insulin (33). The identification of such early markers of renal damage may have important implications for the treatment of early diabetic renal disease and for monitoring the success of therapy. Our findings also indicate that TGF-β1 is an early marker of renal damage in type 1 diabetes. As may be expected from the current understanding of the metabolism of TGF-β1, blockade of the renin angiotensin system and, separately, inhibitors of glycation end products reduce the overexpression of renal TGF-β1 and have a beneficial effect on renal pathology in animal models (34). More unexpected is the finding that (again, in animal models only) lipid-lowering agents, specifically 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors, also reduce the expression of TGF-β1 and reduce albuminuria levels, although this appears not be associated with a reduction in lipid levels (35).

There are certain limitations to this analysis. This is a cross-sectional study, and we can therefore only imply associations rather than causal relationships. However, our results are supported by in vitro and animal studies demonstrating the mechanisms of the associations between glucose, Amadori albumin, TGF-β1, and diabetic complications. Further follow-up data on these and other patient groups would, of course, be invaluable in testing whether our findings are due to causal relationships. Unlike previous studies, a key strength here is the ability to account for confounding by other risk factors and complications, and the large sample size provides sufficient power for these analyses. There are concerns also about the use of plasma to measure circulating TGF-β1. TGF-β1 represents the major isoform of the TGF-β1 superfamily in blood. It is present in platelet granules and is released on platelet activation. However, the degree of sample contamination by platelet release of TGF-β1 in plasma samples appears to be relatively small (36), and because the protocol for sample collection and handling was identical across sites, any bias is likely to be nondifferential.

In conclusion, we demonstrate that urinary TGF-β1 is closely correlated with the degree of albuminuria in type 1 diabetes and confirm that this effect is largely accounted for by hyperglycemia, Amadori albumin, and blood pressure. Circulating TGF-β1, on the other hand, is more closely correlated with retinopathy, and its effect does not appear to be so closely associated with Amadori albumin. These findings underline new possibilities for both the monitoring and treatment of microvascular complications of type 1 diabetes and deserve further exploration.

B. Karamanos, A. Kofinis, K. Petrou, Hippokration Hospital, Athens, Greece; R. Giorgino, F. Giorgino, G. Picca, A. Angarano, G. De Pergola, Istituto di Clinica Medica, Endocrinologia e Malattie Metaboliche, Univesita di Bari, Bari, Italy; C. Ionescu-Tirgoviste, A. Coszma, Clinic of Diabetes, Nutrition, and Metabolic Diseases, Bucharest, Romania; M. Songini, A. Casu, M. Pedron, M. Fossarello, Department of Internal Medicine, *Ospedale San Michele, Cagliari, Italy; J.B. Ferriss, G. Grealy, D.O. Keefe, A. White, P.E. Cleary, Cork University Hospital, Cork, Ireland; M. Toeller, C. Arden, Diabetes Research Institute, Heinrich-Heine University, Dusseldorf, Germany; R. Rottiers, C. Tuyttens, H. Priem, University Hospital of Gent, Gent, Belgium; P. Ebeling, M. Kylliäinen, T. Kyostio-Renvall, University Hospital of Helsinki, Helsinki, Finland; B. Idzior-Walus, J. Sieradzki, K. Cyganek, Department of Metabolic Diseases, Jagiellonian University, Krakow, Poland; H.H.P.J. Lemkes, C. Roest, University Hospital of Leiden, Leiden, the Netherlands; J. Nunes-Correa, M.C. Rogado, L. Gardete-Correia, M.C. Cardoso, Portuguese Diabetic Association, Lisbon, Portugal; G. Michel, R. Wirion, S. Cardillo, Center Hospitalier, Luxembourg, Luxembourg; G. Pozza, R. Mangili, V. Asnaghi, R. Lattanzio, G. Galardi, Ospedale San Raffaele, Milan, Italy; E. Standl, B. Schaffler, H. Brand, A. Harms, City Hospital Schwabing, Munich, Germany; D. Ben Soussan, O. Verier-Mine, Center Hospitalier de Valenciennes, Valenciennes, France; J.H. Fuller, J. Holloway, L. Asbury, D.J. Betteridge, University College London, London, U.K.; G. Cathelineau, A. Bouallouche, B. Villatte Cathelineau, Hospital Saint-Louis, Paris, France; F. Santeusanio, G. Rosi, V.D. Alessandro, C. Cagini, Instituto di Patologia Medica, Policlinico, Perugia, Italy; R. Navalesi, G. Penno, S. Bandinelli, R. Miccoli, Dipartimento di Endocrinologia e Metabolismo, Pisa, Italy; G. Ghirlanda, C. Saponara, P. Cotroneo, A. Manto, A. Minnella, Universita Cattolica del Sacro Cuore, Rome, Italy; J.D. Ward, S. Tesfaye, S. Eaton, C. Mody, Royal Hallamshire Hospital, Sheffield, U.K.; M. Porta, P. Cavallo Perin, M. Borra, S. Giunti, Clinica Medica B, Patologia Medica, Ospedale Molinette, and Ospedale “Agnelli,” Turin, Italy; N. Papazoglou, G. Manes, General Hospital of Thessaloniki, Thessaloniki, Greece; M. Muggeo, M. Iagulli, Cattedra di Malatties del Metabolismo, Verona, Italy; K. Irsigler, H. Abrahamian, Hospital Vienna Lainz, Vienna, Austria; S. Walford, E.V. Wardle, J. Sinclair, S. Hughes, New Cross Hospital, Wolverhampton, U.K.; and G. Roglic, Z. Metelko, Z. Resman, Vuk Vrhovac Institute for Diabetes, Zagreb, Croatia.

J.H. Fuller (London), B. Karamanos, Chairman (Athens), A.-K. Sjolie (Aarhus), N. Chaturvedi (London), M. Toeller (Dusseldorf), G. Pozza (co-chairman; Milan), B. Ferriss (Cork), M. Porta (Turin), R. Rottiers (Gent), and G, Michel (Luxembourg).

J.H. Fuller, N. Chaturvedi, J. Holloway, D. Webb, E. Asbury, University College London, London, U.K.

G.-C. Viberti, R. Swaminathan, P. Lumb, A. Collins, S. Sankaralingham, Guy’s and St Thomas’ Hospital, London, U.K.

S. Aldington, T. Mortemore, H. Lipinski, Imperial College, London, U.K.

The EURODIAB Prospective Complications Study was supported by grants from the Wellcome Trust, the European Community, and Diabetes U.K.

We would like to thank all the investigators and patients who kindly participated in the study.