Demonstration of a Hyperglycemia-Driven Pathogenic Abnormality of Copper Homeostasis in Diabetes and Its Reversibility by Selective Chelation: Quantitative Comparisons Between the Biology of Copper and Eight Other Nutritionally Essential Elements in Normal and Diabetic Individuals

Men (age 30–70 years) with normal electrocardiograms were recruited for this study. Diabetic subjects had been diagnosed at least 6 months prior to the study, and control subjects had normal glucose tolerance. Exclusion criteria included the presence of type 1 diabetes; nephropathy; abnormal hematology or Fe deficiency; a history of significant cardiac disease; previous hepatic, gastrointestinal, or endocrine disease other than diabetes; gangrene or active sepsis; severe retinopathy; nondiabetic renal disease or renal allograft; malignancy, except cutaneous basal cell carcinoma; or known abnormality of Cu or Fe metabolism as well as current treatment with diuretics or calcium-channel blockers.

All protocols received appropriate regulatory approval, and all subjects provided written informed consent.

Type 2 diabetic (n = 20) and control (n = 20) subjects underwent a factorial, randomized, double-blind, placebo-controlled elemental balance study with screening, enrollment, run-in, treatment, and follow-up periods. Participants were resident throughout the study. Elemental intake was controlled by providing all items of food and beverages, which were directly measured (duplicate diets). Diets (Xyris Foodworks) were constructed to adhere to American Diabetes Association recommendations (9).

Elemental excretion was determined throughout the basal period. Balances were calculated from the difference between intake and output. [Elements]_{fasting serum} were determined on the mornings of days 1 and 7, the latter before drugs were administered. After the basal study (days 1–6), subjects were randomized to placebo or trientine (2,400 mg/day; Anstead, Essex, U.K.) and immediately entered the 6-day treatment period (days 7–12). The second part of the study was an otherwise identical regimen, which was completed on the morning of day 13.

Completeness of urine collections was confirmed by a p-aminobenzoic acid (PABA) test. Feces were freeze-dried; completeness was considered to be 93–97% (opaque food-markers; X-ray of fecal samples). Elemental concentrations in urine, feces, and food aliquots were measured (10) using an inductively coupled plasma−mass spectrometer (Perkin-Elmer Sciex Elan 6100). Biochemical and hematological variables (Table 1) were measured in fasting blood samples on days 1 and 7 (pretreatment) and day 13 (treatment). The level of 2-h [Mg]_serum was determined for 10-h after dosing. Data reported for day 1 are herein termed “basal” and those for days 1 and 7 are designated “pretreatment.”

We also measured the effects of increasing dosages of trientine (300, 600, 1,200, and 2,400 mg/day) on urinary elements. Samples were collected before and after each 1-week treatment period from control and diabetic subjects (n = 7 each) after completion of the main study (total duration 28 weeks).

Basal values from healthy (control) and diabetic patients were compared by one-way ANOVA. Restrictive iterative generalized least squares (RIGLS) models were fitted by restricted maximum likelihood (REML) (11). The effects of patient status (control or diabetic) and treatment (placebo or trientine) were analyzed using a factorial experimental design with two time periods, days 0–6 and 7–12, wherein the interaction effect was the term of primary significance. Differences between time periods for each subject were used in a mixed-model ANOVA fitted by REML in which subjects were considered as random (12) and differences between treatments (trientine or placebo) and patient status (diabetic or control) and the interaction between treatment and patient status were tested. Assumptions of normality and homoscedasticity were directly verified. Pearson’s correlation coefficients were calculated to determine the relation between basal Cu and key variables. A model predicting urinary Cu excretion during drug treatment (days 7–12) was constructed by adding predictive basal (day 1) variables into a forward, stepwise, multiple regression model (with α = 0.25 for entry and α = 0.10 for leaving). A significance level of α = 0.05 was used for all statistical tests.

In all, 20 diabetic and 19 control subjects completed the study. Diabetic subjects had greater BMI, fasting plasma glucose (FPG), and HbA_1c than did control subjects (each P < 0.001) (Table 1). Basal [Mg]_serum was lower (P < 0.01) and [ferritin]_serum was elevated (P < 0.001) in diabetic compared with control subjects. Serum levels are usually not a good measure of nutritional status, and measured values of Cu and five other elements did not differ between diabetic and control subjects (Table 1).

Basal extracellular superoxide dismutase (EC-SOD) was elevated in diabetic subjects on days 1 and 7 (both P < 0.01) and was related to [Cu]_serum levels (EC-SOD = 19.6.[Cu]_serum − 228; r² = 0.16, P = 0.011), but no equivalent association was observed in control subjects. Basal EC-SOD did not correlate with the serum concentration of any other element in diabetic subjects. EC-SOD was related to the interaction between [Cu]_serum and HbA_1c in RIGLS models on days 1 ([Cu]_serum, P = 0.022; HbA_1c, P = 0.014; interaction term, P = 0.0096) and 7 ([Cu]_serum, HbA_1c, and interaction term, all P < 0.0001). A three-dimensional spline-surface (on day 7) illustrated that maximum elevations in EC-SOD occurred mainly in subjects in whom both [Cu]_serum and HbA_1c were high (Fig. 1). Thus, elevated EC-SOD may reflect a chemical interaction between these variables. In diabetic subjects, trientine treatment lowered mean serum EC-SOD below baseline levels (31.8 ± 34.7 [day 13] vs. 72.6 ± 72.0 [day 1] units/l; P = 0.034). EC-SOD was not related to interactions between serum concentrations of any other elements and HbA_1c in diabetic subjects (RIGLS; all NS) nor was HbA_1c alone significantly related to EC-SOD in either diabetic or control subjects.

[Ferritin]_serum was elevated in diabetic compared with control subjects (Table 1), but values were uncorrelated with [Fe]_serum, iron-binding capacity (IBC), [Hb]_blood, or packed cell volume in diabetic subjects. These findings are consistent with those in previous reports.

Basal elemental balances are presented in Table 2. Food intake did not differ between groups or as a result of treatment, nor did the basal balance differ between control and diabetic groups. The Cu balance was variable but tended to be more positive in diabetic than in control subjects, although the difference was not significant (P = 0.19) (Table 2). Basal urinary excretion of Cu, Fe, Zn, Ca, Mn, Se, and Cr was higher in diabetic subjects, and basal urinary Cu excretion was closely correlated with that of Fe ([Fe]_{24 h urine} = 1.94.[Cu]_{24 h urine} + 0.53; r² = 0.48, P = 0.00067). Thus, increased basal urinary Cu excretion in diabetes is closely related to increased urinary Fe excretion. Diabetic subjects had elevated urinary volumes, but statistical modeling indicated that the increased urinary output was not a factor in the increased urinary elemental outputs.

The effects of drug treatment (placebo or trientine) and of interactions between trientine and metabolic status (control or diabetes) were examined by ANOVA (Table 3); Table 4 shows the effects of treatment on differences in balance and urinary and fecal excretion of Cu, Fe, Zn, and Ca.

Trientine did not modify the Cu balance in the whole study group, but the interaction term was significant (P = 0.0028) (Table 3). Trientine decreased the Cu balance in diabetic subjects compared with placebo (P = 0.021) (Table 4), but its effect on balance in control subjects was of marginal significance (P = 0.065) (Table 4). Basal urinary Cu was elevated in diabetic subjects (P = 0.0011) (Table 2). Trientine treatment modified urinary Cu excretion in all subjects (P < 0.0001), but the interaction was borderline (P = 0.075) (Table 3). Trientine stimulated urinary Cu excretion in both groups (P < 0.0001) (Table 4), indicating extraction of Cu^II in both. Moreover, its interaction with fecal Cu was significant (P = 0.0034) (Table 3). Fecal Cu effects were evoked largely via decreased control values (P = 0.0041) (Table 4). These observations are consistent with lower fractional absorption of Cu in diabetes.

Trientine modified the Fe balance (P = 0.028), but the corresponding trientine−metabolic status interaction term was not significant (Table 3). Trientine increased the Fe balance in control (P = 0.034) but not in diabetic subjects (Table 4). Basal urinary Fe was elevated in diabetic subjects (P = 0.0001) (Table 2), but was unaffected by trientine, which had only borderline effects in diabetic subjects (P = 0.051) (Table 4). Trientine modified fecal Fe in the whole group (P = 0.019) (Table 3), but the trientine−metabolic status interaction was insignificant. Trientine lowered fecal Fe excretion in control subjects (P = 0.023) (Table 4). These data suggest lower fractional absorption of Fe in diabetes.

Trientine elicited significant modification of the Zn balance in all subjects (P = 0.0021), and the trientine-metabolic status interaction was significant (P = 0.002) (Table 3). Basal urinary Zn was elevated in diabetic subjects (P < 0.0001) (Table 2). Trientine modified urinary Zn in all subjects (P < 0.0001) (Table 3) by stimulation in both control (P < 0.0001) and diabetic (P < 0.0001) subjects (Table 4). Trientine modified fecal Zn excretion (P < 0.0001) (Table 3), and the trientine−metabolic status interaction was significant (P = 0.0045). Trientine decreased fecal Zn in control subjects (P = 0.0001) (Table 4). These data suggest that fractional Zn absorption is lower in diabetes.

Trientine significantly modified the Ca balance (P = 0.0022) (Table 3), mainly by increases in control subjects (P = 0.0012) (Table 4). Drug−metabolic status interactions were significant for Ca balance (P = 0.033) and fecal excretion (P = 0.020) (Table 3). Trientine modified fecal Ca excretion (P = 0.0014) (Table 3), mainly via decreased control values (P = 0.00065) (Table 4). Fractional Ca absorption may thus be lower in diabetes.

The basal Mg balance did not differ between diabetic and control subjects, nor was it modified by trientine treatment. Indexes of Mg balance were uncorrelated with HbA_1c or FPG in either group (data not shown).

Trientine modified the Mn balance (P = 0.028) and fecal Mn excretion (P = 0.017) (Table 3). The trientine−metabolic status interaction was significant for Mn balance (P = 0.043) and fecal excretion (P = 0.034) (Table 3). Trientine altered the Mn balance mainly via decreased fecal excretion. The effects on Mn and Ca balance were similar, being evoked mainly via decreased fecal excretion in control subjects and consistent with decreased fractional gut absorption of Mn in diabetic subjects.

The trientine−metabolic status interaction was significant for Se balance (P = 0.016) and fecal excretion (P = 0.013) (Table 3). The effects on Se balance resembled those for Ca and Mn, being evoked mainly via decreased fecal Se excretion in control subjects. Trientine did not modify [Se]_serum (data not shown).

Neither diabetes nor trientine altered the Mo or Cr balance or their urinary or fecal excretion rates (data not shown). Neither FPG nor HbA_1c correlated with indexes of Cr balance in either diabetic or control subjects (data not shown).

Trientine increased urinary Cu in both diabetic ([Cu]_{24 h urine} = 0.00245.[trientine dosage] + 0.192; r² = 0.66, P < 0.0001) and nondiabetic ([Cu]_{24 h urine} = 0.00183.[trientine dosage] + 0.177; r² = 0.56, P < 0.0001) subjects in a dosage-dependent manner; gradients for Cu excretion did not differ between groups. Trientine stimulated urinary Zn in both diabetic ([Zn]_{24 h urine} = 0.016.[trientine dosage] + 14.5; r² = 0.54, P < 0.0001) and nondiabetic ([Zn]_{24 h urine} = 0.0097.[trientine dosage] + 4.32; r² = 0.42, P < 0.0001) groups in a dosage-dependent manner; the diabetic gradient was significantly greater (P < 0.05). These findings are consistent with those previously reported (13). Trientine had no dosage-dependent effect on 24-h urinary excretion of Fe, Mn, Ca, Mg, Mo, Se, or Cr (data not shown), nor did it modify [Mg]_serum over 10 h (data not shown).

We calculated a multivariate regression model relating basal variables to urinary Cu excretion during drug treatment (Table 5). Trientine and diabetes were positive factors; urinary Cu excretion on days 1–6 and basal [Mg]_serum were significant, positively continuous components and basal [ferritin]_serum was negative. The mechanisms relating basal [Mg]_serum and [ferritin]_serum to urinary Cu are uncertain.

In this study, we found that several aspects of Cu metabolism were altered in diabetic subjects. Basal urinary Cu was 1.4-fold higher in diabetic than in control subjects, whereas [Cu]_serum did not differ significantly between the two groups. These findings are consistent with some previous reports (14,15), but others have reported that diabetic subjects with complications have elevated [Cu]_serum (16).

There was a trend for the Cu balance to be elevated in diabetic subjects, as well as a significant effect of interaction between trientine and diabetes on Cu balance. Trientine markedly stimulated urinary Cu in both subject groups, but lowered the Cu balance only in diabetic subjects. Urinary Cu excretion during drug treatment was positively correlated with pretreatment urinary Cu levels. Thus, elevated basal urinary Cu predicted drug-induced cupriuresis; individual responses may have been determined by prior systemic Cu^II accumulation. In contrast, trientine decreased fecal Cu in control subjects only, possibly through increased uptake, consistent with known actions of polyvalent chelators to increase metal absorption (17). An alternative mechanism that could contribute to the observed effects of trientine on fecal metal excretion is that diabetes may modify hepatobiliary excretion of some or all of the elements concerned, although we are unaware of data to support this hypothesis. Thus, regulation of Cu homeostasis differed significantly between groups, and trientine elicited effects to reverse the elevated Cu balance in diabetic subjects, mainly through stimulation of urinary Cu excretion.

Ceruloplasmin (ferro-O₂-oxidoreductase; EC 1.16.3.1) is a circulating Cu protein present in vertebrate plasma (18). Normally, ≥95% of plasma Cu is ceruloplasmin bound, and levels of circulating Cu and ceruloplasmin are closely related (19). These are the biomarkers most frequently used as measures of Cu status, and both are depressed in severe Cu deficiency (20). However, levels plateau when Cu intake is adequate and do not reflect the magnitude of Cu intake beyond this point, so they are not useful for characterizing Cu excess states (21). In one study, neither biomarker decreased in humans fed a marginally low Cu diet for >3 months (22). Based on our study, we conclude that measurements of isolated [Cu]_serum or [cerulo-plasmin]_serum values are unlikely to be informative concerning the presence of probable systemic Cu^II excess in diabetic subjects.

Although Cu is an essential trace nutrient, it is also a potent cytotoxin when excess amounts accumulate in tissues (23). Because of its redox chemistry, Cu^II readily participates in reactions that elicit ROS production (4,24). Possible roles for Cu-catalyzed redox reactions in diabetes complications have been discussed previously (3,5), and the relative importance of Cu in diabetes in vivo was shown when we demonstrated the reversal of heart failure by systemic Cu^II chelation in diabetic rats (8).

In our study, EC-SOD activity was elevated in diabetic subjects and correlated with an HbA_1c-[Cu]_serum interaction. This finding is consistent with a mechanism whereby elevated EC-SOD is related to a [Cu]_serum−chronic hyperglycemia association, with the latter factor being accepted as a driver of the tendency to develop diabetes complications (25). EC-SOD, a secretory glycoprotein, is the major SOD isoenzyme in extracellular fluids (26) and blood vessel walls (27). Activity and concentrations of serum EC-SOD are known to be elevated in diabetic subjects (26,28), and a significant association between the serum concentration of EC-SOD and the severity of vascular complications has been previously reported (29). A correlation between [Cu]_serum and EC-SOD activity has been reported in humans with essential hypertension (30). Serum EC-SOD activity is reportedly correlated with diabetes duration, carotid artery intimal-media thickness, and severity of nephropathy and retinopathy. It has also been proposed as a marker of vascular injury, possibly reflecting hyperglycemia-induced oxidative injury to the vascular endothelium (29). Nitric oxide (NO), a key physiological vasodilator (31), reacts with superoxide anion at an extremely rapid rate (1,27). The balance between superoxide anion concentrations and cellular antioxidant capacity, particularly SOD activity, is likely to regulate the bioactivity of NO (31). As the major SOD isoform present in vascular endothelium, where it acts to regulate superoxide levels, EC-SOD is a key regulator of endothelium-derived NO bioactivity in blood vessels (27).

Elevated EC-SOD levels (26,28) reflect increased superoxide production in diabetes (32). The finding of decreased EC-SOD after trientine treatment is consistent with the suppression of vascular superoxide production, and our current study implicates an interaction between [Cu]_serum and chronic hyperglycemia in the mechanism by which diabetes causes vascular damage. Trientine treatment could suppress intravascular consumption of NO by lowering vascular superoxide production, thereby enhancing physiological vasodilatation, which is defective in diabetes (32). Trientine treatment lowered EC-SOD activity to control values in diabetic subjects. These data, taken in conjunction with data from our previous study demonstrating that chronic trientine treatment reverses established heart failure in diabetic rats and left ventricular hypertrophy in diabetic humans and rats (8), support the therapeutic lowering of systemic Cu^II for suppression of vascular damage. Because disease mechanisms in hypertensive heart disease, ischemic cardiomyopathy, and atherosclerosis are similar to those in diabetic heart disease (1,7), we expect that our results could also prove relevant in these other related conditions (8).

Basal balances of nine essential elements (33–35) were equivalent between study groups, although the basal urinary excretion rates for Cu, Fe, Zn, Ca, Mn, Se, and Cr were significantly elevated in the diabetic group. Trientine treatment significantly increased the Zn balance in control subjects, mainly via decreased fecal excretion, whereas it stimulated urinary Zn output in both groups. Previous findings of hyperzincuria and lower Zn absorption in diabetic animals and humans have prompted conjecture that diabetic subjects might be more susceptible to Zn deficiency (36); however, others have reported increased tissue Zn values after trientine treatment (37). Apparent absorption and retention of Cu and Zn have been reported in diabetic rats in which food consumption was twice that of controls, and the fractional absorption of Zn and Cu was reported as being lower. In one study (37), net absorption was higher, although it was offset by higher urinary excretion so that the final Cu/Zn retention was similar in both groups. These findings are similar to our results. Low fractional absorption of Zn in diabetic rats has been attributed to lower intestinal transport associated with increased concentrations of intestinal metallothionein, an inhibitor of Cu and Zn transport (36). The increased metallothionein content in the enterocytes of diabetic patients is a plausible mechanism for the lesser degree of stimulation by trientine treatment of Cu and Zn uptake in diabetic subjects. This mechanism could protect against increased systemic uptake that might otherwise result from diabetic hyperphagia. We observed similar disparities in the effects of trientine on the uptake of Ca and Fe between control and diabetic subjects. Similar mechanisms may thus operate in diabetes to lessen the increased dietary Ca and Fe absorption that could otherwise result from hyperphagia. However, these effects are not likely to be mediated through increased enterocyte metallothionein, which appears not to bind to either Fe or Ca but operate through different, specific mechanisms.

Consistent with previous reports (38), mean serum ferritin was increased in diabetic subjects, although other measures of Fe homeostasis including serum Fe and IBC (39) were not different from control values. Basal urinary Fe excretion was also elevated in diabetic subjects, whereas fecal Fe and the Fe balance were similar. These observations could be explained by the increased metal uptake secondary to relative hyperphagia in the diabetic subjects before the imposition of strict dietary control during the controlled balance studies. Trientine’s effects on uptake of Ca, Mn, and Se were similar to those for Fe. At present, however, the significance of high plasma ferritin concentrations in a subset of diabetic patients remains unclear (40), as does the mechanism of the negative correlation between ferritin and [Cu]_urine observed here in trientine-treated diabetic subjects. In the Third National Health and Nutrition Examination Survey (1988–1994), an increased risk of diabetes was concentrated among participants with the lowest transferrin saturation concentrations; it was reported that this association was less likely to be explained by the Fe overload hypothesis and instead may be caused by inflammation as the mechanism by which ferritin becomes elevated in diabetes (38).

In our study we have shown that several alterations in the regulation of Cu homeostasis occur in diabetic humans. How might Cu and chronic hyperglycemia conspire to cause cardiovascular damage in diabetes? Ceruloplasmin and serum albumin are the main Cu-binding proteins in plasma (41), and there is some evidence that chronic hyperglycemia can damage the Cu-binding properties of both (42,43). Results from our recent studies have indicated that diabetes might cause two- to threefold increments in extracellular matrix Cu (8). Furthermore, incubation of ceruloplasmin with glucose reportedly causes fragmentation and time-dependent release of its bound Cu^II, which then appears to participate in a Fenton-type reaction to produce hydroxyl radicals (42). These data were interpreted as consistent with the conjecture that ROS may form by the Maillard reaction and in turn generate hydroxyl radicals via a Cu-dependent Fenton-type reaction (42). In addition, glycated serum albumin reportedly becomes pro-oxidant in the presence of Cu, probably through the generation of ROS (44). Damage by chronic hyperglycemia to Cu-regulating mechanisms through glycation of ceruloplasmin and albumin could lead to elevated concentrations of catalytically active Cu^II in plasma, but this hypothesis requires support by direct observation. Cu binds to transferrin in addition to ceruloplasmin and albumin (45); this interaction could also be important in the association of Cu, EC-SOD, and HbA_1c.

In summary, the regulation of Cu homeostasis was altered in the diabetic subjects in this study, who demonstrated elevated rates of urinary Cu excretion and a tendency to increased Cu balance compared with control subjects. Treatment with the Cu^II-selective chelator, trientine, lowered Cu balance in the diabetic subjects, in whom elevated EC-SOD was strongly correlated with the [Cu]_plasma-HbA_1c interaction, implicating this in diabetic tissue damage. Trientine suppressed elevations of EC-SOD in diabetic subjects. In conjunction with therapies aimed at decreasing hyperglycemia, the use of compounds that can remove systemic Cu^II is proposed as a therapy for suppressing the tendency to cardiovascular complications in diabetes.

This work was supported by the Endocore Research Trust and Protemix.

We thank T.B. Mulvey, G. Allen, S. Thornber, P. Misur, G. Muir, G. Kerr, G. Robinson, and P. Baker for assistance; and M.S. Cameron-Cooper, K.R. Mansford, P.J. Scott, T. Brittain, C.A. Tse, A.R. Phillips, P.D. Boyd, K. Wichmann, and D.J. Crossman for their discussions with us. G.C. acknowledges program support from the Foundation of Research, Science and Technology, New Zealand and the Health Research Council of New Zealand.