The Effect of Cyclooxygenase-2 Inhibition on Renal Hemodynamic Function in Humans With Type 1 Diabetes

Subjects who fulfilled the following inclusion criteria were asked to participate: duration of type 1 diabetes >5 years, Tanner Stage 4–5 puberty, normoalbuminuria (albumin excretion rate [AER] <20 μg/min on two to three overnight urine collections obtained during the month before study), normal clinic blood pressure, no microvascular disease, and absence of chronic illness other than treated hypothyroidism or mild asthma. The Research Ethics Board at the Hospital for Sick Children (Toronto, Canada) approved the protocol. All subjects and their parents gave informed consent. Twenty-one subjects with type 1 diabetes were eligible to participate.

Subjects adhered to a sodium-replete (>150 mmol/day) and moderate protein (1.5–2.0 g · kg⁻¹ · day⁻¹) diet during the 7-day period before each experiment (Fig. 1). A 24-h urine collection on day 6 was used to evaluate dietary adherence through the determination of urinary sodium and urea excretion. Protein intake was calculated from the urea excretion using standard methods (15). Subjects were studied if the excretion of sodium and urea was >150 mmol/day and 3–6 mmol · kg⁻¹ · day⁻¹, respectively. All subjects met these study criteria.

Euglycemic (blood glucose 4–6 mmol/l) and hyperglycemic (blood glucose 9–11 mmol/l) conditions were maintained on 2 consecutive days using a modified glucose clamp technique as described previously (15). Studies were performed after the overnight glucose clamp before and then after 14 days of COX2 blockade using 200 mg celecoxib once daily. During the 14 days of COX2 blockade, subjects continued with their standard diabetes management routines. Euglycemic and hyperglycemic conditions were maintained for ∼10 h preceding and during all investigations. Female subjects were initially studied during the late-follicular phase of the menstrual cycle, determined by counting days and measuring 17β-estradiol levels, and were again studied during the same phase of a subsequent menstrual cycle. None of the female subjects used oral contraceptive medications.

Subjects were admitted to the Clinical Investigation Unit at the Hospital for Sick Children the evening before each day of the study, as described previously (15). A 16-gauge peripheral venous cannula was inserted into a vein in one arm for infusion of glucose and insulin, and a second blood sampling line was inserted into a vein in the contralateral arm. Blood glucose was measured every 30–60 min, and the insulin infusion was adjusted to maintain euglycemia. Studies were conducted the following morning after an overnight fast with subjects lying in the supine position in a warm quiet room. A third intravenous line was inserted into the arm contralateral to the insulin infusion and was connected to a syringe infusion pump for infusions of inulin and paraaminohippurate (PAH). Mean arterial pressure (MAP) and heart rate were measured before each blood sample throughout the study by an automated sphygmomanometer (Dinamapp). After collecting blood for inulin and PAH blank, a priming infusion containing 25% inulin (60 mg/kg) and 20% PAH (8 mg/kg) was administered. Thereafter, inulin and PAH were infused continuously at a rate calculated to maintain their respective plasma concentrations constant at 20 and 1.5 mg/dl. Subjects remained supine at all times. After a 90-min equilibration period, blood was collected for plasma renin, aldosterone and angiotensin II (Ang II), inulin, PAH, and hematocrit (Hct). Blood was further collected every 30 min for 60 min for inulin and PAH, and GFR and effective renal plasma flow (ERPF) were estimated by steady-state infusion of inulin and PAH according to the calculation method described by Schnurr et al. (16).

Subjects were then initiated on celecoxib. After a total of 14 days, subjects were readmitted to the Clinical Investigation Unit and again underwent renal hemodynamic testing using inulin and PAH clearance after overnight euglycemic and hyperglycemic clamping.

Blood samples collected for inulin and PAH determinations were immediately centrifuged at 3,000 rpm for 10 min at 4°C. Plasma was separated, placed on ice, and then stored at −70°C before the assay. Inulin and PAH were measured in serum by colorimetric assays using anthrone and N-(1-naphthy) ethylenediamine, respectively. The mean of two baseline clearance periods represent GFR and ERPF, expressed per 1.73 m². Filtration fraction was determined as the ratio of GFR to ERPF. Renal blood flow (RBF) was calculated by dividing the ERPF by (1 − Hct). Renal vascular resistance (RVR) was derived by dividing the MAP by the RBF. All renal hemodynamic measurements were adjusted for body surface area.

Ang II was measured by radioimmunoassay. Blood was collected into prechilled tubes containing EDTA and angiotensinase inhibitor (0.1 ml Bestatin Solution; Buhlmann Laboratories, Basal, Switzerland). After centrifugation, plasma samples were stored at −70°C until analysis. On the day of analysis, plasma samples were extracted on phenylsilysilica columns. A competitive radioimmunoassay kit supplied by Buhlmann Laboratories AG (Switzerland) was used to measure the extracted Ang II. Aldosterone was measured by radioimmunoassay, using the Coat-A-Count system. Active plasma renin was measured by two-site immunoradiometric assay where two monoclonal antibodies to human active renin are used. One antibody was coupled to biotin while the second was radiolabeled for detection. The sample containing active renin was incubated simultaneously with both antibodies to form a complex. The radioactivity of this complex was directly proportional to the amount of immunoreactive renin present in the sample (17).

To initially determine whether celecoxib had a pharmacological effect on urinary prostanoids, urine samples were collected from a subset of 10 subjects (5 hyperfiltration and 5 normofiltration subjects). Prostaglandin E (PGE) metabolite (11-deoxy-13,14-dihydro-15-keto-11-β, 16 ε-cyclo-PGE₂) levels were detected by competitive enzyme immunoassays (Cayman Chemical) to study the effect of COX2 inhibition on these metabolites, as previously described (18).

To determine urinary between-group differences in the excretion in prostanoids after COX2 inhibition, mass spectrometry, which is a more sensitive technique, was used on urine obtained after the overnight glucose clamps. Prostanoids were quantified using well-established techniques, as described previously (19–23). In brief, 11-deoxy-13,14-dihydro-15-keto-11-β, 16 ε-cyclo-PGE₂ (PGE₂), 9α,11β-dihydroxy-15-oxo-2,3,18,19-tetranor-prost-5-ene-1,20-dioic acid (PGD), 2,3-dinor-6-keto-PGF₁α (PGI), and 11-dehydro-TxB₂ (TXB2) were assayed in urine using a stable isotope dilution method with gas chromatography followed by detection with negative ion chemical ionization mass spectrometry using selective ion monitoring (19–22). The urinary creatinine concentration was measured by the sodium picrate method with an AutoAnalyzer II (Technicon, Tarrytown, NY), and all prostanoid assays were corrected for creatinine (nanograms prostanoid per milligram creatinine).

Urinary AER was determined from three timed overnight urine collections. Urinary albumin concentration was determined by immunoturbidimetry (24). A1C was measured by high-performance liquid chromatography.

Subjects were assessed in the Diabetes Clinic 7 days after starting celecoxib to measure weight and blood pressure and to draw blood for liver and renal function testing, complete blood count and differential, electrolytes, and urine dipstick. No subjects were withdrawn from the study because of adverse effects.

The data were analyzed on the basis of filtration status. Results are presented as means ± SE. Between-group comparisons of all parameters at baseline were made using parametric methods (unpaired t test). Within-subject and between-group differences in the response to COX2 inhibition were determined by repeated-measures ANOVA. All statistical analyses were performed using the statistical package SPSS (SPSS for graduate students, version 14.0).

During the baseline euglycemic clamp, 12 subjects exhibited baseline GFR values of <135 ml/min per 1.73 m² (normofiltration group) and 9 subjects exhibited hyperfiltration (hyperfiltration group, GFR ≥135 ml/min per 1.73 m²) (Table 1). Age, diabetes duration, BMI, AER, 24-h sodium and protein intake, and A1C were similar in the two groups (Table 1). No differences in serum albumin were present before or after COX2 inhibition. The female participants had a mean 17β-estradiol level of 218 ± 67 pmol/l, consistent with the late-follicular phase of the menstrual cycle, and were followed up during the same phase of the menstrual cycle after COX2 inhibition. After the overnight euglycemic clamp, baseline circulating RAS components, including aldosterone, Ang II, and renin, were similar in the hyperfiltration and normofiltration groups (Table 4 ). Insulin levels were similar on the morning of the euglycemic clamp (114 ± 30 in hyperfiltration vs. normofiltration 145 ± 45 pmol/l, NS). As reported in Table 3, the hyperfiltration group exhibited a mean GFR during clamped euglycemia that was significantly higher than in the normofiltration group. Filtration fraction was also significantly augmented in the hyperfiltration group. No significant differences were noted in MAP, RBF, ERPF, or RVR (Table 3).

COX2 inhibition was not associated with changes in urinary sodium excretion in either group (Table 3). While the 2-week period of COX2 inhibition did not change MAP in either group (Table 3), there were marked differences in the renal hemodynamic response between the two groups, in that GFR declined significantly (P = 0.049) in the hyperfiltration group (Table 3; Fig. 2) and increased in the normofiltration group (P = 0.001). The normofiltration group exhibited a significant rise in filtration fraction (P = 0.042). The between-group differences in GFR (P = 0.0001) and filtration fraction (P = 0.046) in response to COX2 inhibition were significant.

After COX2 inhibition, plasma aldosterone levels decreased significantly in normofiltration subjects (P = 0.027) (Table 4). In addition, Ang II levels were significantly lower in normofiltration compared with hyperfiltration subjects after COX2 inhibition (4.7 ± 1.1 pg/ml in hyperfiltration subjects vs. 1.8 ± 0.6 pg/ml in normofiltration subjects, P = 0.017). Renin and aldosterone levels were not significantly different between groups before or after celecoxib. Insulin levels remained similar between the two groups after COX2 inhibition (127 ± 43 pmol/l in hyperfiltration vs. 108 ± 18 pmol/l in normofiltration subjects). By enzyme-linked immunosorbent assay (ELISA), in the subset of subjects, PGE metabolite levels decreased significantly from 3.4 ± 0.4 to 1.8 ± 0.6 pg/ml after COX2 inhibition (P = 0.04). To confirm these findings and to determine the complete prostanoid profile in each group, we used a separate mass spectrometry technique and demonstrated that COX2 inhibition was associated with reductions in prostanoid-related vasodilators, primarily in the normofiltration group. Similar trends were observed in hyperfiltration subjects, but these changes did not reach statistical significance (Table 5). COX2 inhibition did not have a significant effect on the excretion of thromboxane metabolite levels in either group.

Hyperglycemia induced similar increases in urinary sodium excretion in the two groups (Table 2). GFR in the hyperfiltration group increased during clamped hyperglycemia (162 ± 12 ml/min per 1.73 m²) when compared with the euglycemic baseline (150 ± 5 ml/min per 1.73 m²), although this did not reach statistical significance. Hyperglycemia induced a significant rise in ERPF in this group (Table 2). Before COX2 inhibition during hyperglycemia, hyperfiltration subjects also exhibited higher mean values for ERPF and RBF and a lower RVR compared with normofiltration subjects (Table 2). In the normofiltration group before COX2 inhibition, hyperglycemia was associated with a significant rise in GFR (Table 2).

Similar to their euglycemic response to COX2 inhibition, after celecoxib during hyperglycemia, the hyperfiltration group demonstrated declines in GFR and filtration fraction (NS). In the normofiltration group, COX2 inhibition did not abolish the hyperglycemia-mediated increase in GFR. Instead, after COX2 inhibition during hyperglycemia, the normofiltration group exhibited a further augmentation in GFR compared with baseline hyperglycemic parameters (Fig. 3).

COX2 inhibition was also associated with nonsignificant reductions in RAS mediators in both the hyperfiltration and normofiltration groups (NS). PGE metabolite (1.8 ± 0.2 to 1.1 ± 0.3, NS) levels decreased after COX2 under ambient hyperglycemic conditions, although this did not reach statistical significance. Similar to our results during euglycemia, mass spectrometry demonstrated reductions in vasodilatory prostanoids in both groups, but this was only statistically significant in the normofiltration group (Table 5). Thromboxane excretion was not affected in either group.

Previous studies in humans have associated vasodilatory prostaglandins with glomerular hyperfiltration and have used nonspecific COX1/COX2 inhibitors to probe renal hemodynamic function (12). Komers et al. (9) subsequently demonstrated the critical role of specific COX2 inhibition on renal hemodynamic function in experimental animal models of diabetes, suggesting that the inhibition of the COX2 isoform of the enzyme may be critically important in animal and human models of diabetes (13,14). We therefore used a selective COX2 inhibitor as a physiological probe to elucidate the role of renal COX2 metabolites in renal hemodynamic function in a well-characterized (6) group of subjects with uncomplicated type 1 diabetes. In addition, we wanted to examine the role of COX2 in the hyperglycemia-mediated increase in GFR. In accord with our previous studies, we divided the subjects into two groups based on GFR during euglycemic conditions for analysis (6). The major findings were as follows: 1) A 2-week period of COX2 inhibition significantly reduced but did not normalize GFR in the hyperfiltration group during clamped euglycemia. 2) In contrast, the 2-week period of COX2 inhibition significantly increased GFR in the normofiltration group during clamped euglycemia. 3) COX2 inhibition reduced but did not eliminate the hyperglycemia-mediated rise in GFR in the hyperfiltration group and augmented the rise in GFR in the normofiltration group. 4) The pharmacological effect of COX2 inhibition was present and associated with declines in vasodilatory PGE in a subgroup of subjects by ELISA. Mass spectrometric analysis of urine in each group confirmed this finding and demonstrated reductions in vasodilatory prostanoids, a finding that was more prominent in the normofiltration group.

COX2 is constitutively expressed in several locations in the kidney, including the loop of Henle, renal arterioles, and the macula densa, and undergoes inducible expression during inflammation (25,26). Intrarenal COX2 expression is important in the context of diabetes because its expression is augmented in the renal cortex of rats with streptozotocin-induced diabetes, largely due to the effect of hyperglycemia (9,27). Moreover, COX2 inhibition decreases renal hyperfiltration, proteinuria, and glomerulosclerosis in a variety of animal models (28–30). These effects are thought to be due to COX2 inhibition-mediated reductions in vasodilatory prostaglandins, such as prostacyclin and PGE (30), leading to decreased intraglomerular capillary pressure. Some data support the hypothesis that COX2-derived prostaglandins exert vasodilatory actions at the renal arteriolar level in animals and humans (31,32). In humans with type 1 diabetes, previous studies have correlated renal hyperfiltration with elevations in urinary excretion of vasodilatory prostaglandins (33,34). Hommel et al. (12) were also able to reduce GFR in response to nonspecific COX inhibition. Furthermore, studies have shown important COX2-RAS interactions. For example, COX2 inhibition is associated with the inhibition of renin synthesis (35–37), and COX2-RAS interactions have been associated with important hemodynamic effects in animal models (38). Taken together, these studies suggest that COX2 may be an important determinant of renal hemodynamic function in humans with diabetes, although the effect of selective COX2 inhibition on renal hemodynamic function has not been studied in humans with type 1 diabetes. The differential effect of selective COX2 inhibition in hyperfiltration and normofiltration subjects has also not yet been studied and was an aim of the present study.

The first major observation was that the renal hemodynamic response to COX2 inhibition was dependent on the GFR during clamped euglycemia. In accordance with experimental studies, such as the comprehensive series of experiments performed by Komers et al. (9), COX2 inhibition reduced GFR in hyperfiltration subjects and reduced PGE metabolite levels. Despite the maintenance of normal ambient glycemia and the presence of COX2 inhibition, we were unable to normalize the GFR to <135 ml/min per 1.73 m². These findings, in combination with the observation of numerical reductions in vasodilatory prostanoids by mass spectrometry, suggest that, in hyperfiltration subjects, COX2-related factors only partially maintain renal hyperfiltration. Taken together with our previous RAS blockade studies (2,6), this hemodynamic response indicates that COX2 is an additional factor that contributes to the hyperfiltration state in humans with uncomplicated type 1 diabetes (11).

In contrast to the response in hyperfiltration subjects, the 2-week period of COX2 inhibition resulted in a significant rise in GFR in normofiltration subjects, suggesting that COX2 products are important in maintaining normofiltration in normofiltration subjects. This effect could not have been readily predicted based on previous work because nondiabetic control subjects were used in these studies (14,34). The rise in filtration fraction that we observed in normofiltration subjects is suggestive of a vasoconstrictive effect that could be on the basis of either inhibition of vasodilators (such as PGE, PGD, or PGI metabolites), leaving a predominance of vasoconstrictive factors within the renal microcirculation or the activation of vasoconstrictors, such as thromboxanes or Ang II. The dominant effect of COX2 inhibition in animals and humans is a suppression of vasodilatory prostaglandins (25,26,39–42). Consistent with these previous observations and to assess the pharmacological effect of COX2 inhibition on vasodilatory prostanoids, we initially demonstrated that PGE metabolite levels decreased significantly in a sample of our subjects under euglycemic conditions in response to COX2 inhibition. A subsequent between-group analysis using mass spectrometry supported this theory, in that vasodilatory prostanoids, including PGD and PGI, decreased significantly under euglycemic and hyperglycemic conditions after COX2 inhibition. These observations support the hypothesis that the COX2 inhibition-mediated rise in GFR and filtration fraction in normofiltration subjects was due to the blockade of glomerular vasodilators.

Alternatively, from the vasoconstrictor perspective, previous studies in animal models have demonstrated significant reductions, not increases, in the excretion of vasoconstrictive thromboxanes with administration of selective COX2 inhibitors (9,43). The effect of COX2 on thromboxane production is controversial in human studies (44). For example, McAdam et al. (44) recently demonstrated a suppressive effect of rofecoxib on thromboxane in nondiabetic smokers. However, other data have failed to demonstrate a significant effect of COX2 inhibition on thromboxane production, particularly in nonsmokers (26,39,44), such as the subjects in our study. Similar to these previous studies, we failed to observe significant changes in TXB2 levels by mass spectrometry. Regarding RAS activity, COX2 inhibition is thought to block macula densa-derived renin production, thereby reducing intrarenal RAS activity (36,37). Consistent with these physiological effects, we observed numerical decreases in circulating PRA, aldosterone, and Ang II in response to COX2 inhibition. Taken together, these data suggest that the rise in GFR that we observed in the normofiltration group was most likely due to a selective reduction in renal vasodilators rather than an activation of vasoconstrictors. Alternatively, COX2 inhibition may have been associated with changes in the glomerular ultrafiltration coefficient (Kf) or tubular pressures leading to changes in GFR or filtration fraction (10,45,46). However, the role of Kf was likely limited because experimental literature suggests that pharmacological manipulation of the renal microvasculature alters Kf by <5% (47), making this hypothesis less likely.

Based on previous work, hyperglycemia augments COX2 expression and COX2 inhibition reduces hyperfiltration in experimental diabetes (9). We therefore expected COX2 inhibition to blunt any hyperglycemia-related renal hemodynamic responses. Once again, the effect of COX2 inhibition was critically dependent on the GFR during baseline clamped, euglycemic conditions. In the hyperfiltration group, COX2 inhibition only partially blunted the increase in GFR. In the normofiltration group, COX2 inhibition resulted in a further rise in GFR. The response in the normofiltration group could have occurred through several hemodynamic mechanisms. One possible explanation for the rise in GFR that was observed after COX2 inhibition in normofiltration subjects was RAS activation leading to vasoconstriction and a rise in GFR, filtration fraction, and RVR, suggestive of an increase in postglomerular constriction (36,37). This is unlikely because COX2 inhibition was associated with a decline in circulating RAS mediators. Our results are therefore most compatible with the hypothesis that the renal hyperfiltration effect was on the basis of COX2 inhibition-mediated blunting of vasodilators rather than the activation of efferent vasoconstrictors. COX2 inhibition–mediated declines in vasodilatory prostaglandins may have allowed intrarenal vasoconstrictors to then act in an unopposed fashion. This remains speculative, however, and the pathogenesis of this additive effect of COX2 inhibition to hyperglycemia on GFR in normofiltration group requires further investigation. Our observations emphasize that the regulation of renal hemodynamic function is not uniform in subjects with type 1 diabetes (6).

This study has important limitations. The sample size was small, which may have limited our ability to detect significant differences in some parameters, such as within- group changes in urinary prostaglandins (PGE), and also may explain the lack of statistical significance in the suppression of PRA by COX2 inhibition. We attempted to minimize the effect of the small sample size by using homogeneous study groups and by careful prestudy preparation with a focus on known factors such as ambient glycemia and sodium and protein intake, which influence the RAS and GFR (1,15,48–51). We studied women in a defined phase of the menstrual cycle and did not include users of oral contraceptive medications. We also decreased variability by using a study design that allowed each subject to act as his/her own control. Lastly, our findings may not relate to subjects with more advanced type 1 diabetes or microalbuminuria or to subjects with type 2 diabetes. These states require further investigation.

In summary, we used a specific COX2 inhibitor to examine the role of COX2 in the control of the renal microcirculation in subjects with diabetes. We demonstrated that COX2 may in part contribute to the hyperfiltration state in the hyperfiltration group. Moreover, COX2 may be an important factor that serves to maintain normofiltration in normofiltration subjects. Finally, our findings confirm our previous observations and further emphasize that hyperfiltration subjects are physiologically distinct from normofiltration subjects and are not simply divergent based on dietary or glycemic differences.

D.Z.I.C. has received funding from The Kidney Research Scientist Core Education and National Training Program (sponsored by the Canadian Institutes of Health Research, Kidney Foundation of Canada, and Canadian Society of Nephrology) and the Clinician Scientist Program at the University of Toronto. J.A.M. and E.B.S. have received an operating grant from the Juvenile Diabetes Research Foundation. J.W.S. is the CIHR/AMGEN Canada Kidney Research Chair at the University Health Network, University of Toronto.

We thank the nurses in the Clinical Investigation Unit, Hospital for Sick Children, and in particular Maria Maione for her invaluable assistance with the protocol. We also thank Clinalfa for providing PAH for these experiments.