Dissection of Metabolic, Vascular, and Nerve Conduction Interrelationships in Experimental Diabetic Neuropathy by Cyclooxygenase Inhibition and Acetyl-l-Carnitine Administration

Barrier-sustained, Cesarean-delivered male Wistar rats (200–300 g) were acclimatized for 1 week before being fasted overnight and rendered diabetic by an intraperitoneal (i.p.) injection of streptozotocin (STZ) (45 mg/kg) (Upjohn, Kalamazoo, MI) in 0.2 ml of 10 mmol/l citrate buffer, pH 5.5. Diabetes was defined as nonfasting plasma glucose ≥200 mg/dl in tail vein blood (One Touch II; Lifescan, Milpitas, CA) 48 h after STZ injection. Animals were subsequently randomly assigned to the experimental groups listed in Table 1, maintained for 4 weeks in individual air-filtered metabolic cages with ad libitum access to water, and fed a standardized rat diet (ICN Biomedicals, Cleveland, OH). After measurement of MNCV and NBF, animals were killed by exsanguination, and the sciatic nerves were rapidly excised and cleaned for biochemical measurements. Flurbiprofen, a nonselective COX-1 and -2 inhibitor (relative half-maximal inhibitory concentration [IC₅₀], COX-2 vs. COX-1: 5–10 [36,37]), was administered in the drinking water at a dose of 6 mg · kg⁻¹ · day⁻¹, with the concentration reduced in the STZ-induced diabetic (STZ-D) animals (5:1) to compensate for their increased water consumption. ALC was given by daily gavage at a dose of 50 mg · kg⁻¹ · day⁻¹ (using water as vehicle), which is the lowest dose that corrects both nerve Na,K-ATPase activity and MNCV (33) and increases nerve PGE₁ levels (38). Laser-Doppler flux measurements may underestimate the endoneurial nutritive component of NBF, which may critically determine endoneurial oxygenation (3,39, 40). Therefore, separate experiments (experiments 2 and 4) employed hydrogen clearance, which distinguishes nutritive from nonnutritive (shunt) NBF (40), to evaluate total versus endoneurial NBF effects of flurbiprofen, ALC, and meloxicam in nondiabetic (ND) and STZ-D rats.

Additional experiments (3 and 4) were performed to explore the potential role of diabetes-induced increased COX-2 flux on nerve organic osmolyte content, Na, K-ATPase activity, NBF, and MNCV in EDN. The selective COX-2 inhibitor meloxicam (relative IC₅₀, COX-2 vs. COX-1: 0.08 [41,42]) was given for 4 weeks (1 mg · kg⁻¹ · day⁻¹ in drinking water) to ND and STZ-D rats, with the dose adjusted in STZ-D animals to ensure equality of dosing. The initial study evaluated the effects of meloxicam on nerve metabolic end points, and a second study used expanded animal groups to explore potential effects of meloxicam on nerve electrophysiology and NBF. Meloxicam was a gift from Boehringer Ingelheim Pharmaceuticals (Ingelheim/Rhein, Germany).

In all studies, investigators performing the measurements were unaware of treatment group assignments. The animal study protocols were approved by the University of Michigan Committee on Use and Care of Animals.

Sciatic nerve MI, sorbitol, and fructose were determined by gas-liquid chromatography of aldonitrile acetate derivatives as previously described (17). Nerve taurine was measured by reverse-phase high-performance liquid chromatography (43). Standard curves were generated daily, and the recovery-corrected values are expressed as nanomoles per milligram wet weight of tissue.

Ouabain-inhibitable Na,K-ATPase activity, defined as the difference in activity before and after the addition of ouabain, was measured as previously described (17) and is expressed as micromoles ADP formed per hour per gram wet weight of tissue.

MNCV was measured before the determination of NBF in the contralateral limb in rats anesthetized with i.p. urethane (1–1.2 g/kg) as previously described (17). Body temperature was monitored by a rectal probe and maintained at 37°C with a warming pad. Hindlimb skin temperature was also monitored by a thermistor and maintained at 36–38°C by radiant heat.

In experiments 1 and 3, total sciatic NBF was assessed in anesthetized rats using a laser-Doppler blood flow monitor (MBF3/D; Moor Instruments, Devon, U.K.) (44). The left carotid artery was cannulated and connected to a transducer to monitor mean systemic blood pressure, which was recorded and displayed continuously on a computer screen. The sciatic nerve was exposed via a small incision on the left flank, and the laser probe (tip diameter 0.85 mm) was applied just in contact with an area of the sciatic trunk as free as possible from large epi- or perineurial blood vessels. Care was taken not to compress the nerve. The exposed nerve was then covered with liquid paraffin to avoid tissue dehydration, and the flux was allowed to reach a stable baseline over 10–15 min before readings were taken. Body temperature was maintained at 37°C as described above, and the temperature in the vicinity of the nerve was maintained by placing the rat under a source of radiant heat. Biofeedback from the rat was derived from a thermistor placed near the exposed sciatic nerve and was set to maintain local temperature at 36–38°C. Four flux measurements over 2 min each were obtained from the same section of nerve, and the mean value was calculated as arbitrary Doppler units. Each of the four Doppler flux measurements was the mean value of a 2-min recording period; the rats were not artificially ventilated during Doppler measurements. To take into consideration perfusion pressure variations, all flow data are expressed as vascular conductance, calculated by dividing NBF by the mean systemic blood pressure over the recording period.

In experiments 2 and 4, endoneurial nutritive NBF was assessed by hydrogen clearance as previously described (40,45) in rats anesthetized with an i.p. injection of urethane (1 g/kg). After the experiment, mono- or biexponential clearance curves were fitted to the data (Graphpad Software, La Jolla, CA). Nutritive NBF was taken as the slow component of the curve. An average of two determinations at different sites was used to determine nutritive NBF.

Fresh frozen rat sciatic nerve (RSN) was lysed in radioimmunoprecipitation assay (RIPA) buffer (Tris-HCl 50 mmol/l, NaCl 150 mmol/l, EGTA 1 mmol/l, Nonidet P-40 1%, Na-deoxycholate 0.015%, aprotinin, leupeptin, pepstatin 1 μg/ml, and phenylmethylsulfonyl fluoride 1 mmol/l), and protein concentrations were measured by bicinchoninic acid protein assay kit (Sigma, St. Louis, MO). Forty-microgram aliquots of protein extract were separated by electrophoresis on a 7.5% SDS-polyacrylamide gel and transferred to a polyvinylidine difluoride (PVDF) membrane. The blots were then blocked in 5% nonfat dry milk in Tris-buffered saline for 1 h and incubated for 2 h at room temperature with goat polyclonal anti-COX-2 and anti-COX-1 antibodies (Research Diagnostic, Flanders, NJ). After washing, membranes were incubated with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibody and analyzed using enhanced chemiluminescence detection kits (ECL-Plus; Amersham Pharmacia Biotech, Buckinghamshire, U.K.). The specificity of the antibody was confirmed by preincubation with the cognate peptide (which effectively competed) and an irrelevant peptide (which had no effect) (data not shown). The abundance of the sciatic nerve COX-2 and COX-1 proteins in STZ-D and ND rats was compared using a phosphorimager.

Goat polyclonal anti-COX-2 and anti-COX-1 antibodies were used for the immunohistochemistry studies. RSNs were cryoprotected by rinsing in 0.1 mol/l phosphate buffer, pH 7.3, containing 20% sucrose. Sections (15 μm) were cut on a cryostat, thaw-mounted onto slides coated with Superfrost (Fisher, Pittsburgh, PA), and stored at −20°C. Frozen sections were brought to room temperature and rinsed in phosphate buffer and Triton X-100 with 2% nonfat dry milk to reduce nonspecific binding of the antisera. The sections were incubated with primary antibodies (anti-COX-2 and anti-COX-1, 1:1,000 dilution) in a humid chamber for 18–24 h at 22°C. Sections were then rinsed and incubated with the appropriate biotinylated secondary antisera (goat anti-rabbit IgG, 1:500 dilution; Vector Laboratories, Burlingame, CA) followed by avidin-biotin HRP. The chromogen was developed with 3,3′-diaminobenzidine (Sigma) in the presence of H₂O₂. The sections were counterstained with Ehrlich’s hematoxylin then dehydrated through an ethanol series; cover glasses were mounted with Permount (Fisher). Immunostaining for Isolectin B4 (Vector Laboratories) was used as an anatomical marker for endothelial cells.

Data are expressed as means ± SE. Differences between experimental groups were determined by ANOVA, and the significance of between-group differences was assessed by Student-Newman-Keuls multiple range test. Significance was defined as P ≤ 0.05. If the variances for the variables were found to differ significantly, a logarithmic transformation was performed that corrected the unequal variances. All analyses were then performed on the transformed data.

Body weights were similar in all experimental groups at baseline. Four weeks thereafter, body weights were 29–36% lower in STZ-D rats versus ND rats (P < 0.05), 17% lower in flurbiprofen-treated versus untreated ND rats (P < 0.05), and 8% lower in flurbiprofen-treated versus untreated STZ-D rats (NS) (Table 2). Body weights were lower in both ND (21%; P < 0.05) and STZ-D (10%; P = 0.4) rats treated with flurbiprofen plus ALC versus their respective untreated controls. Body weights in STZ-D rats were not significantly decreased by ALC alone. Persistent hyperglycemia in STZ-D rats was unaffected by flurbiprofen or ALC (Table 2).

After 4 weeks, flurbiprofen decreased MNCV in ND rats by 14% versus untreated ND rats (P < 0.05), but this reduction was less than that produced by 4 weeks of diabetes (P = 0.1) (Fig. 1). As expected, MNCV was reduced by 19% (P < 0.05) in STZ-D versus ND rats and was not affected by flurbiprofen. The administration of 50 mg · kg⁻¹ · day⁻¹ ALC corrected MNCV deficits in flurbiprofen-treated ND rats and STZ-D rats irrespective of COX inhibition.

Four weeks of STZ-induced diabetes increased nerve sorbitol levels 4.9-, 4-, 2.4-, and 4.3-fold and nerve fructose levels 4.6-, 4.1-, 2.3-, and 3.1-fold in untreated, flurbiprofen-treated, ALC-treated, and flurbiprofen + ALC-treated rats versus their respective ND controls (all P < 0.05) (Table 3). Nerve sorbitol and fructose levels were significantly lower in the ALC-treated STZ-D group versus the other STZ-D groups. Flurbiprofen tended to lower nerve sorbitol and fructose content in ND and STZ-D rats, but this effect did not achieve statistical significance (P = 0.2).

Flurbiprofen lowered nerve MI and taurine content by 25 and 33%, respectively, in ND rats (both P < 0.05) (Table 3). Although 50 mg · kg⁻¹ · day⁻¹ ALC did not affect the MI decrease, the reduction of taurine content was partially attenuated (11% higher versus flurbiprofen-treated ND; not statistically significant, P = 0.23). As anticipated, STZ-induced diabetes lowered nerve MI content by 47% and taurine content by 19% (both P < 0.05); flurbiprofen potentiated these decreases by 23% (P < 0.05) and 10% (P = 0.34), respectively. Treatment of STZ-D rats with 50 mg · kg⁻¹ · day⁻¹ ALC exacerbated taurine depletion by 28% (P < 0.05); this potentiation was unaffected by flurbiprofen. Treatment of STZ-D rats with ALC tended to ameliorate MI depletion (P = 0.05 vs. untreated STZ-D rats) and prevented further lowering of MI by flurbiprofen.

In ND rats, flurbiprofen profoundly decreased nerve Na,K-ATPase activity by 58% versus nontreated ND rats to levels lower than that in STZ-D rats (both P < 0.05); this decrease was completely corrected by ALC (Table 3). Nerve Na,K-ATPase activity was decreased by 32% (P < 0.01) in STZ-D rats and further decreased by flurbiprofen (P < 0.05). Administration of 50 mg · kg⁻¹ · day⁻¹ ALC significantly increased Na,K-ATPase activity in STZ-rats, and this effect was not blocked by flurbiprofen.

In ND rats, flurbiprofen decreased total NBF and vascular conductance by 29 and 33%, respectively (both P < 0.05) (Table 4). In contrast to the flurbiprofen-induced reductions in MNCV and Na,K-ATPase activity in ND rats, the reduction of vascular conductance by flurbiprofen was not corrected by 50 mg · kg⁻¹ · day⁻¹ ALC. In STZ-D rats, total NBF was reduced by 36% (P < 0.05 vs. ND) and further decreased by flurbiprofen. The deficit of total NBF in STZ-D rats was not affected by 50 mg · kg⁻¹ · day⁻¹ ALC. When corrected for systemic blood pressure (which was 3% lower in STZ-D rats), total vascular conductance was reduced by 44% in STZ-D rats (P < 0.05 vs. ND). Total vascular conductance in STZ-D rats was unaffected by 50 mg · kg⁻¹ · day^{− 1} ALC or flurbiprofen, alone or in combination.

As in experiment 1, 4 weeks of STZ-diabetes interfered with weight gain and produced persistent elevation of plasma glucose that was unaffected by either ALC or flurbiprofen (data not shown). The 25% reduction in MNCV in STZ-D rats was again normalized by 50 mg · kg⁻¹ · day⁻¹ ALC irrespective of flurbiprofen treatment.

In flurbiprofen-treated ND rats, total NBF assessed by hydrogen clearance was reduced by 27% (P < 0.05 vs. untreated ND) (data not shown). These data are consistent with the magnitude of the changes obtained by laser-Doppler flowmetry. Total NBF in STZ-D rats was reduced by 42% (P < 0.05 vs. ND). The deficit of total NBF in STZ-D rats with or without flurbiprofen treatment was not corrected by 50 mg · kg⁻¹ · day⁻¹ ALC. These results were not affected after correction for blood pressure (which was 8–15% lower in untreated and treated STZ-D rats; P < 0.05 vs. ND) (expressed as vascular conductance).

In flurbiprofen-treated ND rats and untreated STZ-D rats, the endoneurial nutritive component of the NBF (Table 5) was found to decrease in parallel to total NBF. But in contrast to the measurements of total NBF, the flurbiprofen-induced reduction of endoneurial NBF in ND rats was selectively corrected by ALC (Table 5). These results were unchanged when expressed as vascular conductance. In contrast to its effects in ND rats, flurbiprofen treatment of STZ-D rats paradoxically prevented endoneurial NBF deficits. Thus, nonspecific COX inhibition in ND and untreated STZ-D rats decreased both total NBF and endoneurial nutritive NBF in parallel. However, the effects of ALC (in flurbiprofen-treated ND and STZ-D rats) and flurbiprofen (in STZ-D rats) were restricted to the endoneurial nutritive component, with total NBF remaining unchanged.

Body weights were similar in all experimental groups at baseline (although ∼100 g higher in the first subset of animals compared with experiments 1 and 2). After 4 weeks of STZ-induced diabetes, body weights were 27% lower versus ND rats in the first subset of animals and 15% lower in the second subset (data not shown). Meloxicam did not affect body weight in ND or STZ-D rats. Plasma glucose was increased by 3.4-fold in untreated STZ-D (ND 127 ± 15 vs. STZ-D 434 ± 23 mg/dl; P < 0.05) and was unaffected by meloxicam in ND rats (99 ± 12 mg/dl) or STZ-D rats (375 ± 38 mg/dl; P < 0.05 vs. ND rats with and without meloxicam treatment).

In ND rats, meloxicam treatment did not significantly effect sciatic nerve MI and fructose content but induced a 25% reduction in taurine content (P = 0.06) and a 60% reduction in sorbitol content (P < 0.05) (Table 6). Four weeks of STZ-induced diabetes increased nerve sorbitol and fructose levels twofold (both P < 0.05) and lowered nerve MI content by 29% and nerve taurine content by 35% (both P < 0.05) versus ND rats. Neither the alterations of nerve polyol pathway metabolites nor nerve osmolytes in STZ-D rats were affected by meloxicam treatment.

In ND rats, meloxicam reduced Na,K-ATPase activity by 33% (P < 0.05), but this reduction was significantly less than that induced by the nonselective COX inhibitor, flurbiprofen (P < 0.05) (Table 6). In STZ-D rats, Na,K-ATPase activity was decreased by 35% (P < 0.05) and was not significantly effected by meloxicam. Therefore, in contrast to flurbiprofen, selective COX-2 inhibition with meloxicam did not worsen metabolic deficits in STZ-D rats.

As observed in experiment 1, 4 weeks of STZ-diabetes significantly slowed MNCV in untreated rats by 16% (P < 0.05) (Fig. 2). Meloxicam did not reproduce the flurbiprofen-induced slowing of MNCV in ND rats and prevented MNCV deficits in STZ-D rats.

In contrast to the effects of flurbiprofen, selective COX-2 inhibition with meloxicam did not decrease total NBF in ND rats (meloxicam-treated ND, 163 ± 13; untreated ND, 164 ± 10 flux units) (Fig. 3). Consistent with the findings of experiment 1, total NBF was decreased by 42% (P < 0.01) after 4 weeks of diabetes (94 ± 11 flux units). In STZ-D rats, in contrast to the effects of flurbiprofen, the selective COX-2 inhibitor meloxicam prevented deficits of total NBF (139 ± 13 flux units; P < 0.05 vs. untreated STZ-D rats). Mean systemic blood pressure was significantly lower in meloxicam-treated ND versus untreated ND rats (102 ± 4 vs. 116 ± 5 mmHg; P < 0.05) and in both STZ-D rat groups (P < 0.05 vs. ND), which did not differ significantly from each other (untreated STZ-D, 100 ± 5; meloxicam-treated STZ-D, 105 ± 5 mmHg). Therefore, when total NBF was corrected for differences in mean systemic blood pressure and expressed as vascular conductance, flow was decreased by 36% (P < 0.01) in untreated STZ-D rats, a deficit that was prevented by meloxicam (Fig. 3A).

A separate experiment (experiment 4) determined the effects of meloxicam on endoneurial NBF in ND and STZ-D rats. As before, body weights of STZ-D rats were decreased by ∼27% compared with ND animals, and meloxicam did not affect body weight or plasma glucose in ND or STZ-D rats (data not shown). Meloxicam did not decrease endoneurial blood flow in ND rats (meloxicam-treated ND, 15 ± 0.7; untreated ND, 15 ± 0.6 flow units). Endoneurial blood flow was decreased by 47% (P < 0.01) after 4 weeks of diabetes (7.9 ± 0.4 flow units). Consistent with the results of experiment 3, meloxicam prevented deficits of endoneurial NBF in STZ-D rats (15 ± 1 flow units; P < 0.01 vs. untreated STZ-D rats). Mean systemic blood pressure was higher in meloxicam-treated ND rats versus untreated ND rats (144 ± 3 vs. 117 ± 5 mmHg; P < 0.05), unaffected in untreated STZ-D rats (112 ± 8 mmHg), and lower in meloxicam-treated STZ-D rats (89 ± 1 mmHg; P < 0.05 vs. other groups). Therefore, when endoneurial NBF was corrected for differences in mean systemic blood pressure and expressed as vascular conductance, flow was decreased by 44% (P < 0.01) in untreated STZ-D rats, a deficit that was prevented by meloxicam (Fig. 3B).

Western blot analysis of composite sciatic nerve homogenates (SNH) showed that the polyclonal anti-COX-2 antibody recognized a single appropriately sized band of ∼75 kDa, which is the expected molecular weight of COX-2 protein (Fig. 4). In ND rats, low-level (n = 3) or undetectable (n = 3) COX-2 expression was found in the SNH, which contrasted markedly with increased expression (94% increase versus ND rats) observed in SNH from all STZ-D rats (Fig. 4A).

In contrast, COX-1 was detectable in SNH from all rats, and there was no difference in COX-1 protein expression between STZ-D and ND rats (Fig. 4B).

By immunohistochemical analysis of STZ-D rat sciatic nerve, COX-2 was found to be mainly localized within the vascular endothelial cells of the peri- and endoneurium (Fig. 5A). Endothelial cell localization was confirmed by labeling serial adjacent sections with an antibody to Isolectin B4 as described in research design and methods (Fig. 5B). In contrast, COX-1 was found to be diffusely distributed across nerve tissue compartments, without specific localization at any site (data not shown).

Alterations in COX pathway activity, potentially secondary to systemic or local defects in fatty acid metabolism in diabetes, have been implicated in endoneurial hypoxia and impaired nerve function in EDN (38,46). However, the role of the COX pathway in the regulation of nerve metabolism and function has not been fully characterized. In the studies reported herein, we explore the relationships between the COX-mediated and ALC-sensitive defects that contribute to functional, metabolic, and vascular abnormalities of EDN.

Nonselective inhibition of the COX pathway with flurbiprofen in ND rats replicated, and in STZ-D rats potentiated, many (but not all) of the biochemical and physiological abnormalities of EDN, i.e., reduced MNCV, Na,K-ATPase activity, and MI and taurine content (3,16, 17,40,46, 47). In ND rats, flurbiprofen reproduced the effects of STZ-diabetes on both total and endoneurial NBF. In STZ-D rats, however, flurbiprofen selectively prevented the deficit of endoneurial NBF (but not MNCV slowing). Co-administration of ALC with flurbiprofen prevented reductions of MNCV and Na,K-ATPase activity, implying that the flurbiprofen-induced deficits were not due to nonspecific drug toxicity. In contrast, selective COX-2 inhibition with meloxicam was without affect on MNCV, NBF, or MI content in ND rats and prevented MNCV slowing and total and endoneurial NBF deficits in STZ-D rats. Western blot analysis showed unchanged sciatic nerve COX-1 protein but increased COX-2 protein abundance in STZ-D rats compared with ND rats. These results imply 1) a tonic role of the COX-1 pathway in the regulation of nerve osmolytes and Na,K-ATPase activity and the maintenance of NBF in ND animals and 2) activation of the COX-2 pathway as an important mediator of NBF and MNCV deficits in EDN.

In animal models of EDN, the effects of COX inhibition on neurovascular and functional deficits have been inconsistent. For example, nonselective COX inhibition with indomethacin has been reported to correct slowed sural sensory NCV and increase caudal motor NCV and NBF (but not endoneurial oxygenation) to levels in excess of ND animals, despite the failure to demonstrate motor NCV slowing or NBF deficits in untreated diabetic rats (48). Additionally, the nonselective COX inhibitor sulindac corrected conduction deficits in sensory caudal and sural fibers in diabetic rats and prevented the reduction of dorsal root ganglia blood flow and sciatic nerve endoneurial oxygen tension but did not affect modest caudal MNCV slowing (49). Piroxicam has been reported to attenuate the diabetes-induced decline of sensory nerve action potential amplitude after 16 weeks of STZ-induced diabetes but was without effect on either motor or sensory NCV (50). In this article, the lack of any beneficial effect of flurbiprofen on MNCV despite preventing endoneurial perfusion deficits in diabetic rodents is consistent with the above reports and suggests that detrimental (but ALC-sensitive) nonvascular effects of nonselective COX inhibition on nerve electrophysiology may negate improvements in endoneurial perfusion. Moreover, the effects of flurbiprofen in ND rats are consistent with rare findings reported in nondiabetic humans, in whom treatment with nonselective COX inhibitors, at doses similar to the ones used herein, has been associated with the development of peripheral somatic neuropathy (51–53). The reasons for the relatively greater susceptibility of rodents compared with humans are, however, unknown.

The explanation for the apparent inconsistency of the effects of different COX inhibitors on NCV deficits in STZ-D rats is unknown but may reflect differences in the relative degree of inhibition of the two COX pathways (or other unspecified metabolic effects). For example, flurbiprofen is a very potent COX-1 inhibitor, its potency being higher than that of indomethacin (54–56), and so any beneficial effects of the other less effective COX-1 inhibitors in diabetic animals may be attributable to the relative degree of COX-1 versus COX-2 inhibition at the dose used. This explanation would be consistent with the beneficial effects of meloxicam in EDN and would implicate an important role for the COX-1 pathway in the maintenance of nerve metabolism, NBF, and ultimately nerve conduction. Alternatively, the beneficial effects of sulindac on MNCV slowing have been attributed to inhibition of the polyol pathway (49). In this report, meloxicam did not affect nerve sorbitol levels in diabetic rats, thereby making unlikely a mechanism attributable to aldose reductase inhibition.

In ND rats, combined COX-1 and COX-2 inhibition with flurbiprofen decreased both total and endoneurial NBF, whereas selective COX-2 inhibition was without measurable effect, suggesting that tonic COX-1, but not COX-2, activity is required for the maintenance of normal NBF. In STZ-D rats, however, flurbiprofen and meloxicam both prevented endoneurial NBF deficits. This is consistent with the construct that diabetes-induced COX-2 activation may predominantly contribute to nerve perfusion deficits. This view is supported by nerve Western blot analysis demonstrating increased abundance of only the COX-2 protein in STZ-D rats and its immunohistochemical localization to the nerve vasculature. COX-1 is constitutively expressed in all cells and tissues, including peripheral nerve (57), whereas COX-2 is inducible by proinflammatory stimuli (58), oxidative stress (59), and hypoxia (30,60). However, the COX-ROS interplay is complex, since activation of COX-2 (61) can itself increase ROS production during PGG₂ to PGH₂ conversion, which generates hydrogen peroxide (58,62), therefore completing a vicious circle. Thus, in EDN, increased oxidative stress (2, 5,9,63, 64) would be expected to induce COX-2 expression and further exacerbate oxidative stress. In addition, induction of COX-2 in the extrinsic epineurial or perineurial penetrating vasculature may increase synthesis of vasomodulatory agents such as TXA2, thereby facilitating vasoconstriction, platelet aggregation, endoneurial hypoxia, and nerve conduction deficits.

Correction of endoneurial blood flow deficits has been viewed as a prerequisite for the correction of MNCV slowing in diabetic rodents (19,40,64). Measurements of NBF are not standardized and are typically performed by either laser-Doppler flowmetry or hydrogen clearance. In this article, the two techniques were consistent with each other in detecting changes of total NBF. However, laser-Doppler flux measurements may underestimate the endoneurial nutritive component of NBF, which may critically determine endoneurial oxygenation (3,40,65). Therefore, hydrogen clearance, which distinguishes nutritive from nonnutritive (shunt) NBF (66), was used to evaluate potential dissociative effects of ALC and flurbiprofen on total and endoneurial NBF in ND and STZ-D rats. In STZ-D rats, ALC and flurbiprofen selectively improved the endoneurial component of NBF, whereas total NBF remained unchanged. However, the salutary effect of ALC but not flurbiprofen in preventing MNCV slowing in STZ-D rats demonstrates the importance of the additional nonvascular, but ALC-sensitive, actions of flurbiprofen on nerve electrophysiology.

Flurbiprofen dissociated nerve perfusion and nerve conduction deficits in STZ-D rats. In 6-week STZ-D rat sciatic nerve, decreased GSH (reduced glutathione), mitochondrial and cytoplasmic NAD⁺/NADH ratios (64,67), and vasodilator-sensitive reductions of PCr/Cr and ATP/ADP ratios (67) are consistent with impaired mitochondrial oxidative phosphorylation, secondary to oxidative stress and ischemic hypoxia. Reduced mitochondrial fatty acid oxidative capacity may exacerbate accumulation of long-chain fatty acids and fatty acid esters (34,35), which are thought to perturb membrane stability and function (34), signal transduction pathways (68), and Na,K-ATPase activity (16). Administration of flurbiprofen to animals could directly reproduce (or exacerbate) the inhibitory effects of diabetes on mitochondrial fatty acid oxidation independently of endoneurial ischemia. In rats, flurbiprofen at doses similar to those used herein induces a potent inhibition of mitochondrial fatty acid oxidation via a nonstereoselective non-CoA mechanism (69,70). However, it is unknown whether the rat nerve accumulates sufficient concentrations of flurbiprofen to exert this effect in vivo. Moreover, the sensitivity of flurbiprofen-induced deficits in nerve metabolic and electrophysiologic function to ALC, which can enhance mitochondrial fatty acid oxidation (33–35,71), supports these hypothetical constructs. Future studies are needed to explore changes in nerve redox potentials in response to nonselective COX inhibition.

Two findings, the unanticipated and profound depletion of nerve MI and taurine content by flurbiprofen and the selective depletion of taurine by meloxicam in ND rats, implicate COX pathway activity in the regulation of nerve organic osmolyte content. The Na⁺-dependent uptake mechanisms for MI and taurine (72,73) and their efflux via volume-sensitive organic osmolyte channels are critically regulated by signal transduction pathways. Depletion of intracellular MI may impair activation of protein kinase C (PKC) in some diabetic tissues, including the neuron (6), whereas depletion of taurine in diabetes (17,18,45) may contribute to oxidative stress (18,45), intracellular calcium overload, and activation of PKC (74). Regulation of nerve organic osmolytes is more complex than originally proposed, since levels of both MI and taurine in the nerve have been shown to be altered by ALC (33), oxidative stress (64), and now COX-pathway activity, in addition to the osmotic consequences of sorbitol accumulation (17).

Nonselective COX inhibition depressed nerve Na,K-ATPase activity in ND rats to levels below that of 4-week STZ-diabetes and further exacerbated the reduction in diabetic rats. Our finding, a reduction in nerve Na,K-ATPase activity induced by selective COX-2 inhibition in both ND and STZ-D rats, confirms the importance of COX-2 products in the maintenance of Na,K-ATPase activity. In addition to regulation through a cAMP-dependent protein kinase pathway (75,76), decreased Na,K-ATPase activity may reflect a COX inhibitor-exacerbated nerve energy deficit (77), secondary to disruption of mitochondrial function. The ability of ALC to ameliorate deficits in Na,K-ATPase activity despite potent COX inhibition is again consistent with a direct mitochondrial action (33–35,71) or possibly an effect mediated via COX-independent production of PGE₁ (78) or reduction in oxidative stress (79,80).

In summary, these studies imply 1) an important role for the COX-1 pathway in the tonic regulation of nerve osmolytes, Na,K-ATPase activity, and nerve perfusion in healthy animals and 2) pathological activation of the COX-2 pathway as an important mediator of NBF and MNCV deficits in EDN. Future studies examining the role of selective COX-2 inhibition in the management or prevention of diabetic peripheral somatic neuropathy appear warranted.

These studies were supported by U.S. Public Health Service Research Grants R01-DK52391 (M.J.S.) and RO1-DK38304 (D.A.G.), a Veterans Administration Career Development Award (M.J.S.), the Juvenile Diabetes Research Foundation (M.J.S.), a Research Award from the American Diabetes Association (M.J.S.), a gift from Hoffman-La Roche Ltd. (M.J.S.), and the Michigan Diabetes Research and Training Center (P60-DK20572). The authors gratefully acknowledge the technical support of L. Beyer.