Interactions Between Hyperglycemia and Hypoxia: Implications for Diabetic Retinopathy

Male Sprague-Dawley rats that weighed 200–250 g were purchased from Sasco (O’Fallon, MO) and housed and cared for in accordance with guidelines of the University Committee for the Humane Care of Laboratory Animals. Rats were housed one per cage in a room with a 12:12-h artificial light cycle at a temperature of 21 ± 2°C and a humidity of 55 ± 2%. They had free access to standard rat chow (Ralston Purina, Richmond, IN) and tap water.

Rats were anesthetized by intraperitoneal injection of pentobarbital (50 mg/kg body wt) or by CO₂ narcosis. The whole eye was quickly removed and placed in Krebs bicarbonate/HEPES buffer at pH 7.45 for further dissection, and the rats were exsanguinated. The buffer composition was 110 mmol/l NaCl, 4.5 mmol/l KCl, 0.5 mmol/l MgCl₂·6H₂O, 1.0 mmol/l CaCl₂·2H₂O, 1.5 mmol/l Na₂HPO₄, 30.0 mmol/l HEPES, 15 mmol/l NaHCO₃, 5.0 mmol/l d-glucose, 0.1 mmol/l l-arginine, and 0.05% BSA.

The cornea was incised with the aid of a dissecting microscope, the lens was rapidly removed, and retinas were gently separated from the choroid with tweezers. Retinas were then incubated for 2 h at 37°C (pH ∼7.40) in the above buffer (10 ml per retina in all experiments) that contained 5 or 30 mmol/l d-glucose ± 70 μmol/l tolrestat (an aldose reductase inhibitor [ARI]; a gift from Wyeth-Ayerst. Retinas that were incubated in 70 μmol/l tolrestat were obtained from rats that were fed tolrestat in the diet to provide a daily dose of 0.2 mmol/kg body wt for 1 week before removal of the retina to achieve inhibitory concentrations of the ARI. One retina from each rat was incubated in 5 mmol/l glucose at 676 torr (normoglycemic aerobic condition) and the other in 30 mmol/l glucose at 676 torr (hyperglycemic aerobic condition) or one retina in 5 mmol/l glucose at 676 torr and the other in 5 mmol/l glucose at 36 torr (normoglycemic hypoxic condition), etc., to permit paired comparisons of effects of hyperglycemia and hypoxia.

The buffer was gassed for 2 h before adding the retinas (and throughout the incubation) with the following humidified gas mixtures that contained 5% CO₂ + various percentages of O₂ with the balance made up of nitrogen: 1) 95% O₂ (676 torr), 2) 50% O₂ (357 torr), 3) 20% O₂ (143 torr), or 4) 5% O₂ (36 torr). The pH and Po₂ of incubation media were measured at the beginning and end of incubations with a Corning 170 pH/Blood Gas Analyzer, and the osmolality (∼290 mOsmol) was measured with a Wescor 5100 C Vapor Pressure Osmometer. None of these parameters changed during the incubation period. Incubations were terminated by removal of retinas for measurement of metabolites.

The time course of 30 mmol/l glucose-induced increases in glycolytic metabolites, free NADHc, and lactate production was assessed in retinas that were incubated for 1, 2, and 3 h at 676 torr. The possibility that the absence of physiological levels of lactate and pyruvate in the incubation medium might have an impact on the effects of elevated glucose levels on these parameters was evaluated by adding 1.0 and 0.1 mmol/l lactate and pyruvate (normal plasma levels), respectively, to the medium. Effects of tolrestat on sorbitol pathway and glycolytic metabolites and on free NADHc were assessed in hypoxic retinas. (Effects of an ARI on these parameters in aerobic retinas have been published [10].)

Glucose, pyruvate, lactate, glucose 6-phosphate (G6P), fructose 6-phosphate (F6P), ATP, glutamate, 2-oxoglutarate, and ammonia in extracts of retina (and pyruvate and lactate in extracts of incubation medium) were quantified by standard enzymatic methods (10,39,40). Because the overall reaction catalyzed by aldolase and triose phosphate isomerase is in near equilibrium under steady-state conditions (7), fructose 1,6-bisphosphate (F1,6-BP) was converted to dihydroxyacetone phosphate (DHAP) by aldolase and measured together with endogenous DHAP and glyceraldehyde 3-phosphate (GA3P) as triose phosphates (7,10). Sorbitol in retinal extracts was quantified by gas chromatography/mass spectrometry as previously described (10). d-Fructose in the incubation medium was quantified by an enzymatic fluorometric method (41).

All metabolite data are reported as moles per microgram of DNA that was quantified as previously described (10). To compare metabolite levels in the present experiments with values from other laboratories reported as moles/tissue wet or dry weight, we determined conversion factors from 10 normal male Sprague-Dawley rats that weighed 336 ± 15 g. The wet weight of the retinas was 12.12 ± 0.98 mg, and the dry weight was 2.55 ± 0.22 mg. The conversion factors (mean ± SD) are 1) 56 ± 6 μg DNA/mg dry wt, 2) 202 ± 22 μg protein/mg dry wt, 3) 42 ± 4 μg protein/mg wet wt, and 4) 0.210 ± 0.008 mg dry wt/mg wet wt. Protein was quantified by the bicinchoninic acid method using the BCA Protein Assay Reagent obtained from Pierce Chemical Company and BSA as the standard.

Free NADHc and NADHm cannot be determined by direct measurement of NADH in whole tissue extracts that contain enzyme-bound as well as free NADHc and NADHm, and free NADm is more than an order of magnitude more reduced than free NADc (6). Free NADHc and NADHm can be estimated only by redox metabolite indicator methods based on near-equilibria between ratios of free NADH to NAD⁺ and reduced to oxidized substrates of dehydrogenase enzymes restricted in distribution to cytosol or mitochondria. Unless otherwise stated, NADHc and NADHm in this report refer to free, not total, NADHc and NADHm.

Free NADHc/NAD⁺c was assessed on the basis of the near equilibrium between free NADHc/NAD⁺c and lactate/pyruvate (L/P) ratios (Eqs. 1 and 2) and an equilibrium constant of 1.11 × 10⁻⁴ (at pH = 7.0) for lactate dehydrogenase (LDH) (6):

Because total retinal NADH and NAD⁺ (and the ratio of total NADH to NAD⁺) are increased in diabetic rats (4,5), it is reasonable to assume that hyperglycemia is unlikely to change substantially total free NADc and total free NADm in these brief experiments. Thus, changes in free NADHc will evoke proportional changes in the ratios of free NADHc to NAD⁺c and L/P; for clarity, changes in measured ratios of L/P are referred to as changes in NADHc∝L/P or NADHc.

Free NADHm/NAD⁺m was estimated on the basis of the near equilibrium between the ratios of NADHm to NAD⁺m and glutamate (GL) to 2-oxoglutarate established by glutamate dehydrogenase (6) and an equilibrium constant of 3.87 × 10⁻³ mmol/l (pH 7.0):

Because (as for changes in free NADHc) changes in free NADHm/NAD⁺m should be proportional to changes in free NADHm, they are referred to as such.

Multiple ANOVA of all parameters was performed with the SAS general linear models procedure as previously described (16) with significance set at P < 0.05. All results are reported as the mean ± SD (n ≥ 6 rats) unless indicated otherwise. Representative results from two to four independent experiments (n = ≤6 rats per group in each experiment) are shown in the figures and tables.

Hyperglycemia increased retinal L/P ratios 1.4- to 2-fold at all Po₂ (Fig. 1, Table 1). Decreasing Po₂ from 676 to 357 torr did not affect L/P (Fig. 1); however, further reduction of Po₂ to 143 and 36 torr increased L/P 1.7-fold (P ≤ 0.035) and >6-fold (P < 0.0001), respectively, both in normoglycemic and hyperglycemic retinas. Tolrestat did not affect the hypoxia-induced increase in L/P in normoglycemic retinas but prevented the additional increase in L/P evoked by hyperglycemia (Fig. 1). NADHm was increased twofold by hypoxia but was unaffected by hyperglycemia in aerobic or hypoxic retinas (Table 1).

Hyperglycemia and hypoxia increased retinal lactate levels ∼3-fold and 1.7-fold, respectively, and the increases were additive (Table 1). In contrast, whereas hyperglycemia increased pyruvate levels approximately twofold, they were decreased ∼90% by hypoxia in normoglycemic and hyperglycemic retinas.

Hyperglycemia and hypoxia increased glycolysis (based on the sum of lactate and pyruvate in retina and medium) and lactate production (sum of lactate in retina + medium) ∼1.5-fold and 1.3- to 2-fold, respectively, and the increases were additive (Fig. 2, Tables 1 and 2). (See online appendix, section 1A, available at http://diabetes.diabetesjournals.org.) Increased lactate production evoked by hypoxia in normoglycemic and hyperglycemic retinas was unaffected by tolrestat. Pyruvate production (sum of pyruvate in retina + medium) by normoglycemic retinas was 2.3 ± 0.3 nmol · μg DNA⁻¹ · 2 h⁻¹ and was unaffected by hyperglycemia, hypoxia, or tolrestat (data not shown).

Lactate levels and L/P ratios in aerobic retinas and lactate production were stable during incubations of 1-, 2-, and 3-h duration under normoglycemic and hyperglycemic conditions and were unaffected by the addition of normal plasma levels of lactate and pyruvate to the medium (data not shown).

Hyperglycemia increased glucose levels approximately eight times in aerobic and hypoxic retinas. G6P and F6P were increased approximately twofold by hyperglycemia but were reduced ∼50% by hypoxia (Table 2); F6P was undetectable in normoglycemic-hypoxic retinas.

Hyperglycemia increased triose phosphates ∼1.4-fold (Fig. 3, Table 2). Triose phosphates were unaffected by decreasing Po₂ from 676 to 143 torr (Fig. 3) but were increased ∼1.7-fold (P < 0.0001) at 36 torr in normoglycemic and hyperglycemic retinas (Fig. 3, Table 2). Tolrestat had no effect on hypoxia-induced increases in triose phosphates in normoglycemic retinas but prevented the additional increase evoked by hyperglycemia (Fig. 3).

Hyperglycemia increased ATP levels 1.6-fold in aerobic retinas and 3.3-fold in hypoxic retinas. Hypoxia reduced ATP 66 and 28% in normoglycemic and hyperglycemic retinas, respectively (Table 2). Thus, ATP levels were the same in hypoxic-hyperglycemic retinas and in aerobic-normoglycemic retinas. Addition of plasma levels of lactate and pyruvate during incubation increased ATP levels twofold (P < 0.01 versus ATP levels in aerobic-normoglycemic and in aerobic-hyperglycemic retinas) (Table 2).

Sorbitol levels in normoglycemic retinas ranged from 5 to 7 pmol/μg DNA and were increased 9- to 18-fold (P < 0.001 at all Po₂) by hyperglycemia (data not shown). Fructose production by normoglycemic retinas ranged from 1.3 to 3.6 pmol/μg DNA at all Po₂ (Fig. 4) and was increased 55- to 74-fold by hyperglycemia (P < 0.001 at all Po₂). Tolrestat reduced fructose production by 65 and 96% in hypoxic-normoglycemic and hypoxic-hyperglycemic retinas, respectively (Fig. 4).

These experiments demonstrate that 1) hyperglycemia and hypoxia increase NADHc, triose phosphates, and glycolysis by different mechanisms that are additive, and 2) hyperglycemia-evoked increases in NADHc∝L/P and triose phosphates are mediated largely by a small increase in glucose metabolism via the sorbitol pathway to fructose that is only ∼2% of the increase in glucose metabolism via glycolysis. These observations have several important implications for the pathogenesis of diabetic complications. Increased ATP production by substrate phosphorylation evoked by hyperglycemia in oxygenated and hypoxic retinas may explain, in part, the transient worsening of retinopathy evoked by rapid normalization of plasma glucose levels.

Increases in NADHc (manifested by an increase in the retinal L/P ratio) and in glycolysis (manifested by increased lactate production) are sequelae of two different mass action effects of hyperglycemia. The increase in NADHc∝L/P evoked by 30 mmol/l glucose at 676 and 36 torr is mediated entirely by increased flux of glucose via the sorbitol pathway to fructose (∼200 pmol glucose · μg DNA⁻¹ · 2 h⁻¹ based on fructose production in oxygenated and hypoxic retinas; Fig. 4) and is independent of the associated increases in glycolysis (∼5,000 pmol glucose · μg DNA⁻¹ · 2 h⁻¹ at 676 torr and ∼10,000 pmol glucose at 36 torr) (Fig. 2). This conclusion is supported by the observations that tolrestat prevented increases in NADHc∝L/P and fructose production at 36 torr (Figs. 1 and 4) but not the associated increase in lactate production (Fig. 2). (See online appendix, section IB) NADHc∝L/P values in normoglycemic retinas and increases in NADHc∝L/P evoked by hyperglycemia agree closely with values in control and diabetic rats (8,11,12).

In contrast, increased glycolysis is mediated by suppression of product inhibition of hexokinase by G6P (42) (and, possibly, by increased phosphorylation of glucose by glucokinase [43]). Elevated levels of G6P and proportional increases in F6P (Table 2) support this conclusion. Increased levels of F6P drive its phosphorylation to F1,6-BP by phosphofructokinase, the rate-limiting step in glycolysis under most physiological conditions. These observations indicate a shift from transport to phosphorylation as the rate-limiting step in hexose uptake like that observed in rat cerebellum and red skeletal muscle under conditions of hyperglycemia and hyperinsulinemia (44).

Some investigators suggest that the increase in NADHc formed by the sorbitol pathway during hyperglycemia is so small (compared with the increase in NADHc formed by glycolysis) that it is unlikely to be of any pathophysiological significance (12,45). Winkler et al. (45) asserted that increased oxidation of sorbitol to fructose must account quantitatively for an increase in retinal lactate content for NADH/NAD⁺c∝L/P ratios to be increased. Both of these notions are unfounded. The finding that tolrestat completely prevented increases in fructose production at 36 torr (which was only ∼2% of the molar increase in flux of glucose via glycolysis) and in NADHc∝L/P evoked by hyperglycemia clearly demonstrates the greater impact of flux of glucose via the sorbitol pathway on NADHc∝L/P than flux of glucose via glycolysis.

Moreover, it is well known that several cytosolic enzymes in addition to LDH catalyze oxidation of NADHc to NAD⁺c, e.g., G3P (glycerol 3-phosphate)-DHc, malate dehydrogenase, NADH oxidase, as depicted in Fig. 5. Lactate is formed only when NADHc is reoxidized by LDH.

Increased NADHc and pyruvate formed in resting cells (as a result of augmented glycolysis evoked by hyperglycemia) are oxidized to NAD⁺c and reduced to lactate, respectively, by LDH. In contrast, because the sorbitol pathway does not form pyruvate, increased NADHc formed by the sorbitol pathway is reoxidized largely by enzymes other than LDH (some of which are implicated in mediating diabetic complications) coupled to reduction of substrates other than pyruvate. (See online appendix, section II.) Thus, a far greater inhibition of aldose reductase than was achieved in past clinical trials with ARI may be needed to prevent complication of diabetes fueled by NADHc formed by sorbitol dehydrogenase.

In contrast to hyperglycemia, increases in NADHc and glycolysis evoked by hypoxia reflect impaired utilization of electrons and protons carried by NADHm for ATP synthesis by oxidative phosphorylation. The resulting increase in NADHm and decrease in NAD⁺m limits 1) reoxidation of NADHc to NAD⁺c (which, in turn, increases NADHc) by the malate-aspartate electron shuttle, which transfers electrons and protons from NADHc to NAD⁺m, and 2) mitochondrial oxidation of glycolysis-derived pyruvate coupled to reduction of NAD⁺m to NADHm by the Krebs cycle (46). Concurrent impaired utilization of ADP for synthesis of ATP by oxidative phosphorylation increases ADP and AMP levels, which activate phosphofructokinase (the key enzyme in regulation of glycolysis [46]), which mediates the hypoxia-evoked increase in glycolysis in normoglycemic and hyperglycemic retinas (Tables 1 and 2).

This scenario is supported by the observation that hypoxia augmented lactate production despite decreased retinal levels of G6P, F6P, and ATP (Table 2). These findings are not surprising but provide new information regarding retinal glucose metabolism. Decreased ATP levels in hypoxic retinas indicate that increased ATP synthesis via substrate phosphorylation did not keep pace with ATP synthesis by oxidative phosphorylation in aerobic retinas. The increased lactate production evoked by hypoxia in normoglycemic retinas is consistent with observations of Winkler et al. (45) and demonstrates (contrary to their conclusion) that glycolysis in normoglycemic-aerobic retinas was not limited by availability of extracellular glucose.

Finally, identical 1.6-fold increases in NADHc evoked by 143 torr (mild hypoxia) in normoglycemic retinas and by hyperglycemia in aerobic retinas (676 and 357 torr; Fig. 1) indicate that relatively mild hypoxia alone increases NADHc to the same extent as hyperglycemia alone; the finding that NADHc at 143 torr was increased 1.8-fold by hyperglycemia demonstrates that the effects of even mild hypoxia on NADHc are additive to those of marked hyperglycemia (Fig. 1).

An increase in NADHc, whatever the cause (e.g., increased sorbitol pathway metabolism, hypoxia, increased oxidation of fatty acids, ketones, lactate, galactose), will tend to inhibit reduction of NAD⁺c to NADHc by some enzymes (e.g., GA3P-DH) and augment oxidation of NADHc by others (e.g., G3P-DHc and NADH oxidase) (Fig. 5). Because the triose phosphates (GA3P, DHAP, and F1,6-BP) are in near equilibrium, their levels will rise with inhibition of GA3P-DH, and the combined mass action effects of increased triose phosphates and NADHc will drive reduction of DHAP to G3P coupled to reoxidation of NADHc to NAD⁺c by G3P-DHc (Fig. 5). This is the first step in one pathway for de novo synthesis of diacylglycerol and activation of PKC (Fig. 5) implicated in mediating several complications of diabetes (22). And, concentration-dependent degradation of triose phosphates to methylglyoxal (a toxic and potent intracellular glycating agent [47]) will be increased. In addition, the mass action effect of an increase in NADHc will augment formation of superoxide by NADH-fueled extramitochondrial NADH oxidase (48).

A growing body of evidence supports an important role for generation of reactive oxygen species (ROS) by PKC-activated, extramitochondrial NAD(P)H oxidases in mediating vascular disease evoked by diabetes. Several lines of evidence indicate that NADHc formed by oxidation of sorbitol 1) fuels de novo synthesis of diacylglycerol that activates PKC and 2) donates electrons that fuel cytosolic superoxide production by extramitochondrial NAD(P)H oxidases. (See online appendix, sections III and IVC.)

In contrast, Brownlee and associates (23) maintain that hyperglycemia-induced ROS are generated exclusively in mitochondria. They attribute increased ROS production to increased oxidation of pyruvate (produced by increased glycolysis), which generates NADHm that promotes superoxide production by the electron transport chain. They propose that mitochondrial superoxide inhibits GA3P-DH, causing triose phosphate levels to rise.

This scenario appears to be largely untenable when viewed in the context of 1) discordant observations of other investigators and 2) important caveats to the interpretation of their data. (See online appendix, sections III and IV.) For example, their scenario suggests that addition of exogenous pyruvate should exacerbate adverse sequelae of hyperglycemia. On the contrary, addition of pyruvate markedly attenuates elevated diacylglycerol levels and associated vascular dysfunction evoked by hyperglycemia in vivo (49). In addition, ARIs prevent hyperglycemia-evoked increases in triose phosphates and NADHc in normoxic and hypoxic retinas (10) (Figs. 1 and 3) without attenuating the associated increase in glycolysis (10) (Fig. 2).

Furthermore, hyperglycemia increases triose phosphates and NADHc in human erythrocytes that lack mitochondria as a source of superoxide to inhibit GA3P-DH (7), and the increase in trioses is prevented by addition of pyruvate and by an ARI. These and other observations support the importance of NADHc formed by the sorbitol pathway (rather than, or in addition to, increased mitochondrial oxidation of pyruvate) in mediating inhibition of GA3P-DH, elevation of triose phosphate levels, and increased superoxide production evoked by hyperglycemia.

Other seemingly discordant reports regarding a role for NADHc in the pathogenesis of diabetic complications are attributable largely to 1) important caveats inherent to experimental paradigms used, e.g., experimental galactosemia, and 2) the short half-life of inhibitors of sorbitol dehydrogenase used to test the hypothesis. (See online appendix, section IV.)

Substantial increases in ATP synthesis via substrate phosphorylation evoked by hyperglycemia (in aerobic and hypoxic retinas) suggest an explanation for transient early worsening of diabetic retinopathy associated with rapid normalization of blood glucose levels in individuals with long-standing poor glycemic control (50). Namely, the mass action effect of elevated glucose levels that increases ATP synthesis by substrate phosphorylation will be attenuated. Adverse effects of lowering glucose levels will be more pronounced to the extent that 1) the retina is hypoxic and more dependent on ATP synthesis by substrate phosphorylation and 2) vascular disease restrains increased retinal blood flows (and associated delivery of glucose and oxygen) evoked by physiological visual stimulation and dark adaptation. (Increased retinal blood flow in response to dark adaptation is impaired in diabetic rats [51]. In normal human subjects, acute hyperglycemia impairs dilation of retinal vessels in response to flicker photostimulation [52].)

In conclusion, these observations attest to the potential of additive increases in NADHc evoked by hyperglycemia and hypoxia to promote the onset and progression of diabetic retinopathy via metabolic pathways fueled by NADHc. This conclusion and recent observations that increased retinal and visual cortex blood flows evoked by photostimulation are mediated by an increase in NADHc∝L/P (53,54) are consistent with the central role of free NADc in fueling energy metabolism and signaling pathways that coordinate blood flow with energy metabolism in resting cells, during physiological work, and in pathophysiological states including diabetes and hypoxia.

This study was supported by National Institutes of Health Grants HL-39934 and EY-06600, Fonden til Laegevidenskabens Fremme (Copenhagen), the Lundbeck Foundation (Copenhagen), and the Kilo Research Foundation (St. Louis, MO).

We thank Eva Ostrow, Judi Burgan, Sam Smith, Joyce Marvel, and Nemani Rateri for excellent technical assistance with these studies.