Hyperketonemia Increases Tumor Necrosis Factor-α Secretion in Cultured U937 Monocytes and Type 1 Diabetic Patients and Is Apparently Mediated by Oxidative Stress and cAMP Deficiency

Written informed consent was obtained from all subjects in accordance with the protocol approved by the Institutional Review Board for the Protection of Human Research Subjects. Blood from patients and healthy siblings was drawn after an overnight fast. Blood samples were collected into precooled tubes with and without EDTA, which was kept in an ice bucket. The EDTA-blood was centrifuged, and the clear plasma was saved for AA, BHB, protein oxidation, and TNF-α assays. All analyses were performed immediately after blood collection. All patients included in this study were type 1 diabetic children and age-matched normal siblings. Diabetic subjects with plasma AA levels ≤0.2 μmol/ml were considered NKD, and those with AA levels >0.2 μmol/ml were considered HKD.

The U937 cell line was obtained from American Type Culture Collection (ATCC, Manassas, VA). These cells were maintained at 37°C in RPMI-1640 medium containing 7 mmol/l glucose, 10% (v/v) heat-inactivated fetal bovine serum, 100 units/ml penicillin, 100 μg/ml streptomycin, 12 mmol/l sodium carbonate, 12 mmol/l HEPES, and 2 mmol/l glutamine in a humidified atmosphere containing 5% (v/v) CO₂. For treatments, cells were washed once in plain RPMI-1640 before being suspended in fresh medium (complete) containing serum and other supplements (28).

U937 (1 million cells/ml) was treated with AA or BHB (0–3 mmol/l). Treatments included use of both normal glucose and HG medium (30 mmol/l) along with AA or BHB. For TNF-α secretion studies, cells were stimulated with phorbol 12-myristate 13-acetate (PMA; 50 ng/ml), and treatments were carried out at 37°C for 24 h. For cAMP assay, cells were not stimulated and treated for 60 min. All experiments were repeated at least four times. α-Ketobutyric acid (AKB) served as an inert ketone control for AA and BHB. Involvement of the cell signaling pathways, including the cAMP-protein kinase A (PKA) and the mitogen-activated protein kinase (MAPK) pathways, were studied using inhibitors/inducers, including forsokolin (inducer of cAMP), H89 (inhibitor of cAMP-PKA), SB203580 (inhibitor of the p38 pathway), and nuclear transcription factor (NF)κB-SB (NFκB inhibitor) (purchased from CalBiochem, San Diego, CA) (29,30). Mannitol was used as an osmolarity control. N-acetylcysteine (NAC) was used to examine the role of oxidative stress. Concentrations of various inhibitors/inducers as well as AA and AKB are given in the figure legends.

Intracellular reactive oxygen species (ROS) generation was detected using dihydroethidium dye (Molecular Probes). When the dye is cleaved by ROS, the fluorescent biproduct ethidium is produced, which is detected using a flow cytometer (31). Carbonyl protein (protein oxidation) was determined to assess oxidative stress levels (32). Plasma levels of AA and BHB were determined by enzymatic methods (26). TNF-α levels in the supernatant of treated cells and in plasma of patients and normal subjects were determined by the sandwich enzyme-linked immunosorbent assay (ELISA) method using a commercially available kit from Neogen (Lexington, KY). All appropriate controls and standards, as specified by the manufacturer, were used, and the data are expressed as picograms TNF-α secreted by 1 million cells. Cell growth was determined using the Alamar Blue bioassay (Alamar Biosciences, Sacramento, CA) as described earlier (28). Cellular cAMP content was measured using a commercially available ELISA kit from Neogen (Lexington, KY). Briefly, the treated cells were washed once in cold PBS (pH 7.4) and suspended in 1 ml ice-cold 65% ethanol. After keeping overnight at −20°C, cell lysates were centrifuged at 2,000g for 10 min and the supernatants collected. The extracts were dried under vacuum and dissolved in 0.5 ml assay buffer, and the intracellular cAMP contents were determined by a competitive immunoassay per the manufacturer’s protocol.

Information on age, duration of diabetes, and clinical laboratory tests, such as HbA_1c, were collected from patients’ medical records in the diabetes clinic. In the oxidative stress and cytokine assays, control plasma samples were analyzed each time to check the variation from plate to plate on different days of analyses. The assays were repeated if the variation in control plasma value from day to day was >7%. All chemicals were purchased from Sigma Chemical (St. Louis, MO) unless otherwise mentioned. Data were analyzed statistically using unpaired Student’s t tests or unpaired nonparametric tests (Mann-Whitney U test) between different groups and the Spearman rank-order correlation using Sigma Plot statistical software (Jandel Scientific, San Rafael, CA). A P value <0.05 was considered significant.

Figure 1 illustrates that TNF-α levels are higher in diabetic patients than in age-matched normal subjects, but this was not statistically significant. However, when diabetic patients were divided into NKD and HKD groups, HKD patients had significantly higher levels of TNF-α than NKD patients (P < 0.01). There was no difference in the levels of TNF-α in NKD patients compared with those of age-matched normal subjects. There was no difference in duration of diabetes (5.1 ± 1 vs. 5.8 ± 1 years) or age (12 ± 1 vs. 13 ± 1 years) between NKD and HKD patients and no difference in age between control subjects (12 ± 1 years) and diabetic patients (13 ± 1 years) (means ± SE). HbA_1c levels between NKD (10.3 ± 1.5%) and HKD subjects (11 ± 1.3%) were not significantly different. Plasma AA levels were 0.17 ± 0.01 mmol/l (range 0.09–0.20) in NKD subjects and 0.29 ± 0.03 mmol/l (0.21–0.73) in HKD subjects. Total ketone body (AA plus BHB) levels were 0.23 ± 0.03 mmol/l (range 0.09–0.41) vs. 0.49 ± 0.08 (0.22–1.95) in NKD versus HKD groups. There was no difference in body weights among children in the NKD (50.2 ± 6.3 kg) and HKD groups (48.0 ± 3.7) groups. This suggests that ketosis contributes to the production of TNF-α in diabetic patients. This study also found a significant relationship (r = 0.36, P < 0.05; n = 34) between ketosis (as determined by AA level) and oxidative stress (as assessed by protein oxidation level) as well as between protein oxidation and TNF-α levels (r = 0.47, P < 0.02; n = 34) in the blood of type 1 diabetic patients. There was no correlation (r = 0.03) between HbA_1c and plasma protein oxidation levels in diabetic patients. There was no evidence of microalbuminuria in either group of diabetic children.

To examine the biochemical mechanisms leading to elevated TNF-α levels in HKD patients, we studied the TNF-α secretion by a U937 monocytic cell line cultured with elevated levels of AA or BHB with and without HG levels in the culture medium. Figure 2A shows that AA, BHB, or AKB alone does not have any effect on TNF-α secretion in unstimulated monocytes. However, in PMA-activated monocytes, AA caused a concentration-dependent increase in TNF-α secretion. However, treatment with elevated levels of BHB or AKB did not have any effect on TNF-α secretion. Figure 2B shows that AA at levels similar to those frequently encountered in diabetic patients can cause ROS production in monocytes. Neither BHB nor AKB had any effect on ROS production.

Figure 3A illustrates that HG also resulted in a modest increase in TNF-α secretion in PMA-activated monocytes. The effect of AA on TNF-α secretion was observed even in the presence of HG. Mannitol (Fig. 3A) did not cause any change in TNF-α secretion compared with basal levels. Figure 3A also shows that AA- and HG-induced TNF-α secretion was completely inhibited in the presence of NAC. Similarly, exogenous addition of dibutyryl cAMP lowered the production of TNF-α in AA-, HG-, and AA plus HG-treated monocytes (Fig. 3B). Forsokolin, which stimulates the endogenous production of cAMP, also caused a significant inhibition of TNF-α secretion by AA, HG, and AA plus HG. This suggests that elevated levels of cAMP inhibit the effect of AA and AA plus HG on TNF-α secretion. The effect of cAMP on TNF-α secretion was not due to cAMP effect on cell survival because there was no difference in cell growth level between cAMP and AA versus AA-treated cells only (data not given). Figure 4 shows that AA and AA plus HG treatment significantly lowered cAMP levels, and forsokolin treatment replenished the cAMP levels in AA- and HG-treated monocytes.

Figure 5A illustrates a significant reduction in cAMP in peroxide-treated cells, and this effect of peroxide was prevented with NAC supplementation. The increase in TNF-α production caused by the peroxide in activated monocytes was also prevented by NAC (Fig. 5B). This suggests a role of oxidative stress in the enhanced TNF-α secretion in AA-treated U937 monocytes and that elevated TNF-α production is mediated by cAMP depletion.

Figure 6 illustrates that H89 and SB203580 inhibited the secretion of TNF-α in AA- and HG-treated monocytes. Figure 7 illustrates that a cell-permeable NFκB inhibitor (NFκB-SN50) completely inhibited the TNF-α secretion in both AA- and HG-treated monocytes. Figures 6 and 7 suggest a role of protein kinase and p38-MAPK in the signal transduction pathways and possible upregulation of NFκB and TNF-α synthesis and secretion by cells treated with AA and HG.

Many biochemical pathways seem to be involved in cellular damage associated with diabetes (33–39). These include increased polyol pathway and associated changes in the intracellular redox state, increased diacylglycerol synthesis with consequent activation of specific protein kinase C (PKC) isoforms, increased nonenzymatic glycation of both intra- and extracellular proteins, and increased oxidative stress (33–39). Oxidative stress may arise from a variety of mechanisms, such as excessive oxygen radical production as a result of the auto-oxidation of glucose (39), the activation of P-450-like activity by the glucose metabolite NADPH (36), glycated proteins (40–42) and the ketone body AA (22–26), and the depletion of NADH by the activation of aldose reductase (35) and glycation of antioxidative enzymes, which limits their capacity to detoxify oxygen radicals (38). Type 1 diabetic patients may experience oxidative stress from both hyperglycemia and ketosis. This study has shown generation of ROS by ketone body AA at concentrations present in uncontrolled diabetic patients. In addition, ketosis can increase extramitochondrial oxidation of fatty acids and generation of hydrogen peroxide and thereby increase oxidative stress in ketotic diabetic patients (43). In HG-treated cultured endothelial cells, the inhibition of superoxide radical generation prevents activation of PKC, formation of advanced glycation end products, sorbitol accumulation, and NFκB activation (35). This suggests a role of excess ROS production in the cellular damage associated with diabetes.

The present study shows that TNF-α levels are higher in HKD patients, but not in NKD patients, than in normal subjects. The diabetic patients in this study were children who did not show any sign of clinical complications. Our data, together with that from previous studies on TNF-α levels in newly diagnosed diabetic children (14,18), suggest that the elevated levels of TNF-α in diabetic patients is not due to the complications associated with diabetes.

To determine the specific mechanisms by which ketosis increases levels of TNF-α in diabetic patients, U937 human premonocyte cells were cultured in vitro with the ketone bodies AA and BHB at concentrations similar to those seen in the blood of diabetic patients. The present study shows that the TNF-α secretion in activated monocytes was stimulated by both AA and HG, separately as well as together, whereas BHB did not have any effect on TNF-α secretion. AA can generate ROS, whereas BHB does not (24,26). This suggests that ROS may be involved in the increased TNF-α secretion in AA-treated monocytes. Indeed, treatment with the standard oxidant H₂O₂ increases TNF-α secretion in U937 cells. The effect of ROS on increased TNF-α secretion was blocked by the antioxidant NAC. Similarly, the effects of AA or AA plus HG were prevented by the antioxidant NAC. Taken together, these studies suggest that ROS generation may be involved in the increased TNF-α secretion in the AA-treated monocytes. Thus, inhibition by antioxidant of TNF-α secretion caused by treatment with AA and HG in a cell culture model, coupled with the significant relationship between blood TNF-α concentrations and oxidative stress levels in vivo, demonstrates that ketosis-induced oxidative stress apparently mediates the elevated TNF-α levels seen in HKD patients. TNF-α treatment by itself can increase cellular ROS production (44). TNF-α is unlikely to have any role in ROS generation in AA-treated cells because ROS production was determined in unstimulated U937 cells. However, we cannot rule out contribution of elevated TNF-α in the increased oxidative stress level in HKD-diabetic patients.

A growing number of studies have shown that moderate noncytotoxic oxidative stress specifically downregulates the expression of various genes involved in the regulation of the immune system, the cardiovascular system, and several other vital organ systems (45,46). Activation of the NFκB induces the expression of several inflammatory cytokines, including TNF-α (45–47). Therefore, it is likely that any ketosis-induced increase in cellular oxidative stress accelerates TNF-α production in diabetes.

High levels of glucose can upregulate expression of transcription factors NFκB and activating protein-1 and the TNF-α gene in monocytes. This expression of the TNF-α gene is mediated by the protein kinases p38 and JNK-1, which are dependent and independent of oxidative stress pathways (20). Several studies advocate the importance of the p38 pathway in diabetes (47,48). cAMP-dependent protein kinases (PKA) can activate phosphorylation of substrate proteins and cross talk with MAPKs pathways and proteins that are involved in signal transduction pathways, leading to altered gene expression (49,50). Whether ketosis affects expression of proteins associated with cytokine signal transduction pathways or has similar effects on endothelial cells, TNF-α secretion is not known and needs to be examined.

In conclusion, this study has demonstrated for the first time that hyperketonemia can increase TNF-α secretion in a U937 monocyte cell culture model and in type 1 diabetic patients. Figure 8 summarizes the proposed mechanism by which ketosis increases blood levels of TNF-α in diabetes. The effect of the ketone body AA on TNF-α secretion is apparently mediated by cAMP deficiency and the activation of PKA, along with p38-MAPK and NFκB. The evidence that the antioxidant NAC can prevent the secretion of TNF-α in AA-treated cultured monocytes needs to be explored at the clinical level to determine whether its supplementation can prevent or delay the excess vascular disease observed among the diabetic patient population.

This study was supported by a Stiles Grant Award by the Louisiana State University Health Sciences Center at Shreveport.

The authors thank Georgia First for editing the manuscript.