Molecular Mechanisms of High Glucose-Induced Cyclooxygenase-2 Expression in Monocytes

Actinomycin D, azaserine, apocynin cycloheximide, glucosamine, and thenoyltrifluoroacetone (TTFA) were from Sigma-Aldrich Chemicals (St. Louis, MO). SB202190, bis-indolylmaleimide (GFx), AG490, and N-acetylcysteine (NAC) were from Calbiochem (San Diego, CA). PD-98059, anti-phospho-cAMP responsive element binding (CREB), and anti-CREB antibodies were from Cell Signaling (Beverly, MA); Bay11-7082 [(E)-3-(4-Methylphenylsulfonyl)-2-propenenitrile] from BIOMOL Research Laboratories (Plymouth, PA); and ³²P-α-UTP (3,000 Ci/mmol) from Perkin-Elmer Life Sciences. RT-PCR reagents were from Applied Biosystems (Foster City, CA), and RPA III kit and Quantum RNA 18S Internal Standards were from Ambion (Austin, TX). Luciferase assay system was from Promega (Madison, WI). PGE₂ enzyme immunosorbent assay (EIA) kit was from Cayman Chemical (Ann Arbor, MI). Anti-COX-2 antibody was from Santa Cruz Biotechnology (Santa Cruz, CA).

Human THP-1 monocytic cells were cultured in RPMI-1640 medium as previously described (8,36). For high glucose culture, cells were treated with 15–20 mmol/l d-glucose (high glucose). In some experiments, THP-1 cells were pretreated with actinomycin D (0.1 μg/ml), cycloheximide (0.5 μg/ml), Bay11-7082 (NF-κB inhibitor, 10 μmol/l), AG-490 (Janus tyrosine kinase [JAK] inhibitor, 100 μmol/l), SB202190 (p38 mitogen-activated protein kinase [MAPK] inhibitor, 1 μmol/l), GFx (PKC inhibitor, 0.5 μmol/l), PD-98059 (MEK inhibitor, 25 μmol/l), azaserine (inhibitor of glutamine: fructose-6-phosphate amidotransferase [GFA], 5 μmol/l), NAC (antioxidant, 100 μmol/l), apocynin (NADPH oxidase inhibitor, 30 μmol/l), or TTFA (mitochondrial complex II inhibitor that blocks mitochondrial superoxide production, 10 μmol/l). They were then incubated in normal glucose medium or with high glucose for various time periods.

Fifty to sixty milliliters of blood from adults with clinically established type 1 diabetes (three subjects, Pn1, Pn2, and Pn3, with HbA_1c levels of 8.0, 8.4, and 6.5%, respectively) or with type 2 diabetes (two subjects, Pn4 and Pn5, with HbA_1c levels of 7.5 and 7.4%), or normal healthy donors (four subjects, N1, N2, N3, and N4) were collected in the presence of anti-coagulant. Informed consent was obtained from all participating subjects in accordance with approved IRB guidelines. Monocytes were isolated from the blood as previously described (36).

THP-1 cells (2 × 10⁶/sample) were cultured in medium containing 5.5 (normal glucose) or 20 mmol/l (high glucose) glucose for various time intervals. Total RNA isolation and relative RT-PCRs were performed as previously described (8,36).

The 5′ and 3′ primers for human COX-2 were 5′-ATCTACCCTCCTCAAGTCCC-3′ and 5-′TACCAGAAGGGCAGGATACAG-3′ (product size 708 bp). The 5′ and 3′ primers for COX-1 were 5′-CCGGATGCCAGTCAGGATGATG-3′ and 5′-CTAGACAGCCAGATGCTGACAG-3′ (529 bp). Results are expressed as fold stimulation over normal glucose after normalizing with paired 18S RNA levels. Ribonuclease (RNase) protection assays (RPAs) were performed as previously described (36).

Construction of the deletion mutants containing specific regions of the human COX-2 gene promoter in the luciferase reporter vector pGL3 Basic (Promega) was accomplished by PCR amplification. Plasmid hCOX-2 (−1,437/+127) was generated by deleting 5,673 bp DNA fragment from upstream of the recombinant plasmid containing firefly luciferase gene under the control of ∼7,273-bp promoter region of the hCOX-2 gene (generous gift from Dr. Thomas McIntyre, University of Utah, Salt Lake City) by digesting with restriction enzyme Kpn-I. The resulting plasmid was religated using T4-DNA ligase, yielding the luciferase gene under the control of −1,437/+127 hCOX-2 promoter region. Deletion constructs containing various promoter regions of hCOX-2, namely from −860 to +127, −360 to +127, −218 to +127, −123 to +127, and −52 to +127 were generated using the plasmid containing ∼7 kb promoter region of the hCOX-2 gene as a template. The following primers were used: upstream primers, from −860, 5′GGTACCCACATTAACTATTTACAG3′; from −360, 5′GGTACCCCAAGGCGATCAGTCCAG3′; from −218, 5′GGTACCTACCCCCTCTGCTCCCAA3′; from −123, 5′GGTACCTTTTTTAAGGGGAGAGG3′, from −52, 5′GGTACCCATGGGCTTGGTTTTC3′; and downstream primer +127, 5′AAGCTTCGGGCAGGGCGCGGCGC3′. All upstream PCR primers contained Kpn-I restriction sites (italic), and the downstream primer contained HindIII recognition site (italic), that forced cloning of the fragments in the desired orientation into the pGL3 Basic vector. Orientation and sequence of all constructs were verified by direct sequencing using the ABI PRISM 377 DNA sequencer.

THP-1 cells in six-well plates were transfected with 1 μg of the indicated COX-2 promoter plasmids, the control pGL3-Luc plasmid, the pCMV-mIκB plasmid (generous gift from Dr. E. Zandi, University of S. California), or the dominant-negative CREB plasmid, KCREB (39) (generous gift from Dr. Jane E.B. Reusch, University of Colorado) using Lipofectamine 2000, and then luciferase activities determined (8,36).

Data are expressed as mean ± SE of multiple experiments. Paired Student’s t tests were used to compare two groups and ANOVA for multiple comparisons. P values <0.05 were considered statistically significant.

We initially evaluated whether high glucose could induce COX-2 and COX-1 mRNAs in THP-1 cells. The cells were treated with or without high glucose (15 mmol/l) for various time intervals from 24 to 72 h (Fig. 1). RNA extracted from these experiments was subjected to relative RT-PCR analyses with 18S as internal control. Results show that high glucose treatment markedly increased COX-2 mRNA expression by 24 h, which declined by 72 h (Fig. 1A). Figure 1B, which contains data from multiple experiments normalized to 18S, shows that the effects of high glucose at 24, 48, and 72 h were statistically significant. Figure 1C shows that high glucose-induced COX-2 mRNA appeared as early as 8 h but not at 4 h. In contrast, COX-1 mRNA was constitutively expressed and not altered during the 24–72 h time periods (Fig. 1D). This suggests that high glucose specifically upregulates the inducible COX-2 isoform and not COX-1. To determine the dose response effects of high glucose, THP-1 cells were treated with different concentrations of high glucose (15, 20, and 25 mmol/l), and COX-2 mRNA levels were determined. Figure 1E shows that the maximum effects of high glucose on COX-2 mRNA induction was at 15–20 mmol/l. Furthermore, Fig. 1F shows that COX-2 mRNA was increased only by high glucose and not by equimolar concentrations of mannitol (osmolality control) or 2-deoxy glucose (control for glucose metabolism), indicating the specificity of high glucose effects. COX-1 mRNA expression was not altered in any of these conditions (Fig. 1F, lower panel).

To further confirm the RT-PCR data, we performed sensitive RPAs using ³²P-labeled COX-2 and 18S antisense riboprobes. Figure 2 shows a representative autoradiogram. A clear protected COX-2 RNA (254 nt) band was seen in the high glucose (20 mmol/l)-treated sample (lane 4), whereas no protected band was seen in the untreated control sample (lane 3). The 18S protected bands (internal control) of the predicted sizes are seen in both control and treated lanes. Lane 1 shows probes used in the experiments, whereas lane 2 is experiment without RNA. Lane 5 is labeled molecular weight markers. These results substatiate the RT-PCR data.

Although THP-1 cells have several properties of monocytes, they may not fully represent peripheral blood monocytes. Hence, to determine the in vivo relevance, we evaluated whether peripheral blood monocytes from diabetic individuals express elevated COX-2 mRNA levels relative to nondiabetic control volunteers. Figure 3A shows that monocytes from normal nondiabetic volunteers (N1, N2, N3, and N4) have extremely low and almost insignificant COX-2 mRNA expression levels. However, strong expression of COX-2 mRNA is seen in the three type 1 diabetic patients (Pn1, Pn2, and Pn3) run in duplicate. Furthermore, RNA from two type 2 subjects (Pn4 and Pn5) also show elevated COX-2 mRNA levels relative to the control subjects. Figure 3B shows bar graph quantitation of the PCR data after normalization for 18S levels. Interestingly, it appears that COX-2 mRNA levels in the type 2 patients (4.5-fold over control average, last bar on right) are somewhat lower than type 1 patients (31.0 ± 3.5-fold over control subjects) since PCR conditions and RNA amounts were kept identical in all samples. With this limited sample size, there seemed to be no specific correlation with HbA_1c levels (see research design and methods). Overall, these results show that diabetic monocytes have high levels of COX-2 expression (20.0 ± 6.6-fold, third bar) that may contribute to several inflammatory events.

The RT-PCR and RPA data clearly showed significant induction of COX-2 mRNA by high glucose. We therefore evaluated whether COX-2 protein levels were also regulated. Western blot analysis with specific COX-2 antibody was carried out using total protein from control and high glucose-treated THP-1 cells. Figure 4A shows that COX-2 protein appeared ∼24 h after high glucose treatment and remains sustained up to 72 h (Fig. 4A, upper panel). This is consistent with the mRNA data and demonstrates that COX-2 protein and its mRNA peak around similar time points. Equal protein loading was confirmed by probing with an anti-actin antibody (Fig. 4A, lower panel). To determine whether the induction of COX-2 mRNA and protein were also associated with increased COX-2 enzyme activity, we examined the levels of the COX-2 product PGE₂. High glucose-treated THP-1 cells had significantly elevated PGE₂ levels relative to normal glucose (410 ± 43 fg/ml, P < 0.001) (Fig. 4B) as measured by specific EIA.

Time course analysis showed that high glucose-induced expression of COX-2 mRNA appears at ∼8 h, peaks at ∼24 h, and then declines by 72 h (Fig. 1). To determine whether high glucose-induced COX-2 mRNA expression at 24 h is due to increases in transcription and/or de novo protein synthesis, THP-1 cells were pretreated for 1 h with actinomycin-D, an inhibitor of transcription, or cycloheximide, a protein synthesis inhibitor, and then treated with high glucose for 24 h. RT-PCR analyses showed that although actinomycin D completely blocked high glucose-induced COX-2 mRNA expression; cycloheximide did not block it (Fig. 5). Thus, COX-2 induction, at least at 24 h, is due to increases in transcription. However, further studies are needed to determine the role of de novo protein synthesis at other time periods because COX-2 is known to be regulated transcriptionally and post-transcriptionally via mRNA stabilization.

To examine whether high glucose at 24 h can induce transcription from the COX-2 promoter, we made several constructs with luciferase gene under the control of various regions of the hCOX-2 promoter. As shown in Figs. 6A and B, deletion constructs pCOX-2 (−1,430/+127), pCOX-2 (−860/+127), pCOX-2 (−360/+127), pCOX-2 (−216/+127), pCOX-2 (−123/+127), and pCOX-2 (−52/+127) were transfected into THP-1 cells and then treated with high glucose for 24 h. High glucose could significantly increase luciferase activity in pCOX2 (−360/+127) and pCOX2 (−123/+127) transfected samples (Fig. 6B) (P < 0.001). However, the longer promoter constructs did not yield significant increase in luciferase activity (Fig. 6B). This suggests that high glucose-induced COX-2 upregulation is mediated by key promoter elements present within the 360-bp promoter region, but certain repressive elements could be present more upstream.

The −360 region of the hCOX-2 promoter (−360/+127) contains several cis-acting elements. We were interested in the potential involvement of two cis-acting elements near the 5′ end of the coding region, namely NF-κB and CRE binding sites (Fig. 6A), both of which have been implicated in COX-2 gene transcription by various agonists. We first explored NF-κΒ involvement by cotransfecting THP-1 cells with phCOX2 (−360/+127) and a NF-κB super-suppressor IκB (mut) plasmid that suppresses NF-κB activation. Figure 6C shows that mutant IκB significantly inhibited high glucose-induced COX-2 promoter activity but did not alter control pGL3-Luc. This indicates involvement of the proximal NF-κΒ site within the −360 region in the high glucose response.

Since significant increase in reporter gene luciferase activity was also observed in the −123/+127 construct that has a CRE site, we examined the potential involvement of CREB transcription factor. This was done by cotransfection of pCOX2 (−123/+127) or pCOX2 (−360/+127) constructs with a mutant CREB plasmid (KCREB) that inhibits the action of CREB by forming heterodimers with the endogenous protein that are incapable of binding target DNA sequences (39). Figure 6D shows that KCREB completely abrogated luciferase activity induced by the −123/+127 construct and partially blocked activity with the −360/+127 construct, thus suggesting involvement of CREB binding to CRE elements in high glucose-induced COX-2 expression. To further validate this, we examined whether high glucose can increase CREB activation. The Western blot in Fig. 6E shows that high glucose led to a marked time-dependent increase in the phosphorylation of CREB at Ser 133, which is known to increase the transcriptional activity of CREB. These results suggest that high glucose-induced COX-2 regulation can be mediated by not only NF-κΒ but also by CREB.

To determine the key signal transduction pathways involved in high glucose-induced COX-2 mRNA, we evaluated the effects of inhibitors of pathways known to be activated by high glucose, including signaling kinases and oxidant stress. We pretreated THP-1 cells with SB202190 (SB, p38MAPK inhibitor), AG-490 (AG, JAK inhibitor), PD98059 (PD, extracellular signal-related kinase (ERK) 1/2 MAPK inhibitor), GFx (GF, PKC inhibitor), or Bay 11-7082 (Bay, NF-κΒ inhibitor) and then treated with high glucose for 24 h. Results (Figs. 7A and B) show that high glucose-induced COX-2 mRNA expression was significantly blocked by inhibitors of p38MAPK, PKC, and NF-κΒ, but not JAK or ERK1/2 MAPK (Fig. 7B). These results implicate the involvement of multiple pathways, including the p38 MAPK, PKC, and NF-κΒ in high glucose-induced COX-2 mRNA expression.

Excess glucose may exert effects via metabolism through the hexosamine biosynthesis pathway by GFA activation and glucosamine formation (40). Hyperglycemia can also increase mitochondrial superoxide production (4,9). In monocytes, high glucose-induced superoxide generation also involves NADPH oxidase (8,41). To elucidate the role of some of these pathways in high glucose-induced COX-2 mRNA expression, cells were preincubated with an antioxidant (NAC), NADPH oxidase inhibitor (apocynin), mitochondrial complex-II inhibitor that blocks superoxide production (TTFA), and GFA inhibitor (azaserine). The representative RT-PCR blot in Figs. 8A and the bar graph quantitation in Fig. 8B show that all four inhibitors significantly blocked high glucose-induced COX-2 mRNA expression in THP-1 cells. Interestingly, direct treatment of THP-1 cells with glucosamine for 24 h could also directly induce COX-2 expression (data not shown), further supporting the idea that glucose metabolism is involved in high glucose-stimulated COX-2 induction.

COX-2 and its products, such as PGE₂, have been implicated in several inflammatory diseases such as atherosclerosis and in the inhibition of insulin secretion and islet dysfunction related to type 1 diabetes (19–22,31,32). Although changes in vascular prostaglandin production are implicated in vascular reactivity derangements associated with diabetes (42), the role of COX-2 in diabetic monocytes and atherosclerosis is not very clear. RAGE ligation can stimulate COX-2 expression, which can induce monocyte activation (36). In the present study, we demonstrated for the first time that treatment of monocytes with high glucose can significantly increase COX-2 expression and activity in human monocytes. In contrast, high glucose did not alter the COX-1 isoform. High glucose-induced COX-2 expression was specific, since mannitol and 2-deoxyglucose had no effect. Furthermore, in vivo relevance to diabetes complications was suggested by our observation that high levels of COX-2 mRNA are present in monocytes from type 1 or type 2 diabetic patients, but not from normal volunteers.

We also demonstrated that high glucose-induced COX-2 mRNA expression at 24 h was transcriptionally regulated because actinomycin D blocked expression; furthermore, high glucose could increase transcription from a minimal human COX-2 promoter luciferase construct. Promoter deletion analysis and use of pharmacological and genetic inhibitors of NF-κB demonstrated the essential role of NF-κB in high glucose-induced COX-2 gene transcription. The human COX-2 promoter has two NF-κB consensus sites (25), one located within −455 to −428 bases and the other within −232 to −205 bases from the transcriptional start site. We noted that high glucose response was confined to a region within the 360 bp upstream of the 5′ end, indicating that the proximal NF-κΒ site (−232 to −205) is essential for high glucose-induced COX-2 transcription. Interestingly, this differs from the actions of the RAGE ligand, S100b, that seemed to induce COX-2 via the distal NF-κΒ site (−455 to −428). Another key difference between the actions of S100b and high glucose is our present observation for the first time that high glucose-induced COX-2 also involves the proximal CRE site and activation of CREB. Whereas reports have shown the effects on high glucose on CREB in vascular smooth muscle cells (43) and mesangial cells (44), our results suggest a novel new role for CREB in monocytes. Thus, whereas both AGEs and high glucose can induce COX-2 expression in monocytes, they differ in the molecular mechanisms of regulation.

We also evaluated the signaling mechanisms involved in high glucose-induced COX-2 mRNA. Using pathway-specific inhibitors, we noted that activation of NF-κB, PKC, and p38 MAPK were involved in high glucose-induced COX-2 mRNA. This indicates the operation of multiple related pathways in COX-2 regulation in monocytes under diabetic conditions. In contrast, the JAK-STAT and ERK pathways did not seem to be involved.

In addition, high glucose-induced COX-2 expression was blocked by NAC, azaserine apocynin, and TTFA, thus suggesting the involvement of oxidant stress, glucosamine production, NADPH oxidase, and mitochondrial superoxide, respectively. These are related pathways since mitochondrial superoxide (O₂^−·) production has been implicated in the activation of several of these downstream pathways by high glucose including PKC (4). NADPH oxidase and PKC-α have also been implicated in monocytic O₂^−· release (41,45). Furthermore, high glucose-induced expression of the inflammatory chemokine MCP-1 in monocytes was regulated by both mitochondrial superoxide and NAPDH oxidase (8). Thus, in monocytes, high glucose-induced O₂^−· release may be via both NADPH oxidase and mitochondrial sources. This increased O₂^−· may trigger the activation of NF-κΒ to induce proinflammatory genes such as COX-2 in diabetic patients.

Some of the downstream consequences of high glucose effects are mediated by the hexosamine biosynthesis pathway, in which fructose-6-phosphate is converted to glucosamine-6-phosphate by rate limiting enzyme GFA (40). Glucosamine can directly induce growth factor gene expression and simulate several glucose effects (44,46). Evidence shows that high glucose and glucosamine-induced PKC and cAMP-dependent protein kinase signaling pathways may participate in hexosamine-induced fibronectin synthesis via CREB phosphorylation and nuclear CREB regulated gene expression via CRE elements (44). In our studies, high glucose-induced COX-2 mRNA was inhibited by a GFA inhibitor azaserine, suggesting the involvement of the hexosamine pathway. Furthermore, our unpublished results indicate that glucosamine can also increase COX-2 mRNA expression in THP-1 cells. Interestingly, since we demonstrated the involvement of a CRE site in the proximal COX-2 promoter, and high glucose also phosphorylated CREB, it is possible that high glucose conversion to glucosamine contributes to COX-2 induction via CREB activation.

In certain cells COX-2 induction is due to increase in post-transcriptional mRNA stabilization rather than increased transcription (29,30). This has been attributed to the presence of several adenylate uridylate-rich elements in the 3′ untranslated region of the COX-2 mRNA. Several mRNA stabilizing proteins have been shown to bind to this region (29,30,47,48). In our studies, it appears that high glucose-induced COX-2 mRNA at 24 h is primarily regulated transcriptionally. However, since COX-2 mRNA is seen at even ∼8 h, some of the early effects of high glucose may be partly mediated by mRNA stabilization. Additional studies are needed to evaluate this.

Evidence shows that high glucose culture of monocytes can lead to the transcriptional regulation of TNF-α and MCP-1 via oxidant stress-dependent key signaling mechanisms and NF-κB (8,12). High glucose culture of THP-1 monocytes also significantly increased their adherence to human aortic endothelial cells (8). Very recently, gene profiling with DNA arrays demonstrated that high glucose regulates several inflammatory genes in THP-1 monocytes (8), many of which are regulated by NF-κB. The present data showing the importance of NF-κB in COX-2 regulation by high glucose further underscores the key role played by this transcription factor in diabetes complications. Our results also implicate a new proinflammatory role for CREB in monocytes. Thus the diabetic environment results in the induction of multiple inflammatory mediators by cells in the vessel wall and in circulation, thereby leading to intra- and intercellular activation. Interestingly a very recent report demonstrated enhanced inflammatory reactions and COX-2 expression in human diabetic atherosclerotic plaque macrophages (49). Thus specific COX-2 inhibitors might be beneficial for diabetes complications. Induction of COX-2 by high glucose suggests that eicosanoid products, apart from their vasoactive roles, can also amplify monocyte activation and related diabetes complications.

This study was supported by grants from the National Institutes of Health (NIH RO1 DK065073, PO1 HL55798, and RO1 DK58191), the Juvenile Diabetes Research Foundation (JDRF 2001-108), and in part by a General Clinic Research Center grant from National Center for Research Resources (MO1RR00043).

We thank Linda Lanting for technical assistance.