Transcriptional Control of Apolipoprotein A-I Gene Expression in Diabetes

The inverse relationship between HDL cholesterol and CHD has been described in several large epidemiological studies (5). In the Framingham study, the protective effect of HDL cholesterol was so remarkable that when the concentrations reached 85 mg/dl or at least 65 mg/dl, the elevated concentrations of LDL cholesterol did not appear to increase CHD risk (7). Whereas every 1% reduction in LDL cholesterol reduces CHD risk by 2%, raising HDL cholesterol level the same amount reduces CHD risk by 3% (7). Although not unequivocally established in clinical trials, it appears that raising HDL is more effective than reducing LDL cholesterol. In support of this notion is the observation that pharmacologic interventions with gemfibrozil in subjects with low HDL cholesterol resulted in a 22% reduction in CHD risk that was attributed to the 7.2% increase in HDL cholesterol level (6). Review of the genetic disorders associated with low HDL cholesterol suggests that when reduced apo A-I expression is the underlying mechanism of low HDL, the risk of CHD is increased.

In addition to the cardioprotective role, HDL cholesterol may have a role in preventing strokes and Alzheimer’s disease and may ameliorate age-related changes in blood-brain barrier integrity (5). However, the evidence in favor of this neuroprotective role of HDL is not as conclusive as the data supporting its cardioprotective properties.

The cardioprotective effect of HDL has been attributed mostly, but not exclusively, to its major protein constituent, namely apo A-I (5,8–10). Some epidemiological studies confirm the inverse correlation between serum apo A-I concentrations and cardiovascular events (5). More importantly, various experimental manipulations targeting increased production of apo A-I are associated with reduced atherogenicity (Table 1).

The potential mechanisms of the cardioprotective effects of apo A-I include enhancement of reverse cholesterol transport, attenuation of oxidative stress, increased peroxonase activity, and enhanced anticoagulant activity (5). Despite the uncertainties as to the mechanisms of cardioprotection, the overall evidence supporting an antiatherogenic role of apo A-I is convincing. Therefore, elucidation of the biochemical pathways regulating apo A-I gene expression is fundamental to the development of new therapies for preventing atherosclerosis.

In type 2 diabetes, the reduced serum HDL cholesterol levels have been attributed to increased fractional clearance of HDL (11). This has been attributed to variety of changes including depletion of HDL of its cholesterol ester content. Thus, increased VLDL production, at least partly because of increased fatty acid flux to the liver, promotes exchange of triglyceride for HDL cholesterol ester through the action of cholesterol ester transfer protein. The triglyceride-enriched HDL is then hydrolyzed through the action of hepatic lipase or lipoprotein lipase. Apo A-I dissociates from smaller HDL and is filtered by the glomerulus and degraded in renal tubular cells (11).

In this article, we will focus on the effects of diabetes-related changes in the metabolic milieu on the transcriptional control of the apo A-I gene.

Expression of the apo A-I gene is regulated primarily at the transcriptional level. The apo A-I gene promoter contains a TATA-like motif close to the transcriptional start site, while further 5′, several cis elements regulate expression of the gene in either a positive or negative manner in response to changes in the hormonal or metabolic status. For example, several hormones have been shown to stimulate transcription of the apo A-I gene. Thyroid hormones, retinoids, estrogens, and glucocorticoids have been shown to induce apo A-I promoter activity and gene expression through proximal promoter elements located between nucleotides −235 and −144 (relative to the transcriptional start site, +1) (12). While thyroid hormones as well as retinoids act on the apo A-I promoter directly through their nuclear receptors, no glucocorticoid or estrogen receptor binding sites are present within this region. Instead, glucocorticoids stimulate binding of the transcriptional activator hepatocyte nuclear factor (HNF)-3β to the promoter (13), leading to an increase in transcription. Likewise, estradiol promotes the interaction between HNF-3β and the orphan nuclear receptor HNF-4 on the apo A-I promoter, most likely due to inactivation of the estradiol-responsive transcriptional corepressor RIP-140, increasing apo A-I gene transcription (14). Furthermore, several transcriptional repressors have been shown to suppress apo A-I gene transcription, including ARP-1 (apo A-I repressor protein-1), HNF-4, and thyroid hormone receptor. However, binding of a thyroid hormone receptor monomer to a negative thyroid hormone response element (nTRE) located 3′ of the apo A-I gene TATA-box may suppress apo A-I gene transcription only when removed from the full-length promoter (15). Further 5′ from the transcriptional start site, an insulin response core element (IRCE) is located between nucleotides −404 and −411 (16). This element binds to the ubiquitous transcription factor Sp1 and is responsible for the induction of the apo A-I gene by insulin. Thus, expression of the apo A-I gene is subject to regulation by several hormone and metabolic signaling pathways, many of which are altered in diabetes.

Hepatic expression of apo A-I protein and its mRNA is reduced in streptozotocin-induced diabetic rats (16). Apo A-I gene expression in insulin-resistant diabetic animal models is not well studied. To further study the mechanisms by which glucose and insulin regulate the levels of apo A-I, we have used the human hepatoma cell line HepG2. This cell line retains the ability to synthesize and secrete apo A-I as well as other proteins produced by normal hepatocytes. In addition, these cells express insulin receptor and glucose transporter proteins and therefore are a suitable model to study the effects of glucose and insulin on apo A-I expression.

Treatment of HepG2 cells with 22.4 mmol/l dextrose for 48 h results in a 50% reduction in apo A-I mRNA levels, whereas treatment with 100 μU/ml of insulin causes a twofold increase in apo A-I mRNA levels relative to control cells maintained in media containing 5.5 mmol/l dextrose (16). Treatment of the cells with both high concentrations of dextrose and insulin was associated with 1.3-fold induction of apo A-I mRNA, which is an intermediate response to that of dextrose and insulin. The transcriptional activity of apo A-I promoter is also suppressed by dextrose and stimulated by insulin in a dose-dependent fashion (16). Deletional analysis of the promoter showed that a 50-bp fragment spanning nucleotides −425 to −376 mediates the effects of both dextrose and insulin. Within this DNA fragment, between −411 and −404, an IRCE was identified. Mutation of this motif abolishes the effects of both insulin and dextrose. Thus, it appears that a single cis-acting element within the promoter regulates the effects of both insulin and dextrose on transcriptional activity of apo A-I gene. However, the fact that insulin but not dextrose alters IRCE binding activity suggests that there are additional carbohydrate-responsive elements remaining to be characterized, and that the IRCE may have only a permissive role for the inhibitory effects of dextrose (16).

The stimulatory effect of insulin on apo A-I promoter is also observed with insulinomimetics such as bisperoxo (1,10-phenathroline) oxovanadate (bpv) and the protein kinase C (PKC) activator phorbol ester (PDBu) (17). However, insulin sensitization with thiazolidendiones may not always be sufficient to induce apo A-I expression (18,19). It appears that insulin acts on the apo A-I promoter through at least two signaling pathways: Ras-raf and phosphatidylinositol 3-kinase (PI3-K), leading to activation of the mitogen-activated protein kinase (MAPK) and PKC kinases, respectively (20). Despite these differences, all these pathways ultimately target the transcriptional factor Sp1 (20,21). Cotransfection of Sp1 expression vector with an apo A-I IRCE reporter construct augments the actions of insulin, while reducing Sp1 levels with an antisense RNA impairs the insulin response (17). In addition, experiments with okadaic acid (a cellular phosphatase inhibitor) or other phosphatase inhibitors suggest that phosphorylation of Sp1 plays a critical role in apo A-I expression by regulating the ability of Sp1 to bind the IRCE. The various molecular signals for Sp1-mediated apo A-I transcription are summarized in Fig. 1 (21). Indeed, numerous domains at Sp1 are targets for phosphorylation by multiple protein serine/threonine kinases. Mutations at threonine 266 to alanine have been shown to inhibit growth factor-mediated induction of apo AI gene expression (22).

One of the complications commonly found in subjects with uncontrolled type1 diabetes is ketoacidosis. Apo A-I gene expression is repressed during ketoacidosis. This change could be correlated with the process of ketogenesis but not with ketone bodies (23). In cell culture studies with hepatocytes and intestinal cells, ketones and their analogs such as butyrate or isobutyramide do not significantly alter apo A-I expression (24). In contrast, acidosis, induced either by changes in the pH of the culture media or by treatment with agents known to alter intracellular pH, consistently suppresses apo A-I promoter activity and apo A-I production (25). These changes were observed in both HepG2 cells and Caco2 cells, a human intestinal epithelial cell line that retains the ability of intestinal cells to produce apo A-I. The effect of acidosis on apo A-I promoter activity was specific and could not be explained by a nonspecific toxic effect of acidosis on protein synthesis (25).

The apo A-I downregulation with acidosis occurs through a pH-responsive element (pH-RE) located within the promoter (25). Acidosis increases the specific DNA binding activity of a putative repressor protein. This pH-RE contains an nTRE and is located 3′ and adjacent to the apo A-I TATA box (15,25,26). The pH-RE identified in HepG2 cells, using transient transfection analysis of the apo A-I promoter with deletion and site-directed mutations, overlaps the apo A-I TATA element (27). This region also contains binding sites (underlined) for other transcription factors including the thyroid hormone receptor (15), δ-EF1 (δ-crystallin enhancer binding protein-1) (28), and AP-1 (activator protein-1) (29) (Table 2). Repression of apo A-I promoter action by acidosis is not altered with cyclohexamide or by actinomycin D treatment, suggesting that this process does not require de novo protein or mRNA synthesis (25). Inhibition of tyrosine kinase activity and diacylglycerol-stimulated PKC signaling with tyrophosphotin A47 and phorbol myristate acetate (PMA), respectively, does not affect the repressive effect of acidosis on apo A-I promoter (25). This is somewhat surprising since acidosis has been shown to alter PKC activation, and others have shown that tyrosine kinase activity is necessary for acidosis to stimulate immediate-early gene expression in renal epithelial cells (30). It is possible that repression by acidosis requires more than one signaling pathway, one of which may involve PKC or tyrosine kinase activity.

Overall it is clear that ketoacidosis, primarily through changes in intracellular pH, suppresses apo A-I gene expression through inhibition of promoter activity. However, not all models of ketosis show repression. For example, apo A-I mRNA and protein levels are elevated in rats placed on a high-fat ketogenic diet (24). This indicates that acidosis is a required element of apo A-I gene suppression by ketoacidosis.

One of the hallmarks of insulin resistance or the metabolic syndrome X is reduced plasma concentration of HDL cholesterol and low plasma levels of its major apoprotein, namely apo A-I (31–33). This change is mostly attributed to increased fractional clearance of HDL without a change in apo A-I production (32,33). However, it is possible that reduced apo A-I expression due to reduced responsiveness of the apo A-I gene to insulin may contribute to the reduced plasma concentrations of HDL.

The potential mediators of insulin resistance in obesity include increased plasma free fatty acid (FFA) concentrations or increased muscle and hepatic tissue content of triglycerides, increased production of leptin, tumor necrosis factor-α (TNF-α), and possible other yet unidentified humoral mediators (34).

The experimental models of insulin resistance and their effects on the apo A-I gene expression are summarized in Table 3. Because FFAs have been implicated as one of the causes of insulin resistance, we studied the effect of FFAs on insulin-stimulated apo A-I promoter activity in HepG2 cells (35). The results indicate that FFA treatment abolishes both insulin and Sp-1-stimulated activation of apo A-I promoter. However, basal apo A-I gene expression was not altered in the presence of FFAs. The FFA stearic acid had no appreciable effect on Sp1-DNA binding (35). Therefore, stearic acid must affect Sp1 function through a mechanism unrelated to regulation of Sp1-DNA binding, most likely by altering posttranslational modification (35).

Another potential mediator of insulin resistance is leptin. Cohen et al. (36) found that leptin antagonizes insulin signaling by decreasing insulin-induced tyrosine phosphorylation of insulin receptor substrate (IRS)-1 and thereby causing a state of insulin resistance (37). Plasma leptin levels correlate with body adiposity, and we recently reported that hyperleptinemia occurs also in diet-induced models of insulin resistance (38,39). However, plasma leptin levels in diabetic subjects do not significantly differ from nondiabetic control subjects when corrected for differences in body fat content (40). In addition, the ob/ob mice, which lack leptin, develop insulin resistance with the onset of obesity (41). This suggests that there are factors other than leptin that contribute to the obesity-related hyperinsulinemia and insulin insensitivity. In our unpublished studies, we have been unable to demonstrate a significant effect of leptin on apo A-I gene expression in HepG2 cells under normal culture conditions. Thus, treatment of HepG2 cells with leptin over a wide concentration range (0–100 ng/ml) did not alter apo A-I promoter activity, apo A-I protein synthesis, or apo A-I mRNA levels (unpublished observations).

Finally, TNF-α, a pluripotent cytokine, is implicated in obesity-related hyperinsulinemia/insulin resistance (42,43). Obese individuals tend to have increased plasma TNF-α levels (44,45), and in a study of a homogenous Native Canadian population, circulating TNF-α concentrations were positively correlated with the degree of insulin resistance (45). Previously published studies as well as our own observations show that TNF-α significantly downregulates apo A-I gene expression in HepG2 cells (46). There is some evidence that Sp1 activity is regulated by TNF-α. Pan et al. (47) reported that insulin treatment of rat hepatoma cells increases Sp1 and Sp3 accumulation by approximately threefold. However, exposure to TNF-α decreases Sp1 level to 50% of control values. In the same study, hepatic Sp1 levels were reduced in diabetic rats relative to control rats but could be normalized with insulin treatment (47). We have recently found that both TNF-α and interleukin (IL)-1β reduce apo A-I gene expression in a dose-dependent fashion (46). This inhibitory effect occurred at the transcriptional level. Treatment with TNF-α and IL-1β resulted in a dose-dependent reduction in apo A-I promoter activity. Using a series of promoter deletion constructs, this effect was found to be mediated through a cytokine responsive element within site A (46).

The role of prostanoids in mediating insulin action and in modulating the apo A-I expression has not been well characterized. Prostanoids are implicated in insulin signaling in the liver (48,49). Therefore, it is possible that interference with prostanoid production within the liver may alter insulin effects on apo A-I expression. In addition, some prostanoids are ligands of peroxisome proliferator-activated receptors (PPARs) (50,51), and PPARs are implicated in the regulation of apo A-I expression (52,53). Thus, it is possible that prostanoids may also have a direct role in modulating apo A-I expression. Recent work in our laboratory shows that cyclooxygenase (COX) inhibition with indomethacin or acetyl salicylic acid downregulates apo A-I protein and mRNA expression at the transcriptional level (54). This effect could not be attributed to either arachidonic acid excess or to a deficiency in various prostanoids tested, including prostaglandin I₂, thromboxane B₂, (±) 5-HETE or (±) 12-HETE, and prostaglandin E₁ and E₂ (54).

The precise underlying mechanism of indomethacin-related downregulation of apo A-I expression is not known. The observed changes could be related to COX inhibition or could be independent of COX inhibition. Although normal hepatocytes do not express COX, it is possible that inhibition of COX activity in nonparenchymal hepatic cells, such as Kupfer cells, could alter paracrine signaling.

It is noteworthy that aspirin, but not indomethacin, has been shown to decrease the expression of apo a in human hepatocytes at the transcriptional level (55). This effect was independent of COX inhibition. It is possible that both the apo A-I and apo a genes have aspirin responsive regions in their promoter. Large clinical studies are needed to address the clinical relevance of the effects of aspirin on apo A-I and apo a expression.

The reduced expression of apo A-I in obese type 2 diabetic subjects, despite the increased levels of plasma insulin, cannot be simply attributed to insulin resistance. Thus, insulin resistance and hyperinsulinemia induced with a high-fructose diet in rats is associated with increased apo A-I levels (56). Similarly, aging in rats is associated with increased expression of apo A-I, although it is accompanied with insulin resistance (57).

Finally, treatment of hepatocytes with glucosamine, a model of cellular insulin resistance, is associated with increased stabilization of apo A-I mRNA and increased expression of apo A-I protein (58).

It appears that the diabetes-related reduction in apo A-I is not only related to increased plasma clearance of the protein but also is the result of downregulation of apo A-I expression at a transcriptional level. Although the precise nature for this change is not known, it is likely that increased cytokine production may contribute significantly to the inhibition of apo A-I gene transcription.

Uncontrolled hyperglycemia is commonly associated with alterations in micronutrient status (59). Some minerals, such as chromium, vanadium, magnesium, and zinc, have either insulinomimetic effects or have permissive effects on insulin action. Deficiencies in some minerals may occur in poorly controlled diabetic subjects. In addition to their effects on insulin action, some micronutrients have been found to have an important modulatory role in apo A-I gene expression (60,61). In particular, zinc deficiency causes downregulation of apo A-I expression (59), whereas supraphysiologic concentrations of zinc, as well as chromium or vanadium, downregulate apo A-I promoter activity (61). Some of the effects of zinc deficiency may be explained by the dependence of zinc finger-containing transcription factors, such as Sp1, for the coordinating effect of zinc ions.

Vitamins, notably those with antioxidative potential, may affect apo A-I expression. The apo A-I promoter is sensitive to the oxidative state (61), and some antioxidants at high concentrations can suppress apo A-I promoter activity in Hep G2 cells (62). Clinical trials have also found that antioxidants may partially blunt HDL induction by simvastatin-niacin combination therapy (63). The precise role of micronutrient status in the diabetes-related reduction of apo A-I expression is not known. It is conceivable that when a significant deficiency state occurs, it may aggravate the lowered plasma HDL concentrations.

The cause of increased cardiovascular morbidity and mortality in diabetic subjects is not entirely clear. Observational studies suggest that it may be related to insulin resistance and associated comorbidities, including hypertension, diabetes, and an atherogenic plasma lipid profile (1).

One of the hallmarks of insulin resistance or the metabolic syndrome X is reduced plasma concentration of HDL cholesterol and low plasma levels of its major apoprotein, namely apo A-I (31). Diabetic subjects with the metabolic syndrome have significantly increased risk of cardiovascular disease compared with diabetic subjects who do not have the metabolic syndrome (64). The reduced plasma levels of apo A-I in this syndrome is mostly attributed to increased fractional clearance of HDL (32,33). However, it is possible that reduced apo A-I expression due to either reduced responsiveness of the apo A-I gene to insulin, or secondary to inhibition by the altered metabolic milieu in diabetes, notably increased cytokine production, may contribute to the reduced plasma concentrations of HDL. In addition, in type 1 diabetic subjects who experience frequent ketoacidosis, the apo A-I gene transcriptional activity is further compromised through a transcriptional repressor protein that interacts with a pH-RE within the promoter. The myriad of changes that may occur in diabetes, culminating in reduced plasma apo A-I concentrations, is summarized in Fig. 3. Future studies elucidating molecular determinants of apo A-I gene expression should help in the rational design of novel therapeutic targets to increase HDL level and reduce cardiovascular disease in diabetes.