APOE Genotype-Function Relationship: Evidence of −491 A/T Promoter Polymorphism Modifying Transcription Control but Not Type 2 Diabetes Risk

Background The apolipoprotein E gene (APOE) coding polymorphism modifies the risks of Alzheimer's disease, type 2 diabetes, and coronary heart disease. Aside from the coding variants, single nucleotide polymorphism (SNP) of the APOE promoter has also been shown to modify the risk of Alzheimer's disease. Methodology/Principal Findings In this study we investigate the genotype-function relationship of APOE promoter polymorphism at molecular level and at physiological level: i.e., in transcription control of the gene and in the risk of type 2 diabetes. In molecular studies, the effect of the APOE −491A/T (rs449647) polymorphism on gene transcription was accessed by dual-luciferase reporter gene assays. The −491 A to T substitution decreased the activity (p<0.05) of the cloned APOE promoter (−1017 to +406). Using the −501 to −481 nucleotide sequence of the APOE promoter as a ‘bait’ to screen the human brain cDNA library by yeast one-hybrid system yielded ATF4, an endoplasmic reticulum stress response gene, as one of the interacting factors. Electrophoretic-mobility-shift assays (EMSA) and chromatin immuno-precipitation (ChIP) analyses further substantiated the physical interaction between ATF4 and the APOE promoter. Over-expression of ATF4 stimulated APOE expression whereas siRNA against ATF4 suppressed the expression of the gene. However, interaction between APOE promoter and ATF4 was not −491A/T-specific. At physiological level, the genotype-function relationship of APOE promoter polymorphism was studied in type 2 diabetes. In 630 cases and 595 controls, three APOE promoter SNPs −491A/T, −219G/T (rs405509), and +113G/C (rs440446) were genotyped and tested for association with type 2 diabetes in Hong Kong Chinese. No SNP or haplotype association with type 2 diabetes was detected. Conclusions/Significance At molecular level, polymorphism −491A/T and ATF4 elicit independent control of APOE gene expression. At physiological level, no genotype-risk association was detected between the studied APOE promoter SNPs and type 2 diabetes in Hong Kong Chinese.


Introduction
Type 2 diabetes is a multi-factorial and polygenic disease which makes up 90% of all cases of diabetes. Dyslipidemia is one of the risk factors for type 2 diabetes as well as for diabetic complications, such as coronary heart disease, diabetic nephropathy and retinopathy [1,2].
Apolipoprotein E (apoE) is a 34 kD protein which plays a central role in lipid metabolism. Two coding polymorphisms of the gene resulting in three protein variants apoE2, apoE3 and apoE4 incur isoform-dependent risk associations with Alzheimer's disease, atherosclerosis and coronary heart disease [3]. ApoE is also an important molecule in the development and progression of diabetes. A recent meta-analysis of genome-wide linkage studies of quantitative lipid traits in families ascertained for type 2 diabetes with diverse ethnic backgrounds identified one of the linkage region for lipid traits on chromosome 19q13. .43 which included the APOE gene locus (19q13.2) [4]. Another metaanalysis on data of 5423 cases and 8197 controls extracted from 30 studies provided evidence that the APOE2 allele carriers have elevated risk for type 2 diabetes [5]. Aside from the isoformdependent effects, plasma apoE has been associated with the risk of cardiovascular diseases in a dose-dependent manner [6]. An increment of plasma apoE in type 2 diabetic patients as compared to healthy controls has been reported [7]. It is conceivable that the transcriptional activity of APOE may affect plasma concentration of the protein. An increasing body of evidence has associated APOE promoter polymorphisms with human diseases. For example, the APOE promoter 2491A genotype has been associated with a higher plasma level of apoE and increased risk for Alzheimer's disease as compared to its 2491T counterpart [8,9]. In spite of the association between APOE promoter polymorphisms and disease risks, the underlying mechanisms responsible for controlling APOE gene expression remain elusive.
In this study, we aimed at elucidating the genotype-function relationship of APOE promoter polymorphism at molecular and physiological levels. At molecular level, we further investigated the transcriptional control mechanism at the 2491A/T-spanning region of APOE. At physiological level, we examined the association of APOE promoter polymorphisms 2491A/T (rs449647), 2219G/T (rs405509) and +113G/C (rs440446) with the risk of type 2 diabetes. These three SNPs were chosen for analysis based on their previously reported association with Alzheimer's disease and coronary heart disease [10,11]. Investigation of association between these SNPs and type 2 diabetes has not been reported. Our molecular studies demonstrate for the first time that ATF4, a key transcription factor mediating ER (endoplasmic reticulum) stress response and regulates lipid and glucose homeostasis in mammals [12], is interactive with the APOE promoter and controls the expression of the gene independent of the control elicited by the 2491A/T polymorphism. At physiological level, no association was detected between the three APOE promoter SNPs and the risk of type 2 diabetes in Hong Kong Chinese.

2491A/T polymorphism regulates APOE promoter activity
The effects of 2491A/T polymorphism on the activities of the APOE gene were analyzed by dual-luciferase reporter assay. Figure 1A shows that 2491A to T single nucleotide substitution significantly changed APOE promoter activity in WRL-68 (human hepatic embryonic) and U-87 (human astrocytic) cell lines (decrease of 14% and 36%, respectively, p,0.05). These results support that the APOE promoter 2491 polymorphism is functionally active and elicits similar regulatory effects on APOE transcription in human cell lines of liver and brain origins. Further analyses by EMSA revealed the interaction of nuclear proteins with APOE promoter 2491A/T-spanning sequence (2521 to 2461) ( Figure 1B). Subsequent studies were carried out to identify the potential interacting transcription factors with APOE promoter 2491A/T-spanning region.
ATF4 is a candidate transcription factor binding to APOE promoter 2491A/Tspanning sequence Yeast one-hybrid screening of a human brain cDNA library identified ATF4 being one of the candidate transcription factors interactive with the APOE 2491-spanning sequence. ATF4 belongs to ATF/CREB transcription factor family that mainly involves in the PERK endoplasmic reticulum (ER) stress response. It is well known that the dysfunction of ER stress responses can result in various diseases including diabetes, Alzheimer's disease and inflammation [13]. Further bioinformatic analysis using the TRANSFAC 6.0 database identified a sequence homologous to the ATF/CRE core binding site (TGACCTTA, 2486 to 2479) adjacent to the studied APOE promoter 2491A/T-spanning region. Taken together, ATF4 was selected as a candidate transcription factor for further investigation of its interaction with APOE promoter and the regulation of APOE gene transcription.

ATF4 interacts with APOE promoter 2491A/T-spanning sequence in vitro and in vivo
To further verify the direct interaction between APOE promoter and ATF4, EMSA was performed using purified recombinant Histagged ATF4 and 2491A or 2491T probes (2521 to 2461). Clear shift of bands were detected for both 2491A and T probes ( Figure 2A). Addition of excessive unlabeled probes completely competed with the labeled probe for ATF4 binding, indicating this binding is specific. Further super-shift assay with ATF4-specific antibody lead to a super-shift band for both 2491A and 2491T probes as shown in Figure 2B. ChIP assays were performed to further confirm the binding of ATF4 and APOE promoter 2491A/T-spanning locus in vivo. Results in Figure 2C showed that APOE promoter 2617 to 2344 region encompassing the 2491A/T-spanning site could be amplified only from the anti-ATF4 antibody immuno-precipitated complex, while PCR product was merely detectable in the IgG control both in 293 cells (2491AA genotype) and WRL-68 cells (2491TT genotype). Taken together, these results further substantiated the physical interaction between ATF4 and the APOE promoter at the 2491A/T-spanning region both in vitro and in vivo.

ATF4 regulates APOE transcription and expression
To elucidate the biological effects of ATF4 interaction with APOE promoter 2491A/T-spanning region, the APOE promoter firefly reporter constructs (with 2491A or 2491T allelic form), pcDNA3.1-ATF4 mammalian expression vector (or pcDNA3.1 empty vector) and the internal control Renilla luciferase reporter vector were co-transfected into mammalian cells. Dual-luciferase reporter assays showed that ATF4 over-expression significantly suppressed the APOE promoter (with 2491A) activity in U-87 cells by about 50% ( Figure 3A). Site-directed mutagenesis was performed to generate an APOE promoter deletion mutant D(2487 to 2469) reporter construct spanning the putative ATF4 binding site. Abolishing this putative ATF4 binding site resulted in 20% increase of APOE promoter activity as compared to the full-length APOE promoter (2491A allelic form) ( Figure 3B). Furthermore, the suppressive effect of ATF4 on APOE promoter activity was dose-dependent with statistical significant effects observed at higher dosages ( Figure 3C). ATF4 over-expression significantly down-regulated the activities of cloned APOE promoters both in 2491A and 2491T allelic forms in U-87 and WRL-68 cells ( Figure 3D) although no apparent allelic-difference was observed. These results strongly support that ATF4 modulates the transcriptional activity of APOE promoter.
We next examined the effects of ATF4 on the expression of endogenous APOE. Blockage of the endogenous ATF4 expression by siRNA in WRL68 human hepatic cells caused a 42% reduction of endogenous APOE expression ( Figure 4A). On the other hand, over-expression of ATF4 enhanced APOE mRNA expression ( Figure 4B). These results further substantiated the functional role of ATF4 in regulating APOE gene expression.

No association was detected between APOE promoter polymorphism and the risk of type 2 diabetes
To investigate the genotype-function relationship of APOE promoter polymorphism at physiological level, we tested the association between APOE promoter SNPs and the risk of type 2 diabetes in Hong Kong Chinese. In our case-control study, the control subjects (41.37610.48 years) were slightly older than the type 2 diabetic patients (40.0768.39 years) but such difference is not statistically significant after adjustment of multiple testing. There is a female preponderance in the type 2 diabetes cohort (55.1% in controls and 61% in type 2 diabetes, p = 0.039). Type 2 diabetic patients had significantly higher BMI, WHR, blood pressure, more adverse lipid profiles (high total cholesterol, high LDL-C, low HDL-c and high TG), and higher plasma glucose levels than non-diabetic control subjects (p,0.001) ( Table S1).
All three APOE proximal promoter polymorphisms were successfully genotyped with high call rates: 95.9% for 2491A/T (rs449647), 99.8% for 2219G/T (rs405509) and 98.8% for +113G/C (rs440446). All genotype distributions were in Hardy-Weinberg equilibrium (HWE) in both type 2 diabetes and nondiabetic control groups. APOE promoter 2491A/T, 2219G/T and +113G/C genotype distributions and allele frequencies were similar between the control and type 2 diabetes groups (Table 1). Analysis with gender stratification did not reveal an association between these SNPs and type 2 diabetes either. Linkage disequilibrium (LD) analysis indicated that 2219G/T and +113G/C polymorphisms form an LD block which is not linked to the APOE 2491A/T polymorphism (Table S2). No association was detected between haplotypes (formed between 2491A/T and the 2219G/T-tagged LD block) and type 2 diabetes.

Discussion
The results from the current and previous studies indicated that the polymorphism of APOE promoter at the 2491 site is functionally significant in modifying APOE gene transcription as well as in the development of diseases [6,8]. However, the transcriptional control mechanism of APOE at this locus has not been well characterized. In this study, we presented evidence that transcription factor ATF4 functionally regulates and physically interacts with the APOE promoter. Such regulation is independent of the 2491A/T polymorphism. ATF4 is widely expressed in different organs including liver, kidney, brain, spleen, heart, thymus, lung, blood cells and fibroblasts [14]. It belongs to the ATF/CREB bZIP transcription factor family and is a key transcription factor controlled under the PERK signaling pathway which is up-regulated by ER stress -a process referring to the excessive cellular protein load relative to the reserve of ER chaperones required for correct folding of newly synthesized membrane or secretary proteins [15]. ATF4 translation is induced upon ER stress followed by its migration to nucleus to induce antioxidant genes and genes of the ER protein maturation machinery [16].
ER stress and APOE have been independently associated with neurodegenerative diseases and atherosclerosis [6,8,9,13]. Also ER stress is related to the increased ATF4 expression as well as b-cell apoptosis and the subsequent development of diabetes [17,18]. Given the important role of ATF4 in ER stress responses, the control of APOE expression by ATF4 demonstrated in this study provides a plausible link between ER stress and APOE-mediated cellular function in disease processes.
Previous studies have reported the difference of APOE promoter transcription activities elicited by the 2491A/T polymorphism in human hepatocellular carcinoma HepG2 cells [19]. We have obtained similar results in other cell types representing the major apoE production tissues including WRL-68 human embryonic hepatocytes and U-87 human astrocytes. APOE promoter in the 2491A allelic form was associated with higher transcriptional activity as compared to its 2491T counterpart in both cell lines. The magnitude of change was greater in astrocytes than that of the liver cells. This may be explained by tissue-specific transcription activities. These observations support that APOE promoter 2491A/T polymorphism is functionally active in regulating APOE promoter in different tissues and justified further characterization of the transcriptional control mechanism within this region. It has been shown that nuclear proteins from HepG2 (hepatoma) and NB (neuroblastoma) cells can bind to the 2491-spanning region of the APOE promoter [20,21] . However, there has been no further resolution of the interacting proteins with the APOE 2491-spanning sequence. In this study, yeast onehybrid screening and EMSA analyses identified and verified the physical interaction between APOE promoter 2491-spanning sequence and the transcription factor ATF4. Using the purified recombinant ATF4 protein and the 2491A/T allelic oligonucleotide probes, we found ATF4 interactive with these probes in vitro. The results of ChIP assays further supported such physical interaction in vivo.
The dual-luciferase assays reported the functional effects of ATF4 on APOE promoter. ATF4 can suppress the activity of APOE promoter in WRL-68 and U-87 cells. This suppressive effect is common to both 2491A and T allelic forms. Consistently, the suppressive effect of ATF4 did not significantly differ in magnitude for the two 2491 allelic forms of the promoter in both cell lines examined.
It is curious that the cloned APOE promoter and endogenous APOE expression showed opposite responses to ATF4 overexpression. It is well known that ATF4 can partner with different transcription factors to act as a transcription repressor or activator [22]. It is likely that the endogenous promoter can recruit a wider spectrum of ATF4 partners as compared to the cloned promoter and thus renders such difference in response. Alternatively, the endogenous APOE promoter may recruit repressors which bind outside of the sequence of the cloned promoter to elicit a different response.
The 2491A/T-dependent difference in APOE promoter activity observed in this study and previous report cannot be explained by ATF4 alone. It is speculated that other transcription factors are involved in the 2491A/T-dependent transcription regulation of APOE. Such speculation is supported by two lines of evidence. First, the EMSA assays using mammalian cell nuclear extracts showed two shift-bands, indicating more than one group of nuclear proteins can bind to the APOE promoter 2491-spanning sequence. Second, ATF4 is known to interact with several transcription factors, i.e., C/EBP and p300 which have putative binding sites within the studied APOE promoter 2491-spanning sequence (predicted by TRANSFAC 6.0 database) [23,24,25]. It is likely that C/EBP, p300 as well as other putative factors interactive with ATF4 can elicit general and/or 2491A/Tspecific control to APOE transcription. Further investigation is required to elucidate the identities and interactions of additional transcription factors with the 2491-spanning region of APOE promoter.
At molecular level, we have demonstrated the relationship between the APOE 2491A/T genotype and the transcription phenotype of the gene. At physiological level, the association between the APOE promoter polymorphism and the risk of type 2 diabetes was tested for the 2491A/T, as well as for two additional SNPs 2291G/T and +113G/C. The fact that the 2291 and +113 loci belong to a different LD block adjacent to the 2491 locus opened the possibility for further exploring the APOE promoter genotype-function relationship by haplotype analysis. No SNP or haplotype association was detected between the APOE promoter and the risk of type 2 diabetes. Since we previously reported an association between APOM gene polymorphism and the disease duration of type 2 diabetes [26], such association was also tested for the three APOE promoter SNPs. Again, we did not detect an  association between APOE promoter SNPs and the duration of type 2 diabetes (data not shown). However, it is important to note that with the given effect and sample size, our study is underpowered to reject the null hypothesis. In conclusion, at molecular level, APOE gene transcription is under the independent control of the promoter 2491A/T polymorphism and the ER stress-responsive transcription factor, ATF4. At physiological level, there is a lack of evidence of association between the three APOE promoter SNPs 2491A/T (rs449647), 2219G/T (rs405509), and +113G/C (rs440446) and the risk of type 2 diabetes. These results encourage further investigation of APOE promoter regulation under ER stress but discourage the application of the three studied SNPs as risk markers for type 2 diabetes.

Yeast one-hybrid screening
Yeast one-hybrid system was adopted to screen for candidate transcription factors interacting with the APOE promoter 2491A/ T-spanning sequence. Yeast strain Saccharomyces cerevisiae YM4271 and reporter vector pHISi-1 carrying the HIS3 reporter gene were obtained from BD Bioscience (Palo Alto, CA, USA). The reporter constructs pHISi-1-491A and pHISi-1-491T were generated with three head-to-tail copies of the 21-bp bait sequence (59-CTG GTC TCA AAC TCC TGA CCT-39) or (59-CTG GTC TCA ATC TCC TGA CCT-39), which corresponds to the human APOE promoter 2491A/T-spanning sequence (2501 to 2481).
Two independent sets of screenings for the GAL4 activating domain AD tagged human brain MATCHMAKER cDNA library (BD Bioscience) were performed with the constructed yeast reporter strains, pHISi-1-491A and pHISi-1-491T, respectively. Following the BD Clontech MATCHMAKER one-hybrid system standard procedure, after three rounds of selections, 25 and 23 positive clones were recovered from pHISi-1-491A and pHISi-1-491T baits screening respectively, with DNA sequencing performed to confirm the identity of the recovered clones (service provided by Macrogen, Seoul, Korea).

Plasmid constructs
Previously characterized APOE proximal promoter region 21017 to +406 containing the 2491A allelic form (GenBank accession No.: AF055343) [19] was amplified by PCR using commercial human genomic DNA as template (Promega, Madison, WI, USA) and subcloned into firefly luciferase reporter pGL3-basic vector (Promega). Site-directed mutagenesis was performed using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA) to introduce A to T substitution at position 2491. APOE promoter 2491A/T-spanning sequence deletion mutant D (2487 to 2469) was generated by PCR overlap extension mutagenesis method.

Cell transfection
For ATF4 overexpression, 1 mg of pcDNA3.1-ATF4 expression vector or the control pcDNA3.1 vector was transiently transfected into the cells using Lipofectamine2000 (Invitrogen) method following manufacturer's instructions. Cells were harvested for RNA extraction 30 hrs post-transfection.

RNA extraction and real-time PCR
Total RNA was extracted from the cells using Tri Reagent (Molecular Research Center, Inc, Cincinnati, OH, USA). Reverse transcription (RT) was performed using 5 mg of total RNA in a total reaction volume of 20 ml by MMLV reverse transcriptase system (GE Healthcare, Buckinghamshire, UK).
Real-time quantitative PCR was performed on the ABI Prism 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The TaqMan probe and primer assays used were Hs00171168_m1 (ABI) for ApoE and Hs00909569_g1 (ABI) for ATF4, respectively. The relative mRNA levels were estimated by the standard method using beta-actin as the reference gene.

Purification of recombinant ATF4 in E.coli
The pET507a-ATF4 expression plasmids were transformed into E.coli BL21 (DE3, pLysS) strain (Novagen) in LB medium containing the appropriate antibiotics (50 mg/ml of chloramphenicol and 100 mg/ml of ampicillin). The expression of His-tagged ATF4 protein was induced by 0.2 mM of isopropyl-b-thiogalactopyranose and harvested after 5 hrs culture at 37uC. The HiTrap TM Chelating HP column (GE Healthcare, Amersham, UK) pre-charged with 0.1 M of NiSO 4 was used for purification of His-tagged ATF4 following the standard method [27].

Cell nuclear extract preparation
Cells nuclear extracts were prepared by NucBuster TM Protein Extraction kit (Novagen). The protein concentrations of nuclear extracts were determined by DC protein assay kit (Bio-Rad, Hercules, CA, USA) using BSA as a standard.

Electrophoretic mobility shift assay (EMSA)
The EMSA assays were performed with the use of mammalian cell nuclear extracts or purified ATF4 and DIG-labeled oligonucleotide probes according to the manufacturer's instructions (Roche, Penzberg, Germany). The 61-bp APOE promoter 2491A/T-spanning sequence were used as the double-stranded probes: 2491A 59-GTT TCA CCA TGT TGG CCA GGC TGG  TCT CAA ACT CCT GAC CTT AAG TGA TTC GCC CAC  TGT G-39 and -491T 59-GTT TCA CCA TGT TGG CCA  GGC TGG TCT CAA TCT CCT GAC CTT AAG TGA TTC  GCC CAC TGT G-39. For the supershift assays, anti-ATF4 antibody (Santa Cruz, CA, USA) was used.