Two Single-Nucleotide Polymorphisms with Linkage Disequilibrium in the Human Programmed Cell Death 5 Gene 5′ Regulatory Region Affect Promoter Activity and the Susceptibility of Chronic Myelogenous Leukemia in Chinese Population

Chronic myelogenous leukemias (CML) are caused by constitutively activated tyrosine kinase BCR/ABL that leads to a proliferative and survival advantage to hematopoietic progenitors (1). In particular, BCR/ABL–transformed cells exhibit enhanced proliferative capacity, altered adhesion properties, and reduced apoptosis (2, 3). It is believed that BCR/ABL–dependent inhibition of apoptosis plays an important role in leukemic cell growth and accumulation in CML patients.

For the past few years, a number of attempts have been made to identify critical genes and factors that regulate survival and apoptosis in CML cells (4). Apart from external factors (such as cytokines; ref. 5), a number of cellular molecules have been implicated in the regulation of apoptosis in BCR/ABL–transformed cells. However, little is known about the relative contribution of each of these molecules to the survival of leukemic cells.

Programmed cell death 5 (PDCD5), also designated TF-1 cell apoptosis–related gene-19 (TFAR19), is a gene cloned from TF-1 cells undergoing apoptosis in our laboratory (6). PDCD5 is evolutionarily conserved among diverse species ranging from yeast to mammals (6, 7). Initial functional studies indicated that recombinant human PDCD5 could accelerate apoptosis of some tumor cells (e.g., HeLa, TF-1, HL60, MCG-803, and MCF-7; refs. 6, 8). A recent study revealed that PDCD5 expression was lower in marrow nucleated cells in CML patients than that in normal controls (9), suggesting that reduced PDCD5 expression may play an important role in the progression of CML.

Single-nucleotide polymorphisms (SNP) are the most common type of human genetic variations. They not only serve as markers for constructing dense genetic maps but also potentially have direct roles in complex diseases as well as in differential drug responses between individuals. Increasing evidence shows that SNPs present in noncoding regions can influence gene expression by affecting regulatory elements. Functional SNPs in promoter regions have been shown to alter the activity of the promoter and subsequent levels of mRNA expression (10–13).

In the present study, we screened 1.1 kb of the 5′ upstream region of the PDCD5 gene to search common genetic variants with distinct effects on the transcriptional activity of the gene. We identified two SNPs, −27A>G and −11G>A, and −11A mutated allele disrupts an AP-2 transcription factor binding site within the promoter. Association studies were done to show the susceptibility of this SNP to CML in a Chinese population. We further analyzed the functional effects of the −27A>G/−11G>A polymorphisms by luciferase reporter assay in HeLa and HEK293 cells after nonapoptotic and apoptotic stimulations. PDCD5 expression analysis was also done in human mononuclear cells (MNC) and cell lines with different −27/−11 genotypes.

A total of 129 CML subjects were recruited from the Institute of Hematology at Peking University People's Hospital for the study. All CML cases were classified according to morphology, immunology, and cytogenetics classification (14). Ethnically and geographically matched 211 healthy controls were obtained from blood DNA bank of Peking University First Hospital and staff of Peking University Center for Human Disease Genomics. Patients and controls were well informed and agreed to participate in the study.

CML bone marrow or peripheral blood samples were collected from Clinical Diagnostic Laboratory. An EDTA anticoagulated peripheral blood sample (2 mL) was obtained from each control by venepuncture. Genomic DNA was extracted using a Blood DNA Mini kit (Watson Biotechnologies, Inc., Shanghai, PRC) according to the instructions of the manufacturer. DNA integrity and quantity were verified by agarose gel electrophoresis.

Identification of PDCD5 polymorphism. DNA samples from 12 controls and 33 patients were subjected to PCR amplification of 1.1 kb DNA fragment upstream of the transcriptional start site of PDCD5. Two pairs of primers were used to amplify two overlapping fragments for accurate sequencing. They are forward 1 [5′-CTTGAGCTCAGGAGATAGAGGCC-3′ (−1,013/−987)] and reverse 1 [5′-TGCAATGTGCTGGTGACCTAGAG-3′ (−298/−276)], forward 2 [5′-CTGCAGCCTCGAACTTCCG-3′ (−643/−625)] and reverse 2 [5′-CAAGCTCCTCGTCCGCCATG-3′ (+72/+91)], respectively. After heating at 95°C for 5 minutes, PCR amplification was done with 35 cycles: 95°C for 30 seconds, 68°C for 60 seconds, and followed by a final extension step at 72°C for 7 minutes. PCR fragments were gel-purified by E.Z.N.A. Gel Extraction kit (Omega Bio-tek, Inc., Doraville, GA) according to the instructions of the manufacturer. Sequencing analyses were done by ABI Big Dye terminator reagents 2.0 (Applied Biosystems, Foster City, CA) on an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems).

PCR-RFLP. PCR-RFLP was used for genotyping of SNPs within the PDCD5 promoter. The shorter fragment (121 bp from −46 to +75, transcription site as position 1) of the experimental subjects was amplified by forward 3 [5′-GCTCCGGGCTGGATTGGTG-3′ (−46/−28)] and reverse 3 [5′-CATGGCTCGGCGTCAGCG-3′ (+58/+75)] primers. PCR products were digested by NarI (New England Biolabs, Beverly, MA) according to the instructions of the manufacturer. Fragments were separated on a 15% polyacrylamide gel stained by ethidium bromide. To verify the genotyping results, direct sequencing of PCR products was done in 20% random samples using the reverse 3 primer.

DNA fragments from the 5′ upstream region of the PDCD5 gene between nucleotides −1,013 to +91 were amplified from an individual with −27GA/−11AG genotype. The PCR products were subcloned into pGEM-T Easy vector (Promega, Madison, WI) to obtain plasmid pGEM-PDCD5. The DNA sequences and orientations of inserts were verified by sequencing using T7 and SP6 primers. The pGEM-PDCD5 plasmids containing −27A/−11G or −27G/−11A were digested with EcoRI and subcloned into the SmaI site of the pGL3-Basic plasmid (Promega) to generate pGL3-PDCD5-11G (−27A/−11G) and pGL3-PDCD5-11A (−27G/−11A).

HeLa and HEK293 cells were cultured in DMEM supplemented with 10% FCS at 37°C in 5% CO₂. Cells were transfected by electroporation (120 V, 20 ms, BTX ECM830, Genetronics, Inc., San Diego, CA). Ten micrograms total of plasmid DNA consisting of 9 μg of reporter plasmid and 1 μg of pGFP-N1 (Clontech, Mountain View, CA; for normalization of transfection efficiency) were used for transfection for 5 × 10⁵ cells. Cells were seeded into 12-well plates at a density of 6 × 10⁴ cells per well in 1 mL medium after electroporation. Luciferase activity was assayed 24 hours after transfection. Cells were washed with PBS and lysed with 50 μL of freshly diluted reporter lysis buffer (Promega). After centrifugation, 10 μL of supernatant was added to 50 μL of the luciferase assay substrate (Promega) and the luminescence of the samples were read immediately with a POLARstar galaxy spectrometer (BMG Labtechnologies, Offenburg, Germany), in which light production (relative light units) was measured for 10 seconds. The results were normalized with equivalent quantities of green fluorescent protein (GFP). The experiment was repeated at least thrice.

HeLa cells were used for induction of apoptosis. After 24 hours of transfection, the cultures were replaced by serum-free medium for induction of apoptosis. Assay of luciferase activity and detection of apoptosis were carried out 0, 12, 24, and 36 hours after serum withdrawal. For tumor necrosis factor (TNF)-α-induced apoptosis (with cycloheximide), after 24 hours of transfection, different dosages of TNF-α with constant concentration of cycloheximide (0, 5 μg/mL cycloheximide + 1 ng/mL TNF-α, 5 μg/mL cycloheximide + 10 ng/mL TNF-α, and 5 μg/mL cycloheximide + 50 ng/mL TNF-α) were added to the culture plate. Cells were collected after 24 hours stimulation and luciferase activity was assayed as above. Apoptosis assay was carried out by detection of phosphatidylserine externalization. In brief, HeLa cells were cultured in serum-free medium, then were washed and resuspended in binding buffer containing FITC-conjugated Annexin V (25 μg/mL) for 15 minutes before analysis. Cells were collected on a FACScan flow cytometer equipped with a 488 nm argon laser and analyzed using the CellQuest software (Becton Dickinson, San Jose, CA). The methods for the luciferase assay have been described above.

Seventeen bone marrow samples from patients with wild-type genotype and 14 from patients with −27AG/−11GA genotype were subjected to mRNA expression analysis. MNCs were isolated from bone marrow aspirates by gradient centrifugation with Ficoll solution (Shanghai Huangjing Biotechnology Co., Shanghai, P.R. China). Total RNA was extracted using the TRIzol reagent (Invitrogen, Carlsbad, CA) following the instructions of the manufacturer. cDNAs were synthesized using the Thermoscript reverse transcription-PCR System (Invitrogen). Real-time quantitative PCR was done with the ABI 7000 Sequence Detection System (Applied Biosystems) using the TaqMan technology. PCR variables were as follows: 50°C, 2 minutes; 95°C, 10 minutes; followed by 50 cycles with 95°C, 15 seconds and 60°C, 1 minute. The standard expression curves for ABL and PDCD5 were first constructed from the results of amplification of serial dilutions of standard cDNA. The expression levels in each patient sample were determined by reference to the corresponding expression level on the standard curve and then normalized by the ABL gene. The following primers and probes were used: PDCD5 forward, 5′-GTGATGCGGCCCAACAG-3′; PDCD5 reverse, 5′-ATCCAGAACTTGGGCTAAGATACTG-3′; PDCD5 probe: 5′-FAM-TCTCATTTCTGCTTCCCTGTGCTTTGCT-TAMARA-3′; ABL forward Ia, 5′-TCCTCGTCCTCCAGCTGTTATC-3′; Ib, 5′-TTATCAAAGGAGCAGGGAAGAAG-3′; ABL reverse primer, 5′-CTCAGACCCTGAGGCTCAAAGT-3′; ABL probe, 5′-FAM-AGCCCTTCAGCGGCCAGTAGCATCT-TAMARA-3′.

MNC samples from some CML patients and normal controls were collected. Genotyping was done before protein extraction. HEK293 and HeLa cells were chosen to represent wild-type cell lines, whereas human monocytic leukemia cells U937 and human Jurkat T cells were chosen as with −27AG/−11GA genotype.

Cells were pelleted by centrifugation, lysed in lysis buffer [0.05% SDS, 10 mmol/L HEPES, 0.15 mol/L NaCl, 1%Triton X-100, 1 mmol/L EDTA, 1 mmol/L EGTA, and 0.5% NP40 (pH 7.4)] and incubated for 30 minutes on ice. Lysates were centrifuged at 18,000 × g for 10 minutes at 4°C and the supernatant was measured using the bicinchoninic acid protein assay reagent (Pierce, Rockford, IL). Equal amounts of protein (10 μg) were separated by SDS-PAGE and transferred onto nitrocellulose membranes (Hybond ECL; Amersham Pharmacia, Little Chalfont, United Kingdom). Membranes were blocked in TBS containing 0.05% Tween 20 containing 5% bovine serum albumin for 1 hour and incubated overnight at 4°C with anti-PDCD5 antiserum at 1:100 dilutions. Blotting with IRDye 800-conjugated anti-IgY was done in the above buffer containing 5% (w/v) bovine serum albumin. The blots were detected by the Odyssey Imaging System (LI-COR Bioscience, Inc., Lincoln, NE).

To confirm that the study groups could be regarded as Mendelian populations, the expected occurrence rates for different alleles were calculated according to the Hardy-Weinberg principle and compared with the observed occurrence rates by χ² test.

Associations for genotype or allele frequencies between different groups were analyzed by χ² test. P values were two sided and defined as P ≤ 0.05 for statistical significance.

Statistical analysis for mRNA expression levels was done with SPSS (Chicago, IL) software version 11.0. Differences between groups were evaluated with adjusted t test. Significance was established if P was 0.05. Data are shown as mean, SD, and SE.

Characterization of the single-nucleotide polymorphisms. After direct sequencing of PCR products in 12 controls and 33 CML patients, a nucleotide substitution (G>A) at 11 bp upstream from the transcriptional start site and a −27 (A > G) substitution were identified simultaneously (Fig. 1A). The mutated allele −11A disrupts a NarI restriction enzyme site (GG′CGCC, italic indicates the SNP site) and the mutated allele −27G creates a NarI site. According to the results of direct sequencing of 45 individuals, eight heterozygous genotypes −11GA were always identified simultaneously with genotype −27AG. Sequencing analysis indicated that −11G is located in CCSCRGGC (italic illustrates SNP site), an AP-2 transcription factor binding site, whereas the −27SNP is not in any potential transcriptional factor binding site (Fig. 2; ref. 15).

Genotyping and identification of linkage disequilibrium. We inferred that there may be the complete linkage disequilibrium between the two SNPs. After the digestion with NarI, target PCR fragment produced two fragments (84 and 37 bp) for wild-type genotype −27AA/−11GG, four fragments (101, 84, 37, and 20 bp) for heterozygous genotype −27AG/−11GA, and two fragments (101 and 20 bp) for homozygous mutant genotype −27GG/−11AA. The 20 bp fragment was not visible due to its small size. We screened another 199 controls and 96 CML patients by PCR-RFLP and the fragment yields were consistent with our prediction of complete linkage disequilibrium between −11 (G>A) and −27 (A>G; Fig. 1B). Thus, the statistical results of the two SNPs were identical. Genotype distributions in the CML group and the control group were in Hardy-Weinberg equilibrium (control group: χ² = 1.331, P > 0.05; CML group: χ² = 0.797, P > 0.05).

Association between −27/−11 single-nucleotide polymorphisms and the susceptibility of chronic myelogenous leukemia. Because of the low frequencies of homozygous mutated genotypes (−27GG/−11AA), the −27GG/−11AA genotype was combined with heterozygous genotype (−27AG/−11GA) as G/A-positive genotype (−27G⁺/−11A⁺). There was a statistically significant difference in the distributions of the genotype and allele in the different groups. The frequency of −27G⁺/−11A⁺ genotype was significantly higher in the CML group (n = 33, 26.36%) than in the control group (n =23, 11.85%, χ² = 11.75, P < 0.01). The frequency of −27G/−11A allele was significantly higher in the CML group (n = 35, 13.57%) than in the control group (n = 27, 6.40%, χ² = 9.48, P < 0.01). The genotype and allele distributions of the two SNPs between different groups are presented in Table 1.

Detection of the basal promoter activity with different−27/−11 variation. Having shown that the −27G/−11A allele of the PDCD5 gene is associated with CML, we sought to determine whether the −27A>G/−11G>A genetic variation has functional consequences. Because these two SNPs lie within the promoter region, we examined, in transiently transfected HeLa and HEK293 cells, the activity of reporter constructs driven by the human PDCD5 gene promoter containing either nucleotide −27G/−11A or −27A/−11G. We generated two different pGL3-PDCD5 constructs for −27G/−11A and −27A/−11G alleles (pGL3-PDCD5-11A and pGL3-PDCD5-11G, respectively), and did three independent transfections with each construct. To correct for differences in transfection efficiency, cells were cotransfected with a plasmid encoding for the GFP (see Materials and Methods). After normalizing luciferase activity by the internal GFP control, the average transcriptional activity of pGL3-PDCD5-11A was 80% less than that of pGL3-PDCD5-11G in HeLa cells and HEK293 cells (Fig. 3).

Up-regulation of PDCD5 gene promoter by serum withdrawal and tumor necrosis factor-α treatment. To assess the effects of −27A>G/−11G>A polymorphisms in the promoter region on the regulation of PDCD5 gene expression during cell apoptosis, we transfected pGL3-PDCD5-11A, pGL3-PDCD5-11G or a promoter control plasmid pGL3-SV40 (containing SV40 promoter) into HeLa cells. Apoptosis was induced by serum withdrawal or TNF-α treatment after 24 hours of transfection. The detections of apoptosis were carried out at 0, 12, 24, and 36 hours after serum withdrawal and 24 hours following treatment with different doses of TNF-α. Results of phosphatidylserine externalization with Annexin-V-FITC are shown in Fig. 4A.

An assay of luciferase activity was carried out 0, 12, 24, and 36 hours after serum withdrawal. The results of the luciferase assay are shown in Fig. 4B. The luciferase activity of the reporter gene construct pGL3-PDCD5-11G was increased constantly over the time course 0, 12, 24, and 36 hours after serum withdrawal, whereas no significant difference was observed in the report gene construct pGL3-PDCD5-11A. The observations of TNF-α-induced apoptosis support the results above (Fig. 4C). The wild-type construct had greatly increased promoter activity with the elevated concentrations of TNF-α (0, 1, 10, and 50 ng/mL) and the mutated one was nearly unresponsive to the induction. Furthermore, the luciferase activity of the control plasmid pGL3-SV40 indicates no significant effect of various time intervals (Fig. 4B) or of different TNF-α dosage (Fig. 4C) after apoptosis induction. Together with the above observations, it is clear that increases in luciferase activity induced by serum withdrawal and TNF-α treatment are specific for the PDCD5 promoter and that −27G/−11A mutated variants are also insensitive to apoptotic induction.

To verify whether the −27/−11 SNPs could influence the endogenetic expression of PDCD5, we examined the mRNA levels in bone marrow MNC cells of 31 CML patients with different genotypes by real-time quantitative PCR. The results of real-time quantitative PCR are shown in Table 2 and Fig. 5. The mean PDCD5 mRNA expression level of 14 samples with −27AG/−11GA genotypes (0.300) are significantly lower than that of 17 samples with wild-type (0.735, t = 2.448, P = 0.024). It shows that −27G/−11A mutated variant cannot effectively up-regulate gene expression in vivo.

The results of Western blot of PDCD5 showed same trends as mRNA expression level (Fig. 6A). The bands of PDCD5 of −27AG/−11GA patients (lanes 1 and 3) were much weaker than those of wild-type patients (lane 2). Interestingly, there is no obvious difference between normal controls with wild-type (lanes 5 and 6) and −27AG/−11GA genotype (lane 4).

We further investigated the effect of −27AG/−11GA on PDCD5 expression after the induction of apoptosis. Two cell lines with wild-type (HeLa and HEK293) and two with −27AG/−11GA genotype (human monocytic leukemia cells U937 and human Jurkat T cells) were induced to apoptosis by serum withdrawal. We observed prominently increased levels of PDCD5 expression in wild-type cell lines after the induction (lane 2 compared with lane 1 and lane 6 compared with lane 5), whereas slight changes happened in −27AG/−11GA cell lines (lane 4 compared with lane 3 and lane 8 compared with lane 7; Fig. 6B).

We identified new functional SNPs in the PDCD5 promoter region that are relevant to PDCD5 expression levels and that may, therefore, be useful genetic markers for detecting the susceptibility for CML. This is, to our knowledge, the first report describing associations between SNPs of apoptosis-related genes and the susceptibility to CML, which accounts for ∼20% of total leukemia cases in adults (16–18). Whereas the pathogenesis of CML was proposed to be involved in abnormal cell differentiation, proliferation, and death that arise from apoptosis (19–21), attempts to investigate possible defects in apoptotic mechanisms will be very helpful for formation of effective diagnostic methods and therapeutic strategy of CML.

PDCD5 was shown to promote apoptosis in some tumor cells (6) and to have lower expression level in marrow-nucleated cells of untreated CML patients than in the ones of normal controls (9). SNPs occurring in upstream promoter regions of genes can potentially affect the process of transcription (10–13). The luciferase assay showed that the wild-type −27A/−11G variants had much higher basal promoter activity and could be remarkably up-regulated by apoptotic inducement but the mutated −27G/−11A variants were much lower and nearly unresponsive to the stimulation. It suggests that the −27G/−11A mutated variant of the PDCD5 gene 5′ upstream region cannot effectively up-regulate expression of the downstream gene under the stress of apoptosis induction. Relevant experiments in other cell lines (U937 and Raji) were done and the results were similar to those presented in this report (data not shown). The observations of endogenetic PDCD5 expression assays were consistent with reporter assays. We have shown that CML patients with −27AG/−11GA genotype had statistically significantly lower mRNA expression level of PDCD5 than wild-type patients by real-time quantitative PCR. We further examined PDCD5 protein levels in cell lysates by Western blot analysis and the results also support our prediction. Thus, reduced PDCD5 expression in −27G/−11A carriers might facilitate antiapoptotic effects and, therefore, accelerate CML progression and increase the susceptibility to CML. This is consistent to our association studies that found that −27G⁺/−11A⁺ genotypes are more susceptible to CML.

According to the database (http://www.ncbi.nlm.nih.gov/SNP/snp_ref.cgi?rs=4723), there is a cSNP (rs4723) located in the exon 5 of PDCD5. The G-to-A substitution at position +5,712 (transcription start site as position 1) results in an amino acid change E104K in PDCD5. To investigate whether the −27G/−11A polymorphisms link to this SNP, we chose 35 samples to screen the SNP of exon 5. Twenty individuals (including 10 normal controls and 10 CML patients) with wild-type genotype −27AA/−11GG and 15 individuals (including 5 normal controls and 10 CML patients) with heterozygous genotype −27GA/−11AG were sequenced and no variations were detected in exon 5. All individuals carried the wild-type genotype. This pilot study suggested that there was no correlation between −27/−11 SNPs and +5,712 SNP in exon 5.

Variations in the promoter region may potentially alter the affinities of existing protein-DNA interactions or, indeed, recruit new proteins to bind to the DNA, altering the specificity and kinetics of the transcriptional process (12, 13, 22–25). According our results, the mutated −11A allele disrupts a putative AP-2 binding site where wild-type −11G is located (CCSCRGGC, italic illustrates SNP site), whereas the −27 SNP is not in any potential transcriptional factor binding site (15). We infer that the −11 SNP may more directly affect transcriptional activity in both basal- and apoptotic-induced situations. Additional studies may be required to conclusively address whether AP-2 regulatory sites are involved in the different transcriptional activation of −11A versus −11G. AP-2 transcription factors have previously been implicated in tumorigenesis. Moreover, they are known to play regulatory roles in apoptosis, cell cycle control, and differentiation (26, 27). Accordingly, regulation of AP-2 sites by different −11 SNP variants may affect their responses to apoptotic stimulation.

As we now know, apoptosis is a fundamental mechanism by which DNA-damaging anticancer agents cause cytotoxicity (28). Indeed, there is now evidence that altered expression or mutation of genes encoding key apoptotic proteins can provide cancer cells with both an intrinsic survival advantage and inherent resistance to chemotherapeutic drugs. For example, it was verified that a polymorphism in wild-type p53 could influence response in cancer chemotherapy in vitro and in vivo (29, 30). In present study, we examined PDCD5 expression under the induction of apoptosis in four cell lines with different −27/−11 genotypes in vitro by Western blot. The results provide clues that the mutated allele carriers may be more resistant upon the apoptotic-inducing factors. We assume that individuals with different PDCD5 SNPs variants may have altered responses to chemotherapy drugs and treatment outcomes.

In summary, our findings indicate that the −27A>G/−11G>A SNP in the promoter region of PDCD5 gene influences PDCD5 expression and suggest that it may be a useful genetic marker for resolving the issue of whether a causal relationship exists between PDCD5 and human disease. The identification of a characteristic variant will benefit individual therapy and will improve diagnosis of the disease. Moreover, recent studies indicate that PDCD5 is involved in paraptosis, another form of programmed cell death (31) that may occur during carcinogenesis (32), and neurodegeneration (33). These studies provide valuable insight not only for CML but also potentially for other diseases, as PDCD5 expression seems to be altered in patients of psoriasis (34). A similar study in psoriasis patients is under way in our laboratory. We speculate that functional SNPs of the PDCD5 gene may participate in the pathophysiologic course of diseases involving abnormal programmed cell death.

We thank Professor Shan-Shan Chen and her colleagues in the Institute of Hematology at Peking University People's Hospital for their cooperation in sample recruiting, and Dr. Mingxu Xu, Yishan Gao, Yingmei Zhang, and Ting Zhang of Peking University Center for Human Disease Genomics for their excellent technical assistance.