Association between p21 Ser31Arg polymorphism and cancer risk: a meta-analysis.

P21 (CDKN1A), a key cell cycle regulatory protein that governs cell cycle progression from G1 to S phase, can regulate cell proliferation, growth arrest, and apoptosis. The Ser31Arg polymorphism is located in the highly conserved region of p21 and may encode functionally distinct proteins. Although many epidemiological studies have been conducted to evaluate the association between the p21 Ser31Arg polymorphism and cancer risk, the findings remain conflicting. This meta-analysis with 33 077 cases and 45 013 controls from 44 published case-control studies showed that the variant homozygous 31Arg/Arg genotype was associated with an increased risk of numerous types of cancers in a random-effect model (homozygote comparison: OR = 1.17, 95% CI = 0.99 to 1.37, P = 0.0002 for the heterogeneity test; recessive model comparison: OR = 1.16, 95% CI = 1.01 to 1.33, P = 0.0001 for the heterogeneity test). Stratified analysis revealed that increased cancer risk associated with the 31Arg/Arg genotype remained significant in subgroups of colorectal cancer, estrogen-related cancer, Caucasians, population-based studies, studies with matching information or a larger sample size. Heterogeneity analysis showed that tumor type contributed to substantial between-study heterogeneity (recessive model comparison: Χ(2) = 21.83, df = 7, P = 0.003). The results from this large-sample sized meta-analysis suggest that the p21 31Arg/Arg genotype may serve as a potential marker for increased cancer risk.

The cell cycle regulates growth and differentiation, and defects in cell cycle control are a hallmark of cancer development[1]. Phase progression is principally regulated by Cyclins, Cyclin-dependent kinases (CDKs), and CDK inhibitors (CKIs)[2]. Cyclins and CDKs form complexes regulating cell growth by cell cycle control, whereas CKIs inhibit the activities of the complexes and induce cell cycle arrest[3],[4]. P21 protein belongs to the Cip/Kip family of CKIs and inhibits the phosphorylation of retinoblastoma (Rb) protein by interfering with cyclin E–CDK2 or cyclin A–CDK2 complex[5],[6].

P21 (Cdkn1a/Waf1/Cip1) is encoded by the CDKN1A locus on chromosome 6p21.2 and has a p53 transcriptional regulatory motif. Studies have shown that p21 is a critical downstream effector in the p53-specific pathway of growth control, and the expression of p21 is directly regulated by p53 in response to DNA damage, leading to cell cycle arrest at the G1/S checkpoint[7]. As p21 inhibits proliferation[8] and acts as one of the major transcriptional targets of p53[7], it was initially considered as a potential tumor suppressor. However, studies have also reported that p21 could act as an Oncogene because of its antiapoptotic activities[9]–[11]. Alterations in p21 expression have been observed in a wide variety of cancers, including breast, lung, cervical, ovarian, liver, uterine, and head and neck cancers[12]–[17], indicating the importance of p21 in Carcinogenesis.

Although the involvement of p21 in tumor formation is evident, mutations in p21 are very rare[18],[19]. Thus, most reports focus on genetic variants of p21, and genotypes of some functional polymorphisms have shown to be associated with a high risk of different types of cancer[20]–[22]. The most frequently investigated polymorphism of p21 is Ser31Arg (rs1801270C > A), with a base change from C to A resulting in a non-synonymous serine to arginine substitution in the protein[23], causing a loss of the Blpl restriction site and affecting the DNA-binding zinc finger motif[22]. Hence, it is likely that p21 Ser31Arg polymorphism may result in the alteration of p21 e xpression and/or activity, thereby affecting susceptibility to cancer.

Many molecular epidemiological studies have been conducted to evaluate the effect of the p21 Ser31Arg polymorphism on cancer risk[19],[24]–[70]. The results, however, remain conflicting, and the underlying heterogeneity between studies still needs to be explored. To estimate the overall cancer risk associated with the p21 Ser31Arg polymorphism and to quantify potential between-study heterogeneity, we conducted a systematic meta-analysis by including the most recent and relevant studies focusing on the association between the p21 Ser31Arg polymorphism and cancer risk.

Materials and Methods

Identification and eligibility of relevant references

We included all references of the case-control studies written in English and published to date on the association between the p21 Ser31Arg polymorphism and cancer risk. Two electronic databases (MEDLINE and EMBASE) were searched (last search update October 2010, using the search terms “p21” or “CDKN1A”, “cancer” or “carcinoma”, and “polymorphism” or “variant”) to identify eligible references. Additional references were identified by a hand search of original papers or reviews. If studies had overlapping subjects, only the one with the larger or largest sample size was selected. Furthermore, the studies including subjects with family history or cancer-prone predisposition were excluded.

Data extraction

The following information was extracted from each report: author, year of publication, country of origin, ethnicity, demographics, cancer type, and detail genotyping information and source of controls (population-based and hospital-based). For studies including subjects of different racial descents, data were extracted separately for each race (categorized as Caucasian, Asian, and others).

Statistical analyses

Genotype frequency was collected from each study to evaluate the risk of cancers [odds ratios (ORs) and 95% confidence intervals (CI)]. For all studies, we evaluated the effects of variant genotypes including Arg/Ser and Arg/Arg, compared with the wild-type Ser/Ser genotype, respectively. Then we calculated the ORs and 95% CI for both dominant and recessive genetic models of the variant Arg allele. In addition, we conducted stratification analysis by tumor type (if one cancer type was investigated in less than 3 studies, it would be merged into the “other cancers” group), ethnicity, control source, matching status (yes or no), and sample size ( < 500, 500 to 1000, and > 1000). Smoking-related cancers included lung, bladder, head and neck, kidney, and pancreatic cancers; estrogen-related cancers included breast, cervical, and ovarian cancers.

The χ2-based Q statistic test was used to assess between-study heterogeneity, and it was considered significant if P < 0.05[71]. The fixed-effects model and the random-effects model were respectively performed to combine values from each of the studies based on the Mantel-Haenszel method and the DerSimonian and Laird method[72]. When the effects were assumed to be homogenous, the fixed-effects model was then used; otherwise, the random-effects model was more appropriate. The inverted funnel plots and Egger's test were used to investigate publication bias (linear regression analysis)[73]. The deviation of genotype distribution from Hardy-Weinberg equilibrium (HWE) among controls was also examined by a goodness-of-fit χ2 test. All analyses were conducted using Review Manage (v.5.0) and Stata 10.0. P values were two-sided.

Results

Characteristics of studies

A total of 48 publications examined the relationship between p21 Ser31Arg polymorphism and cancer risk. Two studies[24],[25] were excluded because they investigated the same or a subset population of reported articles[65],[30]. Another two were also excluded because they did not present detailed genotyping information[69] or had cancer-prone predisposition[70]. The studies investigating different cancers[38], multiple ethnicity[26], or multi-center collaboration[30] were separated into multiple studies in subgroup analysis. In addition, three studies [48],[61],[68] that only provided the total number of variant genotypes (Arg/Ser and Arg/Arg) were included in the analysis for the dominant model but not for other genetic models. Finally, our meta-analysis consisted of 44 publications including 59 case-control studies: 20 breast cancer studies, 5 lung cancer studies, 6 head and neck cancer studies, 7 cervical cancer studies, 3 colorectal cancer studies, 3 skin cancer studies, 5 gastric and esophageal cancer studies, and 10 studies of other cancers (Table 1). Among the 59 studies, 28 were conducted in Caucasian descents, 28 were conducted in Asian descents, 2 were conducted in other descents, and the remaining one by Keshava et al.[26] was divided into two subgroups (Caucasians and other ethnicity), because it included multiple ethnicities. In addition, 11 studies were population-based and 48 were hospital-based; 26 did not provide matching information, while 33 were matched by age, sex, and/or geographic region. The polymerase chain reaction-restriction fragment length polymorphism assay (PCR-RFLP) was the most frequently used method for genotyping. Some other methods were also applied, such as direct sequencing, Taqman, and SNaPshot (Table 1). Overall, most studies indicated that the distribution of genotypes in controls was consistent with HWE with the exception of 6 studies[19],[26],[40],[43],[47],[59].

Table 1.

Characteristics of the 44 references included in the meta-analysis

Reference	Year	Country	Ethnicity	Cancer type	Sample size (case/control)	Matching (yes/no)	Genotyping method	Source of control
Keshava et al.[26]	2002	USA	Multiple	Breast cancer	160/327	Yes	PCR-RFLP	Hospital
Ma et al.[27]	2006	China	Asian	Breast cancer	368/467	Yes	PCR-RFLP	Hospital
Tarasov et al.[28]	2006	Russia	Caucasian	Breast cancer	151/191	No	PCR-RFLP and dCAPs	Hospital
Staalesen et al.[29]	2006	Norway	Caucasian	Breast cancer	547/1006	No	Sequencing	Hospital
Cox et al.[30]	2007	Multiple	Caucasian/Asian	Breast cancer	18 290/22670	Both	Multiple methods	Both
MARIE-GENICA[31]	2010	Germany	Caucasian	Breast cancer	3140/5472	Yes	MALDI-TOF MS and PCR-based fragment analyses	Population
Sjalander et al.[32]	1996	Sweden	Caucasian	Lung cancer	144/761	No	PCR-RFLP	Hospital
Shih et al.[33]	2000	China	Asian	Lung cancer	155/189	Yes	PCR-RFLP	Hospital
Su et al.[34]	2003	USA	Caucasian	Lung cancer	1069/1220	No	PCR-RFLP	Hospital
Popanda et al.[35]	2007	Germany	Caucasian	Lung cancer	402/403	No	Fluorescence-based melting-curve	Hospital
Choi et al.[36]	2008	Korea	Asian	Lung cancer	549/533	Yes	PCR and sequencing	Hospital
Sun et al.[19]	1995	China	Asian	Nasopharyngeal cancer	76/66	No	PCR-SSCP direct sequencing	Hospital
Tsai et al.[37]	2002	China	Asian	Nasopharyngeal cancer	47/119	No	PCR-RFLP	Hospital
Rodrigues et al.[38]	2003	Brazil	Caucasian	Head and neck cancer; skin cancer	73/104;46/104	No	PCR-SSCP	Hospital
Li et al.[39]	2005	USA	Caucasian	Head and neck cancer	712/1222	Yes	PCR-RFLP	Hospital
Bau et al.[40]	2007	China	Asian	Oral cancer	137/105	Yes	PCR-RFLP	Hospital
Gomes et al.[41]	2008	Brazil	Mixed	Oral cancer	80/80	Yes	PCR-RFLP	Hospital
Roh et al.[42]	2001	Korea	Asian	Cervical cancer	111/98	No	PCR-RFLP	Hospital
Harima et al.[43]	2001	Japan	Asian	Cervical cancer	66/108	No	Sequencing	Hospital
Lee et al.[44]	2004	Korea	Asian	Cervical cancer	185/345	No	SNaPshot assay	Hospital
Lee et al.[45]	2004	Korea	Asian	Cervical cancer	81/86	No	PCR-RFLP	Hospital
Bhattacharya et al.[46]	2005	India	Asian	Cervical cancer	148/191	No	PCR-RFLP	Hospital
Tian et al.[47]	2009	China	Asian	Cervical cancer	317/353	Yes	MAMA-PCR	Hospital
Roh et al.[48]	2010	Korea	Asian	Cervical adenocarcinoma	53/286	No	PCR-RFLP	Hospital
Wu et al.[49]	2003	China	Asian	Esophageal cancer	128/178	Yes	PCR-RFLP	Hospital
Wu et al.[50]	2004	China	Asian	Gastric cancer	89/192	Yes	PCR-RFLP	Hospital
Lai et al.[51]	2005	China	Asian	Gastric cancer	123/119	No	PCR-RFLP	Hospital
Taghavi et al.[52]	2010	Iran	Asian	Esophageal cancer	126/100	Yes	PCR-RFLP	Hospital
Yang et al.[53]	2010	China	Asian	Esophageal cancer	80/200	Yes	Sequencing	Hospital
Polakova et al.[54]	2009	Germany	Caucasian	Colorectal cancer	612/611	Yes	Taqman	Hospital
Liu et al.[55]	2010	China	Asian	Colorectal cancer	373/838	No	PCR-RFLP	Population
Cacina et al.[56]	2010	Turkey	Caucasian	Colorectal cancer	53/64	Yes	PCR-RFLP	Hospital
Konishi et al.[57]	2000	Japan	Asian	Skin cancer	113/165	No	PCR-RFLP	Hospital
Li et al.[58]	2008	USA	Caucasian	Cutaneous melanoma	805/838	Yes	PCR-RFLP	Hospital
Hachiya et al.[59]	1999	Japan	Asian	Endometrial cancer	54/55	Yes	Dot Blot Hybridization	Hospital
Chen et al.[60]	2002	China	Asian	Bladder cancer	53/119	No	PCR-RFLP	Hospital
Roh et al.[61]	2004	Korea	Asian	Endometrial cancer	95/285	No	PCR-RFLP	Hospital
Hishida et al.[62]	2004	Japan	Asian	Non-Hodgkin's lymphoma	103/440	No	Duplex PCR-CTPP	Hospital
Huang et al.[63]	2004	China	Asian	Prostate cancer	200/247	Yes	PCR-RFLP	Hospital
Hirata et al.[64]	2007	Japan	Asian	Renal cell carcinoma	200/200	Yes	PCR-RFLP	Hospital
Gayther et al.[65]	2007	Multiple	Caucasian	Ovarian cancer	1491/2463	Yes	Taqman	Population
Rajaraman et al.[66]	2007	USA	Mixed	Brain tumor	594/529	Yes	Taqman	Hospital
Chung et al.[67]	2008	China	Asian	Urothelial carcinoma	169/402	Yes	PCR-RFLP	Hospital
Chen et al.[68]	2010	USA	Caucasian	Pancreatic cancer	509/462	Yes	Pyrosequencing and PCR-RFLP	Hospital

PCR, polymerase chain reaction; RFLP, restriction fragment length polymorphisms; dCAPs, derived cleaved amplified polymorphic sequences; MALDI-TOF MS, matrix assisted laser desorption ionisation time-of-flight mass spectrometry; SSCP, single strand conformation polymorphism; MAMA, mismatch amplification mutation assay; CTPP, confronting two-pair primers.

Quantitative synthesis

The p21 31Arg allele frequency varied in different ethnicities, ranging from 0.04 in a Caucasian population[30] to 0.54 in an Asian population[51]. When the eligible studies were pooled into the meta-analysis, the variant genotypes of p21 Ser31Arg were significantly associated with an increased cancer risk. Specifically, compared to the wild-type homozygotes (31Ser/Ser), the variant homozygotes (31Arg/Arg) had a borderline increased risk of all types of cancers (OR = 1.17, 95% CI = 0.99 to 1.37, P = 0.0002 for heterogeneity test), and the association was significant in the recessive genetic model [Arg/Arg vs. (Ser/Ser + Arg/Ser): OR = 1.16, 95% CI = 1.01 to 1.33, P = 0.0001 for heterogeneity test] (Figures 1 and 2). However, such associations were not found for heterozygous comparison or for dominant model comparison (heterozygote comparison: OR = 1.01, 95% CI = 0.94 to 1.08, P < 0.0001 for the heterogeneity test; dominant model comparison: OR = 0.98, 95% CI = 0.89 to 1.08, P < 0.0001 for the heterogeneity test).

Figure 1.

Forest plot (random-effects model) of overall cancer risk associated with the p21 codon 31 polymorphism: Arg/Arg vs. Ser/Ser. Compared to Ser/Ser, Arg/Arg had a borderline association with increased risk of all types of cancer.

Figure 2.

Forest plot (random-effects model) of overall cancer risk associated with the p21 codon 31 polymorphism: Arg/Arg vs. (Arg/Ser+ Ser/Ser). Compared to Arg/Ser + Ser/Ser, Arg/Arg had an association with increased risk of all types of cancer.

In stratified analysis by tumor type, recessive model comparison with the heterogeneity test showed that individuals with variant homozygous genotypes (31Arg/Arg) had a higher risk for colorectal cancer (OR = 1.39, 95% CI = 1.03 to 1.08, P = 0.25) and estrogen-related cancer (OR = 1.27, 95% CI = 1.01 to 1.60, P = 0.002), but not for other cancers (Table 2). Furthermore, recessive model comparison for the heterogeneity test showed that the risk effect of variant homozygotes (31Arg/Arg) remained significant in studies with Caucasian subjects (OR = 1.41, 95% CI = 1.14 to 1.73, P = 0.34), population-based controls (OR = 1.36, 95% CI = 1.11 to 1.67, P = 0.06), matching design (OR = 1.21, 95% CI = 1.01 to 1.45, P = 0.002), and sample size more than 1000 (OR =1.18, 95% CI = 1.01 to 1.37, P = 0.08).

Table 2.

Summary ORs for association between the p21 Ser31Arg polymorphism and cancer risk

Subgroup	Comparisons	Cases/Controls	Arg/Arg vs. (Arg/Ser + Ser/Ser) OR (95% CI)c	Pd
Totala	56	32 420/43 960	1.16 (1.01–1.33)	0.0001
Tumor type
Breast cancer	20	22 656/30 133	1.25 (0.95–1.63)	0.03
Lung cancer	5	2319/3106	0.92 (0.73–1.17)	0.88
Head and neck cancer	6	1125/1696	1.16 (0.79–1.72)	0.63
Cervical cancer	6	908/1181	1.40 (0.85–2.28)	0.005
Colorectal cancer	3	1038/1513	1.39 (1.03–1.87)	0.25
Skin cancer	3	964/1107	0.64 (0.36–1.16)	0.15
Gastric / esophageal cancer	5	546/769	0.78 (0.58–1.03)	0.25
Other cancers	8	2864/4455	1.43 (1.18–1.73)	0.08
Smoking–related cancer	12	3574/4936	1.05 (0.88–1.27)	0.10
Estrogen–related cancer	29	25 055/33 777	1.27 (1.01–1.60)	0.002
Ethnicityb
Caucasian	28	27 184/36 960	1.41 (1.14–1.73)	0.34
Asian	26	4495/6251	1.09 (0.92–1.28)	< 0.0001
Others	3	741/749	0.87 (0.53–1.42)	0.11
Control source
Population	11	17623/26454	1.36 (1.11–1.67)	0.06
Hospital	45	14797/17506	1.15 (0.99–1.33)	0.0008
Matching status
Yes	32	23 809/27 929	1.21 (1.01–1.45)	0.002
No	24	8611/16031	1.09 (0.88–1.35)	0.01
Sample size
< 500	25	2653/3827	1.05 (0.85–1.30)	0.002
500–1000	8	2455/3522	1.21 (0.87–1.68)	0.01
> 1000	23	27 312/36 611	1.18 (1.01–1.37)	0.08

OR, odds ratio; CI, confidence interval. aThree references that only provided the total number of Arg/Ser and Arg/Arg were excluded from the analysis for the recessive comparison [Arg/Arg vs. (Arg/Ser+Ser/Ser)]. b0ne study by Keshava et al included multiple ethnicities. cRandom effect model was used when P value for heterogeneity test < 0.05; otherwise, fix effect model was used. dTest for heterogeneity.

Heterogeneity and sensitivity analyses

In the recessive model comparison, heterogeneity among all studies on the p21 Ser31Arg polymorphism and cancer risk was observed (χ2 = 98.56, P = 0.0001). We evaluated the source of heterogeneity by tumor type, ethnicity, control source, matching status, and sample size, and found that tumor type contributed to substantial heterogeneity (χ2 = 21.83, P = 0.003), but not ethnicity, control source, matching status, and sample size. The leave-one-out sensitivity analysis indicated that no single study changed the pooled ORs qualitatively. Furthermore, the exclusion of 6 studies [19],[28],[40],[43],[47],[59], whose genotype distributions deviated from HWE, did not affect the results of the meta-analysis (OR = 1.16, 95% CI = 1.01 to 1.34, P = 0.0004).

Publication bias

Funnel plot and Egger's test were conducted to access the publication bias of all studies. The shapes of the funnel plots seemed symmetrical (Figure 3), suggesting that there was no obvious publication bias. Egger's test was used to provide further statistical evidence; similarly, we did not find significant publication bias in this meta-analysis (t = 0.95, P = 0.345).

Figure 3.

Funnel plot analysis to detect publication bias. Each point represents a separate study for the Indicated association.

Discussion

On the basis of 44 independent publications, our meta-analysis provided statistical evidence that variant homozygous Arg/Arg genotype of p21 was significantly associated with an increased risk of cancers, particularly of colorectal cancer and estrogen-related cancer. The stratification analysis also showed that the risk effect of Arg/Arg was more prominent in studies with Caucasian subjects, population-based controls, matching design, and larger sample sizes.

P21 is a Cyclin-dependent kinase inhibitor, causing cell cycle arrest by inhibiting the G1 to S phase checkpoint, and it is up-regulated by the tumor suppressor protein P53[7]. In addition, p21 is frequently down-regulated in human cancer, and the loss of its expression or function has been implicated in Carcinogenesis or the prognosis of multiple cancers[12]–[17],[74]. However, studies also suggest that p21 can promote the development of cancer, indicating a double-edged effect showing tumor-suppressing or tumor-promoting properties[9]–[11]. The most common p21 polymorphism is at codon 31 (C > A) within a highly conserved region of the gene, which causes an amino acid change from Ser to Arg and may encode functionally distinct proteins [23]. Although some functional studies suggest that p21-Ser and p21-Arg variant alleles present similar kinase inhibitory activity and growth suppression ability[19], they have been shown to differ significantly in their transcriptional efficiency. For example, individuals carrying the p21-Arg-encoding allele manifest a lower p21 expression[75]. Our meta-analysis supports that individuals carrying the Arg/Arg genotype have a higher cancer risk as assessed in a recessive model.

Because of the paradoxical role of p21 contributing to both cancer suppressive and promoting effects, it is biologically plausible that multiple tumors with different carcinogenic mechanisms may reflect different susceptibilities conferred by the p21 Ser31Arg polymorphism. In our meta-analysis, we found that the effect of the p21 31Arg/Arg genotype was unfavorable toward the development of breast, head and neck, cervical, and colorectal cancer, but appeared to be favorable toward the development of lung, skin, gastric, and esophageal cancer. Heterogeneity analysis also showed that tumor type contributed to substantial between-study heterogeneity. Thus, inconsistent results among different cancers may involve the mechanisms by which p21 regulates cell proliferation or apoptosis in different cancer cells. However, this difference could also be due to limited statistical power as a result of a small sample size in subgroup analysis. Recently, a meta-analysis investigated the association between the p21 Ser31Arg polymorphism and breast cancer risk among 22 109 cases and 29 127 controls, but no significant associations were found[76], which is consistent with our results in breast cancer-including additional studies (22 656 cases and 30 133 controls). Moreover, studies have reported that estrogen stimulates cell mitotic activity and Carcinogenesis in breast, endometrial, and ovarian cancer[77],[78]. In subgroup analysis, we found that the p21 Ser31Arg polymorphism was significantly associated with risk toward the development of estrogen-related cancer, possibly due to different carcinogenic mechanisms of different cancer types including gene-environment interactions.

Ethnicity may affect tumor susceptibility by different genetic factors and environmental exposures through gene-gene and gene-environment interactions. In our meta-analysis, we observed that the association between the p21 31Arg/Arg genotype and overall cancer risk was significant in Caucasians but not in Asians. Furthermore, our results indicated that the association of significantly increased cancer risk with the p21 31Arg/Arg genotype was more pronounced in studies with population-based controls, matching design, or larger sample sizes. The possible explanation may be that population-based controls were more representative of the general population and that studies with matching design or larger sample sizes may eliminate some bias and thus have a greater reliability or statistical power to detect the moderate effect of this single nucleotide polymorphism, suggesting that some characteristics should be carefully considered in genetic association studies, such as the selection of controls, matching status, ethnicity information, and sample size.

Several potential limitations of the present meta-analysis warrant consideration. First, although the funnel plot and Egger's test showed no publication bias, selection bias might have occurred because only studies published in English were included in our meta-analysis. Second, in the stratification analyses, the numbers of individuals carrying the Arg/Arg genotype in some subgroups were relatively small because of its low allele frequency in Caucasian subjects, which might have a small statistical power to detect the real association. Third, our results were based on unadjusted estimates, because ORs in all studies were not adjusted by the same potential confounders, such as age, sex, and exposure. Thus, a more precise analysis should be conducted, if individual data were available, which would allow for the adjustment by some co-variants and further evaluation of potential gene-environment interactions. In summary, this meta-analysis provides statistical evidence that the p21 Ser31Arg polymorphism may contribute to individual susceptibility to cancer. Future well-designed large studies were warranted to validate our findings in different ethnic populations.