Selective Loss of Codon 72 Proline p53 and Frequent Mutational Inactivation of the Retained Arginine Allele in Colorectal Cancer

Abstract

According to recent reports, some cancer types exhibit nonrandom allele loss at codon 72 in exon 4 of the p53 gene [coding for proline (72Pro) or arginine (72Arg)]. To clarify this phenomenon for colorectal cancer and to find out if this preferential loss might have any functional significance, p53 loss of heterozygosity (LOH) and p53 mutations were investigated in a group of 61 colorectal cancers and 28 liver metastases, and were correlated with clinicopathologic factors. A comparison of a patient's blood codon 72 status with a healthy control group did not reveal an enhanced risk of developing colorectal tumors for one of the two isoforms. p53-LOH and p53 mutations were found in 62.2% and 39.4% of primary tumors, respectively, and in 57.9% and 25% of hepatic metastases, respectively. In 14 heterozygous cases showing exon 4-LOH, only the 72Pro allele was lost and the retained 72Arg was preferentially mutated. In general, p53 mutations were significantly associated with the 72Arg tumor status (P < .001). Distal tumors showed allelic losses of the p53 gene more commonly than proximal tumors (P = .054). The prevalence of 72Arg increased in frequency with higher Dukes stage (P = .056). We suggest that either the preferential loss of 72Pro or the mutation of the 72Arg in colorectal cancer and hepatic metastases is associated with malignant potential and might reflect carcinogenic exposure, particularly in the distal part of the large intestines.

Introduction

The p53 tumor-suppressor gene, located on chromosome 17p13, is one of the most commonly mutated genes in all types of cancer [1,2]. To date, several polymorphisms in the wild-type p53 gene locus have been described [3]. The most important p53 polymorphism is the restriction fragment length polymorphism in codon 72 of exon 4 coding for proline (72Pro: CCC) or arginine (72Arg: CGC). This polymorphism has been reported to be associated with some tumor types [Pro: bladder cancer [4], lung cancer [5,6]; Arg: hepatocellular carcinoma [7], cervical cancer [8], human papillomavirus (HPV)-associated cervical cancer [9].

Functional differences between the two polymorphic forms of the wild-type protein have been described previously [2]. The 72Arg form of wild-type p53 was found to be more sensitive to proteolysis mediated by the E6 protein of the human papillomavirus than the wild-type 72Pro [10]. Moreover, the codon 72 status of tumor-associated p53 mutants decides whether they form heterodimers with p73 or not [11]. Nonrandom allele loss at codon 72 has recently been reported for squamous cell cancers [11–15]. Furthermore, a better clinical response following cisplatin-based chemoradiotherapy has been shown for advanced head and neck cancers expressing 72Pro mutants in comparison to those expressing 72Arg mutants [16]. To the best of our knowledge, data on colorectal cancer are presently not available. Although there are several studies dealing with p53 loss of heterozygosity (LOH) [17–21], colon cancers have not yet been investigated for a preferential loss of one allele. Exon 4 polymorphism thus appears to be a system that is useful in double respect, namely, for studying the frequency of p53-LOH and for addressing the question of polymorphism. In this study, we used a new rapid one-step real-time polymerase chain reaction (PCR) in a LightCycler to analyze the polymorphism [15].

This study was conducted: 1) to test the frequency of p53-LOH and p53 mutations in a group of colorectal carcinomas and liver metastases, and to correlate the genetic findings with clinicopathologic factors; 2) to answer the question of whether one allele is preferentially lost in the case of LOH; 3) to correlate p53 mutations and p53-LOH with 72Arg tumor status; and 4) to test the prevalence of p53 polymorphism at codon 72 compared to a healthy control group.

Materials and Methods

Primary Tumors

The tumor group included 61 patients who had undergone colorectomy for sporadic colon cancer over the last 4 years. The exclusion criteria were familial adenomatous polyposis, simultaneous occurrence of adenomas, and other previous or synchronous adenocarcinomas. The average age of the patients was 63 years (range, 41–89 years). There were 33 males and 28 females. The tumor was localized in the right colon (cecum, ascending) in 29 cases, in the transverse colon in 1 case, and in the left colon (descending, sigmoid, rectum) in 31 cases. Tumors were graded and staged on routinely processed, formalin-fixed, and paraffin-embedded tissue samples. Six tumors were staged Dukes A, 13 as Dukes B, and 42 as Dukes C [22 with Dukes C1 (TN₁M₀) and 20 with Dukes C2 (TN>1M>0)]. Five tumors were well differentiated, 47 tumors were moderately differentiated, and 9 were poorly differentiated adenocarcinomas. Both blood and tumor tissue were obtained immediately after surgical resection (tumor tissue was obtained by the pathologist C.B. from the center of the lesion avoiding necrotic areas), frozen in liquid nitrogen, and stored at -70°C until DNA extraction. Hematoxylin and eosin (H&E)-stained sections obtained from frozen specimens were examined to verify representativity.

Metastases

We investigated another group of 28 liver metastases. In six patients, tissues of the metastases and the corresponding primary tumor were available. Blood and tumor tissue were always collected for DNA analysis. The mean age of the group with metastases was 61 years (range, 41–78 years). Twenty-two tumors were moderately differentiated, and six were poorly differentiated.

Normal Control Group

Blood from 85 Central European Caucasians with homogeneous age and sex distribution (average, 45.6 years; range, 22–88 years; 42 males and 43 females) served as normal control.

DNA Extraction

Tumor and blood DNA were extracted using the proteinase K-phenol-chloroform extraction protocol as previously described [22].

PCR

For PCR of exon 4 of the p53 gene, we used the Light-Cycler-DNA Master Hybridization Probes Kit (Roche Diagnostics, Mannheim, Germany) as described previously by Schneider-Stock et al. [15]. Briefly, the LCRed hybridization probe was designed to detect the C-variant. If G is present, the resulting mismatch with the probe destabilizes the sample probe hybrid, and this effect is further enhanced by inserting two additional inositol bases into the probe, yielding a melting temperature that is lower in mismatched (T_m = 62.5°C) than in normal-matched cases (T_m = 69°C). The different melting curve profiles (Figure 1) allow for exact genotyping without post-PCR handling.

Figure 1.

Visualization of the melting temperature to identify the genotype variant at codon 72 in exon 4 of the p53 gene. If G (72Arg) is present, the resulting sample probe mismatch yields a lower melting temperature (62.5°C) than the normal-matched cases (C: 72Pro, 68.5°C). Two peaks are visible in heterozygous cases (C/G).

p53 Intron 1 Polymorphism

To further enhance the informativity for p53-LOH, a second p53 polymorphic marker was analyzed. The intron 1 region of the p53 gene [Alu repeat with (AAAAT)_n] was amplified using the following touchdown PCR: 95°C for 5 minutes, 20 cycles of 95°C for 10 seconds, 66°C for 10 seconds, and 72°C for 20 seconds, with a stepwise decrease of 2°C in each fifth cycle, and, finally, 30 cycles at 95°C for 45 seconds, 58°C for 1 minute, and 72°C for 1 minute. Primer sequences were sense 5′-ACTCCAGCCTGGGCAATAAGAGCT-3′ and antisense 5′ACAAAACATCCCCTACCAAACAGC-3′. An aliquot of the PCR product was electrophoresed on native 8% polyacrylamide gels, cross-linked with piperazine diacrylamide, and visualized by a silver staining method described by Budowle et al. [23].

An allelic loss was considered in cases when, in comparison with nontumor DNA, the signal of a tumor band disappeared, or signal intensity was reduced by more than 50%. Evaluation was done twice, visually, or by densitometry (VDS; Pharmacia Biotech, Uppsala, Sweden) in ambiguous cases.

p53 Mutation Analysis

Mutations in exons 4 to 8 of the p53 gene were determined using PCR-SSCP sequencing analysis as described by Schneider-Stock et al. [24].

Statistics

Statistical analyses were carried out using the chi-square analysis or Fisher's exact test in cross-tables and one-way ANOVA (for comparison of group means). P values less than .05 were considered statistically significant, and P < .10 was regarded as a statistical trend. All statistical tests were two-sided. Calculations were carried out by using the SPSS 9.0 software.

Results

The distribution of p53 codon 72 genotypes (Arg/Arg, Arg/Pro, and Pro/Pro) is discussed as follows.

Control Group Versus Tumor Group (Blood Analysis)

The distribution of the control group was 44.7% for Arg homozygotes (n = 38), 48.2% for Arg/Pro heterozygotes (n = 41), and 7.1% for Pro homozygotes (n = 6). The cancer group (blood) showed a very similar distribution (47.3% for homozygous Arg, 44.1% for Arg/Pro, and 8.6% for homozygous Pro; df = 2, χ² = 0.366, P = .833; Table 1).

Table 1.

Comparison of the Codon 72 Status in the Blood of Colorectal Cancer, Liver Metastases, and the Healthy Control Group.

Genotype	Control Group [n (%)]	Primary Tumors [n (%)]	Metastases [n (%)]	P
n	85	57*	26*
Arg/Arg	38 (44.7)	26 (45.6)	13 (50)
Arg/Pro	41 (48.2)	26 (45.6)	12 (46.4)	ns
Pro/Pro	6 (7.1)	5 (8.8)	1 (3.6)

In six cases, the primary tumor and the corresponding metastasis were investigated.

^*Blood was collected from 76 patients.

Primary Tumors Versus Metastases (Tissue Analysis)

72Arg tended to predominate in the group of metastases (78.6% for metastases versus 57.4% for primary tumors) and homozygous Pro variants were only rarely found (P = .148; Table 2).

Table 2.

Correlation between Genetic and Clinicopathologic Data of Patients with Colorectal Cancer.

				Status at Codon 72
Factor	Total p53-LOH-Positive [n (%)]	P	n	Arg/Arg [n (%)]	Arg/Pro [n (%)]	Pro/Pro [n (%)]	P
Tumor type
Primary	23/37 (62.2)	0.615	61	35 (57.4)	20 (32.8)	6 (9.8)	0.148
Metastasis	11/19 (57.9)		28	22 (78.6)	5 (17.9)	1 (3.6)
Localization
Right/transverse	7/16 (43.7)	0.054	30	16 (53.3)	11 (36.7)	3 (10)	0.853
Left	16/21 (76.2)		31	18 (58.1)	10 (32.3)	3 (10)
Age (years)	pos. 63.8	0.764		63.1	64.5	63.2	0.901
	neg. 61.9
Sex
Female	14/19 (73.7)	0.184*	28	17 (57.4)	9 (32.1)	2 (7.1)	0.783
Male	9/18 (50)		33	18 (54.5)	11 (33.3)	4 (12.1)
Tumor grade
Well	2/2 (100)	0.320*	5	4 (80)	0 (0)	1 (20)	0.265
Moderate	21/30 (70)		47	27 (57.4)	15 (31.9)	5 (10.6)
Poor	0/5 (0)		9	4 (44.4)	5 (55.6)	0 (0)
Dukes stage
A	3/3 (100)	0.397	6	2 (33.3)	1 (16.7)	3 (50)	0.056*
B	2/4 (50)		13	9 (69.2)	4 (30.8)	0 (0)
C	18/30 (60)		42	24 (57.2)	15 (35.7)	3 (7.1)

^*Fisher's exact test.

Correlation with Other Clinicopathologic Data

Dukes A tumors showed a higher frequency of the homozygous Pro variant, whereas there was preference for the 72Arg variant in Dukes B and C tumors (P = .056; Table 2). There was no significant correlation between codon 72 status and sex, age, localization, and tumor grade.

LOH

Tumor tissue and corresponding blood were collected from a total of 76 patients. Informativity of exon 4 polymorphism were 33 of 76 (43.5%) and 45 of 76 (59.2%) for intron 1 polymorphism. Considering both polymorphisms together, informativity increased to 73.7% (56 of 76 tumors). Twenty cases were excluded because of noninformativity for both polymorphic markers.

Primary Tumors Versus Metastases

Considering the informative cases, 34 of 56 cases (60.7%) showed allelic losses at the p53 gene [23 of 37 primary tumors (62.2%) and 11 of 19 metastases (57.9%)] (Table 2). Thus, for the next analyses, the two groups were pooled and considered together.

Of the 45 informative cases for intron 1 polymorphism, 32 tumors (71.1%) showed allelic losses. Fourteen of 33 informative cases (45.5%) showed allelic losses for exon 4 polymorphism. All 14 cases with exon 4-LOH exclusively showed losses of 72Pro, whereas loss of the 72Arg was never found (Table 3). To validate the new method, all 14 exon 4-LOH cases were confirmed by using the classic restriction digest with BshI (AGS, Heidelberg, Germany). There was a significant correlation between exon 4 and intron 1 polymorphism (P = .041). Five cases (cases 3, 15, 16, 26, and 29) showed partial losses (i.e., both markers were informative, but only one of both showed LOH). Eleven cases exhibited LOH in both markers (Figure 2). Examples of single analyses are given in Figure 2. p53 Mutation Status In total, we found p53 mutations in 31 of 89 cases (34.8%) (Figure 3). There were 24 primary tumors (39.4%) and seven metastases (25%) with p53 mutations.

Table 3.

p53 Status of Tumors in Germline Heterozygous Arg/Pro Patients.

Tumor Number	Exon 4		p53 Mutation
	p53-LOH	Arg/Pro	Exon	Codon	Amino Acid Exchange
1	LOH	Arg	4/42	1-bp insertion	Frameshift
2	LOH	Arg	5/175	CGC to CAC	Arg to His
3	LOH	Arg	5/175	CGC to CAC	Arg to His
4 (M)	LOH	Arg	5/175	CGC to CAC	Arg to His
5 (M)	LOH	Arg	5/146	1-bp insertion	Frameshift
6	LOH	Arg	5/135	TGC to TAC	Cys to Tyr
7 (M)	LOH	Arg	8/306	CGA to TGA	Arg to stop
8	LOH	Arg	8/273	CGT to TGT	Arg to Cys
9	het	Arg/Pro	WT
9A* (M)	LOH	Arg	8/splice donor	G to T	Splicing
10	LOH	Arg	8/273	CGT to TGT	Arg to Cys
11	LOH	Arg	WT
12	LOH	Arg	WT
13	LOH	Arg	WT
14	LOH	Arg	WT
15	het	Arg/Pro	5/175	CGC to CAC	Arg to His
16	het	Arg/Pro	7/splice acceptor	12-bp deletion	Splicing
17	het	Arg/Pro	WT
18	het	Arg/Pro	WT
19	het	Arg/Pro	WT
20	het	Arg/Pro	WT
21	het	Arg/Pro	WT
22	het	Arg/Pro	WT
23	het	Arg/Pro	WT
24	het	Arg/Pro	WT
25	het	Arg/Pro	WT
26	het	Arg/Pro	WT
27	het	Arg/Pro	WT
28	het	Arg/Pro	WT
29	het	Arg/Pro	WT
30 (M)	het	Arg/Pro	WT
31 (M)	het	Arg/Pro	WT
32 (M)	het	Arg/Pro	WT
33 (M)	het	Arg/Pro	WT

^*Primary tumor and metastasis were investigated; M—metastasis; het—heterozygous without LOH; WT—wild type.

Figure 2.

Allelic losses at the p53 locus, codon 72 genotype status, and p53 mutations in the 34 LOH-positive colorectal tumors and liver metastases.

Figure 3.

(A) Two peaks showing the heterozygous status Arg/Pro in the blood (NT) of patient 2; the corresponding tumor (T) shows only the 72Arg peak, confirming an LOH in this case (↓). (B) In accordance with this finding, sequencing analysis of tumor 2 detected the dominance of the mutant adenine allele (CGC to CAC) at codon 175.

p53-LOH and p53 Mutations p53-LOH occurred preferentially in tumors with p53 mutations (P = .029; Table 4).

Table 4.

Correlation between p53-LOH and p53 Mutations in Colorectal Tumors and Metastases.

p53-LOH	p53 Mutations
	-	+	Total	P
-	15	7	22
+	13	21	34	.029
Total	28	28*	56†

^*Three p53 mutations were detected in cases where LOH could not be determined.

^†Correlation was calculated only in cases that were informative for LOH marker.

Arg/Pro Status Versus p53-LOH and p53 Mutations As there was a selective loss of the proline allele in cases of LOH, we assessed the mutational status of the retained arginine allele (Table 3, Figure 2). Examining the 33 germline heterozygous Arg/Pro patients (i.e., 34 tumors), we found that 10 of 14 cases (71.4%) with exon 4-LOH had p53 missense, frameshift, or splicing mutations, whereas such mutations occurred in only 2 of 20 cases (10%) without exon 4-LOH (Table 3). Interestingly, there was a predominance of mutations in codon 175, a common mutation hotspot codon in colorectal and head and neck cancers [25].

Considering all 89 tumors, p53 mutations were significantly correlated with the 72Arg tumor status (P < .001; Table 5). The seven tumors with germline homozygous Pro/Pro never showed p53 mutations.

Table 5.

Correlation between p53 Mutations and Arg/Pro Status in Colorectal Cancer and Metastases.

p53 Mutations	Status at Codon 72
	Arg/Arg [n (%)]	Arg/Pro [n (%)]	Pro/Pro [n (%)]
Primary tumors	22/35 (63)	2/20 (10)	0/6 (0)
Metastases	7/22 (31.6)	0/5 (0)	0/1 (0)
Total	29/57 (50.9)	2/25 (8)	0/7 (0)
		P < .001

Correlation of p53-LOH with Other Clinicopathologic Data There was a tendency for distal tumors to show a higher frequency of p53-LOH than did proximal tumors (43.7% for proximal vs 76.2% for distal tumors, P = .054). Allelic losses were not statistically correlated with sex, age, Dukes stage, and tumor grading (Table 2). In the six cases for which both the primary tumor and the corresponding liver metastases were available, the following observations were made: in one case, in contrast to the primary tumor, metastasis (A) showed an additional allelic loss (Table 3, tumor number 9/9A); in two cases, both primary tumor and metastases showed p53-LOH with p53 mutations (Figure 2, tumor numbers 34/34A and 35/35A); three cases were noninformative for the two polymorphic markers used without p53 mutations.

Discussion

In this study, we investigated p53-LOH frequency, p53 mutations, and p53 codon 72 polymorphism in a group of colorectal carcinomas and colorectal liver metastases. LOH frequency was similarly distributed (i.e., 62.1% in primary tumors and 63.2% in hepatic metastases); exon 4-LOH was associated only with loss of the 72Pro allele. The retained 72Arg was selectively mutated.

Although an association between codon 72 polymorphism and risk of cancer has been reported, the results with regard to colorectal carcinomas remain inconclusive [26,27]. Regarding blood, we did not observe a significant difference between colon cancer cases and ethnically matched controls for the distribution of codon 72 polymorphism. This may be caused by the relatively small number of cases in our study.

Considering tumor tissue DNA, we found a significantly higher frequency of the 72Arg in colorectal tumors. There was an apparent association of the Arg genotype with tumor progression, with metastases showing a bias toward the Arg genotype. There was only one metastasis with Pro/Pro homozygous constitution. This can be easily explained: all allelic losses at codon 72 of p53 exon 4 were directed toward the loss of the proline allele.

There are only a few studies reporting preferential retention of 72Arg in squamous cell carcinomas of the vulva [12], head and neck [15], and esophagus [28]. We suggest that the preferential loss of the 72Pro variant and/or mutations of 72Arg, also seen in our study for colorectal cancer and colorectal liver metastases, is of biologic significance. Thus, we tested all tumors for p53 mutations. Indeed, tumors that had retained the 72Arg allele often carried p53 mutations, with the “gain-of-function” type occurring with remarkably high frequency. In contrast, p53 mutations were found only in a small percentage of LOH-negative tumors in germline Arg/Pro patients and never in the germline homozygous Pro/Pro patients. A favored loss of the 72Pro, together with a selection for mutation in the retained 72Arg, was also reported by Brooks et al. [12] for squamous cell carcinomas of the vulva. Their data and ours seem to be consistent with reports on functionally significant interactions between some p53 mutants and the p53-related protein p73 in head and neck cancers [11].

It was reported that both forms of human p53 can increase the tumorigenicity of NIH 3T3 cells in nude mice and that 72Arg is more oncogenic in this respect than the 72Pro variant of p53 [29]. Thomas et al. [2] showed that the 72Arg form is a better inducer of transcription because of its stronger affinity for the TAFII32 and TAFII70 transcription factors. Therefore, the prevalence of the homozygous 72Arg variant in cervical cancer might be explained by the increased susceptibility of p53 to degradation by E6 protein of HPV16/18 [9]. Considering colorectal cancer and colorectal liver metastases, HPV infection must be excluded as a possible reason for the higher frequency of the homozygous 72Arg variant in the tumors. Otherwise, in a recent study, an enhanced apoptotic potential of the 72Arg variant caused by its preferential mitochondrial localization and its greater ubiquination by the E3 ubiquitin ligase MDM2 was reported [30].

Based on our finding of a significant increase in the 72Arg variant with Dukes stage, we suggest that 72Arg correlates with the malignant potential of the tumor. This seems to be interesting because a similar increase in 72Arg status with more advanced cases has also been reported in urinary tract cancers [31] and lung cancer [14]. Furthermore, it is known that the development of hepatic metastases is largely dependent on certain risk factors, such as diet, and is modulated by different host factors or different colon subsites [32]. Although the number of cases investigated was low, we found in our group of primary tumors and metastases that the original location of metastasized tumors was always on the left (n = 2 sigmoid and n = 4 rectum)—the side of the highest carcinogenic exposure. Exposure to mutagens may first promote p53-LOH with preferential loss of the proline allele; secondly, the remaining Arg allele functions as a mediator responsible for a progressive genetic instability, leading to a more malignant potential, including the capability of invasiveness and metastasization. This is supported by Wynford-Thomas and Blaydes [33], who developed the theory that the genotype of the cell plays a major role in tumor progression and mainly determines the “choice” of further genetic events. Indeed, experiments have demonstrated that the ability of mutant p53 to bind p73, neutralize p73-induced apoptosis, and transform cells in cooperation with EJ-Ras was enhanced when codon 72 encoded Arg [11].

The percentage of p53-LOH in our series of colorectal cancers and liver metastases is in accordance with the data published in the literature [34]. We found a tendency for distal tumors to show allelic losses more frequently than proximal tumors. Our data are in line with Yashiro et al. [35], who found a higher frequency of p53-LOH in distal tumors without reaching significance in their small group of 28 colorectal cancers, so that the authors did not go into this finding. Eguchi et al. [19] confirmed a significantly higher frequency of p53 mutations in left-sided carcinomas than in right-sided carcinomas. We thus confirm the findings that proximal and distal colon carcinomas differ in their genetic mechanisms regulating initiation, morphologic variation, and/or progression [32,34,36,37]. The increased risk for genetic aberrations reflected as p53-LOH can be attributed to the fact that left-sided tumors come into contact with carcinogens more frequently. Here, bowel transit appears to produce a more prolonged aggression of the colon mucosa than in the right colon.

In summary, we found a preferential loss of the 72Pro allele and mutations in the retained 72Arg in colorectal cancer and liver metastases. When determining the risk for a specific cancer type, examinations of tumor tissue alone are problematic because allelic losses remain unmasked and are not considered. The increase in the frequency of p53-LOH, together with a higher frequency of 72Arg homozygotes in left-sided colon cancer, may be due to a high carcinogenic exposure in the distal part of the large intestines. Mutagens may selectively affect the proline allele, and LOH of the proline allele may be specifically advantageous for further tumor progression.