p53 status and response to radiotherapy in rectal cancer: a prospective multilevel analysis

The aim of this study was to evaluate, in a prospective study, the predictive role of p53 status analysed at four different levels in identifying the response to preoperative radiotherapy in rectal adenocarcinoma. Before treatment, 70 patients were staged and endoscopic forceps biopsies from the tumour area were taken. p53 status was assessed by total cDNA sequencing, allelic loss analysis, immunohistochemistry, and p53 antibodies. Neoadjuvant treatment was based on preoperative radiotherapy or radiochemotherapy. Response to therapy was evaluated after surgery by both pathologic downstaging and histologic tumour regression grade. In all, 35 patients (50.0%) had p53 gene mutations; 44.4% of patients had an allelic loss; nuclear p53 overexpression was observed in 39 patients (55.7%); and p53 antibodies were detected in 11 patients (16.7%). In the multilevel analysis of p53 status, gene mutations correlated with both nuclear protein overexpression (P<0.0001) and loss of heterozygosity (P=0.013). In all, 29 patients (41.4%) were downstaged by pathologic analysis, and 19 patients (29.2%) were classified as tumour regression grade 1. Whatever the method of evaluation of treatment response, no correlation between p53 alterations and response to radiotherapy was observed. Our results do not support the use of p53 alterations alone as a predictive marker for response to radiotherapy in rectal carcinoma.

Preoperative radiotherapy for locally advanced rectal cancer is used to reduce local recurrence rates and to improve survival rates (Swedish Rectal Cancer Trial, 1997). The selection of patients for neoadjuvant therapy is currently based on clinical and pathological parameters, rectal endosonography, and CT-scan findings (Becouam et al, 1999). Nevertheless, these parameters do not predict the response to neoadjuvant therapy. Predictor markers of complete responses are thus needed for patients treated with preoperative radiotherapy.
The p53 tumour suppressor gene has been demonstrated to regulate cell cycle progression and apoptosis (Levine, 1997;Polyak et al, 1997). Particularly, p53 plays a major role in the cellular response to DNA damage and is an essential component of the pathway leading from DNA damage produced by ionising radiation to apoptosis (Canman et al, 1998;El Deiry, 2003). Several in vitro (Lowe et al, 1994;Bunz et al, 1999) and in vivo (Lee and Bernstein, 1993;Kemp et al, 2001) studies have demonstrated that p53 dysfunction might restrict therapeutic efficacy. However, depending on tumour site, type of therapy, and methods of p53 detection, controversial results have been obtained (Cote et al, 1997;Pai et al, 1998;Chiarugi et al, 1998;Pruschy et al, 2001). p53 gene mutations have been found in a large number of human cancers (Hainaut et al, 1997) with a frequency of about 60% in rectal cancer (Hollstein et al, 1991). The status of p53 is frequently studied by immunohistochemistry (IHC). This method is rapid and applicable to large-scale samples, but the clinical value of this type of analysis is not always correlated with the data obtained by cDNA sequence analysis (Sjogren et al, 1996;Bazan et al, 2002). In addition, p53 autoantibodies (p53-Ab) have been found in the serum of patients with a variety of human neoplasms (Crawford et al, 1982;Soussi, 2000).
In this study, we attempted to assess whether a complete p53 status analysis, involving molecular, immunohistochemical, and serological studies, correlates with response to preoperative radiotherapy in a prospective study of 70 rectal carcinoma patients.

MATERIALS AND METHODS Patients
In all, 70 patients with rectal adenocarcinoma were included in this study between 1996 and 2001. All of them were treated with preoperative radiotherapy or radiochemotherapy at the Val d'Aurelle Paul-Lamarque Cancer Institute in Montpellier. In this prospective study, the patients were staged using the 1997 TNM classification based on a clinical examination, endoscopic and endorectal ultrasonography evaluation, and computed tomography (CT) of the thorax and abdomen. There were 43 men and 27 women in this prospective study. The median age was 65 years (range 39 -80 years). Clinicopathological characteristics of the 70 patients who underwent radiotherapy or radiochemotherapy are presented in Table 1. All patients had a life expectancy greater than 3 months and a WHO performance status of 0, 1, or 2. Informed consent was obtained from all patients before entering the study.

Therapy
For radiotherapy, patients were treated in the supine position with a three-field isocentric technique using 18 MV photon beams daily, five times a week. The daily dose at the isocenter (in accordance with ICRU 62) was 1.8 Gy; the total dose to the entire pelvis was 45 Gy. In 29 patients, the primary tumour received a boost dose of up to 15 Gy, because a clinical response was observed during the conventional course of treatment. The pelvic target volume encompassed the posterior pelvis, the sacrum, the posterior half of the bladder, the prostate/vagina, and the presacral and low common iliac nodes up to the intervertebral space L4/L5. The boost volume covered the primary tumour plus a 2-cm margin using a three-field technique. Individually shaped blocks were used to shield normal tissues. Four patients received concurrent radiochemotherapy. The chemotherapy regimen consisted of continuous infusion of 5-fluorouracil (5-FU) and intravenous (i.v.) leucovorin beginning on the first day of radiation therapy. One patient received two cycles of 400 mg m À2 day À1 of 5-FU with 10 mg m À2 of leucovorin (days 1 -5 and 29 -33 of radiation therapy). Three patients received three 48-h courses of 5-FU (400 mg m À2 day À1 ) with 100 mg m À2 day À1 of leucovorin (days 1 -2, 15 -16, and 29 -30 of radiation therapy). The median time between the first day of radiotherapy and surgery was 10 weeks (73.9 weeks).

Biopsy collection, processing, and histopathology
Endoscopic biopsies and rectal endosonography were performed by the same physician. All patients were given 250 ml of PEG enema before endorectal examination. Endorectal ultrasonography was performed with an endosonic linear probe (Model EUB-33, Hitachi). Patients were staged according to uTN criteria for ultrasound staging. Four pretherapeutic endoscopic biopsies from the macroscopic tumour area were performed in all cases. The endoscopic biopsies were frozen immediately after resection in liquid nitrogen (N 2 ) and then embedded in Tissue-Tek s OCT compound (Sakura Finetech USA Inc., Torrance, CA, USA). Three consecutive 5-mm-thick sections were cut. One slide was stained with haematoxylin and eosin to estimate the percentage of tumour cells, and the other slides were used for the IHC analyses. The remaining biological material was stored in liquid N 2 until nucleic acid extraction. Two blood samples were collected from each patient on the day of clinical diagnosis for allelic loss and serological analyses.

Nucleic acid extraction
DNA was extracted from the blood sample with QIAamp DNA Blood Maxi Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. For biopsy samples, the OCT compound surrounding the tissue was discarded avoiding thawing in order to prevent RNA degradation. The biopsy was then placed in a polypropylene tube containing 800 ml of extraction solution (TRIZOL s Reagent, Invitrogen, France) and homogenised using an Ultra-Turax apparutus. The RNA and DNA were then coextracted according to the manufacturer's recommendations. The RNA pellet was dissolved in 50 ml of RNAse-free H 2 O, and 25 ml were used for the cDNA synthesis. The DNA pellet was dissolved in 50 ml of Tris-HCl buffer, pH 7.4. cDNA synthesis, PCR amplification, and direct sequencing The tumour cDNA was obtained with a First-Strand cDNA Synthesis Kit (Amersham Pharmacia Biotech, Uppsala, Sweden) as specified by the manufacturer.
For all patients, the entire coding sequence (exons 2 -11) was examined for p53 mutations by direct sequencing. Four sets of primers were designed to cover the entire p53 coding sequence. The four overlapping PCR fragments generated were then solid phase sequenced using an ALF expresst Automated DNA sequencer (Pharmacia Biotech, Sweden). PCR primers and sequencing oligonucleotides were synthesised based on the cDNA sequence of p53 as described previously (Thirion et al, 2002). Briefly, cDNA was subjected to PCR in a 50-ml reaction mixture that contained 10 mmol l À1 Tris-HCl, pH 8.3, 50 mmol l À1 KCl, p53 status and response to radiotherapy in rectal cancer E Lopez-Crapez et al 0.1% Triton X-100, 1.5 mmol l À1 MgCl 2 , 1 U Taq DNA polymerase (Invitrogen, France), 6 pmol of each primer, and 200 mmol l À1 of each dNTP. PCR were carried out in a Touch Down Thermocycler (Hybaid, UK) for 38 cycles: denaturation (15 s at 941C), annealing (30 s at 581C), and elongation (45 s at 721C). A final 5-min elongation was performed at 721C. The Thermo Sequenase Fluorescent-Labelled Primer Cycle Sequencing Kit with 7-deaza-dGTP (Amersham Pharmacia Biotech) was used for direct sequencing as described previously (Thirion et al, 2002). Resolved sequencing products were analysed with the ALFwin Sequencer Analyser 2.00 software (Amersham Pharmacia Biotech); then, the SB mutation analyser software (Amersham Pharmacia Biotech) was used to compare the data obtained with the wild-type p53 sequence. Each mutation identified was confirmed by sequencing an entirely new PCR product, starting from the corresponding cDNA.

Detection of loss of heterozygosity (LOH)
DNA was examined for LOH with four p53 intragenic markers: two restriction sites, namely, MspI in inton 6 (McDaniel et al, 1991) and BstUI in exon 4, (Ara et al, 1990), and two variable numbers of tandem repeats (Jones and Nakamura, 1992;Hahn et al, 1993) as described previously (Thirion et al, 2002). A patient was classified as informative when normal DNA demonstrated heterozygosity for one of the four markers. LOH was evaluated by comparing normal and tumour DNA band intensities by two independent readers. LOH was considered effective when the intensity of one allele in the tumour DNA represented less than 50% of the intensity of the other allele.

IHC analysis
The overexpression of the p53 protein was evaluated on two consecutives 5-mm frozen sections that were fixed in 50% methanol/acetone for 10 min at À201C and then air-dried. Two monoclonal antibodies: Pab 1801 (1 : 25 dilution; Oncogene Research Products, Cambridge, MA, USA) and DO7 (1 : 25 dilution; Dako, Glostrup, Denmark) were used. The alkaline phosphatase and monoclonal anti-alkaline phosphatase staining procedure was used as described previously (Thirion et al, 2002). For each series of IHC analysis, a negative and a positive p53 control slide was included. Nuclear staining of the tumour tissues was scored as follows: o5, 45 -10, 410 -25, and 425%. A specimen was scored positive when more than 5% of the tumour cells showed nuclear staining with at least one of the anti-p53 antibodies (Pab 1801 or DO7). The same pathologist, blinded to the results of the other p53 analyses, reviewed all slides.

Serological analysis
Serum p53-Ab were detected by a semiquantitative enzyme-linked immunosorbent assay commercially available (PharmaCell, France).

Assessment of radiotherapy effects
Rectal endosonography is considered as the standard medical examination for parietal infiltration evaluation before treatment. Reliability determinations of this method, carried out by experts, is about 80 and 75% for uT and uN, respectively (Heriot et al, 1999). In our study, all staging were assessed by the same expert. Response to radiotherapy was based on both comparison of the uT stage vs the pT stage (pathologic downstaging) and histologic tumour regression grade of the surgical samples (Rectal Cancer Regression Grade (RCRG)). For the pathologic staging, patients were considered as responders to neoadjuvant therapy when a downstaging of one T stage was obtained. For the RCRG, the staging proposed by Wheeler et al (2002Wheeler et al ( , 2004 was used. In the RCRG staging, RCRG 1 ¼ 'good' responsiveness, with a sterilised tumour or the presence of remaining microscopic foci of adenocarcinoma; RCRG 2 ¼ marked fibrosis but with macroscopic tumour still present; and RCRG 3 ¼ 'poor' response, little or no fibrosis in the presence of abundant macroscopic tumour. In our study, RCRG 1 patients were considered as histological responders, while histological non-responders corresponded to patients with RCRG 2 or RCRG 3. The same pathologist, blinded to the result of the p53 analysis, classified all tumours.

Statistical analyses
To investigate the association between parameters, univariate statistical analyses were performed using Pearson's w 2 test with exact P computation for categorical variables or Fisher's exact test if applicable.
Multivariate analyses for response, by logistic regression, were carried out to evaluate the effect of interactions between the different variables.
Owing to the small number of patients and the fact that some patients did not have measurements for all variables, the power of analysis was reduced.
All P-values reported are two-sided. For all statistical tests, differences were considered as significant at the 5% level. Statistical analyses were performed on an IBM PC-compatible personal computer using the STATA 7.0 software.

Biological data on preradiotherapy samples
Pathological examinations of slides from pretreatment biopsies In order to obtain a sufficient amount of tumour cells, a preliminary study was performed with three to six endoscopic biopsies from each rectal carcinoma patient. The quality of each endoscopic biopsy was evaluated by histopathological examination of a haematoxylin-and eosin-stained slide. For each patient, dysplasia lesions as well as specimens with o5% of tumour cells were excluded. We decided that four biopsies were necessary to carry out the complete p53 status analyses (sequencing, LOH, and IHC).
Total cDNA sequencing Among the 70 patients analysed, 35 cases (50%) revealed p53 mutations (Table 2) LOH analysis Among 53 patients studied, the use of four intragenic markers allowed us to identify 45 informative cases (84.9%). LOH was detected in 20 patients (44.4%) and each LOH was confirmed by a new PCR.
Multilevel p53 analysis Side-by-side univariate comparison of the four analyses: sequencing, LOH, IHC, and serology revealed a strong correlation between the sequencing results and immunohistochemical data (Po0.001) (Table 3). Nevertheless, seven patients were p53 mutated without p53 overexpression and 11 patients were positive immunohistochemically without any p53 mutation. A significant correlation between p53 LOH and p53 mutations was observed (P ¼ 0.013) (Table 3); mutated patients display majorly an LOH (65%), whereas 72% of nonmutated patients were LOH negative. Association between p53 antibodies positivity in the serum and p53 protein expression in the corresponding rectal carcinoma tissue was assessed. A trend to a correlation between the presence of p53 antibodies and p53 protein overexpression was observed: 72.7% of p53-Ab-positive patients had p53 tumour overexpression vs 56.3% of p53-Abnegative patients. We noticed that one patient with circulating p53-Ab had no gene alteration and no p53 nuclear overexpression. One p53-Ab-positive patient was IHC positive with a wild-type gene, and one patient was IHC negative with a codon 175 mutation.
Response to neoadjuvant therapy and clinicopathological parameters We used a w 2 test to determine whether any of the clinicopathological covariates predicted the response to neoadjuvant therapy. By using pathologic downstaging, only gender and radiotherapy dose correlated with response (Po0.04 and Po0.05, respectively) ( Table 6). In multivariate analysis, sex remained a significant predictor with an odds ratio of 0.33 (95% confidence interval (CI), 0.11 -0.95; P ¼ 0.04). By using the histologic tumour regression grade, no correlation was observed.    Response to neoadjuvant therapy and p53 status Uni-and multivariate analyses did not reveal any correlation between p53 abnormalities and response to therapy, whatever the staging used to evaluate tumour response (Tables 7 and 8).
Nevertheless, studies of the four p53 levels analysed on sterilised surgical samples (pT0N0) and microscopic residual tumours (pT1N0) vs other postsurgical stages demonstrated a trend to correlation (P ¼ 0.059) for LOH detection. Only 12% of patients (one of eight) with a pT0N0 or pT1N0 stage demonstrated an LOH, whereas for advanced stages, there was no difference.

DISCUSSION
The aim of this prospective study was to assess the role of p53 complete analyses in the response to radiotherapy or radiochemotherapy in rectal carcinoma patients. Many other studies have searched for predictive criteria of tumour responsiveness to radiotherapy. In particular, the role of p53 has been investigated, but despite the clear contribution of p53 to the molecular pathogenesis of colorectal tumours, its role in the response to therapy is still unclear. Since the clinical ambiguity of p53 status is largely due to the method used to detect p53 abnormalities, a multilevel analysis of p53 (molecular, protein, and serological) was thus performed to detect all p53 alterations.
Endoscopic biopsies were used as starting material for sequencing, LOH determination, and protein overexpression studies.
Owing to the small size of the biological specimens, an anatomopathological analysis of four different biopsies was performed to prevent contamination by normal cells, which would lead to sequencing or LOH misinterpretations. Biopsies with the highest cellularity and the highest number of tumour cells were selected.
In this study, 50% (35 of 70) of the patients demonstrated a p53 gene mutation. The vast majority of observed mutations (34 of 35) were found to be located in the p53 core domain, affecting the principal functions of p53 as transcription factor. These results are in agreement with other reports where p53 gene mutation rates in colorectal cancers varied from 30 to 63% (Iacopetta, 2003). This wide range of p53 gene alterations depends not only on the sensitivity of the technique used to detected mutations and the number of exons covered but also on the stage of development of the cancer lesions and the localisation of the tumour. Frequently, colorectal cancers are considered as a unique entity, but recent evidence suggests that different genetic pathways are involved in colorectal cancer (Frattini et al, 2004). Particularly, p53 mutation rates vary between the two forms of genomic instability associated with colorectal cancers: microsatellite instability and chromosomal instability (Tang et al, 2004).
Analysis of nuclear p53 protein accumulation was based on the use of two monoclonal antibodies, Pab 1801 and D07 (Baas et al, 1994). We detected 55.7% (39 of 70) of IHC-positive cases when the cutoff value for p53 positivity was set at 5%. The frequency of the positive p53 staining observed was consistent with reported data, ranging from 44.8 to 60.8% (Bosari et al 1994;Poller et al, 1997). In this study, association between p53 protein nucleic overexpression and p53 gene mutation was obtained for 28 patients (71.8%). This percentage is around 70% for colorectal cancer studies reported in the literature (Bosari et al, 1994). Our results revealed 25.7% of nonconcordant results with 11 IHC false-positives and seven IHC false-negatives. p53 protein overexpression analysed by IHC is the    (Greenblatt et al, 1994). IHC-false-negative cases could be due to nonsense mutations or genetic alterations unable to stabilise sufficiently the protein. On the other hand, IHC-falsepositive samples may be attributed to (i) normal cell cycle fluctuations (Delmolino et al, 1993) when a low percentage of nuclear staining was observed or (ii) alternative stabilisation of the protein by alterations of p53 regulatory genes implicated directly in the negative feedback loop such as Mdm2 (Yin et al, 2002) or indirectly such as p14ARF (Esteller et al, 2000).
Overexpression of p53 protein in tumours induces an immune response, and among cancer patients, those with colorectal cancer have the highest prevalence of p53-Ab ranging from 13 to 32% (Angelopoulou et al, 1997). In our series, 16.6% of patients (11 of 66) had detectable levels of p53-Ab. The use of antibodies against p53 as serological marker in the clinical management of colorectal cancer patients has been reported, but the prognostic value of such antibodies (Kressner et al, 1998) and their potential use for prediction of curability (Takeda et al, 2001) or response to adjuvant chemotherapy (Lechpammer et al, 2004) are conflicting. Concerning the detection of p53-Ab in the serum of patients, in our study no clear correlation between tumour progression and p53-Ab presence was found. Lechpammer et al (2004) in a series of 220 colorectal cancer patients detected p53-Ab mainly in Dukes' B and C stages. Moreover, Tang et al (2001), in a large study of 998 colorectal patients demonstrated that the presence of p53-Ab correlates with tumour progression in colorectal carcinogenesis and an increase with advanced node metastasis. However, in these studies, both colon and rectal cancers were analysed as a single entity, whereas our work was focalised only on rectal carcinoma.
Allelic loss of 17p is a frequent event associated with colorectal carcinogenesis . In most cancers, one allele carries a missense mutation and the other allele is lost Nigro et al, 1989). Our series demonstrated that LOH at the p53 locus occurred in combination with sequence alterations. Nevertheless, disruption of p53 function has been described without any loss or inactivation of the intact allele (Inga et al, 1997;Gualberto et al, 1998).
As evaluation of response to therapy is still a matter of debate (Wheeler et al, 2002(Wheeler et al, , 2004, we analysed response to therapy by both pathologic downstaging and histologic tumour regression grade. Our data demonstrated a significant correlation between downstaging and the RCRG stage. Two different doses of radiotherapy were used in our study population (45 and 60 Gy). Regarding p53 abnormalities, these two subgroups were similar. When downstaging was used, a significantly better response was observed for patients receiving high-dose irradiation. This correlation was lost when RCRG stage was used. This discordance may be related to patients demonstrating a 'good' responsiveness (RCRG 1), although residual tumour cells remained in the muscularis and mesorectal tissue leading to an absence of downstaging.
In our multilevel detection of p53 abnormalities, no p53 analysis showed a significant influence on response to preoperative irradiation when pretreatment tissues were analysed, whatever the method used to evaluate treatment response. In the literature, the role of p53 gene mutations in sensitivity or resistance to radiation therapy is still a subject of discussion (Chiarugi et al, 1998). Rebischung et al (2002) showed by sequence analysis in a retrospective series of 86 rectal tumours with 41% responders that the presence of p53 mutations correlated with sensitivity to radiotherapy. However, Rodel et al (2002) analysed the histopathological response to radiotherapy in a series of 44 patients and demonstrated that neither the p53 nor the bcl-2 status was correlated with a response to radiotherapy, but they found that the apoptotic index may help to tailor therapy with regard to neoadjuvant treatment of rectal cancer. Similarly, Saw et al (2003), in a series of 60 low rectal tumours locally advanced, concluded that neither p53 by IHC and PCR -SSCP (single-strand conformation polymorphism), nor DCC (deleted in colon cancer) by IHC was associated with tumour downstaging.
Although no correlation was obtained for pretreated tissues in our study, a trend to correlation was observed on surgical samples where retention of heterozygosity was associated with pT0 -pT1 stages.
When cells are exposed to ionising radiation, a complex response is initiated including cell cycle arrest in the G1 and the G2 phases, apoptosis, and DNA repair. Wild-type p53 is a cell cycle checkpoint determinant following irradiation (Kuerbitz et al, 1992); and in response to ionising radiation (Buschmann et al, 2000), p53 is stabilised through phosphorylation, inhibition of Mdm2-mediated degradation, and reduction in Mdm2 sumoylation. The consequence is promotion of either cell cycle arrest or apoptosis. Following gamma-irradiation-induced cell death, striking tissue specificity is observed, with distinct regulation of target p53-induced genes (Fei and El Deiry, 2003). Instead of static analyses, dynamic immunohistochemical studies, comparing expression of apoptosis-releated genes (Tannapfel et al, 1998;Rau et al, 2003) in pretherapy biopsies and the final resected specimen after neoadjuvant treatment, could contribute to molecular marker positioning. Furthermore, other components such as EGFR and cyclin D1 could play active roles in tumour response to radiotherapy (Milas et al, 2004).
Our analysis of rectal cancers investigated the implication of p53 dysregulation with relation to the response to neoadjuvant therapy. The strength of our study was three-fold. First, p53 was analysed at four different levels; second, it was a prospective study with the criteria for inclusion being (i) a pretreatment biopsy (ii), a complete course of radiotherapy, and (iii) surgical resection. Finally, two methods of response to treatment were used, namely, pathologic downstaging and histologic tumour regression grade. Our results based on a multilevel p53 analysis approach confirm that although p53 appears to be a major regulator, nevertheless it is certainly not the major indicator of tumour radiosensitivity.