Biological predictive factors in rectal cancer treated with preoperative radiotherapy or radiochemotherapy

We analysed the expression of microsatellite instability, p53, p21, vascular endothelial growth factor and thymidylate synthase (TS) in pretreatment biopsy specimens from 57 locally advanced rectal cancers. The aim of the study was to correlate the expression of these markers with pathological response. Nineteen patients were treated with preoperative concomitant radiotherapy (RT) and fluorouracil/oxaliplatin-based chemotherapy (RCT), while 38 had RT alone. Pathological complete remission (pCR) and microfoci residual tumour (micR) occurred more frequently in patients treated with RCT (P=0.002) and in N0 tumours (P=0.004). Among patients treated with RCT, high TS levels were associated with a higher response rate (pCR+micR; P=0.015). No such correlation was found in the RT group. The other molecular factors were of no predictive value. Multivariate analysis confirmed a significant interaction between nodal status and the probability of achieving a pathological response (P=0.023) and between TS expression and treatment, indicating that a high TS level is predictive of a higher pathological response in the RCT subset (P=0.007). This study shows that lymph node status is the most important predictive factor of tumour response to preoperative treatment. Thymidylate synthase expression assessed immunohistochemically from pretreatment tumour biopsies may be a useful predictive marker of rectal tumour response to preoperative RCT.

Preoperative radiation therapy alone (RT) or combined with chemotherapy (RCT) and improvements in surgical techniques, particularly the standardisation of total mesorectal excision (TME), have led to improved outcomes in the management of locally advanced rectal cancer in recent years (Swedish Rectal Cancer Trial, 1997;Sauer et al, 2004). With this approach, pathologic complete response (pCR), which is an important clinical predictor for both local control and disease-free survival, is achieved in 7 -31% of patients (Ruo et al, 2002). In addition, obtaining a complete or near-complete pathologic response before surgery may increase the number of sphincter-sparing procedures (Ruo et al, 2002). The ability to predict tumour response before treatment may significantly impact the selection of patients for preoperative combined-modality therapy as well as potentially modify postoperative treatment plans.
Thymidylate synthase (TS), the target enzyme of the antimetabolite 5-FU, has been shown to be an independent prognostic marker of 5-FU chemotherapy in gastrointestinal tumours. Several preclinical and clinical studies have demonstrated that high TS levels correlate with 5-FU resistance in various malignancies (Leichman et al, 1997;Aschele et al, 1999). However, most of these studies dealt with outcome prediction for postoperative chemotherapy or treatment of metastatic disease rather than preoperative treatment. Therefore, the potential of TS expression levels to predict response to preoperative combined-modality therapy remains unsettled.
Little is known about other potential biological markers, such as p53, p21 and radio-and chemosensitivity of rectal cancer cells. Vascular endothelial growth factor (VEGF) is a potent mediator of tumour angiogenesis and has currently been assessed as a response predictor in rectal cancer. The results of a recently published study show that VEGF assessed immunohistochemically from pretreatment tumour biopsies may be a useful marker for rectal tumour response to preoperative RT (Zlobec et al, 2005a).
The aim of the present study was to test the predictive value of a number of tissue biomarkers, including TS, VEGF, p53, p21, MLH1 and MSH2 with regard to preoperative RT alone or RCT in rectal cancer.

MATERIALS AND METHODS
A total of 57 patients with stage II and III rectal cancer consecutively treated at our Institution were included in this study, provided that adequate archive tumour tissue from pretreatment biopsy was available for biological studies. Disease staging was by computed tomography scan in all patients. Endorectal ultrasonography was performed in 17 patients.
Thirty-eight patients were treated with preoperative RT alone by using high energy Linac (total dose 40 Gy specified to the isocentre, 250 cGy day À1 , four fractions per week), delivered in 16 fractions, to include the true pelvis (rectal volume, perirectal, presacral and iliac nodes) with the three-field technique and shaped portals. Another 19 patients were treated within a phase II trial with preoperative radiation (total dose 45 Gy to the isocentre, 180 cGy day À1 , with the three-field technique and shaped portals to cover the rectal volume and the perirectal, presacral and the internal iliac lymph nodes) and concurrent continuous infusion 5-FU at a daily dose of 200 mg m À2 and weekly oxaliplatin 60 mg m À2 for 5 weeks. The trial was approved by the Ethics Committee of our Institution. Informed consent was obtained from all patients. All the patients underwent surgery regardless of radiological response, which was not routinely performed. Standard surgery, including total mesorectal excision, was performed in all patients after an interval of approximately 6 weeks after completion of treatment. Further adjuvant treatment after surgery was left to the discretion of the treating physician. Only 18 of the 57 patients (32%) received in addition 5-FU-based adjuvant chemotherapy.
Semiquantitative analysis Immunostaining for HMSH2 and HMLH1 was estimated on a semiquantitative score according to the number of positive tumour cells as follows: 0% (0), o10% (1), 10 -50% (2), 51 -80% (3) or 480% (4). The intensity of staining was also evaluated as weak (1 þ ), moderate (2 þ ) or strong (3 þ ). For each tumour case, the values for the two variables were multiplied, resulting in a score ranging from 0 to 12. The 0 -6 scores were considered as altered expression and 7 -12, as preserved expression. Vascular endothelial growth factor staining was considered positive in the tumour cell cytoplasm; immunoreactivity was graded as follows: positive, more than 10% of carcinoma cells stained, and negative less than 10% of carcinoma cells stained. For the evaluation of p53 and p21waf expression, immunostaining were classified into two groups, corresponding to the percentage of nuclear staining: negative expression (less than 10% positive tumour cells) and positive expression (more than 10% positive tumour cells). Thymidylate synthase expression was quantitated using a visual grading system based on the intensity of staining and was classified into groups from 0 to 3, where 0 and 1 were defined as low intensity, and 2 and 3 were defined as high intensity staining.
Assessment of immunoreactivity from pretreatment tumour biopsies was performed independently by two observers (PC and CB), blinded to postoperative tumour response. In case of disagreement, the relevant biopsies were re-examined simultaneously by both pathologists until an agreement was reached.

Response classification
Tumour response was evaluated pathologically on postoperative specimens. Pathological complete remission was defined as no evidence of residual carcinoma or ypT0N0. Partial response was characterised by the presence of residual carcinoma microfoci (micR) typically measuring from 0.3 to 0.9 cm in diameter. Nonresponsive tumours (NR) have large residual carcinoma. For the correlation between biomarkers and response to treatment, we separated the patients into two groups: responding (pCR, and partial response characterised by the presence of micR; pCR þ micR) vs NR (Zlobec et al, 2005b).

Statistics
Fisher's exact test was used to compare the distribution of the dichotomised molecular markers and tumours' characteristics in the two groups (responders and non-responders). Po0.05 was considered statistically significant. To assess the presence of a significant influence of treatment, tumour stage and molecular markers on response, the multivariate logistic regression model was applied, including in the model the appropriate interaction terms between treatment and TS status. The SPSS Software (version 8.0) was used in all analysis.
To assess the effect of treatment and molecular markers on response, a regression logistic model was used, with the probability of achieving a response as dependent variable and TS status, nodal status, treatment and the interaction terms as covariates. Besides nodal status (odds ratio (OR) ¼ 8.0; 95% CI 1.34 -47.8, P ¼ 0.023), only the interaction of TS and treatment was kept in the model as independent predictive factors for response (OR ¼ 30.6; 95% CI 2.5 -371.7, P ¼ 0.007). As a consequence of this interaction, the increased activity seen in tumours expressing high TS levels in patients treated with preoperative RCT was not seen in patients treated with RT only (OR ¼ 1.5; 95% CI 0.14 -16.3).

DISCUSSION
Multidisciplinary treatment of rectal cancer has evolved over the years from adjuvant chemoradiation to preoperative radiation and, more recently, to preoperative combined chemoradiation, in view of an improved toxicity and local control (Sauer et al, 2004).
Oxaliplatin added to FU was found to enhance the efficacy in both metastatic and adjuvant colon cancer setting and may act as a potent radiation sensitiser. For these reasons, this regimen is also being used increasingly in preoperative treatment of rectal cancer. Concomitant RT and oxaliplatin with either FU/LV or capecitabine can achieve pCR rates in 15 -28% in locally advanced rectal cancer (Aschele et al, 2005;Chau et al, 2006). The results of the present study show that, besides disease extent, the type of treatment is the most important predictor of tumour response, with patients treated with combined RCT having a higher probability of achieving a pCR compared to those treated with RT alone (P ¼ 0.002). A pCR rate of 21% with an additional 26% micR rate in the RCT group in this study is encouraging. A high response rate is of clinical significance in rectal cancer because patients who achieve a complete or near-complete pathologic response may experience improved long-term local control and overall survival (Ruo et al, 2002). Therefore, the identification of distinct clinical, pathological and molecular parameters, with the ability to predict   response to preoperative RCT, may become a useful tool in the therapeutic management of rectal cancer. Thymidylate synthase plays an important role in pyrimidine nucleotide synthesis and represents an important chemotherapeutic target for 5-FU. Overall, the proportion of colorectal cancers expressing high levels of TS is about 50% (range, 14 -80%) in both advanced and adjuvant settings (Edler et al, 2002;Johnston et al, 2003). In our study, the percentage of low and high TS expressors was 39 and 56%, respectively.
Several studies have been published during the past few years relating to the measurement of expression of TS to response to fluoropyrimidines as well as to the overall outcome of patients with colorectal carcinoma. A recent meta-analysis showed that colorectal cancers expressing TS at high levels seem to be associated with a poorer prognosis compared with low TS-expressing tumours, although in the adjuvant setting this seems to be the case only for patients treated by surgery alone. In patients treated with both surgery and adjuvant 5-FU, TS expression does not seem to predict clinical outcome (Popat et al, 2004).
A direct correlation between TS expression and response to 5-FU has been reported both in vitro and in vivo. Preclinical studies have demonstrated that an increase in TS expression is associated with resistance to fluoropyrimidines (Johnston et al, 1992). In advanced colorectal cancer, patients with high levels of TS are unlikely to respond to 5-FU, whereas patients with low levels have higher than expected response rates (450%) Cascinu et al, 1999).
To date, only a small number of studies have concentrated on the role of TS and tumour response in rectal cancer patients, especially 5-FU-based chemoradiotherapy. In rectal cancer, low TS gene expression has been found to correlate significantly with tumour response after neoadjuvant 5-FU-based CRT (Jakob et al, 2005). In contrast with these results, the present study indicates that tumours responding to preoperative RCT most often express high levels of TS in their pretreatment biopsies, whereas nonresponding tumours are generally poorly immunoreactive.
Several factors can account for the controversial results on the predictive/prognostic role of TS expression. The first is related to the different techniques used to assess TS expression. The commonest technique used is IHC (Popat et al, 2004). Several semiquantitative methods for dichotomising TS expression have been used. The most common method relates TS expression to chromagen intensity with either a 4-or 5-grade scale. Some studies further subcategorised staining by a either focal or diffuse pattern. In a recently published study, for example, a significant correlation between protein expression and tumour response in rectal cancer patients was seen only when both staining intensity and staining pattern were considered, with a significant association between high TS expression in tumour biopsies and non-response to therapy (P ¼ 0.04) (Jakob et al, 2005). In our study, we used a four chromagen intensity grade, ranging from 0 to 3, where 0 and 1 were defined as low intensity, and 2 and 3 were defined as high intensity staining, as reported previously (Allegra et al, 2002). The existence of such a high variability supports the demand for a more uniform evaluation of IHC expression.
Another reason for the contradictory results may be related to the different FU schedule used. FU may act as two different drugs according to the mode of administration. Bolus FU may exert its major effect on RNA, whereas continuous infusion may have a preferential effect on TS (Sobrero et al, 1997). Most published studies of patients with advanced colorectal cancer indicate that patients with high tumour levels of TS are unlikely to respond to infusional treatment with 5FU (Leichman et al, 1997;Aschele et al, 1999). In contrast, in the present study, FU was administered as continuous infusion at a daily dose of 200 mg m À2 and high TS levels were found to be predictive of response to RCT.
There is some evidence that high TS protein expression may predict high sensitivity to FU in the adjuvant setting. At least four studies have suggested that patients whose primary tumours had high TS expression may benefit from FU-based adjuvant treatment, while those whose tumours had low expression do not (Johnston et al, 1994;Yamachika et al, 1998;Takenoue et al, 2000;Edler et al, 2002). In rectal cancer, adjuvant FU-based chemotherapy demonstrated significant improvement in disease-free and overall survival for rectal cancer patients with high TS levels (Johnston et al, 1994). To our knowledge, this is the first study to suggest that patients who have high TS-expressing rectal cancer have a higher probability to respond to preoperative CRT.
It should also be noted that many more studies were conducted regarding the prognostic value of TS expression in primary colorectal cancer treated with postoperative chemotherapy and the role of TS expression as a predictor of chemotherapeutic benefit in metastatic disease rather than response prediction to preoperative therapy. Importantly, TS expression measured in primary tumours may not reflect TS levels in lymph node or other metastases (Marsh et al, 2002).
Moreover, evidence has accumulated that oxaliplatin treatment results in downregulation of TS expression, providing a possible explanation, although still speculative, for the higher response rate of the combination of oxaliplatin and FU, compared with oxaliplatin alone, in FU-resistant patients (Yeh et al, 2004). In addition, it has been shown that oxaliplatin remains highly cytotoxic in cells that overexpress TS. In breast cancer-derived cell line with tetracycline-regulated expression of TS, TS overexpression confers an increased resistance to TS-targeted drugs, while the growth inhibitory effect of other drugs such as oxaliplatin is unaffected by TS upregulation (Longley et al, 2001). The use of an oxaliplatin-containing chemotherapy in our study could explain our findings. Furthermore, the absence of TS effect on RT treatment makes our hypothesis sound.
When we explored the potential predictive value of other molecular markers, we found no effect on response. Contrary to a previous study on rectal cancer patients receiving preoperative RT, showing that VEGF expression in NR was significantly greater than in completely responsive tumours, the present study does not demonstrate a significant role of VEGF overexpression on clinical response (Zlobec et al, 2005a). Also, the present data do not confirm the previously reported value of high microsatellite instability, p21WAF1/C1PI and p53 for predicting tumour response to preoperative RT or RCT (Charara et al, 2004;Kelley et al, 2005;Komuro et al, 2005). These conflicting results may be attributable to the retrospective nature and the limited detection power inherent in studies that test small subsets of patients.
An interesting finding, although not directly related to the goal of this work, was the observation that lymph node status was significantly associated with tumour response. Lymph node status was first demonstrated to have prognostic value in patients with locally advanced rectal cancer treated on the CAO/ARO/AIO-94 trial of the German Rectal Cancer Study Group (Liersch et al, 2006). All patients who developed cancer recurrence had a persistently positive lymph node status after therapy, which reflects the poor prognosis of these patients. In our study, lymph node status turned out to be the most important predictive factor, with a significantly higher response rate in N0 tumours as compared with N1 -2 tumours (47 vs 9%; P ¼ 0.004), independently of treatment and TS status.
In conclusion, despite the small sample size and the retrospective nature of the study, our data indicate that lymph node status is the most important predictive marker of tumour response to preoperative treatment in rectal cancer. In addition, among a variety of biological markers evaluated, immunohistochemical assessment of TS from pretreatment tumour biopsies was the only one with potential for the prediction of tumour response to preoperative oxaliplatin-based RCT. This observation may deserve further validation in prospective studies of FOLFOX-RT regimens.