The institutional review board of the University of Alabama at Birmingham (UAB) approved these experiments, and the Bioethics Committee reviewed the proposed effort. From the UAB Hospital, we collected data for 650 patients who were diagnosed and underwent surgery for primary colorectal adenocarcinoma with curative intent between 1987 and 1993. Use of patients from this period maximized post-surgery follow-up, because 70% of the patients (78 of 112) had either stage II or III CRCs (Table 1).

During our initial selection process, patients who received radiation and/or any chemotherapy before surgery were excluded. We included only those patients who completed at least 3 months of post-surgical adjuvant chemotherapy and for whom there was complete information on dosage and duration of treatment. With these criteria and the availability of paraffin blocks, the final group consisted of 56 patients who had surgery plus 5-FU-based adjuvant therapy.

Details of the chemotherapy were as follows: Twenty eight patients received 5-FU alone, 12 patients received 5-FU plus levamisole (LV), 10 patients received 5-FU plus leucovorin (LC), 4 patients received 5-FU plus doxorubicin (5-FUDR), 1 patient received 5-FU/1-(2- chloroethyl)-3-cyclohexyl-1-nitrosourea) (CCNU), and 1 patient received 5-FU/LV/LC. The control group of 56 patients, who were matched for age; gender; ethnicity; and tumor stage, location, and histologic differentiation, and who had undergone only curative resection without adjuvant therapy after surgery were selected randomly from the initial patient pool. In the surgery-alone group, patients with stage III or IV CRCs did not receive adjuvant therapy for various clinical and personal reasons but had undergone surgery with a palliative intent. The final study sample consisted of 112 patients; their characteristics are given in Table 1.

The surgical pathology reports were reviewed by three investigators (CS-C, NCJ & CKS), and a pathologist (NCJ) individually reviewed slides stained with hematox ylin and eosin for the degree of histologic differentiation and re-graded lesions as well, moderate, poor or undifferentiated (32, 33). Well and moderately differentiated tumors were pooled into a low-grade group, and poor and undifferentiated tumors into a high-grade group (34). Pathologic staging was performed according to the criteria of the American Joint Commission on Cancer (35). The International Classification of Diseases for Oncology codes were used to specify the anatomic location of the tumor (32). The tumor was considered mucinous if ≥ 50% demonstrated mucinous histology (32). The anatomic sub-sites were the proximal colon, the distal colon, and the rectum. Three-dimensional tumor size was determined; the largest dimension was used for statistical purposes.

Patient demographics, along with clinical and follow-up information, were retrieved retrospectively from medical records, physician charts, and pathology reports and from the UAB tumor registry. Patients were followed, either by their physician or by personnel associated with the tumor registry, until their death or the date of the last documented contact. Through telephone and mail contacts, these personnel ascertained outcome (mortality) information directly from patients (or relatives) and physicians. This information was validated by examination of the state death registry. Demographic data, including patient age at diagnosis, gender, race/ethnicity, date of surgery, date of the last follow-up (if alive), date of recurrence (if any) and date of death, were collected. Collection of followup information, performed every six months, ended in April 2010. Laboratory investigators (VRK & CS-C) were blinded to the outcome information until completion of the assays.

Formalin-fixed, paraffin-embedded archival tissues were collected from the surgical pathology division of the UAB Hospital. From the blocks, tissue sections (5-μm thick), representative of normal mucosa and invasive adenocarcinomas were cut 1 to 2 days before staining to avoid potential problems in antigen recognition due to storage of cut sections on glass slides(39,40). Sections were de-paraffinized in xylene and rehydrated in graded alcohols. For antigen retrieval of Bax and Bcl-2, the slides were microwave boiled in citrate buffer (10 mmol/L, pH 6.0) for 7 min. For p53, antigen retrieval is not required (8,41,42). Endogenous peroxidase activity was quenched with 3% hydrogen peroxide for 5 min. Non-specific binding of the primary antibodies was blocked by incubating the slides in 3% goat serum at room temperature for 1 hr in humidity chambers with the primary mouse monoclonal antibodies for Bax (Clone B9, Santa Cruz Biotechnology Inc, CA, USA) (1:200), Bcl-2 (Clone 124, Roche Diagnostic corporation, Indianapolis, IN, USA) (1:60) and p53 (Clone BP53, BioGenex, San Ramon, CA, USA) (1:80). A biotin-streptavidin horseradish peroxidase detection kit was used as the secondary detection system (BioGenex). The biotinylated goat anti-mouse secondary and avidinhorseradish peroxidase label were each applied for 10 min. The antigen-antibody complex was recognized by incubating with the chromogen, diamino-benzidine, for 7 min. The slides were counterstained with hematoxylin for 1 min. Known positive controls were included in each staining run; negative controls were obtained by omitting the primary antibody. Slides were then dehydrated in graded alcohols, cleared in 3 xylene baths, and mounted with Permount™ mounting media. As we reported earlier (43), these antigens are stable in paraffin blocks.

Stained slides were evaluated under a light microscope, and the staining was scored semi-quantitatively by CS-C, NCJ and UM, CKS together to limit the bias; if there was a disagreement in their scores, they reached to a consensus before proceeding. Obser vers were bl inded for the clinicopathologic data and the treatment status. Phenotypic expression of Ba x and Bcl-2 was present in the cell cytoplasm and accumulation of p53 in the nucleus (p53^nac). As described earlier (8,9,12,13), the percentage of positive cells and staining intensity were taken into consideration for estimation of the final immunostaining score (ISS). The intensity of staining of individual cells was scored on a scale of 0 (no staining) to + 4.0 (strong staining). In addition, each reviewer estimated the proportion of cells stained at each intensity level. The percentage of cells and the corresponding intensity were then multiplied to obtain the ISS. For each case, the final ISS was the average of the values estimated by these three investigators.

Molecular marker expression was dichotomized into high expressors and low expressors, based on the cut-off values discussed below. For Bax expression, the median ISS (1.8) of tumor tissues was taken as the cut-off value; i.e., the tumors expressing ≥ 1.80 were considered as “high expressors” (equivalent to > +1 of routine immunohistochemistry, IHC, scoring in the diagnostic pathology setting) and those CRCs with ISS 27,44,45). For Bcl-2 expression, based on prior studies by us and others (8), we chose 0.5 ISS as the cut-off value. We considered only tumor cells with distinct nuclear immunostaining for p53 as positive and considered the tumor positive only if there was ≥10% positivity of all malignant cells in a tissue section, as described earlier (9). We chose this cutoff because, at this value, there was the highest concordance between immunohistochemical detection of p53^nac and point mutations of the p53 gene detected by single-strand confirmation polymorphism and DNA sequencing analyses. At this cutoff value, IHC detects 95% of point mutations in the p53 gene (42). The cut-off value for Bax/Bcl-2 ratio was based on their levels of expression in benign colonic epithelium. We used the ISS values of Bax and Bcl-2 to determine the Bax/Bcl-2 ratio of each case, then a median value of 0.25 was obtained. This 0.25 value was used as a cut-off for Bax/Bcl-2 ratio to dichotomize CRCs into groups of “high” and “low” ratios.

Correlations between biomarkers and clinical response (overall survival) were evaluated by chi-square tests. The type-I error rate of each test was controlled at < 0.05. All analyses were performed with SAS statistical software, version 9.0 (46). Kaplan-Meier curves and log-rank tests were used to assess the effect of the selected biomarkers in univariate analyses (47). Overall survival was estimated as the number of months from surgery to the date of death or last contact. Patients who were alive at last contact and those who died due to a cause other than colorectal cancer were “right censored.” Only those deaths due to CRCs were considered as events. Multivariate Cox proportional hazards tests (48) were utilized to identify the independent prognostic value of molecules indicators of survival, after controlling for patient age, gender, race, tumor location, tumor size, tumor stage, tumor grade, and the three molecular markers, Bax, Bcl-2, and p53^nac. Models were built separately for each patient group (the group of patients who received chemotherapy after surgery and the group who underwent surgery without any pre- or post-surgery chemoor radiation therapy).

Table 1 shows the patient distribution; their demographic, clinicopathological and molecular characteristics; and their correlation with survival. For both treatment groups, there were similar distributions of patient age, gender, ethnicity, tumor stage, tumor location, tumor size, and tumor grade, in terms of deaths due to CRC. The median follow-up period of the complete study population of 112 patients was 9.31 years (range < 1 – > 20 years).

Univariate Kaplan-Meier survival analysis demonstrated no significant differences in overall survival rates between the surgery-alone and the 5-FU-treated patient groups (log rank, P=0.71) (data not shown).

We analyzed for the presence of mutations in the (G) 8 tract of the Bax gene in a human CRC cell line (LoVo) and in 83 CRCs. The LoVo cells displayed a bi-allelic Bax (G) 8 frame-shift mutation; this status was used as a reference in CRCs for Bax mutations (Fig 1). In our analysis, 23 of 83 (28%) CRCs demonstrated biallelic Bax (G) 8 frame-shift mutations. The majority of CRCs with mutations at the G (8) tract also had low Bax expressing (20 of 23, 87%). CRCs that displayed these mutations were significantly higher for male patients (17 of 23, 74%) and distal tumors (18 of 23, 79%). However, there was no association between the presence of Bax (G) 8 mutations with age, race/ethnicity, depth of wall infiltration, tumor grade, tumor stage, lymph node invasion, or presence of distant metastasis (data not shown). Since the number of CRCs with Bax mutations is small, we have not further analyzed the mutational data to assess correlation between Bax mutations and patient survival in the surgery alone and surgery and 5-FU therapy patient groups separately.

For each patient, the ratio of Bax/Bcl2 expression was determined and correlated with patient survival separately for each treatment categor y. The ratio of Ba x/Bcl2 expression was not correlated with patient survival in surgery alone category (Fig 3C). In contrast, the ratio of Bax/Bcl2 expression significantly correlated with patient survival, indicating that those patients with a high Bax/ Bcl2 ratio value would not benefit from 5-FU treatment; however, those patients with low Bax/Bcl2 ratios were more likely to have better survival when treated with 5-FU based therapies (Fig 3D).

Expression of p53^nac was generally homogenous; however, p53 staining was observed in the cytoplasm of malignant cells in 7% of cases (8 of 112) (Fig 2G-I). The cytoplasmic staining was not further analyzed. High levels of p53^nac were found in tumors of 59% of patients (33 5-FU-treated and 32 surgery-alone). For both categories of patients, there were no significant differences in the incidence of deaths due to CRCs (Table 1), in the median survival (Table 3), or survival rates (data not shown) in relation to p53^nac (Table 1).

As determined by multivariate analyses of Cox proportional hazards, surgery-alone patients with low Bax expression had 5.33 times higher mortality compared to those with high Bax expression (HR: 5.33; CI: 1.78-15.94) when adjusted for demographic and clinicopathological variables and for expression Bcl-2, the Bax/Bcl2 ratio, and p53^nac (Table 4). 5-FU treated patients with low Bax/Bcl2 ratios had 21.55 times higher mortality compared to those with high Bax/ Bcl2 ratios (HR: 21.55; CI: 1.46-31.00), when adjusted for other variables (Table 4). Bcl-2 and p53^nac, however, were not independent prognostic indicators of survival in either group of patients (Table 4).