The transmembrane glycoprotein CD133 (cluster of differentiation 133; also known as Prominin or PROM1) has been described as a potential stem cell marker in colorectal cancer and is associated with higher tumorigenic potential and resistance to radiochemotherapy (RCT). In this study, CD133 expression was evaluated in pre-RCT tumor biopsies and the corresponding post-RCT surgical specimens from patients with locally advanced rectal adenocarcinoma, and expression levels were correlated with histopathologic features and clinical follow-up.
One hundred twenty-six patients with International Union Against Cancer (UICC) stage II/III rectal cancer who received preoperative 5-fluorouracil (5-FU)-based RCT within the German Rectal Cancer Trials were investigated. Pre-RCT and post-RCT CD133 expression levels were determined using immunohistochemistry and were correlated with histopathologic parameters, tumor regression grade, cancer recurrence, and patient survival.
Compared with pre-RCT biopsies, significantly higher CD133 expression was observed in tumor specimens (P = .01). However, no correlations were observed for either biopsies or tumor specimens between CD133 expression levels, histopathologic characteristics, or survival. In matched analyses of corresponding biopsy/tumor pairs, patients who had an increased fraction of CD133-expressing (CD133+) cells after preoperative RCT had significantly higher residual tumor stages (P = .02) and lower histopathologic tumor regression (P < .01). Moreover, these patients had significantly reduced disease-free survival and cancer-specific overall survival in univariate analysis (P < .001 and P = .004, respectively) and multivariate analysis (P = .003 and P = .024, respectively).
Preoperative radiochemotherapy (RCT) followed by total mesorectal excision (TME) is currently regarded as standard treatment for locally advanced International Union Against Cancer (UICC) stage II/III rectal cancer.1 However, individual response to neoadjuvant RCT is considerably heterogeneous, and despite enhanced local control, particularly patients who had a pathologic complete response (pCR) had significantly enhanced disease-free survival rates compared with intermediate responders and nonresponders.2 Nevertheless, distant metastatic disease remains a challenging problem in the treatment of primary rectal adenocarcinoma. Valid molecular markers are needed for patient-based risk assessment and for individually tailored, multimodal treatment approaches.
The transmembrane glycoprotein CD133 (cluster of differentiation 133; also known as Prominin or PROM1) is localized in the apical plasma membrane of epithelial cells and initially was described as a surface antigen characteristic of human hematopoietic stem cells.3 Successive CD133 also has been reported as a potential cancer stem cell marker in various solid tumors, including colorectal cancers.4-6 The cancer stem cell (CSC) model indicates that only a small subpopulation of cancer cells is responsible for tumor growth and expansion as well as metastatic behavior and dissemination.7, 8 The main characteristic properties of CSCs include the ability of unlimited self-renewal, pluripotence, and intrinsic resistance to various exogenic noxa.9 Tumor-initiating CSCs in colon cancers (Co-CSCs) are highly enriched in a subpopulation of cells that express CD133 (CD133+). Injected into immunodeficient mice, CD133+ cancer cells have demonstrated the ability to repopulate tumors with homologous phenotypic characteristics; whereas CD133-negative (CD133−) cancer cells failed to effect tumor initiation.4, 5 Furthermore, CD133+ cells have distinct properties of resistance towards radiation and (radio)-chemotherapy in vitro10, 11 and in patients with colon cancer,12 indicating their potential to serve as a prognostic factor for rectal cancer patients undergoing multimodality cancer treatment. By contrast, the role of CD133 expression in rectal cancer treated with neoadjuvant RCT is widely unknown. Whereas, in primary colon cancers, CD133 expression has been strongly correlated with distant metastases,13, 14 only limited data derived from small cohort studies are available from patients with rectal cancer who received preoperative RCT.15-18
In those small groups, however, CD133 expression reportedly was increased in surgical specimens after preoperative RCT compared to the expression in pre-RCT biopsies and tumors from patients who underwent primary surgery.16, 18 In addition, patients who had elevated CD133 expression in residual tumor tissues showed significantly worse oncologic outcomes after RCT compared to patients who had low or absent CD133 expression.17 Recent in vitro investigations further indicated that HT29 human colorectal cancer cells with acquired resistance to 5-FU/Oxaliplatin chemotherapy expressed significantly higher CD133 levels than the nonresistant parental HT29 cells.19 Moreover, CD133 expression in circulating colorectal carcinoma cells from patients with Duke's B and C disease has been identified as a strong marker of poor prognosis and decreased survival.20
In the current study, CD133 expression was evaluated in a large set of prospectively collected tissue samples from patients with locally advanced rectal cancer (UICC stage II/III) who received identical neoadjuvant RCT regimens within randomized clinical phase 3 trials. CD133 expression characteristics in pretreatment cancer biopsies and in corresponding post-RCT surgical specimens were assessed and correlated with clinicopathologic tumor characteristics and clinical follow-up.
MATERIALS AND METHODS
In total, 126 patients with locally advanced (UICC stage II/III) rectal adenocarcinoma were included in this study. All patients received standardized preoperative RCT followed by TME-based curative surgery within the randomized clinical phase 3 German Rectal Cancer Trials by the Working Group of Surgical Oncology/Working Group of Radiation Oncology/Working Group of Medical Oncology of the German Cancer Society (CAO/ARO/AIO) (studies CAO/ARO/AIO-941 and CAO/ARO/AIO-04). The current study was approved by the Ethics Committee of the University of Goettingen. Informed consent for additional translational research on biopsy and tumor samples was obtained from all patients before enrollment into this study, in accordance with German Clinical Research Unit 179 standard operating procedures (available at: www.kfo179.de; [access date]). Pretreatment staging procedures included rigid rectoscopy with endorectal ultrasonography, chest x-ray, contrast-enhanced computed tomography scans of the abdomen and pelvis, and magnetic resonance imaging studies of the pelvis to confirm locally advanced tumor stage and to exclude patients who had evidence of distant metastases at the time of initial rectal cancer diagnosis.
Treatment Procedures, Pathologic Staging, and Tumor Regression Grading
According to the protocols of the German Rectal Cancer Trials (CAO/ARO/AIO-94 and CAO/ARO/AIO-04), all patients received a cumulative irradiation dose of 50.4 grays (Gy) in 28 fractions of 1.8 Gy and concomitant application of chemotherapy with either 5-FU (78 patients; 62%) or combined 5-FU/oxaliplatin (48 patients; 38%). Patients subsequently underwent total mesorectal excision within 6 weeks after completion of RCT. Surgical procedures comprised anterior resections, abdominoperineal resections, and discontinuous resections (Hartmann procedure). Perioperative assessment of TME quality was performed in all surgical specimens.
Pretreatment biopsies from different representative luminal tumor sites were obtained during staging rectoscopy and stored in the prospective translational biobank in the Department of General and Visceral Surgery of the University Medical Center Goettingen. Pathologic staging in post-treatment surgical specimens was performed according to the respective TNM classification.21 Histopathologic tumor regression grading (TRG) was based on the semiquantitative 5-point grading system initially described by Dworak et al,22 in which TRG 0 indicates no regression, TRG 1 indicates a dominant tumor mass with obvious fibrosis or mucin, TRG 2 indicates a dominantly fibrotic or mucinous alteration with few tumor cells or cell groups; TRG 3 indicates very few tumor cells in fibrotic or mucinous tissue, and TRG 4 indicates no tumor cells and complete histopathologic tumor regression. Tumor downstaging was defined as RCT-induced reduction by 1 or more UICC stages comparing clinical UICC (cUICC) stages with pathologic UICC (ypUICC) stages.
Immunohistochemical Staining of CD133
Immunohistochemical staining was performed using a monoclonal rabbit antihuman CD133 antibody (clone C24B9; Cell Signaling Technology Inc., Danvers, Mass). Two-micrometer sections from formalin-fixed, paraffin-embedded rectal cancer tissues were mounted on microscope slides (Starfrost; Light Laboratories, Dallas, Tex) and subjected to an automated staining procedure with standardized tissue preparation, epitope retrieval (at 98°C for 60 minutes), and antibody incubation (at 37°C for 32 minutes; dilution 1:100) using a benchmark XT autostainer (Ventana Medical Systems, Tucson, Ariz). Alkaline phosphatase (Red Detection Kit; Ventana Medical Systems) was used for color development and visualization of the epitope-antibody product, and hematoxylin was used for counterstaining.
CD133 Expression Patterns and Scoring of CD133 Immunoreactivity
One medium-power field (×20) in tumor biopsies and 1 to 5 medium-power fields (×20) were investigated in post-treatment surgical specimens, depending on the amount of residual tumor cells. We observed 3 different staining patterns of CD133, which were defined as follows: CD133+ 1, staining of the apical/luminal membrane with/without intraluminal cell debris; CD133+ 2, staining of intraluminal cell debris without membranous staining; and CD133+ 3, cytoplasmic staining in nonglandular residual tumor cell clusters. The percentage of CD133+ cells was recorded for each specimen, and a defined semiquantitative scoring system was applied with 0% to ≤40% positive cells indicating low CD133 expression and >40% positive cells indicating high CD133 expression, independent of the staining pattern. The cutoff of 40% positive cells was based on the median CD133 expression in the biopsy samples. According to Horst et al,14 in cases of exclusive staining of intraluminal cell debris without membranous positivity, cells from the respective glands are considered CD133+. All specimens were evaluated by 2 independent observers who were blinded to clinicopathologic parameters and patient outcomes.
Clinical Follow-Up and Criteria for Relapse
According to the study protocol of the German Rectal Cancer Trials, all patients were re-evaluated at 3-month intervals postsurgery within the first 2 years and at 6-months intervals thereafter. After-care consisted of pertinent medical history, physical examination, and blood examination, including carcinoembryonic antigen (CEA) levels and abdominal ultrasound at every follow-up visit. Rigid rectoscopy was performed every 3 months in the first year, every 6 months in the second year, and once annually thereafter. Chest x-rays and computed tomography scans of the abdomen and pelvis were obtained at regular intervals within the study protocols and if patients had suspect findings during routine examination or increased CEA levels. Local relapse was defined as tumor recurrence within the pelvis after curative resection, and distant relapse was defined as recurrence outside the pelvis after curative resection. Histologic confirmation of any tumor relapse was encouraged. Alternate acceptable criteria included sequential enlargement of the lesion on radiologic images with simultaneous increase of serum CEA. Disease-free survival (DFS) was defined as the time from surgery to the initial diagnosis of any tumor relapse. Cancer-specific overall survival (CSS) was defined as the time from surgery to death from rectal cancer. The median follow-up in this study was 45 months.
Statistical analysis was performed using the open-source statistical computing software R (R Foundation for Statistical Computing, Vienna, Austria) with the packages exactRankTests and survival. The CD133 ratio was scored as the log2 fold-change in percentage scores of CD133+ cells in tumor and biopsy specimens. For both the nominator and the dominator, 0.1 was added to avoid division by 0. Correlation of the CD133 ratio (ie, the increase or decrease of CD133+ cells in residual tumors) with clinical outcome variables was computed either with a Wilcoxon 2-sample test for comparison of 2 groups or as a Kendall tau correlation for continuous or ordinal variables with several groups. Time-to-event data were observed using Kaplan-Meier analysis with a cutoff log2 fold-change >0 (ie, CD133 increases vs no increase or decrease). Significant association of the binary CD133 classes was assessed using the log-rank test. The level of significance was set at α = 5% for all tests.
Clinical Results and Cancer Relapse
Clinical and pathologic staging characteristics of all 126 patients are summarized in Table 1. In the initial pretreatment staging, 92 patients (73%) had mesorectal lymph node involvement (cN+, cUICC III). After preoperative RCT, 45 patients (36%) had evidence of residual lymph node metastases (ypN+, ypUICC III). Pathologic lymph node staging was based on a median lymph node yield of 19 nodes per specimen. Thirteen patients (10%) achieved a pCR (TRG 4) after neoadjuvant RCT. An evaluation of tumor regression revealed TRG 3 in 66 patients (52%), TRG 2 in 28 patients (22%), and TRG 1 in 15 patients (12%). In another 4 patients (3%), regression grading was not evaluable. The therapy-related and surgery-related 30-day morbidity and mortality rates were 23% and 0%, respectively.
Abbreviations: 5-FU, 5-fluorouracil; CRM, circumferential resection margin; Gy, grays; NA, not available; SD, standard deviation; TME, total mesorectal excision; UICC, International Union Against Cancer.
Percentages may not add up to 100% because of rounding.
Tumor distance from anal verge, cm
Clinical tumor classification
Clinical lymph node status
Neoadjuvant treatment regimen
50.4 Gy & 5-FU monotherapy
50.4 Gy & 5-FU/oxaliplatin
Surgical procedure, including TME
Low anterior resection
CRM <1 mm
Tumor regression grade
UICC pathologic tumor classification
UICC pathologic lymph node status
UICC pathologic metastasis classification
ypM1: Not detected before surgery
Distant recurrence only
Local recurrence only
Simultaneous local and distant recurrence
Cancer recurrence developed in 36 patients (29%) during follow-up. In 30 patients (24%), isolated distant metastases were detected, and simultaneous local and distant recurrences were observed in 3 patients (2%). Another 3 patients (2%) were diagnosed with local recurrence only. During follow-up, cancer-related death occurred in 16 patients (13%), all of whom had developed distant metastatic disease.
CD133 Expression in Pre-radiochemotherapy Biopsies and Post-radiochemotherapy Surgical Tumor Specimens
Immunohistochemical analyses were performed on 126 pretreatment rectal cancer biopsies. In 13 surgical specimens from patients who achieved a pCR and in 14 tumors with TRG 3 from patients who had near-total (95%) regression after RCT, immunohistochemical evaluation of CD133 expression failed or was unreliable because of absent or insufficient residual tumor tissue. Thus, 99 post-RCT tumor specimens were accessible for immunohistochemical analyses. Accordingly, 99 matched pairs of native pre-RCT biopsies and the corresponding post-treatment tumor specimens qualified for evaluation of CD133 levels and statistical analyses (Fig. 1A).
In pretreatment biopsy samples, CD133 immunoreactivity was observed in 69% of all 126 specimens analyzed. Thirty-nine rectal cancers (31%) did not initially show CD133 expression. Of the 87 patients who had CD133+ tumors, 27 (21%) had low CD133 expression (≤40% of tumor cells), and 60 (48%) had high CD133 expression (>40% of tumor cells). In pre-RCT biopsy samples, staining was detected only in the apical/luminal membrane and in intraluminal cell debris, whereas cytoplasmic immunoreactivity was not observed.
In post-treatment tumor specimens, 15 of 99 rectal cancers (15%) were negative for CD133, whereas 84 (85%) had membranous staining and/or staining of extracellular cell debris or cytoplasmic positivity. Nine post-RCT tumor specimens without CD133 expression also had negative corresponding pre-RCT biopsy specimens. Three tumors had low (≤40% of tumor cells) initial CD133 expression, and 2 tumors had high (≥40% of tumor cells) expression levels in the corresponding pre-RCT biopsy specimen.
After preoperative RCT, 24 patients (24%) had low CD133 expression (≤40% of tumor cells), and 60 patients (61%) had high CD133 expression (>40% of tumor cells). CD133 was detected exclusively in tumor tissue. No immunoreactivity was observed in normal mucosa epithelium or inflammatory cells. CD133 staining characteristics for all 126 rectal cancer biopsies and the corresponding 99 post-treatment tumor specimens are provided in Table 2.
Table 2. Results of CD133 Immunoreactivity
No. of Patients (%)
Proportion of CD133+ Cells
Biopsy, n = 126
Tumor, n = 99
Abbreviations: +, positive expression; CD133, cluster of differentiation 133 (a transmembrane glycoprotein; also known as Prominin or PROM1).
Low expression: 1%-40% CD133+ cells
High expression: >40% to 100% CD133+ cells
CD133 Expression Before and After Neoadjuvant Radiochemotherapy
Surgical tumor specimens had significantly higher global CD133 expression compared with the expression in initial pre-RCT biopsies. This became evident in both pairwise analyses of the 99 patients who had corresponding pretreatment and post-treatment cancer tissues (P = .013) and when comparing all 126 biopsy specimens with the 99 corresponding post-RCT tumor specimens (P = .015) (Fig. 1A). An exemplary corresponding biopsy/tumor pair with low initial and high post-RCT CD133 expression is depicted in Fig.1C.
Association of CD133 Expression Levels Before and After Radiochemotherapy With Clinicopathologic Parameters and Survival
For the purpose of survival analyses, rectal cancer specimens in which ≤40% of all tumor cells expressed CD133 were regarded as low CD133-expressing tumors and were compared with tumors in which >40% of all tumor cells expressed CD133 (based on the median CD133 expression level of 40% in the 126 tumor biopsies). None of the clinicopathologic parameters were correlated with CD133 expression (dichotomized as high vs low CD133 expression) in pretreatment biopsies or in residual post-treatment tumor specimens. In addition, pre-RCT and post-RCT CD133 expression was not significantly associated with cancer relapse, cancer-related death, DFS, or CSS.
Association of the CD133 Ratio With Clinicopathologic Parameters and Survival
Comparing pretreatment and post-treatment CD133 expression in matched tumor pairs, patients who had an increased proportion of CD133+ cells after neoadjuvant RCT had significantly higher residual tumor stages (P = .033) and a diminished TRG (P = .009). In addition, these patients displayed significantly higher rates of distant metastatic relapse (P = .002) and cancer-related death (P < .001).
In more detail, patients who did not develop metastases had no median increase in the CD133+ cell fraction, whereas patients who developed distant metastases during follow-up had a 2.4-fold median increase in CD133+ cells during RCT. Moreover, patients who died of rectal cancer had a 42-fold median increase in CD133+ cells compared with patients who survived the follow-up and had no median increase (Fig. 2). In addition, we observed a significant correlation between an enrichment of CD133+ rectal cancer cells and reduced DFS (P < .001) and CSS (P = .004) (Fig. 3A,B). Particularly, patients who received an intensified preoperative treatment regimen (5-FU/oxaliplatin) and had no increase in CD133 expression had excellent oncologic outcomes and an absence of cancer-related death in long-term follow-up (Fig. 3C,D). All parameters that had a significant influence on survival in univariate analyses were included in a multivariate Cox regression model. RCT-induced enrichment of CD133+ cells retained a significant correlation with DFS (P = .003) and CSS (P = .024) in this model, further indicating CD133 as an independent prognostic marker in rectal cancer after preoperative RCT (Table 3).
Table 3. CD133 Expression and Survival: Multivariate Analyses
Significant Variables in Univariate Analyses
HR [95% CI]
HR [95% CI]
Abbreviations: +, positive expression; 5-FU, 5-fluorouracil; CI, confidence interval; CSS, cancer-specific overall survival; DFS, disease-free survival; HR, hazard ratio; UICC, International Union Against Cancer.
In the current study, we investigated the expression of CD133 in locally advanced rectal cancers from patients who received 5-FU-based, preoperative RCT. In detail, we assessed the fraction of CD133+ tumor cells in pretreatment biopsies and in corresponding post-treatment surgical specimens that were available for immunohistochemical analyses. To the best of our knowledge, this was the largest investigation to date of CD133 expression in rectal cancers based on a well documented patient cohort that received standardized treatment within the 2 phase III randomized German Rectal Cancer Trials.1 We observed that 69% of all primary rectal cancers had CD133 expression. Although Shmelkov et al reported that CD133 is expressed ubiquitously in differentiated human colon epithelium,23 we observed a strong restriction of CD133 positivity to tumor cells. Furthermore, we observed that, in nonirradiated rectal adenocarcinomas, CD133 was expressed exclusively on the luminal surface of the tumor glands and/or on tumor cell debris shed into the glandular lumen. These findings are in line with recent results reported in colorectal cancer cell lines and primary colon cancers14, 24 and also with results reported in pre-RCT rectal cancers18 indicating comparable staining characteristics. Nevertheless, cytoplasmic immunoreactivity for CD133, which initially was described in pancreatic cancers,25, 26 became evident in several tumors within our collective but was strictly limited to post-RCT specimens. However, because of the rare occurrence of this phenomenon, a correlation with a distinct biologic behavior or survival was not evaluable.
In recent studies, investigators observed that CD133 expression was inversely correlated with survival in patients with colorectal cancer who underwent primary surgery.14, 27 Compared with other potential Co-CSC markers, such as CD44 CD133+ cells and CD166, CD133 expression proved to be the strongest predictor of impaired outcome. In a smaller cohort of rectal cancers, Saigusa et al investigated CD133 expression before and after neoadjuvant RCT. However, preoperative treatment protocols varied considerably in their study, and no significant correlation was demonstrated between post-RCT CD133 levels and clinicopathologic parameters, although patients who had elevated residual CD133 expression had a significantly shorter DFS.28 In our investigation, we also could not demonstrate a significant association between the absolute CD133 expression in pre-RCT and post-RCT rectal cancer specimens and clinicopathologic factors or patient survival. However, we observed a significant global increase in the fraction of CD133+ cancer cells after RCT. In vitro experiments on various colon cancer cell lines revealed analogous results with increased mRNA levels of stem cell-associated proteins after irradiation, including CD133, sex determining Y-box 2 (SOX2), and octamer-binding transcription factor 4 (OCT4), implicating an up-regulation of genes with potential impact on tumor progression during RCT.18, 28 In addition—and, to our knowledge, for the first time—our results demonstrate that those patients who had an increase in the CD133+ tumor cell fraction after RCT had significantly minor local tumor responses and lower histopathologic tumor regression, indicating an enhanced resistance to preoperative RCT. Moreover, these patients developed significantly more distant metastases and had lower DFS and CSS rates. Accordingly, our results suggest that it is not the absolute fraction of CD133+ cancer cells before or after neoadjuvant treatment that may be indicative of tumor progression and development of metastases but, rather, it is the dynamic alterations of CD133 expression under RCT. Nevertheless, it is widely accepted today that CD133 expression is not a highly specific marker for Co-CSC14, 27 and that only 1 in 262 CD133+ cells has cancer-initiating capabilities (in contrast to 1 in 5.7 × 104 unfractionated colon cancer cells).4 However, concluding that only a subpopulation of CD133+ cancer cells has a CSC phenotype, the total amount of CD133+ cells still may be correlated with the number containing “true” CSC. We have demonstrated that CD133 was not expressed in single tumor cells or in defined tumor cell clusters but, instead, was expressed predominantly in coherent areas of the tumor specimen in both pre-RCT and post-RCT rectal cancers, indicating that differentiated non-CSC also may display a CD133+ phenotype. This may explain why, in our current study, pre-RCT and post-RCT CD133 expression levels were not associated with clinicopathologic parameters or survival, in contrast to previous studies with smaller subsets of patients.18, 28 Our results demonstrating that the up-regulation of CD133 expression during RCT is correlated independently with distant metastases and survival indicate that the biologic basis of dynamic alterations of CD133+ cell populations in rectal cancer should be considered for future investigation. Kemper et al29 recently reported that the AC133 epitope of CD133 characterizes an undifferentiated CSC with unique tumor-initiating properties. During the course of differentiation, the AC133 epitope is modulated and is successively glycosylated, rendering it recognizable by the AC133 antibody, whereas the CD133 protein itself is not down-regulated. To specify the role of CD133 in patients with rectal cancer who receive neoadjuvant RCT, monitoring the expression and regulation of the AC133 epitope may be useful for a better understanding of its cellular functions. To date, it remains unclear which intracellular and intercellular processes result in tumor progression and metastatic spread after RCT-induced up-regulation of CD133. In addition to cancer-initiating capabilities, an enhanced radiation-induced DNA damage-repair capacity30 and the presence of aggressive circulating colorectal cancer cells20 have been discussed as potential mechanisms to render CD133+ cells radioresistant and to trigger their metastatic behavior. In addition, the RCT-induced relocalization of the cell-surface protein CD133 to subcellular compartments with the detection of cytoplasmic CD133 immunoreactivity after RCT and its biologic implications require further studies both in vitro and in vivo.
In conclusion, we have demonstrated that it is not the absolute fraction of CD133+ cells in pre-RCT and post-RCT rectal adenocarcinomas but, rather, the enrichment of CD133+ cancer cells during RCT that is correlated with lower local tumor response, increased distant cancer relapse, and impaired survival. The assessment of CD133 expression in patients with rectal cancers who receive preoperative RCT, thus, may be a useful tool to individualize multimodal treatment regimes. Because the role of adjuvant chemotherapy in patients with rectal cancer after neoadjuvant RCT and curative surgery is currently under discussion and its administration is very inconsistent,31 the monitoring of CD133 expression during RCT may help to stratify patients who have a higher risk for recurrence and who qualify for intensified adjuvant treatment as well as after-care programs.
We thank Birgit Jünemann for excellent technical and organizational support as well as high-quality performance of all laboratory procedures and Dr. Ulrike Dürr for critical and helpful comments on the article.
This study was supported by the Deutsche Forschungsgemeinschaft (KFO 179).