Immunohistochemical detection of CD133 expression in colorectal cancer: A clinicopathological study

Abstract

CD133 has been reported to be a cancer‐initiating cell marker in colorectal carcinoma. The objective of this study was to evaluate the frequency of CD133 expression in colorectal cancer, the distribution of CD133‐positive cancer cells, and their relationship to clinicopathological features, including survival. An immunohistochemical examination of CD133 expression and a clinicopathological analysis were performed in the 189 consecutive colorectal cancer patients. CD133 expression was seen at the luminal surface of cancer glands mainly with cribriform features. Expression was detected in only 29 of the 189 tumors (15.3%). Of these, 21 tumors (11.1%) showed CD133 overexpression. All 21 tumors with CD133 overexpression were diagnosed as well‐ or moderately‐differentiated adenocarcinoma. There was no difference in the distribution of CD133 expressing cells between the invasive area and surface area. Although there was no difference in recurrence‐free survival between patients with CD133 overexpression and without, the patients with CD133 overexpression had significantly poorer overall survival (P = 0.03). CD133 overexpression is a risk factor for poorer overall survival in patients with well‐ and moderately‐differentiated adenocarcinoma. Expression of this cancer‐initiating cell marker may vary with the histological type of the cancer, and further investigation of the relationship between its expression and clinicopathological features may be necessary. (Cancer Sci 2008; 99: 1578–1583)

For many years, since the finding that only a minority of the cancer cells in a transplanted animal model formed the tumor, the existence of small populations of cancer cells with the potential to initiate and sustain tumor growth has been hypothesized.^{( 1} , ² , ^{3 )} The cancer stem cell theory based on these findings assumed the existence of a small population of cancer‐initiating cells (C‐IC), which is exclusively responsible for the growth and maintenance of the entire heterogeneous tumor,^{( 4} , ^{5 )} and recent advances in flow cytometry and cell‐sorting systems had made it possible to collect such small distinct populations by using fluorescence‐labeled antibodies or dyes such as Hoechst 33342.^{( 6} , ^{7 )} Many C‐IC markers have been reported in different cancers, and cells positive for these markers that have been sorted from them have shown the C‐IC properties of ‘self‐renewal’ and ‘high tumorigenic potential’.^{( 8} , ⁹ , ¹⁰ , ¹¹ , ^{12 )} As evidences of self‐renewal in C‐IC, implanted or cultured C‐IC marker‐positive cells self‐renew to generate C‐IC marker‐positive cells and differentiate to generate marker‐negative non‐tumorigenic cells, and which is consistent with the hierarchical model in heterogeneous tumor formation. The histological similarity between the original tumors and implanted tumors supports the concept of self‐renewal. Next, the limiting dilution method was used to investigate tumorigenic potential and C‐IC marker‐positive cells were found to generate a new tumor after injection of a very small number of cells, whereas negative cells rarely generated tumors even when a large number of cells were injected. One of the C‐IC markers, CD133, has been reported to be a C‐IC marker in central nervous system tumors and hepatocellular carcinoma,^{( 13} , ^{14 )} and Ricci‐Vitiani and O’Brien recently reported that the CD133‐positive cells in colorectal cancer exhibit the C‐IC properties of self‐renewal and high tumorigenic potential.^{( 15} , ^{16 )} However, no immunohistochemical and clinicopathological studies of CD133 have been performed in a large series of colorectal cancer patients. Immunohistochemical analysis of the C‐IC marker in a large series of cancers may bring us more morphological information including the distribution of C‐IC or the relationship between the C‐IC marker and clinicopathological features.

We focused on the following three points. First, like previously reported in primary (de novo) and secondary glioblastoma, the populations of CD133‐positive cells in different subtypes of colorectal cancer may also be different.^{( 17 )} Second, in normal colon tissues, the somatic stem cells are located near the base of the crypts, and mucosal renewal is maintained within a mesenchymal niche,^{( 18 )} but the polarity of C‐IC in a cancer microenvironment is unknown. Third, although C‐IC seem to be associated with tumor initiation and maintenance in in vitro or mouse models, it is not clear whether the amount of C‐IC is associated with clinical behavior, including patient's survival.

Our aim of this study was to solve these issues.

Materials and Methods

Patients and patient follow up.  A series of 189 consecutive colorectal cancer patients who were treated surgically between January 1996 and September 1997 were entered in this study. In 155 patients, the operation was curative, and in the other 34 patients, it was not because of distant metastasis, including to the liver, lung, or lymph nodes other than regional nodes, or because of peritoneal dissemination. The 35 International Union Against Cancer (UICC) stage I patients were followed up every 12 months by computed tomography and colonoscopy, and the other patients were followed up in the same manner at least every 6 months. Multiple colon cancer and multiple primary cancer patients were excluded from this study. None of the patients received preoperative adjuvant therapy, but all the stage III and IV patients received postoperative adjuvant chemotherapy.

All patients were followed up for survival, and the 155 stage I–III patients whose operation was curative were also followed up for recurrence‐free survival. The follow‐up period was calculated from the date of surgery until January 31, 2006. Recurrence was defined as the initial tumor recurrence. Metastasis or local recurrence was considered evidence of tumor relapse, and only deaths from cancer were considered in the analysis of survival in this study.

Histological examination.  The resected specimens were fixed in 10% formalin, and the size and gross appearance of the tumors were recorded. The entire tumor was cut into 5‐mm slices, and representative slices were embedded in paraffin for histological and immunohistochemical examination. The slices were cut into 3‐µm sections and stained with hematoxylin–eosin (HE) for histological evaluation. Elastica staining was used to examine vascular invasion by the tumor cells.

Immunohistochemical examination.  Sections of 5 µm from paraffin embedded slices that included the most representative area were chosen in each of the 189 patients, and used for immunohistochemical staining of CD133. To investigate the characteristics of CD133‐positive cells, immunohistochemical staining of p53 and Ki‐67 were performed in CD133‐positive cases. The sections were deparaffinized in xylene, dehydrated in a graded ethanol series, and immersed in a 0.3% hydrogen peroxide solution in methanol for 15 min to inhibit endogenous peroxidase activity. After washing with phosphate‐buffered saline (PBS), the sections were placed in ethylenediaminetetraacetic acid buffer at pH 8.0 for CD133, or citrate buffer for p53 and Ki‐67 stainings. For antigen retrieval, the slides were heated at 95°C for 20 min in a microwave oven and allowed to cool for 1 h at room temperature. The slides were then washed three times with PBS, and nonspecific binding was blocked by pre‐incubation with 2% normal swine serum in PBS for 60 min at room temperature. Each slide was then incubated overnight at 4°C with primary anti‐CD133 antibody (AC133; Miltenyi Biotec, Auburn, CA, USA) at a dilution of 1:100, anti‐Ki‐67 antibody (MIB‐1; Dakocytomation, Glostrup, Denmark) at a dilution of 1:100, or antip53 antibody (polyclonal antibody; Nichirei, Tokyo, Japan) at a dilution of 1:1000 in the blocking buffer. After washing the slides three times with PBS, they were incubated with EnVision (DAKO A/S, Glostrup, Denmark) for 1 h at room temperature, and after three washes with PBS, each slide was incubated for 3 min in 2% 3,3′‐diaminobenzidine tetrahydrochloride 50 mM Tris‐buffer (pH 7.6) containing 0.3% hydrogen peroxidase as a chromogen and then counterstained with hematoxylin. Renal tubules near a renal cell carcinoma and a cell block from human colorectal cancer cell line, caco‐2, were used as positive control.^{( 19 )}

First, the proportion of CD133‐positive area in the entire tumor was calculated. CD133 overexpression was defined as a CD133‐positive area that occupied more than 10% of the entire tumor area, and when CD133 overexpression was present the case was classified as CD133‐positive. Next, we evaluated CD133 expression in the invasive area, under the half of the tumor from the invasive front, and in the surface area, over the half of the tumor from the mucosal surface, respectively.

In CD133‐positive cases, the photographs of Ki‐67 and p53 immunostaining from three CD133‐positive and ‐negative areas with high‐power magnification were taken using a digital camera (DXM1200F; Nikon, Tokyo, Japan) mounted on an Nikon Eclipse E1000M microscope. The average of percentage in three areas from the CD133‐positive and ‐negative areas was determined, and the differences were examined.

Statistical analysis.  CD133 expression and clinicopathological features were tested for association by the χ²‐test, Fisher's exact test or Student's t‐test, as appropriate. Cumulative survival curves were drawn by the Kaplan–Meier method, and the differences between the curves were analyzed by the log–rank test. The Cox proportional hazard model was used in the multivariate analysis. All P‐values were two‐sided, and the significance level was set at P < 0.05. The statistical analysis was performed by using Stadt Flex ver. 5 for Windows software (Artec, Osaka, Japan).

Results

Clinicopathological findings.  The clinicopathological findings of the 189 patients are listed in Table 1. The median age of the patients was 62.1 years (range, 40–85 years; standard deviation, 9.7), and there were 120 males and 69 females. According to location, 40 tumors were in the right colon, 66 were in the left colon and 83 were in the rectum, and the median size of the tumors was 4.7 cm. According to UICC stage, 35 cases were stage I, 58 stage II, 64 stage III, and 32 stage IV; additionally, 23 of the 32 stage IV patients had liver metastasis, five had extrahepatic metastasis, and intraoperative cytology was positive for carcinoma in six cases.

Table 1.

Clinicopathological data

Characteristics		No. of patients (%)
n		189 (100)
Age: mean age ± SD, year		62.1 ± 9.7
Gender	Male	120 (63.5)
	Female	69 (36.5)
Location	Right	40 (21.2)
	Left	66 (34.9)
	Rectum	83 (43.9)
Tumor size: mean ± SD, cm		4.7 ± 2.8
Depth of tumor	T1	20 (10.6)
	T2	21 (11.1)
	T3	21 (28.0)
	T4	20 (10.6)
Lymph node stage	N0	109 (57.7)
	N1	63 (33.3)
	N2	11 (5.8)
	N3	6 (3.2)
Distant metastasis	M0	164 (86.8)
	M1	25 (13.2)
Liver metastasis	Negative	166 (87.8)
	Positive	23 (12.2)
UICC stage	I	35 (18.5)
	II	58 (30.7)
	III	64 (33.9)
	IV	32 (16.9)
Histology	Well/moderate	160 (94.7)
	Poor	29 (15.3)
Lymphatic invasion	Positive	82 (43.4)
	Negative	107 (56.6)
Vascular invasion	Positive	147 (77.8)
	Negative	42 (22.2)

Poor, poorly‐differentiated adenocarcinoma; SD, standard deviation; UICC, International Union Against Cancer; Well/moderate, well‐ or moderately‐differentiated adenocarcinoma.

The pathological diagnosis was well‐ or moderately‐differentiated adenocarcinoma in 160 cases, and poorly‐differentiated adenocarcinoma in 29 cases. Lymphatic invasion was present in 82 cases (43.4%), and vascular invasion was present in 147 cases (77.8%).

Immunohistochemical CD133 expression and clinicopathological features.  CD133 expression was detected in 29 of the 189 tumors (15.3%) and was seen exclusively on the cell membrane at the luminal surface of cancer gland with mainly cribriform features. Cytoplasmic expressions of cancer cells were very few and equivocal (Fig. 1). Of the 29 tumors with CD133 expression, 21 cases with luminal expression more than 10% of cancer area were assigned as positive cases. To know the distribution of CD133‐positive cells, we calculated the proportion occupied by the CD133‐positive cells in the invasive area and the surface area of the tumor, separately. However, there was no difference in proportion of CD133‐positive area between invasive and surface area (Fig. 2). Next, we compared the clinicopathological features of the CD133‐positive and ‐negative cases (Table 2). All CD133‐positive cases were diagnosed as well‐ or moderately‐differentiated adenocarcinoma, and only tumor areas exhibiting ductal structures had positive expression. Poorly‐differentiated adenocarcinoma, or so‐called “budding” components,^{( 20 )} were never positive for CD133 expression. We therefore also analyzed CD133 expression in the 160 well‐ to moderately‐differentiated adenocarcinomas, 21 of which were CD133‐positive and 139 of which were CD133‐negative (also in Table 2). Although the results showed that lymph node metastasis, distant metastasis and liver metastasis were more common in the CD133‐positive cases than in the CD133‐negative cases, the differences were not statistically significant. On the other hand, the difference of frequency of distant metastasis between the CD133‐positive and ‐negative cases in well‐ and moderately‐differentiated adenocarcinoma cases was statistically significant (P = 0.02) (Table 2). Most CD133‐positive cases were found in advanced cancer; more specifically, there was only one CD133‐positive case for both T1 and T2 tumors, and there was only one stage I case in 21 CD133‐positive cases. However, the difference in frequency of CD133‐positive cases in the patients with well‐ and moderately‐differentiated adenocarcinomas between stages I and II and stages III and IV were marginal, and not statistically significant (P = 0.05). Lymphatic and vascular invasion was also seen more frequently in CD133‐positive cases than in CD133‐negative cases, but the differences were not statistically significant. There were no significant differences in age, gender, size or location of the tumors between the CD133‐positive cases and CD133‐negative cases.

Figure 1.

Hematoxylin–eosin (a,b,d) and immunohistochemical expression of CD133 (c,e) in colorectal cancer. Arrow head is the CD133‐positive surface area, and arrow is the CD133‐positive invasive area. Note the uniform distribution of CD133‐positive area between the invasive area and the surface area (c,e). In high magnification, luminal expression of CD133 in the cancer gland with cribriform features were seen. Cytoplasmic expression was not defined (c,e).

Figure 2.

Difference of proportion occupied by CD133‐positive area between the invasive area and the surface area in 21 CD133‐positive cases. There was no difference in the proportion of CD133 expression between the invasive area and the surface area.

Table 2.

CD133 expression in UICC stage I–IV patients

		CD133‐positive in all cases	CD133‐negative in all cases	P‐value	CD133‐negative in well/mod cases	P‐value
n		21	168		139
Age: mean age ± SD, year		61.0 ± 9.0	62.3 ± 9.9	0.54	62.6 ± 9.8	0.41
Gender	Male : female	11:10	109:59	0.26	88:51	0.34
Tumor size: Mean ± SD, cm		4.7 ± 1.9	4.7 ± 2.8	0.82	4.6 ± 2.1	0.90
Location	Right	5	35	NS	25	NS
	Left	8	58		50
	Rectum	8	75		64
Histology	Well/mod	21	139	0.03
	Poor	0	29
Depth of tumor	T1	1	19	T1 + T2 vs T3 + T4	17	T1 + T2 vs T3 + T4
	T2	1	20	P = 0.15	18	P = 0.11
	T3	8	45		40
	T4	11	84		64
Lymph node metastasis	Negative	9	91	0.33	84	0.13
	Positive	12	77		55
Distant metastasis	Negative	16	147	0.16	128	0.02
	Positive	5	21		11
Liver metastasis	Negative	17	149	0.31	128	0.10
	Positive	4	19		11
UICC stage	I	1	34	Stage I + II vs III + IV	32	Stage I + II vs III + IV
	II	7	58	P = 0.14	52	P = 0.05
	III	8	49		41
	IV	5	27		14
Lymphatic invasion	Positive	11	72	0.41	49	0.13
	Negative	10	96		90
Vascular invasion	Positive	19	128	0.14	100	0.07
	Negative	2	40		39

Poor, poorly‐differentiated adenocarcinoma; SD, standard deviation; UICC, International Union Against Cancer; Well/moderate, well‐ or moderately‐differentiated adenocarcinoma.

To examine the difference of characteristics between CD133‐positive and ‐negative cancer cells, we performed immunostaining of Ki‐67 and p53 in 21 CD133‐positive cases. However, the proportion of nuclear Ki‐67 in CD133‐positive and ‐negative areas were 11.7 ± 4.9% and 11.7 ± 3.2%, respectively, and the proportions of nuclear p53 expression in CD133‐positive and ‐negative areas were 15.8 ± 18.8% and 14.4 ± 15.9%, respectively, in CD133‐positive cases. The differences between the two areas were not statistically significant (Table 3).

Table 3.

Ki‐67 and p53 expression in CD133 positive and negative area

	CD133‐positive area	CD133‐negative area	P‐value
Ki‐67 expression (%)	11.7 ± 4.9	11.7 ± 3.2	0.89
p53 expression (%)	15.8 ± 18.8	14.4 ± 15.9	0.79

Immunohistochemical CD133 expression and overall survival and recurrent‐free survival.  All 189 patients were followed up for survival to assess CD133 expression as a prognostic factor. The median follow‐up period was 2024 days. Of the 189 patients, 54 died of their cancer and 10 died of other causes. Of the 155 curative resection patients, 32 were diagnosed with recurrence during the follow‐up period. There were three cases of local recurrence, 13 of liver metastasis, nine of lung metastasis, three of synchronous liver and lung metastasis, one of synchronous local recurrence and liver metastasis, and three of peritoneal dissemination and distant lymph node metastasis other than regional lymph nodes.

Follow up for overall survival in all 189 patients showed no statistical difference between the CD133‐positive cases and CD133‐negative cases (P = 0.10, data not shown). Next, to avoid the statistical confounding by CD133‐negative poorly‐differentiated adenocarcinoma, the 160 patients with well‐ and moderately‐different adenocarcinomas were analyzed. The results showed that the CD133‐positive patients had worse prognosis than CD133‐negative patients in the 160 patients with well‐ and moderately‐differentiated adenocarcinoma (P = 0.03) (Fig. 3a). Among the other clinicopathological characteristics assessed in the patients with well‐ and moderately‐differentiated adenocarcinoma, tumor size, depth of tumor, lymph node status, distant metastasis, liver metastasis, UICC stage, lymphatic invasion and vascular invasion were significantly associated with patient survival in the univariate log–rank test. In the multivariate analysis, however, UICC stage was the only independent factor associated with patient survival (Table 4). Accordingly, the relationship between CD133 expression and recurrence were examined in 140 well‐ and differentiated‐adenocarcinoma patients who underwent curative resection. Recurrence was diagnosed in 27 of the 124 CD133‐negative cases (21.8%), and in five of the 16 CD133‐positive cases (31.3%), and the difference was not statistically significant. Recurrence‐free survival curves were not different between CD133‐positive and ‐negative patients (Fig. 3b).

Figure 3.

(a) Overall survival curves of CD133‐positive and ‐negative patients in 160 well‐ and moderately‐differentiated adenocarcinoma patients. CD133‐positive cases showed significantly shorter survival than CD133‐negative cases by log–rank test (P = 0.03). (b) Recurrence‐free survival curves of CD133‐positive and ‐negative patients in 140 curatively resected well‐ and moderately‐differentiated adenocarcinoma patients. The difference between CD133‐positive cases and ‐negative cases was not statistically significant by log–rank test (P = 0.47).

Table 4.

Univariate and multivariate Cox analyses for survival of patients with colorectal cancer

		Univariate analysis		Multivariate analysis
		n	P‐value	HR	CI (95%)	P‐value
n		160
Age, years	≤62	86	0.62	1.04	0.53–2.01	0.92
	>62	74		1
Gender	Male	99	0.82	1
	Female	61		0.95	0.48–1.86	0.88
Tumor size, cm	≤4.7	89	0.04	1.58	0.79–3.13	0.19
	>4.7	71		1
Location	Colon	88	0.48	1
	Rectum	72		1.11	0.57–2.17	0.76
Depth of tumor	T1 + T2	37	>0.01	ND
	T3 + T4	123
Lymph node metastasis	Negative	93	>0.01	ND
	Positive	67
Distant metastasis	Negative	144	>0.01	ND
	Positive	16
Liver metastasis	Negative	145	>0.01	ND
	Positive	15
UICC stage	I + II	92	>0.01	1
	III + IV	68		5.05	2.11–12.09	>0.01
Lymphatic invasion	Positive	60	0.02	1.40	0.70–2.80	0.34
	Negative	100		1
Vascular invasion	Positive	119	0.01	2.59	0.78–8.57	0.12
	Negative	41		1
CD133 expression	Positive	21	0.03	1.74	0.78–3.89	0.18
	Negative	139		1

CI, confidence interval; HR, hazard ratio; ND, not done.

Discussion

CD133 is the epitope of a glycosylated form of human prominin‐1, which is a member of the pentaspan transmembrane glycoprotein family,^{( 21 )} and expression of CD133 has been reported to be localized in normal neonatal and adult precursor cells as well as in C‐IC.^{( 22} , ²³ , ^{24 )} The results of this study revealed a low incidence of CD133‐positive cases in a large series of colorectal cancer patients, and no CD133 expression was detected in poorly‐differentiated carcinoma. This finding may mean that the populations of CD133‐positive cells in poorly‐differentiated adenocarcinoma are too small to detect immunohistochemically or that CD133 expression varies with the histological subtype. Bier et al. observed the CD133‐positive cells with stem cell‐like properties in primary glioblastomas, but did not detect any in secondary glioblastomas derived from gliomas.^{( 17 )} Similarly, colorectal cancer may also include heterogeneous subtypes with and without CD133‐positive cells with stem cell‐like properties. Alternatively, the pathogenesis of poorly‐differentiated adenocarcinoma might not depend on CD133‐positive cell population. Therefore, although immunohistochemical staining does not allow demonstration of self‐renewal or the tumorigenic potential of CD133‐positive cells, our findings suggested the need to further analyze associations between expression of C‐IC markers and clinicopathological features.

Our detailed morphological study of surface area and invasive area revealed equal distribution of CD133‐positive cells throughout the tumors. We did not detect any CD133‐positive cells in 10 cases of colorectal adenoma (data not shown). Detailed morphological, biochemical and physiological studies have clearly shown the presence of stem cells near the base of the normal crypts.^{( 25 )} On the other hand, the morphogenesis of colorectal adenoma and cancer remain elusive.^{( 26} , ^{27 )} Our finding of equal distribution of CD133‐positive cells between surface area and invasive area may depend on the uniform distribution of the C‐IC in the cancer microenvironment. On the other hand, stem cell‐like properties have recently been reported in CD133‐negative cells, as well.^{( 28} , ^{29 )} Therefore, the CD133‐positive tumor cells may not represent the entire cancer‐initiating population.

Another interest is whether CD133‐positive cells can be a prognostic factor. This report is the first of a study that immunohistochemically analyzed CD133 expression and assessed survival in a large series of human tumors. Although there were only 21 (11.1%) CD133‐positive cases in our series, and further investigation in a larger series of CD133‐positive cases may be necessary, CD133‐positive cases of well‐ and moderately‐differentiated adenocarcinoma had a worse outcome than the CD133‐negative cases. However, CD133 expression was not an independent risk factor associated with patient survival in multivariate analysis. Although CD133‐positive cases showed higher incidence of distant metastasis, there was no association with vascular invasion. Moreover, CD133 was not a risk factor for recurrence. These contradicting results seem not to provide enough evidence of the malignant potential of CD133‐positive cells. The poorer patient survival seems to be consistent with the results in a xenograft mouse model in which a larger CD133‐positive fraction of glioma cells decreased tumor latency and increased tumor growth and vascularity.^{( 30 )} Although biological function of CD133 was not known, chemoresistance of CD133‐positive cells may also be associated with poorer patient survival of CD133‐positive patients.^{( 31 )} Larger series of CD133‐positive cases may reveal a more accurate association between CD133 expression and clinicopathological features. On the other hand, CD133 expression was seen more frequently in advanced cancers. Also, Ki‐67 expression was not different between CD133‐positive and ‐negative areas. Therefore, as one of possibility, CD133‐positive cells may only be increased during tumor progression. CD133‐positive cells themselves may not have biological ability associated with malignant potential in colorectal cancer. A similar phenomenon was shown in the repeated implantation model of glioblastoma.^{( 28 )}

In conclusion, CD133 expression was seen in only a small number of well‐ and moderately‐differentiated adenocarcinomas of the large intestine. Although the CD133‐positive cases had a poorer outcome than the CD133‐negative cases, CD133 expression was not an independent factor associated with patient survival or a risk factor to estimate patient recurrence. Further immunohistochemical study of C‐IC markers and comparison of the results with clinicopathological features are necessary.