Colon cancer stem cells (CSC) can be identified with AC133, an antibody that detects an epitope on CD133. However, recent evidence suggests that expression of CD133 is not restricted to CSCs, but is also expressed on differentiated tumor cells. Intriguingly, we observed that detection of the AC133 epitope on the cell surface decreased upon differentiation of CSC in a manner that correlated with loss of clonogenicity. However, this event did not coincide with a change in CD133 promoter activity, mRNA, splice variant, protein expression, or even cell surface expression of CD133. In contrast, we noted that with CSC differentiation, a change occured in CD133 glycosylation. Thus, AC133 may detect a glycosylated epitope, or differential glycosylation may cause CD133 to be retained inside the cell. We found that AC133 could effectively detect CD133 glycosylation mutants or bacterially expressed unglycosylated CD133. Moreover, cell surface biotinylation experiments revealed that differentially glycosylated CD133 could be detected on the membrane of differentiated tumor cells. Taken together, our results argue that CD133 is a cell surface molecule that is expressed on both CSC and differentiated tumor cells, but is probably differentially folded as a result of differential glycosylation to mask specific epitopes. In summary, we conclude that AC133 can be used to detect cancer stem cells, but that results from the use of this antibody should be interpreted with caution. Cancer Res; 70(2); 719–29
Cancer stem cells (CSC) are thought to be responsible for tumor growth. These CSCs can drive tumor growth and possess multilineage differentiation potential (1, 2), allowing them to reform the original malignancy on transplantation in mice. In several cancers, CSCs have been identified using one or multiple markers, like CD24 (3, 4), CD29 (4), CD44 (5–7), CD133 (8–12), ALDH1 (13, 14), or Hoechst exclusion (15–17). However, the appropriateness of these markers is an ongoing discussion.
The pentaspan membrane protein CD133, originally identified as a marker for CD34+ hematopoietic stem and progenitor cells (18, 19), has been used for CSC identification in several types of cancer, such as glioblastoma (11, 20), melanoma (8), liver cancer (21), osteosarcoma (22), and colon cancer (9, 10, 12). Several monoclonal antibodies have been developed against CD133. The most commonly used are AC133 (CD133/1) and 293C/AC141 (CD133/2), which are reported to recognize distinct epitopes (23). AC133 is frequently used to isolate CSCs and suggested to recognize a glycosylated epitope on CD133 (18), which contains eight putative N-linked glycosylation sites. However, the use of CD133 as a marker for identifying and isolating CSCs is controversial because its expression pattern is debated. Several groups have shown that AC133+, but not AC133−, cells sorted from primary colon carcinomas can form tumors in immunodeficient mice that recapitulate the morphology of the original tumor (9, 10, 12). In addition, expression of mouse CD133 (prominin-1) has been shown to mark stem cells in the small intestine (24, 25). In contrast, CD133 mRNA expression was found in cells other than the (cancer) stem cell fraction (25–28). For example, insertion of a LacZ reporter directly after the ATG start site of mouse CD133 (prominin-1), or between exons 3 and 8, showed CD133 expression throughout the mouse colon (25, 28), including differentiated goblet cells (25), and also in mouse colon adenocarcinoma (28). Moreover, Shmelkov and colleagues (28) described that CD133 does not uniquely mark CSCs from human colon carcinomas but is also expressed on differentiated tumor cells.
The above-mentioned data clearly show that there is a contradiction between AC133 as a CSC marker and the broad CD133 mRNA and protein expression found in colon (cancer). The functionality of CD133 as a CSC marker is therefore unclear and should be clarified. In this study, we addressed this question by analyzing CD133 promoter activity and mRNA, splice variant (Sv), protein, and AC133 epitope expression between stem cells and differentiated cells in normal colon epithelium and colon cancer. In addition, we addressed the possibility that posttranslational modification, such as glycosylation, might play a role in recognition of CD133 by the AC133 antibody. We show that CD133 is expressed on the cell surface of CSCs and differentiated tumor cells but is differentially glycosylated. Nevertheless, AC133 does not recognize a glycosylated epitope or a specific Sv. Instead, we show that the failure to recognize CD133 is due to the inaccessibility of the epitope, which we hypothesize to depend on the tertiary structure of CD133 on differentiated cells.
Materials and Methods
Tissue collection, CSC isolation, culture, and xenografting
Samples of human colon carcinomas were obtained according to standard medical ethical procedures of the Academic Medical Center or the University of Palermo. CSC cultures were derived and cultured as described previously (12). C001 and C002 were derived from the primary cancer, whereas LMIV and LMV were obtained from liver metastases. All lines were from different patients. Cells were differentiated on adherent plates (Corning) by withdrawal of epidermal growth factor and fibroblast growth factor and addition of non–heat-inactivated 2% FCS. Primary colon tumor pieces of 1 mm3 were implanted s.c. in nonobese diabetic/severe combined immunodeficient mice. Before tumors reached 1 cm3, the mice were sacrificed and tumors were digested as described in ref. 12.
The antibodies used were FITC-anti-ESA (Biomeda), anti-AC133, anti-CD133/2 (AC141), anti-CD133 (W6B3C1), anti-CD133/2 (293C; Miltenyi), anti–extracellular signal-regulated kinase (ERK; a kind gift from B. Burgering, Physiological Chemistry, University of Utrecht, Utrecht, the Netherlands), anti-actin (Santa Cruz Biotechnology), anti- Epcam (Abcam), IRDye 680 anti-mouse and IRDye 800 anti-rabbit (LI-COR Biosciences), horseradish peroxidase–labeled anti-mouse IgG (Southern Biotechnologies), anti–cytokeratin-20 and anti–intestinal alkaline phosphatase (Genetex), anti–Muc-2 (Abcam), mouse IgG1 (DakoCytomation), rhodamine red–conjugated anti-mouse antibodies (Molecular Probes), Alexa546-anti-mouse IgG and Alexa546-anti-rabbit IgG (Invitrogen).
Cells were stained with directly labeled antibodies for 30 min at 4°C in PBS containing 1% bovine serum albumin (BSA) and 0.02% sodium azide. Intracellular fluorescence-activated cell sorting (FACS) staining was done with Cytofix/cytoperm (BD Biosciences) according to the manufacturer's protocol. Analysis was done on a FACSCalibur, and sorting on a FACSAria.
Laser-aided microdissection of villus and crypt regions
Sections of snap-frozen colonic tissue of familial adenomatous polyposis (FAP) patients were stained with hematoxylin and digitally scanned with a Veritas Microdissection System (Molecular Devices Corporation). Epithelial cells from the base and the upper part of normal crypts distant from adenomatous tissue were cut and collected into Capsure Macro LCM caps (Molecular Devices).
Protein isolation and immunoblotting
Immuoblotting was done as described in ref. 29. For deglycosylation, lysates were treated overnight with 500 units/μL of peptide-N-glycosidase F (PNGase F) at 37°C.
RNA isolation and PCR analysis of CD133 mRNA and CD133 promoter activity
Total RNA was isolated by TRizol extraction (Invitrogen). RNA quality and quantity were assessed using Nanodrop technologies, and cDNA was prepared with reverse transcriptase III (Invitrogen). The following intron-spanning primers were used: CD133, 5′-TTCTATGCTGTGTCCTGGGGC-3′ and 5′-TTGTTGGTGCAAGCTCTTCAAGGT-′3; actin, 5′-ATGGAAGAAGAGATCGCCGC-3′ and 5′-TCGTAGATGGGCACCGTGTG-3′; exon 3, 5-ATATCTTTCTCTATGTGGTACAGCC-3 and 5′-AATAAACAGCAGCCCCAGGA-3′ (30); exons 25–29, 5′-AAAACTGGCTAAGTACTATCG-3′ and 5′-AGACCCAGAAACTACCAAAA-3′. The CD133 promoter activity was determined as described in ref. 31.
Cell surface protein isolation
C002 cells were cultured under cancer stem cell conditions and plated adherently (Corning) for the last 20 h or differentiated for 10 d. Cell surface protein isolation was done with the Pierce Cell Surface Protein Isolation Kit, except that columns were washed twice with radioimmunoprecipitation assay (RIPA) buffer (Thermoscientific) and the biotinylated protein was eluted with RIPA buffer containing 50 mmol/L DTT.
Bacterial expression of CD133
DNA encoding CD133 from start to end of the NH2-terminal part and the first and second extracellular loops was acquired by PCR performed with the following reverse primers: NH2 terminus, 5′-TTACTACTCGAGCTACCCTGCTTCATAGTAGAC-3; half first loop, 5′-TTACTACTCGAGCTAGTTCATGTTCTCCAACGCC-3′; first loop, 5′-TTACTACTCGAGCTAGTATGAATCATACTCTTC-3′; second loop, 5′-ATCTTGGATCCGGAGGGCAGCCTTCATCCACAGATG-3′, and with the forward primer, 5′-TTACTACTCGAGCTAATTCAAGGGGTCGATAAT-3′. PCR product was cloned into a pET-28 vector with BamH1 and Xho1 restriction sites and transformed into BL21(DE3) bacteria (Stratagene). Bacterial expression of CD133 was done according to the manufacturer's protocol.
Immunofluorescence staining of colon cancer specimens
Human colon carcinoma samples were obtained from 15 patients. Immunofluorescence was done on fresh frozen tissues post-fixed in 2% paraformaldehyde for 20 min at 37°C (Shmelkov's procedure) or on fresh tissue specimens embedded in liquid optimum cutting temperature compound, solidified gradually in liquid nitrogen vapor, and fixed in acetone. Sections of 5 μm were blocked with TBS containing 3% AB human serum for 10 min. Primary antibody was incubated overnight at 4°C (1:5) and secondary antibody for 1 h at room temperature (1:300 in 1% BSA/PBS). Counterstaining was done using Toto-3.
Promoter regulation of CD133 in normal human colon
Human CD133 is a highly complex gene that contains five promoters, differing in the 5′ untranslated region (ref. 31; Fig. 1A), and seven Svs, which differ in coding exons and generate distinct protein isoforms (Fig. 2A). In previous studies, LacZ was inserted into a part of CD133 that was unaffected by differential splicing, thereby generating a setting that allows for detection of all CD133 promoter activities and Svs (25, 28). However, this approach is not adequate to detect distinct CD133 promoter activity or isoform expression within separate regions of colon epithelium and carcinoma. Promoter regulation exists for human CD133 (31) and could result in differential Sv expression (32–34). Moreover, several mouse CD133 Svs reportedly cannot reach the cell surface (32). We therefore hypothesized that differential Sv expression influences CD133 recognition, and we analyzed promoter activity and Sv expression to gain more insight into the regulation and detection of CD133.
We detected CD133 mRNA expression in the upper and lower parts of the crypt in normal human colon derived from FAP patients (Fig. 1B), in agreement with earlier findings that CD133 mRNA was expressed in all epithelial cells of mouse colon (25, 28).
All five CD133 promoter activities could be detected using nested PCR (ref. 31; Supplementary Fig. S1A–C), and all promoters, except P4, showed activity in colon cancer cell lines. Sequencing of the products revealed all reported as well as two additional alternatively spliced noncoding exons for promoter 3 (sequences to be submitted), which are consistent with the splice acceptor-donor site rules (31), and additional alternative Svs for exons transcribed from promoter 5 (Supplementary Fig. S1D).
Next, we analyzed the differential CD133 promoter usage on microdissected crypt surface and base and showed that promoters 1, 2, and 3 were active in the human normal colon, whereas promoter 4 and 5 activities were not detected (Fig. 1B). Besides a small variation in splicing of exons transcribed from promoter 3, we observed no major changes in promoter usage. Although this nested PCR only allows for semiquantitative measurements of promoter activity, we conclude that differential CD133 promoter activity is not a dominant effect in differential CD133 protein expression in normal colon.
CD133 promoter regulation in cancer stem cells
Because CD133 is applied as a marker for CSCs in colon cancer, we analyzed CD133 promoter activity in colon CSCs versus differentiated cancer cells (DCC). Therefore, CSCs derived from primary colon carcinomas were cultured as spheroids or adherent under differentiation-inducing conditions to generate DCCs (4, 12). CSCs had high expression of surface CD133 as detected with the AC133 antibody, whereas differentiation decreased this detection significantly (Fig. 1C). Previously, we have shown that these high AC133-expressing spheroid cultures retain the capacity to induce colon adenocarcinomas on xenotransplantation, whereas DCCs lose this capacity (12). In addition, differentiation markers such as cytokeratin-20, Muc-2, and intestinal alkaline phosphatase are only expressed in DCCs (Supplementary Fig. S2A). More importantly, high AC133 expression selected for the clonogenic population as shown by xenotransplantation (10, 12, 35) as well as clonogenic growth (ref. 4; Supplementary Fig. S2B), confirming that AC133 expression can identify CSCs in these cultures.
CSCs as well as DCCs showed promoter 1, 2, and 3 activities, whereas promoter 4 was not active in either population (Fig. 1C). Although promoter 5 was active in colon CSCs, its activity was very low and not changed upon differentiation. Combined, we conclude that CD133 promoter activity does not differ substantially between CSCs and DCCs in vitro and thus cannot explain differential recognition of AC133 during differentiation. Surprisingly, CD133 mRNA expression was not decreased upon in vitro differentiation (Fig. 1C), which indicates that protein detection by FACS, which is decreased by almost 10-fold in DCCs, and mRNA expression do not correlate.
In vitro culture of CSCs is a powerful method, but could influence expression of many genes including CD133. To exclude culture artifacts, we studied CD133 promoter and mRNA expression of a colon cancer xenograft, grown by direct implantation of a primary human colon carcinoma. The tumor-derived cells were FACS sorted into Epcam+AC133+ and Epcam+AC133− fractions, while nonepithelial cells were discarded (Fig. 1D). Importantly, promoter activities were not significantly different in ex vivo sorted AC133+ and AC133− fractions (Fig. 1D). Moreover, despite a more than 50-fold difference in AC133 reactivity, CD133 mRNA levels were only slightly lower in the AC133− fraction (Fig. 1D). This clearly shows that AC133 staining is not at all correlated to CD133 mRNA expression.
CD133 Svs in cancer stem cells
Although CD133 mRNA levels are fairly comparable in CSCs and DCCs, the PCR used does not discriminate between alternative Svs in the coding sequence. In addition, regulation of differential CD133 splicing by distinct promoters has not been shown yet. The seven reported Svs of CD133 (Fig. 2A; Table 1) generate different CD133 protein isoforms and could potentially encode proteins that lack the AC133 epitope. The Svs differ mainly in the presence of exon 5 (30) or in splicing of exon 26a, 26b, and/or 27. Analysis of the different Svs using PCRs directed at the spliced regions (Fig. 2A) indicated that several CD133 Svs could be detected in both the base and the upper surface of the crypt. However, it was also quite evident that Sv2 was by far the most prominent form present. Importantly, differential expression of Svs between the base and the upper part of the crypt was not found (Fig. 2B). Similarly, CSCs as well as DCCs almost only displayed expression of one Sv that lacked exon 5 and contained exons 26b and 27, consistent with the expression of Sv2 (Fig. 2C). This was confirmed by sequencing, which revealed only one major sequence (data not shown). This is not unique to in vitro cultured cancer cells, but also freshly isolated AC133− and AC133+ cancer cells from xenografts mainly expressed Sv2 at the mRNA level (Fig. 2D). Although the function of the different CD133 Svs remains enigmatic, it is clear that several of these are expressed at the mRNA level in colon cancer samples. Nevertheless, the vast majority of the expressed CD133 mRNAs encode for Sv2. More importantly, no change is observed during CSC differentiation.
We therefore conclude that the presence of the AC133 epitope can define colon CSCs, as determined by in vitro clonogenicity and in vivo tumor growth. However, the promoter activity and mRNA and Sv expression did not vary between CSCs and DCCs, indicating that expression of the AC133 epitope is likely regulated at the translational or posttranslational level.
CD133 protein is not downregulated upon differentiation, but its glycosylation is changed
Downregulation of the AC133 epitope during differentiation of CSCs could be caused by a reduction in translation and thus total CD133 protein. To monitor this, the effect of differentiation on the recognition of CD133 by other antibodies was studied. The antibodies AC141 and 293C (CD133/2) have been reported to detect overlapping epitopes distinct from the AC133 epitope, which was confirmed by cross-blocking studies (Supplementary Fig. S3). Nevertheless, they display similar downregulation as AC133 on CSC differentiation (Fig. 3A). For W6B3C1, an antibody normally used for immunoblotting, we also observed epitope downregulation on CSC differentiation (Fig. 3A). In contrast, analyzing CD133 expression by immunoblotting revealed no decrease of total CD133 protein during differentiation (Fig. 3B), confirming our mRNA data and indicating that CD133 protein levels are not modified upon CSC differentiation but that either CD133 is retained inside the cell or specific epitopes are shielded. Interestingly, we observed an enhanced mobility of CD133 derived from DCCs on immunoblot, which is likely caused by a change in posttranslational modification and which could determine protein folding or trafficking.
CD133 is a highly glycosylated protein (Supplementary Fig. S4A; refs. 18, 36), and several groups have suggested that the AC133 epitope is a glycosylated epitope (18, 37, 38), which could be lost upon differentiation. We therefore analyzed whether the observed change in mobility was due to glycosylation differences. Lysates of CSCs and DCCs were treated with PNGase F to remove N-linked glycans, which resulted in a mobility shift equaling around 30 kDa, confirming that CD133 is heavily glycosylated (Fig. 3C). Intriguingly, we also observed that deglycosylated CD133 from CSCs and DCCs comigrate at the exact same height, whereas the non–PNGase-F-treated samples displayed a clear migration shift upon differentiation (Fig. 3C). Therefore, we conclude that the change in molecular weight of CD133 induced by differentiation reflects a change in glycosylation.
AC133 does not recognize a glycosylated epitope
The previous experiments indicated that altered glycosylation of the CD133 protein during differentiation coincided with decreased recognition of at least two different epitopes of CD133 (CD133/1 and CD133/2), hinting that these epitopes are glycosylated, as has been suggested for AC133 (18, 37, 38). To study this possibility and to largely map the epitopes, we bacterially expressed CD133 because eukaryotic proteins are not glycosylated in bacteria. DNAs encoding different C-terminal truncated forms of His-tagged CD133 were cloned into a pET vector and expressed in bacteria (Fig. 4A). Immunoblotting with an antibody against the His tag confirmed the expression and predicted size of the recombinant proteins. However, all CD133 antibodies tested only recognized the recombinant protein that contained the second extracellular loop (Fig. 4A). First of all, these findings show that AC133 does not recognize a glycosylated epitope, in contrast to previous suggestions (18, 37, 38). Second, because the epitopes were mapped to the second extracellular loop (Fig. 4A), which is present in all known CD133 Svs, these data confirm that differential splicing is not the cause of differential AC133 recognition.
Intracellular retention is not the cause of differential epitope expression
Previously, CD133 was shown to reside mainly in microvilli of epithelial cells (18, 39), which seemed to depend on cholesterol (40). Although our observations indicate that glycosylation is not directly influencing epitope expression, it could orchestrate differential trafficking of protein, causing intracellular retention or localization to subdomains of the cell membrane. To test this hypothesis, we performed cell surface biotinylation and subsequent isolation of proteins present on the cell membrane. Using this approach, we observed that CSCs and DCCs display clear expression of CD133 on the cell surface (Fig. 4B). Moreover, these data showed that differentially glycosylated forms of CD133 can reach the cell surface (Fig. 4B) and therefore make intracellular retention an unlikely explanation for epitope loss. In agreement, intracellular FACS analysis confirmed that AC133 and 293C did not show enhanced detection of CD133 when DCCs were permeabilized, whereas detection in CSCs increased (Fig. 4C). Combined, this indicates that CD133 is not retained intracellularly in DCCs but is transported to the cell surface, in spite of differential glycosylation.
AC133 epitope is masked on differentiation
Our results indicate that differentiation decreased AC133 detection by FACS, coinciding with a change in glycosylation, but that this is not due to the loss of total membrane CD133 protein or of a glycosylated epitope. Differential glycosylation therefore seems to result in a distinct overall tertiary structure or localization of CD133 on the membrane that disallows the antibodies to access their epitopes. We hypothesized that AC133 epitope recognition could be enhanced by unfolding the CD133 protein chemically. In agreement, CD133 derived from DCCs or bacteria is recognized by AC133 on immunoblot, suggesting that this is indeed feasible. To directly analyze this possibility, we compared the immunofluorescence protocol used by Shmelkov and colleagues, who found CD133 staining on all epithelial tumor cells (28), to our procedure that showed isolated CD133+ cells within colon carcinomas (4, 12). Treatment of the same colon carcinoma specimen with these two different procedures revealed that only a small percentage of tumor cells stain positive for AC133 when samples are treated mildly, but that all cells showed clear AC133 positivity when the samples were treated more harshly (Fig. 4D). Although these data do not formally prove our model, they lend further support to the idea that CD133 mRNA and protein are not decreased when CSCs differentiate and lose their stemness, but that the epitope for AC133 is lost due to shielding.
Several recent publications have addressed the expression pattern of the CSC marker CD133 in different tissues and discovered that CD133 could also be detected in more differentiated cell types, questioning its function as a marker for (cancer) stem cells (25–28, 36). In contrast, the AC133 epitope has been convincingly used to sort CSCs from primary colon carcinomas (9, 10, 12). In this article, we addressed these conflicting data by studying possible regulation mechanisms for expression of the AC133 epitope. We showed that the AC133 and 293C/AC141 epitopes were downregulated during CSC differentiation, whereas total CD133 protein expression remained equal. The loss of these epitopes was not induced by a switch in activation of the different CD133 promoters or by differential Sv expression that could influence antibody detection. Previously, two mouse models using CD133 promoter–driven LacZ showed that CD133 mRNA is expressed throughout the colon (25, 28), supporting our data. However, mouse CD133-LacZ expression marked the stem cell and progenitor population in the small intestine (24, 25), indicating that CD133 regulation differs between the small intestine and colon in mouse. Additionally, mouse CD133 was shown to be a CSC marker for tumors induced in the small intestine (25) and could not be found in differentiated tumor cells. Although this points to CD133 as a direct CSC marker in mouse small intestinal tumors, care should be taken when translating these data to human because human colorectal cancer tumors obviously do not emanate from the small intestinal cells and mouse colon does not seem to display the same expression pattern.
Next to unchanged CD133 mRNA and protein expression, our data reveal that differential splicing could not explain the disappearance of the epitope. On one hand, differential splicing was, if at all present, not very prominently observed upon differentiation of CSCs. On the other hand, epitope mapping indicated that the AC133 antibody recognizes extracellular loop 2, which does not contain alternatively spliced exons (Fig. 5). Moreover, our data indicate that the CD133 protein was also not retained intracellularly upon differentiation but remained expressed on the membrane. In contrast, differentiation was associated with reduced glycosylation of CD133, marked by a small loss in molecular weight. Interestingly, a microarray study comparing spheroid cultured glioblastoma cells with serum-cultured differentiated glioblastoma cells revealed that expression of glycosylation enzymes differed significantly between these two populations (41), confirming that differentiation of CSCs is associated with a change in glycosylation enzymes. Alternatively, the change in mobility could be caused by a change in sialylation of CD133, which has been reported for malignant colon tissue (42, 43).
Although decreased glycosylation correlated with reduced detection of the AC133 and 293C/AC141 epitopes after differentiation, we found that this was not due to loss of glycosylated epitopes by immunoblotting of CD133 from CSCs and DCCs or by using completely unglycosylated bacterially expressed CD133 protein. In addition, mutation of each of the eight glycosylation sites of CD133 did not prevent recognition of the protein on an immunoblot (Supplementary Fig. S4B). Combined with the fact that the CD133 protein (a) did not decrease upon differentiation, (b) is still present on the cell surface, and (c) is also not detected when DCCs are permeabilized, this indicates that the epitope is not deleted upon differentiation but is more likely shielded. Epitope masking could be caused by differential folding of the protein or, for instance, be the result of a CD133 binding partner that masks the AC133 epitope. Although our data would suggest that epitope masking is a result of differential glycosylation, alternative explanations are possible. For instance, cholesterol plays a role in localization of CD133 on the membrane because mild depletion of cholesterol induced redistribution of CD133 from the microvilli to the entire apical plasma membrane (40). Differential localization could also determine protein folding and, therefore, recognition of CD133 by AC133. In addition, Taïeb and colleagues showed that a glycosylation-independent epitope on the N-terminal part of CD133 was also lost during differentiation of Caco-2 colon cancer cells. The N-terminal epitope was found to be masked by ganglioside binding during differentiation (44). Although this seems to be similar to our observations, the AC133 epitope resides in the second extracellular loop and therefore is not related to the N-terminus. However, disruption of lipid rafts with β-methyl-cyclodextrin recovered the N-terminal epitope, but not the AC133 epitope (44), which confirms that CD133 is also still present on differentiated Caco-2 cells but cannot be detected with AC133.
Importantly, our observations can explain the contradictions found in literature between AC133 as a CSC marker and the broad expression of CD133 in differentiated cell types. For example, Caco-2 cells, expressing high levels of AC133, showed reduced staining with the AC133 antibody when cells were differentiated (45). However, a homemade antiserum against CD133 (αhE2) detected CD133 on differentiated Caco-2 cells. Moreover, this antibody detected CD133 at the apical membrane in a range of primary tissues, whereas AC133 did not. The hypothesis that CD133 is present on DCCs but that the AC133 epitope is lost after differentiation is further substantiated by the observation that CD133 is detected on the cell surface of DCCs using cell surface biotinylation and can be unmasked by immunofluorescence using differential fixing procedures.
Although our data provide an explanation on why AC133 can be used as a bona fide CSC marker when using FACS (or magnetic activated cell sorting)-based isolation, they also call for caution when using this marker. For instance, our data help explain the vastly different, sometimes uniform, stainings of CD133 seen in tissues (26–28). Because CD133 mRNA seems to be constant, its usage as a CSC marker when analyzing microarray is clearly flawed. Multiple studies have now reported on the clinical significance of CD133 mRNA based on such analyses. Although these associations are undeniable, we believe they do not reflect a difference in CSC incidence but point to a different feature of these subsets of tumors. In this light, it is important to note that the CD133 promoter region can also be inactivated by hypermethylation (46) and that knockdown of CD133 is apparently not affecting CSC features (47), potentially explaining why Shmelkov and colleagues (28) observed that CD133− colon cancer cells derived from liver metastasis can initiate tumor growth.
To conclude, we have shown that the colon CSC marker CD133 should be used with caution because it is widely expressed throughout the colon and colon carcinomas when analyzed for mRNA expression, by immunoblotting, or by immunohistochemical staining. However, our data also show that the differential accessibility of the AC133 epitope makes CD133 a bona fide CSC marker when using the right conditions.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant Support: VICI NWO grant (J.P. Medema), the Academic Medical Center Graduate School (K. Kemper), the EMBO Long-term Fellowship (M.R. Sprick), and Associazione Italiana per la Ricerca sul Cancro (G. Stassi).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received May 20, 2009.
- Revision received November 13, 2009.
- Accepted November 16, 2009.
- ©2010 American Association for Cancer Research.