In this study, we investigated the expression of CD133 in colorectal cancer using an immunohistochemical approach. As the localisation of CD133+ colorectal cancer cells has not yet been characterised in situ
, we had to verify that our approach truly detects the CD133 antigen. Thus, we used three different antibodies that were specific on the protein level and showed comparable staining patterns on serial tumour sections. In addition, we demonstrate that CD133+ cell groups were CK20− and EpCAM+, which is in accordance with previous studies that characterised the phenotype of CD133+ colon cancer cells using tumour cell suspensions (Dalerba et al, 2007
; Ricci-Vitiani et al, 2007
). Moreover, we found the CD133 antigen at the luminal surface of epithelial tumour glands with shedding into the lumina, while no other staining pattern was observed. This complies with previous results of CD133 detection in embryonal tissue and on the apical surface of the cultured colon cancer cell line Caco-2 (Corbeil et al, 2000
). Taken together, the specificity of our results is supported by previous studies in colorectal cancer. However, staining results, which differ from the positivity of the luminal surface in colorectal cancer, have recently been reported in pancreatic cancer, where a cytoplasmic expression of CD133 was shown (Hermann et al, 2007
; Maeda et al, 2008
). Such organ-specific patterns of CD133 expression still have to be clarified in further studies.
The CD133+ subpopulation of colon cancer cells was recently demonstrated to be highly enriched in tumour-initiating Co-CSCs, whereas in contrast, CD133− tumour cells did not have tumour-initiating capabilities (O'Brien et al, 2007
; Ricci-Vitiani et al, 2007
). CD133 is therefore currently one of the best markers for the detection of Co-CSCs. As Co-CSCs are hypothesised to be central for tumour relapse and progression (Brabletz et al, 2005
; Ricci-Vitiani et al, 2008
), we investigated CD133 expression in a collection of colorectal adenocarcinomas that was highly stratified towards patient outcome. Expectedly, CD133 expression was strongly associated with patient survival and proofed to be an independent prognostic marker.
Another prognostic marker that has previously been described for colon cancer is tumour ‘budding'; that is, small clusters of undifferentiated cells ahead of the invasive front (Hase et al, 1993
; Ueno et al, 2002
). We therefore expected to find CD133+ budding tumour cells in colorectal cancer. Although CD133 expression was pronounced in tumour glands close to the invasive margin, CD133 expression was restricted to glandular differentiated cells and budding tumour cells were generally CD133−. As both makers correlate with patient survival, it might be speculated whether budding tumour cells contain a population of Co-CSCs that are not characterised by CD133 expression.
Although CD133 is currently one of the best markers to characterise Co-CSCs, by far not every CD133+ cell is a Co-CSC. It was estimated that 1 in 262 CD133+ colon cancer cells actually has tumour-initiating capabilities (O'Brien et al, 2007
). We demonstrate that in some tumours, several cells or even whole tumour glands are CD133+, whereas the actual Co-CSCs among them cannot be further confined using this marker. Thus, although the immunostaining for CD133 might correlate with the amount of Co-CSCs and can be used to estimate their impact on prognosis, CD133 appears to be inappropriate for the morphological characterisation of single Co-CSCs and their niche.
Taken together, we provide further evidence that the CD133+ tumour cell population in colorectal cancer is specifically, although not exclusively, important for colon cancer progression, and we present an approach that can be applied for further characterisation of these cells in situ.