We read with interest the article entitled “Expression profiling of GIST: CD133 is associated with KIT exon 11 mutations, gastric location and poor prognosis” by Arne et al.,1 who used a tissue macro arrays (TMAs)-based approach to demonstrate that the expression of CD133 in gastrointestinal stromal tumor (GIST) segregates with c-kit exon 11 deletion and poor prognosis, and suggested that over-expressing CD133 cells might be GIST cancer stem cells (CSCs).1
The results of our own flow cytometry (FC) analysis of the expression of CD133 in naïve and imatinib-treated GISTs were partially different and led us to different conclusions.2 Like Arne et al., we observed a trend toward higher CD133 expression in naïve GISTs carrying c-kit exon 11 mutations in aggregate in a gastric location than in those with an intestinal location but, because of the small number of c-kit deleted cases (n = 4), we can neither confirm nor deny the association between high CD133 levels and the c-kit exon 11 deletions.
However, unlike Arne et al. (who found CD133 expression in only 28% of their cases), we observed CD133 expression in all our (untreated and treated) samples, even though with some variability in intensity and the percentage of positive cells. In relation to the naïve cases, these apparently conflicting findings are probably due to our different methodological approach (FC rather than TMA), as it is well known that FC using fresh material allows the better preservation of CD133 epitopes. Furthermore, although interchangeable, the two antibodies (AC133 used by Arne et al. and 293C3 used by us) were mainly validated by FC and Western blots of hematopoietic stem cells,3–5 whereas the relatively poorer efficiency of CD133 antibody when using formalin-fixed material is indirectly confirmed by the immunophenotype read-out of the TMA assay, which shows the expected immunoprofile for KIT, DOG1 and CD34, but a definitely low level of immunolabeling for CD133.
In comparison with the untreated cases, we observed lower CD133 immunolabeling in all nine of our imatinib-treated cases carrying primary c-kit mutations (c-kit exon 11) or primary and secondary mutations (c-kit exon 13 and exon 17). This finding led us to draw different conclusions concerning the significance of CD133 expression as a surrogate marker of putative GISTs CSCs, because the retained CD133 expression in untreated and imatinib-treated GISTs suggest a lineage marker similar to KIT rather than true CSCs.
Translating the CSC hypothesis into the GIST paradigm, secondary mutations responsible for imatinib resistance should occur in putative imatinib-selected self-renewing GIST CSCS that should highly express CD133.6 In line with a lineage marker rather than a CSCs marker, we found decreased CD133 expression in our imatinib-treated GISTs and no subpopulation of cells expressing high levels of CD133—and the same was true of KIT.2 It is also worth noting that our findings are also in line with those of a recent study showing that the expression of CD133 in GIST is positively regulated by ETV1, a transcription factor that cooperates with KIT in GIST tumorigenesis, and is expected to be coexpressed with CD133 and KIT in all KIT-addicted GISTs.7
Finally, in an elegant study of postnatal murine stomach,8 Bardsley et al. found that the interstitial cells of Cajal (ICC) progenitors (and ICC-derived GISTs) have low levels of KIT expression and do not depend on KIT signaling for their survival, and (in the light of the recently proposed ETV1 model) are, therefore, unable to stabilize ETV1 and induce CD133 transcription and expression.
On the basis of these preclinical findings, it seems that we cannot rule out the provocative hypothesis that, if GIST CSCs exist, they may be CD133 null.