Studies related to the cancer stem cell hypothesis are challenging because of the imperfect tools to identify cell populations of interest and controversy on the usefulness of established cancer cell lines. We previously found CD133 to not be selective for a tumor-propagating or radioresistant population in a near-diploid, microsatellite-instable colorectal carcinoma (CRC) cell line. Because of discrepant literature data, we herein systematically analyzed the behavior of microsatellite-stable cell line subpopulations reflecting the more frequent carcinogenesis pathway in spontaneous CRC. CD133⁺ and CD133(-/low) populations were isolated by fluorescence-activated cell sorting and further processed. HT29 and SW620 cells were studied in detail in monolayer and/or spheroid culture assays and upon subcutaneous injection in NMRI (nu/nu) mice using a limiting dilution approach. CD133(-/low) HT29 cells showed a significantly lower clonogenic survival and reduced spheroid formation capacity than their CD133⁺ counterparts. However, the cell populations neither differed in growth kinetics and response to treatment in vitro nor in tumor formation capacity when injecting as low as 10 cells. CD133(-/low) HT29 cells rapidly re-expressed CD133 protein in vitro and in vivo as shown by flow cytometry and/or western blot analyses, and they also showed a particular survival benefit under tissue normoxic conditions. In contrast, CD133 protein in the CD133⁺ population was quite stable throughout culturing. The observation of CD133 re-expression and lack of difference in tumor take rate of subpopulations was confirmed in SW620 cells. Here, we found cell density to affect CD133 re-expression in the CD133(-)-sorted population. And even SW480 cells, classified as a CD133⁻ cell line, presented some CD133 protein on their surface upon in vivo engraftment. We conclude that (i) CD133 protein expression shows high plasticity in CRC cell lines, and (ii) in vitro CD133 status on the cell surface neither determines tumorigenic potential nor CD133 profile in vivo.