In recent years there has been an increasing focus on the cancer stem cell (CSC) theory. More and more CSCs in solid tumors have been prospectively identified based on expression of cell surface markers 12345 or Hoechst 33342 dye efflux 6789. According to CSC hypothesis, a CSC is a cancer-initiating cell that possesses the ability to recreate a new tumor when transplanted into a suitable microenvironment. However, to date, there have been no studies that convincingly put this proposition into practice. CSCs identified from several solid tumors are only a small subpopulation of tumorigenic cells that most probably represents a heterogeneous mixture of CSCs and their differentiated progeny. More primitive, undifferentiated CSCs (also called true CSCs) remain hidden among them.

Since a single stem cell has been demonstrated to have the ability to reconstitute an entirely functional organ in a mouse model 101112, a single CSC with stem-like properties should also generate a tumor when injected into a congenetic animal model. It is still unclear whether true CSCs have unique cell surface antigens. To address this basic issue in CSC hypothesis, we explored the possibility of exploiting the feature of drug resistance in cancer cells for screening CSCs on the basis of serum-free suspension culture for isolation and in vitro propagation of tumorigenic cancer cells. By applying suspension culture combined with anticancer regimens, we obtained single breast cancer stem cell (BrCSC), which was able to generate new tumor when injected into congenetic mice.

Since BrCSCs have been identified as CD44⁺CD24^- subpopulation from breast cancer patient samples 1, much progress in research on this field has been achieved. However, the isolation and purification of BrCSCs is still hindered by their scarcity in quantity and continuous differentiation. To the development of targeted therapy against breast cancer, the key is to solve the problem from the identification of true BrCSCs. At present, identification of CSCs is mainly based on cell surface antigens sorted by fluorescence-activated cell sorting or magnetic immunosorting. As the exact molecular signature of CSCs is not known, it seems somewhat blind and impractical in order to find CSCs by screening large numbers of known markers. Hoechst 33342 dye efflux, another sorting approach in using functional properties of CSCs, are being utilized to isolate side population cells enriched in CSCs. However, cytotoxicity of Hoechst-33342 may restrict its further applications 131415. The above two methods both use flow cytometry, and cell sorting may cause the trauma of stained cells.

The CSC theory holds that CSCs reside in a largely quiescent state with regard to cell-cycle activity 1617 and express high levels of ATP-binding cassette (ABC) transporters such as the multidrug resistance gene 18, so that they can be more resistant to chemotherapeutics than tumour cells with limited proliferative potential. Furthermore, it has been demonstrated that BrCSCs can be propagated and passaged in vitro under serum-free suspension culture and sustain the undifferentiated state 19. Based on the above theory and research results, we propose a novel strategy-suspension culture combined with anticancer regimens – to screen BrCSCs. At first, we followed the approach of Ponti et al. 19 to expand and enrich tumorigenic breast cancer cells under suspension culture conditions. Cytotoxic regimens were then used to eradicate rapidly dividing tumor cells but spare BrCSCs. Finally, tumor xenograft experiments were performed to evaluate tumorigenicity of the sorted cells. Recently, Yu et al. 20 have successfully taken advantage of mammospheres in suspension culture and chemotherapy to enrich for self-renewing breast cancer cells, which further confirms our viewpoint.

Tumorspheres could sustain long-term suspension culture and serial passage in vitro. In our laboratory, they have been through 110 successive passages. During the period of serum-free culture, we observed that cells were not able to proliferate in serum-free medium without EGF and/or bFGF. Tumorspheres only formed in the presence of EGF and/or bFGF. Similar results have also been in culture of mammospheres reported by Dontu et al.21. We speculate it is EGF and/or bFGF that induce cell proliferation under suspension culture conditions. As described by Ponti et al. 19 and Yu et al.20, tumorspheres were undifferentiated cells because they failed to express mature markers associated to myoepithelial cells (cytokeratin 14 and alpha-smooth muscle actin) and luminal/ductal cells (cytokeratin 14 and MUC-1). However, it is the relative degree of undifferentiation. In this experiment we found that the sorted CD44⁺CD24^- subset could generate CD44^-CD24^-, CD44^-CD24⁺, and CD44⁺CD24⁺ subpopulation besides CD44⁺CD24^- phenotype in serum-free culture (data not shown). Even if under suspension culture, tumorsphere cells propagate always followed by differentiation. Before chemotherapy, more than 70% of cells in serum-free media possessed CD44⁺CD24^- phenotype. One week after chemotherapy, almost all cells expressed CD44⁺CD24^-.(Note early flow cytometry analysis after anticancer application may affect test results as a large number of necrotic cells and debris). The findings showed that chemotherapy drugs killed non-CD44⁺CD24^- cells and only retained a very small number of CD44⁺CD24^- cells in 0.35 PPC of TE (less than 2%). After about 2 weeks of quiescence and recovery, a single surviving CD44⁺CD24^-cell would propagate as tumorspheres and differentiate into a heterogenous population. These cells exhibted the ability of self-renewal and multipotent differentiation in virto suspension culture.

In order to obtain a true, single BrCSC, we have proposed in vitro suspension culture growth pattern to describe the properties of BrCSCs from suspension culture in vitro 22. Tumorspheres consist of BrCSCs and transit amplifying cells/progenitors, which may be compared to the structure of an atomic model. A few BrCSCs locate inside the core of tumorsphere (viewed as an atomic nucleus) and a large amount of relatively differentiated progenies surround the outer layer (viewed as electrons). BrCSCs simultaneously perpetuate themselves through symmetric division and generate differentiated progeny through asymmetric division 2324. Self-renewal of BrCSCs is limit and tumorpsphere formation is mainly due to expansion of their progeny (rapidly dividing cells, also termed transit amplifying cells). A possible mechanism may involve that BrCSCs have a relatively slow turnover rate to ensure uniform genetic background and rapid expansion of transit amplifying cells prevent neighboring BrCSCs proliferating via unknown regulation. As gentle mechanical aspiration can disperse tumorspheres, surface tension may play a key role in tumorpsphere formation rather than cell-cell interactions. From this hypothesis model, we can speculate that a certain size tumorsphere should contain the largest proportion of BrCSCs. To obtain a large number of BrCSCs, it is necessary to dissociate the optimal size of tumorspheres into single cells. In this way, single BrCSC is exposed again and another self-renewal process will continue. By measuring CD44⁺CD24^- expression of tumorsphere cells during serial passages, we found that it had the highest proportion of CD44⁺CD24^- cells at passage 10. Cells of passage 10 were therefore the most desirable research subjects. As shown in Fig. 2, the proportion of CD44⁺CD24^-subpopulation exhibited a rough S-shaped curve. This suggests that CD44⁺CD24^- cells has achieved stable proportion after some passages of suspension culture and BrCSCs may reside in a relatively slow proliferation period or a quiescent state, whereas their differentiated progeny may sustain expansion in number. During this time, most BrCSCs can survive the attack of chemotherapy drugs. Pacilitaxel and epirubicin, which mainly target rapidly dividing tumor cells, are widely used against breast cancer as a combined chemotherapy protocol. It is a critical issue to determine the optimal concentrations of TE. The criteria is required to comply with the requirements of eradicating the vast majority of dividing cells and retaining the minimum number of dormant cells. Less than 2% of the cells treated with 0.35 PPC of TE survived the double sorting of suspension culture and anticancer regimens. In theory, these sorted cells were much closer to the source of primitive BrCSCs. In contrast, the other concertrations of chemotherapy drugs would kill all tumor cells or would not stop cell proliferation in our study.

Tumorigenicity test are currently been regarded as the gold standard for proving the existence of CSCs. The key to a successful transplantation is to keep cell viability. Before single-cell inoculation, each cell was evaluated by trypan blue staining to exclude dead cells. At the same time matrigel and estrogen were both applied. Matrigel can provide a more favourable microenvironment niche for cell colonization and growth. Exogenous estrogen supplementation can promote tumor formation. In this study, in order to avoid transplantation rejection in the use of human breast cancer cells and immunocompromised mice, we adopted the established breast cancer cell line TM40D from BALB/C mice. Although congenic transplantations were used, we applied VP-16 before transplantation in order to reduce immune and inflammation response and guarantee the viability of cells after inoculation under the consideration of transplantation of a small number of tumor cells at initial design stage, especially for a single-cell transplantation. In fact, the transplantation results showed there were no distinct difference between the non-immunosupressed and immunosuppressed group in this study. It suggests that immune function may have no apparent effect on syngeneic tumor transplantation of inbred mice. The immunochemical analysis of the resulting tumors driven by a single sorted cell showed a heterogenous tumor in vivo, with the similar features compared with primary tumors from TM40D cells. Taken in vitro and in vivo experimental studies together, we conclude that the sorted cells are true BrCSCs.

As a functional sorting method, suspension culture combined with anticancer regimens provides us a novel approach to enrich and purify CSCs. At the pre-experiment, we adopted CD29 and CD24 as the presumed markers of BrCSCs according to CD29^hiCD24⁺ from mouse mammary epithelium stem cell identified by Shackleton et al. 12. We found no tumors were formed by injecting 10⁴ CD29^hiCD24⁺ cells or CD29^hiCD24^- into BALB/C mice. The result reminded us that there may be unknown or no specific cell surface antigens highly expressed in BrCSCs. So we bagen to enrich tumorgenic breast cancer cells by serum-free suspension culture and purify BrCSCs by chemotherapy drugs. The results showed BrCSCs in mice derived from CD44⁺CD24^- subpopulation and had the same markers of puative BrCSCs in human. By applying this sorting stragegy, we could expand and purify CSCs under the condition of unknown putative markers in vitro. Similarly, we could also obtain CSCs much closer to the source under the condition of known putative markers. It could not be made only by flow sorting of the original cultures.

In this study, we identified BrCSCs by suspension culture combined with anticancer regimens and confirmed that a single BrCSC can form breast cancer in mice. We showed for the first time, to our knowledge, that a single tumor cell can repopulate a new tumor in mice with normal immune function. The sorting method of suspension culture combined with anticancer regimens for BrCSCs may be applied to identify other CSCs in solid tumors.

CSCs – cancer stem cell, BrCSCs – breast cancer stem cells, EGF – epidermal growth factor, bFGF – basic fibroblast growth factor, PE – phycoerythrin, CY5 – cyanin 5, FITC – fluorescein isothiocyanate, PI – propidium iodide, PPC – peak plasma concentration, FACS – fluorescence-activated cell sorting.

The authors declare that they have no competing interests.

Most experiments and the manuscript were made by H–ZL. T–BY performed immunocytochemistry staining. Z–YW was involved in designing all experiments and revised the manuscript. All authors read and approved the final manuscript.