Background

Recent strategies to activate and engage the immune system have led to significant breakthroughs in cancer therapy. The recent success of checkpoint blockade therapy, targeting the immune-inhibitory proteins PD-1 and/or CTLA-4, has initiated a paradigm shift in the clinical and preclinical investigation and use of immunotherapy for solid tumors. However, despite the impressive clinical efficacy of the most recent iterations of these therapies, there are still significant issues that remain. Stem-like tumor cells, also known as cancer stem cells (CSCs), are present in numerous hematological and solid malignancies [1, 2]. However, there is considerable debate regarding how to properly characterize these cells, whether they exist as a discrete subpopulation or are a plastic phenotype under epigenetic control, and a variety of correlative marker systems have been suggested for various tissue types. Classically, a very small fraction of CSCs possesses the ability to form aggressively growing, heterogeneous, tumors in immunodeficient mice or in long-term in vitro outgrowth assays. Depending on the tissue of origin, the cell surface markers CD133, CD24/CD44, as well as high expression of aldehyde dehydrogenase (ALDH) delineates these cells from their more differentiated and rapidly dividing progeny. Additionally, the use of extensively cultured commercial cell lines to study these cells does not accurately represent the heterogeneity found within primary human tumors, where the epigenetic influence of certain microenvironmental factors, such as hypoxia and/or inflammation exists [3]. To combat this discrepancy, conclusions drawn from tumor cell lines are being increasingly supported by data using human primary tumor tissues. We, and others, observed that despite reduction in overall tumor size with conventional therapies, such as radiotherapy, chemotherapy, and small molecule inhibitors, residual stem-like tumor cells remain, which have the ability to repopulate the tumor mass and seed metastases [4–7]. An immune approach to target this population of cells is likely to fill an unmet need, since unlike chemotherapy, targeted small molecule therapy, and radiotherapy; immune cells do not need tumor cell targets to be actively cycling in order to mediate their effects. We postulated that the use of immunotherapy after such regimens are applied may allow for greater CSC targeting, particularly when used after cytoreductive therapies.

Main text

Our laboratory has explored the use of natural killer (NK) cells in the recognition and targeting of CSCs in several human solid cancers. NK cells are terminally differentiated innate lymphoid cells with the ability to spontaneously kill virally-infected or malignant cells without prior immunization in an MHC-unrestricted manner. As an adoptive immunotherapy, NK cells have primarily been used against hematologic cancers as results with solid tissue tumors have been disappointing [8–10]. The goal of our studies was to determine if NK cells would selectively identify and lyse stem-like tumor cells and to understand how this potential recognition occurs. NK cells have been long known for their unique ability to reject allogeneic hematopoietic stem cells, but not solid tissue allografts. Previous data from our laboratory demonstrated that NK cells reject embryonic stem cells in an NKG2D-dependent manner and thus, we sought to apply this same rationale to evaluate their potential to target CSCs [11]. We have demonstrated that certain small molecule drugs and radiation therapy enrich for populations of tumor cells that harbor stem-like qualities and may possibly sensitize these CSCs to NK cell attack [4, 6].

In our initial evaluation, we screened several human breast, brain, pancreas, and sarcoma tumor cell lines for their susceptibility to cytokine-activated NK cell-mediated cytotoxicity [12]. Using flow cytometry, we observed that activated allogeneic NK cells significantly killed stem-like tumor cells expressing high amounts of the CSC-associated proteins ALDH, CD133, CD24, or CD44. Importantly, our cell line data was mirrored when we evaluated a panel of primary tumor cells derived from freshly excised patient samples. These primary tissue samples were tested within 48 h after surgical resection and were comprised of unsorted cells from multiple subtypes of sarcoma as well as multiple pancreatic adenocarcinoma samples. It is important to note that while NK cells demonstrated the capacity to kill both the CSC and non-CSC populations, the greatest decreases in the overall frequency of cells before and after NK killing occurred within the CSCs. In order to address the possibility of patient/donor MHC mismatching contributing to the CSC killing we observed, we also compared autologous versus allogeneic NK cell activity. We observed similar efficacy in the ability of activated NK cells to lyse stem-like targets in vitro although since these cells were highly cytokine-activated first, it does not rule out potential differences with lower activation. Other groups have noted that certain populations of stem-like tumor cells express various ligands for NK cell activating receptors [13]. We observed that CSCs from multiple solid tumors expressed ligands for the NK cell activating receptor NKG2D, and that blocking NKG2D using Fc-chimeric proteins abrogated the NK killing in vitro. Lastly, in order to evaluate the translational potential of the effects we observed in vitro, we implanted NSG mice with human tumor xenografts and evaluated the effects of adoptively transferred NK cells to eliminate stem-like cells in vivo. Paralleling our in vitro results, we observed that activated allogeneic NK cells significantly reduced the frequencies of CSCs, reduced the colony forming capacity of treated tumor cells ex vivo, and delayed the growth of orthotopic pancreatic xenograft tumors. Importantly, we have also recently observed that combining NK cell therapy with prior local tumor irradiation markedly enhances these effects, indicating that for optimal efficacy that elimination of the non-CSC populations needs to occur [4].