Breast cancer stem-like cells (BCSC) are crucial for metastasis but the underlying mechanisms remain elusive. Here, we report that tumor-infiltrating natural killer (NK) cells failed to limit metastasis and were not associated with improved therapeutic outcome of BCSC-rich breast cancer. Primary BCSCs were resistant to cytotoxicity mediated by autologous/allogeneic NK cells due to reduced expression of MICA and MICB, two ligands for the stimulatory NK cell receptor NKG2D. Furthermore, the downregulation of MICA/MICB in BCSCs was mediated by aberrantly expressed oncogenic miR20a, which promoted the resistance of BCSC to NK cell cytotoxicity and resultant lung metastasis. The breast cancer cell differentiation–inducing agent, all-trans retinoic acid, restored the miR20a–MICA/MICB axis and sensitized BCSC to NK cell–mediated killing, thereby reducing immune escape–associated BCSC metastasis. Together, our findings reveal a novel mechanism for immune escape of human BCSC and identify the miR20a–MICA/MICB signaling axis as a therapeutic target to limit metastatic breast cancer. Cancer Res; 74(20); 5746–57. ©2014 AACR.

The escape of cancer cells from immune surveillance has been considered as a prerequisite for metastasis (1), the major cause of death in breast cancer. Cancer cells are highly heterogeneous and a small fraction of cancer cells with stem cell properties, known as tumor-initiating cells, or cancer stem cells (CSC), may be responsible for tumorigenesis (2), angiogenesis (3) and metastasis (4). Recently, selection by adaptive immunity has been shown to generate or enrich cancer cells with stemness (5–9), suggesting the presence of intimate interaction between CSCs and the skewed adaptive immunity in patients with cancer. However, the role of innate immunity in affecting CSCs and their metastasis remains unclear.

Cancer immunoediting, defined by phases of elimination, equilibrium, and escape, consists of both tumor-limiting and -promoting effects (10). Selection by adaptive and innate immunity components, together with clonal evolution of malignant cells, contributes to the generation of cancer cells with better survival advantages and functional heterogeneity (10). Recent studies using mice deficient in adaptive immunity demonstrate that cancer immunoediting may occur in the absence of adaptive immune cells through IFNγ-producing natural killer (NK) cells (11). Thus, we hypothesize that improper immunoediting by NK cells may result in the failure of cancer cells with stem cell property to respond to anticancer immunity and, thus, promote cancer metastasis.

As innate immune effectors, NK cells are one of the major infiltrating immune cells in breast cancer and are endowed with the capability of recognizing nascent transformed cells to prevent tumorigenesis and subsequent metastasis (12–15). Breast cancer stem-like cells (BCSC) are resistant to therapies and are highly metastatic (16–18). However, it remains unclear whether BCSCs succumb to NK cell cytotoxicity. In this study, we attempt to address this issue and report that BCSCs escape from NK cell cytotoxicity and metastasize through miR20a-mediated downregulation of MICA and MICB, two ligands for the NK cell–activating receptor NKG2D.

Breast cancer specimens and immunohistochemistry

Breast cancer tissues were obtained from 591 consecutive consented patients who underwent surgical resection at the Breast Surgery Center at Southwest Hospital from 2006 to 2007, with approval from the Institutional Ethics Committee. All specimens were formalin-fixed and paraffin-embedded in the Institute of Pathology at Southwest Hospital. The median age of patients at diagnosis was 45 years (range, 23–87 years). The patients received radical mastectomy, or modified radical mastectomy, or breast-conserving surgery. The axillary lymph nodes were routinely examined for metastasis. Tumor size, location, distal metastasis, and final tumor–node–metastasis stage were determined. The histologic diagnosis was made according to the WHO standard. The expression of the breast cancer marker aldehyde dehydrogenase 1 (ALDH1) and the NK cell marker CD56 was examined by immunohistochemistry (IHC; Supplementary Materials and Methods).

Cell sorting by FACS

Primary breast cancer cells at 90% confluence were suspended in ALDEFLUOR assay buffer containing an ALDH substrate and incubated for 45 minutes at 37°C (ALDEFLUOR assay kit; Stem Cell Technologies; refs. 18, 19). Negative control cells were incubated with an ALDH inhibitor, diethylamino-benzaldehyde (DEAB). The cells were then suspended in assay buffer for sorting ALDH+ and ALDH cells on a BDAria II sorter (BD Biosciences). ALDH+ cells were maintained as mammospheres for further analysis.

Culture of peripheral blood NK cells and NK cell cytotoxicity assay

Peripheral blood NK cells (pNK cells) from patients with breast cancer and healthy donors were isolated for direct analysis or for further expansion (Supplementary Materials and Methods). NK cell cytotoxicity was measured by a 51chromium (51Cr) release assay. The target cells (1 × 106) were labeled with 250 μCi Na251CrO4 for 2 hours at 37°C, seeded into 96-well round bottom plates at 5,000 per well, and cocultured with pNK cells (effector cells) at indicated effector:target (E/T) ratios. K562 cells were served as a positive control for killing. After incubation for 4 hours at 37°C, cell supernatants were collected, and spontaneous and maximal releases of 51Cr were measured. The percentage of 51Cr release (cytotoxicity) was calculated as 100 × [(experimental release − spontaneous release)/(maximum release − spontaneous release)]. In blocking experiments, pNK cells were pretreated with 20 μg/mL anti–NKG2D-neutralizing mAb (R&D Systems) for 20 minutes at room temperature before the assay.

Cell death of targets induced by NK cell cytotoxicity was also examined by flow cytometry (Supplementary Materials and Methods).

IFNγ quantification by ELISA

Briefly, target cells were seeded in laminin-coated 96-well plates and cultured for 24 hours. pNK cells (5 × 105/mL) were incubated with the targets (2.5 × 105/mL) in a final volume of 200 μL of fresh NK cell medium (Supplementary Materials and Methods) for 24 hours (E/T = 2:1). Blockade of NKG2D receptor before incubation was performed using an anti–NKG2D-neutralizing mAb (R&D Systems). To test whether cancer cells released IFNγ, we included target cells alone as control. After centrifugation, the supernatant was collected to determine the level of IFNγ using a human IFNγ ELISA Kit (Invitrogen). Because the levels of IFNγ secretion varied among donors, we showed the representative data using NK cells derived from one donor for presentation, and all of the assays were repeated using NK cells from at least three different donors.

Clearance by NK cells and lung metastasis

The elimination of target cells by NK cells in vivo and immunoevasion-associated metastasis were assessed in a short-term lung clearance assay and an experimental lung metastasis model in NK cell (mNK)–depleted NOD/SCID mice transferred with human pNK cells (Supplementary Materials and Methods).

Statistical analysis

All experiments were performed at least in triplicates and representative data were shown. The correlation between the number of CD56+ NK cells or ALDH1 positivity and clinicopathologic parameters was assessed by the nonparametric Wilcoxon rank-sum test, the χ2 test, or the Fisher exact test as appropriate. Overall survival (OS) and metastasis-free survival (MFS) rates were plotted as Kaplan–Meier curves and analyzed using the log-rank test. The Cox proportional hazard regression model was used to determine the influence of various parameters on patient survival in univariate and multivariable analysis. The Student t test or ANOVA was used when appropriate. All data were analyzed using SPSS 13.0 statistical software. Data were shown as the means ± SD. A P value of <0.05 was considered statistically significant.

Failure of tumor-infiltrating NK cells to prevent the metastasis of BCSC-rich human breast cancers

We first examined the presence of CD56+ NK cells and BCSCs in breast cancer specimens from 591 patients using ALDH1 as a stem cell marker (18–20). On the basis of the IHC staining, 212 cases of breast cancers were ALDH1high and 379 cases were ALDH1low (Supplementary Table S1). High level of infiltrating CD56+ cells was detected in 33.0% of cancers. Of note, we frequently detected codistribution of CD56+ cells in the areas with either non-BCSCs (ALDH1low) or BCSCs (ALDH1high) cancer cells (Fig. 1A). In ALDH1low cancers, the level of infiltrating CD56+ cells was lower in cancer tissues with ≥4 lymph node metastasis (LNM) than those with <4 LNM (Fig. 1B). Although close localization of infiltrating CD56+ cells with BCSCs was observed (Fig. 1C), there was no significant difference in the number of tumor-infiltrating CD56+ cells between cases with ≥4 LNM and those with <4 LNM in ALDH1high tumors (Fig. 1D).

Figure 1.

Infiltrating NK cells fail to limit metastasis of human breast cancers with high level of BCSCs. A, in ALDH1low cancers (top), low or high levels of infiltrating CD56+ cells (bottom) were detected by IHC staining on serial sections, demonstrating codistribution of NK cells with non-BCSCs. B, average numbers of infiltrating CD56+ cells in ALDH1low cancers with ≥4 or <4 LNM. C, in ALDH1high cancers (top), low or high levels of infiltrating CD56+ cells (bottom) were also revealed, showing codistribution of NK cells with BCSCs. D, average numbers of infiltrating CD56+ cells in ALDH1high cancers with ≥4 or <4 LNM. LN, lymph node. E and F, Kaplan–Meier plot of OS and MFS of the patients with ALDH1low (E) or ALDH1high cancers (F), stratified by the levels of infiltrating CD56+ cells. Mean ± SD (B and D); bar, 100 μm.

Figure 1.

Infiltrating NK cells fail to limit metastasis of human breast cancers with high level of BCSCs. A, in ALDH1low cancers (top), low or high levels of infiltrating CD56+ cells (bottom) were detected by IHC staining on serial sections, demonstrating codistribution of NK cells with non-BCSCs. B, average numbers of infiltrating CD56+ cells in ALDH1low cancers with ≥4 or <4 LNM. C, in ALDH1high cancers (top), low or high levels of infiltrating CD56+ cells (bottom) were also revealed, showing codistribution of NK cells with BCSCs. D, average numbers of infiltrating CD56+ cells in ALDH1high cancers with ≥4 or <4 LNM. LN, lymph node. E and F, Kaplan–Meier plot of OS and MFS of the patients with ALDH1low (E) or ALDH1high cancers (F), stratified by the levels of infiltrating CD56+ cells. Mean ± SD (B and D); bar, 100 μm.

Close modal

We then analyzed the correlation of infiltrating CD56+ cells with patient survival in ALDH1low and ALDH1high subgroups. Significantly longer OS and MFS were found in the patients with higher level of infiltrating CD56+ cells than those with lower level of infiltrating CD56+ cells in ALDH1low, but not ALDH1high cancers (Fig. 1E and F). Moreover, increased level of infiltrating CD56+ cells in tumors was a favorable prognostic factor for OS and MFS in ALDH1low, but not the ALDH1high patient subgroup (Supplementary Table S2). Furthermore, increased level of infiltrating CD56+ cells was able to independently predict fewer distal metastases in the ALDH1low tumors (Supplementary Table S3). However, in the ALDH1high subgroup, no significant difference in the relative risk to develop metastasis was observed in patients with CD56high tumors as compared with those with CD56low tumors. These results indicate that high level of infiltrating NK cells in breast cancer fails to prevent ALDH1high BCSCs from metastasis in patients.

BCSCs are resistant to NK cell cytotoxicity and contribute to lung metastasis

We isolated ALDH+ cells from primary breast cancer cells and found that they exhibited potent self-renewal capacity with multiple differentiation potentials and markedly enhanced tumorigenicity in vivo (Supplementary Fig. S1), confirming the BCSC property of ALDH+ cells (18, 19). We next examined activation and degranulation of NK cells cocultured with either BCSCs or non-BCSCs (ALDH cells). Flow cytometric analysis showed reduced expression of CD107a, a surface marker for NK cell degranulation, on human pNK cells cocultured with BCSCs (Fig. 2A and Supplementary Fig. S2A). Because IFNγ released by activated NK cells is important for immunoregulation, we determined the level of IFNγ and found significantly reduced production of IFNγ by pNK cells cocultured with BCSCs (Fig. 2B and Supplementary Fig. S2B). Thus, primary BCSCs are unable to activate normal NK cells.

Figure 2.

Resistance of BCSCs to the cytotoxicity of autologous or allogeneic pNK cells and the contribution of BCSCs to lung metastasis. A, detection of CD107a on pNK cells by FACS in pNK cells cocultured with primary BCSCs or non-BCSCs (E:T = 5:1). K562 cells served as positive control. Alternatively, pNK cells were pretreated with anti-CD16 mAb (3 μg/mL) to demonstrate the specificity for the function of NK cells. pNK represents human peripheral blood NK cells that were expanded for 2 to 3 weeks. Fresh pNK represents freshly isolated NK cells. B, IFNγ released by pNK cells in the coculture supernatant (E:T = 2:1). C, lysis of primary BCSCs and non-BCSCs by autologous pNK cells as measured by 51Cr release. Data from six patients are shown (triple repeats). D, lysis of BCSCs and non-BCSCs by allogeneic pNKs from healthy donors. Anti-CD16 mAb was applied to specifically inhibit NK cell activity. E, flow cytometric analysis of target cell death cultured with pNK. The middle panel shows representative flow cytometric graph and total dead cells (%). F, the percentages of ALDH+ cells (left) or efficacy of sphere formation (right) of cancer cells that survived NK cell–mediated lysis. G, the relative survival ratio of BCSCs or non-BCSCs (DiD-labeled) in relation to HeLa cells (GFP-labeled, internal control) in a short-term lung clearance assay in NOD/SCID mice (n = 3). H, the numbers of pulmonary metastasis established by BCSCs and non-BCSCs (n = 3); mean ± SD (A–I); *, P < 0.05; **, P < 0.01; ns, no significance.

Figure 2.

Resistance of BCSCs to the cytotoxicity of autologous or allogeneic pNK cells and the contribution of BCSCs to lung metastasis. A, detection of CD107a on pNK cells by FACS in pNK cells cocultured with primary BCSCs or non-BCSCs (E:T = 5:1). K562 cells served as positive control. Alternatively, pNK cells were pretreated with anti-CD16 mAb (3 μg/mL) to demonstrate the specificity for the function of NK cells. pNK represents human peripheral blood NK cells that were expanded for 2 to 3 weeks. Fresh pNK represents freshly isolated NK cells. B, IFNγ released by pNK cells in the coculture supernatant (E:T = 2:1). C, lysis of primary BCSCs and non-BCSCs by autologous pNK cells as measured by 51Cr release. Data from six patients are shown (triple repeats). D, lysis of BCSCs and non-BCSCs by allogeneic pNKs from healthy donors. Anti-CD16 mAb was applied to specifically inhibit NK cell activity. E, flow cytometric analysis of target cell death cultured with pNK. The middle panel shows representative flow cytometric graph and total dead cells (%). F, the percentages of ALDH+ cells (left) or efficacy of sphere formation (right) of cancer cells that survived NK cell–mediated lysis. G, the relative survival ratio of BCSCs or non-BCSCs (DiD-labeled) in relation to HeLa cells (GFP-labeled, internal control) in a short-term lung clearance assay in NOD/SCID mice (n = 3). H, the numbers of pulmonary metastasis established by BCSCs and non-BCSCs (n = 3); mean ± SD (A–I); *, P < 0.05; **, P < 0.01; ns, no significance.

Close modal

To examine the capacity of NK cells to destroy BCSCs, we used pNK cells from both patients with breast cancer and healthy donors. Activated pNK cells showed a weaker lytic activity on autologous BCSCs than non-BCSCs (Fig. 2C). BCSCs were also resistant to killing by either long-term expanded or fresh pNKs from healthy donors (Fig. 2D and Supplementary Fig. S2C and S2D). Pretreatment of NK cells with an anti-CD16 mAb inhibited both degranulation and cytotoxicity by NK cells that were cocultured with BCSCs or non-BCSCs (Fig. 2A and D, Supplementary Fig. S2A, S2C, and S2D), indicating the specificity of the function of NK cells. Moreover, the levels of cell death in BCSCs incubated with either long-term expanded or freshly isolated pNK cells were reduced (Fig. 2E and Supplementary Fig. S2E). In addition, short-term exposure to pNK cells did not significantly affect cell-cycle distribution and proliferation of BCSCs (Supplementary Fig. S2F and S2G). These results suggest that the resistance of BCSCs to NK cell cytotoxicity is due to reduced cell death. To test possible immunoselection of BSCSs versus non-BCSCs by NK cells, we cultured primary cancer cells with allogeneic pNK cells and found that the residual viable cancer cells were enriched with BCSCs, as indicated by increased number of ALDH1+ cells and high efficacy of tumorsphere formation (Fig. 2F and Supplementary Fig. S2H). Thus, BCSCs are more resistant than non-BCSCs to lysis by NK cells.

To test the consequences of BCSC resistance to NK cell cytotoxicity, we used an in vivo short-term lung clearance assay of cancer cells in mice (Supplementary Fig. S3A; refs. 21, 22). Although no significant difference of cell survival rates was found between BCSCs and non-BCSCs i.v. injected into NOD/SCID mice depleted of NK (mNK) cells (mNK cell–depleted mice), less BCSCs were eliminated in the mice transferred with human pNK cells (Fig. 2G and Supplementary Fig. S3B). These results suggest that circulating BCSCs are resistant to elimination by NK cells in vivo. In a lung metastasis model (Supplementary Fig. S3C), i.v. injected BCSCs formed higher numbers of metastases in the lungs of normal mice as compared with non-BCSCs (Fig. 2H and Supplementary Fig. S3D). In contrast, there was no difference in the number of metastatic foci formed by either BCSCs or non-BCSCs in mNK cell–depleted mice. However, i.v. injected BCSCs established more metastatic foci in the lungs of mice adoptively transferred with human pNK cells. The results indicate that BCSCs are resistant to NK cell cytotoxicity in vivo, thus are prone to form lung metastasis.

Reduced expression of MICA and MICB by BCSCs contributes to the resistance to NK cell cytotoxicity and the resultant metastasis

Tumor-infiltrating immune cells release cytokines such as IFNγ to regulate leukemic stem cells (8). However, we found that the expression of IFNγ receptor α chain, an essential component of IFNγ receptor for ligand binding and trafficking, was comparable between BCSCs and non-BCSCs (Supplementary Fig. S4A). Because NK cell receptor ligands expressed by target cells determine the sensitivity to NK cell cytotoxicity, we compared the expression of inhibitory receptor ligands HLA-A/B/C, HLA-E, HLA-G by BCSCs and non-BCSCs and also found no significant difference (Supplementary Fig. S4B). We then compared the levels of activating receptor ligands (Supplementary Fig. S4C–S4E) on BCSCs and non-BCSCs, and found that the expression of both MICA and MICB, two ligands for the activating receptor NKG2D, was significantly lower on BCSCs as compared with non-BCSCs (Fig. 3A and Supplementary Fig. S4F).

Figure 3.

The expressions of NGK2D ligands MICA and MICB on BCSCs and their contribution to NK cell cytotoxicity and tumor metastasis. A, representative data of MICA and MICB expression by primary BCSCs and non-BCSCs (left). Right, quantitative data from six cases of primary cancers. B, detection of CD107a on pNK cells cocultured with BCSCs overexpressing MICA or MICB (E:T = 5:1). To abrogate MICA/B–NKG2D interaction, pNK cells were pretreated with an NKG2D-blocking mAb (20 μg/mL). C, IFNγ levels in the coculture supernatant of pNK cells with NKG2D ligands transfected with BCSCs (E:T = 2:1). Experimental controls included pNK cells alone, or nontransfected BCSCs alone, or NKp30 mAb (2 μg/mL)- or NKG2D mAb-pretreated pNK cells with BCSCs. D, the cytotoxicity of pNK cells to BCSCs (mean ± SEM). Controls included K562 cells, nontransfected or vector-transfected tumor cells, or pNK cells pretreated with NKG2D or NKp30 mAb. E, the survival of BCSCs overexpressing MICA or MICB in vivo (n = 3). F, numbers of lung metastatic nodules formed by BCSCs (n = 3). Mean ± SD (A–C, E, and F); *, P < 0.05.

Figure 3.

The expressions of NGK2D ligands MICA and MICB on BCSCs and their contribution to NK cell cytotoxicity and tumor metastasis. A, representative data of MICA and MICB expression by primary BCSCs and non-BCSCs (left). Right, quantitative data from six cases of primary cancers. B, detection of CD107a on pNK cells cocultured with BCSCs overexpressing MICA or MICB (E:T = 5:1). To abrogate MICA/B–NKG2D interaction, pNK cells were pretreated with an NKG2D-blocking mAb (20 μg/mL). C, IFNγ levels in the coculture supernatant of pNK cells with NKG2D ligands transfected with BCSCs (E:T = 2:1). Experimental controls included pNK cells alone, or nontransfected BCSCs alone, or NKp30 mAb (2 μg/mL)- or NKG2D mAb-pretreated pNK cells with BCSCs. D, the cytotoxicity of pNK cells to BCSCs (mean ± SEM). Controls included K562 cells, nontransfected or vector-transfected tumor cells, or pNK cells pretreated with NKG2D or NKp30 mAb. E, the survival of BCSCs overexpressing MICA or MICB in vivo (n = 3). F, numbers of lung metastatic nodules formed by BCSCs (n = 3). Mean ± SD (A–C, E, and F); *, P < 0.05.

Close modal

To examine whether the lower expression of MICA and MICB was responsible for the resistance of BCSCs to NK cell cytotoxicity, we transfected BCSCs with MICA and MICB (Supplementary Fig. S4G). Coculture of transfected BCSCs with pNK cells significantly increased the expression of CD107a by pNK cells, which was inhibited by pretreatment of pNK cells with an NKG2D-blocking mAb (Fig. 3B and Supplementary Fig. S4H). In addition, these cocultured pNK cells showed increased IFNγ secretion (Fig. 3C). BCSCs with increased MICA or MICB, but not nontransfected parental BCSCs or vector control, were significantly more sensitive to allogeneic pNK-mediated lysis (Fig. 3D and Supplementary Fig. S4I). Increases in IFNγ secretion and NK cell–mediated cytotoxicity were inhibited by blocking NKG2D but not NKp30 receptor on NK cells (Fig. 3C and D, Supplementary Fig. S4I). Short-term lung clearance assay showed a reduced survival of MICA- or MICB-transfected BCSCs in the lungs of mNK cell–depleted mice transferred with human pNK cells. In contrast, blocking NKG2D on NK cells promoted the survival of BCSCs (Fig. 3E and Supplementary Fig. S4J). Moreover, BCSCs transfected with NKG2D ligands formed fewer metastatic foci in the lungs of these mice. The NKG2D-blocking mAb restored the lung metastasis of MICA- or MICB-transfected BCSCs (Fig. 3F and Supplementary Fig. S4K). These results indicate that lower levels of MICA and MICB contribute to BCSC resistance to NK cell lysis, thus increasing their lung metastasis.

To further examine the relevance of MICA and MICB to tumor cell sensitivity to NK cell cytotoxicity, we used siRNA to downregulate the ligands originally expressed at higher levels on non-BCSCs (siRNA–non-BCSCs; Supplementary Fig. S5A). Coculture of pNK cells with siRNA–non-BCSCs reduced their CD107a expression (Fig. 4A and Supplementary Fig. S5B), accompanied by reduced IFNγ secretion (Fig. 4B), suggesting that cancer cells with lower levels of MICA or MICB are unable to trigger NK cell killing. These nonactivated NK cells, therefore, were less effective in lyzing siRNA–non-BCSCs, but not nontransfected non-BCSCs or scramble siRNA control (Fig. 4C). NKG2D mAb, but not isotype IgG or NKp30 mAb, cast the effect similar to MICA/MICB knockdown on the IFNγ secretion and cytotoxicity (Fig. 4B and C), suggesting an important role of MICA/MICB–NKG2D interaction in the immune surveillance of non-BCSCs by NK cells. Moreover, siRNA–non-BCSCs with downregulated NKG2D ligands showed improved survival (Fig. 4D and Supplementary Fig. S5C), and, thus, metastasized more frequently to the lungs of mNK cell–depleted mice transferred with human pNK cells (Fig. 4E and Supplementary Fig. S5D). These in vivo results demonstrate that loss of MICA or MICB results in enhanced survival and metastasis of breast cancer cells.

Figure 4.

The effect of MICA or MICB knockdown on the sensitivity of non-BCSCs to NK cell–mediated lysis and their metastatic potential. A, FACS analysis of CD107a expression in pNK cells that were incubated with non-BCSCs treated with indicated siRNAs (E:T = 5:1). B, IFNγ levels in the supernatant were determined by ELISA (E:T = 2:1). Controls included NK cells alone, nontransfected non-BCSCs, and siRNA–non-BCSCs in the presence of effector cells or pNK cells pretreated with NKG2D mAb, NKp30 mAb, or isotype mAb. C, killing efficacy of non-BCSCs by activated pNK cells (mean ± SEM). Experiment controls were identical to B. D, the survival of non-BCSCs transfected with indicated siRNAs in vivo (n = 3). E, the numbers of metastatic nodules formed by non-BCSCs (n = 3) mean ± SD (A, B, D, and E); *, P < 0.05.

Figure 4.

The effect of MICA or MICB knockdown on the sensitivity of non-BCSCs to NK cell–mediated lysis and their metastatic potential. A, FACS analysis of CD107a expression in pNK cells that were incubated with non-BCSCs treated with indicated siRNAs (E:T = 5:1). B, IFNγ levels in the supernatant were determined by ELISA (E:T = 2:1). Controls included NK cells alone, nontransfected non-BCSCs, and siRNA–non-BCSCs in the presence of effector cells or pNK cells pretreated with NKG2D mAb, NKp30 mAb, or isotype mAb. C, killing efficacy of non-BCSCs by activated pNK cells (mean ± SEM). Experiment controls were identical to B. D, the survival of non-BCSCs transfected with indicated siRNAs in vivo (n = 3). E, the numbers of metastatic nodules formed by non-BCSCs (n = 3) mean ± SD (A, B, D, and E); *, P < 0.05.

Close modal

High level of miR20a in BCSCs downregulates MICA/B and promotes metastasis

The mechanisms underlying the loss of MICA and MICB in BCSCs were explored. We failed to detect differences in the mRNA levels of either MICA or MICB between BCSCs and non-BCSCs (Supplementary Fig. S6A). However, the level of one microRNA, miR20a, was found significantly higher in BCSCs than non-BCSCs (Fig. 5A). We, therefore, hypothesized that miR20a might be involved in the downregulation of MICA and/or MICB on BCSCs (Supplementary Fig. S6B). Dual luciferase reporter assays showed that integration of MICA and MICB (Fig. 5B) 3′ untranslated region (UTR) markedly inhibited the luciferase activities. However, the inhibition was reversed when miR20a-binding sites were mutated, suggesting that endogenous miR20a binds to 3′-UTR of MICA and MICB to regulate their expression. These results were confirmed by the findings that the miR20a inhibitor (Supplementary Fig. S6C) markedly upregulated the expression of MICA and MICB on BCSCs (Fig. 5C). Thus, miR20a appears to interfere with the posttranscriptional modification of MICA and MICB.

Figure 5.

Regulation of MICA/MICB expression, the sensitivity to NK cell–mediated lysis, and metastatic potential by miR20a in BCSCs. A, the expression of miR20a in BCSCs and non-BCSCs by cells from six primary cancers was examined by qRT-PCR (U6B as a loading control; horizontal bars, means). B, luciferase activity was detected in BCSCs transfected with Mock or p-MICA 3′UTR (or mutant) or p-MICB 3′UTR (or mutant) reporters in the presence of control miRNA (ctrl-miRNA) or an miR20a inhibitor (anti-miR20a). C, FACS analysis of MICA or MICB expression by BCSCs transfected with anti-miR20a or ctrl-miRNA for 48 hours. D and E, pNK cells, with or without preincubation using NKG2D-blocking mAb, were cocultured with BCSCs transfected with anti-miR20a. CD107a expression on pNK cells was detected (D) and IFNγ levels in the supernatant were analyzed (E). F, cytotoxicity of pNK cells to CSCs (mean ± SEM). G, the short-term lung survival of BCSCs (n = 3). H, numbers of lung metastasis formed by BCSCs (n = 3); mean ± SD (B, D, E, G, and H); *, P < 0.05.

Figure 5.

Regulation of MICA/MICB expression, the sensitivity to NK cell–mediated lysis, and metastatic potential by miR20a in BCSCs. A, the expression of miR20a in BCSCs and non-BCSCs by cells from six primary cancers was examined by qRT-PCR (U6B as a loading control; horizontal bars, means). B, luciferase activity was detected in BCSCs transfected with Mock or p-MICA 3′UTR (or mutant) or p-MICB 3′UTR (or mutant) reporters in the presence of control miRNA (ctrl-miRNA) or an miR20a inhibitor (anti-miR20a). C, FACS analysis of MICA or MICB expression by BCSCs transfected with anti-miR20a or ctrl-miRNA for 48 hours. D and E, pNK cells, with or without preincubation using NKG2D-blocking mAb, were cocultured with BCSCs transfected with anti-miR20a. CD107a expression on pNK cells was detected (D) and IFNγ levels in the supernatant were analyzed (E). F, cytotoxicity of pNK cells to CSCs (mean ± SEM). G, the short-term lung survival of BCSCs (n = 3). H, numbers of lung metastasis formed by BCSCs (n = 3); mean ± SD (B, D, E, G, and H); *, P < 0.05.

Close modal

To study the contribution of increased miR20a in BCSCs to their escape from NK cell killing, the cells were transfected with an miR20a inhibitor (miRi-BCSCs). After coculture with miRi-BCSCs, pNK cells expressed a significantly higher level of CD107a (Fig. 5D and Supplementary Fig. S6D) and secreted increased amounts of IFNγ (Fig. 5E). miRi-BCSCs were also more sensitive to the cytotoxicity (Fig. 5F and Supplementary Fig. S6E) and the clearance by NK cells in vitro and in vivo (Fig. 5G and Supplementary Fig. S6F). Masking the NKG2D receptor on NK cells reversed the effect of the miR20a inhibitor. Moreover, inhibition of miR20a reduced the metastasis of BCSCs in the lungs of the mice, which was also reversed by blockade of the NKG2D receptor on pNK cells (Fig. 5H and Supplementary Fig. S6G). Although the miR20a inhibitor slightly reduced sphere formation, ALDH expression, and proliferation by BCSCs (Supplementary Fig. S6H and S6I), inhibition of miR20a alone did not significantly increase the apoptosis of the cells without the presence of NK cells (Supplementary Fig. S6J). Inhibition of miR20a alone did not significantly attenuate cell survival and metastatic capacity of BCSCs in the absence of NK cells in vivo (Fig. 5G and H and Supplementary Fig. S6F and S6G). Therefore, miR20a downregulates MICA and MICB and increases the capacity of BCSCs to metastasize by evading NK cell clearance.

Targeting the miR20a–MICA/B axis by all-trans retinoic acid sensitizes BCSCs to NK killing and prevents metastasis

Because differentiation therapy targeting CSC represents a promising therapeutic approach (2, 23, 24), we tested the capacity of all-trans retinoic acid (ATRA), a well-established differentiation-inducing agent for CSCs from leukemia and glioblastoma (25, 26), to affect immune evasion by BCSCs. ATRA markedly enhanced the expression of both MICA and MICB proteins (Fig. 6A) but not their mRNA levels (Supplementary Fig. S7A) by BCSCs. Moreover, ATRA treatment downregulated miR20a (Fig. 6B), whereas its restoration with miR20a mimics (miR20a M) in ATRA-treated BCSCs (Supplementary Fig. S7B) reduced MICA and MICB on BCSCs (Fig. 6C). These results suggest that ATRA regulates MICA and MICB expression by repressing aberrantly expressed miR20a.

Figure 6.

The effect of ATRA on the sensitivity to NK cell–mediated lysis and the metastatic potential of BCSCs. A and B, BCSCs were treated with ATRA for 48 hours. The expression of MICA, MICB (A), and miR20a (B) was examined. C, BCSCs with or without transfection of miR20a mimics (miR20aM) were treated with ATRA for 48 hours, and MICA or MICB expression was detected. D, lysis of BCSCs by pNK cells (mean ± SEM). E, the survival of BCSCs in vivo; (n = 3). F, numbers of lung metastatic foci formed by BCSCs in vivo (n = 3); mean ± SD (B, E, and F); *, P < 0.05.

Figure 6.

The effect of ATRA on the sensitivity to NK cell–mediated lysis and the metastatic potential of BCSCs. A and B, BCSCs were treated with ATRA for 48 hours. The expression of MICA, MICB (A), and miR20a (B) was examined. C, BCSCs with or without transfection of miR20a mimics (miR20aM) were treated with ATRA for 48 hours, and MICA or MICB expression was detected. D, lysis of BCSCs by pNK cells (mean ± SEM). E, the survival of BCSCs in vivo; (n = 3). F, numbers of lung metastatic foci formed by BCSCs in vivo (n = 3); mean ± SD (B, E, and F); *, P < 0.05.

Close modal

The enhancing effect of ATRA on MICA and MICB expression prompted us to investigate the potential of induced differentiation to sensitize BCSCs to NK cell cytotoxicity, thus preventing metastasis. BCSCs treated with ATRA showed reduction of miR20a and were more sensitive to NK cell–mediated lysis. Restoration of miR20a in ATRA-treated BCSCs or blocking NKG2D on pNK cells attenuated the sensitivity of ATRA-treated BCSCs to the lysis by NK cells (Fig. 6D). Furthermore, pretreatment of BCSCs with ATRA reduced their survival (Fig. 6E and Supplementary Fig. S7C) and metastatic capacity (Fig. 6F and Supplementary Fig. S7D) in mice transferred with human pNK cells. The effect of ATRA was reversed by introduction of exogenous miR20a in BCSCs or blockade of NKG2D on NK cells. In addition, ATRA treatment resulted in measurable reduction in the propagation and stem-like properties of BCSCs (Supplementary Fig. S7E and S7F). However, ATRA alone neither induced significant apoptotic death of BCSCs (Supplementary Fig. S7G), nor attenuated their survival/metastatic capacity in vivo (Fig. 6E and F), indicating that ATRA inhibits BCSC metastasis mainly through facilitating elimination of BCSCs by NK cells. Moreover, ATRA showed less effect on the expression of miR20a and MICA/MICB as well as the sensitivity to NK cell cytotoxicity in non-BCSCs (Supplementary Fig. S7H–S7J), suggesting that ATRA preferentially targets BCSCs. These observations indicate that targeting the miR20a–MICA/B axis by induction of differentiation sensitizes BCSCs to NK cell killing and prevents their metastasis.

CSCs are the cellular origin of multiple malignant properties in cancer, including unrestrained self-renewal (2), neovascularization (3), resistance to chemo- or radiotherapy (2, 27), dissemination (4), and evasion of antitumor immunity (5–9). The metastasis of BCSCs is a result of epithelial-to-mesenchymal transition (EMT) induced by genetic/epigenetic alterations or microenvironmental cues (4). In this study, we demonstrate that BCSCs evade NK cell surveillance to establish tumor metastasis, representing a revelation of novel mechanisms by which BCSCs establish distant metastasis.

Tumor heterogeneity has been recognized as the consequence of genetic diversity within cancer and the presence of CSCs (2). Immunoediting may also promote cancer progression and select tumor cells with enhanced malignancy (10). The potential for CSCs to avoid destruction by adaptive immunity has been documented in the context of CTL-mediated cytotoxicity (5–7, 9, 28). In addition, IFNγ released by CTL promotes the proliferation of CSCs and contributes to leukemia progression (8). These observations demonstrate that tumor metastasis is associated with dynamic regulation of CSC evasiveness from adaptive immunity.

The impact of innate immunity on cancer progression and enrichment of CSCs has largely been unclear. It has been reported that several types of normal stem cells were efficiently lyzed by fresh allogeneic NK cells (29–31). Similarly, stem-like cells in oral squamous carcinoma (31), colorectal carcinoma (32), and glioblastoma (33) were susceptible to NK cell–mediated cytotoxicity. However, leukemic stem cells were resistant to allogeneic NK cell cytotoxicity in vitro (34). Evasion of BCSCs from NK cell surveillance was implicated in trastuzumab resistance, in which treatment of MCF-7 cells with trastuzumab resulted in enrichment of CD44+CD24 cells by NK cells (35), although BCSCs and remaining cells from other cell lines were equally sensitive to allogeneic NK cell–mediated lysis in vitro (13). Our findings, using primary breast cancer cells, demonstrate that ALDH+ BCSCs were resistant to both autologous and allogeneic NK cell cytotoxicity, and that cancer cells surviving NK cell killing were enriched with ALDH+ BCSCs. Therefore, CSCs from neoplastic tissues of different origin exhibit diverse sensitivity to lysis by NK cells. Understanding this diversity has important implication for the development of rational immune therapy against CSCs.

NK cell anergy is detected in a great number of patients with cancer, especially those with late-stage cancer (15, 36). By using in vitro expanded anergic NK cells, we demonstrate that immunoevasion of BCSCs from NK cell cytotoxicity is essential for establishment of breast cancer metastasis. Therefore, resistance to NK cell lysis may be an intrinsic property of BCSCs to initiate metastasis in the patients with dysfunctional NK cells. In cancer tissues, we observed codistribution of CD56+ cells with ALDH1+ BCSCs, indicating that these cells may play wider roles in tumor microenvironment. IFNγ released by immune cells regulates the biology of CSCs in chronic myeloid leukemia (8). Nevertheless, no significant difference of IFNγ receptor α chain expression between BCSCs and non-BCSCs indicates that the immune evasion of BCSCs is not likely due to the differential activation of IFNγ signaling by infiltrating NK cells.

We found that the lack of stimulatory signals in BCSCs for NK cells was associated with reduced expression of MICA and MICB, two ligands for the NK-activating receptor NKG2D (37), in BCSCs. MICA and MICB are stress-inducible antigens rarely expressed by healthy cells but are frequently upregulated on transformed or virus-infected cells (38). Reduced expression of MICA/B facilitates the evasion of epithelial tumors from surveillance by NK cells and subsets of T cells (39). We demonstrated that reduced MICA/B expression by BCSCs contributed to their resistance to NK cell cytotoxicity and enhanced metastatic capacity in vivo. Therefore, it is plausible that reduction in MICA/B expression promotes BCSC metastasis by evading NK cell killing.

Despite the lower expression of MICA/B proteins by BCSCs, no substantial difference in MICA/B mRNA levels was detected between BCSCs and non-BCSCs, suggesting posttranscriptional regulation of MICA/B in tumor cells. Small noncoding RNAs, miRNAs, are important posttranscriptional regulators in CSCs (40, 41). miRNAs targeting MICA or MICB have been identified in studies of viral infection and IFNγ treatment (42–45). As a member of the oncogenic miR17-92 cluster, the expression of miR20a is elevated in several malignancies (46–48). We demonstrated that overexpression of miR20a reduced the levels of both MICA and MICB in BCSCs, resulting in their immune evasion with increased metastasis. Thus, miR20a is a pivotal regulator of immune evasion by BCSCs, and represents a potential target for strengthening immunosurveillance to inhibit tumor metastasis.

Promoting differentiation is a promising therapeutic approach against CSCs. Induced differentiation of CSCs results in their loss of chemoresistance and tumorigenicity (24). One such clinically applied differentiation agent is ATRA, which is well-known for its potent differentiation-inducing effect on CSCs from leukemia and glioblastoma (25, 26). In the present study, BCSC differentiated by ATRA showed diminished immune evasiveness with reduced expression of miR20a but enhanced MICA/B expression. Thus, the resistance of BCSCs to NK cell surveillance is reversible by differentiation and increasing the levels of ligands for NK cell stimulating receptor NKG2D.

In conclusion, our present study reveals that an aberrant miR20a–MICA/B axis is responsible for the escape of BCSCs from innate immunity, which is critical for increased metastatic capacity of BCSCs. Thus, targeting the miR20a–MICA/B axis may represent a promising immune therapeutic strategy to prevent breast cancer metastasis.

No potential conflicts of interest were disclosed.

Conception and design: B. Wang, Z. Wang, X.-W. Bian

Development of methodology: B. Wang, Z. Wang, J. Jiang, S.-C. Yu, Y.-H. Cui

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Wang, Q. Wang, Z. Wang, J. Jiang, S.-C. Yu, Y.-F. Ping, J. Yang, S.-L. Xu, X.-Z. Ye, C. Xu, L. Yang, C. Qian, Y.-H. Cui, X.-W. Bian

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Wang, Z. Wang, S.-C. Yu, L. Yang, Y.-H. Cui, X.-W. Bian

Writing, review, and/or revision of the manuscript: B. Wang, J.-M. Wang, Y.-H. Cui, X. Zhang, X.-W. Bian

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Wang, S.-C. Yu, Y.-H. Cui, X.-W. Bian

Study supervision: Y.-H. Cui, X.-W. Bian

The authors thank Dr. Peng Tang for collecting fresh specimens from patients with breast cancer (Breast Surgery Center, Southwest Hospital, the Third Military Medical University) and Qing-hua Ma for assistance in FACS analyses (Institute of Pathology and Southwest Cancer Center, Southwest Hospital, Third Military Medical University).

This work was supported by grants from the National Basic Research Program of China (973 Program, no. 2010CB529403; X.-W. Bian), the National Natural Science Foundation of China [nos. 81071771, 30930103, and 30725035 (X-W. Bian); 81101631 (B. Wang); and 81101606 (Z. Wang)], the National Science and Technology Major Program (no. 2011ZX09102-010-02; X.-W. Bian), and the Natural Science Foundation Project of CQ CSTC (no. CSTC2012JJA10124; B. Wang). J.-M. Wang was supported by the Intramural Research Program of the National Cancer Institute, NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bidwell
BN
,
Slaney
CY
,
Withana
NP
,
Forster
S
,
Cao
Y
,
Loi
S
, et al
Silencing of Irf7 pathways in breast cancer cells promotes bone metastasis through immune escape
.
Nat Med
2012
;
18
:
1224
31
.
2.
Visvader
JE
,
Lindeman
GJ
. 
Cancer stem cells: current status and evolving complexities
.
Cell Stem Cell
2012
;
10
:
717
28
.
3.
Ping
YF
,
Bian
XW
. 
Consice review: contribution of cancer stem cells to neovascularization
.
Stem Cells
2011
;
29
:
888
94
.
4.
Baccelli
I
,
Trumpp
A
. 
The evolving concept of cancer and metastasis stem cells
.
J Cell Biol
2012
;
198
:
281
93
.
5.
Noh
KH
,
Kim
BW
,
Song
KH
,
Cho
H
,
Lee
YH
,
Kim
JH
, et al
Nanog signaling in cancer promotes stem-like phenotype and immune evasion
.
J Clin Invest
2012
;
122
:
4077
93
.
6.
Noh
KH
,
Lee
YH
,
Jeon
JH
,
Kang
TH
,
Mao
CP
,
Wu
TC
, et al
Cancer vaccination drives Nanog-dependent evolution of tumor cells toward an immune-resistant and stem-like phenotype
.
Cancer Res
2012
;
72
:
1717
27
.
7.
Santisteban
M
,
Reiman
JM
,
Asiedu
MK
,
Behrens
MD
,
Nassar
A
,
Kalli
KR
, et al
Immune-induced epithelial to mesenchymal transition in vivo generates breast cancer stem cells
.
Cancer Res
2009
;
69
:
2887
95
.
8.
Schurch
C
,
Riether
C
,
Amrein
MA
,
Ochsenbein
AF
. 
Cytotoxic T cells induce proliferation of chronic myeloid leukemia stem cells by secreting interferon-gamma
.
J Exp Med
2013
;
210
:
605
21
.
9.
Akalay
I
,
Janji
B
,
Hasmim
M
,
Noman
MZ
,
Andre
F
,
De Cremoux
P
, et al
Epithelial-to-mesenchymal transition and autophagy induction in breast carcinoma promote escape from T-cell–mediated lysis
.
Cancer Res
2013
;
73
:
2418
27
.
10.
Schreiber
RD
,
Old
LJ
,
Smyth
MJ
. 
Cancer immunoediting: integrating immunity's roles in cancer suppression and promotion
.
Science
2011
;
331
:
1565
70
.
11.
O'Sullivan
T
,
Saddawi-Konefka
R
,
Vermi
W
,
Koebel
CM
,
Arthur
C
,
White
JM
, et al
Cancer immunoediting by the innate immune system in the absence of adaptive immunity
.
J Exp Med
2012
;
209
:
1869
82
.
12.
Dewan
MZ
,
Terunuma
H
,
Takada
M
,
Tanaka
Y
,
Abe
H
,
Sata
T
, et al
Role of natural killer cells in hormone-independent rapid tumor formation and spontaneous metastasis of breast cancer cells in vivo
.
Breast Cancer Res Treat
2007
;
104
:
267
75
.
13.
Mamessier
E
,
Sylvain
A
,
Bertucci
F
,
Castellano
R
,
Finetti
P
,
Houvenaeghel
G
, et al
Human breast tumor cells induce self-tolerance mechanisms to avoid NKG2D-mediated and DNAM-mediated NK cell recognition
.
Cancer Res
2011
;
71
:
6621
32
.
14.
Sceneay
J
,
Chow
MT
,
Chen
A
,
Halse
HM
,
Wong
CS
,
Andrews
DM
, et al
Primary tumor hypoxia recruits CD11b+/Ly6Cmed/Ly6G+ immune suppressor cells and compromises NK cell cytotoxicity in the premetastatic niche
.
Cancer Res
2012
;
72
:
3906
11
.
15.
Tai
LH
,
de Souza
CT
,
Belanger
S
,
Ly
L
,
Alkayyal
AA
,
Zhang
J
, et al
Preventing postoperative metastatic disease by inhibiting surgery-induced dysfunction in natural killer cells
.
Cancer Res
2013
;
73
:
97
107
.
16.
Kasimir-Bauer
S
,
Hoffmann
O
,
Wallwiener
D
,
Kimmig
R
,
Fehm
T
. 
Expression of stem cell and epithelial-mesenchymal transition markers in primary breast cancer patients with circulating tumor cells
.
Breast Cancer Res
2012
;
14
:
R15
.
17.
Liu
H
,
Patel
MR
,
Prescher
JA
,
Patsialou
A
,
Qian
D
,
Lin
J
, et al
Cancer stem cells from human breast tumors are involved in spontaneous metastases in orthotopic mouse models
.
Proc Natl Acad Sci U S A
2010
;
107
:
18115
20
.
18.
Charafe-Jauffret
E
,
Ginestier
C
,
Iovino
F
,
Wicinski
J
,
Cervera
N
,
Finetti
P
, et al
Breast cancer cell lines contain functional cancer stem cells with metastatic capacity and a distinct molecular signature
.
Cancer Res
2009
;
69
:
1302
13
.
19.
Ginestier
C
,
Hur
MH
,
Charafe-Jauffret
E
,
Monville
F
,
Dutcher
J
,
Brown
M
, et al
ALDH1 is a marker of normal and malignant human mammary stem cells and a predictor of poor clinical outcome
.
Cell Stem Cell
2007
;
1
:
555
67
.
20.
Ginestier
C
,
Liu
S
,
Diebel
ME
,
Korkaya
H
,
Luo
M
,
Brown
M
, et al
CXCR1 blockade selectively targets human breast cancer stem cells in vitro and in xenografts
.
J Clin Invest
2010
;
120
:
485
97
.
21.
Elboim
M
,
Gazit
R
,
Gur
C
,
Ghadially
H
,
Betser-Cohen
G
,
Mandelboim
O
. 
Tumor immunoediting by NKp46
.
J Immunol
2010
;
184
:
5637
44
.
22.
Tsukerman
P
,
Stern-Ginossar
N
,
Gur
C
,
Glasner
A
,
Nachmani
D
,
Bauman
Y
, et al
miR-10b downregulates the stress-induced cell surface molecule MICB, a critical ligand for cancer cell recognition by natural killer cells
.
Cancer Res
2012
;
72
:
5463
72
.
23.
Wang
B
,
Yu
SC
,
Jiang
JY
,
Porter
GW
,
Zhao
LT
,
Wang
Z
, et al
An inhibitor of arachidonate 5-lipoxygenase, Nordy, induces differentiation and inhibits self-renewal of glioma stem-like cells
.
Stem Cell Rev
2011
;
7
:
458
70
.
24.
Sachlos
E
,
Risueno
RM
,
Laronde
S
,
Shapovalova
Z
,
Lee
JH
,
Russell
J
, et al
Identification of drugs including a dopamine receptor antagonist that selectively target cancer stem cells
.
Cell
2012
;
149
:
1284
97
.
25.
Jiao
B
,
Ren
ZH
,
Liu
P
,
Chen
LJ
,
Shi
JY
,
Dong
Y
, et al
8-CPT-cAMP/all-trans retinoic acid targets t(11;17) acute promyelocytic leukemia through enhanced cell differentiation and PLZF/RARalpha degradation
.
Proc Natl Acad Sci U S A
2013
;
110
:
3495
500
.
26.
Ying
M
,
Wang
S
,
Sang
Y
,
Sun
P
,
Lal
B
,
Goodwin
CR
, et al
Regulation of glioblastoma stem cells by retinoic acid: role for Notch pathway inhibition
.
Oncogene
2011
;
30
:
3454
67
.
27.
Wang
B
,
Liu
J
,
Ma
LN
,
Xiao
HL
,
Wang
YZ
,
Li
Y
, et al
Chimeric 5/35 adenovirus-mediated Dickkopf-1 overexpression suppressed tumorigenicity of CD44+gastric cancer cells via attenuating Wnt signaling
.
J Gastroenterol
2013
;
48
:
798
808
.
28.
Volonte
A
,
Di Tomaso
T
,
Spinelli
M
,
Todaro
M
,
Sanvito
F
,
Albarello
L
, et al
Cancer-Initiating cells from colorectal cancer patients escape from T cell-mediated immunosurveillance in vitro through membrane-bound IL-4
.
J Immunol
2014
;
192
:
523
32
.
29.
Spaggiari
GM
,
Capobianco
A
,
Becchetti
S
,
Mingari
MC
,
Moretta
L
. 
Mesenchymal stem cell-natural killer cell interactions: evidence that activated NK cells are capable of killing MSCs, whereas MSCs can inhibit IL-2-induced NK-cell proliferation
.
Blood
2006
;
107
:
1484
90
.
30.
Sotiropoulou
PA
,
Perez
SA
,
Gritzapis
AD
,
Baxevanis
CN
,
Papamichail
M
. 
Interactions between human mesenchymal stem cells and natural killer cells
.
Stem Cells
2006
;
24
:
74
85
.
31.
Tseng
HC
,
Arasteh
A
,
Paranjpe
A
,
Teruel
A
,
Yang
W
,
Behel
A
, et al
Increased lysis of stem cells but not their differentiated cells by natural killer cells; de-differentiation or reprogramming activates NK cells
.
PLoS ONE
2010
;
5
:
e11590
.
32.
Tallerico
R
,
Todaro
M
,
Di Franco
S
,
Maccalli
C
,
Garofalo
C
,
Sottile
R
, et al
Human NK cells selective targeting of colon cancer-initiating cells: a role for natural cytotoxicity receptors and MHC class I molecules
.
J Immunol
2013
;
190
:
2381
90
.
33.
Castriconi
R
,
Daga
A
,
Dondero
A
,
Zona
G
,
Poliani
PL
,
Melotti
A
, et al
NK cells recognize and kill human glioblastoma cells with stem cell-like properties
.
J Immunol
2009
;
182
:
3530
9
.
34.
She
M
,
Niu
X
,
Chen
X
,
Li
J
,
Zhou
M
,
He
Y
, et al
Resistance of leukemic stem-like cells in AML cell line KG1a to natural killer cell–mediated cytotoxicity
.
Cancer Lett
2012
;
318
:
173
9
.
35.
Reim
F
,
Dombrowski
Y
,
Ritter
C
,
Buttmann
M
,
Hausler
S
,
Ossadnik
M
, et al
Immunoselection of breast and ovarian cancer cells with trastuzumab and natural killer cells: selective escape of CD44high/CD24low/HER2low breast cancer stem cells
.
Cancer Res
2009
;
69
:
8058
66
.
36.
Orr
MT
,
Lanier
LL
. 
Natural killer cell education and tolerance
.
Cell
2010
;
142
:
847
56
.
37.
Long
EO
,
Sik Kim
H
,
Liu
D
,
Peterson
ME
,
Rajagopalan
S
. 
Controlling natural killer cell responses: integration of signals for activation and inhibition
.
Annu Rev Immunol
2013
;
31
:
227
58
.
38.
Raulet
DH
,
Gasser
S
,
Gowen
BG
,
Deng
W
,
Jung
H
. 
Regulation of ligands for the NKG2D activating receptor
.
Annu Rev Immunol
2013
;
31
:
413
41
.
39.
Champsaur
M
,
Lanier
LL
. 
Effect of NKG2D ligand expression on host immune responses
.
Immunol Rev
2010
;
235
:
267
85
.
40.
Liu
S
,
Patel
SH
,
Ginestier
C
,
Ibarra
I
,
Martin-Trevino
R
,
Bai
S
, et al
MicroRNA93 regulates proliferation and differentiation of normal and malignant breast stem cells
.
PLoS Genet
2012
;
8
:
e1002751
.
41.
Yu
Z
,
Willmarth
NE
,
Zhou
J
,
Katiyar
S
,
Wang
M
,
Liu
Y
, et al
microRNA 17/20 inhibits cellular invasion and tumor metastasis in breast cancer by heterotypic signaling
.
Proc Natl Acad Sci U S A
2010
;
107
:
8231
6
.
42.
Nachmani
D
,
Lankry
D
,
Wolf
DG
,
Mandelboim
O
. 
The human cytomegalovirus microRNA miR-UL112 acts synergistically with a cellular microRNA to escape immune elimination
.
Nat Immunol
2010
;
11
:
806
13
.
43.
Nachmani
D
,
Stern-Ginossar
N
,
Sarid
R
,
Mandelboim
O
. 
Diverse herpesvirus microRNAs target the stress-induced immune ligand MICB to escape recognition by natural killer cells
.
Cell Host Microbe
2009
;
5
:
376
85
.
44.
Stern-Ginossar
N
,
Gur
C
,
Biton
M
,
Horwitz
E
,
Elboim
M
,
Stanietsky
N
, et al
Human microRNAs regulate stress-induced immune responses mediated by the receptor NKG2D
.
Nat Immunol
2008
;
9
:
1065
73
.
45.
Yadav
D
,
Ngolab
J
,
Lim
RS
,
Krishnamurthy
S
,
Bui
JD
. 
Cutting edge: down-regulation of MHC class I-related chain A on tumor cells by IFN-gamma-induced microRNA
.
J Immunol
2009
;
182
:
39
43
.
46.
Matsubara
H
,
Takeuchi
T
,
Nishikawa
E
,
Yanagisawa
K
,
Hayashita
Y
,
Ebi
H
, et al
Apoptosis induction by antisense oligonucleotides against miR-17-5p and miR-20a in lung cancers overexpressing miR-17-92
.
Oncogene
2007
;
26
:
6099
105
.
47.
Huang
G
,
Nishimoto
K
,
Zhou
Z
,
Hughes
D
,
Kleinerman
ES
. 
miR-20a encoded by the miR-17-92 cluster increases the metastatic potential of osteosarcoma cells by regulating Fas expression
.
Cancer Res
2012
;
72
:
908
16
.
48.
Conkrite
K
,
Sundby
M
,
Mukai
S
,
Thomson
JM
,
Mu
D
,
Hammond
SM
, et al
miR-17∼92 cooperates with RB pathway mutations to promote retinoblastoma
.
Genes Dev
2011
;
25
:
1734
45
.

Supplementary data