Cancer Stem Cell Vaccination Confers Significant Antitumor Immunity

Clinical trials to treat patients with cancer using adoptively transferred T cells (1–3) or dendritic cells (DC; refs. 4–6) have shown therapeutic efficacy for patients with advanced diseases. However, the clinical responses to such immunotherapeutic approaches have been confined to a limited percentage of treated patients. To date, bulk tumor masses with heterogeneous populations of cancer cells have been used as a source of antigen either to generate effector T cells or to prime DC vaccines. Human tumors are composed of heterogeneous tumor cell clones that differ with respect to proliferation, differentiation, and ability to initiate daughter tumors. The inability to target cancer stem cells (CSC) with current immune approaches may be a significant factor for treatment failures.

The identification of human CSCs (7–17) presents a new paradigm for the development of cancer treatments. These stem cells have been shown to be relatively resistant to conventional chemotherapeutic regimens and radiation (18, 19) and are postulated to be the cells responsible for the relapse and progression of cancers after such therapies. In an analogous fashion, the CSC phenomenon may adversely affect the development of effective immunotherapies for cancer. These therapies have involved targeting cells that express differentiated tumor antigens. However, such antigens may be selectively expressed on differentiated tumor cells. CSCs that do not express these antigens may thus escape these immunologic interventions.

While a few studies have evaluated the resistance of CSCs to the cytotoxic effects of chemotherapy (18, 20–23) and low-dose radiation treatment (19), the immunogenicity of CSCs and their susceptibility to immune-based therapy have not been determined. So far, the majority of CSC studies have been conducted using human tumors inoculated into severely immunosuppressed hosts [e.g., severe combined immunodeficient (SCID) mice]. These hosts represent very useful models for the studies of the biology, tumorigenicity, and signaling pathways of human CSCs as well as for the screening of small molecules which may lead to the development of new drugs that target CSCs. A very recent report described the isolation of cancer-initiating cells (CIC) using ALDEFLUOR/ALDH as a marker from human head and neck, breast, and pancreatic carcinoma cell lines and the generation of ALDH1A1-specific CD8 T cells in vitro (24). These T cells eliminated CICs in vivo by adoptive transfer to immunodeficient (SCID) mice bearing human tumor xenografts. However, the absence of adaptive immune responses in the SCID mouse precludes the ability to investigate the host immune response to CSCs. Although normal mouse mammary stem cells have been isolated (25), there is a need to develop model systems where CSCs can be isolated in the immunocompetent host to evaluate the immunogenicity of CSCs.

In this study, we isolated and assessed the tumorigenicity of murine CSCs in 2 histologically different tumors from 2 genetically distinct immunocompetent hosts. From there, we evaluated the immunogenicity induced by purified CSCs used as a source of antigen to prime DCs as a vaccine. We found that CSC-based vaccines conferred effective protective antitumor immunity which was associated with the induction of humoral and cellular responses that directly targeted CSCs via complement-dependent cytotoxicity (CDC) and CTLs, respectively.

Female C57BL/6 (B6) and C3H/HeNCrMTV (C3H) mice were from Charles River Laboratories. All the animals were maintained in a pathogen-free environment and used at age 8 weeks or older. The University of Michigan Laboratory of Animal Medicine (Ann Arbor, MI) approved all the animal protocols.

D5 is a clone which our laboratory produced (26) from the B16-BL6 tumor line that is a poorly immunogenic melanoma of spontaneous origin syngeneic to B6 mice (27, 28). SCC7 is a spontaneously arising squamous cell cancer syngeneic to C3H mice also described in our previous report (29).

The ALDEFLUOR Kit (StemCell Technologies) labels the ALDEFLUOR⁺/ALDH^high population including the stem/progenitor cells (30–33). The ALDEFLUOR assay uses a fluorescent substrate of the enzyme (BAAA) freely diffusible across cell membranes. Polar fluorescent products (BAA) accumulate when this substrate is oxidized in cells that express aldehyde dehydrogenase (ALDH). Consequently, cells with high levels of ALDH enzymatic activity stain more brightly (ALDEFLUOR⁺ also referred to as ALDH⁺ or ALDH^high) than cells with lower ALDH (ALDEFLUOR⁻ also referred to as ALDH⁻ or ALDH^low). The fluorescent product BAA is trapped in the cells, due to its negative charges. In each experiment, a sample of cells was stained under identical conditions with specific ALDH inhibitor diethylaminobenzaldehyde (DEAB) as negative control. Flow cytometry–based sorting is conducted using a FACStarPLUS. The sorting gates are established using as negative controls the propidium iodide (PI)-stained cells for viability and the ALDEFLUOR-stained cells treated with DEAB.

Equal number of ALDEFLUOR⁺ or ALDEFLUOR⁻ tumor cells mixed with Matrigel (BD Biosciences; 1:1) were injected into the opposite side of the syngeneic mice. Tumor size was measured every 3 to 4 days.

To examine the protective antitumor immunity induced by vaccination with DCs pulsed with the lysate of ALDEFLUOR⁺ cells (CSC-TPDC), ALDEFLUOR⁺/ALDH^high and ALDEFLUOR⁻/ALDH^low cells were isolated as described above either from cultured D5 and SCC7 cells or from freshly harvested growing tumors from initial respective ALDEFLUOR⁺ D5 or SCC cell injection. ALDEFLUOR⁺, ALDEFLUOR⁻, and unsorted cells were frozen and thawed 3 times to make cell lysate. Bone marrow–derived DCs were cultured in interleukin (IL)-4 and granulocyte macrophage colony—stimulating factor (GM-CSF) as previously described in our laboratory (5, 27) and were pulsed with tumor lysate to generate tumor lysate–pulsed DCs (TPDC). After 24 hours of coculture, normal animals were vaccinated with CSC-TPDC or DC pulsed with lysate from unsorted heterogeneous tumor cells (H-TPDC) or DCs pulsed with sorted ALDEFLUOR⁻ cell lysate (ALDH^low-TPDC) at the same DC to tumor cell lysate ratio as CSC-TPDC.

After vaccination, the B6 mice were challenged with the heterogeneous D5 tumor cells intravenously and the lungs harvested 20 days later to enumerate lung metastases. In SCC7 model, the C3H mice were challenged with the heterogeneous SCC7 tumor cells subcutaneously on the opposite side of the vaccine and the tumor size was monitored.

To test systematic immune responses conferred by CSC-based vaccine, spleens were harvested at the end of experiments. Spleen B cells were activated with lipopolysaccharide (LPS) plus anti-CD40 (FGK45) monoclonal antibody (mAb) ascites as previously described (28). After activation, supernatants were collected and analyzed for IgG production.

Plasma was collected from vaccinated hosts at the end of the experiments. IgG level was tested using ELISA. ALDEFLUOR⁺ cells were washed with fluorescence-activated cell-sorting (FACS) buffer, blocked with anti-CD16/CD32 (BD biosciences), and incubated with the plasma with equal quantity of IgG for 60 minutes on ice. Cells were washed again and incubated with fluorescein isothiocyanate (FITC) anti-mouse IgG (0.5 μg/10⁶ cells) for 30 minutes on ice. Cells were then washed and their binding to plasma IgG was detected using flow cytometry.

CSC lysis mediated by antibodies in plasma was assessed by incubation of 10⁵ viable ALDH⁺ CSCs or ALDH⁻ non-CSCs (serving as control) with plasma in test tubes on ice for 1 hour followed by culture in the presence of rabbit complement (Calbiochem) in a 37ºC water bath for another 1 hour. Viable cells were then counted after trypan blue staining to calculate CSC lysis: % of viable cells = viable cells counted after plasma and complement incubation/10⁵. Lower percentage of viable cells at the end of incubation indicates more cell lysis.

CTLs were generated from the peripheral blood mononuclear cells (PBMC) or splenocytes harvested from vaccinated animals by anti-CD3/CD28 activation and IL-2 expansion, which consistently results in more than 90% of CD3⁺ T cells (data not shown). CTL-mediated CSC cytotoxicity was tested using the lactate dehydrogenase (LDH) Release Assay (CytoTox 96 Non-Radioactive Cytotoxicity Assay, Promega) according to the manufacturer's protocol. The following formula was used to calculate cytotoxicity:

The significance of difference in tumorigenicity, metastatic nodules, tumor size, the concentration of IgG, and CSC lysis by antibodies or CTLs was determined using unpaired t test. P values <0.05 were considered statistically significant between the experiment groups.

We have previously described the isolation of stem cell–enriched populations using ALDEFLUOR/ALDH as a marker (30, 33, 34). Using this technique, we identified CSC-enriched populations in 2 different immunocompetent murine tumor models. As indicated in Fig. 1, approximately 4% to 6% of the cultured murine melanoma D5 and squamous cancer SCC7 cells are ALDEFLUOR⁺/ALDH^+/high; with the rest (∼95%) being ALDEFLUOR⁻. The existence of a small percentage of ALDEFLUOR⁺ cells in established murine tumors was confirmed by analyzing freshly harvested tumor cells. Processed tumor cells from in vivo established D5 and SCC7 murine tumors also revealed approximately 5% of the ALDEFLUOR⁺ cells (Fig. 1). To determine the purity of the sorted cells, the whole ALDEFLUOR⁺/ALDH^+/high cells and an approximately equal number (5%–15%) of the ALDEFLUOR⁻ cells used for the subsequent immunogenicity analyses (ALDH^−/low cells) were collected and restained with ALDEFLUOR using the same staining protocol. High percentages (>90% in 7 of the 8 restains) of the ALDH^−/low cells (in blue) and ALDH^+/high cells (in red) after restaining confirmed the purity of originally sorted cells (Fig. 1).

The tumorigenicity of sorted D5 melanoma ALDEFLUOR⁺ cells was evaluated in the syngeneic immunocompetent C57BL/6 host. ALDEFLUOR⁺ D5 cells (5,000 per inoculum) generated large size tumors in 19 days (Fig. 2A), whereas equal numbers of ALDEFLUOR⁻ cells injected into the opposite side of the same mouse failed to generate tumors. Separate mice were injected with much lower numbers of ALDEFLUOR⁺ D5 cells. In 4 weeks, 500 injected ALDEFLUOR⁺ cells formed tumors (Fig. 2A). In contrast, the curve in Fig. 2A shows that as many as 50,000 ALDEFLUOR⁻ D5 cells did not grow. The tumorigenicity of sorted SCC7 ALDEFLUOR⁺ population was evaluated in the syngeneic immunocompetent C3H host. As was the case for the D5 model, only the ALDEFLUOR⁺ (as few as 2,000) cells grew into tumors whereas equal numbers or even much greater numbers (as many as 200,000) of ALDEFLUOR⁻ cells failed to generate tumor (Fig. 2B). These results indicate that the ALDEFLUOR⁻ tumor cells are less tumorigenic than ALDEFLUOR⁺ cells.

Collectively, these data indicate that ALDEFLUOR/ALDH can serve as a reliable marker for the enrichment of murine D5 and SCC7 CSCs. This has allowed us to characterize CSC-induced antitumor immunity in the immunocompetent murine host in the subsequent experiments.

DCs pulsed with whole tumor lysate have been reported to be an effective vaccine for cancer both in animal studies (35) and in clinical trials (4) including the findings reported by our own group (5, 27). To examine whether DCs pulsed with the lysate of CSCs are more effective in inducing antitumor immunity, we evaluated the protective antitumor immunity induced by vaccination with DCs pulsed with the lysate of ALDEFLUOR⁺ cells (CSC-TPDC) and used DCs pulsed with the lysate of whole unsorted heterogeneous tumor cells (H-TPDC) as a positive control. In the D5 melanoma model, we used D5 CSCs as a source of antigen. D5 subcutaneous tumors were established and used as a source of CSCs by sorting for ALDEFLUOR⁺ cells. DCs were pulsed with tumor lysate to generate TPDC to be used as a vaccine. Normal immunocompetent B6 mice were immunized with D5 CSC-TPDC or D5 H-TPDC (at the same lysate:DC ratio as D5 CSC-TPDC). Control groups received PBS. The TPDCs were inoculated subcutaneously for 2 doses (10⁶/each) given 1 week apart. Seven days after the last vaccine, the mice were challenged with the heterogeneous D5 tumor cells intravenously and the lungs harvested 20 days later to enumerate lung metastases. The study scheme and results are illustrated in Fig. 3. Compared with nonvaccinated, PBS-injected animals (PBS), H-TPDC induced protective immunity against tumor growth, which corroborates with previous observations (27, 35). Of note, the pulsing of tumor cell lysate to DC to generate H-TPDCs was suboptimal compared with what has been reported in the past (4, 5, 27, 34), which may partially explain why H-TPDCs did not immunize effectively (Fig. 3). In these experiments, DCs were pulsed with the lysate of ALDEFLUOR⁺ cells to generate CSC-TPDCs at the ratio of DCs to ALDEFLUOR⁺ cells 3:1 (D5) and 10:1 (SCC7), respectively. This ratio is much lower than the DC:unsorted tumor cell ratio (1:3) as previously described by our group (4, 5, 27). We used fewer tumor cells to pulse DCs to generate CSC-TPDC due to the following reasons: (i) the number of ALDEFLUOR⁺ cells obtained was limited and (ii) we wanted to see the antitumor potential of DCs pulsed with this limited number of ALDEFLUOR⁺ cells compared with the DCs pulsed with the same number of unsorted cells. Importantly, mice that received CSC-TPDC prepared at an identical low lysate to DC ratio resulted in significantly fewer lung metastases than the PBS control group as well as the H-TPDC vaccine group. These results suggested that D5 CSCs are immunogenic and can induce an immune response, even under a suboptimal CSC to DC pulsing condition, which led to decreased lung colonization upon tumor challenge. Significantly, these experiments showed a not yet recognized gain and beneficial effect against tumor growth mediated by CSC-TPDC versus H-TPDC (P = 0.018 in experiment 1 and P = 0.001 in experiment 2).

In the SCC7 model, subcutaneous SCC7 tumors were established and used as a source of CSC by sorting for ALDEFLUOR⁺ cells. Normal C3H animals were vaccinated with CSC-TPDC (DCs pulsed with ALDEFLUOR⁺ SCC7 cells) or DCs pulsed with lysate from unsorted heterogeneous SCC7 cells (H-TPDC). The TPDCs were inoculated subcutaneously for 3 doses given 1 week apart on days −14, −7, and 0 in the right flank of C3H mice. CSC-TPDC- or H-TPDC–vaccinated mice were challenged with unsorted SCC7 tumor cells subcutaneously into the left flank on day 0 and tumor growth monitored. Compared with nonvaccinated control animals, H-TPDCs induced protective immunity against tumor growth to a modest extent (Fig. 4A–C). In contrast, mice that received CSC-TPDC showed significant inhibition of tumor growth compared with the no-treatment group (Fig. 4A; P = 0.003), and the tumors were much smaller than those growing in the H-TPDC–treated hosts (Fig. 4B). These results confirmed our previously observed CSC-induced protective immunity against D5 melanoma in a second tumor model syngeneic to a different immunocompetent host, as well as in a different tumor setting (subcutaneous tumor growth) that enriched CSCs are immunogenic and more effective as an antigen source than unsorted heterogeneous tumor cells in inducing immunity of the host to reject the challenge of tumor cells.

We repeated the experiment as shown in Fig. 4A, but added one control group: ALDH^lowTPDC, for example, DCs pulsed with sorted ALDH^−/low SCC7 cell lysate. As shown in Fig. 4C, vaccination with ALDH^highTPDC (CSC-TPDC) induced significantly higher protective immunity against tumor than H-TPDC as well as ALDH^lowTPDC. Furthermore, we have carried out additional experiments with the D5 tumor model by evaluating the efficacy of CSC-TPDCs in the protection of hosts against subcutaneous tumor challenge. As shown in Fig. 4D, vaccination of DCs pulsed with ALDH^+/high D5 cell lysate (ALDH^highTPDC) induced significantly higher protective immunity against subcutaneous tumor challenge than H-TPDCs as well as ALDH^lowTPDC vaccination. We analyzed the phenotype of the tumors growing from ALDFLUOR⁺ cells. The ALDEFLUOR⁺ population regenerated the initial heterogeneity of the tumor by reconstitution of both ALDEFLUOR⁺ and ALDEFLUOR⁻ cell populations. To verify our findings using the ALDELFUOR⁺/ALDH^+/high cells isolated directly from the cultured cell lines, we separated ALDELFUOR⁺ cells isolated from freshly harvested growing tumor in murine hosts which resulted from previous ALDELFUOR⁺ cell injection. The stem cell nature of the ALDELFUOR⁺ cells isolated from freshly harvested growing tumors was verified both in vitro and in vivo; in both D5 and SCC7 models (Supplementary Figs. S1 and S2 and Table S1). The data shown in Fig. 4D were generated by pulsing DCs with lysates from ALDELFUOR⁺/ALDH^+/high cells isolated freshly from the growing D5 tumor followed by D5 subcutaneous challenge. Consistent with our early findings using cultured tumor cells as a source of CSCs, these experiments showed that vaccination of DCs pulsed with the lysate of ALDELFUOR⁺/ALDH^+/high cells isolated from growing tumor induced significantly higher protective immunity against tumor than H-TPDCs as well as DCs pulsed with the lysate of ALDH^−/low cells also isolated from growing tumors (Fig. 4D). It is worth noting that the P values of the difference between PBS versus H-TPDCs and PBS vs. ALDH^lowTPDC growth curves in Fig. 4C using SCC7 was <0.05 at all time points except for day 11 (P = 0.523 and 0.308, respectively) after tumor inoculation. In contract, there was no significant difference between PBS versus H-TPDC and PBS versus ALDH^lowTPDC growth curves in Fig. 4D using D5. However, Fig. 4C and D both showed that vaccination of DCs pulsed with the lysate of ALDH^+/high cells induced significantly higher antitumor immunity than PBS, H-TPDC, as well as DCs pulsed with the lysate of ALDH^−/low cells in 2 tumor models. These data further highlight the advantage of using CSCs in vaccination versus the traditional vaccine strategy using bulk unsorted tumor cells (H-TPDC) or using ALDH^−/low cells.

To define possible mechanisms underlying CSC-induced protective antitumor immunity, we harvested the splenocytes from the animals subjected to H-TPDC and CSC-TPDC vaccination. These cells were secondarily activated in vitro and the culture supernatants collected for antibody detection. We found significantly higher IgG production by LPS/anti-CD40–activated splenocytes isolated from the animals vaccinated with D5 CSC-TPDC or SCC7 CSC-TPDC than D5 H-TPDC (P = 0.004) or SCC7 H-TPDC (Fig. 5A; P = 0.031). These data showed systemic humoral responses in CSC-TPDC–vaccinated immunocompetent hosts.

In the experiments shown in Figs. 3 and 4, the enhanced CSC-induced protective antitumor immunity is postulated to occur by inhibiting the growth of CSCs present in the unsorted tumor inoculums. To examine this hypothesis, we harvested the sera and PBMCs from the animals subjected to vaccination to determine the specificity of the immune responses to CSCs. Using flow cytometry (Fig. 5B), we observed that immune sera from D5 CSC-TPDC–vaccinated hosts bound to D5 CSCs (>80%) much more efficiently than the binding of the sera from D5 H-TPDC–vaccinated hosts or sera from PBS-injected hosts to D5 CSCs (30.6% and 30.1%, respectively). Similarly, immune sera from SCC7 CSC-TPDC–vaccinated hosts bound to SCC7 CSCs (26.4%) significantly more than the binding of the sera from SCC7 H-TPDC–vaccinated hosts (4.0%) or the background binding by sera from PBS-injected hosts (0.9%) to SCC7 CSCs (Fig. 5B).

To evaluate the immunologic significance of the binding of CSC-primed antibody to CSCs, we examined antibody and CDC of CSCs. Immune sera from D5 CSC-TPDC–vaccinated hosts mediated significantly more efficient D5 CSC lysis than the sera collected from D5 H-TPDC–vaccinated (P = 0.002) or PBS-treated (P = 0.004) hosts (Fig. 6A). Such CDC mediated by CSC-primed antibody was CSC specific because sera from the same D5 CSC-TPDC–vaccinated hosts resulted in minimal lysis of ALDH⁻ D5 cells (Fig. 6A). Of note, this enhanced D5 CSC lysis correlated with the significantly increased protective antitumor immunity induced by D5 CSC-TPDC vaccination (Fig. 3). Similarly, immune sera from SCC7 CSC-TPDC–vaccinated hosts mediated significantly increased SCC7 CSCs lysis compared with the sera collected from SCC7 H-TPDC–vaccinated hosts (P = 0.001) or from PBS-treated hosts (P = 0.001), but not the ALDH⁻ SCC7 cells (Fig. 6B), indicating again the relative specificity of the CSC-TPDC–induced humoral response toward ALDH⁺ CSCs. This enhanced SCC7 CSC destruction correlated with the increased protective antitumor immunity induced by SCC7 CSC-TPDC vaccination (Fig. 4).

To provide further evidence that the enhanced CSC-induced antitumor immunity is due to direct targeting of CSCs, we harvested PBMCs from D5 or SCC7 CSC-TPDC–vaccinated animals to generate CTLs by activation in vitro with anti-CD3/CD28 mAb in the presence of IL-2. These activated cells were assessed for cytotoxicity against ALDH⁺ or ALDH⁻ tumor cells. D5 CSC-TPDC–primed CTLs killed D5 CSCs efficiently (approximately 60%) and significantly more than D5 H-TPDC–primed CTLs (∼20%; Fig. 7A; P = 0.003). Concurrently, the killing of ALDH⁻ D5 cells by D5 CSC-TPDC–primed CTLs was significantly less effective (∼20%; P = 0.005). In contrast, D5 H-TPDC–primed CTLs killed ALDH⁻ D5 cells (∼40%) more than D5 CSC-TPDC–primed CTLs (P = 0.03). SCC7 CSC-TPDC–primed CTLs killed SCC7 CSCs efficiently (∼60%) and significantly more than their killing of unsorted SCC7 cells (∼20%; P = 0.001) or ALDH⁻ SCC7 cells (∼25%; Fig. 7A; P = 0.002). We also carried out CTL experiments using the effector cells generated from the splenocytes harvested from the vaccination experiments shown in Fig. 3. As revealed in Fig. 7B, D5 CSC-TPDC–primed CTLs killed ALDH⁺ D5 CSCs significantly higher (P < 0.01 at all effector:target ratios) than ALDH⁻ D5 non-CSCs. In addition, we have observed a reduction of residual ALDH^high CSCs within the tumors growing in the hosts subjected to ALDH^high-TPDC vaccination compared with ALDH^low-TPDC vaccination (Supplementary Table S2).

Together, these results provide direct experimental evidence that CSCs can be destroyed by CSC vaccine–primed antibodies and T cells. The immunologic targeting of CSCs may provide a novel approach for the development of more effective cancer immunotherapies.

ALDEFLUOR/ALDH has been successfully used as a marker to isolate stem cell–enriched populations in human cancers (30–33). Using this marker, we enriched CSCs from 2 histologically distinct murine tumors to investigate immunologic strategies which may specifically target CSCs in 2 genetically different immunocompetent hosts. Schatton and colleagues recently identified a subpopulation enriched for human malignant melanoma–initiating cells (MMIC) defined by expression of the chemoresistance mediator ABCB5 (17). We have reported that ALDH is a CSC marker for human head and neck squamous cell cancers (36). Controversy exists about the existence of CSCs in human melanomas (37) as tumorigenicity of non-CSCs is evident when extreme immunocompromised hosts (NOD/SCID/IL-2rγ^null) are used. However, Civenni and colleagues recently found that the avoidance of trypsin for isolating CSCs from human melanomas which leaves intact surface epitopes on tumor cells results in the ability to isolate cells with genuine CSC properties (38). In our murine studies, we have avoided the use of trypsin to isolate CSCs from freshly harvested growing tumors. These reports that CSCs have been identified in human melanoma and squamous cell cancers provide a rationale to explore CSCs as a target for immunotherapies in our murine models.

Studies have shown that CSCs are resistant to chemotherapy (18, 20–23). These studies highlight the limitation of current cancer chemotherapies that are unable to target CSC populations. Novel therapeutic strategies are needed, particularly by targeting the CSCs. We have recently reported that CXCR1 blockade can selectively target human breast CSCs in vitro and in xenografts (34). CXCR1 is a receptor for IL-8 which stimulates the self-renewal of CSCs. In the present study, we have identified a different method to selectively target CSCs using the host immune system. We have been able to show that CSCs can be selectively targeted by both B- and T-cell mechanisms in an active immunization protocol. This was done in 2 genetically distinct immunocompetent hosts using a pulmonary metastasis model and a subcutaneous tumor model, respectively. Cancer immunotherapy using DNA vaccine (39), adoptively transferred T cells (40) or DCs (27, 41), has shown the involvement and modulation of host immune effector cells. T- and B-cell depletion in the recipient mice to show absent effects on tumor behavior would confirm that CSC vaccination induces cellular and humoral anti-CSC responses and underscore the relative role of host T and B cells in the immune response induced by CSC-based vaccines.

Several reports have described the killing of CSCs via nonspecific immune effector cells, such as natural killer (NK) cells (42). Contag and colleagues reported a modified approach to kill CSCs which involved cytokine-induced killer cells (IFN-γ and anti-CD3 activated) that were used as cellular vehicles to deliver oncolytic virus to CSCs but no CSC-specific antibodies or T cells were identified, even though immune components necessary for targeting the virus itself were examined and identified (43). Interestingly, Todaro and colleagues described killing of human colon CSCs by γδ T lymphocytes in vitro (44), but no in vivo protective or therapeutic experiments were carried out to correlate the immunologic significance of the in vitro CSC killing with potential in vivo antitumor immunity against tumor growth or development.

Our immune monitoring studies revealed direct targeting of CSCs by antibody and CTLs. This was evident by the production of IgG by the splenocytes isolated from the hosts subjected to CSC-TPDC vaccination and the binding of the antibody to the CSCs which resulted in the CSC lysis via CDC. In addition, CTLs generated from the PBMCs and splenocytes obtained from CSC-vaccinated hosts selectively killed CSCs. Pellegatta and colleagues used a murine brain tumor cell line GL261 and reported that neurospheres (NS) from glioblastoma multiforme are enriched in CSCs and that DC loaded with GL261-NS protected mice (45). However, no evidence was described for the direct targeting of CSCs, and no immunologic evaluation was conducted to elucidate the mechanisms potentially involved in the protection. Xu and colleagues reported a human glioblastoma-derived stem-like cell study in vitro (46). They used a rat 9L CSC brain tumor model and found that neurosphere-pulsed DCs prolonged survival in rats bearing 9L tumors and induced IFN-γ mRNA expression in CD8 cells. Nevertheless, no direct and specific targeting and killing of the CSCs, either by antibody-mediated cytotoxicity or by T-cell–mediated cytotoxicity was examined. Our studies provide the first direct experimental evidence that CSCs can be selectively targeted and destroyed by CSC vaccine–primed antibodies and T cells, and such immunologic targeting of CSCs is associated with enhanced in vivo CSC vaccine–conferred antitumor immunity.

The immunogenic response elicited against bulk tumor cells may be skewed toward differentiation antigens expressed on these cells. This might mask immunologic responses to CSCs which represent only a minor percentage of tumor cells. Our results suggest that CSCs, which may not express immunogenic differentiation antigens (47), can elicit an anti-CSC response when presented as a vaccine. Furthermore, as this immune response is specifically directed against the CSC populations, it may have a greater biologic effect than one directed at bulk tumor cell populations. Schatton and colleagues reported that human ABCB5⁺ melanoma-initiating cells manifested immunomodulatory functions that were immunosuppressive to T-cell activation in culture (48). Instead of an in vitro system, we used an in vivo approach to activate host T-cell and antibody responses which resulted in the induction of specific anti-CSC immunity. Furthermore, we used CSC lysate for vaccine preparation instead of using intact CSCs. Identification of the respective murine melanoma D5 and SCC7 CSC antigens responsible for the observed effects in this study warrants further investigation. We have shown herein preferential induction by ALDH⁺ cell–based DC vaccines of a humoral response to ALDH⁺ CSCs. Use of polyclonal sera from CSC-vaccinated recipients in immunoprecipitation assays may lead to the identification of differentially expressed candidate antigens in CSCs that can be validated experimentally in vaccination studies.

It is important to determine the therapeutic efficacy of established tumors with CSC-based vaccines. Because established tumors are composed of a very small percentage of CSCs, we postulate that the use of CSC-based vaccines alone will have minimal effect on established tumors and will require other adjunctive therapies. Furthermore, CSCs isolated from highly passaged cell lines or from freshly harvested tumors after inoculation may possess immune activity that is not a feature of CSCs obtained directly from spontaneously arising tumors. The lack of using spontaneous tumors remains one of the limitations of the findings in this study. Nevertheless, evaluation of the immunogenicity of CSCs from tumors derived from immunocompetent hosts is novel, and the use of CSCs in such syngeneic hosts permits the assessment of immune responses against these cells. Our animal models will allow us to explore combinatorial approaches for the therapy of more established tumors by selectively targeting CSCs.

No potential conflicts of interests were disclosed.

This work was supported by The Will and Jeanne Caldwell Endowed Research Fund of the University of Michigan Comprehensive Cancer Center, and in part by NIH grant CA82529, and the Gillson Longenbaugh Foundation.

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.