Induction of cytotoxic T lymphocytes against ovarian cancer-initiating cells



The majority of patients with stage III/IV ovarian carcinoma that respond initially to standard therapies ultimately undergo relapse due to the survival of small populations of cells with tumor-initiating potential. These ovarian cancer (OVCA)-initiating cells (OCIC) are sometimes called cancer stem cells (CSC) because they express stem cell markers, and can survive conventional therapies such as chemotherapy, which usually target rapidly replicating tumor cells, and give rise to recurrent tumors that are more chemo-resistant and more aggressive. Thus, it would be desirable to develop a therapy that could selectively target OCIC and be used to complement the conventional therapies. In this study, we isolated a subset of OVCA cells with a CD44+ phenotype in samples from patients with OVCA that possess CSC properties including the formation of spheroids in culture, self-renewal and the ability to be engrafted in immune-compromised mice. We next explored the use of immunotherapy using fusions of dendritic cells and OCIC to specifically target the OCIC subpopulations. Fusion cells (FCs) prepared in this way activated T cells to express elevated levels of IFN-γ with enhanced killing of CD44+ OVCA cells. We envision a combined approach where conventional therapies such as chemotherapy kill the bulk of tumor cells, whereas OCIC-reactive cytotoxic T lymphocytes target the resistant OCIC fraction. A combined therapy such as this may represent a promising approach for the treatment of OVCA.

Epithelial ovarian cancer (EOC) is responsible for the majority of gynecologic cancer deaths despite recent advances in adjuvant chemotherapy. Indeed, the majority of patients with Stage III/IV ovarian cancer (OVCA) that respond initially to standard therapies ultimately relapse due to drug-resistant disease refractory to further treatment with chemotherapy.1 Development of resistance is a property of a heterogeneous tumor cell population. One key tumor subpopulation in many tumors appears to be a cancer-initiating/cancer stem cell (CSC) fraction. Such CSC have been implicated in drug-resistance.1 Thus, treatment of the drug-resistant CSC fraction may be critical for the improvement of patient survival. CSC are a rare group of cancer cells found in both hematopoietic malignancies2, 3 and in solid tumors.4–13 It is believed that a hierarchal cell population exists in leukemia. At the apex of the hierarchical tree, CSC undergo asymmetrical cell division, giving rise to one daughter cell that becomes a committed progenitor, proliferating rapidly and differentiating into a heterogeneous tumor. The remaining daughter cell enters a quiescent state and exhibits CSC characteristics.14 The outlines of similar hierarchies have also been found in solid tumors and CSC have been identified in established OVCA cell lines as well as in samples from patient with OVCA.8, 9, 15 Zhang et al. obtained self-renewing and anchorage-independent spheroids by culturing patient-derived OVCA cells in stem cell-selective conditions.15 The spheroid cells preferentially expressed stem cell markers including Oct-4, nestin, Notch-1, Bmi-1, CD44 and CD177 surface markers and were able to differentiate into epithelial morphology when cultured under differentiating conditions.15 In addition, transplantation of as few as 100 such spheroid cells resulted in the formation of tumors that recapitulated the heterogeneous population of the original tumor.15 In contrast, injection of more than 105 nonselected cancer cells failed to form tumors. CSC were also isolated from malignant ovarian ascites.9 CD44+ cells isolated from such ascites possessed CSC properties including formation of spheroids even after over 20 passages and development of tumors after injection into nude mice. In addition, ovarian CSC were also identified using CD133 as marker, suggesting the heterogenous nature of CSC.16, 17 Thus, the identification of CSC described above makes it possible to develop a CSC-targeted immunotherapy. Indeed, Pellegatta et al. reported that dendritic cells (DC) loaded with lysates from GL261 neurospheres induced robust CD4 and CD8 T cells that were infiltrated into the tumor and led to the cure of 60–80% of animals with glioma.18, 19 Xu et al. found that CSC from glioblastoma multiforme expressed increased levels of tumor associated antigens as well as major histocompatibility complex (MHC) molecules and vaccination with DC pulsed with CSC antigens induced cytotoxic T lymphocytes (CTLs) response that recognized CSCs and prolonged the survival of animals bearing 9L CSC brain tumors.20 These results indicate that certain targets for immunotherapy against CSC are already known, and others, although they remain unidentified, presumably exist.

Recently, we isolated a subset of OVCA-initiating cells (OCIC) with CD44+ phenotype in samples from patients with OVCA. These cells possess the capacity of self-renewal and formed spheroids in serum-free culture. In addition, these CD44+ cells grew in immune compromised mice at low cell inocula, suggesting their high-tumor-initiating potential. We also tested the potential strategy of targeting such cells with immunotherapy. Indeed, CTL were induced capable of preferably killing the CD44+ OCIC. This is the first time that OCIC-targeted CTL have been induced.


CSC: cancer stem cells; CTL: cytotoxic T lymphocyte; DC: dendritic cell; EOC: epithelial ovarian cancer; FC: fusion cells of DC and ovarian cancer cells; MUC1: mucin 1; OCIC: ovarian cancer-initiating cells; OVCA: ovarian cancer; OVCADrug-Res: OVCA drug resistant cells; TAA: tumor-associated antigen

Material and Methods

Isolation of ovarian carcinoma cells

OVCA cells were isolated from samples derived from patients with OVCA. Briefly, the tumor tissues were weighed and minced to small pieces (1–3 mm), and then mashed through a sterile 50-μm nylon mesher (Sigma, St. Louis, MO) in tissue culture hood. Single-tumor cell suspensions or tumor cells from ascites were obtained by passing through a filter and purified through a serum column to remove dead tumor cells and other types of cells. OVCA cells (about 80–95% purity) were cultured in Dulbecco's Modified Eagle's medium (DMEM) supplemented with 10% heat-inactivated human AB serum, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin until they were sorted to subpopulations, and used as fusion partner or CTL targets.

Generation of human DC, monocytes and T cells

Peripheral blood mononuclear cells (PBMC) were isolated from leucopack obtained from healthy donors (HLA-A*0201) using Ficoll density gradient centrifugation (Ficoll-Paque™ plus, GE healthcare Bio-Sciences AB, Sweden). PBMC were collected and plated in tissue culture dish with 5% human AB male serum (Sigma) in RPMI 1640 medium containing 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin for 1 hr in a humidified CO2 incubator. The loosely adherent fraction was cultured in 1,000 U/ml GM-CSF (Genzyme, Framingham, MA) and 500 U/ml IL-4 (R&D Systems, Minneapolis, MN) RPMI/AIM-V (1:1) medium with 1% of human AB male serum for 5 days. On day 4, the loosely adherent cells were collected and further enriched through repeated adherence method for fusion use. The nonadherent cells were frozen with 10% dimethylsulfoxide (DMSO) in human AB serum and used as source of T cells, and adherent cells used as monocytes for CTL control target.

Cell sorting to select CD24+ or CD44+ cells for fusion with DC

The OVCA cells were collected and dual-stained with Fluorescein isothiocyanate (FITC)-anti-CD24 (ML5) and PE-anti-CD44 (G44-26, BD Pharmingen, San Diego, CA) monocolonal antibody (mAbs) for 1 hr on ice. The stained OVCA subsets (CD24 or CD44 cells) were sorted into separate tubes by MoFlo (Cytomation, Fort Collins, CO) with Summit v3.0 analysis software. The purity of the sorted cells was checked by fluorescence-activated cell sorting (FACS) analysis. The CD24+, CD44+ and bulk OVCA cells were fused to DC to generate FC/CD24, FC/CD44 and FC/OVCA, respectively, using the method previously described.21 Briefly, the mixture of DC and subsets of OVCA cells was resuspended in a polyethylene glycol solution (PEG, MW1450, Sigma-Aldrich) for 3–5 min at room temperature, and then progressively added prewarmed serum-free RPMI medium to dilute the PEG in the next 5 min. After washing, FCs were resuspended in RPMI 1640 medium supplemented with 5% human AB serum and 500 U/ml GM-CSF, and cultured in 5% CO2 at 37°C for 5 days. In some experiments, DC were also fused to OVCA cells surviving carboplatin treatment [FC/OVCA drug resistant cells (OVCADrug-Res)].

Dose response of ovarian cancer cells to carboplatin

To determine the sensitivity of CSCs to chemo-agent, ovarian CD44+ and non-CD44 cells were seeded in four replicates into 96-well tissue culture plates (1 × 104/well), carboplatin (CARB, MP Biomedicals, Solon, OH) was added to each well at concentrations of 5, 10, 25, 50, 100, 250 and 500 μM. Cells incubated in culture medium alone served as a control for cell viability (untreated wells). Cells were incubated at 37°C. Seventy-two hours after incubation, 20 μL aliquot of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) solution (5 mg/mL in PBS) was added to each well and reincubated for 4 hr, followed by centrifugation at 800 rpm for 5 min. The supernatant in the culture was carefully aspirated, and then 200 μL of DMSO was added to each well and shaked for 10 min to dissolve the formazan crystals. The absorbance was measured at 590 nm by a micro-plate reader.22, 23 Cell viability rate was calculated as the percentage of MTT absorption as follows: % survival = (mean experimental absorbance/mean control absorbance) × 100.

To select drug resistant ovarian tumor cells, carboplatin at 50 μM, the concentration that causes 50% growth inhibition (IC50) were added into the culture medium and then incubated for 48 hr. Live cells were selected by Ficoll purification and then used for immunocytochemical staining and FACS.

Immunohistochemical staining

Cytocentrifuged cells or frozen sections were air-dried, fixed in acetone and incubated with anti-CD44, CD133 (M-286, Santa Cruz, CA), CK19 (BA17, Santa Cruz), CA-125 (NCL-L-CA125, Novocastra Laboratories, UK) or mucin 1 (MUC1) (HMPV, BD Pharmingen) antibodies at 1:100 dilutions for 1 hr at room temperature. After the cells were washed, biotinylated horse anti-mouse IgG was applied for an additional 45 min. The cells were stained red with ABC solution (Vector Laboratories, Burlingame, CA). The slides were then incubated with anti-HLA-DR for 1 hr followed by alkaline phosphatase (AP-ABC) solution (Vector Laboratories) to generate a blue color.

Phenotype of OVCA and fusion cells

Sorted CD24+ and CD44+ cells were incubated with anti-human mAbs against MUC1, CA-125, CD133, CK19, ESA (VU-1D9, Millipore) or HLA-A2 (0791HA) (One lambda, Canoga park, CA) for 1 hr, and then stained with FITC-conjugated anti-mouse IgG (Chemicon International, Temecula, CA). In addition, the FCs were dual-stained with FITC-conjugated anti-CD44 and PE-conjugated anti-HLA-DR (TU36), anti-CD86 (IT2.2) or CD80 (L307.4) (BD Pharmingen). Cells were fixed in 2% paraformaldehyde and analyzed by flow cytometry using CellQuest software (BD Biosciences).

T cell proliferation

T cells (nonadherent cell population) were cocultured with OVCA, FC/OVCA, FC/CD24 or FC/CD44 in RPMI-1640 medium containing 10% human AB serum, 10 U/ml human IL-2, 15 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin and 5 × 10−5 M β-mercaptoethanol at 96-well U-bottom plates in a volume of 200 μl/well for 5 days. T-cell proliferation was assessed by [3H]thymidine incorporation after an additional 12-hr incubation with 1 μCi/well of [3H]thymidine. Radioactivity (mean ± SD of triplicates) was measured by liquid scintillation counting.

IFN-γ or tetramer staining

T cells (HLA-A*0201) were stimulated by OVCA, FC/OVCA, FC/CD24 or FC/CD44 cells in the presence of 10 U/ml human IL-2 in RPMI medium containing 10% human AB serum for 5 days. On day 5, the T cells were collected and purified through a nylon wool column to remove OVCA or FCs, and then stained with anti-human CD4/interferon (IFN)-γ or anti-human CD8/IFN-γ (BD Pharmingen) according to the manufacture instruction. For tetramer staining, T cells obtained after passing through nylon wool column were incubated with PE-conjugated MUC1 tetramer (HLA-A*0201, STAPPVHNV) (PROIMMUNE, Oxford, UK) for 1 hr at 4°C. After wash, the T cells were further stained with FITC-conjugated anti-CD8 mAb for 40 min at 4°C. Cells were washed and fixed with 2% paraformaldehyde, and analyzed by flow cytometry using CellQuest software (BD Biosciences).

CTL assay

T cells (HLA-A*0201) were stimulated by OVCA cells or various DC-based vaccines in the presence of 10 U/ml human IL-2 in RPMI medium containing 10% human AB serum. After 5 day culture, T cells were purified through a nylon wool column and used as effector cells. Patient-derived OVCA, sorted CD24+ and CD44+ cells (HLA-A*0201), and monocytes were labeled with 100 μCi Namath imageCrO4 at 37°C for 60 min, and then washed to remove unincorporated isotope. In some experiments, OVCA cells surviving the treatment of carboplatin (OVCADrug-Res) were also used as a target. The effector T cells (E) or tumor target cells (T) were resuspended in CTL assay medium at 50:1 (E:T) ratios and placed in 96-well V-bottom plates for 5 hr at 37°C. After incubation, the supernatants were collected and radioactivity was quantitated in a gamma counter. Spontaneous release of 51Cr were determined by incubation of targets in the absence of effectors, and maximum or total release of 51Cr by incubation of targets in 0.1% Triton X-100. Percentage-specific 51Cr release was calculated using the following equation: percent specific release = [(experimental − spontaneous)/(maximum − spontaneous)] × 100.

Statistical analysis

Statistical significance was analyzed using χ2, Student's t test or One-way ANOVA. p < 0.05 indicates statistical significance.


Selection and characterization of ovarian OCIC derived from patients

We first examined primary human OVCA tissue for cells with OCIC potential. Single cell suspensions of human OVCA cells were obtained by mechanical disaggregation of EOC samples as previously described.24, 25 CD44 was chosen as a marker for OCIC, whereas high expression of CD24 was chosen as a marker for more differentiated cancer cells. CD44 has been extensively reported as a marker for CSC in hematologic and solid tumors.26 The CD44+ and CD24+ subsets were selected by cell sorting and then cultured in vitro under conditions that promote spheroid formation (Fig. 1a). Spheroids were observed after culture of CD44+ cells under these conditions, but not in CD24+ cells (Fig. 1b), indicating that these cell populations have different proliferative characteristics.

Figure 1.

Phenotypic and functional properties of OCIC from human ovarian cancer (tumor) and malignant ascites (ascites). Ovarian cancer cells were isolated and cultured in 10% FCS in DMEM for 1–2 days. (a) The cells were stained with anti-CD24 and anti-CD44 mAbs and sorted into CD44+ and CD24+ subsets. (b) Spheroid formation by CD44+ cells. CD24+ (lower panel) and CD44+ (upper panel) cells were cultured, respectively, and spheroids were observed under inverted microscopy. (c and d) Distinct phenotypes of CD44+ and CD24+ cells in solid tumor or malignant ascites. Ovarian cancer cells from recurrent solid ovarian tumor (upper panel) or ascites (lower panel) were isolated, and the CD44+ and CD24+ cells were selected by cell sorting. The sorted CD44+ and CD24+ cells were further analyzed by FACS for co-expression of CD133, CK19, ESA and CA125. lgG (green line) was used as a control. (d) Percentage of cancer cells double positive for CD44 and CD133, CK19, ESA, MUC1 or CA125 in samples from three patients was presented.

CD44+ OVCA cells have been reported to be enriched in the malignant fluid of hosts with OVCA.9 We, therefore, next compared the percentage of CD44+ cells from ovarian solid tumors and malignant fluid. OVCA cells were isolated both from solid tumor and malignant fluid samples derived from the same patient, stained with anti-CD44 and anti-CD24 mAbs and analyzed by FACS. As shown in Figure 1c, ∼2.66% and 18.6% of CD44+ cells were found in solid tumor and malignant fluid, respectively, suggesting the enrichment of CD44+ cells in malignant fluid (Fig. 1c). The sorted CD44+ and CD24+ subsets from both solid and ascites were next stained with mAbs against stem cell markers CD133, CK19 or ESA and expression of these proteins analyzed by flow cytometry. High levels of expression of CD133, CK19 or ESA were observed on CD44+ cells (Figs. 1c and 1d). By contrast, these markers were expressed at minimal levels on CD24+ cells, indicating the existence of distinct cell surface protein profiles between CD44+ and CD24+ cells. In addition, the tumor-associated antigens (TAA) CA125 and MUC1 were also expressed on the surface of CD44+ cells (Fig. 1d).

Engraftment of CD44+ ovarian cancer cells

Next, to determine the tumor initiating ability of CD44+ cells, cancer cells isolated from patient-derived ovarian tumors were sorted into CD44+ and CD24+ subsets. The cells were injected subcutaneously into NCR Nude mice. Inoculation of such mice with as few as 100 ovarian CD44+ cells resulted in growth of tumor subcutaneously (Fig. 2a and Supporting Information Table 1). In contrast, there was no tumor growth in mice injected with as many as 10,000 CD24+ cells, suggesting differential tumor initiating capacity between CD44+ and CD24+ cells. In addition, intraperitoneal injection of 10,000 CD44+, but not non-CD44, cells resulted in ovarian tumor growth in the abdominal cavity (Fig. 2b). Furthermore, the engrafted tumors derived from these cells resembled the primary tumors morphologically and phenotypically indicating the potential of these OCIC to give rise to tumors with a mixed population of cancer stem and differentiated cells (Fig. 2c). The tumor cells were round or oval with dark stained nuclei, formed glandular structure and infiltrated the basal membrane. In addition, both primary and engrafted tumors contained cells positive for CD44, CD133, CK19, CA125 or MUC1. These experiments indicate that ovarian tumor-derived CD44+ cells are endowed with tumorigenic potential and are thus functionally distinct from CD24+ cells.

Figure 2.

Growth of ovarian cancer-initiating cells in NCR Nude mice. (a) Groups of mice were subcutaneously injected with 102, 103 or 104 CD44+ cells mixed with Matrigel (1:1 ratio) on the left back of NCR Nude mice, and on the opposite side were injected with the same number of CD24+ cells. (b) The 104 CD44+ or non-CD44 cells were intraperitoneally injected into NCR Nude mice. The mice were followed for 1–2 months and tumor growth was then determined. Tumor growth was indicated by arrow. Subcutaneous (a, right panel) or intraperitoneal engraftment (b, right panel) was summarized from three and two samples, respectively. (c) The engrafted tumor tissue collected from NCR Nude mice (low panel) and the primary OVCA tumor from the same patient (upper panel) were processed for frozen sections, which were stained with H&E or the indicated antibodies by immunohistochemical (IHC) staining. Red color indicates positive tumor cells for individual antibody (×60).

Differential chemoresistance between CD44 and non-CD44 subpopulations

CSC are implicated in the development of chemoresistant tumor cell populations.9 We next assessed the sensitivity of CD44 positive and negative cells to a drug commonly used in the treatment of OVCA. OVCA cells were treated with carboplatin at various doses and the cell survival was measured by the MTT assay. Inactivation of OVCA cells was correlated with the dose of carboplatin (Fig. 3a). Next, the unsorted OVCA as well as CD44 positive and CD44 negative cells were incubated with carboplatin at the indicated concentrations. The IC50 for unsorted OVCA, CD44 positive and CD44 negative cells was 50.7, 114.4 and 28.2 μM, respectively, suggesting that CD44 positive cells were more resistant to carboplatin (Fig. 3b). We also examined whether cancer cells expressing stem cell markers were enriched in cells surviving chemo-treatment. OVCA cells isolated from the primary tumor sample, without a previous history of chemotherapy were treated with 50 μM carboplatin, the drug dose that 50% of tumor cells survived, for 2 days. On day 3, surviving cells or untreated controls were stained with anti-CD133, anti-CK19, anti-MUC1 and anti-CA125 antibodies (Fig. 3c). From Figures 3c and 3d, we can observe that stem cell marker CD133 or CK19 positive cells are significantly increased in the carboplatin-treated cells compared with the control group, suggesting the enrichment of the subpopulation expressing stem cell marker in the chemoresistant clones. In addition, the fraction of cells expressing cell surface tumor antigens MUC1 and CA125 was also increased after chemo-treatment.

Figure 3.

Differential drug sensitivity of CD44+ or non-CD44 OVCA cells. (a) Dose response of OVCA cells treated with carboplatin. OVCA cells isolated from solid tumors were treated with indicated concentrations of carboplatin and cell inactivation was determined by MTT assay. (b) The sorted CD44+, non-CD44 cancer cells or unsorted OVCA cells were cultured with the indicated concentrations of carboplatin. Drug-induced growth inhibition is shown 72 hr later by MTT assay in repeated experiments. The statistical significance of the differences between the three groups was determined by One-way ANOVA. (c and d) Ovarian cancer cells isolated from primary tumor samples without prior history of chemotherapy were treated with 50 μM carboplatin for 48 hr. After a 2 day treatment, the OVCA cells without or with drug treatment were collected and dead cells removed by Ficoll. The cells were stained with the indicated antibodies by IHC staining and photographed under microscopy. Red color indicates positive cells. (d) The cells positive for the indicated antigens were counted and the percentage of positive cells from two samples without or with drug treatment was presented as a bar graph. Significance was determined by χ2-test.

Generation of DC-OCIC fusion cells

We next examined the possibility of selectively targeting these chemotherapy resistant OCIC by immunotherapy. We used the tumor-DC fusion approach to generate anti-OVCA cell immunity. Human DC were generated from PBMC as previously described.24, 25 Such DC expressed the surface marker HLA-DR, whereas OVCA cells were negative for HLA-DR but positive for surface antigen MUC1. Fusion of DC and OVCA cells created a heterokaryon that expressed both tumor cell and DC markers HLA-DR and MUC1 (Fig. 4a). In addition, CD44+ and CD24+ OVCA cells were selected by cell sorting and then fused to DC to generate both FC/CD44 and FC/CD24 FCs, respectively (Fig. 4a). FC/CD44 FCs were distinguished from other FCs through phenotypic markers that are preferentially expressed on CD44+ tumor cells. Figure 4b shows that HLA-DR (DC marker) and CD133 or CK19 (CSC markers) were co-expressed on the FC/CD44 cells. By contrast, minimal levels of co-expression of HLA-DR and CD133 or CK19 were observed on FC/CD24 cells. In addition, an increase in the number of cells double-positive for CD44 and DC cell surface markers MHC Class II, CD80 or CD86 was observed in FC/CD44 FCs greatly exceeded those in FC/CD24 cells by FACS analysis (Fig. 4c). FC/CD44 cells were shown to contain high levels of double-positive cells when CD133 or CK19 were used as alternative ovarian CSC markers (Fig. 4d). Thus, fusion of DC to different subsets of OVCA cells resulted in FCs expressing different markers, and we could obtain heterokaryons enriched in CSC markers when the CD44+ subpopulation was used. Such FCs could thus be used to potentially target OCIC population.

Figure 4.

Characterization of DC-tumor fusion cells. (a) DC (HLA-A*0201+) were fused with unsorted ovarian cancer cells (OVCA), CD24+ or CD44+ cells from patients to generate FC/OVCA, FC/CD24 and FC/CD44 fusion cells, respectively. The cells were double stained with anti-MUC1 (red color) and anti-HLA-DR (blue color) mAbs using IHC staining and examined under microscopy (×60, ×100). (b) Coexpression of HLA-DR and CK19 or CD133 on fusion cells. The indicated fusion cells were stained with mAbs against HLA-DR (blue color) and CK19 or CD133 (red color) using IHC staining. (c) Coexpression of CD44 and HLA-DR, CD80 or CD86 on fusion cells. FC/OVCA, FC/CD24 or FC/CD44 fusion cells were double stained with mAbs against CD44 and HLA-DR, CD80 or CD86, and then analyzed by FCAS. (d) The percentages of cells double positive for HLA-DR and CD133 or CK19 in representative experiment are presented in the bar graph.

Differential quality and quantity of T cells stimulated by fusion cells

DC-tumor FCs efficiently process and present tumor antigens including those previously unknown to CD4 and CD8 T cells.21 To determine the ability of the DC-tumor FCs described above to activate T cells, we first measured their capacity to increase T cell proliferation. T cells (HLA-A*0201+) isolated from the same donor from whom the DC described above were generated and used as fusion partners were stimulated by FC vaccines. FC vaccines stimulated efficient T cell proliferation, which was greater in magnitude than the lower levels of cell division stimulated by OVCA cells or medium plus IL-2 (Fig. 5a). Although higher levels of T cell proliferation were observed in T cells stimulated by FC/CD44 than those stimulated by FC/CD24, no statistical difference was found in the magnitude of T cell proliferation between these different vaccine groups (Fig. 5a). We next examined the numbers of effector T cells after incubation with the various vaccines using IFN-γ expression as a marker of functionally active T cells. In general, more IFN-γ expressing T cells were stimulated by FCs than those stimulated by tumor cells (Fig. 5b). However, FC/CD44 stimulated the highest fraction of T cells to express IFN-γ (Fig. 5b). FC/CD44, FC/OVCA or FC/CD24 cells stimulated 27.5, 12.5 and 2.63%, respectively, of CD8 T cells expressing IFN-γ (Fig. 5b). The difference between IFN-γ expressing CD4 (Fig. 5c, left panel) and CD8 (Fig. 5c, right panel) T cells stimulated by FC/CD44 and the other vaccines was statistically significant. Moreover, such FC/CD44 stimulated antigen-specific T cells as demonstrated by MUC1 tetramer analysis, with the highest level of MUC1-specific T cells induced by the FC/CD44 FCs (Figs. 5d and 5e). By contrast, T cell populations stimulated by OVCA cells contained minimal antigen-specific T cells. Together, these results indicate that FC vaccines are capable of induction of tumor antigen-specific T cells.

Figure 5.

Characterization of T cells stimulated by DC-CSC fusion vaccines. (a) Proliferation of T cells from three samples. T cells (HLA-A*0201+) isolated from the same donor from whom DC were generated and used for fusion partner were stimulated by FC/OVCA (▪), FC/CD24 (▴), FC/CD44 (•) or OVCA cells alone (♦) for 5 days followed by pulsing with 1 μCi/well of 3H-thymidine. T cell proliferation was measured by liquid scintillation counter and mean ± SD of triplicates was presented. (b and c) Expression of IFN-γ in stimulated CD4 and CD8 T cells. T cells purified through nylon wool after 5 days stimulation with FC/OVCA, FC/CD24, FC/CD44 or OVCA cells alone, respectively, were double stained with mAbs against IFN-γ and either CD4 or CD8, and analyzed by FACS. The IFN-γ-positive CD4 (c, left panel) or CD8 (c, right panel) cells from three samples are presented in the bar graph. (d and e) Antigen-specific T cells induce by FC vaccines. T cells (HLA-A*0201+) were stimulated by FC/OVCA, FC/CD24, FC/CD44 or OVCA cells for 5 days, purified through nylon wool and then stained with MUC1 peptide tetramer. The T cells were further stained with FITC-conjugated anti-CD8 antibody and analyzed by flow cytometry. T cells positive for a MUC1 peptide tetramer from three samples were presented in the bar graph. Statistical significance was determined by χ2-test. [Color figure can be viewed in the online issue, which is available at]

Preferential killing of CD44+ ovarian cancer cells by T cells stimulated by FC/CD44 fusion cells

We next measured the CTL activity of FC-stimulated T cells against unsorted OVCA cells as well as selected subpopulations. CD8 T lymphocytes were incubated with labeled tumor cells from the different groups and levels of lysis assayed. CTL induced by FC/OVCA vaccine lysed bulk population of OVCA cells. However, their lysis of CSC was significantly reduced (Fig. 6a). By contrast, CTL stimulated by FC/CD44 vaccine consistently exhibited the highest CTL activity against CSC and a lesser activity against the bulk population of OVCA cells (Figs. 6a and 6b). The lysis of CTL induced by FC/CD24 fusion or OVCA cells against CD44+ OVCA cells was relatively low, although variable levels of killing of the other targets, CD24+ or unsorted OVCA cells were observed (Figs. 6a and 6b). We also examined killing of an irrelevant cell line that does not share tumor antigens with OVCA. A slightly above-background killing of K562 cells was also observed, suggesting the involvement of natural killer cells in toxicity. However, there was only background level of CTL activity against CD44 positive monocytes, suggesting minimal cross reactivity of FC/CD44-activated CTL with normal cells even when expressing the CD44 marker (Fig. 6b, second right panel). To determine whether CTL induced by FC/CD44 can preferentially kill CD44+ cancer cells as well as chemo-resistant cancer cells, T cells stimulated by FC/CD44 or fusions of DC and drug resistant OVCA cells (FC/ OVCADrug-Res) were incubated with OVCA cells pretreated with carboplatin or CD44 positive cancer cells. High levels of lysis of chemo-resistant or CD44+ cancer cells by T cells induced by FC/CD44 or FC/OVCADrug-Res was observed, suggesting the existence of overlapping of antigens expressed by the chemo-resistant population and the sorted CD44+ population (Fig. 6c). In contrast, much low levels of CTL activity were observed against CD44+ or drug resistant OVCA (OVCADrug-Res) cells in T cells induced by OVCA cells, FC/OVCA, FC/CD24 or DC pulsed with cell lysates from OVCA cells pretreated without or with carboplatin. Therefore, these experiments indicate that we can generate a FC vaccine that is able to induce T cells with superior immune killing activity compared with those stimulated by other vaccines. More importantly, the FC/CD44 vaccine induced CTL are able to preferentially kill OVCA cells expressing stem cell marker and cells that have become chemoresistant. These findings suggest that these CTL can be used to target CSCs or chemo-resistant cancer cells within the OVCA, and thus may be complementary to conventional therapies. These results provide a proof of principle to support our hypothesis that OCIC-reactive CTL can be induced by a tumor-DC FC vaccine approach.

Figure 6.

Preferential killing of OCIC by T cells induced by FC/CD44. (a) T cells stimulated by FC/OVCA, FC/CD24, FC/CD44, OVCA cells or medium alone, respectively, for 5 days, were incubated with 51Cr-labeled OVCA (unsorted, hatched bar), sorted CD44 (black bar), CD24 (gray bar) cells or monocytes (white bar) at a ratio 50:1. CTL activity was measured in triplicate and is presented as the mean ± SD. (b) T cells (HLA-A*0201+) stimulated by FC/OVCA, FC/CD24, FC/CD44 or OVCA alone were incubated with 51Cr-labeled OVCA, sorted CD44+, CD24+ cells, CD44-positive monocytes and K562 target. CTL activity was determined by the standard 51Cr release assay in triplicates and data are presented as the mean ± SD. (c) Enhanced lysis of drug resistant tumor cells by CTL induced by FC/CD44. The HLA-A*0201+ T cells were co-cultured with HLA-A2 matched OVCA cells, OVCADrug-Res and their respective fusion cells (FC/OVCA and FC/OVCADrug-Res). In addition, T cells were also co-cultured with FC/CD44, FC/non-CD44 and DC pulsed with OVCA cell lysates from cells pretreated without or with carboplatin (DC/OVCALys or DC/OVCADrug-Res-Lys). Four days after culture, T cells were collected and analyzed for killing ability against OVCA CD44+ or drug resistant (OVCADrug-Res) cell targets at 50:1 ratio by standard 51Cr release assay. The percentage of CTL activity from two blood samples is presented in bar graph and statistical analysis was determined by χ2-test. [Color figure can be viewed in the online issue, which is available at]


OVCA cells expressing stem cell markers have been identified in established OVCA cell line and patient-derived ovarian tumor samples.8, 9, 15 This study shows differential tumor initiation capacity between subpopulations of OVCA cells. The subpopulation of cancer cells bearing stem cell marker, although variable among samples, can be identified and found to possess high tumorigenicity. In contrast, OVCA cells expressing CD24 failed to be engrafted (Fig. 2). These results suggest that OVCA is heterogeneous and the subpopulation of cells expressing CD44 are tumor-initiating cells. Furthermore, these cancer-initiating cells are more resistant to carboplatin, a commonly used drug in treatment of OVCA (Fig. 3). Treatment with this drug preferentially selected cells expressing stem cell markers (Fig. 3). Thus, OVCA cells expressing stem cell markers are linked to chemo-resistance, a major challenge in the treatment of OVCA.1 The mechanisms underlying the chemo-resistance in CSC have not been fully elucidated. However, members of the ATP-binding cassette transporter (ABC transporters) super family have been implicated in the development of drug resistance.27–29 These energy-dependent transporter proteins are highly expressed in CSCs and work individually or in concert to pump cytotoxic drugs out of cells, thus reducing intracellular levels and diminishing the ability of these agents to cause tumor cell death. CSC also tend to proliferate slowly, thus reducing their sensitivity to agents that target proliferating cells. Thus, CSC may be important factors in the development of chemoresistance and in the relapse of cancers. Thus, it would be desirable to develop therapies that specifically kill CSC. Ideally, CSCs could be targeted through the molecules in their intrinsic signaling pathway essential for their accumulation and survival. However, such molecular-targeted therapy remains to be developed.14 Alternatively, CSC-targeted immunotherapy including DC fused to CSC or chemo-resistant tumor cells could be used because this approach represents a noncross-resistant strategy that could be complementary to conventional therapy.

The potential of immunotherapy in the management of OVCA is based on the findings that EOC expresses well-defined target antigens that are capable of stimulating an antitumor immune response.30 OVCA cells express mutated forms of p5331–33 and/or BRCA-134, 35 tumor suppressor genes. Ovarian tumor cells also express high levels of the CA-12536–38 and MUC139 TAA. In addition, these tumors express abundant levels of the HER2/neu (c-erB2) and epidermal growth factor receptors.40–42 Consistent with these findings, recent studies show that the number of tumor infiltrating lymphocytes (TILs) is an independent indicator of prognosis for patients with OVCA.43–45 Further studies show that both the quantity and the quality of TILs are important factors in determining the prognosis in OVCA.46–48 Thus, increase in the potency and specificity of T cells has potential to improve the clinical outcomes for patient with OVCA.

Our strategy is based on the observation that tumorigenicity of cancer cells is not created equal and only a subpopulation of tumor cells are endowed with capacity to initiate tumor and implicated in chemo-resistance. We hypothesize that CTL targeting this subpopulation of OVCA cells can be induced with enhanced therapeutic value. OCIC or chemo-resistant tumor-specific antigens would be an attractive target for immunotherapy. However, such antigens remain to be identified. In this context, DC-CSC fusion vaccine as described here could be a promising alternative approach because the whole CSC antigenic repertoire including those specific to chemo-resistant tumor cells can potentially be channeled to the endogenous antigen processing and presenting machinery of DC without the need to identify them.21 In addition, polyclonal CTL can be induced to maximize the potency of cellular immunotherapy. In the current study, we enriched a subset of OVCA cells that expressed stem cell markers and were resistant to chemo-agents (Fig. 3). These cells, which we refer to as OCIC or CSC were fused to DC to produce FCs that differ from previous DC-tumor FC vaccines. When OCIC were isolated, enriched and fused to DC to generate DC-OCIC FCs, CTL induced by these FCs were highly selective for OCIC killing and were less effective against the bulk tumor cell population. By contrast, when bulk tumor cell populations isolated from the OVCAs were fused to DC, the CTL induced by such FCs were directed mainly against the bulk of OVCA cell population which presumably contains progenitor cells, differentiated or differentiating tumor cells. These preparations were less effective against the cancer-initiating CSC.

One major concern for this approach is the specificity of T cells stimulated by DC-OCIC fusion vaccine. It is possible that OCIC share common antigens with normal cells, similar to the TAA such as MUC1 shared by OVCA and normal gland tissues. Indeed, MUC1 was expressed in MCF7 side population cells with stem cell characteristics, suggesting that there is overlapping expression of TAA between CSC and more differentiated subpopulations.49 In this scenario, T cells activated by FC/CD44 would be cross reactive with normal tissues that express the same antigen, resulting in autoimmune disease. Such complications, although theoretically possible, rarely occur in immunotherapy against TAA. For example we failed, in both animal and human studies50 to observe symptoms of autoimmune disease against MUC1 using a tumor vaccine targeting MUC1 (J. Gong, unpublished data). In this study, we observed minimal cross-reactivity of T cells induced by FC/CD44 FCs to CD44-positive monocytes (Fig. 6). Nevertheless, further study would be needed to allay the concern of autoimmune disease before OCIC-reactive T cells were used in the clinics.

We have shown, therefore, that human ovarian tumors contain cell populations with tumor initiating properties and that these are enriched after chemotherapy. We have also shown that such cells can be targeted by tumor-DC fusion vaccines offering the promise of selective targeting of cancer-initiating cell/CSC in cancer as an adjuvant treatment for OVCA.