Eradication of Ovarian Tumor Xenografts by Locoregional Administration of Targeted Immunotherapy

Oligodeoxynucleotides containing dinucleotides with unmethylated CpG motifs (CpG-ODN) are strong activators of both innate and adaptive immune systems (1, 2). Recognition of CpG-ODN is mediated by Toll-like receptor 9 (TLR9), a member of the TLR family, which is critically important in detecting microbial pathogens. TLRs, initially identified on cells of the immune system, are also expressed by nonprofessional immune cells such as endothelial cells, fibroblasts, and epithelial cells (3, 4). Both bone marrow and non-bone marrow-derived cells are thought to be involved in the response induced by TLR agonists.

Successes in preclinical studies using CpG-ODN and early indications of its safe use in humans have led to considerable interest in the clinical development of these agents in the treatment of cancer patients (1, 5). Preclinical studies (6–8) have shown superior antitumor effects after intratumor administration of CpG-ODN; thus far, a mechanistic explanation of the route-dependent effects remains to be elucidated. The critical role ascribed to the adaptive immunity in CpG-ODN antitumor effects have led to clinical trials conducted using systemic administrations (9–11). In clinical practice, ovarian cancer is among the few tumor types suitable for intratumor treatment, because its growth is mostly confined to the peritoneal cavity. In such patients, locoregional delivery of therapeutic agents may be a suitable option.

Ovarian cancer is the fifth most frequent cancer, accounting for about 6% of all cancer death in females. Progress in the treatment of the disease has been made in recent years, with the current 5-year survival rate around 45% (12). Cytoreductive surgery and systemic combination chemotherapy with a platinum drug and a taxane represent the standard of care for ovarian cancer patients (13). The role of immunotherapy for ovarian cancer patients has been addressed in an early clinical trial (14), but to our knowledge the efficacy of CpG-ODN against ovarian tumors has not been addressed clinically.

Here, we examined the effects of CpG-ODN on human ovarian cancers in the peritoneal cavity of nude mice and the role of the administration route in determining these effects. The results indicate the superior antitumor activity after i.p. administration compared with other routes of delivery, with increased mouse survival time and tumor-free rates. Analyses of the cellular and immunologic bases of this effect point to the crucial role of innate effector cells amplification at the site of tumor growth.

Mice. All experiments were carried out using 8- to 10-week-old female Swiss nude mice (Charles River). Mice were maintained in laminar flow rooms at constant temperature and humidity, with food and water given ad libitum. Experiments were approved by the Ethics Committee for Animal Experimentation of the Istituto Nazionale Tumori of Milan according to institutional guidelines.

Oligodeoxynucleotides. Purified, phosphorothioated ODN1826 (5′-TCCATGACGTTCCTGACGTT-3′) containing CpG motifs and control ODN2243 lacking a CpG motif were synthesized by Coley-Pharmaceutical Group. Phosphorothioate modification was used to reduce susceptibility of the oligodeoxynucleotide to DNase digestion, thereby significantly prolonging its in vivo half-life. CpG-ODN was administered at a dose of 20 μg/mouse in 0.2 mL saline based on a dose-response curve assessing the effects on keratinocyte chemokine (KC) production (15).

Tumors. IGROV-1 (16), SK-OV-3 (17), OVCAR-5 (18), and OvCa432 (19) human ovarian carcinoma cells were used in the study. The IGROV-1 tumor was adapted to grow i.p. and maintained by serial i.p. passage of ascitic cells into healthy mice as described (20). For experiments here, 2.5 × 10⁶ ascitic cells in 0.2 mL saline were injected i.p. into mice. For experiments with SK-OV-3, OVCAR-5, and OvCa432 tumors, 10⁷ exponentially growing cells from culture, suspended in 0.4 mL of saline, were inoculated i.p. In three of the four tumor systems, hemorrhagic ascites with diffuse peritoneal carcinomatosis develops and the animals eventually die. In OvCa432-bearing mice, only 75% developed peritoneal tumors and no mouse died for tumor burden by 180 days.

Experimental groups (5-7 mice) were inspected daily and weighed three times weekly. Survival was the endpoint of the study and the median day of death (median survival time) was calculated for each group. Antitumor activity was assessed as T/C%, the ratio of median survival time in treated mice with that in control mice × 100. Mice alive at the end of experiment were considered “long-term survivors” (LTS) and “cures” if tumor-free at necropsy.

KC quantitation. Sera from healthy or tumor-bearing mice (injected 10 days before with IGROV-1 cells) were collected 90 min (15) after a single CpG-ODN treatment and evaluated for KC levels using ELISA kits (R&D Systems) according to the manufacturer's recommendations. From healthy mice, parietal mesothelial peritoneal tissue under abdominal transversal muscle near to the linea alba was also recovered. Samples were washed twice in saline before chemokines were extracted using a modification of the PERFEXT method (21). KC concentration was normalized to total protein content. Contamination of intestinal tissues with blood was evaluated based on hemoglobin levels assessed using a Hitachi Modular Analyzer with optimized reagents (Roche Diagnostics).

Isolation of peritoneal cells. Healthy nude mice treated with saline or CpG-ODN were sacrificed by cervical dislocation and peritoneal washings were done by i.p. injection and withdrawal of saline with a 22-gauge needle. The lavage fluid was transferred to a centrifuge tube maintained on ice. RBC were lysed by incubation in ammonium chloride lysing solution [0.15 mol/L NH₄Cl, 10 nmol/L KHCO₃, 1 nmol/L Na₄EDTA (pH 7.2)] for 10 min at 4°C and preparations were washed twice in PBS. Cells were counted by the trypan blue exclusion assay.

Flow cytometry. Total peritoneal cells were incubated for 30 min at 4°C with antibodies PE-F480 (Invitrogen, Caltag Laboratories), PE-NK1.1 (PharMingen), or PE-GR-1 (Southern Biotech) to detect macrophages, natural killer (NK) cells, or granulocyte markers, respectively. After washing with PBS-5% FCS, cells were assessed by flow cytometry (FACSCalibur, BD Biosciences) and CELLQUEST software (BD Biosciences) for data analysis.

Cytotoxicity assay. IGROV-1 cells were labeled with 100 μCi ⁵¹Cr for 1 h at 37°C, washed three times with PBS-5% FCS, and coincubated with effector cells (mouse spleen or peritoneal cells treated with hypotonic shock to lyse red cells) in 200 μL RPMI 1640, 10% fetal bovine serum, and 1% glutamine, in triplicate 96-well U-bottomed plates for 4 h at 37°C. Effector-to-target ratio was 50:1. Radioactivity of the supernatant (70 μL) was measured with a Trilux Beta Scintillation Counter (Wallac). Specific lysis was calculated as % specific lysis = 100 × (experimental cpm - spontaneous cpm) / (maximum cpm - spontaneous cpm).

Proliferation assay. Total peritoneal cells, obtained by peritoneal washing, and adherent macrophages, obtained by seeding total peritoneal cells in 96-microwell cell culture at 1.5 × 10⁵ per well and removing nonadherent cells after 90 min, were incubated with IGROV-1 tumor cells for 48 h. Tumor cells were pulsed with [³H]thymidine at 1 μCi/well for 18 h and radioactivity was measured with a Trilux Beta Scintillation Counter (Wallac). Under these conditions, total peritoneal cells and macrophages incorporate negligible amounts of [³H]thymidine, so that the radioactivity measured reflects the proliferation levels of the tumor cells. Control wells were cultured with peritoneal cells or adherent macrophages collected from untreated mice to determine background proliferation. Antiproliferative activity was assessed as 100 - [(cpm from treated × 100) / (cpm from untreated)].

Statistical analysis. Percent survivorship over time was estimated by the Kaplan-Meier product limit method and compared with the log-rank test. Results of in vitro assays were compared using unpaired, two-tailed Student's t test, and differences at P < 0.05 were considered significant. Analyses were done with GraphPad Prism (GraphPad Software).

CpG-ODN activity against i.p. ovarian tumors and influence of administration route. Mice xenografted i.p. with IGROV-1 human ovarian carcinoma cells were injected i.p., i.v., or s.c. with CpG-ODN delivered q7dx4 at a dose of 20 μg/mouse, corresponding to about 1 mg/kg, starting the day after tumor cell inoculation. Control mice were treated i.p. with saline. Mice were evaluated for survivorship and tumor-free rates (Table 1; Fig. 1A). Saline-treated mice developed ascites, with increased abdominal volume and body weight, in a median time of 11 days and died by day 24. Negligible effects on survival were achieved by administering the drug s.c., whereas i.v. delivery of CpG-ODN led to a significantly increased survival time compared with controls (P = 0.0004), with a T/C value of 190%, and to 1 LTS of 6 mice. However, the most dramatic effect on survival was observed after i.p. treatment (P = 0.0004 versus controls and P = 0.0005 versus i.v. treated mice) where most mice were LTS. In two replicate experiments, with the i.p. treatment, totally comparable results were obtained (Table 1). Overall, a total of 12 of 17 mice were LTS at the end of the experiment (120-130 days), and necropsy revealed a few small solid tumors in 4 mice, whereas 8 were tumor-free. Treatment using ODN2243 lacking a CpG motif did not increase mouse survival (T/C value of 112%). Analysis of the effect of i.p. CpG-ODN in mice with advanced IGROV-1 tumor burden, with treatment starting 8 days after cell injection, revealed a significant increase in lifespan (T/C value of 162%; P = 0.0044 versus control mice), but all mice died with ascites by day 60 (Table 1; Fig. 1A).

The antitumor efficacy of CpG-ODN delivered i.p. q7d was also assayed in mice bearing the human ovarian tumor cells SK-OV-3, OVCAR-5, and OvCa432. Treatment was continued for 7 weeks, because such tumors grew slower than IGROV-1 tumor. In mice injected with SK-OV-3 cells, a slow-growing tumor developed with ascites onset in a median time of 40 days. CpG-ODN treatment starting 1 day after tumor cell inoculation delayed the onset of ascites and significantly prolonged survival (P = 0.0001), with 5 LTS of 6 mice at day 185. A strong antitumor effect was also observed when treatment started 8 days after cell injection, with 3 of 5 mice alive at the end of the experiment (day 185). In total, 8 of 11 treated mice were LTS, and at necropsy, 7 were tumor-free (Table 1; Fig. 1B). OVCAR-5-bearing mice developed ascites in a median time of 17 days. CpG-ODN treatment starting the day after cell inoculation led to significantly increased lifespan (T/C value of 306%; P = 0.0001). Four of 7 mice died from tumor, with solid nodules and low or no ascites, whereas the 3 LTS at day 180 were tumor-free at necropsy (Table 1; Fig. 1C). The OvCa432 tumor grew very slowly and only 75% of control mice developed i.p. tumors by 180 days after inoculation; at that time, none of the mice treated with CpG-ODN starting the day after cell injection revealed evidence of tumor (data not shown).

CpG-ODN delivered by all routes was well tolerated, although brief diarrheic events were often observed 2 to 4 h after i.p. or i.v. administration. At necropsy, tumor-free mice showed no sign of peritoneal damage.

Immunologic responses induced by CpG-ODN administered by different routes. To explore the mechanisms underlying the significant differences in antitumor activity observed among the three routes of CpG-ODN administration, several immunologic variables were evaluated at both systemic (blood and spleen) and local (peritoneal washings and peritoneal tissue) levels in mice treated with CpG-ODN by the different routes at the same dose (20 μg/mouse) used for the antitumor efficacy studies.

KC levels. Murine KC is the functional counterpart of human interleukin-8, a chemokine that induces a cascade of events leading to the attraction and accumulation of activated polymorphonuclear neutrophils. In healthy nude mice, ELISA analysis of serum from blood collected 90 min after CpG-ODN treatment by the three different routes revealed higher KC levels in all three groups compared with those in untreated mice, in which KC was barely detectable. The highest KC levels were found in mice treated i.p., in which levels were 2-fold greater than those in i.v. or s.c. treated mice (P = 0.008 versus i.v. treated mice and P = 0.006 versus s.c. treated mice; Fig. 2A). Even in tumor-bearing mice, the i.p. injection of CpG-ODN induced a 2-fold greater KC levels ± SD than those observed in i.v or s.c. treated mice (4,322 ± 962, 1,919 ± 1,337, and 2,041 ± 823 pg/mL, respectively).

Because KC is produced by immune and nonimmune cells in response to TLR9 activation (22), we have also carried out ELISA analysis using peritoneal tissue extracts. Again, the highest level of KC was detected in mice treated by the i.p. route; although the value differed significantly compared with that of control mice (P = 0.0098), it was not significantly higher than those in i.v. or s.c. treated mice (Fig. 2B). The KC detected in peritoneal tissue extracts was not due to blood contamination, because no hemoglobulin was detected in these samples (data not shown).

Spleen cell cytotoxic activity. Spleen cells from mice (4 mice per group) treated with CpG-ODN by the three routes and recovered at 3, 24, and 72 h after treatment were tested for cytotoxic activity in a standard 4-h ⁵¹Cr release assay using IGROV-1 as targets. At all times evaluated, no significant cytolytic activity was observed (<4% at effector-to-target ratio of 50:1; data not shown).

Functional and phenotypic profile of peritoneal cells. Peritoneal cells obtained from mice 72 h after CpG-ODN treatment by the three routes were evaluated for cell number (9 mice per group), surface phenotype (3 mice /group), and cytotoxic (3 mice per group) and antiproliferative activity (3 mice per group). Cell counting in peritoneal washings indicated a significantly higher cell number in i.p. treated mice compared with controls (P = 0.0027), whereas cell numbers in mice treated i.v or s.c. were similar to those in control mice (Fig. 3). Fluorescence-activated cell sorting analysis of the peritoneal cells using direct staining with antibodies PE-F480, PE-NK1.1, or PE-GR-1 to detect markers of macrophages, NK cells, or granulocytes, respectively, revealed a marked increase in the percentage of both NK cells (P = 0.0056 versus control mice) and granulocytes (P = 0.0127 versus control mice) in mice treated i.p. with CpG-ODN, whereas mice treated s.c. or i.v. exhibited only a slight increase in the percentage of cells positive for the granulocyte marker (P = 0.0366 s.c. treated mice and P = 0.0266 i.v. treated mice versus control mice; Fig. 4A-D).

Peritoneal cells showed no significant cytotoxic activity against IGROV-1 target cells in a standard 4-h ⁵¹Cr release assay (<4% at effector-to-target ratio of 50:1; data not shown). On the other hand, by [³H]thymidine incorporation assay, the antiproliferative activity against IGROV-1 cells by total peritoneal cells derived from i.p., i.v., or s.c. treated mice revealed 32%, 24%, and 2% proliferation inhibition; similar analysis using adherence-purified mononuclear macrophages revealed 25%, 16%, and 2% inhibition, respectively.

The present study shows the powerful antitumor effect of CpG-ODN delivered i.p. against human ovarian tumor xenografts growing i.p in nude mice and shows that this effect is highly dependent on the administration route. Indeed, against IGROV-1 tumor CpG-ODN had less effect when administered i.v. and essentially no effect when delivered s.c., in sharp contrast with the significantly increased survival time after i.p. administration, with most mice still alive at the end of experiments and nearly half of them (8 of 17) tumor-free at necropsy. The dramatic effect of CpG-ODN delivered i.p. was observed regardless of the ovarian tumor model used, because in mice implanted with four different tumors most of treated mice resulted “cure” at experiments end (4-6 months after tumor cell implant), thus indicating that the antitumor effect was not cell specific. Such findings, observed when CpG-ODN treatment started 1 day after tumor cell inoculum, raised the possibility that the CpG-ODN induced immunologic effects that impair cancer cell adhesion and migration. However, in the slow-growing SK-OV-3 tumor, a marked increase in mouse survival was observed even when treatment started 8 days later, thus indicating the ability of i.p. CpG-ODN to affect the growth of already established tumors.

We reported previously the antitumor effect of cytotoxic drugs delivered i.p. according to optimal schedules against the i.p. IGROV-1 tumor xenograft (20). In that study, which included drugs of different mechanistic classes and in clinical use, all of the drugs increased mouse survival and some of them, especially topotecan, achieved a certain number of LTS. However, none of the treated mice was tumor-free at necropsy. Our present findings of “cure” in a sizable proportion of ovarian tumor-bearing mice after i.p. CpG-ODN treatment raise the possibility that this immunomodulating reagent might be superior to current chemotherapies for ovarian cancer.

The i.p. administration of antineoplastic agents has been investigated in clinical studies of peritoneal tumors for more than 40 years (23) based on the rationale that this route of delivery increases exposure of cancer cells to the drug while minimizing systemic toxic effects (24). Clinical advantages of i.p. chemotherapy have been reported in ovarian carcinoma patients with minimal residual disease after debulking (25). Such treatment modality has not received a general acceptance because of technical problems, risk for poor (drug-dependent) tolerance, and reduced quality of life during treatment. However, the clinical advantage of i.p. therapy in selected patients supports the interest of this approach as an effective treatment option. In contrast to chemotherapy, i.p. administration of immunostimulating agents, such as IFN-α, showed good tolerability in ovarian cancer patients, although no conclusions about efficacy could be drawn in that study (14). In our present preclinical model, locoregional administration of CpG-ODN was not only well tolerated by the mice but also remarkably efficacious.

In past and ongoing clinical trials, CpG-ODN is administered s.c. because this route has been reported to effectively activate adaptive immunity (1). However, the immune-modulating activities of CpG-ODN are not restricted to the instruction and reprogramming of adaptive immunity but may also involve amplification of the innate effector cell response. Unlike cells of the adaptive immune response, which can reach the antigen wherever they are activated, innate effector cells gather where the antigen induces their activation. The strong antitumor effect observed in our models only after CpG-ODN i.p. treatment in athymic mice points to the importance of amplification of the innate effector cell response at the site of tumor growth. Consistent with this notion, mice treated by the i.p. route showed a strong increase in peritoneal cell number, especially in granulocytes neutrophils and NK cells, and produced the highest levels of serum and peritoneal tissue KC, which plays a pivotal role in the recruitment and activation of neutrophils and monocytes (26–28). Other cytokines, such as IFN-γ and tumor necrosis factor-α, were not differently modulated by CpG-ODN delivered accordingly the three routes.

Whereas in vitro cytotoxicity assays using spleen cells or cells from peritoneal washing as effector cells and tumor cells as targets revealed no significant cytotoxic activity, total peritoneal cells, containing NK cells, neutrophils, and mononuclear macrophages, inhibited the in vitro proliferation of IGROV-1 cells as did adherence-purified macrophages to a lesser extent. Although interspecies NK reactivity is lower than intraspecies (29), altogether these results suggest that NK cells might not play a prominent role in the CpG-induced antitumor activity in ascitic ovarian carcinomas xenograft, whereas multiple and presumably synergistic antineoplastic cell populations are induced by CpG-ODN to mediate its antitumor effect. Indeed, a critical role for phagocytes (macrophages and granulocytes) has been reported in the antitumor effects induced by CpG-ODN without T cells in murine tumors (30). Although the antitumor activity of NK cells and macrophages is well known, insights into the potential role of neutrophils in anticancer therapy have only recently begun to emerge (31–38), because these cells have been long recognized only as key players in innate defenses against microbial infections. Neutrophils express TLR9 mRNA and protein (39–41), but little is known about the effect of TLR9 signaling on neutrophil functions (42).

It is likely that KC is the main cytokine involved in the recruitment and activation of innate immune effector cells at tumor site and that TLR9-expressing nonprofessional immune cells, such epithelial and endothelial cells, play a prominent role in the cytokine production. Accordingly, high levels of serum KC were observed after i.p. administration of CpG-ODN in chimeric C57BL/6 mice reconstituted with TLR9^−/− bone marrow-derived cells and therefore expressing TLR9 exclusively in nonhematopoietic cells.⁴

The powerful antitumor effect of the TLR9 agonist CpG-ODN delivered locoregionally in a panel of ascitic models of human ovarian carcinoma resulted in a tumor-free status for a sizable proportion (about 50%) of mice with a genetically limited adaptive immune response. Such an effect, superior to those obtained by different established cytotoxic drugs, provides evidence of the importance of innate effector cells in the antitumor activity induced by the CpG-ODN immunomodulating agent. Clinical trials of i.p. CpG-ODN treatment in ovarian carcinoma patients with minimal residual disease might now be contemplated.

No potential conflicts of interest were disclosed.