CXCL12/CXCR4 blockade induces multimodal anti-tumor effects that prolong survival in an immunocompetent mouse model of ovarian cancer

Animal models and cell lines

Tumor generation involved a single intraperitoneal injection of the syngeneic cancer cell line, BR5-1(1×10⁷ cells per mouse), into five-week-old FVB/NJ mice(Bar Harbor, ME)(28), and was consistently first evident after three to five weeks via abdominal distension secondary to malignant ascites. Tumor-bearing mice were euthanized at the endpoint when there were signs of distress, including fur ruffling, rapid respiratory rate, hunched posture, reduced activity, and progressive ascites formation. At necropsy, tumor weight, volume of ascites, and peritoneal dissemination were measured and lungs, brain, and liver were examined for metastasis by visual inspection. A second tumor model, B16F10 murine melanoma, was used as described previously re described (see Supplemental Methods)(29). All measurements were performed in a blinded manner, (to control or experimental group), and all animal experiments were executed according to Public Health Service Policy on Humane Care of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Massachusetts General Hospital.

CXCL12 knockdown using RNAi vector

CXCL12 expression knockdown in BR5-1 cells was achieved by transducing the cells with MISSION lentiviral transduction particles (Sigma-Aldrich) through the use of five different shRNA sequences targeting CXCL12 (see Supplemental Methods).

Cell proliferation and CXCL12 quantitation

Cell proliferation was measured by mannual counting and by using the Premix WST-1 Cell Proliferation Assay System(Clontech Laboratories, Mountain View, CA). In certain experiments, 100 ng/ml of exogenous recombinant murine CXCL12(Peprotech, Rocky Hill, NJ) was added to the culture of CXCL12 knocked-down BR5-1 cells(kdBR5-1) in fresh media at 0, 24, 48, and 72 hr. CXCL12 was measured in cancer cell culture supernatants using a CXCL12-ELISA kit(R&D).

AMD3100 treatment in vivo

Administration of AMD3100 into mice was achieved using osmotic minipumps(Alza, Palo Alto, CA) implanted dorsolaterally under the skin. For the murine ovarian cancer model, animals were treated with AMD3100 or vehicle (PBS) using two different schedules; schedule A - treatment began upon the development of ascites either for 3 days (for time-matched study) or until evident signs of distress as described above (for survival study), schedule B - treatment began two weeks after intraperitoneal injection of tumor cells either for 14 days (for time-matched study) or until evident signs of distress as described above (for survival study). For the murine B16F10 melanoma model, AMD3100 treatment was initiated when tumors reached 4 mm in mean diameter either for 4 days (for time-matched study) or until tumors reached a size >12mm in diameter or the mice showed signs of distress (for survival study). The details of the treatment are described in the Supplemental Methods. All monitoring of animal survival and tumor progression was performed in a manner that was blinded to the observer.

Quantitation of tumor vessel density by immunofluorescence

Mice were euthanized at day 3 after tumor development, at which point the sizes of the tumors analyzed were comparable in the treatment group vs. controls. Tumor vessel density was determined and scored in a blinded manner using fluorescein-labeled lectin (Vector lab, Burlington, MA) and image analysis(Carl Zeiss LSM5 Pascal, Carl Zeiss North America, MO Zeiss) and Image J 1.43 freeware(NIH, MD) as previously described(30).

Quantitation of tumor infiltrating lymphocytes (TILs) by immunofluorescence

Immunofluorescence analysis of tumor infiltrating CD8(CTLs) and FoxP3(T-regs) positive cells was performed. T-cells were visualized using confocal microscopy and scored in a blinded manner using Image J 1.43 freeware (see Supplemental Methods).

Quantitation of apoptosis using TUNEL staining

TUNEL taining was performed on 10μm-thick tumor sections prepared as described in the previous section using the In situ Cell Death Detection Kit TMR Red(Roche). Slides were imaged and analyzed using Image J and were blinded to the analyzer.

Flow cytometry

For TIL subpopulation analysis, mononuclear cells including tumor cells, stromal cells and immune cells derived from disaggregated tumor tissue were separated by centrifugation on Ficoll Hypaque, and directly stained with fluorophore-conjugated anti-CD3, anti-CD8, anti-CD4, anti-CD25, anti-FoxP3, and anti-CXCR4 antibodies(BD Bioscience). Cell subpopulations, including CD3+CD4+ T helper cells, CD3+CD8+ cytotoxic T-cells(CTL) and CD4+CD25+FoxP3+(T-regs) were quantified as a percentage of cell per gram of tumor tissue through examination of both positive control and fluorescence-minus-one gating strategy (FMO)(31). For in vitro assessment of tumor-specific T-cell function, splenocytes and TILs isolated and prepared as single-cell suspensions were plated and pulsed with tumor lysates, Her2/neu peptide(0.1 μg/ml, PDSLRDLSVF, EZBiolab, Westfield, IN)(32), or medium alone for 18 hr when Golgi Plug(BD Bioscience) was added (0.75 μl/well) for 6 hr, stained with anti-Granzyme B(R&D Systems) , and analyzed on a 3 laser LSRII BD Biosciences)(see Supplemental Methods).

T regulatory cell depletion and CD8+ cell isolation

Magnetic beads were used to separate T-regs and CD8+ cells (MACS, Milteny Biotech, Auburn, CA) yielding a population with >95% purity. T-reg depleted cells(CD4+CD25-) were plated, stimulated and stained as described above.

Transmigration assay

T-regs and CD8+ cell migration was measured using Transwells (96-well format, 3-micron pore size; ChemoTx System, Neuro Probe Inch, Gaithersburg, MA) as previously described(33) in CXCL12 concentrations(5ng/ml, 50ng/ml, 500ng/ml, and 5μg/ml), and in the presence or absence of AMD3100(1μg/ml) or pre-incubation with 100 ng/ml pertussis toxin(PTX; Sigma) for 1hr at 37°C.

Statistical analysis

All of the results are expressed as mean ± SEM. Differences between treatment groups were determined by the Mann Whitney test and in other numerical variables by unpaired t test with a P value of <0.05 considered significant. All statistical analyses were independently performed on datasets by Dr. Gebremichael at the Harvard School of Public Health.

CXCL12 knockdown impacts BR5-1 proliferation in vitro and tumor growth in vivo

CXCL12 has been shown to stimulate ovarian cancer cell proliferation in vitro(16). We used the murine ovarian cancer cell line BR5-1, which constitutively expresses CXCL12 and CXCR4, to observe the effect of RNAi-mediated CXCL12 silencing on tumor growth. CXCL12 levels from wild-type BR5-1 cells(wtBR5-1), RNAi-transfected, CXCL12 knocked-down BR5-1 cells(kdBR5-1), and scrambled-RNAi transfected BR5-1 cells(scrBR5-1) were quantitated using an ELISA immunoassay(Figure 1A). Similar levels of CXCL12 were released by wtBR5-1 and scrBR5-1 (1.05 ± 0.19 ng/ml and 0.85 ± 0.09 ng/ml) while CXCL12 from kdBR5-1 was negligible (<0.02 ng/ml, P < 0.0001). CXCR4 expression was not significantly different among wtBR5-1, kdBR5-1, and scrBR5-1 (Figure 1B). KdBR5-1 proliferated at a significantly slower rate compared to wtBR5-1 and scrBR5-1 (P < 0.0001, Figure 1C). The results were confirmed using a WST-1 proliferation assay, showing a significantly higher absorbance in wtBR5-1 compared to kdBR5-1 at 48 hr (absorbance ratio 12.5 ± 0.23 vs. 3.37 ± 0.30 respectively, P <0.001, not shown). Addition of exogenous CXCL12 significantly increased kdBR5-1 proliferation to a level comparable to wild type (Figure 1D). Apoptosis was excluded as a mechanism responsible for the reduced kdBR5-1 cell proliferation (Figure 1E), thus confirming direct effect of CXCL12 on ovarian cancer cell proliferation(16).

To test if defective CXCL12 production was maintained in vivo, cancer cells derived from wtBR5-1(WT) and kdBR5-1 tumors(RNAi) were grown for 48 hr in media. Quantitation of CXCL12 in supernatants showed results comparable to in vitro data, with CXCL12 < 0.02 ng/ml in tumor cells extracted from kdBR5-1 injected mice and CXCL12 levels of 1.2 ± 0.10 ng/ml in tumor cells extracted from wtBR5-1 tumors. Furthermore, RNAi tumors displayed a marked reduction of CXCL12 (stained red in Figure 1F) compared to WT tumors via immunofluorescence. Small numbers of cells expressing CXCL12 were detected and thought to be due to the infiltration of CXCL12-expressing stromal and endothelial cells (white arrows, Figure 1F) in the tumor microenvironment.

FVB/NJ mice injected with wtBR5-1 normally generate intraperitoneal tumors in three to five weeks. The time interval between tumor implantation and appearance of ascites was found to be significantly higher in kdBR5-1 injected mice compared to wtBR5-1 and scr-BR5-1 injected mice (53 ± 5.5 days vs. 33 ± 2.1 and 32 ± 3.3 days respectively, P < 0.05, Supplementary Figure 1A). A reproducible ‘survival window’ was defined between the first clinical sign of disease (abdominal swelling secondary to ascites formation) and euthanasia required by advanced clinical condition. The overall survival was significantly longer in kdBR5-1-injected mice compared to wtBR5-1-injected mice (8.7 ± 1.8 days vs. 3.4 ± 0.4 days respectively, P < 0.005, Figure 1G). Mice injected with scrBR5-1 had similar survival windows (3.3 ± 0.5 days) compared to wtBR5-1-injected mice.

AMD3100 treatment prolongs mouse survival and reduces dissemination of tumors

Mice were treated with AMD3100 either at the onset of ascites until the end-stage time point (schedule A) in order to simulate the treatment in human disease(34), or from day 15 after the implantation of tumor cells (schedule B). AMD3100 treatment schedules were based on previous pharmacokinetic evaluations to maintain steady serum levels (35). AMD3100 treatment on schedule A resulted in a two-fold increase in survival compared to PBS-treated mice (6 days vs. 3 days, P <0.05) (Figure 2A), and higher overall survivals were reported in the AMD3100 treated mice vs. controls, (38 days ± 1.1. vs. 32 days ± 1.5, P < 0.05) (Figure 2B). In a separate experiment, AMD3100-treated and PBS-treated mice were euthanized at day three from the onset of clinical signs to compare total tumor mass, volume of ascites, and tumor dissemination (Figure 2 C-E). Mean tumor weight and amount of ascites were lower in AMD-treated compared to PBS–treated mice (1.3 ± 0.23 g and 4.4 ± 0.4 ml vs. 0.8 ± 0.17 g and 3.6 ± 0.4 ml, respectively) (Figure 2 D- E), although these differences were not significant.

BR5-1 injection robustly results in multiple tumor nodules growing on organ surfaces without significant invasion. Sites of tumor spread include the mesothelial lining of the peritoneum, intestines, spleen, and diaphragm, but not the lungs(28). Tumor dissemination was compared between AMD3100-treated vs. PBS-treated mice (Figure 2 F-G) and classified as localized (localizations ≤ 3), limited (6 < localizations > 3, with no visible masses on mesentery and diaphragm), or disseminated (localizations ≥ 6, including mesentery and diaphragm). The percentage of mice with tumor deposits on organ surfaces at necropsy was lower in AMD3100-treated compared to PBS-treated mice. While all AMD3100-treated mice had localized or limited tumors, 70% of the tumors in PBS-treated mice fell under the disseminated category.

In schedule B, when the treatment was started at day 15 after tumor implantation, a significant delay in the onset of ascites was recorded in the AMD3100 treated mice vs. controls (21 ± 0.5 days vs. 24 ± 1.3 days, P < 0.05). An increase in survival in AMD3100 treated versus control treated animals was observed (3.6 days vs. 7 days, Figure 2H) as well as higher overall survival times (41 ± 2.1 days vs. 36 ± 0.5 days, P < 0.05) (Supplemental Figure 1B). Tumor dissemination in schedule B examined five weeks after the tumor injection was comparable to those on schecdule A (Figure 2I).

CXCL12/CXCR4 manipulation markedly reduces tumor angiogenesis

CXCL12 is a pro-angiogenic factor in ovarian cancer via upregulation of VEGF, which in turn increases CXCR4 expression by endothelial cells (19, 36). Vessel density was significantly lower in AMD3100 tumors and RNAi tumors compared to PBS-tumors (Figure 3A and B: 91.23 ± 17.57 vessels/mm² and 71.11 ± 28.73 vessels/mm² vs. 167.84 ± 27.92 vessels/mm² respectively, AMD3100 vs. PBS - P < 0.05; RNAi vs. PBS - P < 0.05). SCR tumors displayed similar vessel density (111.11 ± 11.07 vessels/mm²) to PBS-tumors.

AMD3100 treatment increases tumor apoptosis and necrosis

It has been previously reported that CXCL12-mediated inhibition of the NF-kB/TNFα apoptotic pathway (17, 18) affords in vivo protection from tumor apoptosis and that this can be impaired via CXCR4 inhibition (37, 38). Here, significantly higher numbers of apoptotic cells were found in AMD3100 tumors compared to PBS-tumors (90.86 ± 17.99 cells/mm² vs. 37.18 ± 11.60 cells/mm², P < 0.05) (Figure 4 A-B). RNAi tumors showed a non significant increase in apoptotic cells compared to controls (97.35 ± 46.5 cells/mm²), suggesting a compensatory role for stromal-derived CXCL12 in the these tumors. H&E stained selected tumor tissues from AMD3100-treated and PBS-treated mice showed areas of necrosis in 50.0% of the AMD3100 tumor sections, while PBS tumor sections appeared microscopically negative for necrotic tissue (Figure 4C).

CXCR4 antagonism selectively reduces intratumoral FoxP3+ T-cell infiltration in comparison to CD8+ T-cells

CXCL12 can act as both a chemoattractant and a chemorepellent for T-cell subpopulations in a concentration and CXCR4 dependent manner in tumor models(25). Therefore, we examined the effect of the manipulation of the CXC12/CXCR4 axis on intratumoral T-cell trafficking in this model of ovarian cancer. CD8+ T-lymphocyte infiltration was not significantly different in PBS-treated, kdBR5-1-injected, and AMD3100-treated mice via immunofluorescence (Figure 5A, upper panels and 5B) and flow cytometry (CD3+CD8+: 11.09 ± 0.20%, 10.44 ± 1.18 %, 9.54 ± 0.16 % and CD3+CD4+: 42.1 ± 2.39 %, 40.53 ± 0.15 %, 40.33 ± 2.30 %, respectively). In contrast, RNAi tumors and AMD3100 treatment resulted in a significant 4 to 5-fold reduction in FoxP3+ T-cells compared to PBS-tumors as determined by immunohistochemistry (18.77 ± 7.29 cells/mm² and 14.46 ± 4.48 cells/mm² vs. 76.0 ± 25.02 cells/mm², P < 0.05, Figure 5A, lower panels and 5C) and flow-cytometry (Figure 5D). No differences were detected in the spleen for CD3+CD8+, CD3+CD4+ and CD4+CD25+FoxP3+ cells among PBS-treated, kdBR5-1-injected, and AMD3100-treated mice (data not shown). In order to determine whether this was the result of selective FoxP3+ T-cell redistribution(39) or an alteration in TIL trafficking(25), thymus and spleens were obtained from tumor-bearing mice and stained for FoxP3 expression. No significant differences were detected in PBS-treated (1008.23 ± 196.15 cells/mm² and 69.54 ± 21.77 cells/mm²) vs. AMD3100-treated mice with regards to the number of FoxP3+ T-cells in these tissues (948.15 ± 156.03 cells/mm² and 153.40 ± 86.65 cells/mm², Figure 5E). AMD3100 treatment resulted in a 6-fold increase in the CD8+/FoxP3+ ratio, a parameter correlated with increased survival in ovarian cancer(40), compared to control PBS treated animals (3.12 and 3.50 vs. 0.58 respectively, P < 0.01, Figure 5F).

AMD3100 treatment selectively increases tumor-specific T-cell responses and impairs T-reg migration in vitro

To test if the differential distribution of T-regs was associated with impaired tumor-specific T-cell function, we analyzed the in vitro production of Granzyme B in T-lymphocytes extracted from spleen and tumors in AMD3100-treated and PBS-treated mice as described above. Splenocytes and TILs were stimulated with tumor lysates and Her2/neu specific peptide for 24 hr and then analyzed for the intracellular expression of Granzyme B using flow-cytometry. Her2/neu was shown to be expressed by BR5-1 using flow-cytometry (data not shown). Granzyme B production was significantly higher in TILs from AMD3100-treated compared to PBS-treated mice for both tumor lysates and Her2/neu peptide stimulation for CD4+ and CD8+ T-cells (Figure 6 A-B). To assess the impact of T-regs on T-lymphocyte function, splenocytes and TILs were depleted of T-regs using magnetic beads. Depleted CD4+CD25- and non-depleted CD4+CD25+ cells were stimulated as before, and Granzyme B expression was determined using flow cytometry. Non-depleted CD4+ TILs from AMD3100 tumors showed higher Granzyme B production compared to PBS-tumors, whereas T-reg depleted TILs from PBS-tumors demonstrated increased Granzyme B production compared to non-depleted TILs from PBS treated tumors (Figure 6 C-D). No differences in Granzyme B production were observed in the spleen derived T-cell populations for either AMD3100 or PBS-treated tumors (Figure 6D and data not shown).

We observed that intratumoral T-regs consistently express higher CXCR4 levels compared to CD4+CD25− and CD8+ cells (Figure 6E-F). We then examined the migratory responses of T-regs and CD8+ cells to gradients of CXCL12 in vitro. We showed that T-regs from both tumors and spleen underwent chemotaxis in response to chemotactic gradients of the CXCL12 with a classical bell shaped dose-response curve. Addition of AMD3100 or pre-treatment of cells with Pertussis toxin (PTX) significantly impaired G-coupled, CXCR4-mediated T-reg cell chemotaxis (Figure 6G-H), while AMD3100 did not prevent CD8+ migration towards CXCL12 (Figure 6I). These data support the view that differential expression of CXCR4 may selectively affect the intratumoral trafficking of regulatory versus effector cells towards intratumoral CXCL12 gradients.

CXCL12/CXCR4 antagonism using AMD3100 selectively impacts T-regs infiltration in a second murine tumor model

To test the impact of CXCL12/CXCR4 axis blockade on T-regs infiltration in a second solid tumor, we employed the B16F10 murine melanoma model. Previous reports reveal that B16 melanoma cells express both CXCR4 and CXCL12 (41) and have assessed the effect of CXCR4 antagonists on tumor spread (10). To test the effect of CXCL12/CXCR4 manipulation on T-cell infiltration into melanomas, osmotic pumps delivering AMD3100 were implanted as described in the Methods section. We observed a significant decrease of CD4+CD25+ and CD4+CD25+FoxP3+ T-cells in the AMD3100 treated tumors compared to controls (1277.28 ± 36.73 T-cells/g and 202.56 ± 29.14 T-cells/g vs. 3494.27 ± 831.46 T-cells/g and 573.55 ± 111.40 T-cells/g P < 0.05, Figure 7A), while the CD8+ TIL count was not significantly different between the experimental group and controls (1569.28 ± 599.99 T-cells/g vs. 3337.51 ± 687.12 T-cells/g, P = 0.89, Figure 7A). Intratumoral T-regs were shown to express higher levels of CXCR4 compared to CD8+ cells (Figure 7B). These data are consistent with those generated in the ovarian cancer model and supports the view that the antagonism of the CXCR4/CXCL12 axis results in selective T-reg depletion compared to CD8+ T-cells potentially as a result of higher levels of CXCR4 expression on the regulatory T-cell subpopulation. Tumor growth delay at the levels of 4 and 8 times the initial volume was 2.9 ± 0.6 days and 4.7 ± 1.5 days in control tumors (n=11), and 2.7 ± 0.7 days and 4.3 ± 1.5 days in AMD treated tumors (n=12), respectively. There was no significant difference in the rate of tumor growth or in overall survival between the groups, which is consistent with reports that the CXCL12/CXCR4 axis does not impact B16F10 melanoma cell proliferation(42) and that depletion of T-regs alone is not sufficient to promote tumor eradication in the B16 melanoma model.