Anti-tumor innate immunity activated by intermittent metronomic cyclophosphamide treatment of 9L brain tumor xenografts is preserved by anti-angiogenic drugs that spare VEGF receptor 2

MDSCs are increased in tumor-bearing mice, and in cancer patients, and have been implicated in promoting tumor growth and suppressing anti-tumor immunity [35]. Given the ability of MDSCs to suppress NK cell activity [22], which contributes functionally to metronomic CPA-induced tumor regression [10], we investigated whether MDSCs are also recruited into CPA-treated tumors, where they could counter the innate immune response to metronomic chemotherapy. FACS analysis of MDSCs was performed on single-cell suspensions prepared from untreated and metronomic CPA-treated spleens, bone marrow and 9L tumor xenografts grown in scid mice. CD11b⁺ was used as a marker of bone marrow-derived cells, including monocytes, macrophages, dendritic cells and NK cells, while CD11b⁺Gr1⁺ co-positive cells marked MDSC populations [36]. The presence of 9L tumors had no effect on the distribution of either single-positive CD11b⁺ cells or double-positive CD11b⁺Gr1⁺ cells in either spleen or bone marrow (Figure 1, left vs. middle column). Single-positive CD11b⁺(Gr1⁻) cells were increased significantly – by ~2-fold in spleen and bone marrow and by ~8-fold in tumor after 4 cycles of CPA treatment (day 24) (Figure 1, middle vs. right column, top left quadrant). A time-dependent increase in CD11b⁺ tumor-infiltrating cells was seen from 2 to 4 CPA cycles (Additional file 1). Metronomic CPA significantly decreased CD11b⁺Gr1⁺ MDSC populations in treated bone marrow (2-fold decrease) and in treated spleens (4.7-fold decrease), with no significant increase in the treated tumors (Figure 1, middle vs. right column: top right quadrant). Thus, metronomic CPA suppresses CD11b⁺Gr1⁺ MDSC populations in spleen and bone marrow without significantly increasing the intratumoral MDSC population.

Metronomic CPA treatment on an intermittent, 6-day repeating schedule regressed large, established 9L gliosarcoma xenografts in scid mice after 3–4 cycles of CPA administration (Figure 2A), in agreement with earlier findings [37]. Combination of metronomic CPA with the VEGFR2-specific monoclonal antibody DC101 (22.5 mg/kg) resulted in tumor stasis but little or no tumor regression over the 39-day observation period (Figure 2A). A very similar tumor growth static response was seen previously when metronomic CPA was combined with the VEGF receptor-selective inhibitor axitinib [38]. DC101 was a highly effective anti-angiogenic agent, as shown by the large decrease in CD31 immunostained blood vessels in the CPA and DC101 co-treated tumors (Figure 2B), but caused only a modest tumor growth delay, consistent with the relative insensitivity of 9L tumors to angiogenesis inhibition [38] (also see Figure 3A, below). DC101 significantly inhibited the CPA-stimulated tumor recruitment of macrophages (CD68 marker), dendritic cells (CD74 marker), and NK cells (NKp46 marker) and their cytotoxic effectors, perforin, granzymes, and lysozymes (Figure 2C; Additional file 2). These findings were confirmed by immunohistochemical staining for macrophages, NK cells, and the NK cytotoxic effector perforin 1 (Additional file 3). Metronomic CPA-induced expression of CXCL14, an NK cell chemoattractant, was not significantly affected by DC101 (Figure 2C). In a separate experiment where the DC101 dose was increased to 28.6 mg/kg, the inhibition of immune cell recruitment was even more complete but was accompanied by host toxicity in the CPA combination group (i.e., internal bleeding and death in 2 of 8 mice by treatment day 24; data not shown). Given the high specificity of DC101 for VEGFR2 [29], these studies demonstrate that VEGFR2 signaling contributes to metronomic CPA-induced anti-tumor innate immunity, and is likely the target in the previously observed inhibition of immune recruitment and tumor regression by three VEGF receptor-selective RTKIs [10].

Sorafenib is a multi-RTKI with an IC₅₀ for VEGFR2 > 100-fold higher than VEGF receptor-selective RTKIs (Table 1). When given at a dose of 25 mg/kg/day, sorafenib alone exhibited little or no activity against 9L xenografts when compared to vehicle-treated controls (Figure 3A), similar to DC101 (Figure 2A) and other anti-angiogenic drugs [38]. Importantly, when combined with metronomic CPA, sorafenib slightly delayed but did not block tumor regression (Figure 3A). Tumor regression was slightly less complete in the sorafenib + CPA combination group, but the overall anti-tumor response was not significantly different from that of metronomic CPA alone (p > 0.05, one-way ANOVA). This contrasts with the significant inhibitory effects of DC101 (Figure 2A) and the VEGFR-selective RTKIs axitinib, cediranib, and AG-028262 on metronomic CPA-induced tumor regression in the same tumor model [10, 38]. The time of onset of tumor regression (~4-6 days after the third CPA treatment on day 12) was only slightly delayed (~3-6 day delay; range for individual tumors) by sorafenib co-treatment. The absence of a synergistic anti-tumor response to the sorafenib + metronomic CPA combination may reflect the low intrinsic sensitivity of 9L tumor growth to anti-angiogenesis (Figures 2A, 3A, and [38]) together with the strong tumor cytotoxic activity of CPA in conjunction with the potent anti-tumor immune response that it activates, which may already induce a maximal regression response. Both CPA alone and the CPA + sorafenib combination were well tolerated, with no significant toxicity apparent (data not shown).

Sorafenib exhibited strong anti-angiogenic activity under these treatment conditions, as indicated by significant decreases in tumor microvessel density (CD31 immunostaining; Figure 3B). Tumor vascularity was also significantly decreased by sorafenib combined with metronomic CPA, albeit less completely than following sorafenib treatment alone (Figure 3B). This may reflect concomitant chemotherapy-induced circulating endothelial progenitor mobilization [41]. Metronomic CPA treatment alone did not decrease 9L tumor microvessel density (Figure 3B), as also seen earlier [9]. Immunostaining of 9L tumor samples with antibody to phospho-VEGFR2 (Tyr1214), a major site of VEGF-induced auto-phosphorylation, verified that sorafenib did not inhibit tumor blood vessel VEGFR2 phosphorylation (Figure 3C; Additional file 4). In contrast, DC101 and axitinib both inhibited VEGR2 phosphorylation when compared to the drug-free tumor controls (Figure 3C; Additional file 4). Thus, the anti-angiogenic activity of sorafenib under the in vivo treatment conditions used here likely involves one or more of its non-VEGFR targets.

Next, we examined the effects of sorafenib on metronomic CPA induction of several factors associated with the innate immune response linked to tumor regression. Figure 4A shows that sorafenib does not block the strong increase in host (mouse) expression of thrombospondin-1 (TSP1), which occurred as early as 6 days after the second metronomic CPA treatment cycle. Sorafenib also did not interfere with NK cell recruitment, as determined by NK1.1 expression (Figure 4B) and NK1.1 immunostaining (Figure 4D). A similar time course of induction and lack of inhibition by sorafenib characterized perforin 1 (Figure 4C), an essential NK cell cytotoxic effector [42, 43]. These findings contrast with the strong inhibition of metronomic CPA-induced NK cell recruitment by VEGF receptor-selective inhibitors, including VEGFR2-targeted DC101 (Figure 2C; Additional file 2, Additional file 3) and the pan-VEGFR inhibitory small molecules axitinib, cediranib and AG-028262 [10]. Furthermore, sorafenib did not inhibit tumor recruitment of macrophages (CD68 and F4/80 markers; Figure 5A), Fas, and lysozymes 1 and 2 (Figure 5B, 5C), which are important for macrophage anti-tumor activity [44, 45]. Sorafenib also did not block the metronomic CPA-stimulated increases in dendritic cells (CD74), NK and dendritic cell hybrid interferon-producing killer dendritic cell factor B220 [46], platelet-associated platelet factor 4 (PF4), stromal-derived factor 1α (SDF1α), and CD11b, a general marker for bone marrow-derived cells (Additional file 5).

The rat 9L gliosarcoma cell line, authenticated by and obtained from the UCSF Neurosurgery Tissue Bank (San Francisco, CA), was grown at 37°C in a humidified, 5% CO₂ atmosphere in 10% FBS, 100 units/ml penicillin, and 100 μg/ml streptomycin containing DMEM culture medium. CPA was purchased from Sigma Chemical Co. (St. Louis, MO), sorafenib was purchased from LC Labs (Woburn, MA), axitinib was a gift from Pfizer (New York, NY), and DC101 was a gift from ImClone Systems (New York, NY). Fetal bovine serum (FBS) and DMEM were purchased from Invitrogen (Frederick, MD).

Isolation of total RNA from frozen tumor tissue, reverse transcription, and qPCR analysis were carried out using primer sets described previously [10] or shown in Additional file 6. Primers were designed using Primer Express (Applied Biosystems, Carlsbad, CA) and evaluated using LaserGene software (DNAStar, Madison, WI) to ensure mouse-specificity. The absence of cross-species amplification was verified by testing each primer set using a panel of RNA samples isolated from rat and mouse liver, human HUVEC cells, rat 9L cells and human U251 cells. Results were analyzed using the comparative C_T (ΔΔC_T) method and are presented as relative RNA level compared to untreated tumors after normalization to the 18S RNA content of each sample.

ICR/Fox Chase immune deficient male scid mice were purchased from Taconic Farms (Germantown, NY) at the age of 5 to 6 weeks (24 to 26 g) and were housed in the Boston University Laboratory of Animal Care Facility and treated in accordance with approved protocols and federal guidelines. 9L cells (4 × 10⁶ cells) were injected s.c. on each posterior flank in 0.2 ml serum-free DMEM using a 0.5-inch 29-gauge needle and a 0.3 ml insulin syringe. Tumors volumes and body weights were measured at least twice/wk [10] and treatment groups were normalized (each tumor volume set to 100%) once average volumes reached 500 mm³. Mice were treated with CPA on a 6-day repeating metronomic schedule at 140 or 160 mg CPA/kg body weight (BW)/injection (equal to doses of ~150 and 170 mg/kg, respectively, based on CPA-monohydrate [9]), which induce indistinguishable immune responses and anti-tumor activity in this model [9]. Sorafenib and axitinib were administered daily at 25 mg/kg BW/day i.p. for up to 36 days. The VEGFR2-specific monoclonal antibody DC101 was administered i.p. every 3 days at 22.5 mg/kg BW, or as specified, once the tumors reached an average volume of ~360 mm³. On co-treatment days, DC101, or sorafenib, was administered 4 hr after CPA to minimize the potential for drug-drug interactions. Day 0 marks the first day of drug treatment. 9L tumors are well tolerated when grown s.c. in this animal model and show little or no toxicity to the mouse host, even at much larger tumor sizes [37].

Tumors were collected on day 0 (first day of drug treatment) and 6 days after the second and fourth cycles of CPA treatment, i.e., days 12 and 24, or as indicated. Tumors were excised and portions were frozen in liquid nitrogen (for RNA) or in 2-methylbutane (for tissue sectioning and immunohistochemistry). Cryosections (5–10 μm) were prepared and fixed in acetone (for NK1.1 and Prf1 immunostaining) or 1% paraformaldehyde (for CD31 staining). Slides were stained using goat anti-mouse NK1.1 (CD161) antibody (clone M-15, cat. #sc-70150, Santa Cruz Biotech, Santa Cruz, CA) at a dilution of 1:25, or rat anti-mouse CD31 (Cat. #557355, BD Biosciences Pharmingen, San Jose, CA) at a dilution of 1:1000, or rat anti-mouse Perforin 1 (CB5.4) antibody (cat. #sc-58643, Santa Cruz Biotech) at a dilution of 1:100. An avidin-biotin blocking kit (cat. #SP-2001, Vector Labs, Burlingame, CA) was used for all incubations. Incubation with biotinylated rabbit anti-goat (cat. #BA-5000; 1:250 dilution) (for NK1.1) or rabbit anti-rat (cat. #BA-4000; 1:200 dilution) (for CD31 and Perforin 1) secondary antibody was followed by ABC signal amplification with Vectastain ABC peroxidase reagent (cat. #PK-4000 or cat. #PK-6100), and then color development with VIP (cat. #SK-4600) or DAB (cat. #SK-4100) substrate (all from Vector Labs). For paraffin-embedded samples (DC101 study), CD31 antibody (Dianova, Hamburg; cat. #DIA 310) was used at a dilution of 1:40 along with ABC Elite (Vector, cat. #PK-6100) for enhancement and ImmPACT VIP for detection (Vector, cat. #SK-4605). The washed slides were briefly countered stained with Hematoxylin Solution, Harris Modified (Sigma, cat. #HHS128). Stained tumor section images were quantified using NIH ImageJ, typically based on ≥10-20 images per group [10]. Data shown are treatment group mean values ± S.E.

Mice bearing 9L tumors ~500 mm³ in size were treated with a single i.p. injection of DC101 (22.5 mg/kg BW), sorafenib (25 mg/kg BW), or axitinib (25 mg/kg BW). Tumors were collected after 2 hr (sorafenib, axitinib), based on the reported half-lives of ~ 2 hr for those drugs [31, 61], or after 4 hr (DC101), to give sufficient time for DC101 absorption and distribution [62]. The effects of these drug treatments on the activity state of functional tumor blood vessel VEGFR was assessed by immunohistochemical staining for VEGFR phosphorylation. VEGFR phosphorylation of tumor blood vessels could not be reliably assayed after multiple days of anti-angiogenic drug treatment owing to the extensive loss of tumor blood vessels (e.g., Figure 2B and Figure 3B). 9L tumor cryosections were fixed with 4% paraformaldehyde for 15 min, permeabilized with 1% Triton X-100 in 1% sodium citrate for 5 min at 4°C, and then blocked with fresh 3% H₂O₂ for 5 min at room temperature, followed by 5% normal horse serum for 20 min at room temperature. Slides were incubated with anti-phospho-VEGFR2 antibody (rabbit anti-mouse phospho-Fik-1, Tyr1214, 1:25 dilution; sc-101820, Santa Cruz Biotechnology) overnight at 4°C, followed by biotinylated horse anti-rabbit secondary antibody (BA-1100, Vector Labs; 1:250 dilution) for 60 min at room temperature, then ABC Vectastain Elite for 30 min at room temperature, DAB (SK-4100, Vector labs) for 10 min for detection, followed by hematoxylin counterstaining.

Single-cell suspensions of freshly excised spleens and tumors were prepared in a GentleMACS Dissociator (Miltenyi Biotec, Auburn, CA) using the manufacturer’s protocols for mouse spleens and implanted tumors. Bone marrow was isolated from excised mouse femurs and tibiae, which were cut open at the ends with surgical razor blades and washed out with phenol red-free alpha-MEM (Cat. #41061, Gibco, Grand Island, NY). Tumor single-cell suspensions (5 ml) were diluted to 15 ml with PEB dissociation buffer (1X PBS, pH 7.2, 0.5% BSA, and 2 mM EDTA). Diluted tumor, bone marrow, and spleen suspensions were passed through 70 μm filters (Cat. #22363548, Fisher Scientific, Pittsburgh, PA). All three tissue-derived, single-cell populations were subjected to red blood cell lysis with 5 ml of 1X RBC lysis buffer (Cat. #00-4333, eBioscience, San Diego, CA) for 5 min at room temperature. Reactions were terminated by addition of 20 ml of sterile 1X PBS. Cells remaining were centrifuged at 300-400 g at 4°C and resuspended in ~50 μl of eBioscience Staining Buffer for antibody incubation. Spleen samples were incubated with 0.5-1 μg of anti-CD16/CD32 antibody (Cat. #14-0161-82, eBioscience) for 5–10 min on ice to neutralize non-specific Fc antibody interactions. All samples were incubated in the dark for 25 min at 4°C with fluorescently tagged monoclonal antibody specific for the cell markers Ly-6G (Gr-1) (1 μl (0.5 μg) per sample; Ly-6G-PE, Clone RB6-8C5, Cat. #11-5931, eBioscience) and CD11b (1 μl (0.2 μg) per sample; CD11b-PE, Clone M1/70, Cat. #12-0112, eBioscience). Background samples were stained with FITC- and PE-labeled Rat IgG (1 μl per sample, Cat. #553929 and Cat. #12-4321, eBioscience). Samples were washed, filtered, resuspended and analyzed [10].

Data were analyzed for statistical significance by one-way ANOVA with Bonferroni multiple comparison correction (for comparisons involving 3 or more groups) or by t-test, two-tailed, unpaired comparison, as implemented in GraphPad Prism 4.1.