Introduction

The goal of tumor immunotherapy is the development of modalities given alone or in combination that result in the generation of an effective systemic tumor-specific immune response by the host that can overcome immune escape mechanisms. The identification of immunogenic targets unique to or overexpressed by tumor cells is critical to achieving this goal. The HER2/neu oncogene encodes human epidermal growth factor receptor 2 (HER2/neu), a member of the epidermal growth factor receptor family of transmembrane tyrosine kinase receptors, which participates in processes including physiology, proliferation and differentiation of various human tissues.1, 2 Overexpression of HER2/neu is found in ~20% of invasive breast cancers, and is associated with a more invasive phenotype and a poorer prognosis.3 Development of an active immune response using a vaccine targeting HER2/neu represents an attractive immunotherapeutic strategy for overcoming immune escape mechanisms induced by the tumor microenvironment.

Myeloid-derived suppressor cells (MDSCs), a population of immature myeloid cells that are increased systemically and in the tumor microenvironment of both murine cancer models and human malignancies, are prominent contributors to tumor immune escape.4, 5 This heterogeneous population is characterized phenotypically in mice by the cell surface antigens CD11b and Gr-1.5 Gr-1 encompasses two subtypes, Ly-6C and Ly-6G, which have been used to further differentiate MDSCs into CD11b+Ly-6Chigh Ly-6G− monocytic (mMDSC) and CD11b+Ly-6ClowLy-6G+ granulocytic (gMDSC) subpopulations, respectively.6, 7 Consistent with their heterogeneous phenotype, MDSCs suppress the antitumor immune response through multiple mechanisms.8 MDSCs interfere with lymphocyte proliferation via deprivation of essential amino acids, such as arginine and cysteine.7, 9, 10 They also mediate oxidative stress via production of reactive oxygen species and peroxynitrate. This leads to nitration of tyrosine in CD8 and the T-cell receptor, ultimately changing the rigidity of the T-cell receptor.11 Furthermore, MDSCs support induction of other immune inhibitory populations such as regulatory T cells (Tregs) through the production of transforming growth factor-β and interleukin-10.12, 13, 14, 15 Given these immune-suppressive effects, therapies that can overcome systemic anergy induced by MDSCs have generated great interest.

Studies from our group were the first to develop and test recombinant vaccinia vectors encoding the immune-enhancing granulocyte-macrophage colony-stimulating factor (GM-CSF) for the localized treatment of solid tumors. In preclinical studies, we demonstrated that vaccinia and vaccinia recombinants were effective in infecting/transfecting tumors and, importantly, that despite the immunogenicity of the vaccinia vector, high levels of transfection could be obtained following repeated injections of tumor in mice16 and subsequently in patients with recurrent superficial melanoma.17 We developed and took clinical VVGMCSF into phase I trials in melanoma.18 Subsequent to our studies, this recombinant (JX-594) was shown to have antitumor activity in preclinical models and clinical trials in a number of diseases.19, 20

In the present study using orthotopic growth of an aggressive HER2/neu-expressing murine tumor characterized by high levels of CD11b+Gr-1+ MDSCs in the tumor microenvironment and systemically that suppressed HER2/neu-specific Th1, we show that intratumoral treatment with the oncolytic VVGMCSF is ineffective at reducing tumor growth nor does it lead to the development of a systemic tumor-specific immune response. However, when combined with a neu-encoding vaccinia VVneu and administered into the tumor microenvironment, mice develop systemic anti-neu immunity, significant reduction in tumoral and systemic MDSC and manifest a major antitumor response. The same virus combination (vaccine) fails to generate a similar response when given systemically in NBT1-bearing mice. We characterize the presence and function of the MDSC response and the resultant cytotoxic T lymphocyte (CTL) and interferon responses to HER2/neu. These results point to the ability of the tumor microenvironment to both promote immune escape and act as an effective vaccination site for tumor antigen-encoding oncolytic viruses that can result in a systemic immune response capable of mediating tumor regression.

Materials and methods

Cell culture

The HER2/neu-expressing NBT1 mammary tumor cell line was derived by serial cell culture of tumor cells obtained from a spontaneous breast tumor that arose in a FVB/neuT mouse transgenic for a HER2/neu receptor that is constitutively activated by a point mutation in the transmembrane region. NBT1 was maintained in media composed of CMRL-1066 (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum, 2 mM l-glutamine, 50 IU ml−1 penicillin/streptomycin and 4 μM dexamethasone. Splenic cell cultures used in immune analyses were carried out using a supplemented RPMI-1640 media as described.21 The Sf9 and Sf21 cell lines (Invitrogen) were maintained in Grace’s supplemented medium (Invitrogen) supplemented with 10% fetal bovine serum and 50 IU ml−1 penicillin/streptomycin. HER2/neu Sf21 cell lysates used for in vitro restimulation of the CTL assay described below were prepared by resuspending cells in RPMI medium with protease inhibitor cocktail (Sigma-Aldrich, St Louis, MO), lysing via four freeze/thaw cycles, sonicating for 2 min and centrifugating for 15 min at 13 500 rcf.

Construction of recombinant vaccinia virus expressing HER2/neu (VVneu)

The rat neu complementary DNA (cDNA) was cloned from pSV2-neuNT (provided by RA Weinberg), ligated into pBluescript (Stratagene, La Jolla, CA, USA) using SalI and Hind III restriction sites, and then removed from pBluscript using SalI and NotI. It was cloned into vaccinia recombination plasmid pSC11/9. VVGMCSF was prepared as previously described.21, 22, 23 VVβGal was provided as a negative control by Dr Laurence Eisenlohr (Thomas Jefferson University, Philadelphia, PA, USA).

Construction of recombinant Baculovirus expressing rat HER2/neu

The rat ne u cDNA-coding sequence was assembled in the pFASTBac/CT-TOPO recombination vector (Invitrogen). Rat neu-transforming cDNA was excised from a vaccinia recombination plasmid containing the full-length rat neuNT cDNA sequence24 in the vector pSC11/9 using NotI and SalI restriction enzymes. The 4727 bp rat neu DNA was gel purified and then ligated to pBluescript II SK(+) (Stratagene) at Not/Sal to produce plasmid BSratneuNT. This neuNT-containing plasmid was used to generate three DNA fragments which, when ligated together, prepared the rat neuNT-coding sequences for ligation into pFASTBac/CT-TOPO. pBSratneuNT was first digested with HincII, which cut at two sites: a HincII site at residue 1599 and the SalI site at the 3'-end of the neuNT cDNA, residue 4622. HincII digestion released a 3023-bp fragment, residues 1599–4622, which contained two problematic NcoI sites. The HincII-cut vector was gel purified, religated and cloned and then digested with NcoI. Filling in the NcoI site, located at position −19, with T4 DNA polymerase followed by ligation transformed the NcoI site into a NsiI site, creating plasmid BSNeuNsi. Plasmid BSNeuNsi was digested with both SbfI and SalI releasing a 1366-bp fragment and leaving 338 bp of cDNA containing 233 bp of the 5'-coding sequence. Plasmid pBSratneuNT was next digested with both SbfI and SalI releasing a 4389-bp fragment, residues 234–4622, which, after gel purification, was cut with NcoI to generate a 2775-bp middle fragment comprising residues 234–3008. Lastly, pBSratneuNT was PCR-amplified using primers NTNco (5'-gcatagcggccgccatggacagtaccttctaccgtt-3') and NTSal (5'-ctacgcgtcgacacaggtacatccaggcctaggtac-3') to produce a 768-bp fragment of the neu-coding sequence comprising residues 3009–3777. At its 3' end, this amplicon contained a SalI site for cloning and a final residue codon mutation of GTA to GTG. The NTNco/NTSal PCR amplicon—residues 3009–3777—was used without subcloning in a three-way ligation, with the SbfI/NcoI middle fragment—residues 234–3008—and the plasmid BSNeuNsi—residues −105 to 233—cut with SbfI/SalI. Following transformation, clone selection and DNA sequencing, a correct full-length neuNT cDNA clone was identified, pBSNeuNsiSal. This plasmid was cut with both NsiI and SalI to release the neuNT cDNA and was treated with mung bean nuclease to remove 3' and 5' overhangs, generating a blunt-end cDNA. Treatment with Antarctic alkaline phosphatase removed terminal phosphates allowing neuNT cDNA to be TOPO cloned into pFASTBac/CT-TOPO vector. The recombinant bacmid was transfected into the Sf9 cell line to produce P1baculovirus stock expressing HER2/neu (bac HER2/neu), which was further amplified with several rounds of infection of Sf21 cells. Bac HER2/neu was used to infect Sf21 cells. Infected HER2/neu Sf21 cells were pelleted and used to generate Sf21neu lysate for restimulation and enzyme-linked immunosorbent assay as described below.

Animal experiments

Four to 6-week-old FVB/N mice (Jackson Labs, Bar Harbor, ME, USA) were maintained in a high-efficiency particulate air-filtered cage system for at least 1 week prior to use. For in vivo experiments involving tumor-bearing mice, anesthetized FVB/N mice were injected using a 27-gauge tuberculin syringe (Becton Dickinson, Franklin Lakes, NJ, USA) into the right second mammary fat pad with 2 × 106 NBT1 cells suspended in Hanks balanced salt solution (Sigma-Aldrich) as previously described.25 For in vivo vaccination studies in naive FVB/N mice, mice received 2 subcutaneous (s.c.) or intra-mammary fat pad (i.m.f.) injections, 2 weeks apart, using a cocktail of 1 × 106 plaque-forming unit (p.f.u.) VVGMCSF plus either 1 × 106 p.f.u. VVneu or VVβGal using a 27-gauge tuberculin syringe. In in vivo vaccination studies of NBT1 tumor-bearing mice, mice were treated with 2 s.c. or intratumoral (i.t.) injections, 2 weeks apart, of 1 × 106 p.f.u. VVGMCSF plus 7.5 μg keyhole limpet hemocyanin (KLH) (Sigma-Aldrich) with either 1 × 106 p.f.u. VVneu or VVβGal. For vaccinations, i.m.f. injections were made directly into the number 2 fat pad. All i.t. injections were made directly into the mammary fat pad tumors. S.c. injections were placed in the contralateral (left) groin. All viral injections used the same total of 2 × 106 vaccinia p.f.u. Five animals per group were used for each independent in vivo experiment. Animal experiments were conducted in accordance with protocols approved by the Rutgers Institutional Animal Care and Use Committee.

CTL assays

CTL assays were performed as previously described,21, 26 with several modifications. Briefly, effector cells were prepared from splenocytes, vaccination site-draining lymph nodes (VDN) or tumor-draining lymph nodes of treated/vaccinated FVB/N mice. Spleens or lymph nodes were homogenized and red blood cells were lysed using ammonium chloride buffer (ACK buffer, 0.15M NH4Cl, 1.0 mM KHCO3 and 0.1 mM EDTA), washed with tissue culture medium (TCM) and filtered through a 70-μm nylon mesh (BD Biosciences, San Jose, CA, USA). Effector cells were resuspended at 7 × 106 cells ml−1 and cultured with 3 × 106 cells ml−1 of irradiated (25 Gray) splenocytes from naive FVB/N mice that had been incubated overnight in 50 ml conicals at 37 °C at 4 × 106 cells ml−1 in TCM+2-mercaptoethanol with 300 μg ml−1 of the RNEU420–429 (PDSLRDLSVF) (Genscript, Piscataway, NJ, USA) immunodominant epitope of the rat neu protein,27 lymphocytic choriomeningitis virus nucleoprotein NP118–126 (RPQASGVYM) control peptide (Genscript) or indicated cell lysates. Restimulation cultures were in a total of 2 ml of TCM+2-mercaptoethanol in 24-well plates for 5 days at 37 °C, 5% CO2. On day 3 of restimulation, 50 μl of supernatant was assayed for interferon-γ using an enzyme-linked immunosorbent assay, as described previously.28 On day 5, stimulated effector cells were collected and cultured for 4 h with 51Cr-labeled NBT1 cells as targets. Hundred μl of supernatant were removed and 51Cr release of target cells was measured with a gamma counter (Packard Bioscience, Meriden, CT, USA). Percent-specific lysis was calculated from the formula [(experimental release−spontaneous release) × 100/(maximal release in 1% SDS−spontaneous release)].

Magnetic bead depletion of MDSCs

MDSCs were depleted from the splenocyte population of NBT1 tumor-bearing mice using a mouse MDSC isolation kit (Miltenyi Biotec, Auburn, CA, USA) as per the manufacturer’s instructions. Briefly, 1 × 108 splenocytes were resuspended in 350 μl MACS buffer (phosphate-buffered saline, 0.5% bovine serum albumin and 2 mM EDTA) and blocked for 10 min with 50 μl FcR blocking reagent at 4 °C. Hundred μl of anti-Ly6G-biotin was added and incubated for 10 min at 4 °C. Cells were washed in 10 ml MACS buffer and resuspended in 100 μl MACS buffer. Two hundred μl anti-biotin microbeads were added and incubated at 4 °C for 15 min. Magnetic separation was performed using LS disposable column and MACS Magnetic separator (Miltenyi Biotec). The negative fraction was collected and restimulated as described above. MDSC depletion was validated using flow cytometry.

Flow cytometry

Flow cytometry for RNEU420–429 tetramer (provided by the NIAID Tetramer Facility, Atlanta, GA, USA) was conducted as described previously.21 Briefly, effectors from treated mice were restimulated for 5 days as described above. Effectors were collected, washed and resuspended in phosphate-buffered saline/5% fetal bovine serum with 0.1% w/v sodium azide at 1 × 106 cells per 100 μl. Cells were stained with anti-CD8α-fluorescein isothiocyanate at the manufacturer’s recommended concentration (BD Biosciences) and the RNEU420–429-specific tetramer labeled with phycoerythrin (1:100 dilution). Flow cytometry for MDSC was conducted on splenocytes, lymph node cells and tumor suspensions straight ex vivo. Tumor suspensions were prepared as described previously,29 with modifications. Tumors were cut into small pieces and incubated at 37 °C with gentle shaking for 20 min in 5 ml RPMI with collagenase D (1 mg ml−1, Roche, Diagnostics, Indianapolis, MO). Tumor pieces were dissociated with a metal strainer, washed in TCM+2-mercaptoethanol, red blood cells were lysed using ACK buffer, washed again with TCM and filtered through a 70-μm nylon mesh (BD Biosciences). Cells were stained using anti-Gr-1 PE (eBioscience, San Diego, CA, USA), CD11b fluorescein isothiocyanate (eBioscience), anti-Ly6G APC (BD Biosciences) and anti-Ly6C PE-Cy7 (eBioscience). Flow cytometry data were acquired using a FC-500 flow cytometer (Beckman Coulter, Miami, FL, USA) and analyzed using CXP (Beckman Coulter) provided by the CINJ flow cytometry shared resource.

Statistical analysis

Results were expressed as mean±s.e. Significance for experiments with >2 conditions (P<0.05) was determined by analysis of variance with the post hoc Tukey multiple comparisons test using the InStat software package (GraphPad Software, La Jolla, CA, USA). Significance for experiments with two conditions was determined using the Student’s T-test, two-tailed with unequal variances in Microsoft Excel (Microsoft, Redmond, WA, USA). Kaplan–Meier survival curves were generated using Prism (GraphPad Software), with significance calculated using the log-rank (Mantel–Cox) test and correction for multiple comparisons using the Bonferroni method.

Results

Vaccination of naive FVB/N mice results in a systemic HER2 MHC class I epitope-specific CTL response

We hypothesized that vaccination of a naive host with a vaccinia construct expressing HER2/neu would induce an antigen-specific Th1 immune response, regardless of the site of vaccination. We utilized i.m.f. injection or contralateral s.c. injection of VVneu+VVGMCSF in naive FVB/N mice. Two weeks after the final vaccination, splenocytes and draining lymph nodes of the vaccination site (VDN) were restimulated with irradiated splenocytes from naive female FVB/N mice that had been pulsed with immunodominant RNEU420–429 peptide (Figure 1a). Both s.c. and i.m.f. vaccination with VVneu and VVGMCSF resulted in increased percent-specific lysis of 51Cr-labeled NBT1 tumor cells by splenic CTLs in vitro, when compared with controls (Figure 1b).

Figure 1
figure 1

Vaccination of naive FVB/N mice with VVneu induces a systemic HER2-specific Th1 response. (a) Female FVB/N mice were injected twice, 2 weeks apart, with VVneu+VVGMCSF (either subcutaneous (s.c.) or intra-mammary fat pad (i.m.f.)), VVBGal+VVGMCSF (i.m.f.) or vehicle (i.m.f.). Two weeks after the second and final vaccination, spleens were restimulated with irradiated splenocytes from naive female FVB/N mice that had been pulsed with immunodominant RNEU420–429 peptide. (b) NBT1 tumor-specific systemic (spleen) cytotoxic T lymphocyte (CTL) activity against NBT1 target cells after restimulation. Differences of both s.c. and i.m.f. VVneu+VVGMCSF groups compared with controls were significant (**P<0.01, *P<0.05). Results are representative of three independent experiments.

Full size image

MDSCs suppress the systemic antitumor immune response

MDSCs are a heterogeneous population of immature granulocytic and monocytic cells that have inhibitory effects on tumor-specific T-cell activation and function.5, 6, 30 Given previous studies demonstrating the importance of a type 1 antitumor response in effective immune therapies,21, 28 we hypothesized that MDSCs inhibited a systemic antitumor immune response in the NBT1 model. To determine whether the presence of orthotopic NBT1 tumor affected levels of MDSC in FVB/N mice, we injected FVB/N mice with NBT1 cells in the right number 2 mammary fat pad. After 4–5 weeks of tumor growth, we used flow cytometry to measure levels of Gr-1, Ly-6G, Ly-6C and CD11b on cells from spleen and tumor. The CD11b+Gr1+ population was significantly increased in the systemic population of tumor-bearing mice compared with naive mice (Figure 2a). A large population of CD11b+Ly-6CintLy6G+ cells was detected, a phenotype consistent with granulocytic MDSCs (gMDSCs, Figure 2b).

Figure 2
figure 2

NBT1 growth in female FVB/N mice leads to increased levels of Ly6G+ gMDSC, both systemically and in the tumor microenvironment. (a) Flow cytometry was performed on spleen, tumor-draining lymph node (TDN) and tumor samples stained for Gr-1, CD11b, Ly-6C and Ly-6G. Levels of CD11b+Gr1+ cells in spleen, TDN and tumor were compared with levels in spleen and axillary lymph node (LN) of naive female FVB/N mice. Cumulative results are presented from five independent experiments (*P<0.05). (b) Cells from spleen and tumor were gated to include monocytic and granulocytic populations for evaluation of Gr-1 and CD11b. This population was then further gated for CD11b+ cells, which were evaluated for Ly-6G and Ly-6C. Results are representative of three independent experiments. NS, not significant.

Full size image

To demonstrate the immune-suppressive nature of the MDSCs induced by NBT1, we used a MDSC magnetic bead isolation kit (Miltenyi Biotec, Auburn, CA) to deplete Ly6G+ cells from the splenocyte population from NBT1 tumor-bearing mice. We validated depletion of MDSCs from the negative fraction using flow cytometry staining for CD11b, Gr-1, Ly6G and Ly6C (Figure 3a). We then restimulated this MDSC-depleted splenocyte population with irradiated splenocytes from naive female FVB/N mice that had been pulsed with the immunodominant major histocompatibility complex (MHC) class I RNEU420–429 peptide or HER2/neu protein in Sf21 neu baculovirus-infected lysate, with NP118–126 peptide and unmodified Sf21 cells used as restimulation controls. Interferon-γ production by MDSC-depleted splenocytes was increased in most conditions compared with pre-depletion controls, with HER2/neu-restimulated effectors significantly increased compared with all other conditions (Figure 3b).

Figure 3
figure 3

Myeloid-derived suppressor cells (MDSCs) suppress a HER2/neu-specific Th1 response in NBT1 tumor-bearing mice. MDSCs from spleen of NBT1 tumor-bearing mice were depleted using a MDSC (Ly6G+) magnetic bead isolation kit. (a) Flow cytometry was performed on pre- and post-depletion samples stained for Gr-1, CD11b, Ly-6C and Ly-6G. (b) The MDSC-depleted negative fraction was restimulated with irradiated naive splenocytes pulsed with control NP118–126 peptide, RNEU420–429 peptide, lysate from Sf21 insect cells that had been infected with baculovirus expressing HER2/neu and control Sf21 cell lysate. Interferon (IFN)-γ enzyme-linked immunosorbent assay was performed in triplicate on supernatant from MDSC-depleted effectors. Results are representative of three independent experiments. (†P<0.001 compared with all other stimulation conditions.).

Full size image

Vaccination with VVneu+VVGMCSF into the tumor microenvironment, but not VVGMCSF alone, results in a systemic HER2-specific CTL response, decreased MDSCs and tumor regression

We asked whether vaccination with HER2/neu-antigen-encoding VVneu and VVGMCSF could overcome MDSC-associated anergy against HER2/neu induced by the NBT1 model. After 2 weeks of orthotopic NBT1 tumor growth, two doses of VVneu+VGMCSF+KLH, 2 weeks apart, were administered subcutaneously or by injection into the tumor microenvironment (i.t., Figure 4a). KLH was added to all treatment groups and controls, as previous studies by our group and others had demonstrated its effectiveness as a vaccine adjuvant.21, 31, 32 Two weeks after the final vaccination, splenocytes and VDN were restimulated with irradiated splenocytes from naive female FVB/N mice that had been pulsed with immunodominant RNEU420–429 peptide. Splenocytes from mice vaccinated with i.t. VVneu+VVGMCSF+KLH showed a significantly increased ability to lyse target NBT1 cells after in vitro restimulation when compared with both equivalent s.c. vaccination and i.t. VVBGal+VVGMCSF+KLH (Figure 4b). Similarly, cytolytic activity of the VDN population was highest in mice given i.t. VVneu+VVGMCSF+KLH (Figure 4c). Increased CTL activity was specific to the RNEU420–429 MHC class I epitope, as restimulation with control NP118–126 did not result in increased cytolytic activity (not shown). The percentage of restimulated CD8+ lymphocytes that were RNEU tetramer-positive from both spleen and VDN was significantly higher in mice that were treated with i.t. VVneu+VVGMCSF+KLH than in all other groups (Figure 4d).

Figure 4
figure 4

Vaccination with VVneu into the tumor microenvironment generates a systemic antitumor cytotoxic T lymphocyte (CTL) response. (a) Female FVB/N mice were injected with 2 × 106 NBT1 tumor cells into the right second mammary fat pad. Mice were treated twice (days 14 and 28) with VVneu+VVGMCSF+KLH (subcutaneous (s.c.) or intratumoral (i.t.)), VVBGal+VVGMCSF+KLH (i.t.) or vehicle control (i.t.). On day 42, spleens and vaccination site-draining lymph nodes (VDN) were restimulated with irradiated naive splenocytes that had been pulsed with RNEU420–429 peptide. (b) Ability of splenocyte effectors to lyse target NBT1 cells was measured by the 51Cr release assay. Difference between i.t. VVneu+VVGMCSF+KLH and all other conditions was significant (**P<0.01). (c) Tumor-specific CTL activity of VDN effector cells was measured by 51Cr lysis. Differences between i.t. VVneu+VVGMCSF+KLH and all other groups were statistically significant (**P<0.01, *P<0.05). (d) Restimulated effectors were evaluated for RNEU tetramer-positive CD8+ cells using flow cytometry. The percent of CD8+ cells that are RNEU tetramer positive was measured, with representative dot plots and cumulative results of three independent experiments presented (**P<0.01). KLH, keyhole limpet hemocyanin.

Full size image

MDSC levels were evaluated using flow cytometry. In mice treated with i.t. VVneu+VVGMCSF+KLH, MDSC levels in spleen on day 28 were significantly decreased compared with i.t. vehicle (Figure 5). Systemic MDSC levels were not significantly decreased in mice treated with s.c. VVneu+VVGMCSF+KLH. Notably, tumor-bearing mice treated with i.t. VVBGal+VVGMCSF+KLH exhibited a significant increase in intratumoral, although not systemic, MDSC levels (Figure 5). A potential explanation may be that, in addition to its enhancing effects on antigen presentation, GM-CSF is one of multiple cytokines that support development of MDSCs.30

Figure 5
figure 5

Vaccination with VVneu into the tumor microenvironment results in decreased levels of systemic and tumor microenvironment myeloid-derived suppressor cells. Splenocytes and tumor were stained ex vivo for Gr-1 and CD11b using flow cytometry on day 28. Representative dot plots and combined mean %CD11b+Gr-1+ cells from three independent experiments are shown (*P<0.05 compared with intratumoral (i.t.) vehicle tumor, **P<0.01 compared with i.t. vehicle spleen, ¥P<0.05 compared with i.t. VVBGal+VVGMCSF+KLH tumor; all other comparisons are not significant). KLH, keyhole limpet hemocyanin; s.c., subcutaneous.

Full size image

Primary tumors in mice vaccinated with i.t. VVneu+VVGMCSF+KLH regressed from peak size and were significantly smaller than all other treatment conditions on day 42, including equivalent s.c. VVneu and i.t. VVGMCSF without antigen (Figures 6a–c). S.c. VVneu+VVGMCSF+KLH vaccination or i.t. VVBGal+VVGMCSF+KLH did not inhibit NBT1 growth.

Figure 6
figure 6

Vaccination with VVneu into the tumor microenvironment results in regression of NBT1 tumor. (a) Mean tumor size and s.d. from five combined independent experiments are shown. (b) Day 42 tumor size from five combined experiments, each point representing one mouse from the indicated treatment condition (†P<0.001). (c) Kaplan–Meier survival curve from five combined experiments (†P<0.001). KLH, keyhole limpet hemocyanin; i.t., intratumoral; s.c., subcutaneous.

Full size image

Discussion

The studies presented here demonstrate, using an aggressive orthotopic model of HER2/neu-driven mammary tumor (NBT1), that growth of the tumor results in a significant infiltration of intratumor and systemic MDSC, resulting in the lack of development of systemic immunity and anergy to peripheral s.c. immunization using a neu-encoding vaccinia vaccine that is effective in naive (non-tumor bearing) mice. Treatment of tumor-bearing mice using i.t. oncolytic VVGMCSF fails to induce systemic immunity nor does it lead to tumor regression. In fact, intratumoral VVGMCSF and resulting GM-CSF expression in tumor leads to a significant expansion of tumor resident MDSC, as suggested by the Vonderheide group.33 However, when tumor antigen-encoding VVneu is combined with VVGMCSF, thus providing a virally expressed tumor antigen to the tumor microenvironment, the combination results in the generation of systemic neu-specific immunity, a significant reduction in tumor and systemic MDSC and a significant antitumor response. We confirmed the contribution of MDSCs to immune escape in our model using flow cytometry to identify a population of CD11b+ Ly-6CintLy-6G+ cells. We also demonstrated that these cells have suppressive effects by finding, despite an overall increased background of IFN-γ in MDSC-depleted cultures, increased HER2/neu-specific IFN-γ production of the systemic splenocyte population after depletion of Ly-6G+ cells.

The goal of tumor immunotherapy is the development of modalities given alone or in combination that result in the generation of an effective systemic tumor-specific immune response by the host that can overcome immune escape mechanisms. Original studies from our laboratory using a murine bladder cancer model demonstrated the unexpected expansion of antigen-specific (tetramer positive) CD8+ T cells in the tumor microenvironment, but not systemically.21 This led us to immunize intratumorally using a vaccinia-encoding vaccine with resultant induction of systemic T-cell immunity.21 Coincidently, studies from the Schlom group using a colon cancer model demonstrated the induction of significant antitumor immunity, when priming or boosting immunizations were given intratumorally.34 These initial responses to intratumoral poxvirus-based vaccines have been translated to phase I trials in prostate cancer35 and by us in pancreatic cancer36 using antigen-encoding non-replicating fowlpox vectors.

The strategy of delivering an oncolytic virus to a tumor has long been studied with the overall hypothesis being that virus-mediated oncolysis could have primary antitumor effectiveness and as a sequellae of tumor lysis, antigen released into the tumor microenvironment could lead to a consolidating systemic antitumor response active against metastases. As recently reviewed by Lichty et al,37 preclinical and clinical studies are ongoing that use a variety of oncolytic vectors and encoded immune regulatory molecules. Our initial studies focused on the potential of generating systemic immunity using the oncolytic VVGMCSF, which would have the potential of lysing tumors, eliciting antigens and enhancing antigen presentation via the GM-CSF. This agent studied by others as JX-594 was subsequently characterized as to its bioavailability, and antitumor and immune responses.19, 20, 38 Of particular note in these studies, JX-594 was shown to preferentially persist and replicate in the tumor microenvironment following intravenous administration in preclinical models and in patients.19, 20 Given that accessibility of tumor for local administration is a limitation to the use of these approaches, this finding suggests that the uniqueness of the vaccinia platform may allow systemic delivery of virus, immune-modulating encoded cytokines, as well as antigen as we present here. Clinical studies using another DNA viral vector HSV-1-GMCSF construct (OncoVEXGM-CSF) have also demonstrated local infectivity, antitumor activity and local and systemic immune responses39, 40 further supporting the potential of this approach. Ongoing preclinical and clinical studies combining the oncolytic virus with other modalities such as chemotherapy and immune checkpoint inhibitors are also under development and ongoing by a number of investigators.41, 42

In conclusion, the studies presented here provide support for delivering to the tumor microenvironment a combination of virally encoded tumor antigen and immunomodulating GM-CSF, both encoded by the oncolytic vaccinia vector. Our studies demonstrate the requirement for antigen delivered to the microenvironment in overcoming MDSC-based immune escape and subsequent development of a systemic antitumor immune response in our model. Our and others’ studies have proven the feasibility of using DNA viral vectors to deliver multiple gene constructs to the tumor microenvironment via direct injection18, 19, 20, 38, 39, 40 and in some cases following intravenous administration.19, 20 These combined studies support the conclusion that this approach has the potential of impacting tumor alone, and supports future approaches combining this with added immune targeted and antitumor approaches.