Intra-tumoral injection of CpG oligonucleotides induces the differentiation and reduces the immunosuppressive activity of myeloid-derived suppressor cells

Abstract

Immunostimulatory CpG oligonucleotides (ODN) activate cells that express TLR 9 and have been shown to improve the host’s response to tumor antigens. Unfortunately, the immunosuppressive microenvironment that surrounds many cancers inhibits Ag-specific cellular responses and thus interferes with CpG-mediated immunotherapy. Myeloid-derived suppressor cells (MDSC) represent an important constituent of this immunosuppressive milieu. Large numbers of MDSC are present in and near tumor sites where they inhibit the activity of antigen-specific T and NK cells. Current studies indicate that the delivery of CpG ODN directly into the tumor bed reduces the immunosuppressive activity of monocytic (CD11b⁺, Ly6G⁻, Ly6C^high) MDSC. Monocytic MDSC express TLR9 and respond to CpG stimulation by i) losing their ability to suppress T cell function, ii) producing Th1 cytokines and iii) differentiating into macrophages with tumoricidal capability. These findings provide insight into a novel mechanism by which CpG ODN contribute to tumor regression, and support intra-tumoral injection as the optimal route for their delivery.

Introduction

Synthetic oligodeoxynucleotides (ODN) containing unmethylated CpG motifs mimic the ability of bacterial DNA to stimulate the innate immune system. CpG ODN trigger cells that express toll-like receptor (TLR) 9, thereby promoting the maturation and improving the function of professional antigen-presenting cells (APCs) while supporting the generation of Ag-specific B cells and CTL (1-3). Preclinical and clinical trials indicate CpG ODN have potent immunostimulatory effects that enhance the host’s response to cancer (4,5). Kawarada et al and Heckelsmiller et al showed that CpG ODN facilitated the induction of tumor-specific immunity and memory (6,7). This involved both improved pDC entry into the tumor site and the activation of tumor-specific CD8⁺ CTL and NK cells. This activity was optimized by direct injection of CpG ODN into the tumor, as CpG DNA was far less effective when delivered systemically (6,7). Virtually all studies to date examined the effect of CpG ODN on nascent tumor foci and tumors <5 mm in diameter. This work extends those studies to better understand the effect of CpG ODN on tumors of clinically relevant size (>1 cm diameter).

Despite evidence that tumor-specific CTL are expanded in the periphery, immune-mediated tumor destruction is difficult to achieve by any form of immunotherapy. For example, CpG ODN administered alone or in combination with vaccines promote the induction of tumor-specific cellular and humoral immune responses yet rarely lead to prolonged tumor regression (4,5,8). Analysis of the tumor microenvironment indicates that the lytic activity of CTL and NK cells is suppressed by regulatory T lymphocytes (Treg), myeloid-derived suppressor cells (MDSC) and/or M2 macrophages surrounding the tumor (9,10). Thus, it appears that successful immunotherapy will require both the amplification of tumor-specific immunity plus a means of reversing tumor-associated immune suppression.

MDSC are key contributors to the inhibitory microenvironment found at the tumor site. MDSC are a heterogeneous population of early myeloid progenitors that arise in the bone marrow (11,12). Their numbers are expanded in the peripheral lymphoid organs of cancer patients and they frequently constitute a majority of tumor-infiltrating cells. Two distinct subpopulations of MDSC have been identified: both are Gr-1⁺ and CD11b⁺ with granulocytic MDSC being Gr-1^hi, Ly6g⁺ and Ly6c^low while monocytic MDSC (mMDSC) are Gr-1^intermediate, Ly6g^lo and Ly6c^hi. Although both subsets suppress T and NK cell responses through the production of arginase 1 and/or inducible nitric oxide synthase (iNOS), mMDSC show greater suppressive activity on a per cell basis (13-15). In addition, mMDSC promote the generation and/or expansion of Tregs (16). An agent capable of blocking the immunosuppressive activity of mMDSC might therefore improve the efficacy of tumor immunotherapy.

This study examines the effect of CpG ODN on mMDSC. Consistent with earlier work, intra-tumoral injection of CpG (but not control) ODN promoted tumor regression. Within the tumor microenvironment, CpG ODN treatment increased the number of tumor infiltrating T and NK cells while decreasing the frequency and inhibitory activity of resident mMDSC. Further results showed that the effect of CpG ODN on TLR9-expressing mMDSC included i) triggering their rapid production of Th1-type cytokines (including IL-6, IL-12 and TNFα), ii) impairing their ability to secrete arginase 1 and nitric oxide (factors critical to their suppression of T cell activity) and iii) inducing their differentiation into tumoricidal macrophages. These results suggest additional mechanisms though which CpG ODN could promote tumor regression.

Materials and methods

Animals and tumor cell lines

BALB/c and C57Bl/6 mice were obtained from the National Cancer Institute (Frederick, MD) and studied at 6-10 weeks of age. CD8 TCR Tg mice specific for peptide 518-526 of PR8 HA were a gift from Dr. T. Sayers (NCI). OVA 257-264 specific CD8 TCR Tg mice (OT-I) were obtained from the Jackson Laboratory (Bar Harbor, ME). All studies were approved by the NCI Frederick Animal Care and Use Committee (ACUC). The CT26 colon cancer cell line was a gift from Dr. Z. Howard (NCI). The following cell lines were purchased from American Type Culture collection (Manassas, VA): E.G7, which is a CD8⁺ T cell line derived from the EL-4 thymoma transfected to express ovalbumin; TC-1, which is a lung epithelial tumor cell line that expresses the E7 oncoprotein from human papilloma virus 16; and 4T1, which is a breast cancer cell line.

Oligodeoxynucleotides and reagents

Phosphorothioate ODNs were synthesized at the Core Facility of the Center for Biologics Evaluation and Research, FDA (Bethesda, MD). The following ODNs were used: CpG ODN 1555 (GCTAGACGTTAGCGT) and control ODN 1612 (GCTAGAGCTTAGCGT). All ODN were free of detectable protein or endotoxin contamination.

Peptideglycan, MPL and imiquimod were purchased from Invivogen (San Diego, CA). Poly(I:C) was purchased from Sigma-Aldrich (St. Louis, MO)

In vivo tumor studies

Mice were injected s.c. with 10⁵ CT26 tumor cells. Solid tumors formed that reached a diameter of .1 cm after 2 wk, at which time they were injected with 200 ug of CpG or control ODN. Tumor size was calculated by the formula: (length × width × height)/2. Tumor growth curves were generated from 3 - 5 mice per group and all results were derived by combining data from 2 - 3 independent experiments. Any animal whose tumor exceeded a diameter of 2.0 cm was immediately euthanized as per ACUC protocol.

To deplete NK and CD8⁺ T cells, mice were injected i.p. with 500 ug of rat anti-mouse CD8 (53.6.72) Ab, 25 ul of ascites containing anti-asialo GM1 Abs (Wako Pure Chemical Industries, Ltd., Japan) or 500 ug of control Ab (LTF-2) from Bio X Cell (West Lebanon, NH). These Abs were delivered i.p. on days 12, 14, 17 and 20 post tumor challenge.

Preparation of monocytic MDSC

Two techniques were used to prepare mMDSC from the spleens of tumor bearing mice. Single spleen cell suspensions were layered onto Ficoll (density 1.083 g/ml) and centrifuged for 20 min at 2000 rpm. Cells at the interface were isolated, stained, and FACS sorted to isolate CD11b⁺, Gr-1^intermediate, Ly6c⁺ MDSC (purity by this method was > 98%). Alternatively, mMDSC were purified by magnetic cell sorting using the mouse MDSC isolation kit according to the manufacturers’ instructions (MACS; Miltenyi Biotec, Auburn,CA). The purity of CD11b⁺/Ly6c⁺ cells by this method was 90 - 95% as determined by flow cytometry.

Flow cytometry

Cells were washed with PBS, fixed in 4% paraformaldehyde for 10 min, and stained with anti −CD11b, −CD8, −Dx5, −Gr-1, −Ly6c, −Ly6g, −F4/80, and/or −CD45 Abs for 30 min at 4° C. All Abs were obtained from BD Pharmingen (San Diego, CA). Stained cells were washed, re-suspended in PBS/0.1% BSA plus azide and analyzed by FACS Calibur (BD Pharmingen, San Diego, CA).

Detection of intra-cytoplasmic arginase I and IL-12

Single cell suspensions were prepared from the spleens of tumor bearing mice. These cells were surface stained to detect CD11b, Gr-1 and Ly6c triple positive cells. They were then treated with cell permeabilization solution (BD Pharmingen) followed by anti-arginase I, anti-IL12 or control Ab followed as needed by PE-conjugated anti-goat Ab. The frequency of internally stained cells expressing the surface markers of mMDSC was determined by FACS.

RT-PCR and quantitative RT-PCR

Total RNA was extracted from target cells using TRIzol reagent (Life Technologies Inc., Carlsbad, CA) as recommended by the manufacturer. 1 ug of total RNA was reverse-transcribed in first strand buffer (50 mM Tris-HCl, pH 7.5, 75 mM KCl, and 25 mM MgCl2), containing 25 ug/ml oligo-(dT), 200 U Moloney leukemia virus reverse-transcriptase, 2 mM dinucleotide triphosphate, and 10 mM dithiothreitol. The reaction was conducted at 42 C for 1 hr. A standard PCR was performed on 1 ul of the cDNA synthesis using TLRs primer pairs (Invivogen, San Diego, CA). Aliquots of the PCR reactions were separated on a 1.2% agarose gel and visualized with UV light after ethidium bromide staining.

Arginase I mRNA levels were examined using the Applied Biosystems StepOne RT-PCR system, in which primers obtained from the Gene Expression Assay set (Applied Biosystems, Foster City, CA) were amplified using the TaqMan Gene Expression Master Mix kit. mRNA expression levels were then calculated by step-one software (Applied Biosystems) after correction for GAPDH expression independently for each sample.

T cell proliferation assay

HA or OVA-specific CD8 T cells were purified by magnetic cell sorting (MACS; Miltenyi Biotec) using mouse anti-CD8 beads and were labeled with carboxyfluorescein succinimidyl ester (CFSE) (Invitrogen, Carlsbad, CA) as previously described (17). 5 × 10⁵ CD8 T cells were co-cultured with 10⁶ mitomycin C-treated naive spleen cells plus 5 × 10⁵ Ly6c⁺ MDSC in the presence of peptide (0.1 ug/ml) for 3 days. Cell division was analyzed using a FACSCalibur to monitor cellular CSFE content.

Nitrite assay

NO levels in culture supernatants were assessed using the Griess reagent (Sigma-Aldrich). Nitrite concentration was calculated by comparison to a standard curve generated by sequentially diluting sodium nitrite.

CTL assay

mMDSC were incubated with 1 uM CpG or control ODN for 48 hr. Various numbers of these ODN-pulsed MDSC were added to 10⁴ CT26 target cells. After 4 h in culture, supernatants were recovered and released lactate dehydrogenase (LDH) assayed as recommended by the manufacturer (Roche Diagnostics, Indianapolis, IN). Mean percent specific lysis of triplicate wells was determined by the formula: % cytotoxicity = [(experimental - spontaneous LDH release)/(maximum - spontaneous LDH release)] × 100.

Statistical analysis

A two-sided unpaired Student’s t test was used to analyze tumor growth and cellular responses. P values < 0.05 were considered to be statistically significant.

Results

Intra-tumoral injection of CpG ODN reduces tumor growth

Previous studies established that intra-tumoral injection of CpG ODN could slow and in some cases reverse the growth of nascent tumors (<5 mm diameter) (6,7,18). Yet tumors of such small size are difficult to detect clinically and can lack the immunosuppressive microenvironment that inhibits the efficacy of immunotherapy directed against larger tumors (> 1 cm diameter).

To evaluate the effect of CpG ODN on tumors of clinically relevant size, BALB/c mice were injected s.c. with CT26 colon tumor cells. These tumors grew to >1 cm in diameter within 2 wk, at which time they were injected twice with 200 ug of CpG ODN. This treatment reduced the rate of tumor growth by >50%, whereas control ODN had no effect (p < 0.01, Fig 1A). The efficacy of intra-tumoral CpG injection significantly exceeded that of systemic ODN delivery (p < 0.01, Fig 1B). The impact of CpG treatment on established tumors involved both CD8 and NK cells, as co-administration of neutralizing Abs against either cell type abrogated CpG-dependent tumor regression (Fig 1C). In addition to delaying tumor progression, CpG treatment significantly increased the fraction of DX5⁺ NK cells and CD8⁺ T cells infiltrating the tumor (data not shown).

Figure 1.

Intra-tumoral injection of CpG ODN impacts tumor immunity.

10⁵ CT26 colon cancer cells were injected s.c. into BALB/c mice on day 0. These formed solid tumors that reached an average diameter of >1 cm by day 14. A-C) On days 14 and 15, mice were injected i.p. or intra-tumorally with 200 ug of CpG or control ODN, and tumor size monitored. C) Mice were also injected i.p. with anti-CD8, anti-asialo GM1 (NK cell) or control Ab on days 12, 14, 17 and 20. Data represent the mean + SE of 5-8 mice/group from 2 independent experiments.

* p < 0.05 compared with the untreated (tumor alone) group

** p < 0.01 compared with the untreated (tumor alone) group

Monocytic MDSC express TLR9 and produce cytokines in response to CpG stimulation

A majority of the immune cells infiltrating established tumors are of myeloid lineage and bear the CD11b surface marker (Fig 2A). This includes both myeloid derived suppressor cells (MDSC) that are Gr-1⁺/CD11b⁺/Ly6c⁺ and granulocytic MDSC that are Gr-1⁺/CD11b⁺/Ly6g⁺. mMDSC are extremely potent suppressors of tumor-specific immune responses (15). To examine the effect of CpG ODN treatment on MDSC, ODN were injected 14 days after the initiation of CT26 tumors (at which time tumor volumes ranged from 200 - 400 mm³). CpG treatment had no effect on granulocytic MDSC or on the total number of CD45⁺ cells infiltrating the tumor bed (Fig 2 A and Supplemental Fig 1). However, CpG treatment reduced the frequency of mMDSC infiltrating the tumor site by >3-fold (p. <0.01, Fig 2 A,B). This reduction in mMDSC number persisted for 4 days (Fig 2C) and was independent of tumor size (Supplemental Fig 2A). To determine whether intra-tumoral injection of CpG ODN had systemic as well as local effects on mMDSC, mice were injected bilaterally with CT26 cells. The infiltration of mMDSC was reduced in tumors injected with CpG ODN but remained high in contra-lateral tumors treated with control ODN (supplemental Fig 2B).

Figure 2.

Effect of intra-tumoral CpG ODN on CD11b⁺ cells.

A) Mice were treated as described in Fig. 1. The number of tumor infiltrating CD11b⁺ cells that expressed Ly6c and Ly6g was determined on day 17 by FACS. A) Representative results from one mouse/group and B) mean + SD from 5 independently analyzed mice/group, showing Ly6c and Ly6g expressing cells as a percentage of all tumor infiltrating CD45⁺ lymphocytes. C) The frequency of CD11b, Ly6c double positive cells as a percentage of CD45⁺ tumor infiltrating cells was examined 2 - 5 days post CpG administration. Mean + SD from 4-7 independently analyzed mice/group.

* p < 0.01 versus untreated group

CpG treatment also reduced the number of MDSC in the spleen (Supplemental Fig 2D). However this systemic effect was influenced by tumor volume, consistent with previous studies (19,20). Thus, when spleens from mice with tumors of equal size were compared, no effect of CpG treatment on mMDSC frequency was observed (Supplemental Fig 3C, E). This result indicates that the reduction of splenic MDSC number is a consequence of CpG-induced reductions in tumor size rather than an independent effect on mMDSC.

To better understand the effect of local CpG ODN treatment on tumor infiltrating mMDSC, the expression of TLR9 (the cognate receptor for CpG DNA) by mMDSC was examined. As seen in Fig 3A, TLR9 mRNA was readily detected in Gr-1⁺/CD11b⁺/Ly6c⁺ cells. Consistent with that observation, FACS purified mMDSC responded to in vitro stimulation with CpG ODN by secreting a variety of cytokines (including IL-6, TNFα and IL-12, Fig 3B and data not shown). The mMDSC origin of these cytokines was confirmed by FACS analysis, in that Gr-1⁺/CD11b⁺/Ly6c⁺ cells isolated from either the tumor or spleen of CpG treated mice contained high levels of intra-cytoplasmic IL-12 (Fig 3C). In the same studies, control ODN did not induce cytokine production. By comparison, the level of TLR9 mRNA expression by purified Gr-1/CD11b/Ly6g cells was very low, and granulocytic MDSC failed to secrete cytokines in response to CpG ODN stimulation (Fig 3A and Supplemental Fig 3).

Figure 3.

Monocytic MDSC express TLR9 and respond to CpG ODN.

Spleens were removed from CT26 tumor bearing mice and CD11b⁺, Ly6c⁺, Gr- 1^intermediate or CD11b, Ly6g⁺, Gr-1^hi MDSC were FACS sorted to >98% purity. (A) TLR9 mRNA levels were determined by RT-PCR in comparison to the RAW 264.7 macrophage cell line (positive control) and EL4 thymoma cell line (negative control). (B) 10⁵ FACS sorted MDSC were cultured with 1 uM CpG or control ODN for 24 hr. Culture supernatants were assayed for IL-12 levels by ELISA. Data represents the mean + SD from 3 independent experiments. (C) MDSC isolated independently from the spleen or tumor of CT26 tumor bearing mice were cultured with 1 uM CpG or control ODN for 8 h. Brefeldin A was added during the final 4 h of incubation. Cells were stained to identify CD11b⁺, Ly6c⁺, Gr-1^intermediate mMDSC for the presence of intracytoplasmic IL-12. All experiments were repeated three times with similar results.

* p < 0.01 versus untreated group

CpG-treated monocytic MDSC fail to inhibit T cell activation

mMDSC suppress the proliferation and functional activity of Ag-stimulated T cells (11,12). To examine the effect of CpG treatment on this inhibitory activity, mMDSC were isolated from tumor bearing mice. The studies shown utilized splenic rather than tumor infiltrating mMDSC, as the former could be isolated at higher yield and purity. However, all results involving splenic mMDSC were confirmed in more limited studies of tumor infiltrating cells (Supplemental Fig 4). mMDSC were mixed with HA-specific CD8⁺ T cells (isolated from HA TCR Tg mice). The vast majority of these T cells proliferated when stimulated with HA peptide (Fig 4 A,B). This proliferation was reduced by >74% when mMDSC were added to the culture (p. <0.01, Fig 4 A,B). Yet when the mMDSC were pre-treated with CpG ODN, their ability to suppress T cell proliferation was abrogated (p. <0.01, Fig 4 A,B). In contrast, no reduction in suppressive activity was observed when MDSC were pre-treated with control ODN (Fig 4B).

Figure 4.

CpG ODN treatment inhibits the suppressive activity of monocytic MDSC

A,B) mMDSC from the spleen of CT26 tumor bearing mice were isolated by MACS (final purity 90 - 95%). These cells were cultured for 3 h with 1 uM CpG or control ODN and then washed. 5 × 10⁵ ODN-pulsed MDSC were cultured for 3 days with 5 × 10⁵ CFSE-labeled CD8⁺ HA-specific Tg T cells plus 10⁶ mitomycin C treated spleen cells (as APCs) in the presence of 0.1 ug/ml HA peptide. CD8 T cell proliferation was monitored by CFSE dilution. A) Representative example, B) mean + SD (N = 5 independent MDSC preparations in 2 independent experiments). C) Mice bearing 1.5 cm diameter CT26 tumors were injected i.p. with 300 ug of CpG or control ODN. Three hours later, spleens were removed and Ly6c^high, Gr-1^intermediate MDSC isolated by MACS sorting. These MDSC were co-cultured with CFSE-labeled CD8⁺ HA-specific Tg T cells plus HA peptide for 3 days, as described above. The proliferation of CD8⁺ T cells was monitored by CFSE dilution. Results represent the mean + SD of 4-7 independent mice/group studied in 3 independent experiments.

* p < 0.01 versus CpG-treated MDSC.

This set of findings led us to examine the effect of CpG ODN treatment on the behavior of MDSC in vivo. Mice bearing CT26 tumors were injected with CpG or control ODN. mMDSC isolated from tumor bearing mice treated with control ODN (or left untreated) significantly inhibited the proliferation of Ag-stimulated CD8⁺ T cells (p. <0.01, Fig 4C). In contrast, MDSC isolated from tumor bearing mice treated with CpG ODN failed to suppress T cell proliferation (p. <0.01). The same pattern was observed using MDSC isolated from CpG-treated mice bearing EL4 tumors, establishing the consistency of this general finding (data not shown).

The ability of mMDSC to suppress T cell activation is linked to their metabolism of l-arginine via arginase-1 and release of inducible nitric oxide synthase (iNOS) (13,14). The amount of iNOS and arginase-1 produced by MDSC was therefore evaluated. MDSC from tumor bearing mice were stimulated to produce iNOS by culture with Ag-activated T cells (Fig 5A). The addition of CpG ODN to these cultures reduced iNOS production by half (p. <0.01). Control ODN had no such effect (Fig 5A). Similarly, CpG but not control ODN significantly reduced the level of arginase-1 present in culture supernatants of purified mMDSC (Fig 5B). Intra-cytoplasmic staining of Ly6C⁺ cells confirmed that mMDSC were the source of this arginase-1 (Fig 5C).

Figure 5.

CpG ODN treatment down-regulates the production of NO and arginase by mMDSC.

mMDSC purified by MACS sorting were cultured with HA-specific Tg CD8⁺ T cells plus 10⁶ mitomycin C treated spleen cells (as APCs) in the presence of HA peptide for 1 day as described in Fig 4. NO levels in culture supernatants were quantified using the Griess reagent. mMDSC purified by MACS were pulsed with 1 uM CpG or control ODN for 36 to 48 hr and analyzed B) for arginase 1 mRNA levels by real time qPCR and C) for expression of Ly6c and arginase I by flow cytometry. Results represent the mean + SD of results from 5 independent MDSC preparations.

* p < 0.01 versus untreated MDSC group

CpG ODN induce monocytic MDSC to differentiate into macrophages

The mechanism by which CpG treatment reduced the ability of MDSC to suppress T cell activation was investigated. mMDSC were isolated from the spleens of tumor bearing mice and cultured in vitro with CpG or control ODN for 48 hr. Cell surface staining of CpG-treated cultures revealed that their expression of F4/80 increased by 3-fold while expression of Ly6c and Gr-1 decreased by 3-fold over this period (p. < 0.01 for both parameters, Fig 6 A,B and data not shown). No such effects were observed when MDSC were cultured with control ODN. These results suggest that Ly6c⁺ MDSC were induced to differentiate into F4/80⁺ macrophages by CpG ODN, a finding consistent with the shared lineage of these two cell types.

Figure 6.

Differentiation of monocytic MDSC.

mMDSC were MACS purified as described in Fig 4. The cells were incubated with 1 uM CpG or control ODN for 48 hr and then analyzed for the expression of Ly6c and F4/80 by flow cytometry. Representative results are shown by dot plot (A) and histogram (B) (shadow; untreated, dotted line; control ODN, solid line; CpG ODN). The mean + SD in MFI from 4 independent experiments is shown in (C). D) MDSC were pulsed with CpG or control ODN for 3 hr, washed extensively, and then transferred to the upper well of a transwell plate. Untreated MDSC were added to the lower well of the same plate. After 48 hr, both cell populatons were analyzed for the expression of Ly6c and F4/80 by flow cytometry. Experiments were repeated three times with similar results.

* p < 0.01 versus untreated group.

We sought to determine whether CpG ODN were directly inducing TLR9-expressing mMDSC to differentiate into macrophages or were stimulating other cell types to produce factors that induced such differentiation. To distinguish between these alternatives, highly purified mMDSC were pulsed with CpG ODN, the ODN was then washed away, and the cells then cultured in transwell plates with unstimulated mMDSC. As seen in Fig 6D, the CpG-pulsed mMDSC in the upper well down-regulated Ly6c and increased F4/80 expression, demonstrating that even short-term exposure to CpG ODN triggered their differentiation into macrophages. In contrast, naive mMDSC in the lower well (exposed to factors secreted by the CpG-pulsed cells in the upper well) did not alter their expression of Ly6c or F4/80. Similarly, culture supernatants from CpG-activated MDSC had no effect on the differentiation of naive mMDSC (data not shown). These findings suggest that CpG ODN directly induce the differentiation of MDSC into macrophages.

The effect of other TLR ligands on mMDSC maturation was also investigated. Initial studies examined TLR mRNA levels in highly purified mMDSC. As seen in Fig 7A, mRNAs encoding TLRs 2, 3, 4, 7 and 8 (in addition to TLR9, Fig 3A) were expressed by these cells. When stimulated with ligands directed against each receptor, MDSC responded by proliferating and secreting TNFα, IL-12, IL-6and/or IL-10 (Fig 7B). Yet only ligands targeting TLR7 and TLR9 (Imiquimod and CpG DNA) induced mMDSC to differentiate into macrophages (Fig 7C).

Figure 7.

Effect of TLR ligands on monocytic MDSC

A) Spleens were removed from CT26 tumor bearing mice. mMDSC (identified as CD11b, Ly6c⁺, Gr-1^intermediate) were FACS sorted to >98% purity. TLR 2, 3, 4, 7 and 8 mRNA levels were determined by RT-PCR in comparison to the RAW 264.7 macrophage cell line (positive control) and EL4 thymoma cell line (negative control). B,C) Purified mMDSC were cultured with 1 uM ODN, 1 ug/ml peptideglycan (PGN), 30 ug/ml Poly(I:C), 1 ug/ml MPL 10ug/ml Imiquimod. B) Culture supernatants were assayed after 24 hr for IL-6, TNFα, IL-12 and IL-10 levels by ELISA (all values in ng/ml). Data represents the mean + SD from 3 independent experiments. C) CD11b⁺cells stimulated for 48 hr were analyzed for the expression of Ly-6c and F4/80 by flow cytometry. Experiments were repeated three times with similar results.

Macrophages derived from CpG-treated monocytic MDSC support tumor elimination

Macrophages are broadly classified into two distinct types that have opposite effects on tumor growth. M1 macrophages contribute to Th1-associated responses and thus improve host resistant to cancer (21). In contrast, M2 macrophages suppress the host’s inflammatory response and enhance angiogenesis, thereby supporting tumor development (22,23). To determine which type of macrophage was generated by CpG ODN treatment, MDSC derived from tumor bearing mice were cultured in vitro with ODN and then co-injected with CT26 tumor cells into BALB/c mice. As seen in Fig 8A, tumor cell growth was significantly reduced when co-administered with macrophages derived from CpG treated MDSC. Similarly, when mMDSC pulsed in vitro with CpG ODN were injected into established CT26 tumors, tumor growth was again significantly slowed (p. <.05, Fig 8B). In contrast, MDSC that were untreated or cultured with control ODN had no significant effect on tumor growth.

Figure 8.

Effect of CpG treated monocytic MDSC on tumor growth.

10⁵ (A) or 3×10⁴ (B) CT26 colon cancer cells were injected s.c. into BALB/c mice. mMDSC were MACS purified from other donor mice as described in Fig 4, pulsed with CpG or control ODN for 3 hr and then washed extensively. 2 × 10⁶ ODN-treated MDSC were injected into the tumor site on days 0 and 2 (A) or into established tumors on days 12 and 13 (B). Data show tumor size in 6 - 7 mice/group from 2 independent experiments. C) ODN-pulsed MDSC were mixed with CT26 target cells at various E:T ratios in vitro. Specific lysis was determined by quantitative measurements of LDH. Results are the mean from 3 independently studied cell populations/group.

* p < 0.01 compared with the untreated treatment group or untreated MDSC.

To investigate the mechanism underlying the protective effect of CpG-activated/differentiated mMDSC, their cytotoxic activity was examined. Tumor-derived mMDSC treated with control ODN manifest little cytotoxic activity against CT26 target cells (Fig 8C). In contrast, mMDSC incubated in vitro for 48 hrs with CpG ODN exhibited significant CTL activity against CT26 targets (Fig 8C).

Discussion

It is well established that CpG ODN treatment can improve the host’s immune response to tumor challenge (5,24). This activity was historically attributed to CpG-mediated activation of DC and macrophage leading to improved NK and CD8 T cell killing. Current results suggest that CpG ODN can also act on mMDSC, further aiding tumor elimination. mMDSC express TLR9 (Fig 3), and treating such cells with CpG DNA rapidly and significantly increased their production of Th1 cytokines (most notably IL-6, IL-12 and TNFα, Fig 3 and 7), while reducing their capacity to inhibit CTL activity. Subsequently, CpG-treated MDSC differentiated into macrophage with tumoricidal activity (Figs 6,8). These findinds demonstrate that CpG ODN reduce the immunosuppressive activity of mMDSC. While the experiments presented herein used MDSC from CT26 tumor bearing mice, the findings were reproduced in studies of MDSC generated in mice challenged with other tumor types (including 4T1 breast cancer, Renca renal tumor, EL4 thymoma and TC-1 tumor lines, data not shown).

The ability of MDSC to down-regulate tumor specific immune responses is well established (11,12). mMDSC are particularly potent, as they suppress the activity of tumor infiltrating CTL (15). In this context, MDSC are considered a major impediment to immunotherapy, and multiple strategies are being pursued to reduce their activity (25). The successful reduction in MDSC function by treatment with PDE5, ATRA, Gemcitabine, 5-Flurouracil, Sunitinib or other agents is frequently associated with an improvement in tumor-specific immunity (26-30). The novel observation that CpG-treated mMDSC lose their ability to inhibit CD8 T cell activation is thus of considerable interest. The mechanism by which mMDSC reduce CTL activity has been attributed to the depletion of L-arginine, an amino acid critical to the maintenance of T cell activation (13,14). mMDSC produce arginase I and the inducible form of NOS, both of which contribute to L-arginine catabolism (13,14). Current findings show that CpG treatment reduces the expression of NO and arginase I by mMDSC, allowing tumor-specific CTL to remain active (Fig 5). Supporting the potential clinical utility of this approach, the number of tumor infiltrating CTL and NK cells rose after CpG treatment, coincident with a significant reduction in the rate of growth of large, established tumors.

An additional effect of CpG ODN treatment was to induce the differentiation of mMDSC into M1-like macrophages (Fig 6). This process was characterized by the loss of Ly6c and Gr-1 (markers expressed by immature monocytes and mMDSC) and the acquisition of F4/80 (Fig 6). Macrophages are typically categorized as being of either the M1 or M2 type. M2 macrophages are generated in a Th2 polarized environment (found in many tumor sites) and are characterized by their production of arginase-1 and IL-10. In contrast, M1 macrophages i) express toll-like receptors, ii) produce pro-inflammatory cytokines (such as IL-12) when stimulated, iii) express diminished levels of arginase 1 and iv) help protect the host from infectious pathogens and tumors (21,24). This constellation of activities was exhibited by CpG treated mMDSC (Figs 3,5). Such activity is consistent with the findings of Colombo et al, who reported that co-administering CpG ODN plus IL-10 receptor Ab re-programed M2-like macrophages and dendritic cells towards an M1-like phenotype, thereby improving anti-tumor immunity (31). The mechanism underlying this MDSC differentiation and the signals that control M1 vs M2 commitment are poorly defined and require further study.

Previous studies demonstrated that intra-tumoral injection of CpG ODN could slow and in some cases reverse the growth of small/nascent tumors (6,7,18,32). In those experiments, CpG treatment led to a significant increase in the fraction of NK cells and CD8⁺ T cells infiltrating the tumor (6,7,18,32). The current work confirms and extends those findings by showing that CpG treatment impacts mMDSC at the tumor site: reducing their number and blocking their suppressive activity. This, in turn, facilitated the expansion and infiltration of cytotoxic CD8 T cells and thus supported immune mediated tumor regression. Of note, our ability to characterize the effect of CpG ODN on mMDSC was facilitated by the study of large established tumors, where the number and immunosuppressive activity of MDSC was enhanced.

The results of human clinical trials in which CpG ODN were administered systemically for cancer therapy were somewhat disappointing (33-35). Preliminary studies from our lab and elsewhere indicate that the route of ODN administration critically impacts outcome, in that local but not systemic delivery of CpG DNA altered the tumor microenvironment (6,7,36). Intra-tumoral injection was key to increasing the number of tumor infiltrating T and NK cells and reducing the number and suppressive activity of mMDSC (Figs 1, 2, 4 and data not shown). In this context, two phase I clinical trials in which CpG ODN were delivered intra-tumorally to treat malignant skin tumors yielded promising results. Hofmann et al used this strategy to induce complete or partial tumor remission in half of all subjects while Molenkamp et al and Brody et al showed that intra-tumoral CpG administration (alone or combined with radiation therapy) could induce systemic tumor regression by improving the generation of tumor-specific CD8 T cells (37-39).

Zoglmeier et al. recently reported that CpG ODN treatment reduced the suppressive activity of granulocytic (CD11b⁺/Ly6g⁺) MDSC (40). In those studies, CpG ODN had no direct on MDSC but instead stimulated pDC to produce type I IFN which indirectly promoted MDSC differentiation (40). This contrasts to our results which show that CpG ODN directly induced mMDSC to differentiate into macrophages. Data provided herein demonstrate that: i) mMDCS express TLR9 (Fig 3), ii) no other cells are required for such differentiation to occur (Fig 6) and iii) CpG ODN, but not the factors produced by CpG-stimulated MDSC, mediate this differentiation (Fig 6). In contrast to the work of Zoglmeier et al, we did not detect the proliferation of granulocytic MDSC following intra-tumoral injection of CpG ODN (Fig 2) and were unable to induce mMDSC to differentiate into macrophages by adding type I or type II IFNs (data not shown).

Of interest, we found that mMDSC expressed multiple TLRs (TLR 2, 3, 4, 7 and 8, Fig 7). When cultured with ligands directed against each of those receptors, MDSC responded by proliferating and secreting cytokines (including TNFα and IL-6). Yet only ligands targeting TLR7 and TLR9 (Imiquimod and CpG DNA) induced mMDSC to differentiate into macrophages (Fig 7 and data not shown). Further study is needed to clarify why TLRs that signal via the same MyD88, NF-kb and MAP kinase pathways (including TLRs 2, 4, 7 and 9) have divergent effects on MDSC differentiation. Hopefully, such studies will clarify the signaling pathway(s) that underlie the changes in MDSC function described in this report, and lead to improved strategies to support immune-mediated anti-tumor therapy.

Glossary

Abbreviations

ODN

phosphorothioate oligodeoxynucleotide

iNOS

inducible nitric oxide synthase

MDSC

myeloid-derived suppressor cells

mMDSC

monocytic MDSC

TLR

toll like receptor

Treg

regulatory T lymphocytes