Multimodal intralesional therapy for reshaping the myeloid compartment of tumors resistant to anti-PD-L1 therapy via IRF8

Abstract

Intralesional therapy is a promising approach for remodeling the immunosuppressive tumor microenvironment (TME) while minimizing systemic toxicities. A combinatorial in situ immunomodulation (ISIM) regimen with intratumoral administration of Fms-like tyrosine kinase 3 ligand (Flt3L), local radiation, and toll-like receptor 3 (TLR3)/CD40 stimulation induces and activates conventional type 1 dendritic cells (cDC1s) in the TME, and elicits de novo adaptive T cell immunity in poorly T cell-inflamed tumors. However, the impact of ISIM on myeloid-derived suppressor cells (MDSCs) that may promote treatment resistance remains unknown. Here, we examined changes in the frequencies and heterogeneity of CD11b⁺Ly6C^loLy6G⁺ polymorphonuclear (PMN)- and CD11b⁺Ly6C^hiLy6G⁻ monocytic (M)-MDSCs in ISIM-treated tumors using mouse models of triple-negative breast cancer (TNBC). We found that ISIM treatment decreased intratumoral PMN-MDSCs, but not M-MDSCs. Although the frequency of M-MDSCs remained unchanged, ISIM caused a substantial reduction of CX3CR1⁺ M-MDSCs that express F4/80. Importantly, these ISIM-induced changes in tumor-residing MDSCs were not observed in Batf3^−/− mice. ISIM upregulated PD-L1 expression in both M-MDSCs and PMN-MDSCs, and synergized with anti-PD-L1 therapy. Furthermore, ISIM increased the expression of interferon regulatory factor-8 (IRF8) in myeloid cells, a known negative regulator of MDSCs, indicating a potential mechanism by which ISIM decreases PMN-MDSC levels. Accordingly, ISIM-mediated reduction of PMN-MDSCs was not observed in mice with conditional deletion of IRF8 in myeloid cells. Altogether, these findings suggest that ISIM holds promise as a multimodal intralesional therapy to alter both lymphoid and myeloid compartments of highly aggressive poorly T cell-inflamed, myeloid-enriched tumors resistant to anti-PD-L1 therapy.

Introduction

Myeloid-derived suppressor cells (MDSCs) are immature, pathologically activated myeloid cells, which exhibit a potent ability to suppress innate and adaptive immune responses (1–6). MDSCs are heterogeneous cell populations (7–9), and divided into two main subgroups with different phenotypic and biological properties: CD11b⁺Ly6C^hiLy6G⁻ monocytic (M)-MDSCs and CD11b⁺Ly6C^loLy6G⁺ polymorphonuclear (PMN)-MDSCs in mice (10). The differentiation of myeloid cell subsets is coordinated by transcriptional regulators, and one integral player is interferon regulatory factor-8 (IRF8). IRF8 promotes monocyte/dendritic cell (DC) differentiation while limiting granulocyte development (11–13), suppresses differentiation of granulocyte progenitors (GPs) to PMN-MDSCs (14), and decreases the frequency of peripheral PMN-MDSCs (15).

MDSCs are one of the several major components of the tumor microenvironment (TME) that promote suppression of T-cell responses (3, 4). Suppression of T-cell responses caused by MDSCs is often associated with treatment resistance, reduction in the efficacy of immunotherapies, and ultimately diminished patient outcomes (7, 16). This has been evident most recently with programmed death 1 (PD-1)/ligand-1 (PD-L1) blockade therapy in which the majority of patients demonstrate primary or acquired resistance, due in part to tumor-driven mechanisms of T-cell exclusion and/or tolerance (17). Therefore, there is a critical need for novel treatment modalities that concurrently increase T-cell infiltration and responsiveness, and decrease the immunosuppressive nature of the TME to improve immunotherapy efficacy.

A novel approach recently reported by our group to overcome PD-1/PD-L1 blockade therapy resistance is in situ immunomodulation (ISIM) (18). ISIM is a multimodal method utilized to convert poorly T cell-inflamed to T-cell inflamed tumors by induction and activation of tumor-residing Batf3-dependent conventional type 1 dendritic cells (cDC1s). Such cDC1s display an enhanced ability to phagocytize dead or dying cells, cross-present exogenous antigens (Ags) onto MHC class I molecules (19), and prime and expand tumor-specific CD8⁺ T cells (20–23). This combinatorial intralesional therapy consists of intratumoral: (i) injection of Fms-like tyrosine kinase 3 ligand (Flt3L) to recruit cDC1s to the TME (24–26); (ii) delivery of local irradiation to induce immunogenic cell death of tumor cells and maturation of cDC1s (27–30); and (iii) administration of dual toll-like receptor 3 (TLR3) and CD40 agonists to activate Ag-loaded cDC1s and facilitate their migration to tumor-draining lymph nodes (18). ISIM elicits cDC1, IFN-γ- and IL-12-dependent de novo adaptive T cell immunity, mediates effective regression of poorly T cell-inflamed tumors, and renders them responsive to anti-PD-L1 therapy (18). ISIM also decreases tumor-associated macrophages (TAMs) within the myeloid compartment of the TME (18). Given that the myeloid phenotypes within the TME are complex, and that MDSCs represent a significant component of that complexity, we tested the hypothesis that ISIM mediates antitumor immunity through modulation of MDSC burden within the TME.

In this study, we sought to investigate changes in the frequencies and phenotypes of MDSCs by ISIM in mouse models of TNBCs. We show that the frequency of PMN-MDSCs decreases in the TME in response to ISIM treatment. Reduction of intratumoral PMN-MDSCs requires all components of ISIM; Flt3L, radiotherapy, and dual stimulation of CD40 and TLR3. While ISIM does not alter the frequency of M-MDSCs, subsets of M-MDSCs expressing the CX3C chemokine receptor 1 (CX3CR1) substantially decrease in the TME. Both M- and PMN-MDSCs express high levels of PD-L1 in ISIM-treated tumors, which respond to anti-PD-L1 therapy. Moreover, ISIM upregulates IRF8 expression, a negative regulator of MDSCs, in tumor-infiltrating myeloid cells, and ISIM-induced reduction of PMN-MDSCs was not observed in mice with genetic deletion of IRF8 in myeloid cells, suggesting an IRF8-mediated remodeling of myeloid cells within the TME. Collectively, these findings show ISIM could reshape the myeloid compartment of the immunosuppressive TME via IRF8, and potentiate antitumor efficacy of PD-L1 blockade therapy.

Methods

Mice

Female C57BL/6 and Batf3^−/− mice were purchased from the Jackson Laboratories. IRF8^fl/fl, Lyz2-Cre, and Lyz2-Cre-IRF8^fl/fl (IRF8-cKO) on a C57BL/6 (B6; H-2^b) background mice were described elsewhere (31). Homozygous IRF8–enhanced GFP (EGFP) mice on a B6 background were generous gift from Dr. H. Morse (National Institutes of Health, Bethesda, MD) and have been described (14, 32). Batf3^−/−, IRF8-EGFP, IRF8^fl/fl, Lyz2-Cre, and IRF8-cKO mice were bred in-house. All mice were age matched (7–10 wk old) at the beginning of each experiment and kept under specific pathogen-free conditions and housed in the Laboratory Animal Resources. All animal studies were conducted in accordance with and approved by the Institutional Animal Care and Use Committee (IACUC) at Roswell Park Comprehensive Cancer Center.

Cell lines

The AT-3 tumor cell line was derived from PyMT-MMTV transgenic mice (33). The 4T1 breast cancer cell line was purchased from the American Type Culture Collection (ATCC), authenticated at ATCC and maintained. AT-3 cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Sigma), 0.5% penicillin/streptomycin (Gibco), 2 mM L-glutamine (Gibco), 1% NEAA (Gibco), and 55 μM 2-mercaptoethanol (Gibco). 4T1 cells were cultured in RPMI (Gibco) supplemented with 10% FBS, 0.5% penicillin/streptomycin, 2 mM L-glutamine, 1% NEAA, and 55 μM 2-mercaptoethanol. These cell lines were authenticated by morphology, phenotype and growth, and routinely screened for Mycoplasma, and were maintained at 37°C in a humidified 5% (4T1) or 7% (AT-3) CO₂ atmosphere.

Tumor inoculation

AT-3 (5 × 10⁵) and 4T1 (1 × 10⁵) tumor cells were surgically implanted under anesthesia with isoflurane into the fourth mammary gland of female mice.

In situ immunomodulation (ISIM)

Tumor-bearing mice were treated with hFlt3L (30 μg/dose; Celldex Therapeutics, Inc.) in 30uL PBS or control PBS intratumorally for 9 consecutive days. Irradiation of mammary tumors was conducted as recently described (18, 34). Briefly, the mice were anesthetized with isoflurane and positioned for exposure to radiation (9 Gy) under a 2 mm thick lead shield containing 1 cm² hole; limiting exposure to the tumors. Irradiation was performed with an orthovoltage X-ray machine (Philips RT250, Philips Medical Systems) at 75 kV using a 1 × 2 cm cone. One day after radiotherapy (RT), mice were treated with injection of agonistic anti-CD40 monoclonal antibody (mAb) (50 μg/dose; clone FGK4.5, BioXcell), high molecular weight poly(I:C) (50 μg/dose; InvivoGen), or combination of these at the peritumoral site subcutaneously (18, 35). Tumor growth was measured 3–5 times a week, and the volumes were calculated by determining the length of short (l) and long (L) diameters (volume =l² ×L/2). Experimental endpoints were reached when tumors exceeded 20 mm in diameter or when mice became moribund and showed signs of lateral recumbency, cachexia, lack of response to noxious stimuli, or observable weight loss.

In vivo antibody (Ab) treatment

For anti PD-L1 therapy, anti-PD-L1 Ab (clone 10F.9G2, BioXcell) was given intraperitoneally (i.p.) every third day from the day RT performed at a dose of 200 μg/mouse. Rat IgG2b Ab (clone LTF-2, BioXcell) was used as a control. To deplete CD8⁺ and CD4⁺ T cells, 200 μg of anti-CD8β (clone Lyt 3.2, BioXcell) and anti-CD4 (clone GK1.5, BioXcell) Abs were administrated i.p. every third day for three times from the day when RT was given, respectively. To neutralize IL-12 and IFN-γ, 1 mg of anti-IL-12p40 (clone C17.8, BioXcell) and anti-IFN-γ (clone R4-6A2, BioXcell) Abs were injected by i.p. at the day when RT was performed, with follow-up doses of 500 μg for 5 consecutive days.

Flow cytometry, cell sorting, and imaging flow cytometry

Single cell suspensions of mouse tumors were prepared for flow cytometric analysis. Cells were blocked with anti-mouse CD16/32 (BioLegend) and surface stained with indicated markers. Live/dead cell discrimination was performed using LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Life Technologies). To evaluate change in the frequency of myeloid cells and MDSCs in the TME by ISIM, we used 12-color flow cytometry gating strategy (18, 22). The following Ab were used; anti-CD45 (clone 30-F11, Invitrogen), anti-CCR2 (R&D systems), anti-Ly6C (clone HK1.4, BioLegend), anti-CD11b (clone M1/70, BD Biosciences), anti-I-A^b (clone AF6-120.1, BD Biosciences), anti-PD-L1 (clone MIH5, BD Biosciences), anti-CD62L (clone MEL-14, BioLegend), anti-CX3CR1 (clone SA011F11, BioLegend), anti-F4/80 (clone BM8, BioLegend), anti-CD24 (clone M1/69, BD Biosciences), and anti-Ly6G (clone 1A8, BioLegend). LIVE/DEAD Fixable Near-IR Dead Cell Stain kit (Thermo Fisher Scientific)-stained cells were excluded from analysis. Samples acquired on LSRII or Fortessa (BD Biosciences) cytometers were analyzed with FlowJo software (Treestar).

Statistical analysis

Statistical analysis was performed using two-tailed Student’s t-test or Mann-Whitney U test for comparisons between two groups, or Kruskal-Wallis with Dunn’s multiple comparisons for comparisons more than 2 groups using GraphPad Prism 9.02 (GraphPad Software). Data are presented as mean ± SEM.

Results

In situ immunomodulation (ISIM) decreases PMN-MDSCs, but not M-MDSCs in mouse models of triple-negative breast cancer (TNBC).

We have recently reported that a multimodal intralesional therapy comprised of intratumoral administration of Flt3L for 9 days followed by RT and dual TLR3/CD40 stimulation (Fig. 1A) could facilitate infiltration of tumor-specific CD8⁺ T cells, and decrease TAMs in preclinical models of poorly T cell-inflamed tumors (18). However, these tumors are also enriched with MDSCs, similar to human breast cancer (36). Since MDSCs represent a potentially significant barrier to antitumor immunity, we first tested whether ISIM alters the frequency of tumor-residing MDSCs. To this end, we used two orthotopic mouse models of TNBC (AT-3 and 4T1) that are poorly T cell-infiltrated tumors and refractory to anti-PD-L1 treatment (18). We evaluated the frequency of M- and PMN-MDSCs in untreated and ISIM-treated tumors by flow cytometric analysis (Supplementary Fig. 1). Tumor-bearing mice were treated with ISIM and tumors were collected 5 days (4T1) or 7 days (AT-3) after completion of the treatment (Fig. 1A). Control mice received intratumoral PBS injections. In agreement with our previous study (18), both AT-3 and 4T1 tumors responded to ISIM with rapid tumor regression, which was followed by acquired resistance to ISIM (Fig. 1B, C). Although we found that ISIM substantially decreased CD11b⁺ myeloid cells, the frequency of CD11b⁺Ly6C^hiLy6G⁻ M-MDSCs among CD45⁺ cells remained unchanged (Fig. 1D, E). In contrast, CD11b⁺Ly6C^loLy6G⁺ PMN-MDSCs in the ISIM-treated tumors markedly decreased among CD45⁺ cells (Fig. 1D, E).

Figure 1.

In situ immunomodulation (ISIM) decreases PMN-MDSCs, but not M-MDSCs in mouse models of triple-negative breast cancer. (A) Experimental set-up. Mice bearing AT-3 or 4T1 tumors were treated with intratumoral (i.t.) Flt3L injection followed by radiotherapy (RT) and in situ TLR3/CD40 stimulation. (B, C) Tumor volume curves (mean) in AT-3 (B) or 4T1 (C) tumor-bearing mice treated with PBS (NT) or ISIM. n = 5–6 (B) and 10 (C) mice per group. (D, E) Representative flow cytometric plots showing CD11b and Ly6G/Ly6C expression of total CD45⁺ cells and CD11b⁺ cells, respectively, in AT-3 (D) and 4T1 (E) tumors. Data panels show frequency of CD11b⁺ cells, M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻), and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) of CD45⁺ cells. n = 6–7 (D) and 5–8 (E) mice per group. NS: not significant, *p < 0.05, **p < 0.01, ****p < 0.0001 by a Mann–Whitney U test. Data shown are representative of three independent experiments. Mean ± SEM.

In situ Flt3L administration, local radiation and TLR3/CD40 stimulation cooperate to decrease intratumoral PMN-MDSCs.

Next, we examined the contribution of Flt3L administration, RT or TLR3/CD40 stimulation to the reduction of tumor-infiltrating PMN-MDSCs. To this end, we treated mice bearing AT-3 tumors with Flt3L and RT alone or TLR3/CD40 stimulation alone, or a combination of all components (ISIM) and evaluated the frequency of MDSCs within the tumor. While we observed modest tumor growth delay after treatment with Flt3L and RT or TLR3/CD40 stimulation alone compared to untreated tumors, effective tumor regression manifested with combining all components of ISIM (Fig. 2A).

Figure 2.

In situ Flt3L administration, radiotherapy (RT) and TLR3/CD40 stimulation corporate to decrease PMN-MDSCs. Mice bearing AT-3 tumors were treated with ISIM as described in Fig. 1A. (A) Tumor volume curves (mean) in AT-3 tumor-bearing mice in different treatment as indicated. (B, C) Representative flow cytometric plots showing expression of CD11b (B) and Ly6G/Ly6C (C) among CD45⁺ cells and CD11b⁺ cells, respectively, in AT-3 tumors. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. Data panels show frequency of CD11b⁺ cells (B), M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻), and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) (C) of CD45⁺ cells. n = 5 mice per group (A-C). NS: not significant, *p < 0.05, **p < 0.01 by a Mann–Whitney U test (A) and Kruskal-Wallis with Dunn’s multiple comparisons (B, C). Data shown are representative of two independent experiments. Mean ± SEM.

Consistent with these observations, we found a significant reduction of CD11b⁺ myeloid cells among the CD45⁺ tumor-infiltrating cells in response to all ISM components, but not in response to Flt3L and RT alone or TLR3/CD40 stimulation alone, compared to the non-treatment controls (Fig. 2B). This reduction of CD11b⁺ myeloid cells was at least partly due to the decrease in PMN-MDSCs but not M-MDSCs in the tumors (Fig. 2C). Notably, we did not observe a significant decrease of PMN-MDSCs, unless mice were treated with all ISIM components (i.e., Flt3L, RT and TLR3/CD40 stimulation; Fig. 2C). These results suggest the requirement of combining Flt3L, RT and TLR3/CD40 agonists to enhance antitumor efficacy, while concurrently reducing PMN-MDSCs in the TME for significant tumor control.

Reduction of PMN-MDSCs requires dual TLR3/CD40 stimulation following Flt3L administration and local irradiation.

We previously demonstrated that dual TLR3/CD40 stimulation was required for the conversion of poorly T-cell-infiltrated to T-cell inflamed tumors (18). However, whether combined TLR3/CD40 stimulation is required to decrease tumor-residing PMN-MDSCs remains unknown. To this end, we investigated the role of dual TLR3/CD40 stimulation after Flt3L injections and RT in remodeling the MDSC load. We treated AT-3 tumor-bearing mice with agonistic anti-CD40 mAb, poly(I:C), or their combination following intratumoral injections of Flt3L and RT. Antitumor efficacy was superior when both TLR3 and CD40 agonists were administered in combination with intratumoral injections of Flt3L and RT (Fig. 3A). Anti-CD40 mAb or poly(I:C) alone did not significantly alter the frequency of CD11b⁺ cells in the tumors when combined with Flt3L and RT. However, dual TLR3/CD40 stimulation synergistically decreased CD11b⁺ cells in the tumors (Fig. 3B). Regarding MDSCs, the frequency of M-MDSCs was not changed in response to treatment with agonistic anti-CD40 mAb, poly(I:C), or their combination (Fig. 3C). In contrast, PMN-MDSCs were significantly decreased by dual TLR3/CD40 stimulation compared to CD40 or TLR3 agonist alone (Fig. 3C). Consistent with these findings in the AT-3 tumor model, the reduction of intratumoral CD11b⁺ cells and PMN-MDSCs required both anti-CD40 mAb and poly(I:C) in mice bearing 4T1 tumors (Supplementary Fig. 2A, B). These results indicate that dual TLR3/CD40 stimulation of Flt3L and RT-induced DCs is essential in tumor control and for the reduction of PMN-MDSCs in tumors.

Figure 3.

Reduction of PMN-MDSCs requires dual TLR3/CD40 stimulation following Flt3L administration and radiotherapy (RT). Mice bearing AT-3 tumors were treated with ISIM as described in Fig. 1A. (A) Tumor volume curves (mean) in AT-3 tumor-bearing mice in different treatment as indicated. (B, C) Representative flow cytometric plots showing expression of CD11b (B) and Ly6G/Ly6C (C) among CD45⁺ cells and CD11b⁺ cells, respectively, in AT-3 tumors. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. Data panels show frequency of CD11b⁺ cells (B), M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻), and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) (C) of CD45⁺ cells. n = 6–8 mice per group (A-C). NS: not significant, *p < 0.05, **p < 0.01, ***p < 0.001 by a Mann–Whitney U test (A) and Kruskal-Wallis with Dunn’s multiple comparisons (B, C). Data shown are representative of two independent experiments. Mean ± SEM.

Non-redundant requirement of Batf3-dependent cells for ISIM-mediated reduction of intratumoral CD11b⁺ myeloid cells and PMN-MDSCs.

ISIM elicits tumor-specific adaptive T cell immunity mediated by Batf3-dependent cells (18). To examine the role of Batf3-dependent cells in ISIM-induced remodeling of the myeloid enriched TME, we treated Batf3^−/− mice bearing orthotopic AT-3 tumors with or without ISIM. Antitumor efficacy of ISIM was abrogated in Batf3^−/− mice (Fig. 4A). Furthermore, we did not observe a reduction of CD11b⁺ myeloid cells or PMN-MDSCs (Fig. 4B), suggesting a key role of Batf3-dependent cells in ISIM-induced remodeling of the tumor-infiltrating myeloid compartment.

Figure 4.

Batf3-dependent cells are required for ISIM-induced reduction of intratumoral CD11b⁺ myeloid cells and PMN-MDSCs. (A) Tumor volume curves (mean) in AT-3 tumor-bearing Batf3^−/− mice treated with ISIM as described in Fig. 1A. (B, C) Representative flow cytometric plots showing expression of CD11b (B) and Ly6G/Ly6C (C) among CD45⁺ cells and CD11b⁺ cells, respectively, in AT-3 tumors. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. Data panels show frequency of CD11b⁺ cells (B), M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻), and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) (C) of CD45⁺ cells. n = 6–8 mice per group (A-C). NS: not significant, **p < 0.01 by a Mann–Whitney U test (A-C). Data shown are representative of two independent experiments. Mean ± SEM.

ISIM upregulates PD-L1 expression in both intratumoral M-MDSCs and PMN-MDSCs mediated by CD4⁺ T cells and IFN-γ.

Our recent study demonstrated that ISIM activates the PD-1/PD-L1 axis in tumors, characterized by the upregulation of PD-1 in CD8⁺ tumor infiltrating lymphocytes (TILs), and PD-L1 in tumor cells and TAMs (18). However, whether PD-L1 is upregulated in MDSCs remains unknown. To this end, we examined PD-L1 expression in MDSCs, and antitumor efficacy of ISIM and anti-PD-L1 therapy in AT-3- or 4T1 tumor-bearing mice. We found that PD-L1 expression was significantly increased in both M-MDSCs and PMN-MDSCs in AT-3 and 4T1 tumors, and that ISIM-treated but not untreated mice responded to anti-PD-L1 blockade therapy (Fig. 5A–D). Therefore, although the frequency of M-MDSCs remains unchanged by ISIM, M-MDSCs might play a role in inhibiting T-cell function in ISIM-treated tumors, which could be alleviated by anti-PD-L1 therapy. To gain mechanistic insights into the upregulation of PD-L1 in MDSCs, we examined the role of CD4⁺ and CD8⁺ T cells and cytokines, IFN-γ and IL-12, which contribute to antitumor efficacy of ISIM (18). Depletion of CD4⁺ T cells but not CD8⁺ T cells correlated with the upregulation of PD-L1 in MDSCs (Fig. 6A). IFN-γ was also needed for this effect, while there was a moderate correlation of IL-12 and increased PD-L1 expression in MDSCs (Fig. 6B).

Figure 5.

ISIM increases PD-L1 expression in both M-MDSCs and PMN-MDSCs in the tumor. (A-D) Mice bearing AT-3 (A, B) or 4T1 (C, D) tumors were treated with ISIM as described in Fig. 1A. Anti-PD-L1 Ab or isotype Ab was given every third day from the day RT performed. (A, C) Representative flow cytometric plots showing PD-L1 expression and median fluorescence intensity (MFI) of PD-L1 in M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻) and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) of CD45⁺ cells in AT-3 (A) and 4T1 (C) tumors. n = 5 (A) and 5–8 (C) mice per group. AT-3 and 4T1 tumors were harvested 7 and 5 days after vaccination with TLR3/CD40 agonists, respectively. Numbers denote MFI of PD-L1 expression. (B, D) Individual tumor volume curves of AT-3 (B) and 4T1 (D) tumor-bearing mice in different treatment as indicated. n = 6–7 (B) and 4–5 (D) mice per group. NS: not significant, *p < 0.05, **p < 0.01, ***p < 0.001 by a Mann–Whitney U test. Data shown are representative of three independent experiments. Mean ± SEM.

Figure 6.

ISIM-induced upregulation of PD-L1 in MDSCs depends on CD4⁺ T cells and IFN-γ. Mice bearing AT-3 tumors were treated with ISIM as described in Fig. 1A. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. Mice were depleted for CD4⁺ T cells or CD8⁺ T cells (A) and IFN-γ or IL-12 (B). Representative flow cytometric plots showing PD-L1 expression and median fluorescence intensity (MFI) of PD-L1 in M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻) and PMN-MDSCs (CD11b⁺Ly6C^loLy6G⁺) of CD45⁺ cells in AT-3 tumors. Numbers denote MFI of PD-L1 expression. n = 5–7 mice (A) and 5 mice (B) per group. *p < 0.05, **p < 0.01 by Kruskal-Wallis with Dunn’s multiple comparisons. Data shown are representative of two independent experiments. Mean ± SEM.

ISIM decreases frequency of intratumoral CX3CR1⁺ M-MDSCs.

A large body of evidence supports the notion for phenotypic and functional heterogeneity in monocytes in tumor-bearing hosts, both systemically and in the tumor (5–9, 37–39). Although the frequency of CD11b⁺Ly6C^hiLy6G⁻ cells was similar in untreated and ISIM-treated mice, it is possible that ISIM might alter the heterogeneity of M-MDSCs in the TME. First, we evaluated CX3CR1 expression, which is a monocyte marker (40), and could be upregulated on M-MDSCs as they differentiate towards TAMs (10). Within the intratumoral CD11b⁺Ly6C^hiLy6G⁻ population of both untreated AT-3- and 4T1 tumor-bearing mice, two subsets were observed that were stratified by the expression of CX3CR1 (Fig. 7A). We observed that CX3CR1⁺ M-MDSCs contained more F4/80⁺ and less CD62L⁺ populations compared with CX3CR1⁻ M-MDSCs (Fig. 7B) although the level of F4/80 in M-MDSCs was lower than in Ly6C⁻class II⁺CD24⁻F4/80⁺ TAMs (Fig. 7C). ISIM treatment substantially decreased the frequency of CX3CR1⁺ M-MDSCs, a precursor of TAMs (41), in AT-3 and 4T1 tumors (Fig. 7D). Notably, ISIM-induced reduction of CX3CR1⁺ M-MDSCs was not observed in Batf3^−/− mice (Fig. 7E).

Figure 7.

ISIM decreases frequency of the CX3CR1⁺ M-MDSCs. C57BL/6 (A-D) and Batf3^−/− (E) mice bearing AT-3 (A-E) and 4T1 (B, D) tumors were treated with ISIM as described in Fig. 1A. AT-3 and 4T1 tumors were harvested 7 and 5 days after vaccination with TLR3/CD40 agonists, respectively. (A) Expression of CD62L and F4/80 in CX3CR1⁻ (lower) and CX3CR1⁺ (upper) M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻) in untreated AT-3 tumors. (B) Data panels show frequency of CD62L (left) and F4/80 (right) in CX3CR1⁻ and CX3CR1⁺ M-MDSC subsets in AT-3 and 4T1 tumors. n = 10 mice per group for both AT-3 and 4T1 tumors. (C) Expression of CX3CR1 and F4/80 in M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻), and macrophages (Ly6C⁻class II⁺CD24⁻F4/80⁺) in untreated AT-3 tumors. M-MDSCs (gray) and macrophages (black) are overlaid. (D, E) Representative flow cytometric plots showing CX3CR1 expression of M-MDSCs in AT-3 and 4T1 tumors in C57BL/6 mice (D) and in AT-3 tumors in Batf3^−/− mice (E). Data panels show frequency of CX3CR1⁺ cells of M-MDSCs. n = 8–10 (D: AT-3), 5 (D: 4T1), and 6–8 (E) mice per group. NS: not significant, *p < 0.05, **p < 0.01, ****p < 0.0001 by a two-tailed paired t-test (B) and a Mann–Whitney U test (D, E). Data shown are representative of three (A-D) and two (E) independent experiments. Mean ± SEM.

ISIM upregulates IRF8 in tumor-infiltrating myeloid cells.

Next, we sought to investigate the mechanisms underlying the reduction of PMN-MDSCs by ISIM treatment. IRF8 is a transcription factor expressed in monocytes/macrophages and DCs, and certain early myeloid progenitors. IRF8 is essential for the development of monocytes and DCs, but it suppresses the generation of neutrophils (11). This is evident in Irf8^−/− mice, which develop a myeloproliferative disorder characterized by a profound accumulation of both immature and mature granulocytes at the expense of monocytes and DCs (12, 13). We previously reported that IRF8 inversely controls PMN-MDSC burden in mouse tumor models, and that transgenic overexpression of IRF8 in myeloid cells decreased the frequency of PMN-MDSCs (15). These findings led us to hypothesize that ISIM treatment decreases the frequency of PMN-MDSCs by enhancing IRF8 expression. To begin to test this hypothesis, we utilized transgenic knock-in mouse expressing an IRF8-EGFP fusion protein at the endogenous IRF8 locus (Irf8^Irf8Gfp/WT; henceforth referred to as IRF8-EGFP) (32). IRF8-EGFP mice bearing AT-3 tumors were treated with ISIM, and IRF8 expression of tumor-infiltrating myeloid cells was evaluated (Supplementary Fig. 3A, B). Immunophenotyping revealed increased IRF8 expression in macrophages and DCs, but not M-MDSCs in ISIM-treated tumors (Fig. 8). Although the majority of PMN-MDSCs had low levels of IRF8 expression, ISIM increased the frequency of IRF8⁺ PMN-MDSCs.

Figure 8.

ISIM upregulates interferon regulatory factor-8 (IRF8) in tumor-infiltrating myeloid cells. IRF8–enhanced GFP (EGFP) mice bearing AT-3 tumors were treated with ISIM as described in Fig. 1A. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. Representative flow cytometric plots showing IRF8 expression (EGFP) of macrophages (Ly6C⁻class II⁺CD24⁻F4/80⁺), dendritic cells (Ly6C⁻class II⁺CD24⁺F4/80⁻), CX3CR1⁻ M-MDSCs, CX3CR1⁺ M-MDSCs, and PMN-MDSCs in AT-3 tumors of IRF8-EGFP mice. Data panel shows % IRF8 (EGFP)⁺ (left) and median fluorescence intensity (MFI) (right) of IRF8 (EGFP) in myeloid cell subsets in AT-3 tumors. n = 6 mice per group. NS: not significant, *p < 0.05, **p < 0.01 by a Mann–Whitney U test. Data shown are representative of two independent experiments. Mean ± SEM.

IRF8 plays a critical role in ISIM-mediated remodeling of the myeloid compartment of the TME.

Lastly, we employed a genetic loss-of-function approach using Lyz2-Cre-IRF8^fl/fl (IRF8-cKO) mice (31) to determine a potential role of IRF8 in the reduction of PMN-MDSCs within the TME. Lyz2 encodes lysozyme M, which is exclusively expressed in myelomonocytic cells including monocytes, macrophages and granulocytes in mice (42). Because such IRF8-cKO mice do not exhibit developmental defects in myeloid progenitors of the bone marrow, this allows us to study IRF8 function in differentiated myelomonocytic cells (31). IRF8^fl/fl mice (WT) were used as a control. We treated WT and IRF8-cKO mice bearing AT-3 tumors with or without ISIM. We found an increase in tumor burden in IRF8-cKO mice compared with WT mice (Fig. 9A). Additionally, IRF8-cKO mice responded poorly to ISIM with negligible tumor regression compared with WT mice (Fig. 9A), suggesting the significance of IRF8 signaling in myeloid cells in the antitumor efficacy of ISIM.

Figure 9.

IRF8 plays a critical role in ISIM-mediated remodeling of the myeloid compartment of the TME.

IRF8^fl/fl (WT) or Lyz2-Cre-IRF8^fl/fl (IRF8-cKO) mice bearing AT-3 tumors were treated with ISIM as described in Fig. 1A. Tumors were harvested 7 days after vaccination with TLR3/CD40 agonists. (A) Tumor volume curves (mean), tumor weight and waterfall plots in AT-3 tumor-bearing mice in different treatment as indicated. Waterfall plots show maximal change of tumor volume at the day when TRL3/CD40 agonists were given. (B) Representative histograms and flow cytometric plots showing CX3CR1, CCR2, CD62L and F4/80 in M-MDSCs (CD11b⁺Ly6C^hiLy6G⁻) in untreated AT-3 tumors. Data panel shows % CX3CR1⁺, CCR2⁺, CD62L⁺ or F4/80⁺ among M-MDSCs. (C, D) Representative flow cytometric plots and frequency of class II⁺ cells (Ly6C⁻class II⁺), dendritic cells (Ly6C⁻class II⁺CD24⁺F4/80⁻), macrophages (Ly6C⁻class II⁺CD24⁻F4/80⁺), and PMN-MDSCs of CD45⁺ cells in AT-3 tumors. n = 5–6 mice per group (A-D). NS: not significant, *p < 0.05, ***p < 0.001, ****p < 0.0001 by a Mann–Whitney U test (B) and Kruskal-Wallis with Dunn’s multiple comparisons (A, C, D). Data from an experiment are shown. Mean ± SEM.

Untreated AT-3 tumors from IRF8-cKO mice were further characterized by increased CX3CR1⁺ and F4/80⁺ M-MDSCs and had decreased frequency of M-MDSCs expressing CD62L and CCR2, a marker of mature monocytes (43) compared with tumors in WT mice (Fig. 9B). In line with this finding, TAMs increased while tumor-residing MHC class II⁺ cells and DCs decreased in IRF8-cKO mice (Fig. 9C). Furthermore, ISIM-mediated reduction of TAMs (18) was not observed in IRF8-cKO mice. Importantly, untreated AT-3 tumors in IRF8-cKO mice had a higher frequency of PMN-MDSCs than tumors in WT mice, which was not mitigated by ISIM treatment (Fig. 9D). In contrast, ISIM treatment significantly reduced PMN-MDSCs in tumor-bearing WT mice. Taken together, these data suggest that IRF8 plays a critical role in ISIM-induced remodeling of the myeloid-enriched TME.

Discussion

Myeloid cells, which include monocytes/macrophages, DCs, neutrophils, and monocytes abundantly infiltrate the TME and play important roles, both positive and negative, in regulating tumor outcome (5–7). As barriers to host defense against cancer, these populations contribute to tumor progression by driving tumor cell proliferation, angiogenesis, and metastasis, and by mediating immune suppression and treatment resistance (5–7). Using a novel multimodal intralesional therapy, our study identifies remodeling of the myeloid landscape of the TME, impacting not only DCs and TAMs but also MDSCs. We demonstrate that: (i) in situ administration of Flt3L, local irradiation, poly(I:C) and anti-CD40 mAb cooperatively decreases PMN-MDSCs, but not M-MDSCs within the TME mediated by Batf3-dependent cells; (ii) although the frequency of CD11b⁺Ly6C^hiLy6G⁻ M-MDSCs remains unchanged in the TME, ISIM preferentially decreases the CX3CR1⁺ subset of M-MDSCs, thereby, potentially decreasing the development of immunosuppressive TAMs; (iii) both M-MDSCs and PMN-MDSCs upregulate PD-L1 expression in ISIM-treated tumors, which respond to anti-PD-L1 therapy; and (iv) upregulation of IRF8 in the myeloid compartment of ISIM-treated tumors correlates with decreased PMN-MDSC levels.

Previous studies have shown that increased CX3CR1 expression occurs in progressively growing tumors in preclinical models and humans (44, 45). Furthermore, upregulation of CX3CR1 in M-MDSCs by cyclin-dependent kinase (CDK) inhibitors causes an accumulation of M-MDSCs in tumors and subsequent acceleration of tumor growth in allograft mouse models (46). In line with these findings, we identified an abundant presence of CX3CR1⁺ M-MDSCs in mouse models of TNBC, and further demonstrated that ISIM treatment decreased the frequency of CX3CR1⁺ M-MDSCs in the TME. Future studies, however, are needed to clarify the monocytic populations that increased after ISIM since total numbers of CD11b⁺Ly6C^hiLy6G⁻ cells remained unchanged. Moreover, although we observed two subsets of M-MDSCs characterized by the expression of CX3CR1, CD62L and F4/80 in two separate mouse models of TNBC, their function and role in the TME remain to be determined. Evidence suggests that substantial heterogeneity exists in circulating and tumor-residing monocytes (5–9, 37–39), and that M-MDSCs and TAMs share many characteristics while they are regarded as separate cellular entities (39). Although mechanisms underlying the differentiation of M-MDSCs to TAMs were not formally investigated in this current study, it is conceivable that the reduction of CX3CR1⁺ M-MDSCs, which contained more F4/80⁺ subsets, might be associated with decreased TAMs by ISIM (18). Nevertheless, additional work is warranted to interrogate the associations between MDSCs and TAMs and their functional relevance in poorly T cell-inflamed tumors.

Our results showing a substantial inverse association between the reduction of PMN-MDSCs with increased IRF8 expression align with a previous study reporting significantly more neutrophils in Irf8^−/− mice compared with the wild-type controls (12), and with our previous work demonstrating that IRF8 negatively regulates the expansion of PMN-MDSCs in tumor-bearing mice (15). We further demonstrated that selective deletion of IRF8 in myelomonocytic cells increased PMN-MDSC burden, and abrogated ISIM-induced reduction of PMN-MDSCs in the TME. ISIM increased IRF8⁺ subsets in DCs, TAMs, and PMN-MDSCs, but not in M-MDSCs. The reasons for this remain unclear but might be related to the differentiation status or heterogeneity of different subpopulations of M-MDSCs in the TME before and after ISIM. IRF8 expression decreases when monocyte-DC progenitors (MDPs) and common monocyte progenitors (cMoPs) differentiate into monocytes (13). Heterogeneous M-MDSCs might contain cMoPs expressing high levels of IRF8 in untreated tumors, which could differentiate with decreased IRF8 expression.

A previous study revealed that IFN-γ could drive the phenotypic differentiation of immature neutrophils progenitors, such as PMN-MDSCs, into APC-like hybrid neutrophils (47). Our depletion study showed an important role of IFN-γ for upregulating PD-L1 in MDSCs in line with our recent study (18). Therefore, although reductions of PMN-MDSCs and increased IRF8 expression in myeloid cells in ISIM-treated tumors are in line with our previous observations that IRF8 inversely controls PMN-MDSC burden in preclinical models (14), we cannot rule out the possibility that a reduction of PMN-MDSCs in the TME might be associated with the differentiation of PMN-MDSCs. Further studies are required to address the states and potential plasticity of MDSCs in the TME.

Our study has some limitations. More work is warranted to determine the impact of changes in MDSC frequency and function in response to ISIM. For example, it remains unclear the degree to which ISIM-induced reductions of PMN-MDSCs and CX3CR1⁺ M-MDSCs contributes to therapeutic efficacy, although our previous studies and those of other laboratories have demonstrated that MDSCs in these tumor models are immunosuppressive (2, 3, 14, 15, 31, 41). CX3CR1 is expressed not only on M-MDSCs and TAMs but also on T and NK cells (10, 23, 35, 48–50); therefore, determining the role of CX3CR1⁺ M-MDSCs by selective depletion of this subset might be difficult. With the advent of high-dimensional single-cell profiling and lineage tracing technologies, future work is needed to elucidate the complexities of the tumor-infiltrating myeloid landscape, and to trace myeloid suppressive cell lineages across tissue and circulation. Additionally, the implantable tumor models of TNBC might not fully recapitulate the complexity of the host-tumor dynamic of spontaneous tumorigenesis. As such, the role of MDSCs after ISIM treatment remains to be elucidated in those cancer settings.

Our study is relevant to poorly T cell-inflamed, myeloid-enriched highly aggressive tumors such as TNBC. TNBC is an aggressive form of invasive cancer, affecting 12 to 17% of women with breast cancer, and remains a significant challenge in cancer care (51). As a group, patients with TNBC have relatively poor outcomes and because they lack expression of estrogen receptor (ER), progesterone receptor (PR), and HER2/neu (HER2) and, they are resistant to therapies that target those surface receptors which otherwise have been successful in other breast cancer subtypes. Chemotherapy and chemo-immunotherapy remain the standard of care, yet patients still suffer from a high rate of relapse and metastatic recurrence. The TME and its immunosuppressive milieu is thought to contribute to treatment resistance. The presence of TILs correlates with response to anti-PD-L1 therapy (52), and increased pathologic complete response rates to chemotherapy, and improved survival in TNBC (53, 54). Therefore, ISIM may represent a novel strategy that allows remodeling of both lymphoid and myeloid compartments of the TME to potentially further increase clinical benefit to at least some patients with TNBC.

In conclusion, our data presented here identify changes in the frequency and phenotype of MDSCs by multimodal intralesional therapy in preclinical mouse models of TNBC. These insights support the concept of devising therapies or therapeutic combinations that target both innate and adaptive immune elements to overcome poorly T cell-inflamed, myeloid-enriched highly aggressive tumors.

Key points.

• In situ immunomodulation (ISIM) reduces PMN-MDSCs and CX3CR1⁺ M-MDSCs in the tumor

• ISIM upregulates PD-L1 in intratumoral MDSCs and synergizes with anti-PD-L1 therapy

• Deletion of IRF8 in myeloid cells increases PMN-MDSCs and causes poor outcome