Selective targeting of different populations of myeloid-derived suppressor cells by histone deacetylase inhibitors

Abstract

Myeloid-derived suppressor cells (MDSCs) are widely implicated in negative regulation of immune responses in cancer. Inhibition of class I histone deacetylases (HDAC) with entinostat has anti-MDSC activity. However, as single agent, it did not delay tumor growth in EL4 and LLC tumor models. Here, we found that entinostat reduced immune suppressive activity of only one type of MDSC—polymorphonuclear, PMN-MDSC, whereas it had no effect on monocytic M-MDSC or macrophages. M-MDSC had high amount of class II HDAC—HDAC6, which was further increased after the treatment of mice with entinostat. Inhibition of HDAC6 with ricolinostat reduced suppressive activity of M-MDSC, but did not affect PMN-MDSC or delayed tumor growth. However, combination of entinostat and ricolinostat abrogated suppressive activity of both populations of MDSC and substantially delayed tumor progression. Thus, inactivation of MDSC required targeting of both major subsets of these cells via inhibitors of class I and class II HDAC.

Electronic supplementary material

The online version of this article (10.1007/s00262-020-02588-7) contains supplementary material, which is available to authorized users.

Introduction

Myeloid-derived suppressor cells (MDSCs) are pathologically activated neutrophils and monocytes with potent immune suppressive activity [1]. These cells accumulate in almost all types of cancer as well as in most chronic infections, chronic inflammatory conditions, and many autoimmune diseases. MDSCs are implicated in negative regulation of immune responses by suppressing functions of T cells, NK cells, B cells. Accumulation of MDSC was directly linked with negative clinical outcome in cancer, as well as chronic infections and sepsis [2–5]. MDSCs are comprised of two large group of cells: pathologically activated mostly immature neutrophils—PMN-MDSC and pathologically activated monocytes—M-MDSC. Although MDSC and their classical counterparts neutrophils (PMN) and monocytes (MON) share many phenotypic and morphological characteristics, they have distinct transcriptomic and proteomic profiles, metabolism, biochemical features as well as functions [1, 6–9]. Because of MDSC involvement in regulation of immune responses in cancer, these cells are attractive targets for therapy. One of the approaches to target MDSC is epigenetic modulatory drugs, specifically inhibitors of histone deacetylase (HDAC).

Entinostat is an oral, class I-specific HDAC inhibitor, shown to disrupt the dynamic interactions between the tumor microenvironment and host immune surveillance [10, 11]. HDAC inhibitors can increase immunogenicity of tumor cells by activating expression of tumor antigen, antigen presentation, and co-stimulation molecules in tumor cells [11, 12]. Entinostat also inhibits the function of immunosuppressive T regulatory (Treg) cells through acetylation of the STAT3 transcription factor [10]. We and others have recently demonstrated that entinostat can decrease MDSC activity in different tumor models and enhance the effect of check point inhibitor therapy [13, 14]. Decrease in circulating MDSC was found in breast cancer patients treated with entinostat and aromatase inhibitor [15]. However, treatment of immune competent mouse tumor models with entinostat as single agent showed no antitumor activity [13, 14]. First results of clinical trials of combination of entinostat with check-point inhibitors showed only modest antitumor activity [16, 17]. This raised question whether the effect of HDAC inhibition on MDSC was sufficiently strong. In this study, we discovered that entinostat affected only one group of MDSC—PMN-MDSC and did not block suppressive function of M-MDSC and tumor-associated macrophage (TAM). We found that the mechanism of M-MDSC escape from entinostat was mediated by a high level of HDAC6 in these cells. Combination of HDAC1 and HDAC6 inhibitors provided for strong antitumor effect in several tumor models.

Material and methods

Mice All procedures were performed according to NIH Guide for the Care and Use of Laboratory Animal guidelines and were approved by the Institutional Animal Care and Use Committee (IACUC) at the Wistar Institute. Female six-week-old C57BL/6 mice were purchased from Charles River Labs. B6.Cg-Thy1^a/Cy Tg(TcraTcrb)8Rest/J (PMEL) mice were obtained from the Jackson Laboratory. All mice were maintained in a temperature-controlled room and pathogen-free conditions with a 12/12 h light/dark schedule. Food was provided ad libitum.

Cell lines EL4 lymphoma and LLC lung carcinoma cell lines were obtained from ATCC and cultured in DMEM (Corning Incorporated) supplemented with 10% FBS (Atlanta Biologicals, Inc.) and 1% antibiotics (Penicillin–Streptomycin, Thermo Fisher Scientific Inc.). The cells were harvested using 0.25% trypsin (Thermo Fisher Scientific Inc.), suspended in DPBS (Corning) as 200 μL containing 5 × 10⁵ cells, and injected s.c. into the mice. After tumors were established, the mice were randomized into groups and used for the studies. The tumor diameters (width and length) were measured using digital calipers and used for the calculation of tumor area (width × length).

Compounds and antibodies Entinostat (MS275) and anti-CSF1R mIgG2 antibody (SNDX-ms6352) were obtained from Syndax Pharmaceuticals, Inc. Ricolinostat (ACY-1215) was purchased from Selleck Chemicals. For the in vivo studies, entinostat was dissolved in DMSO and then diluted with ddH₂O. Ricolinostat was dissolved in DMSO, and then diluted with PEG300 (Sigma-Aldrich) and ddH₂O to achieve 2% DMSO/30%DMSO in H₂O. Entinostat was orally administered to the mice at 10 mg/kg every day, and 50 mg/kg of ricolinostat was administered daily i.p. Anti-CSF1R antibody was diluted with PBS and injected i.p. to the mice (30 mg/kg) twice a week.

Cell isolation Single-cell suspensions were prepared from spleens and bone marrow. Tumor tissues were cut into small pieces and digested with the mouse tumor dissociation kit (Miltenyi Biotec). Red blood cells in the cell suspensions were lysed using ammonium chloride lysis buffer.

Flow cytometry Monoclonal antibodies specific to the mouse cell surface markers CD45, CD11b, CD11c, Ly6G, Ly6C, F4/80, I-Ab, and CD16/32 were purchased from BD bioscience. Flow cytometry data were acquired using a BD LSR II flow cytometer and analyzed using FlowJo software (Tree Star).

T cell suppression assay PMN-MDSCs (Ly6G⁺) were purified from spleens and tumors by subsequently reaction with biotinylated Ly6G antibody and streptavidin microbeads (Miltenyi) and pass through the MACS column according to the manufacturer’s recommendation (Miltenyi). The purity of cell populations was > 95%. M-MDSC (CD45 + , CD11b + , Ly6G-, Ly6C +) cells and TAM (CD45 + , CD11b + , F4/80 + , Ly6C-, Ly6G-) were isolated from spleens (M-MDSC) and tumors by cell sorting on FACSAria cell sorter (BD Biosciences). CD8⁺ T cells from PMEL mice that recognize gp100-derived peptide were used as responders. Spleen cells from PMEL mice were suspended in complete RPMI media and plated into 96-well U-bottom plates at 10⁵ cells/well. PMN-MDSC, M-MDSC, or TAM was added to the wells at 12,500 to 10⁵ cells/well (1:8–1:1 ratios). Murine gp100 peptide [25-33] EGSRNQDWL (AnaSpec, Inc.) was added into the wells at the final concentration of 0.1 μg/mL. After 48 h of culture, cells were pulsed with ³H-thymidine (1 μCi/well; GE healthcare) for 16 h. ³H-thymidine uptake was counted using a liquid scintillation counter as counts per minute (cpm) and calculated the percentage of proliferation to the positive control (responder cells and peptide without MDSC or TAM).

Quantitative RT-PCR RNA was extracted with total RNA extraction kit (Zymo research). cDNA was synthesized using cDNA reverse transcriptase kit (Applied Biosystems), and then real-time PCR was performed in triplicate for each sample with 10 μl SYBR Master Mixture (Applied Biosystems) and the following primers.

Arg1, 5′-GCTGTCTTCCCAAGAGTTGGG-3′ and 5′-ATGGAAGAGACCTTCAGCTAC-3′; iNOS, 5′-AACGGAGAACGTTGGATTTG-3′ and 5′-CAGCACAAGGGGTTTTCTTC-3′; COX2, 5′-CCAGCACTTCACCCATCAGTT-3′ and 5′-ACCCAGGTCCTCGCTTATGA-3′; PTGES, 5′-GCACACTGCTGGTCATCAAG-3′ and 5′-ACGTTTCAGCGCATCCTC-3′; HDAC1, 5′- CCAATGCTGAGGAGATGACCA-3′ and 5′-GCTTCACAGCACTTGCGAC-3′; HDAC2, 5′- AAGTGTGCTACTACTATGATGGTGA-3′ and 5′-TGAGGCTTCATGGGATGACC-3′; HDAC3, 5′- CACTATGGAGCTGGACACCC-3′ and 5′-TCAGAATGGAAGCGGCACAT-3′; HDAC8, 5′- CCTCAACTACATCAAAGGGAATCTG-3′ and 5′-ACAAACCGCTTGCATCAACA-3′; HDAC6, 5′- ATGGACGGGTATTGCATGTT-3′ and 5′-GCGGTGGATGGAGAAATAGA-3′; HDAC10, 5′- TCCATCCGAGTACCTTCCAC-3′ and 5′-GGCTGCTATGGCCACACTAT-3′; TGFb, 5′- AGCTGCGCTTGCAGAGATTA-3′ and 5′-ATTCCGTCTCCTTGGTTCAGC-3′; Actb, 5′-ATGGAGGGGAATACAGCCC-3′ and 5′-TTCTTTGCAGCTCCTTCGTT-3’.

Expressions of the different genes were normalized to β-actin by the 2^−ΔCt method.

Western blotting Total proteins were extracted from spleen PMN- and M-MDSC sorted from spleens of EL4 tumor bearing mice. RIPA buffer supplemented with 1% of protease inhibitor cocktail (P8340-5ML; Sigma-Aldrich) was used. Extracted proteins were resolved by electrophoresis using 10% SDS-PAGE gels (Invitrogen). Ten micrograms of protein was loaded per lane, and proteins were electroblotted onto PVDF membrane (Immobilon-P, Millipore). The membranes were blocked with 5% non-fat dry milk in TBS plus 0.1% Tween20 (TBS-T) for 1 h at room temperature and then incubated with the primary antibodies in 5% non-fat dry milk in TBS-T overnight at 4 °C. The primary antibodies used (dilution 1:1000) were anti-HDAC1 (5356S), anti-HDAC6 (7612S), and anti-β actin (4970S) from Cell Signaling Technology, Inc., and anti-HDAC2 (32,117), anti-HDAC3 (32,369), anti-HDAC8 (187,139) and anti-HDAC10 (108,934) from Abcam. After washing, the membranes were incubated with HRP-linked secondary antibodies specific to rabbit and mouse antibodies (Cell Signaling Technology, Inc.) for 1 h at room temperature. Detection was visualized by Pierce ECL Western Blotting Substrate (Thermo scientific).

HDAC activity Nuclear protein was extracted from splenic PMN- and M-MDSC using CelLytic NuCLEAR Extraction Kit (Sigma-Aldrich), and HDAC activities were measured by HDAC Activity Assay Kit (Abcam).

Combination therapy EL4 tumor-bearing mice were randomized into 4 groups (n = 4—10) on Day 6 or 7. LLC tumor-bearing mice were randomized into 4 groups (n = 5) on Day 5. Entinostat solution (10 mg/kg) or vehicle (1% DMSO) was orally treated to the mice every day. Anti-CSF1R antibody or isotype control antibody was administered i.p. on days 6, 10, 13, 17, 20 and 24. Ricolinostat solution or vehicle (2% DMSO and 30%DMSO in H₂O) was administered i.p. every day.

Results and discussion

Two subcutaneous tumor models (EL4 lymphoma and LLC lung carcinoma) were used in this study. Treatment with entinostat (10 mg/kg) started on day 7 (EL4 model) or on day 11 (LLC model) once tumors became palpable and daily administration of drug continued for 10 days. Treatment with entinostat alone did not affect tumor growth (Fig. 1a, c), but caused significant increase in the populations of PMN-MDSC, M-MDSC, and macrophages in spleens. In tumors, however, only PMN-MDSC were increased, whereas populations of M-MDSC and TAM were not changed (Fig. 1b, d). Remarkably, entinostat had different effect on suppressive activity of PMN-MDSC, M-MDSC, and TAM. It abrogated suppressive activity of PMN-MDSC in spleens and tumors but had no effect on suppressive activity of M-MDSC and TAM (Fig. 2a-c). In PMN-MDSC entinostat significantly decreased expression of Arg1, Nos2¸and Ptgs2 known to be involved in suppressive function of MDSC [6]. In contrast, no effect on the expression of these genes was observed in M-MDSC (Fig. 2d).

Fig. 1.

Entinostat did not show anti-tumor activity but affected the presence of myeloid cell. a, b EL4 tumor-bearing mice were treated with vehicle or entinostat (10 mg/kg) daily on days 7—17. a Kinetics of tumor progression (n = 10). b The presence of indicated cell populations in spleens and tumors measured on day 17 by flow cytometry. N = 5. P values were calculated in two-tailed Student’s t-test. Results are shown as mean ± SD. c, d, LLC tumor-bearing mice were treated with vehicle or entinostat daily on days 11—28. c Kinetics of tumor progression (n = 5). d The presence of indicated cell populations in spleens and tumors measured on day 28 by flow cytometry. N = 11 for vehicle-treated group, n = 13 for entinostat-treated group. P values were calculated in two-tailed Student’s t-test. Results are shown as mean ± SD. PMN-MDSC: CD45⁺CD11b⁺Ly6G⁺Ly6C^low, M-MDSC: CD45⁺CD11b⁺Ly6G⁻Ly6C^high, macrophage: CD45⁺Ly6C⁻F4/80⁺Ly6G⁻

Fig. 2.

Entinostat impairs the immunosuppressive capacity of PMN-MDSC but not of M-MDSC. PMN-MDSC or M-MDSC was purified from spleens and tumors of EL4 tumor-bearing mice (a, b) or LLC tumor-bearing mice (c), which were treated with vehicle or entinostat (10 mg/kg) for 2 weeks. MDSCs were cocultured with PMEL splenocytes and gp100 peptide for 2 days. T cell proliferation was measured in triplicate by ³H-thymidine uptake. T cell proliferation in the absence of MDSC was set as 100%, and percent of change in each experiment was calculated. N = 5–11 in different experiment (shown on each panel). P values were calculated in two-tailed Student’s t-test. Results are shown as mean ± SD. *p < 0.05. d RNA was extracted from MDSC from spleens of EL4 tumor-bearing mice, and expression of indicated mRNA was analyzed by qRT-PCR (n = 4–5). P values were calculated in two-tailed Student’s t-test. Results are shown as mean ± SEM

We hypothesized that PMN-MDSC and M-MDSC may have different expression of class I HDACs and this could explain different sensitivity of these cells to entinostat. Indeed, expression of HDAC1, HDAC2, and HDAC3 was significantly higher in M-MDSC than in PMN-MDSC (Fig. 3a). These differences were also seen in the amount of proteins (Fig. 3b). Activity of HDACI in M-MDSC was higher than that in PMN-MDSC (Fig. 3c). However, the treatment of M-MDSC with entinostat caused marked decrease in the activity of class I HDAC, so it became comparable with the activity observed in entinostat-treated PMN-MDSC (Fig. 3c). These results indicated that higher expression of class I HDACs in M-MDSC did not prevent inactivation of HDAC I activity in these cells. In addition to class I HDACs, we observed that M-MDSC had much higher amount of HDAC6 than PMN-MDSC (Fig. 3b). HDAC6 is a class II HDAC and is not a target for entinostat. Treatment of mice with entinostat did not affect expression of HDAC1, HDAC2, HDAC3, and HDAC8 in PMN-MDSC or M-MDSC (Fig. S1). In contrast, it caused substantial up-regulation of HDAC6 expression in M-MDSC (Fig. 3d).

Fig. 3.

HDAC expression in MDSC. a Total RNA was extracted from splenic PMN-MDSC and M-MDSC purified from EL4 tumor-bearing mice and analyzed by qRT-PCR (n = 4). P values were calculated in two-tailed Student’s t-test. b HDAC protein expression in whole cell lysate of splenic PMN-MDSC and M-MDSC from EL4 tumor-bearing mice (n = 2) was analyzed by western blot. c Class I HDAC activity was measured in PMN-MDSC and M-MDSC from spleen of EL4 tumor bearing mice treated with vehicle or entinostat (10 mg/kg) for 2 weeks (n = 5). P values were calculated in two-tailed Student’s t-test. dHDAC6 expression in splenic PMN-MDSC and M-MDSC from EL4 tumor-bearing mice treated with vehicle or entinostat for 2 weeks was analyzed by qRT-PCR (n = 4–5). P values were calculated in two-tailed Student’s t-test

Since entinostat affected suppressive activity of PMN-MDSC but not M-MDSC or TAM, we tested the possibility of targeting both arms of immune suppressive myeloid cell network by combining entinostat with a CSF1R antibody. Treatment with CSF1R antibody depleted TAM but caused accumulation of tumor PMN-MDSC (Fig. S2). CSF1R blockade is known to deplete TAM [18–24]. Accumulation of tumor-associated PMN-MDSC after CSF1R blockade was previously associated with regulatory effect of CSF1 on chemokine (CXCL1) production by cancer-associated fibroblasts [25].

Entinostat alone as well as CSF1R antibody alone did not affect tumor growth. In contrast, their combination caused significant antitumor effect (Fig. 4a). Thus, inactivation of PMN-MDSC in combination with targeting monocytic myeloid cells may be therapeutically valuable strategy.

Fig. 4.

Anti-tumor effect of HDAC inhibitors. a EL4 tumor-bearing mice were treated with entinostat in combination with anti-CSF1R antibody (30 mg/kg, twice a week). N = 4. P values were calculated in two-way Anova test. Results are shown as mean ± SEM. b, c. EL4 (n = 9) or LLC (n = 5) tumor-bearing mice were treated with entinostat in combination with ricolinostat. Tumor size was measured. Results are shown as mean ± SEM. P values were calculated in two-way Anova test. d M-MDSC and PMN-MDSC were purified from the spleens of EL4 tumor-bearing mice treated with entinostat and ricolinostat and cocultured with PMEL splenocytes and gp100 peptide. T cell proliferation was measured by ³H-thymidine uptake. N = 3–8 shown on each plot. P values were calculated in two-tailed Student’s t-test. Results are shown as mean ± SD

Our data also suggested that lack of the effect of entinostat on M-MDSC could be the result of up-regulation of HDAC6. Therefore, we tested the effect of HDAC6 inhibitor ricolinostat on tumor growth. Neither entinostat nor ricolinostat alone had a direct effect on EL4 or LLC tumor cells (Fig S3) and had undetectable or very minimal antitumor effect in LLC or EL4 TB mice (Fig. 4b, c). In contrast, combination of entinostat and ricolinostat substantially reduced tumor progression in both models (Fig. 4b, c). In contrast to the treatment with entinostat, treatment with ricolinostat did not cause accumulation of PMN-MDSC in tumors (Fig. S4). This was consistent with previous observation implicating HDAC2 in regulation of Cxcl1 expression in cancer-associated fibroblasts and increased accumulation of PMN-MDSC in entinostat-treated mice [25]. No effect on the presence of M-MDSC and TAM was observed with either of the treatment. Combined treatment caused decrease in the number of TAM (Fig. S4). We assessed the effect of the treatment on the suppressive activity of M-MDSC and PMN-MDSC. In contrast to entinostat, which showed potent decrease in the suppressive activity of PMN-MDSC, treatment with ricolinostat had very little effect on PMN-MDSC (Fig. 4d). However, ricolinostat significantly reduced suppressive activity of M-MDSC. Combination of entinostat and ricolinostat significantly reduced suppressive activity of both MDSC subsets (Fig. 4d). Consistent with the effect on suppressive activity, ricolinostat and combination of ricoliniostat and entinostat markedly reduced the expression of nos2, ptgs2, and tgfβ in M-MDSC (Fig. S5).

Immune suppressive myeloid cells include populations of PMN-MDSC, M-MDSC, and macrophages. Targeting just one group of cells was not sufficient to cause potent antitumor effect. For instance, targeting of CSF-1R caused substantial decrease in TAM; however, it did not affect tumor growth [18–21]. One of the reasons for that could be accumulation of PMN-MDSC in tumors of mice treated with CSF1R inhibitor [25]. In contrast, entinostat targeted only PMN-MDSC, without affecting TAM and M-MDSC. Combination of entinostat with CSF1R inhibitor resulted in substantial antitumor effect.

Class I HDACs are involved in immune suppressive activity of different cells, including myeloid cells [13, 14]. Our data indicate that in contrast to PMN-MDSC, M-MDSC has much higher expression of HDAC6 that was not a target for entinostat and thus may explain the resistance of suppressive activity of M-MDSC to class I HDAC inhibitor entinostat. HDAC6 was previously implicated in immune suppressive activity of macrophages [26]. HDAC6 acts as a transcriptional activator of IL-10 [27]. Our data indicate that treatment of mice with selective class I HDAC inhibitor caused compensatory up-regulation of HDAC6 that maintained suppressive activity of M-MDSC. As a result, inhibition of class I HDACs was not sufficient to abrogate immune suppressive activity of all myeloid cells and induce antitumor effect. Our data suggest that combination of two selective HDAC inhibitors may overcome this resistance mechanism and be a potentially promising therapeutic option.

Author contributions

AH performed most experiments and wrote manuscript, TF performed some experiments, RZ participated in research design and reviewed manuscript, and DIG obtained financial support for the study, designed overall concept and specific experiments, supervised experiments, and wrote manuscript.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Funding

This work was supported by Syndax Pharmaceuticals and by animal, genomics, and flow cytometry core facilities of Wistar Cancer Center Support NIH Grant P50 CA168536. All authors state no competing financial interests.

Compliance with ethical standards

Ethical approval

All procedures were performed and approved in strict accordance with the Institutional Animal Care and Use Committee (IACUC) at the Wistar Institute, and with the NIH Guide for the Care and Use of Laboratory Animal guidelines.