STAT3 Inhibition Combined with CpG Immunostimulation Activates Antitumor Immunity to Eradicate Genetically Distinct Castration-Resistant Prostate Cancers

This article is featured in Highlights of This Issue, p. 5785

Emerging immunotherapies exhibit robust clinical activity across a broad spectrum of late-stage tumor types, with notable exception of prostate cancers (1–3). Advanced prostate tumors are dependent on multiple cancer cell–intrinsic and –extrinsic mechanisms to promote tumorigenesis, escape immunosurveillance, and actively block immune responses (2, 4, 5). Genetic drivers of carcinogenesis, such as expression of MYC oncogene overexpression or PTEN tumor-suppressor deficiency, can differentially shape immunocyte composition of the tumor microenvironment and drive distinct mechanisms of immune evasion in prostate cancer and in other solid tumors (6–8). Tolerogenic effects of prostate tumors extend beyond the PD-1 immune checkpoint regulation. Therefore, there is a need for combinatorial immunotherapeutic approaches to disrupt complex signaling networks in the tumor microenvironment (9–13).

The STAT3 transcription factor is a multifaceted oncogene and a master regulator of immunosuppression commonly activated in the majority of human cancers (14–16). Extensive evidence suggests that tumors, such as advanced prostate cancers, critically depend on STAT3 for their survival, vascularization, and metastasis, whereas normal cells do not (17–19). The role of STAT3 in determining prostate cancer cell fate seems to be context dependent. Constitutive STAT3 activity can result in tumor progression toward hormone-refractory/castration-resistant prostate cancer (CRPC) phenotype and correlate with poor overall survival (4, 20, 21). Conversely, in PTEN-null prostate cancer cells, STAT3 activation mediates cell senescence and restricts tumor growth (22). However, STAT3 serves a consistent role as a key mediator of tumor immune evasion regardless of tumor genotype as well documented in prostate cancers, as well as other human malignancies (21, 23–25).

Previous studies demonstrated that stress and inflammation can trigger and/or sustain STAT3 activity in prostate tumors and especially in tumor-associated myeloid cells, such as macrophages (MAC) and myeloid-derived suppressor cells (MDSC; refs. 25–27). Cell death causes the release of Toll-like receptor 9 (TLR9) ligands, such as mitochondrial DNA, and TLR9/NF-κB–induced secretion of IL6-type cytokines. IL6 and Leukemia Inhibitory Factor (LIF) in turn stimulate STAT3 activity in cancer cells and myeloid cells in the tumor microenvironment (25–27). More recently, we confirmed high TLR9 expression and STAT3 activation in polymorphonuclear MDSCs (PMN-MDSC), potently immunosuppressive cells, which accumulate in circulation in prostate cancer patients during disease progression from localized to metastatic/CRPC (mCRPC; refs. 25, 28). The TLR9⁺ PMN-MDSCs relied on the STAT3-mediated expression of Arginase-1 (ARG-1) to block T-cell proliferation and activity (28). These effects underscore well-known pivotal role of MDSCs in prostate cancer progression and poor overall survival (25, 28–32).

As a master regulator operating in both cancer cells and in tumor-associated immune cells, STAT3 is a unique and highly desirable target for cancer therapy (14, 16, 29, 33). Due to the lack of enzymatic activity, pharmacologic inhibition of STAT3 is challenging (33, 34). Emerging oligonucleotide-based strategies to inhibition of STAT3 signaling, such as decoy and antisense oligonucleotides (ASO), showed promise in phase I clinical trials (35, 36). However, the efficacy of STAT3ASO is limited by the lack of cell-selectivity and targeted delivery, which decrease oligonucleotide penetration into the solid tumor microenvironment and efficacy (29, 36). To improve the pharmacologic properties of oligonucleotide therapeutics (ONT), we previously developed a strategy for targeted delivery of oligonucleotides, such as siRNA, specifically to tumor-associated TLR9⁺ immune cells and TLR9⁺ cancer stem–like cells in prostate tumors (27–29, 37, 38). Here, we describe the conjugate of CpG oligodeoxynucleotide (ODN), a synthetic TLR9 ligand, with chemically modified STAT3ASO molecules, with improved nuclease resistance suitable for systemic administration. We characterized effects of CpG-STAT3ASO on both human and mouse cellular targets in vitro and in two syngeneic models of bone-localized prostate tumors. Our studies assessed two-pronged effects of the conjugate, directly on prostate cancer cells and indirectly, through immune-mediated antitumor immune responses.

Human DU-145 prostate cancer (HTB-81) and mouse RAW264.7 cells (TIB-71) were purchased from the American Type Culture Collection. Mouse DC2.4 cells were originally from Dr. K. Rock (University of Massachusetts Medical School, Worcester, MA), RM9 cells were kindly provided by Dr. T. Thompson (The University of Texas MD Anderson Cancer Center, Houston, TX), and PPS (Pten^pc−/− Trp53^pc−/− Smad4^pc−/−) cells were recently generated by Drs. R.A. DePinho and X. Lu (6). All tested cells were cultured for less than 6 months before experiments and tested bimonthly to be free from mycoplasma infections. For oligonucleotide uptake studies, DU-145 cells were pretreated for 1 hour in the presence of 50 μg/mL dextran sulfate (#42867), 100 μg/mL of fucoidan (#F8190), 1 mmol/L of amiloride (#A7410), 150 μmol/L of genistein (#G6649), 2 μg/mL of filipin (#F4767), 100 μmol/L of cadaverine (#33211), 50 μg/mL of chondroitin sulfate (#4384; Sigma-Aldrich), or at +4°C prior to treatment with oligonucleotides followed by flow cytometric analysis.

The CpG-ODN conjugates were synthesized in the DNA/RNA Synthesis Core (COH) by linking CpG-ODNs to STAT3 ASO similarly as previously described (37). The resulting ODN conjugates are shown below (x indicates a single C3 unit; asterisk indicates 2′O methylation; and underline indicates phosphorothioation site):

CpG-STAT3ASO - (CpGD19 ODN + human STAT3 ASO targeting sequence):

5′ GGTGCATCGATGCAGGGGGG-xxxxx-C*A*G*C*A*GATCAAGTCCA*G*G*G*A* 3′.

STAT3 ASO (human STAT3 ASO targeting sequence):

5′ C*A*G*C*A*GATCAAGTCCA*G*G*G*A* 3′.

CpG-scrON - (CpGD19 ODN + scrambled ASO sequence):

5′ GGTGCATCGATG CAGGGGGG-xxxxx-A*G*A*G*C*CTAACGGAAGG*C*A*C*T* 3′.

CpG-mSTAT3ASO - (CpG1668 ODN + mouse STAT3 ASO targeting sequence):

5′ TCC ATG ACG TTC CTG ATG CT-xxxxx-G*A*C*T*C*TTGCAGGAATC*G*G*C*T* 3′

mSTAT3ASO - (mouse STAT3 ASO targeting sequence):

5′G*A*C*T*C*TTGCAGGAATC*G*G*C*T*3′

CpG-scrON (CpG1668 ODN + scrambled ASO targeting sequence):

5′ TCC ATG ACG TTC CTG ATG CT-xxxxx-A*G*A*G*C*CTAACGGAAGG*C*A*C*T* 3′

For internalization studies, oligonucleotides were labeled on 3′ ends using Cy3 or Alexa488 fluorochromes.

NOD/SCID/IL-2RγKO (NSG) and C57BL/6 mice, aged between 6 and 8 weeks, were purchased from the Jackson Laboratory. Mouse care and experimental procedures were performed under pathogen-free conditions in accordance with established institutional guidance and approved protocols from Institutional Animal Care and Use Committees. For subcutaneous tumor growth experiments, RM9 (1.5 × 10⁵) or PPS (8 × 10⁵) cells were injected s.c., and the tumors size was measured every other day. When tumors reached approximately 150 mm³ size, mice were injected i.t. every other day using 5 mg/kg CpG-STAT3ASO, STAT3ASO or CpG-scrON. For the intratibial tumor experiments, the animals were injected intratibially with 1 × 10⁴ of RM9- or PPS mCherry/luciferase-expressing cells. Tumor engraftment and progression were monitored using bioluminescent imaging (BLI) on the AmiX (Spectral Instruments). After tumor engraftment, mice were injected systemically using 5 mg/kg of CpG-STAT3ASO, STAT3ASO, or CpG-scrON, every other day.

Total RNA was extracted from cultured or in vivo grown tumor cells using Maxwell system (#AS1390; Promega) and then transcribed into cDNAs using the iScript cDNA Synthesis Kit (#1725064; Bio-Rad). The qPCR was carried out using specific primers for STAT3, 18S, actin, and TBP, as previously described (38, 39), using CFX96 Real-Time PCR Detection System (Bio-Rad).

Total cellular lysates were prepared and analyzed using antibodies specific to tyrosine 705-phosphorylated or total STAT3 (#9131; Cell Signaling Technology) and β-actin (#A3854; Sigma-Aldrich). Immunohistochemistry using anti-pSTAT3 and anti-Ly6B (clone 7/4, # CL8993AP; Cedarlane) antibodies was performed as previously described (37). For immunohistochemical staining, the formalin-fixed paraffin-embedded tumor sections were stained using primary antibodies and horseradish peroxidase–conjugated secondary antibodies and then analyzed on the Observer II microscope (Zeiss).

Single-cell suspensions were prepared by mechanic dispersion and enzymatic digestion of tumor tissues. Extracellular staining was performed using fluorochrome-labeled antibodies to MHC class II (AF6-120.1, #553552), CD3 (145-2C11, #11-0031-85), CD4 (RM4-5, #12-0042-82), CD8 (53-6.7, #12-0081-83), CD11b (M1/70, #45-0193-82), CD40 (1C10, #12-0401-83), CD80 (16-10A1, #11-0801-85), CD86 (GL1, #25-0862-82), Ly6C (HK1.4, #53-5932-82), or Ly6G (RB6-8C5, #25-5931-82; eBioscience) after anti-FcγIII/IIRBlock were used (#558636; BD Biosciences). Human immune cells and peripheral blood mononuclear cells (PBMC) were analyzed using the following antibodies: HLA-DR (LN3, #12-9956-42), CD1c (L161, #17-0015-14), CD3 (OKT3, #11-0037-42), CD14 (61D3, #11-0149-42), CD15 (MC-480, #12-8813-42), CD16 (CB16, #17-0168-42), CD19 (HIB19, #11-0199-42), and CD303 (201A, #11-9818-42). For intracellular staining, cells were fixed/permeabilized and immunostained for pSTAT3 (4/P-STAT3, #562071; BD Biosciences), FoxP3 (236A/E7, #53-4777-41; eBioscience), or Arginase-1 (SL6ARG, #IC5868F; R&D Systems) as previously described (28). Fluorescence data were analyzed on BD Fortessa and an AccuriC6 Flow Cytometer (BD Biosciences) using FlowJo software (TreeStar).

DU-145 cells were cultured in RPMI 1640 medium with 10% FBS using 24-well plates on top of laminin-coated coverslips, and then treated using 250 nmol/L CpG-STAT3ASO^Cy3. The coverslips were fixed in 2% paraformaldehyde (#15710; EMS), permeabilized with 0.1% Triton X-100 (#X100-500ML; Sigma-Aldrich), and stained using primary anti-EEA1 (sc-33585; Santa Cruz Biotechnology) or anti–RNase H1 (#NBP2-38501; Novus Biologicals) and Alexa488-coupled secondary antibodies (#A32723; Sigma-Aldrich). The proximity ligation assays were performed using Cy3- and RNase H–specific antibodies as reported (40). Slides mounted in Vectashield Hard-Set medium (#H-1400; Vector Laboratories) were visualized on an LSM510-Axiovert inverted confocal microscope (Zeiss) and analyzed using LSM ImageBrowser (version 4.2.0.121; Zeiss).

Studies using human blood samples were performed in accordance with the ethical standards and according to the Declaration of Helsinki under the institutional review board approvals (IRB13141 and IRB 12367) with written, informed consent of all patients. For studies on human PMN-MDSCSs, CD15⁺ cells were isolated from peripheral blood of patients with metastatic prostate cancers or from healthy subjects and analyzed as previously described (28). Flow cytometric analysis was performed to assess T-cell proliferation using CFSE dilution (#C34554; Sigma-Aldrich) and IFNγ and granzyme B production by CD8⁺ T cells using specific antibodies to CD3, CD8 (RPA-T8, #25-0088-42), IFNγ (#13-1191-82; eBioscience), and granzyme B (GB11, #561998; BD Biosciences).

The Unpaired t test was used to calculate two-tailed P value to estimate statistical significance of differences between two treatment groups. One- or two-way ANOVA plus Bonferroni posttest were applied to assess differences between multiple groups or in tumor growth kinetics experiments. Statistically significant P values are indicated in figures as follows: *, P < 0.05; **, P < 0.01; and ***, P < 0.001. Data were analyzed using Prism software v. 6.01 (GraphPad).

To improve the efficiency and selectivity of STAT3 inhibition in the microenvironment of prostate tumors, we conjugated a STAT3 ASO with a TLR9 agonist, single-stranded CpG(D19) ODN, using a synthetic carbon linker (Fig. 1A; ref. 37). The STAT3ASO has a standard gapmer design, with targeting sequence flanked by 2′-O-methyl–modified nucleotides, to induce RNase H1–dependent knockdown as previously characterized (41). For enhanced nuclease resistance, the sugar backbone of the conjugate was also partly phosphorothioated. The half-life of CpG-STAT3ASO in 50% human serum exceeded 4 days (T_1/2 = 106 hours; Supplementary Fig. S1). The conjugation of STAT3ASO to CpG moiety improved oligonucleotide delivery to target primary human immune cells compared with the antisense molecule alone (Fig. 1B, left). When incubated for 1 hour with PBMCs, fluorescently-labeled CpG-STAT3ASO^Alexa488 was internalized by dendritic cells (DC) and B cells more effectively than STAT3ASO^Alexa488 alone at 500 nmol/L concentration. The uptake of CpG-STAT3ASO was dose-dependent and detectable even at 50 nmol/L in human and mouse myeloid cells and B cells, but not in T cells (Fig. 1B, right; Supplementary Fig. S2). For testing on mouse target cells, we used conjugates with sequences optimized for immunostimulation and targeting STAT3 in mice (CpG1668-mSTAT3ASO), but these modifications did not affect the pattern of conjugate uptake (Fig. 1C; Supplementary Fig. S2). The CpG-STAT3ASO was also effectively internalized by TLR9⁺ prostate cancer cells such as human DU-145 or mouse RM9 cells (Fig. 1C; ref. 27). Cultured prostate cancer cells also internalized STAT3ASO alone but required significantly higher concentrations (500 nmol/L) compared with CpG-STAT3ASO conjugates.

Uptake of ASO can occur through various surface receptors and endocytic mechanisms, but some of these internalization pathways are “non-productive” as they do not result in target gene knockdown (42). To identify these mechanisms, we compared the effect of endocytosis inhibitors on the uptake of fluorescently-labeled CpG-STAT3ASO versus STAT3ASO by DU-145 cells, at 250 and 750 nmol/L, respectively, for comparable baseline level of uptake. Target cells were preincubated at reduced temperature (4°C) or in the presence of dextran sulfate, a general inhibitor of scavenger receptor (SR)–mediated internalization, or with other more specific inhibitors of endocytosis, such as fucoidan, amiloride, cadaverine, and filipin. We observed that both oligonucleotides underwent active and mainly SR-A–mediated internalization (Fig. 1D; Supplementary Fig. S3). However, the presence of CpG moiety enhanced clathrin-mediated uptake as indicated by the increased sensitivity to cadaverine, which is one of the “productive” uptake pathways for ASO molecules (42). The improved internalization of CpG-STAT3ASO could be facilitated by the increased size and the additional phosphorothioate modifications of the conjugate (43). When assessed using confocal microscopy, most of CpG-STAT3ASO^Cy3 was found located in early endosomes within 30 minutes of incubation (Fig. 1E). After 1 hour, the endosomal signal of CpG-STAT3ASO^Cy3 decreased, which likely indicates cytoplasmic release of the conjugate (Supplementary Fig. S4; ref. 40). At 4 hours, CpG-STAT3ASO^Cy3 partially colocalized and directly interacted with RNase H1 in the cytoplasm as detected using confocal microscopy (Fig. 1F) and in situ proximity ligation assay (PLA) using antibodies specific to Cy3 and RNase H1 (Fig. 1G). In contrast, we did not detect interaction of the RNase H1 with control Cy3-labeled conjugates (CpG-STAT3siRNA or CpG-STAT3dODN), or with the Alexa488-labeled CpG-STAT3ASO (Fig. 1G).

The ASO induces RNase H1-dependent cleavage of the specific target mRNA, which is then degraded by cytoplasmic nucleases (44, 45). To verify whether a CpG-STAT3ASO conjugate retains similar activity as the STAT3ASO alone, we tested STAT3 knockdown in several immune and prostate cancer cells (Fig. 2). Consistent with the pattern of uptake (Fig. 1C), both CpG-STAT3ASO and STAT3ASO induced similar levels of STAT3 knockdown in mouse immune cells, such as DCs and MACs (Fig. 2A). Both oligonucleotides had similar activity also in mouse RM1/9 prostate cancer cells and to a lesser extent in PPS (Pten^pc−/− Trp53^pc−/− Smad4^pc−/−) cells (6), likely due to poor internalization of both molecules (Fig. 2B; Supplementary Fig. S5A). At lower concentrations, CpG-STAT3ASO was clearly more effective in reducing STAT3 protein levels in mouse immune cells (Fig. 2C). The improved inhibitory effect of CpG-STAT3ASO was more pronounced in human prostate cancer cells such as LAPC4, CWR-22rv1 (Fig. 2D), and DU-145 cells (Fig. 2E). In DU-145 cells, CpG-STAT3ASO showed accelerated kinetics of STAT3 knockdown, which was already detectable at 8 to 12 hours compared with 24 hours required for STAT3ASO (Fig. 2E). At the protein level, CpG-STAT3ASO reduced STAT3 activation (Y705-phosphorylation) after 24 hours, and the total protein levels were significantly reduced after 48 hours (Fig. 2F). Correspondingly, CpG-STAT3ASO showed increased cytotoxicity against STAT3-dependent prostate cancer cells such as DU-145 and RM9 compared with STAT3ASO (Supplementary Fig. S6). Together, these results show that in vitro CpG-STAT3ASO conjugate shows better or at least comparable potency against human and mouse target cells compared with STAT3ASO.

As noted, STAT3 serves multiple context-specific roles in tumor biology (22). We first assessed local and systemic antitumor effects of CpG-STAT3ASO versus STAT3ASO or CpG immunostimulation alone, using the syngeneic Ras/Myc-driven (Pten intact) RM9 prostate cancer model, which generates potently immunosuppressive tumor microenvironment (46). In vitro, RM9 cancer cells showed sensitivity to STAT3 inhibition (Supplementary Fig. S6B). Mice with RM9 tumors engrafted s.c. into opposite flanks were treated using i.t. injections of either CpG-STAT3ASO, CpG-scrambled conjugate (CpG-scrON), STAT3ASO alone (5 mg/kg each), or vehicle only into only one tumor site. As shown in Fig. 3A, both CpG-STAT3ASO and STAT3ASO induced comparable STAT3 knockdown. Both treatments inhibited growth of RM9 tumors in the locally treated site, but the antitumor effect of STAT3ASO was transient and followed by tumor regrowth (Fig. 3B). The control CpG-srcON conjugate, which is immunostimulatory but nontargeting, did not affect growth of the potently immunosuppressive RM9 tumors (Fig. 2B; ref. 25). Only CpG-STAT3ASO reduced tumor progression in the distant/noninjected site, which suggests generation of systemic antitumor effects (Fig. 3C).

Next, we compared antitumor efficacy of these oligonucleotides in the genetically distinct, PPS (Pten^pc−/−Trp53^c−/−Smad4^pc−/−) metastatic prostate cancer model, which is engineered with mutations commonly found in human tumors (6). In contrast to RM9, our preliminary in vitro studies found that PPS cells were resistant to STAT3 knockdown and did not reduce their proliferation and viability (Supplementary Fig. S5B and S5C). As before, PPS tumors were engrafted s.c. in mice on both flanks, but only tumors in the left site were treated. Local injections of CpG-STAT3ASO and STAT3ASO alone reduced STAT3 expression in whole tumors, although less efficiently than in RM9 tumors (Fig. 3A and D). In the PPS model, only CpG-STAT3ASO abrogated tumor growth in both treated and distant locations, with complete rejection of 7 of 10 in primary and 3 of 10 in distant site. Moreover, in the treated site, STAT3ASO injections seemed to stimulate tumor progression (Fig. 3E and F). Tumor-promoting effect of STAT3ASO at high local oligonucleotide concentrations may indicate the role of STAT3 in regulating cancer cell senescence in PTEN-deficient prostate tumors by yet unclear p53-independent molecular mechanism (22). The inhibitory effect of control CpG-scrON was minimal, indicating failure of TLR9 stimulation alone to overcome tolerogenic effects of the tumor microenvironment. Importantly, our results suggest that the concomitant STAT3 inhibition and TLR9 stimulation are necessary for the generation of systemic antitumor immune effects against prostate tumors, independently from the intrinsic sensitivity of prostate cancer cells to STAT3 inhibition.

Prostate tumors are known for harboring potently tolerogenic microenvironments, which thwarts antitumor immunity. Thus, we assessed whether CpG-STAT3ASO administration was capable of alleviating tumor immunosuppression and stimulating systemic T-cell activity. Previous studies from our group and others documented the critical role of tumor-associated myeloid cells, and specifically PMN-MDSCs, in sustaining prostate cancer immune evasion (6, 25, 27–29, 47). We used flow cytometry to characterize changes in immune cell populations infiltrating RM9 or PPS tumors after local i.t. administration of either 5 mg/kg CpG-STAT3ASO conjugates, STAT3ASO alone, control CpG-scrON, or vehicle as before. The immunophenotypic analysis revealed treatment-dependent differences in the composition of the tumor microenvironment in the two tumor models (Fig. 4; Supplementary Fig. S7). The untreated RM9 tumors showed low level of infiltration by immune cells, mainly immature myeloid cells with almost undetectable levels of T cells compared with PPS tumors (Fig. 4A and B). Only CpG-STAT3ASO treatment, but not CpG-scrON or STAT3ASO alone, significantly altered the RM9 microenvironment by inducing recruitment of CD11b⁺Ly6G⁺ myeloid cells into primary tumors and CD8 T cells into the distant tumor site (Fig. 4A). Immunohistochemical analysis indicated that these CD11b⁺Ly6G⁺ cells represented activated Ly6B.2⁺ neutrophils rather than PMN-MDSCs (Fig. 4C). This observation corresponds to our original findings of increased neutrophil activity in mice with hematopoietic cell–selective Stat3 ablation, especially in response to local TLR9 stimulation (48, 49). In fact, the CD11b⁺Ly6G⁺ myeloid cells infiltrating RM9 tumors in CpG-STAT3ASO–treated mice showed reduced phospho-STAT3 (pSTAT3) levels compared with mice from other treatment groups as assessed by intracellular staining and flow cytometric analysis (Fig. 4D). Although STAT3ASO decreased STAT3 activity in the treated tumor site as assessed by immunohistochemistry (Fig. 4C), it did not significantly inhibit STAT3 in CD11b⁺Ly6G⁺ myeloid cells recruited into tumors (Fig. 4D). STAT3 is known to be a direct upstream regulator of PDL1 transcription in human or mouse monocytes and cancer cells (50, 51). Consistent with the level of STAT3 inhibition, injections of CpG-STAT3ASO, but not STAT3ASO, reduced PD-L1 levels by approximately 80% in RM9 tumor–associated CD11b⁺Ly6G⁺ cells (Fig. 4E).

We next compared effects of locally administered (i.t.) oligonucleotides on the PPS tumor microenvironment, which is potently tolerogenic due to high percentage of tumor-associated PMN-MDSCs (6, 47). Local i.t. injections of CpG-STAT3ASO reduced the population of putative PMN-MDSCs, CD11b⁺Ly6G⁺Ly6C^LO granulocytic myeloid cells, in treated tumors from 26% to 14% on average, while stimulating recruitment of CD8⁺ T cells to both tumor sites (Fig. 4B). Unexpectedly, treatment with STAT3ASO alone had the opposite effect. It increased the percentage of CD11b⁺Ly6G⁺Ly6C^LO myeloid cells in treated site but failed to recruit CD8⁺ T cells into tumors (Fig. 4B; refs. 6, 47). The expansion of tumor-associated myeloid cells by the local STAT3ASO treatment could support the unexpected effect of the accelerated progression of primary PPS tumors (Fig. 3E). The effect of CpG-scrON on myeloid and T-cell populations was negligible (Fig. 4B), indicating failure of TLR9 stimulation alone to counteract the immunosuppressive effect of the tumor microenvironment. While inhibiting tolerogenic CD11b⁺Ly6G⁺ cells, CpG-STAT3ASO also augmented activation of DCs in tumor-draining lymph nodes as detected by the elevated levels of MHC class II, as well as CD40 and CD80 costimulatory molecules, and decreased percentage of regulatory T cells (Treg; Supplementary Fig. S8A and S8B). Further flow cytometric analysis confirmed that that CpG-STAT3ASO and to a lesser extent CpG-scrON, but not STAT3ASO, increased percentage of CD8⁺ T cells while reducing Tregs in distant tumors in both RM9 and PPS models. The significant rise of CD8:Treg ratio is a critical indicator of successful generation of systemic T-cell–mediated response (Fig. 4F and G).

To further assess the contribution of granulocytic myeloid cells and T-cell lymphocyte populations in the antitumoral immune responses induced by CpG-STAT3ASO, we individually depleted Gr1⁺, CD4⁺, or CD8⁺ cells in RM9 tumor–bearing mice. Following the antibody-mediated neutralization, we treated the mice using 5 mg/kg of CpG-STAT3ASO at the left tumor site, as before. The CD8 depletion partly impaired antitumor CpG-STAT3ASO activity against treated tumors, while completely eliminating any effects against distant tumors (Fig. 4H). The depletion of Gr1⁺ and CD4⁺ cells had a strong negative impact on the antitumor effects of CpG-STAT3ASO against tumors in both treated and the distant sites (Fig. 4H). These results suggest that only the combination of STAT3 inhibition with TLR9 stimulation had potential to disrupt tumor immune evasion while effectively engaging multilayered cellular immune network, thereby combining both innate, neutrophil-dependent, and adaptive, T-cell–mediated antitumor immunity.

The potential to induce systemic antitumor immune responses using CpG-STAT3ASO conjugates prompted us to test the feasibility of using this strategy to treat metastatic disease. First, we compared biodistribution of CpG-STAT3ASO versus STAT3ASO alone after systemic administration into tumor-bearing mice. Mice with established intratibial RM9 tumors were injected i.v. using 2.5 mg/kg of fluorescently Cy3-labeled oligonucleotides. Cellular biodistribution of both oligonucleotides was measured using flow cytometry in various immune cell populations, such as DCs, MACs, total MDSCs (CD11b⁺Gr1⁺), and T cells in bone marrow and spleen after 3 hours. As shown in Fig. 5A, CpG-STAT3ASO was internalized by >80% of MDSCs and approximately 45% of DCs in the bone marrow and at a lower percentage (∼30%) in the spleen but not by T cells in any of the tested organs. Although STAT3ASO showed similar biodistribution pattern, it penetrated a significantly lower percentage of MDSCs (∼65%) and DCs (∼30%) in the bone marrow. Given the superior uptake of ASOs by bone marrow–localized myeloid cells, we decided to test their efficacy against prostate tumors in the clinically relevant localization (52). Mice were injected intratibially using luciferase-expressing RM9-Luc cancer cells. After tumors were established, as verified using BLI, mice were i.v. injected with every other day using 5 mg/kg CpG-STAT3ASO, STAT3ASO, CpG-scrON, or PBS. The levels of STAT3 activity were assessed in tumor sections using immunohistochemistry and flow cytometry day after the third injection. Repeated systemic administration of CpG-STAT3ASO strongly reduced the overall pSTAT3 levels in bone-localized RM9 tumors in contrast to both control treatments (Fig. 5B). Flow cytometric analysis indicated reduction of pSTAT3 levels in various tumor-infiltrating immune cell populations, such as DCs and MDSCs, after i.v. treatment using CpG-STAT3ASO and STAT3ASO (Fig. 5C). Consistent with the lack of oligonucleotide uptake, STAT3 activity was not reduced but slightly elevated in tumor-resident T cells after CpG ODN and CpG-STAT3ASO treatments (Fig. 5C). It remains to be tested whether such STAT3 upregulation could indicate CD8 T-cell activation and expansion (53).

Based on these results, we decided to compare the efficacy of CpG-STAT3ASO, STAT3ASO alone, and control CpG-scrON against bone-localized RM9 prostate tumors. Mice with established (day 13) intratibial RM9-Luc tumors were treated every other day using i.v. injections of 5 mg/kg of above-mentioned oligonucleotides or PBS for a total of six injections. As shown in Fig. 5D, tumors progressed rapidly in all control experimental groups except for CpG-STAT3ASO. Already after the second injection (day 15), CpG-STAT3ASO treatment halted tumor progression, and within the following week, tumors regressed in the majority of treated mice (Fig. 5D). The surviving CpG-STAT3ASO–treated mice remained tumor free until the end of experiment at 58 days, whereas survival of mice in other treatment groups did not exceed 30 days (Fig. 5E). Importantly, the coinjection of unconjugated CpG ODN together with STAT3ASO failed to significantly extend animal survival (Fig. 5E and F).

Our earlier results (Fig. 5B) indicated that the efficacy of systemic administration of CpG-STAT3ASO could depend on direct STAT3 targeting in RM9 prostate cancer cells as well as in myeloid cells in the tumor microenvironment. In contrast, the PPS tumor model allows for more selective analysis of cancer cell–extrinsic effects of tested oligonucleotides due to the resistance of cancer cells to direct cytotoxic effect of the conjugate (Supplementary Fig. S5). Mice with established intratibial, luciferase expression PPS tumors (PPS-Luc) were injected i.v. every other day using CpG-STAT3ASO, STAT3ASO alone, and CpG-scrON (5 mg/kg) or PBS and monitored using BLI as before. Treatment with CpG-STAT3ASO led to complete regression of PPS tumors and resulted in tumor-free survival of the 90% of treated mice for up to 230 days after engraftment (Fig. 5G). None of other treatments resulted in statistically significant extended animal survival compared with PBS group. Finally, to assess the contribution of immune responses to antitumor effects of CpG-STAT3ASO, we engrafted RM9-Luc tumors into the tibia of immunodeficient NOD/SCID/Il2rγKO (NSG) mice. Similarly as before, after tumors were established as verified by BLI, mice were treated every other day using i.v. injections of 5 mg/kg of CpG-STAT3ASO, STAT3ASO alone, CpG-scrON, or PBS. As shown in the Fig. 5H, in the absence of functional immune cells, CpG-STAT3ASO injections failed to control RM9 tumor progression and did not extend mice survival. These results suggested that the presence of functional immune cells is critical for the antitumor efficacy of CpG-STAT3ASO. In agreement with these results, flow cytometric analysis confirmed that i.v. injections of CpG-STAT3ASO, but not control oligonucleotides, improve about -fold the ratio of effector to Treg populations in the RM9 prostate tumor microenvironment (Supplementary Fig. S8C–S8E). These data further underscore therapeutic potential of combining systemic disruption of STAT3-mediated prostate tumor immune evasion with concomitant CpG immunostimulation in order to unleash antitumor immunity against bone-localized prostate cancers.

We previously described accumulation of potently immunosuppressive population of TLR9⁺ PMN-MDSCs in prostate cancer patients during disease progression (28). Because human PMN-MDSCs rely on STAT3 signaling for their tolerogenic effects on T cells, we tested whether CpG-STAT3ASO can target human PMN-MDSCs. We first assessed the spontaneous uptake of CpG-STAT3ASO and unconjugated STAT3ASO by primary PBMCs from patients with advanced prostate cancers. Within 1 hour of incubation with fluorescently-labeled oligonucleotides, 80% to 90% of PMN-MDSCs (Lin⁻HLA⁻DR⁻CD15⁺) internalized both CpG-STAT3ASO^Alexa488 and STAT3ASO^Alexa488, although the intracellular levels of the CpG-STAT3ASO conjugate were significantly higher as indicated by mean fluorescent intensity (Fig. 6A). Subsequently, the level of target gene silencing was assessed in CD15⁺ PMN-MDSCs after 72-hour incubation with CpG-STAT3ASO, STAT3ASO, or CpG-scrON using real-time PCR. The results showed >90% reduction in STAT3 mRNA for both STAT3 targeting oligonucleotides compared with controls (Fig. 6B). Given the comparable effect of CpG-STAT3ASO and STAT3ASO oligonucleotides on STAT3 knockdown ex vivo, we assessed the effect of both treatments and control CpG-scrON on the immunosuppressive properties of PMN-MDSCs. CD15⁺ PMN-MDSCs enriched from blood of advanced prostate cancer patients were preincubated for three days in the presence of 500 nmol/L of CpG-STAT3ASO, STAT3ASO, or CpG-scrONs. Viable PMN-MDSCs were then cocultured (3:1) with allogeneic CD3⁺ T cells and CD3/CD28 costimulation for additional 3 days. As indicated in Fig. 6C, only the CpG-STAT3ASO conjugate was able to alleviate the multiple immunosuppressive effects of PMN-MDSCs on CD3⁺ T-cell proliferation. The CpG-STAT3ASO conjugate also stimulated the production of granzyme B and IFNγ by CD8⁺ T cells (Fig. 6D and E). Neither STAT3ASO alone nor immunostimulatory CpG-scrON resulted in significant improvement of CD8⁺ T-cell activity. Therefore, we conclude that, as in mouse tumor models, overcoming the MDSC-dependent immunosuppression in the prostate cancer microenvironment requires the concomitant STAT3 inhibition and TLR9 stimulation.

Here, we demonstrate the feasibility of using bifunctional oligonucleotide strategy to overcome de novo therapeutic resistance of potently immunosuppressive and immunologically “cold” prostate tumors. The combination of targeted STAT3 inhibition and TLR9 immunostimulation was required to disrupt myeloid cell–dependent immunosuppression and to unleash antitumor immune responses, eradicating aggressive, bone-localized prostate tumors. Importantly, we showed that neither treatment with STAT3ASO or CpG ODN as single treatments, nor their coinjection was sufficient for tumor eradication. The combination of STAT3 inhibitor and TLR9 agonist in a single molecule maximized treatment efficacy by several mechanisms. First, it enhanced and accelerated cell-selective uptake of systemically administered oligonucleotide, thereby improving STAT3 knockdown. Secondly, TLR9 immunostimulation combined with the STAT3 inhibition resulted in the “push & release” effect, essential for jump-starting antitumor immunity. As shown by our studies comparing the i.t. injections of CpG-STAT3ASO versus STAT3ASO alone, even when STAT3 knockdown levels were comparable, only the CpG-STAT3ASO conjugate had potential to overcome tumor immune tolerance and induce systemic antitumor immune responses. This is consistent with our previous report, which demonstrated that combined STAT3 inhibition/TLR9 stimulation specifically in myeloid cells and not in cancer cells is sufficient for eradication of solid tumors through efficient recruitment of innate and adaptive effector cells, such as neutrophils and CD8/CD4 T cells (49).

CpG-STAT3ASO strategy provides an opportunity for the development of immunotherapies broadly applicable to therapy of genetically diverse prostate cancers. As recently described, Myc oncogene expression as well as combined deletion of Pten/Smad4 or Pten/Tp53 causes expansion of tumor-associated MACs and MDSCs, and thereby promotes tumor immune tolerance and vascularization (6, 8, 47). Both Ras/Myc-driven and Pten deletion–dependent prostate tumors were sensitive to CpG-STAT3ASO–induced immune responses, even though cancer cells in both models differed dramatically in their intrinsic sensitivity to STAT3 inhibition. As discussed above, due to pharmacokinetic properties, systemic administration of CpG-STAT3ASO had limited direct effect on cancer cells, as indicated by the lack of significant antitumor activity against STAT3-dependent RM9 tumors in immunodeficient mice. This is of benefit considering that STAT3 may gain tumor-suppressor function in Pten-loss prostate tumors (15, 22). STAT3 inhibition in Pten-deficient cancer cells interfered with the induction of p53-dependent cell senescence, promoting tumor growth in vivo. The transplantable PPS (Pten^pc−/−Trp53^c−/−Smad4^pc−/−) model of castration-resistant prostate tumors used in our studies lacks p53. Therefore, the accelerated PPS tumor progression after local STAT3ASO treatment suggests that STAT3 may have additional p53-independent tumor-suppressor functions in Pten-deficient prostate cancers (6). Our future studies will interrogate molecular mechanisms underlying these effects, which can be of concern for a number of patients with PTEN-deficient prostate cancers. Importantly, in contrast to STAT3ASO alone, CpG-STAT3ASO showed consistent if not higher antitumor activity against Pten-deficient prostate tumors, as observed in Ras/Myc-driven tumors. These results underscore broad therapeutic potential of targeting myeloid cells in the prostate tumor microenvironment, which likely reflects the fundamental role of myeloid cell–dependent immune evasion for tumor progression (31).

To date, prostate cancers have proven to be difficult targets for emerging immunotherapies, including cancer vaccines and immune checkpoint blockade (5, 9). Recent studies highlighted the role of the tumor microenvironment, especially MDSCs, in shielding prostate cancers from antitumor immunity (25, 31, 47). Lu and colleagues demonstrated in mouse tumor models that targeting MDSCs, predominantly of the PMN-MDSCs subtype, can augment the modest efficacy of immune checkpoint inhibitors against mCRPC (47). We recently found that potently tolerogenic TLR9⁺ PMN-MDSCs, with high levels of STAT3 activity, accumulate in blood of prostate cancer patients with progression of the disease (28). STAT3 activation in PMN-MDSCs correlated with elevated plasma levels of IL6-type cytokines, such as LIF, suggesting a potential cross-talk mechanism promoting tumor immune evasion (25). IL23, another cytokine and STAT3-regulated target in tumor-associated myeloid cells (54), was recently identified as a potential driver of CRPC (55). Both RM9 and PPS prostate tumor models replicate well these clinical findings and show high percentage of pSTAT3⁺ tumor-infiltrating PMN-MDSCs (6, 25, 47). STAT3 activity in these tumor-infiltrating MDSCs correlated with elevated levels of PD-L1, a known a STAT3 target gene (50, 51) and a key immune checkpoint regulator (56, 57). However, targeting MDSCs in clinical setting remains a challenge. The lack of distinct surface markers for these immature myeloid cells complicates their neutralization using antibodies. Recent studies focused on small-molecule drugs inhibiting immunosuppressive functions of MDSCs. Blocking Jak/STAT3 signaling using broadly specific tyrosine kinase inhibitors, such as sunitinib or cucurbitacin B, or in vitro inhibition of STAT3 was shown to alleviate tolerogenic MDSCs activity (30, 58). PI3K inhibitors, such as BEZ235 and cabozantinib (HGF/PI3K), can also overcome MDSC-mediated immunosuppression, while promoting innate immune responses against prostate tumors (47, 59). At the same time, targeting signaling molecules across the immune network can result in contradictory outcomes in various immune cell populations. This is of concern for both PI3K/mTOR and Jak/STAT3 inhibitors due to complex role of these target pathways in T lymphocytes. For example, STAT3 seems to reduce antitumor activity of CD8 T cells and expand tumor-promoting Th17 lymphocytes, but it is also indispensable for the generation of memory T cells and long-term antitumor immunity (60–62). Targeting Jak1/2 kinases upstream from STAT3 was shown to reduce numbers of MDSCs, while paradoxically increasing their immunosuppressive activity and blocking T-cell proliferation (63). Similarly, PI3K and PI3K/mTOR inhibitors can interfere with T-cell activation and induce tolerance (64). Therefore, targeting tolerogenic signaling in tumor-associated myeloid cells demands cell-selective strategies to avoid potential adverse effects and toxicities, while maximizing therapeutic efficacy.

ONTs provide an alternative and clinically relevant strategy for targeting tumor-associated myeloid cells (29). The sensitivity of innate immune cells to ONTs has long been a major obstacle to the clinical development of these reagents beyond oncology. Paradoxically, the myeloid cells may be essential cellular targets for cancer immunotherapy. This is best exemplified by the recently reported results from the phase I clinical trial on STAT3ASO molecule (AZD9150) in patients with various relapsed/refractory tumors (36, 65). Similar as in preclinical models, i.v. injected STAT3ASO was found to primarily target nonmalignant, stromal cells rather than cancer cells. Importantly, nonautonomous effects of STAT3ASO in tumor cells were likely essential for the observed antitumor effects (65, 66). Consistent with our preclinical data, partial clinical responses to STAT3ASO alone occurred in patients with B-cell lymphoma but not with the advanced prostate cancer (36). The limited penetration of the STAT3ASO into tumors was likely a result of the slow uptake kinetic and the limited circulatory half-life of ASOs due to renal clearance rather than degradation. Targeted and rapid cellular internalization of CpG-STAT3ASO can improve therapeutic efficacy in vivo compared with the unconjugated STAT3ASO even at significantly lower dosing (36). The new types of chemical modification of ASO molecules, such as locked nucleic acids, are likely to further improve the potency, while reducing potential toxicities of these ONTs. Both CpG ODNs and STAT3ASO molecules were well tolerated by patients, when tested as single agents in clinical trials (36, 67). The most common adverse effects for both reagents were flu-like symptoms likely related to production of interferons and proinflammatory cytokines. In addition, the whole class of PS-modified ONTs can potentially trigger platelet activation and thrombocytopenia at high dosing. In case of CpG-STAT3ASO, such risks are mitigated by the significantly lower effective dosing and reduced extent of PS modifications compared with unconjugated and fully phosphorothioated ASOs. The preclinical studies to fully assess safety and pharmacokinetic/pharmacodynamic properties of CpG-STAT3ASO are ongoing. We believe that STAT3 inhibition/TLR9 stimulation targeted to myeloid cells using a single-oligonucleotide agent can provide broadly applicable, yet effective and safe strategy to overcome resistance of metastatic prostate cancers to immunotherapy.

R.A. DePinho has ownership interest (including patents) in Tvardi. S.K. Pal is a consultant/advisory board member for Bristol-Myers Squibb, Eisai, Genentech, Ipsen, and Pfizer. M. Kortylewski has a patent on STAT3 inhibitors and uses thereof, and a patent on methods and compositions for the treatment of cancers and other diseases. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: D. Moreira, S.K. Pal, M. Kortylewski

Development of methodology: D. Moreira, T. Adamus, P. Swiderski, S.K. Pal, M. Kortylewski

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Moreira, T. Adamus, Y-L. Su, Z. Zhang, R.A. DePinho, S.K. Pal, M. Kortylewski

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Moreira, T. Adamus, S.K. Pal, M. Kortylewski

Writing, review, and/or revision of the manuscript: D. Moreira, S.V. White, P. Swiderski, R.A. DePinho, S.K. Pal, M. Kortylewski

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Moreira, X. Zhao, Y-L. Su, P. Swiderski, X. Lu, S.K. Pal, M. Kortylewski

Study supervision: S.K. Pal, M. Kortylewski

The authors would like to thank Dr. S. Vonderfecht (Veterinary Pathology) for his assistance and acknowledge the dedication of staff members at the Analytical Cytometry, Analytical Pharmacology, Light Microscopy, Pathology Cores, and Animal Resources Center (City of Hope).

This work was supported by the Department of Defense (Prostate Cancer Research Program) grant number W81XWH-16-1-0499, the Prostate Cancer Foundation, the Israel Cancer Research Fund “Jacki and Bruce Barron Cancer Research Scholars Program” (M. Kortylewski), and the NCI of the NIH under grant number P30CA033572 (City of Hope).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.