Pathogen Molecular Pattern Receptor Agonists: Treating Cancer by Mimicking Infection

The interplay between cancer and the immune system is a double edged sword; the inflammation that recruits and activates intratumoral immune cells can either eliminate cancer cells or drive tumor progression in a context-dependent manner (1). Pattern recognition receptors (PRR) are a key family of proteins involved in the inflammatory response. They are expressed on a wide variety of innate and adaptive immune cells, as well as tumor cells, and recognize both foreign pathogen-associated molecular patterns (PAMP) and self-derived damage-associated molecular patterns (DAMP) resulting from injury or cell death (1–3). There are five families of PRRs: toll-like receptors (TLR), nucleotide-binding oligomerization domain (NOD)-like receptors (NLR), C-type lectin receptors (CLR), RIG-I–like receptors (RLR), and cytosolic DNA sensors (CDS; ref. 2). Each PRR family possesses distinct immunomodulatory properties, making them attractive immunotherapeutic targets. Here, we discuss PRR mechanisms and clinical implications to provide a detailed overview of the role of PRRs in immuno-oncology.

TLRs are the most widely studied PRR family, acknowledged in the 2011 Nobel Prize awarded to Drs. Steinman, Beutler, and Hoffman. There have been 10 TLRs identified in humans and 13 in mice (4); here we focus on the former. Structurally, TLRs are type I transmembrane proteins characterized by a ligand-binding N terminal ectodomain containing leucine-rich repeats, a single transmembrane domain, and a cytosolic Toll/IL1R homology domain responsible for signal transduction (2). TLRs 1, 2, and 4–6 are located on the cell surface and recognize bacterial membrane components such as lipids, proteins, lipoproteins (Fig. 1; refs. 2, 3), as well as several self-molecules, including extracellular matrix components, HSPs, and nuclear high-mobility group box 1 (HMGB1), often released as DAMPs from apoptotic or necrotic cells (1, 3).

Intracellular TLRs 3 and 7–9 are located within endosomes, and recognize viral and bacterial nucleic acids resulting from microbial replication or degradation upon entry into the cell (Fig. 1; refs. 1–3). Their localization normally prevents intracellular TLRs from binding self-nucleic acids. However, breakdown of this spatial separation may trigger autoimmune disease through recognition of self-nucleic acids (3, 5). Although TLR10 exists in humans, it has been difficult to study as it is nonfunctional in mice. Recent work suggests it may serve as a negative regulator of TLR signaling (6, 7). TLRs dimerize upon binding their cognate ligand (Fig. 1), causing conformational changes that allow for the recruitment of adapter molecules (MyD88, TIRAP, TRIF, and TRAM), initiating signaling cascades that ultimately induce transcription of inflammatory mediators (2, 3, 8).

As TLRs are highly expressed on antigen-presenting cells (APC), targeting TLRs can activate APCs and trigger adaptive immune responses; intratumorally, this may shift a tolerogenic tumor microenvironment (TME) to become immunogenic. However, because TLR signaling triggers inflammatory and cell survival mechanisms, and certain tumors express TLRs, TLR activation could instead be tumorigenic in certain settings (1, 9). Both TLR7 and TLR8 signaling have been implicated in driving lung cancer cell survival and chemotherapy resistance mechanisms (10, 11). TLR4 signaling in breast cancer both enhances chemotherapeutic resistance and promotes angiogenesis and lymphatic metastasis (12, 13). Tumor cells may also secrete HSPs and extracellular matrix factors as DAMPs, stimulating an immunosuppressive program in tumor-associated macrophages (TAM) to promote angiogenesis and metastasis (14, 15). One recent study demonstrated that mice lacking TLR3/7/9 cleared implanted tumors through spontaneous induction of an adaptive antitumor response (16). Similar effects are observed with other PRR families; galectin-9 signaling through the CLR dectin-1 on TAMs is protumorigenic in mouse models of pancreatic ductal adenocarcinoma, although signaling through this receptor in other cancers may have the opposite effect (17). While several of these reports implicate DAMPs in tumor initiation and progression through chronic inflammation, other studies demonstrate that DAMPs released from dying tumor cells are the hallmark of immunogenic cell death, activating APCs in the TME to present tumor antigen (9, 18). Despite the nuanced role of PRR signaling in cancer, in many contexts, therapies targeting PRR pathways have the ability to overcome immunosuppression or drive a de novo antitumor response by activating APCs to enhance tumor antigen presentation (Fig. 2). For the remainder of this review, we will focus on clinical trials and preclinical studies utilizing PRRs in this setting.

One of the few FDA-approved TLR-targeting therapies in oncology is Bacillus Calmette-Guérin (BCG), a strain of Mycobacterium bovis initially developed as a tuberculosis vaccine. Used as a urogenital cancer therapeutic for over 35 years, a large body of work has dissected its mechanism, demonstrating that BCG triggers an immune response by activating TLRs 2, 4, and 9, and the NLR NOD2 (1, 19, 20). Several trials are combining BCG therapy with checkpoint blockade or have expanded BCG to other cancers, with varying degrees of success (21).

TLR3 is one of the most actively explored TLR targets, with 54 ongoing clinical trials using TLR3 agonists as single agents or in combination with other therapies to treat a broad list of malignancies (clinicaltrials.gov). TLR3 recognizes viral double-stranded RNA (dsRNA), and can be targeted using synthetic dsRNA analogs such as polyinosinic-polycytidylic acid (poly-IC; refs. 2, 3, 22). Poly-IC initially showed high toxicity and limited therapeutic benefit, but several poly-IC derivatives were subsequently created to improve efficacy (22). One such derivative is poly-ICLC, modified with poly-l-lysine and carboxymethylcellulose (Hiltonol, Oncovir) to increase stability in vivo, improving its interferon (IFN) response to levels similar to those seen with attenuated viral infections (23, 24). Another derivative, poly-IC₁₂U (rintatolimod/Ampligen, Hemispherx Biopharma), adds unpaired bases that reduce stability, effectively reducing toxicity while generating robust dendritic cell (DC)/T-cell responses (25). Preclinical data also suggests that rintatolimod recruits fewer regulatory T cells to the TME versus unmodified poly-IC, possibly by losing ability to bind cytosolic RLRs (26). Recently completed trials using poly-ICLC demonstrate its potent ability to induce adaptive immune responses against a range of solid and hematopoietic cancers. A phase I study evaluating peptide-pulsed DC vaccination in combination with poly-ICLC for pancreatic cancer showed promise with a 7.7 month median survival, an improvement over the 4.2–4.9 month survival seen with second-line chemotherapy in metastatic pancreatic cancer (27). A phase I/IIa trial in smoldering multiple myeloma recently demonstrated that peptide + poly-ICLC vaccination increased numbers of antigen-specific CD8 T-cells with an effector memory phenotype (28). Another study pinpointed TLR3⁺ DCs as key mediators of tumor antigen cross-presentation, where an in situ vaccination combining poly-ICLC, Flt3 ligand, and local irradiation induced both partial and complete responses in patients with non-Hodgkin lymphoma (29). A pilot study in patients with transplant-ineligible hepatocellular carcinoma showed survival benefit compared with historical controls using local tumor irradiation followed by intratumoral poly-ICLC administration (30). Similarly, studies treating patients with glioblastoma demonstrated impressive survival outcomes by combining poly-ICLC with irradiation and/or alkylating chemotherapy (31, 32). In the preclinical setting, next-generation DC vaccines are being explored, employing nanoparticles to selectively deliver tumor antigens + poly-IC to DCs in vivo, eliminating the need for ex vivo DC manipulation (33). In addition, several groups have also demonstrated that TLR3 activation can help overcome resistance to checkpoint blockade, leading to ongoing academic and pharma trials planning to accrue >400 patients studying combinations of poly-ICLC with PD-1, PD-L1, or CTLA-4 blockade to treat various cancers (e.g., NCT03121677 and NCT03633110).

TLR4 is canonically involved in the recognition of bacterial lipopolysaccharide (LPS), although it is also indirectly involved in viral infection by recognizing DAMPs, such as HMGB1, HSPs, and extracellular matrix components released from infected or dying cells (1, 3, 8). Ongoing clinical efforts with TLR4 ligands in cancer immunotherapy include the FDA-approved TLR4 agonist AS04 (GlaxoSmithKline), a monophosphoryl lipid A (MPLA) LPS derivative in alum. AS04 is used as an adjuvant in the (human papillomavirus) HPV-16/18 vaccine Cervarix, which in a landmark trial was shown to not only protect women from cervical cancer from the vaccine-inclusive strains, but was also cross-reactive against other oncogenic forms of HPV (34). Interestingly, two recent phase III trials of a MAGE-A3 vaccine with AS15 (GlaxoSmithKline), an adjuvant containing MPLA and a TLR9 agonist, both failed to improve patient survival, citing low CD8 T-cell responses in patients (35, 36). Another TLR4 agonist, a synthetic analog of glucopyranosyl lipid A engineered to decrease heterogeneity and minimize toxicity over natural lipid A formulated in a stable emulsion (GLA-SE/G100; Immune Design), showed promise in a phase I study of Merkel Cell Lymphoma, where 2 of 10 patients had durable sustained antitumor responses, whereas 2 others had complete responses (37). Similarly, encouraging clinical responses were seen in 26 patients with follicular lymphoma receiving intratumoral G100 and radiotherapy with or without PD-1 blockade, where >80% disease control rates were seen in both groups, and addition of anti-PD-1 benefited relapsed and chemo-refractory patients (38).

TLR5 is unique in that it does not recognize DAMPs as its only ligand is bacterial flagellin, making it a potentially useful immunotherapeutic target (39). Two TLR5 agonists in clinical development, entolimod and mobilan (Cleveland Bio Labs), have shown preclinical efficacy in several tumor models (39–41). Entolimod is a flagellin derivative engineered to reduce toxicity, currently being investigated in a phase II trial as a neoadjuvant therapy for colorectal cancer (NCT02715882; ref. 42). Mobilan is an adenovirus construct that upon infection induces co-expression of TLR5 and a secreted form of entolimod, creating an autocrine signaling loop and inflammatory signature in the TME (41). Mobilan has shown preclinical efficacy for prostate cancer, which expresses high levels of the adenovirus receptor necessary for its entry, and is now in a phase I/II trial (NCT02844699; ref. 41).

Both TLR7 and TLR8 are functional in humans, whereas mice only have functional TLR7. TLR7/8 recognizes single-stranded RNA from RNA viruses, although RNA from certain bacterial strains may also ligate these TLRs (8). TLR7/8 also recognizes purine analogs such as imidazoquinolines, as well as guanine derivatives and certain siRNA (3). Ligation of TLR7/8 triggers robust proinflammatory cytokine production, and is critical for the activation of plasmacytoid DC (pDC), a key source of type 1 IFNs (3, 8). The only FDA-approved TLR7/8 agonist is imiquimod (Aldara; 3M Pharmaceuticals), an imidazoquinoline topical agent for the treatment of basal-cell carcinoma (BCC) that both enhances local immune response and directly induces apoptosis in BCC cells (43). Imiquimod is being investigated in phase III studies as a treatment for gynecologic cancers with promising early results (44), and in dozens of phase I and II trials in various cancers, either alone or combination (clinicaltrials.gov). Other imidazoquinoline derivatives in the clinic include a topical gel formulation of resiquimod, a more potent imidazoquinoline investigated as an adjuvant to NY-ESO-1 vaccination for patients with melanoma that has been shown to induce NY-ESO-specific CD8 T-cell responses (45). DSP-0509 (Boston Biomedical), a TLR7/8 agonist formulated for intravenous delivery, has shown preclinical efficacy in several tumor models and is now being investigated in a phase I trial (NCT03416335; ref. 46). MEDI9197 (3M-052; Medimmune) is an imidazoquinoline formulated for intratumoral injection and optimal tumor retention to improve safety. Preliminary phase I results (NCT02556463) demonstrate intratumoral immune cell infiltration and low serum MEDI9197 levels, indicating effective retention in the TME (47). Similarly, NKTR-262 (Nektar Therapeutics) is a TLR7/8 agonist formulated for intratumoral retention to minimize systemic exposure (48) and has shown potent efficacy, where treatment of one tumor site led to clearance of untreated contralateral tumors in multiple preclinical models, an abscopal effect often considered the holy grail of intratumoral immunotherapy. These promising results led to a recently opened phase I/II study of NKTR-262 in combination with a CD122 agonistic antibody and checkpoint blockade (NCT03435640; ref. 49). A recent randomized study of platinum-based chemoimmunotherapy for head/neck cancers demonstrated no overall survival benefit by adding the TLR8 agonist motolimod (Array Biopharma/Celgene), although motolimod did improve survival in subsets of patients with HPV⁺ tumors (50).

TLR9 recognizes unmethylated 2′-deoxyribo(cytidine-phosphate-guanosine) (CpG) motifs, which occur more frequently in prokaryotic DNA. Similar to TLRs 7/8, TLR9 is highly expressed on pDCs, as well as on B cells, and is critical in the immune response to DNA viruses (3, 8). Synthetic CpG oligodeoxynucleotides (ODN) potently activate TLR9-expressing immune cells and have been divided into four classes: Class A, B, C, and P (51, 52). Class A ODNs contain palindromic phosphodiester CpG central sequences with phosphorothioate G rich ends, allowing tetrad formation, enhanced stability, endosomal uptake, and robust activation of pDC type 1 IFN responses. Class B ODNs are short, linear phosphorothioate backbone ssDNA strands, and potent activators of B and natural killer (NK) cells. Class C ODNs combine properties of class A and B, activating both B and NK cells, as well as type 1 IFN pDC responses. Class P ODNs feature multiple palindromic sequences and form multimeric structures, enhancing stability and immunostimulatory responses (52). The first ODN in human trials was CpG7909 (agatolimod/PF-3512676/ProMune; Pfizer), a class B ODN, which showed early promise both as an in situ vaccination and chemotherapy adjuvant (53–55). However, a phase III lung cancer trial of chemotherapy with or without this ODN concluded that CpG7909 increased adverse events without benefiting survival, curtailing its development. Two other phase III trials that investigated CpG7909 as part of a MAGE-A3 vaccination also failed to demonstrate clinical benefit (35, 36).

Despite failures with CpG7909, several CpG ODNs modified to enhance efficacy and safety are in development. CMP-001 (Checkmate Pharmaceuticals), a class A ODN packaged within a virus-like particle, potently activates intratumoral pDCs and overcomes resistance to checkpoint blockade; five trials with this ODN are ongoing for various solid tumors (clinicaltrials.gov; ref. 56). Tilsotolimod (IMO-2125; Idera Pharmaceuticals) is another TLR9 agonist that is being investigated in checkpoint blockade refractory patients. In patients who failed anti-PD-1 therapy, tilsotolimod combined with ipilimumab CTLA-4 blockade improved objective tumor responses over ipilimumab alone, and this combination has entered a phase III trial (NCT03445533; ref. 57). Lefitolimod (MGN1703, Mologen AG) is a novel class of ODN that lacks phosphorothionate backbone modifications and is instead “dumbbell shaped” to prevent degradation (51). Demonstrating favorable safety and clinical efficacy, lefitolimod has initiated a phase III trial in metastatic colorectal cancer (NCT02077868; refs. 58, 59). Another class C agonist, SD-101 (Dynavax), is being investigated in several trials, after showing both preclinical and clinical efficacy in melanoma and as an in situ vaccination for lymphoma, in combination with checkpoint blockade and agonistic antibodies for T-cell costimulation (55, 60–62). DV281, a TLR9 agonist formulated for inhalation, is being investigated as an adjuvant for PD-1 checkpoint blockade therapy in lung cancer (NCT03326752), where intratumoral injection of adjuvant is more challenging. Notably, another Dynavax TLR9 agonist tested in a large randomized trial did demonstrate superior immunogenicity (seroconversion) when combined with HBsAg as compared with standard HBV vaccination, leading to its FDA approval. Potentially, this immunostimulatory effect could portend success in cancer therapy as now shown with pathogen vaccines.

NLRs are intracellular PRRs that recognize a diverse set of ligands including bacterial and viral PAMPs, as well as DAMPs (reviewed in ref. 61; refs. 2, 3, 8, 63). Of several NLR families, the NLRC and NLRP families are the most well-studied (2, 63). NOD1 and NOD2 are prominent NLRC family members, which all contain N-terminal CARD domains. Similar to TLR2, NOD1 and NOD2 recognize components of the peptidoglycan bacterial cell wall, where NOD1 specifically recognizes gamma-d-glutamyl-meso-diaminopemelic acid and NOD2 recognizes muramyl dipeptide (Fig. 1; refs. 8, 63). NLRPs, NLRP3 being the most well characterized, form part of the inflammasome, which leads to production of proinflammatory IL-1β/IL-18 (Fig. 1). Besides several bacterial ligands, environmental pollutants such as asbestos and silica are known to initiate NLR inflammasomes (64).

Mifamurtide, a synthetic analog of muramyl tripeptide and NOD2 agonist, is approved in the European Union in combination with chemotherapy to treat osteosarcoma (4, 65). The TLR8 agonist, motolimod, is also a potent stimulator of the NLRP3 inflammasome, likely because of the molecule's lipophilic structure; however the specific mechanism is still under investigation (50, 66). In addition, particulate adjuvants such as alum and saponins, often used in cancer vaccine formulations including HPV and the previously mentioned MAGE-A3 vaccine studies (35, 36), are potent activators of the NLRP3 inflammasome, producing inflammatory cytokines to engender adaptive immune responses (67, 68).

CLRs are a large family of receptors that contain at least one carbohydrate recognition domain, recognizing mannose, fructose, and glucans present on pathogens (reviewed in ref. 67; refs. 2, 8, 69). Although classically associated with antifungal and mycobacterial immune responses, more recent evidence suggests CLRs are involved in sensing numerous pathogens including bacteria, viruses, and helminths, as well as DAMPs (69–71). CLRs are mainly expressed by DCs, although monocytes/macrophages, B cells, and neutrophils may also express CLRs. Most CLR family members are transmembrane receptors, although a few may be released as soluble proteins, such as mannose-binding-lectin (4, 69). Upon ligation, CLRs transduce signal by either associating with kinases and phosphatases directly, or by recruiting ITAM-containing co-receptors such as FcRγ (Fig. 1; refs. 8, 69). Signaling ultimately converges on MAPK and NF-κB, allowing CLRs to influence signaling cascades from other PRRs, tailoring the immune response against specific pathogens. In addition, many CLRs are internalized after activation, bringing their ligand cargo within the cell for degradation and subsequent antigen presentation, a critical process in activating the adaptive immune response (69).

Strategies targeting CLRs date back over two decades. Randomized studies with a mannan-MUC1 fusion protein targeting mannose receptor (MR), vaccinating patients with breast cancer after surgical resection, showed significant protection from recurrence, demonstrating the efficacy of CLR targeting and the importance of adaptive immunity in preventing recurrence (72). Anti-CLR antibodies have also been used to target CLR-expressing DCs. CDX-1307 (Celldex Therapeutics) is an MR-specific antibody fused to human chorionic gonadotropin beta-chain (HCG-β), commonly overexpressed in various cancers (73). Vaccinating with CDX-1307, GM-CSF, poly-ICLC, and/or resiquimod to mature DCs, most treated patients developed humoral response against HCG-β, and some developed T-cell responses. Combining CDX-1401 (Celldex Therapeutics), an anti-DEC-205 antibody fused to NY-ESO-1 antigen, with poly-ICLC and/or resiquimod, yielded humoral and T-cell responses in most patients that correlated with stable disease (74). In addition, several patients who progressed saw dramatic benefit with subsequent checkpoint blockade, warranting studies of this combination therapy. CDX-1401 is currently being investigated in gynecologic (NCT02166905) and hematologic (NCT03358719) malignancies. CMB305 (Immune Design) is a Sindbis virus engineered to use DC-SIGN as an attachment receptor, selectively infecting and expressing NY-ESO-1 protein in DCs for antigen presentation (75). CMB305 is being investigated in a phase III trial in synovial sarcoma (NCT03520959). Another therapy, Imprime PGG (Biothera), uses IV yeast-derived soluble β-glucan, a dectin-1 ligand, to sensitize patients and boost efficacy of targeted therapy and anti-PD-1 blockade, and is being investigated in several phase I and II studies (clinicaltrials.gov) with promising early results. Interestingly, Imprime PGG shows a moderate toxicity profile, with 22% of patients discontinuing treatment due to adverse events, possibly because of the route of administration and the drug's potent activation of the complement cascade (76).

While TLR3 is responsible for detecting viral dsRNA within endosomal compartments, RLRs retinoic acid–inducible gene I (RIG-I), melanoma differentiation–associated gene 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2) recognize cytosolic dsRNA (Fig. 1; reviewed in ref. 75; ref. 77). These sensors are critical in the host antiviral response and are expressed within most cell types, including cancer cells (4). Structurally, RLR family members contain a C-terminal RNA binding domain, a DExD/H central domain for ATP catalysis and activation, and an N-terminal CARD domain that interacts with the downstream effector molecule MAVS (also referred to as IPS-1) to convey signaling. Short dsRNA with 5′ triphosphate (5′-PPP) ends are preferentially recognized by RIG-I, whereas MDA5 recognizes longer dsRNA fragments, including poly-IC. While LGP2 also recognizes dsRNA, this family member cannot convey downstream signaling because it lacks a CARD domain, and is instead important in regulating RIG-I and MDA5 activation (8, 77). Stimulation of epithelial ovarian cancer cells with a RIG-I–specific agonist triggers type I IFN release and immunogenic apoptosis, which effectively matures DCs upon phagocytosis of these apoptotic cancer cells (78). In addition, activation of RIG-I using 5′-PPP RNA or MDA5 using poly-IC causes apoptosis in human melanoma cells both in vitro and in vivo, whereas adjacent nonmalignant cells are spared because of intact antiapoptotic BCL-XL signaling (79).

The multimodal action of RLRs in immune cell activation while simultaneously triggering immunogenic cell death in cancer cells makes this pathway a particularly attractive immunotherapeutic target. The successes of poly-ICLC in clinical trials can in part be attributed to its dual agonistic activity on TLR3 and MDA5. BO-112 (Bioncotech) is another poly-IC derivative that potently activates RLR signaling in addition to TLR3, and is currently in phase I trials with promising early results (NCT02828098; ref. 80). One synthetic RIG-I–specific ligand, RGT100/MK-4621 (Merck) has initiated human trials, after preclinical data demonstrated potent antitumor activity in melanoma and colon carcinoma models (81). A phase I/II trial in solid tumors began in 2017 (NCT03065023), yielding only stable diseases as best response with intratumoral therapy (82). Pharmacokinetic studies show intratumorally administered MK-4621 is well retained in the TME, helping minimize adverse events due to systemic toxicity. Numerous RLR agonists are currently in active preclinical development and will likely be seen in the clinic soon.

As cellular DNA is ordinarily restricted to the nucleus and mitochondria, aberrant cytosolic DNA arising from viral infection or cell damage triggers immunogenic signaling by activating ubiquitously expressed CDS. To date, several pathways for sensing cytosolic DNA have been described. DNA can be transcribed in the cytosol, generating a dsRNA molecule that can be recognized by RIG-I, triggering an inflammatory response to cytosolic DNA in an RLR-dependent fashion (8). In addition, cytosolic DNA can be recognized by absent in melanoma 2 (AIM2), prompting inflammasome assembly, resulting in IL-1β/IL-18 production (Fig. 1; ref. 8). Perhaps the most impactful CDS is the stimulator of interferon genes (STING) pathway. Knockout studies indicate that STING is critical for host type I IFN and NF-κB responses to synthetic and viral DNA, whereas STING deletion had no impact on AIM2-mediated IL-1β production and the TLR9 CpG DNA response (83). STING is also essential for a successful adaptive immune response to DNA vaccination. Several other CDSs such as DNA-dependent activator of IFN-regulatory factors (DAI) have been identified; however, deletion studies indicate they may serve redundant function (8, 83, 84). In the context of infection, STING is a key mediator of the immune response against intracellular bacteria, DNA viruses, and retroviruses; however, its ability to detect genomic DNA from dying tumor cells makes the STING pathway potentially important for antitumor immune responses.

The ability of the STING pathway to drive adaptive antitumor responses has generated significant interest, and recent studies suggest that efficacy of numerous DNA-damaging cancer therapies can in part be attributed to STING signaling. Chemotherapeutic agents cause DNA leakage into the cytosol, triggering a STING-dependent type I IFN response (Fig. 2; ref. 85). STING signaling is also required for successful activation of adaptive immunity and tumor clearance in response to both radiotherapy (86) and T-cell checkpoint blockade, as cGAS-STING signaling enhances DC-mediated T-cell priming. Administration of adjuvant cGAMP synergized with checkpoint blockade in vivo, presumably by increasing the tumor reactive T-cell pool (87). Such studies highlight the importance of STING and type I IFNs in DC cross-presentation of tumor antigens for antitumor T-cell priming.

Several STING-specific agonists recently entered clinical development. Originally investigated as a vascular disrupting agent, STING agonist DMXAA (ASA404/vadimesan, Antisoma/Novartis) showed preclinical efficacy, but failed in a pivotal phase III trial as a combination treatment with chemotherapy in non–small cell lung cancer (NSCLC; ref. 88). It was later shown to be ineffective in patients due to STING polymorphisms that prevent DMXAA binding (84, 89). MIW815 (ADU-S100, Aduro Biotech) is a cyclic dinucleotide human STING agonist currently in phase I trials in combination with PD-1 (NCT03172936) or CTLA-4 blockade (NCT02675439). MK-1454 (Merck), a similar cyclic dinucleotide agonist currently in a phase I trial in combination with PD-1 blockade (NCT03010176) has shown favorable safety profiles and an objective response rate of 20% across several cancer types, with a median depth of response of approximately 80% (90). In addition, tumor clearance mediated by antibody blockade of CD47, a classical “do not eat me” signal, is STING dependent, where enhanced phagocytosis resulting from CD47 blockade ultimately requires STING and type I IFN signaling to prime T cells and inhibit tumor growth (91). Blockade of CD47 is currently the focus of several clinical trials in both hematopoietic and solid tumors (clinicaltrials.gov).

Among the most rapidly evolving therapeutic approaches in immuno-oncology is the use of oncolytic viruses (OV). Either through the intrinsic permissiveness of tumor cells for unchecked replication (including viral replication) or by directly engineering the viral genome, OVs can selectively infect and kill tumor cells. Tumors are specifically susceptible to viral infection and replication, as many of the pathways required for oncogenesis can be coopted by OVs. While loss of tumor suppressors such as p53 and RB, activation of RAS and similar oncogenes, disruption of IFN signaling components, as well as a generally immunosuppressive TME all enable immune escape and promote tumor growth, these pathways concurrently promote OV infection, replication, and inhibit viral clearance, creating a permissive space for OV growth that is preferential to nontransformed tissue (92). OV infection in turn results in the immunogenic cell death of infected tumor cells, initiating de novo antitumor immune responses or boosting existing responses through mechanisms discussed in the above sections (Fig. 2). Cancer cells infected with the polio OV PVSRIPO (Istari Oncology) release DAMPs (HMGB1, HSP60/70/80) and PAMPs (viral dsRNA), activating DCs to drive a tumor-antigen–specific cytotoxic T-cell response (93). Similar to other intratumorally delivered PRR agonists, the innate–adaptive immune axis is critical for OV therapy, as the ability to induce systemic antitumor immunity is antigen restricted to the OV infected site. Using a Newcastle disease OV and contralateral B16 and MC38 tumors, Zamarin and colleagues demonstrate that OV injection of one tumor results in immunity only against that same tumor type (94). Talimogene laherparepvec or T-VEC (Amgen), a modified herpes virus expressing GM-CSF, was approved in 2015 for the treatment of late-stage metastatic melanoma, earning a place for OV therapy in the clinic. In a landmark phase III study, intratumoral injection of T-vec caused complete resolution in 47% of injected lesions, as well as 22% of noninjected visceral lesions, highlighting the power of OV therapy to induce systemic antitumor immunity (95). T-VEC has already been effectively combined with CTLA-4 blockade, where a randomized phase II study demonstrated an increase in objective response rates from 18% with CTLA-4 monotherapy to 39% in the combination group (96), and is being investigated aggressively, including combinations with PD-1 blockade (NCT02965716), with neoadjuvant chemotherapy (NCT02779855), and with preoperative radiotherapy (NCT02453191). Other promising OV platforms in late development include Pexa Vec (JX-594, SillaJen), a vaccinia virus also engineered to express GM-CSF, currently in a phase III trial for hepatocellular carcinoma in combination with the kinase inhibitor sorafenib (NCT02562755). A modified Coxsackie virus, CAVATAK (Viralytics), is currently in phase II trials for several indications (clinicaltrials.gov). Moving beyond GM-CSF as the genetic payload, Swedish biotech Lokon recently opened a trial of their lead candidate LOAd703, an adenovirus encoding the costimulatory ligands CD40L and 4-1BBL (NCT03225989). Upon infection, tumor and other cells in the TME begin to express costimulatory ligands, helping to activate NK effector cells and remodel the TME (97). OVs without a therapeutic payload, including Pelareorep (Reolysin, Oncolytics Biotech) and PVSRIPO, are in active clinical development as well. Recently published phase I data shows PVSRIPO increased 36 month overall survival to 21% in patients with recurrent glioblastoma, a major increase from the 4% survival seen in historical controls (98). Taken together, all of these different successful approaches with OVs substantiate the idea that induction of immunogenic tumor cell death in combination with PRR agonism can drive effective adaptive immune responses.

Pattern recognition receptors present potentially powerful weapons in the cancer immunotherapy armory. Their ability to modulate numerous aspects of the tumor microenvironment, from APCs and their cross-talk with T cells, to directly modulating cancer cells themselves, allow these pathways to shape and ultimately drive an antitumor immune response. As PRR agonists and oncolytic viruses trigger innate cells to activate adaptive immunity, combining these approaches with checkpoint blockade therapy effectively presses the gas pedal while cutting the brakes, unleashing the full potential of immune effector cells. PRR agonism could additionally reverse resistance in checkpoint refractory tumors (99), and synergize with standard-of-care therapies including chemotherapy (100) and anti-CD20 targeting against lymphoma (101); a variety of combinatorial approaches are being actively explored in clinical trials (Tables 1,Table 2,Table 3–4). Preclinical approaches focused on developing next-generation agonists are underway; one group recently developed an OV-packaged anti-CTLA4 antibody construct, effectively combining OV and checkpoint therapy into a single injection (102). Others have fused Resiquimod nanoparticles to PD-1–targeting antibodies, allowing PD-1⁺ T cells to selectively deliver the TLR7 agonist to the tumor, reshaping the TME to improve T-cell infiltration and disease control (103). These approaches highlight the immunomodulatory potency of PRRs, where their ability to overcome immunosuppression and drive adaptive immunity effectively enhances efficacy of concurrently administered therapies. With so many novel agonists being investigated preclinically and in the clinic, continued exploration and understanding of PRR pathways and targeting will help to shape treatment paradigms in immuno-oncology.

C. R. Flowers reports receiving other commercial research support from Abbvie, Acerta, BeiGene, Celgene, Gilead, Genentech/Roche, Janssen Pharmaceuticals, Millennium/Takeda, Pharmacyclics, and TG Therapeutics, and is a consultant/advisory board member for Abbvie, AstraZeneca, Bayer, BeiGene, Celgene, Denovo Biopharma, Genentech/Roche, Gilead, Karyopharm, Pharmacyclics/Janssen, and Spectrum. A. Marabelle reports receiving commercial research grants from Transgene, speakers bureau honoraria from Amgen and MSD, and is a consultant/advisory board member for AstraZeneca, Lytix Pharma, MSD, Bioncotech, and Oncovir. J.D. Brody reports receiving commercial research grants from Merck, Genentech, and Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other authors.