Integrative Development of a TLR8 Agonist for Ovarian Cancer Chemoimmunotherapy

Most chemotherapeutic agents are believed to work independently of the host's immune system and some even appear to exert immunosuppressive functions. However, select chemotherapeutic agents such as anthracyclines may promote tumor cell death through mechanisms that activate antitumor immune response (1), thereby converting the tumor microenvironment (TME) from an immunologic sanctuary to a setting characterized by immunogenic tumor cell death. This transition is associated with the surface expression of stress-associated markers such as calreticulin on injured cells, which induce active phagocytosis by myeloid dendritic cells (mDC) and promote antigen processing and cross-presentation (2). Furthermore, the release into the TME of high-motility group box 1 (HMGB1) protein, a danger-associated molecular pattern (DAMP) molecule, triggers the innate immune response (3, 4). Despite these immune-activating processes, tumor-infiltrating mDCs are likely to be suppressed by factors within the TME (5), and phagocytosis of necrotic cells may not be a sufficient stimulus for optimal cross-presentation of tumor antigens by tumor DCs to trigger protective cytotoxic T-cell responses.

Toll-like receptors (TLR) comprise a family of 13 surface or endosomal receptors that recognize ligands that are broadly shared by pathogens, collectively referred to as pathogen-associated molecular pattern (PAMP) molecules. PAMPs activate DCs and stimulate the release of cytokines and chemokines, thus facilitating the development of adaptive immune responses (6, 7). In humans, TLR8 is distinguished from other endosomally expressed TLRs by its unique appearance on mDCs, monocytes, and natural killer (NK) cells. The natural ligand of TLR8 is foreign single-stranded RNA; engagement triggers cellular activation and maturation, including the release of helper T (T_H)1-polarizing cytokines. On the basis of a small genetic deletion in the murine TLR8 gene (8), the activity of TLR8 agonists in mice is significantly attenuated compared with that in humans and other species. Consequently, the in vivo biologic activity of TLR8 agonists has not been extensively studied to date. Nonetheless, the specific expression of TLR8 within human myeloid cells offers unique opportunities for anticancer therapy.

We have developed a synthetic, small-molecule agonist of TLR8 (motolimod; previously identified as VTX-2337) comprising a 2-aminobenzazepine core (9). Motolimod is selective for TLR8 and potently stimulates CD14⁺ monocytes and HLA-DR⁺CD11c⁺ mDCs but not HLADR⁺CD11c⁻CD123⁺ plasmacytoid DCs (9). In mDCs, motolimod activated NF-κB and induced expression of IL12 and TNFα. A completed phase I clinical study has demonstrated the tolerability and biologic activity of motolimod monotherapy and revealed predictable pharmacokinetic and pharmacodynamic profiles (10).

Given the stimulatory effects of motolimod on mDCs and monocytes, we hypothesized that motolimod would be an optimal partner for immunomodulatory chemotherapies such as anthracyclines, as simultaneous activation of PAMP and DAMP pathways during tumor antigen exposure could produce positive interactions. To test this hypothesis, we chose pegylated liposomal doxorubicin (PLD) because of favorable pharmacokinetics and pharmacodynamics, limited hematopoietic toxicity, and preferential accumulation in tumors (11). We chose ovarian cancer because of the implicated role of immune mechanisms (12–14) and because PLD is standard of care in relapsed disease (15).

Because the activity of TLR8 agonists in mice varies from that in humans (16), we pursued an integrative pharmacologic approach to study the combination of PLD and motolimod in parallel models including in vitro human, in vivo non-human primate, and in vivo humanized mouse to acquire knowledge on the mechanism of action of the combination, which was then evaluated in a phase Ib clinical study in women with recurrent ovarian cancer. Motolimod plus PLD induced potent immune activation, which was similar to that observed in the NSG-HIS mouse and in healthy human volunteers, and was well tolerated without unexpected or synergistic toxicities.

Motolimod (VTX-2337) was supplied by VentiRx Pharmaceuticals, Inc. For preclinical studies, PLD (Ben Venue Laboratories Inc.) was purchased from the University of Pennsylvania Hospital (Philadelphia, PA) pharmacy. For the GOG-9925 study, PLD was supplied by each clinical site's pharmacy.

Five men and 5 women healthy volunteers enrolled in a randomized, open-label, phase I study of motolimod (Study VRXP-A105) received 2 doses of motolimod 2.5 mg/m² subcutaneously separated by 1 week (17). Plasma samples were collected predose and 6 hours after dosing and analyzed for mediator levels as described above. This clinical study was conducted in accordance with Good Clinical Practice guidelines and the ethical principles on the basis of the Declaration of Helsinki. Approval for study procedures was obtained from the institutional review boards of each study site, and all subjects provided written informed consent upon study enrollment.

Studies in cynomolgus monkeys were conducted at Charles River Laboratories, Preclinical Services, in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the NIH (Bethesda, MD). Study protocols were reviewed and approved by the Charles River Institutional Animal Care and Use Committee. Three male cynomolgus monkeys were administered a single dose of motolimod 3.6 mg/m². Plasma samples were collected predose and 6 hours postdose and analyzed as described above. For in vitro studies, 7 mL of peripheral blood was collected from 3 males and 3 females and added to TruCulture tubes (Myriad-RBM) containing 2 mL of culture media plus 300 nmol/L motolimod. Tubes were incubated for 24 hours; supernatants were stored at −35°C until analysis for mediator levels as described above.

We used NOD-scid IL2rγ^null (NSG) mice reconstituted with human immune system (NSG-HIS) to test the biologic effects of motolimod and the combination with PLD. To generate human ovarian tumors, we injected mice with OVCAR5 cells. Details of mouse experiments as well as flow cytometric analysis of human leukocytes harvested from NSG-HIS mice, and testing of reactive human tumor-infiltrating lymphocytes (TIL) from mouse tumors are provided in Supplementary Material.

Human and non-human primate cytokines in biologic samples (TruCulture supernatants and plasma samples) were measured using Luminex-based technology (HumanMAP panel; Myriad RBM). Plasma samples from CD-1 mice were pooled at each time point, and cytokines were measured using the Myriad-RBM RodentMAP panel.

OVCAR5 cells were treated with PLD (0.1, 1, and 10 μg/mL) and/or recombinant human TNFα (20 ng/mL). Cell viability was assessed by Annexin V staining (Pharmingen #559763). Protein extraction and Western blot analysis were performed as described elsewhere (18). Primary antibody against c-FLIP (Cell Signaling #3210 rabbit) and secondary antibody (BioRad 172-1019) were used at room temperature. Protein bands were visualized using ECL (Amersham #RPN2132) with X-films (Bioexpress # F-9023).

GOG-9925 was a phase Ib, open-label, dose-escalation study of motolimod in combination with either PLD or paclitaxel in subjects with advanced ovarian cancer. The trial utilized a traditional 3 + 3 dose-escalation design to identify the maximum tolerated dose (MTD) of each combination (19). Primary objectives were to assess safety and tolerability, secondary objectives were to evaluate pharmacokinetics and pharmacodynamics, and exploratory objectives included assessment of antitumor activity. The study was conducted at 4 Gynecologic Oncology Group centers in the United States (June 2011 to September 2013) in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki. Approval was obtained from each site's Institutional Review Board. All subjects provided written informed consent. For each subject in cohort 1, PLD (40 mg/m²) was administered on day 1, followed by 2.5 mg/m² motolimod on day 3; each subject was assessed over ≥24 hours for safety. In the absence of overt toxicity, the next subject in the cohort was enrolled. Enrollment of subsequent cohorts proceeded without such staggering. All enrolled subjects were assessed for dose-limiting toxicities (DLT), defined as treatment-related adverse events (AE) that occurred during the first cycle of treatment and met standard predefined toxicity criteria. DLTs and other toxicities were the basis for decisions regarding dose escalation, de-escalation, and cohort expansion. Details on eligibility criteria and study assessment are provided in Supplementary Material.

Details on statistical analyses are provided in Supplementary Material.

To screen for suitable preclinical models, we compared the response to motolimod in vitro of peripheral blood mononuclear cells (PBMC) obtained from human healthy volunteers, non-human primates, and wild-type CD1 mice. Confirming the reported phylogenetic differences in TLR8, leukocytes from wild-type mice were minimally responsive to motolimod relative to non-human primate and human leukocytes (Supplementary Fig. S1). We thus sought an immunocompetent murine model that would allow us to explore the therapeutic potential of motolimod for tumor immunotherapy in vivo at clinically relevant doses. We selected the NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NOD-scid-IL2rγ^null NSG) mouse which, when reconstituted with human donor hematopoietic stem cells (HSC), develops all lineages of the human immune system (NSG-HIS; ref. 20). NSG mice were given 2.75-Gy whole-body irradiation and then intravenously injected with about 1 × 10⁵ human donor cord blood CD34⁺ cells. Consistent with published data (21), a high level of human hematopoietic cell engraftment was seen in NSG-HIS mice within 12 weeks, giving rise to human immune T- and B-lymphocytes as well as monocytes and dendritic cells repopulating peripheral blood, spleen, and lymph nodes (Supplementary Fig. S2).

We next tested whether human immune cells in NSG-HIS mice were sensitive to the effects of motolimod. Compared with untreated controls, mice treated with motolimod at 1.5 or 15 mg/m² showed a dose-dependent increase in plasma levels of human cytokines and chemokines, including IL6, IL12p70, TNFα, MCP-1, and MIP-1β 6 hours following treatment that declined to baseline by 24 hours (Fig. 1A). Although no individual cytokines were significantly increased (due to the small sample size), a mixed-effect linear model combining all analytes showed a significant treatment effect at the highest dose (P = 0.01). Consistent with activation of human monocytes in vivo, motolimod upregulated the co-stimulatory markers CD86 in human CD14⁺ and CD11c⁺ cells harvested from NSG-HIS spleens (Fig. 1B and C).

To further assess the suitability of the NSG-HIS mouse model, we conducted comparative pharmacodynamic studies assessing the in vivo plasma biomarker response to motolimod in NSG-HIS mice, non-human primates and healthy human volunteers. Doses that were proven to be safe and biologically active for each species were used (10, 17). A dose-dependent increase in cytokines and chemokines indicative of TLR8-induced monocyte activation—including IL6, MCP-1, and MIP-1β—were observed in all species (Fig. 2). Collectively, these data demonstrate that NSG-HIS mice exhibit sensitivity to TLR8 stimulation similar to that of non-human primates and humans.

Given the immunomodulatory effects of doxorubicin (22, 23) and prior evidence of positive interactions with innate immunity activation (24), we next tested whether concomitant administration of PLD would attenuate motolimod-induced TLR8 activation in vivo. NSG-HIS mice were administered intraperitoneally PLD once followed by subcutaneous motolimod on the same day or 2 to 5 days later. Whether a concurrent or sequential dose schedule was utilized, the acute response in plasma cytokines and chemokines and in human monocytes that was previously observed with motolimod monotherapy was not significantly attenuated by the addition of PLD (Fig. 3A and Supplementary Fig. S3). Given the slow release of doxorubicin in the TME and to maximize the timing of drug interactions, we adopted a staggered schedule for all subsequent murine experiments, with administration of PLD (50 mg/m²) on day 1, followed by motolimod at either 1.5 or 15 mg/m² on days 5, 7, and 9 of each 14-day cycle. This combination was administered for 3 cycles. In pilot experiments, both dose levels of motolimod were biologically active, eliciting cytokine and chemokine responses comparable to those shown in Figs. 1 and 2, with no acute toxicities. The 1.5 mg/m² dose level was selected for use in our subsequent studies, as it more closely approximated the relevant human dose levels identified in a previous phase I study of motolimod (10).

The potential antitumor effects of PLD plus motolimod in ovarian cancer were next evaluated using NSG-HIS mice bearing HLA-matched human ovarian tumor/human bone marrow chimeric xenografts. NSG-HIS mice reconstituted with CD34⁺ cord blood cells from HLA-A2⁺ donor neonates (NSG-HIS-A2) were inoculated with subcutaneous HLA-A2⁺ OVCAR5 tumors. Tumor grafts were accepted at a rate >95% and treatment began when tumors were approximately 50 mm³ (∼day 28). As expected, mice treated with PLD alone showed reduced tumor growth compared with untreated animals. Motolimod alone did not significantly affect tumor growth relative to untreated mice. However, the combination of PLD and motolimod markedly reduced tumor growth compared with either agent alone, including 2 of 10 mice who experienced complete disappearance of their tumors (Fig. 3B).

To test whether the drugs mobilized immunity against tumors, we examined tumors for overall human CD45⁺ leukocyte infiltration. Rare human CD45⁺ cells were detected in OVCAR5 tumors of untreated mice by immunohistochemical (IHC) analysis, indicating a baseline low-level tumor inflammation. Mice treated with either PLD or motolimod monotherapy or with the combination exhibited a marked increase in CD45⁺ cells (Fig. 3C). Consistent with activation of human tumor macrophages and DCs, there was an increase in the frequency of CD40-expressing human CD11b⁺ and CD11c⁺ cells in tumors following treatment with either drug alone or the combination relative to controls (Fig. 3D, top). An increase in human CD8⁺ TILs was also seen in mice receiving motolimod alone or motolimod in combination with PLD, whereas activation of CD8⁺ T lymphocytes, judged by the upregulation of CD69, was noted following treatment with motolimod in combination with PLD (Fig. 3D, bottom). Thus, the addition of PLD did not attenuate the effect of motolimod, and together the two drugs activated tumor-infiltrating monocytes and T cells, which was associated with significant tumor control.

Activated macrophages secrete high amounts of TNFα, which can induce apoptosis in susceptible cells expressing the TNF receptor 1 (TNFR1; ref. 25). We confirmed by Western blotting that OVCAR5 cells express TNFR1 (Supplementary Fig. S4). Given the potent activation of tumor macrophages by PLD plus motolimod, we hypothesized that the therapeutic interaction could be partly explained by PLD rendering tumor cells susceptible to death induced by TNFα. In agreement with prior observations (26), OVCAR5 cells were initially resistant to TNFα even at high levels (Fig. 4A). However, cells pre-exposed to PLD became sensitive to TNFα killing in vitro. Consistent with their increased susceptibility to TNFα-induced cell death, and in agreement with prior evidence (27), treatment of OVCAR5 cells with PLD significantly reduced cellular FADD-like IL1β-converting enzyme-inhibitory protein (c-FLIP), an inhibitor of procaspase-8 recruitment, which prevents apoptosis induced by TNF family death ligands (Fig. 4B; ref. 28).

In view of the observed increase in activated human CD8⁺ TILs following PLD plus motolimod, we also asked whether TILs were in part responsible for the tumor growth restriction in this model. Human TILs were expanded ex vivo from each treatment group using IL2 and tested against OVCAR5 cells, using TILs from the same human donor in each experiment. Consistent with the significant CD69 upregulation seen in TILs following combination treatment, TILs from mice treated with PLD plus motolimod exhibited high levels of cytotoxicity against OVCAR cells ex vivo, whereas TILs from mice treated with monotherapy exhibited modest cytotoxicity by chromium release assay ex vivo (Supplementary Fig. S5). There was no cytotoxicity from TILs expanded from control mice. These results were consistent among different human CD34⁺ cell donors. To further test the contribution of TILs to restrict tumor growth, TILs were harvested from tumors at the end of treatment from each experimental group and from control mice, expanded ex vivo using IL2, and adoptively transferred to treatment-naïve NSG-HIS-A2 mice bearing established OVCAR5 tumors. TILs from animals treated with PLD plus motolimod conferred a highly significant antitumor effect (Fig. 4C), suggesting that the increased efficacy of the PLD plus motolimod combination was, in part, also due to the cytolytic activity of TILs.

On the basis of the above preclinical data, a phase Ib study of PLD plus motolimod was initiated in women with recurrent or persistent ovarian cancer following platinum-based chemotherapy (Study GOG-9925). A standard dose-escalation design was used to evaluate the combination of 40 mg/m² PLD with 3 dose levels of motolimod (2.5, 3.0, and 3.5 mg/m²), which were selected on the basis of the results of a previous phase I dose-seeking study of single-agent motolimod (doses of 0.1 to 3.9 mg/m²) in adults with advanced cancer (10). PLD was administered intravenously on day 1 of a 28-day cycle, followed by subcutaneous motolimod on days 3, 10, and 17. Cycles of therapy were repeated until disease progression or unacceptable toxicity.

Following enrollment of the first and second cohorts [PLD + 2.5 mg/m² motolimod (n = 3) and PLD + 3.0 mg/m² motolimod (n = 3), respectively], study accrual was suspended because of a global shortage in the commercial supply of PLD. After a 6-month disruption, PLD availability returned and enrollment proceeded [cohort 3; PLD + 3.5 mg/m² motolimod (n = 7)]. During the interim absence of PLD, there was a need to identify alternate combinations wherein motolimod could augment the activity of existing chemotherapeutics. Because paclitaxel is an alternate standard-of-care chemotherapy for this patient population, the protocol was amended to include a second treatment arm evaluating motolimod plus paclitaxel. In this regimen, a single-dose level of motolimod (3.0 mg/m²) was combined with paclitaxel (80 mg/m²). Both agents were administered on days 1, 8, and 15 of a 28-day cycle. Cycles were repeated until disease progression or unacceptable toxicity.

Twenty subjects were evaluated on Study GOG-9925: 13 treated with PLD plus motolimod and 7 treated with paclitaxel plus motolimod. The majority of participants (n = 16, 80%) were Caucasian and aged between 50 and 69 years (80%) (Supplementary Table S1).

Motolimod in combination with PLD or paclitaxel was generally safe and well tolerated. There were no DLTs and no serious, unexpected drug-related AEs. There was no evidence of toxicity interactions between motolimod and PLD or motolimod and paclitaxel. No significant drug-related gastrointestinal, neurologic, or cardiac toxicities were observed. Clinical tolerability of motolimod was better at the 2.5 and 3.0 mg/m² dose levels compared with 3.5 mg/m². The most common AEs on each treatment regimen are provided in Table 1. Eleven of the 20 subjects experienced at least 1 AE ≥ grade III in severity; AEs in 7 of these subjects were considered related to combination therapy. Treatment-related events ≥ grade III observed in subjects receiving PLD plus motolimod included neutropenia (n = 3); anemia (n = 2); mucositis oral (n = 2); lymphopenia, palmar–plantar erythrodysesthesia syndrome, urinary tract infection, and leukopenia (n = 1 each); and chills, fever, and vomiting (all in one subject). All of these events were attributed to PLD except for the chills, fever, and vomiting. Treatment-related events ≥ grade III observed in subjects receiving paclitaxel plus motolimod included bladder infection, fatigue, febrile neutropenia, neutropenia, and leukopenia (all reported in 1 subject and attributed to paclitaxel).

Standard pharmacokinetic evaluations of each drug were performed to rule out any alternations in the metabolism and elimination due to drug–drug interactions. In both regimens, pharmacokinetic parameters associated with motolimod were within the range of previously reported values (Supplementary Fig. S6A). Exposure to PLD was consistent with historical data (Supplementary Fig. S6B). Plasma levels of paclitaxel were also within the expected range.

The pharmacodynamic effect of motolimod in combination with PLD or paclitaxel was evaluated in ovarian cancer patients by quantitative assessment of more than 90 cytokines, chemokines, and other immune inflammatory mediators. These data confirmed that motolimod was biologically active when combined with either PLD or paclitaxel, with the induction of multiple inflammatory biomarkers indicative of TLR8 stimulation and immune activation. Motolimod was biologically active at all 3 dose levels tested (Fig. 5A). Specifically, motolimod plus PLD increased plasma levels of G-CSF, IL6, MCP-1, and MIP-1β at one or more of the motolimod dose levels. Importantly, the magnitude of biomarker response was maintained over multiple treatments over 2 cycles, with no evidence of augmentation or attenuation (Fig. 5B). Results with the paclitaxel regimen were similar (Supplementary Fig. S7). Taken together, these results indicated that, as seen in the NSG-HIS mouse, the combination of cytotoxic chemotherapy plus motolimod did not attenuate the immunostimulatory effects of motolimod in humans.

Next, we conducted comparative pharmacodynamic assessments in women enrolled in the GOG-9925 study and in NSG-HIS mice bearing ovarian tumor xenografts treated with PLD plus motolimod. In both clinical study participants and NSG-HIS mice, plasma levels of IL6, MCP-1, and MIP-1β increased in a dose-related manner in response to motolimod (Fig. 5C). The magnitude of the pharmacodynamic effects seen in NSG-HIS mice dosed with 15 mg/m² motolimod plus PLD was comparable with the levels seen in study participants who received 3.5 mg/m² motolimod plus PLD.

The antitumor activity of the PLD/motolimod combination was assessed as an exploratory objective of the GOG-9925 study. At the time of study entry, subjects had evidence of recurrent disease either by biochemical criteria (elevated CA-125) or by imaging (measurable or nonmeasurable disease on CT or MRI). Of the 13 subjects treated with PLD plus motolimod, one subject with nonmeasurable disease treated with 2.5 mg/m² of motolimod achieved a complete response (CR; Fig. 5D). Interestingly, this subject received PLD plus 2.5 mg/m² motolimod for 2 cycles on study, with stable disease (SD) and interval improvement on interval imaging. Because of the shortage of PLD, she was unable to receive additional study drugs and elected to forego additional anticancer therapy. A CA125 level within normal range by cycle 4 and a repeat PET/CT confirming ongoing stabilization of disease 5 months later in the absence of therapy suggest a durable benefit from the 2 cycles of PLD/motolimod. She subsequently received 4 additional cycles of PLD plus motolimod (6 cycles total) and achieved a CR, then elected to discontinue study treatment. Post-study follow-up showed the CR was maintained for 10 weeks without additional therapy. (Fig. 5D). Another subject enrolled on the basis of biochemical evidence of disease (see Materials and Methods) and treated at the 3.0 mg/m² dose level achieved a biochemical CR after 3 cycles. She continued treatment for 3 additional cycles, for a total of 6 cycles, and then elected to go off-treatment. The CR endured for 16 additional weeks without anticancer therapy. Moreover, SD was seen in 8 subjects, including 1 subject with biochemical disease only; 3 additional subjects had progressive disease (PD; Fig. 5E). Overall, the response rate of subjects treated with PLD plus motolimod was 15% (n = 2) and the disease control rate (DCR) was 69% (n = 9; Supplementary Table S2). Of the 7 subjects who received paclitaxel plus motolimod, one achieved a partial response (PR), 2 experienced SD, and 4 had PD, with a DCR of 43% (n = 3; Supplementary Table S2). Notably, the 2 subjects who achieved CR with PLD plus motolimod and the subject who achieved a PR with paclitaxel and motolimod achieved their best response after receiving ≥3 cycles of therapy.

Cancer immunotherapy has resulted in major successes in a variety of solid tumors. Durable objective responses and long-term survival benefits have been achieved with T-cell–based adoptive therapy as well as using antibodies that activate T cells in vivo by blocking key lymphocyte-inhibitory receptors (29). Although the field of T-cell activation has significantly advanced, activation of innate immunity still remains an elusive yet key goal of immunotherapy.

TLR8 is an innate sensor that may facilitate the transition from innate to adaptive immune responses owing to its ability to activate monocytes and mDC. Because of phylogenetic differences in TLR8 and the relative lack of activity in mice, TLR8 agonists have not been evaluated as cancer therapies to date. We therefore sought to assess for the first time, the application of a novel small-molecule TLR8 agonist, motolimod, in combination with immunogenic chemotherapy using a humanized mouse model. Comparative pharmacodynamic studies confirmed that the human immune system in non–tumor-bearing NSG-HIS mice responds to motolimod in a similar manner to non-human primates and normal human volunteers, with clear dose-dependent activation of plasma and immune cell biomarkers. Having identified an active dose range that closely recapitulated the optimal dose identified in a phase 1 study in patients with cancer (10), we examined the combination of motolimod with PLD, a cytotoxic drug with favorable pharmacokinetic and immunomodulatory profiles. The NSG-HIS model proved a clinically relevant system for evaluating human immunotherapies such as motolimod and helped determine the pharmacodynamic effects and test the antitumor efficacy of the combination.

Clinical development of immunomodulatory agents that recognize human targets but not their murine counterparts is challenging. Although the NSG-HIS-A2 tumor model has limitations related to the partial HLA mismatch and multiple minor H mismatching between bone marrow and tumor allograft, it offers an acceptable immunocompetent surrogate mouse model with which to examine the effects of therapy on the TME. Despite the potential allorecognition, tumors engrafted successfully and grew in NSG-HIS-A2 mice, indicating local tolerance mechanisms enacted by the tumor. These tolerance mechanisms were effectively disrupted by the combination of PLD plus motolimod, which resulted in significant activation of tumor-infiltrating CD11b⁺ monocytes and CD11c⁺ DCs, which in turn were associated with activation of alloreactive CD8⁺ T cells. A concomitant elevation of human T_H1-polarizing and other inflammatory cytokines in the plasma documented effective systemic activation of immune cells with the PLD/motolimod combination, which preserved the potency of single-agent motolimod therapy by eliciting IL12 and TNFα responses. Additional distinct favorable features of the combination included enhanced IFNγ and lower levels of IL10 relative to motolimod monotherapy.

Activated TILs were, in part, responsible for the observed therapeutic effect of the PLD/motolimod combination in this mouse model, which shows that PLD plus motolimod significantly reprogrammed the TME through activation of monocytes to overcome local T-cell tolerance. It is also possible that tumor restriction was mediated, in part, by activated macrophages. Our in vitro experiments demonstrated a synergistic interaction between TNFα, a cytokine released by activated macrophages (and effector T cells), and PLD. Pretreatment of tumor cells with PLD was required to induce susceptibility to TNFα-mediated death, which was explained by downregulation of c-FLIP induced in tumor cells stressed by PLD. An additional interaction could entail the upregulation of the proapoptotic molecule Bfk by both TNFα and cytotoxic chemotherapy (30). These results prove the principle of a positive interaction of the combination due to independent effects of PLD and motolimod on tumor cells and immune effector cells respectively. It is likely that in tumors TNFα is derived from both activated tumor monocytes as well as activated effector TILs. Additional and complementary positive immunomodulatory mechanisms may be envisioned in human tumors in vivo, including the induction of immunogenic cell death induced by PLD, which could further contribute to proper antigen presentation by DCs, the production of T-cell recruiting chemokines by activated tumor macrophages, and a possible reprogramming of additional immunosuppressive monocyte populations such as myeloid-derived suppressive cells (MDSC), which should be further investigated if the combination proves effective in the clinic.

On the basis of the promising preclinical results, we developed the GOG-9925 phase I clinical study, adopting a phased administration schedule with motolimod initiated 2 days after PLD in each cycle. This schedule was selected to maximize the PLD-associated antigen-shedding effects on the tumor. Despite the absence of DLTs, the clinical tolerability of motolimod was considered better at 3.0 mg/m² compared with 3.5 mg/m². As all dose levels were associated with robust pharmacodynamic responses, and evidence of antitumor activity was observed at the 2.5 and 3.0 mg/m² dose levels, 3.0 mg/m² was selected as the recommended phase II dose. Similar to observations in the mouse, PLD did not attenuate the immune signature of motolimod in patients. Similar biomarker responses were observed in ovarian cancer subjects as in the NSG-HIS mice, indicating the accuracy of pharmacodynamics predictions in the NSG-HIS mouse. Notable AEs including thrombocytopenia, palmar–plantar erythrodysesthesia syndrome, and peripheral sensory neuropathy were typically attributed to PLD and/or prior therapy rather than to motolimod and were generally grade I or II in severity. Furthermore, the incidence of these events was similar to those reported in other trials of PLD in this setting (31–33). Finally, the clinical benefit observed with PLD plus motolimod suggests that the combination may improve clinical outcomes. The time to best response following treatment with motolimod plus chemotherapy was consistent with reports of delayed clinical response that have been observed with other immunotherapies such as ipilimumab (34, 35).

Although the phase I safety data are encouraging, it is important to note that PLD shortage during the study likely limited the understanding of the clinical effect of the chemoimmunotherapy combination and may have tainted the toxicity endpoints, as PLD is known to have delayed toxicity as well as efficacy. On the basis of these preclinical and initial clinical data, a randomized, placebo-controlled trial comparing PLD to PLD plus motolimod in platinum-resistant ovarian cancer was initiated (GOG-3003, NCT01666444). This trial enrolled nearly 300 women; data for the primary endpoint of overall survival are expected before the end of 2016. In addition, the immune effects elicited by the combination suggest that PLD/motolimod chemoimmunotherapy could be effectively combined with approaches further enhancing T-cell function, such as checkpoint blockade immunotherapy. This hypothesis is being tested in a phase I/IIa study assessing the PLD/motolimod combination with an anti–PD-L1 neutralizing antibody in the same patient population (NCT02431559).

In summary, we followed an integrative pharmacologic approach that used traditional approaches, as well as a humanized mouse model to advance the clinical development of motolimod for cancer therapy. The NSG-HIS model mimicked human donors and non-human primates and proved useful in exploring combinations of motolimod with chemotherapy drugs, allowing identification of appropriate combination schedules and dissecting interactions at the level of the TME. Despite its limitations, this model offers important advantages including experimental throughput and cost containment relative to non-human primates. In addition, this model solves a major limitation in cancer drug development by allowing assessment of the effects of drug combinations on the human peripheral immune system and in the TME. Further optimization of mice reconstituted with a human immune system and bearing human tumor grafts can be envisioned, including the use of autologous tumor and peripheral HSCs to obviate current limitations from alloreactive graft models.

B.J. Monk reports receiving commercial research grants from Amgen, Array, Eli Lilly, Genentech, Janssen/Johnson & Johnson, Morphotek, Novartis, and TESARO, Inc.; reports receiving speakers' bureau honoraria from AstraZeneca, Janssen/Johnson & Johnson, Myriad, and Roche/Genentech; and is a consultant/advisory board member for Amgen, AstraZeneca, Bayer, Cerulean, Clovis, Gradalis, ImmunoGen, Insys, Merck, NuCana, Oxigene, Pfizer, Roche/Genentech, TESARO, Inc., and Vermillion. R.M. Hershberg holds ownership interest (including patents) in VentiRx Pharmaceuticals. G. Coukos reports receiving commercial research grants from Boehringer-Ingelheim and Celgene; reports receiving speakers' bureau honoraria from Genentech and Roche; and is a consultant/advisory board member for Genetech, Novartis, Roche, and Sanofi-Aventis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.J. Monk, A. Facciabene, W.E. Brady, C.A. Aghajanian, P.M. Fracasso, J.L. Walker, H.A. Lankes, K.L. Manjarrez, R.M. Hershberg, G. Coukos

Development of methodology: B.J. Monk, A. Facciabene, C.A. Aghajanian, P.M. Fracasso, J.L. Walker, H.A. Lankes, G.H. Danet-Desnoyers, R.M. Hershberg, G. Coukos

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.J. Monk, A. Facciabene, C.A. Aghajanian, J.L. Walker, H.A. Lankes, G.H. Danet-Desnoyers, K.M. Bell-McGuinn, C.K. McCourt, A. Malykhin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.J. Monk, W.E. Brady, C.A. Aghajanian, P.M. Fracasso, J.L. Walker, H.A. Lankes, K.L. Manjarrez, G.H. Danet-Desnoyers, A. Malykhin, R.M. Hershberg, G. Coukos

Writing, review, and/or revision of the manuscript: B.J. Monk, A. Facciabene, W.E. Brady, C.A. Aghajanian, P.M. Fracasso, J.L. Walker, H.A. Lankes, K.L. Manjarrez, G.H. Danet-Desnoyers, K.M. Bell-McGuinn, C.K. McCourt, R.M. Hershberg, G. Coukos

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.J. Monk, W.E. Brady, J.L. Walker, H.A. Lankes, K.L. Manjarrez, G.H. Danet-Desnoyers

Study supervision: B.J. Monk, W.E. Brady, C.A. Aghajanian, P.M. Fracasso, J.L. Walker, H.A. Lankes, G. Coukos

We thank Gregory N. Dietsch, PhD, for providing the analysis of pharmacokinetic and pharmacodynamic samples; Raphael Gottardo, PhD, for providing statistical analyses of the preclinical data; and Eilidh Williamson, PhD, for providing medical writing assistance.

This study was supported by National Cancer Institutegrants: GOG Tissue Bank (U10 CA27469, U24 CA114793, and NRG Oncology Operations grant number U10 CA180868) as well as NRG SDMCgrant U10 CA180822,) and the GOG Statistical and Data Center (CA37517). GOG-9925 was an NRG Oncology and Gynecologic Oncology Group study. Support was also provided by VentiRx Pharmaceuticals.

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.