Intraperitoneal Monocytes plus IFNs as a Novel Cellular Immunotherapy for Ovarian Cancer: Mechanistic Characterization and Results from a Phase I Clinical Trial

This article is featured in Highlights of This Issue, p. 297

Ovarian cancer is the leading cause of death due to gynecologic malignancy, and the fifth leading cause of cancer death in women (1). The peritoneal cavity is the earliest and primary site of metastatic spread in ovarian cancer as well as the most common site of relapse (2). Patients with peritoneal disease relapse have a poor prognosis, particularly once their disease becomes resistant or refractory to platinum-based chemotherapy. These observations led to trials of intraperitoneal (IP) chemotherapy, thereby focused to the primary site of tumor, which has resulted in an overall survival advantage in phase III clinical trials (3–6).

The efficacy of IP chemotherapy raises the possibility of IP administration of novel cellular and/or immune therapies. Because immune checkpoint therapy has limited efficacy in ovarian cancer (7), we took the alternative approach of pursuing a therapy exploiting the innate immune system. Monocytes are innate immune cells which, when exposed to certain cytokines, can differentiate into a tumoricidal state in preclinical studies (8–13). Work from our group and others has demonstrated that the combination of IFNα and IFNγ plus monocytes has potent synergistic antitumor activity in a xenograft model of ovarian cancer (11), and that this combination is capable of inducing differentiation of monocytes into tumoricidal M1 macrophages (9, 10). We hypothesized that intraperitoneal administration of the combination of both IFNs with autologous monocytes, a combination not previously tested, would lead to an antitumor response in patients with ovarian cancer, and that this combination would be safe and well-tolerated.

Here we characterize for the first time the mechanism of an innate immunotherapy using a patient's own IFN-activated monocytes to kill cancer cells and demonstrate that killing is dependent on cell contact and induction of TRAIL and death receptor (DR) 4/5-mediated apoptosis. We translated these results into a first-in-human trial in patients with platinum-resistant or platinum-refractory ovarian cancer. We demonstrated that the therapy is well tolerated, with evidence of clinical activity by both imaging and CA-125 responses. Exploratory analyses demonstrated induction of TRAIL expression in monocytes and upregulation of proinflammatory cytokines in blood. Long-term responders displayed different baseline immunologic characteristics from other patients, including lower level of T-regulatory cells and neutrophil-to-lymphocyte ratio at baseline. However, these same patients exhibited a marked increase in myeloid-derived suppressor cells (MDSC) at the time of clinical and/or tumor progression, suggesting an opportunity for improving this immunologic therapy platform with an additional agent targeting MDSCs.

OVCAR8, OVCAR5, and OVCAR3 ovarian serous epithelial cell lines were obtained from the NCI-60 repository (for OVCAR 8 and 5) and ATCC (for OVCAR 3), respectively. Both lines were re-validated by short tandem repeat profiling. Cells were cultured in complete medium RPMI1640 (cRPMI; Quality Biological) supplemented with 10% heat inactivated FBS (Quality Biological) without antibiotics in a humidified incubator with 5% CO₂ warmed at 37°C.

Human IFNα2a, was a gift of Hoffmann LaRoche, and IFNγ was purchased from Intermune Pharmaceutical Inc. All dilutions were performed in cRPMI. Propidium iodide solution was purchased from Sigma-Aldrich.

Peripheral blood was isolated by percutaneous venipuncture and collected in two 10 mL EDTA tubes (Becton Dickinson). Peripheral blood mononuclear cells (PBMC) were isolated as described previously (14). Blood was obtained from patients attending the Women's Malignancy Branch Ovarian Cancer Clinic under IRB approval through protocol 14-C-0056. De-identified age-matched control blood was obtained through the Department of Transfusion Medicine (DTM), Clinical Research Center, NIH, under protocol 99-CC-06168.

De-identified primary human monocytes from healthy donors were obtained from the DTM, Clinical Research Center, NIH, and cultured as described previously (15).

Monocytes were co-cultured at a ratio of 10:1 for 3 days with ovarian cancer cell lines and without IFNs as described previously (15).

Monocytes were cultured for 24 hours with or without IFNα2a and IFNγ. Monocytes were then washed of culture media and subsequently cultured with OVCAR8 cells at a ratio of 10:1 for an additional 48 hours on 0.4-μm pore size polycarbonate transwell inserts (Corning, #7910) or normal 10 cm dishes. Cells were diluted 1:10,000 and stained with Sytox green dye. Flow cytometry was then performed and percentage of live cells was calculated.

All CRISPR HDR-plasmid constructs were purchased from Santa Cruz Biotechnology, used for targeted deletion of specific genes as stated in the product manual.

siRNA to DR5 (Hs_TNFRSF10B_6), Catalog No sI03038665 (Qiagen) were validated for knockdown by Western blot analysis prior to use in biological assay. Sequence: 5′-CTGCAAATATGGACAGGACTA-3′. siRNAs showing greater than 80% knockdown were used.

siRNA-mediated knockdown of primary monocytes using reverse transfection was slightly modified from the original method (16). Primary human monocytes were isolated by counter elutriation and further isolated using CD14-positive selection beads (Miltenyi Biotech). TNFSF10B target sequences 5′-CCAAACAUACUUUCGAUUU-3′, 5′-GCAACUCCGUCAGCUCGUU-3′, 5′-CAAGUUAUCCUGACCCUAU-3′, 5′-GAGGAAUGGUGAACUGGUC-3′. Knockdown was confirmed by Western blot analysis. More complete details can be found in the Supplementary Files.

At indicated time points cells were pelleted by centrifugation and snap frozen and stored at −80°C. Cell pellets were lysed with M-PER lysis buffer (Thermo Fisher Scientific) with 1× protease and phosphatase inhibitors (Thermo Fisher Scientific). Lysates were clarified by centrifugation at 20,000 RPM for 20 minutes at 4°C. Details, include product vender information and antibody clone number can be found in the Supplementary Information.

5 × 10⁵ cells were added to 5 mL flow cytometry tubes, spun down at 1,000 RPM (epithelial cells) 1,250 RPM (monocytes) for 5 minutes at 4°C, and resuspended in 100 μL ice cold PBS. A Fc blocking step (eBioscience) was performed for staining of monocytes for 15 minutes on ice. Antibodies were added at the manufactures suggested concentration for 30 minutes on ice in the dark. Cytometry was performed according to eBioscience protocols on a Becton Dickinson FACSCalibur. Antibodies used are listed in Supplementary Materials.

Detailed materials and methods are in the Supplemental Materials.

Mouse experiments were approved by the NIH Animal Care and Use Committee. Briefly, 10 Foxn1nu athymic mice were injected intraperitoneally with 2 × 10⁶ OVCAR8 cells. Fourteen days later, mice were either given PBS or a single dose of intraperitoneal monocytes at a dose of 20 × 10⁶ cells/mouse. Mice were also given PBS or 1 μL of IFNα2a (20 μg/mL) and γ1b (20 μg/mL), for a final dose 20 ng/mouse. Mice were euthanized at first sign of illness.

Cells were harvested at the indicated time points and pelleted by spinning at 1,000 rpm at 4°C using a Beckman Coulter Allegra X-15R centrifuge. The supernatants were removed, and the pellets were washed in 1 mL of PBS (pH 7.4). The pellet was then resuspended in 350 μL of RLT buffer plus from the RNeasy Plus Kit (Qiagen). The RNA mixture was then stored at −80°C until RNA isolation. RNA isolation was performed using the RNeasy Plus Kit according to the manufacturer's instructions. To remove contaminating DNA, two DNase steps were performed. DNA was first removed using the genomic DNA (gDNA) eliminator column (Qiagen) prior to RNA binding to the RNeasy column. A second, on-column DNase step was performed using 10 μL of DNase 1 with 70 μL of buffer RDD (Qiagen) and 2 μL of Turbo DNase (Thermo Fisher Scientific) per column. The columns were incubated at room temperature for 30 minutes. The RNA samples were quantitated using the Nanodrop 1000 (Thermo Fisher Scientific).

The Ambion Illumina TotalPrep RNA Amplification Kit (Ambion, Inc.) was used similarly to manufacturer instructions and as described previously (17) to prepare biotinylated cRNA from isolated total RNA. Input was normalized to 300 to 500 ng per sample and yields ranged from 10 to 70 μg. Briefly, the kits employ reverse transcription to produce high yields of first-strand cDNA using a reverse transcriptase (RT) that catalyzes the synthesis of near full-length cDNA. The cDNA was subjected to second-strand synthesis to convert to a double-stranded DNA (dsDNA) template for transcription for in vitro transcription with T7 RNA polymerase. This method leverages DNA polymerase and RNase H to simultaneously degrade the RNA and synthesize the second-strand cDNA. Clean-up is performed to remove RNA, primers, enzymes, and salts, which could inhibit transcription. Next, biotinylated antisense RNA copies of each mRNA in a sample are generated. Labeled cRNA (0.75 μg) was then used for hybridization with Illumina HumanHT-12 v4.0 beachchips, scanned on HiScan-SQ imager.

Signal data were extracted from the image files with the gene expression module (v. 1.9.0) of the GenomeStudio software (v. 2011.1) from Illumina, Inc. Signal intensities were converted to log2 scale. Calculation of detection P values is described in the GenomeStudio Gene Expression Module User Guide. Data for array probes with insufficient signal (probes not having at least two arrays with detection P value < 0.1) were removed from the dataset. After dropping nonperforming probes, quantile normalization was applied across all arrays. A total of 35,680, length 50 nt probes had sufficient signal for ANOVA computation. The single probe with the most statistically significant result was taken to represent the gene's expression level. Expression estimates are Log2 of BeadChip signal intensity after quantile normalization with JMP/Genomics software (v 6.0; SAS Institute). Expression profiles from for (n = 3) technical replicates were averaged together before computing ANOVA across data from (n = 3) different donors. Mixed effects ANOVA was calculated using JMP/Genomics software with treatment condition (fixed) and donor ID (random) effects to test for gene expression differences. N = 3 monocyte donors per ex vivo treatment group. A P value of 0.05 was used for the statistical significance cutoff, after adjusting for the family-wise error rate (FWER) using Benjamini–Hochberg method to account for multiple testing. Additional requirement of log2 expression ratio greater than 1.0 (2-fold change of IFN-treated vs. control, positive or negative) was used to select differentially expressed genes for heatmap. The microarray data have been deposited in NCBI's Gene Expression Omnibus (18) and are accessible through GEO Series accession no. GSE190594 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE190594).

Student t test, one-way ANOVA, and two-way ANOVA with Bonferroni posttest analysis were performed were indicated. The number of samples chosen for each comparison was determined on the basis of past similar experiments or by performing pilot experiments to assess the expected magnitude of differences. This approach was duplicated from another publication (19). The number of experiments performed is indicated in the figure legends. Biological assays were performed in triplicate, with separate donors before statistical analysis was performed.

This was an open-label, single-center phase I study comprised of a conventional 3+3 dose escalation to determine the recommended phase II dose (RP2D). This clinical trial was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from each subject. As previously described, on day 0, patients underwent apheresis of peripheral blood mononuclear cells (PBMC), from which monocytes were purified by counter-flow elutriation, quantified, and cryopreserved (20). On day 1 of each 28-day cycle, 3 to 6 patients at each DL received infusion of pegylated IFNα2b (Sylatron, Merck; 25–250 mcg) and IFNγ1b (Actimmune, Horizon; 5–50 mcg), with or without their own autologous monocytes (75–750 × 10⁶ cells). Monocytes and IFN were combined in the infusion product for no more than 4 hours. No differences were noted in studies comparing viability and functionality of fresh monocytes to cryopreserved monocytes in the final product (20). Patients were not premedicated with nonsteroidal anti-inflammatory drugs to prevent temperature elevation. Doses of each cytokine corresponded to the following in international units (IU): 1.75 × 10⁶ IU or 1.5 × 10⁷ IU for Sylatron, 1 × 10⁵ IU or 1 × 10⁶ IU for Actimmune (see study schema and figures as well as previously published descriptions of monocyte isolation procedure (20) and trial design; ref. 21). Treatment was administered by either intraperitoneal catheter or surgically implanted intraperitoneal port. Monocytes that were not immediately needed were cryopreserved and used as need in later cycles; additional monocytes were collected in later cycles as needed. Patients continued to receive treatment until disease progression, intolerable toxicity, or withdrawal of consent. In addition to the above dose levels, an optional DL of 3b was available for patients at higher dose levels unwilling or unable to tolerate subsequent apheresis procedures, which was comprised of the highest levels of IFNα (250 mcg) and IFNγ (50 mcg), but without the addition of monocytes.

Toxicities were assessed and graded each cycle according to CTCAE 4.0 criteria (22). Dose-limiting toxicities (DLT) were determined during the first cycle. DLTs were defined as ≥Grade 3 fatigue lasting >7 days, ≥Grade 3 cytopenias lasting >7 days, ≥Grade 3 transaminitis lasting >7 days, ≥Grade 3 depression lasting >7 days, ≥Grade 3 abdominal pain lasting >7 days, or any other grade ≥3 toxicity lasting >72 hours despite maximal medical intervention.

The RP2D was defined as the dose at which 0 to 1 of 6 patients experienced a DLT, or the highest planned dose level. The RP2D was expanded to 6 patients to get additional information on the safety, tolerability, and efficacy of the combination. Response was determined every 8 weeks by CT imaging according to RECIST criteria v1.1 and/or CA-125 level according to Gynecological Cancer Intergroup (GCIG) criteria (23).

Before each cycle, patients underwent collection of peripheral blood for multiparameter flow cytometry analysis of immune cell populations including T-cell subsets, B-cell subsets, myeloid subsets (M1 and M2 macrophages, monocytes, dendritic cells), and NK cells, as well as molecules associated with checkpoint blockade (PD1, PD2, PDL1, PDL2, CTLA4). Patient samples were stored at −80°C and thawed for all analyses except for peripheral blood cell count and differential, which was performed on fresh plasma according to conventional clinical assays. Cell staining and flow cytometry was performed as described previously (24). For full list of markers and populations, please see the Supplementary Data.

Presence of anti-IFN antibodies in plasma was determined using a high-throughput particle-based assay (25), with healthy age-matched plasma as negative controls and known autoantibody-positive patient plasma as positive controls. Briefly, plasmas were diluted 1:100 in PBS and incubated for 30 minutes with differentially fluorescing magnetic beads that had been covalently coupled to 2.5 μg recombinant human IFNα (PBL Biomedical Laboratories; Catalog No. 11101–2) or IFNγ (R&D Systems; Catalog No. 285-IF-100/CF). Beads were washed then incubated with 1 μg/mL PE-labeled goat anti-human IgG (eBioscience; Catalog No. 12–4998) for an additional 30 minutes before being washed again and run in a multiplex assay on the BioPlex X200 instrument (BioRad).

Plasma cytokine levels were assessed at baseline, 3 to 5 days after administration of first dose of treatment, and directly before administration of second dose of treatment using the V-PLEX Proinflammatory Panel 1 (Meso Scale Diagnostics), which assays for levels of IFNγ, IL1β, IL2, IL4, IL6, IL8, IL10, IL12p70, IL13, and TNFα.

For RT-PCR, total RNA from the patient monocytes or monocyte plus IFN product was isolated with an miRNeasy Mini Kit (Qiagen). cDNA was synthesized using a high capacity cDNA Reverse Transcription Kit (Applied Biosystems) with oligo(dT) priming and amplification of gene-specific products was achieved with TaqMan probes (Applied Biosystems). The relative fold change in gene expression was calculated using the delta-delta Ct method (2^−ΔΔCt). Probes and target mRNAs were as follows: TNFSF10 - Hs00921974_m1, STAT1 - Hs01013996_m1, GAPDH - Hs99999905_m1.

Safety analyses included all patients who were enrolled on study. Efficacy analyses were descriptive and included patients who had either RECIST-quantifiable lesions or measurable CA-125 response as defined by GCIG criteria (23). For statistical analyses of correlative endpoints, t tests for paired or unpaired data were used as specified in the individual figure legends (significance level, P ≤ 0.05).

Raw data for this study were generated at the NIH. Derived data supporting the findings of this study are available from the corresponding author upon request. Microarray data generated in this study are publicly available at GEO (GSE190594) as described above.

Spheroid cultures from the human ovarian cancer cell line OVCAR5 were cultured with monocytes and IFNs, both separately and in combination, or treated with carboplatin-based chemotherapy drugs as a positive control for inducing cell death (Fig. 1A). Monocytes and IFNs alone resulted in significant cell death as compared with no treatment, but death did not exceed that induced by chemotherapy. The combination of monocytes and IFNs resulted in an approximately 80% reduction in viability compared with control, significantly greater than chemotherapy. Staining with hematoxylin and eosin (H&E) and CD68 IHC were then performed to identify monocytes/macrophages in the tumor spheroid cultures (Fig. 1B). The untreated spheroids are complex structures defined by cells with fused nuclei, large vacuole spaces, and extracellular matrix. The addition of monocytes to the culture resulted in migration of CD68-positive monocytes into the spheroids. IFN treatment by itself caused some disaggregation of the spheroid. Remarkably, the combination of both monocytes and IFNs resulted in extensive migration into the spheroids, destruction of most of the spheroid structures, and loss of viable CD68-negative tumor cells. Real-time single-photon confocal microscopy assay utilizing dye-labeled monocytes confirmed the migration of monocytes into the spheroids (Fig. 1C and Supplementary videos S1–S4). Cell tracking software used to quantify monocyte movement demonstrated that when treated with IFNs, monocytes exhibited decreased track length, displacement, and speed, consistent with a prolonged and close interaction between monocytes and the cancer cells (Supplementary Figs. S1A–S1D).

To determine whether direct cell contact is required for cell killing, we used a transwell assay in which monocytes were treated with IFNs for 24 hours, then washed, and then added to wells with OVCAR8 (permissive of direct cell contact) or separated by a 0.4-micron membrane (too small for cell contact; Fig. 1D). The 0.4-μm membrane prevented cell killing by the monocytes, demonstrating the importance of a cell contact-dependent mechanism. Importantly, without direct contact, the presence of the monocytes appeared to stimulate cancer cell growth by more than 2-fold, but when in direct contact, they triggered cell death, decreasing cancer cell viability by 35% alone and 60% with IFNs present in this experiment. The requirement for cell–cell contact for tumor cell killing suggested that expression of death-receptor ligands on the monocytes might be responsible for the observed tumor cell death by activation of the extrinsic apoptosis pathway. Consistent with this idea, co-culture of OVCAR3 cells with IFNs and monocytes led to activation of the extrinsic apoptosis pathway, showing an increase in cleavage of the initiator procaspase-8 and downstream “executioner” procaspases-3 and -7 (Fig. 1E–G; ref. 26). Conversely, an OVCAR3 cell line CRISPR-deleted for caspase-8 was largely immune to monocyte killing (Fig. 1H; Supplementary Fig. S1E). Taken together, these data implicate a mechanism by which monocytes activated by IFNs kill ovarian cancer cells by direct contact, triggering the extrinsic apoptosis pathway.

We next sought to identify a ligand expressed by activated monocytes that could mediate cell contact-dependent killing of ovarian cancer cells. Global gene expression analysis performed on monocytes treated with both IFNs versus either IFNα or IFNγ alone revealed that one of the most upregulated genes was TRAIL (Fig. 2A), a molecule expressed on activated monocytes as well as other immune cells, which is capable of inducing apoptosis through the extrinsic signaling pathway (see also Supplementary Table S1 for complete list of differentially expressed genes and Supplementary Table S2 for complete list of genes in heatmap; ref. 15). This is in line with previous reports that TRAIL is synergistically upregulated in cell lines by the combination of IFNα and IFNγ (27), at least in part due to direct transcriptional upregulation mediated by an IFN-stimulated response element in the TRAIL promoter (28). We cultured IFN-treated monocytes from healthy human donors with OVCAR3 ovarian cancer cells and confirmed this upregulation of TRAIL at the protein level using flow cytometry (Fig. 2B), and Western blotting of cell lysates (Fig. 2C).

To confirm that TRAIL is required for IFN-activated monocytes to induce cell death, we knocked down TRAIL in IFN-activated healthy donor monocytes using siRNA and measured killing of OVCAR3 cells compared with control (Fig. 2D). RNA interference was only partially effective in knocking down TRAIL in primary monocytes (Supplementary Fig. S2A). Even with partial knockdown, it was apparent that control-transfected monocytes could kill OVCAR3 cells at increasing IFN concentrations, whereas knockdown of TRAIL rescued OVCAR3 cell viability to a level not statistically different from that of IFN-treated cells. TRAIL has been shown to initiate apoptosis in OVCAR3 cells by binding to its receptors DR4 and DR5, and OVCAR3 cells are known to respond equally through both DR4 and DR5 (29). To measure the effect of each death receptor, we performed CRISPR-Cas9 deletion of DR4 and/or siRNA knockdown of DR5 in OVCAR3 cells (Fig. 2E; Supplementary Figs. S2B–S2E). DR4 knockout significantly decreased cytotoxicity compared with control OVCAR3 cells co-cultured with monocytes and IFNs. Although knockdown of DR5 resulted in cytotoxicity that was decreased but not statistically significant, the double DR4/DR5 depletion rescued cell death to the levels of OVCAR3 treated with IFNs only, suggesting a greater role for DR4-mediated signaling versus DR5.

We confirmed the clinical applicability of these results by testing the ability of IFN-activated monocytes derived from patients with ovarian cancer to kill OVCAR3 ovarian cancer cells (Fig. 2F). Monocytes were isolated from the peripheral blood of healthy donors or patients with ovarian cancer by elutriation, as described previously (20). Interestingly, IFN-activated monocytes from these heavily pretreated patients with ovarian cancer were not only effective at killing OVCAR3 cells, but we found that they were actually significantly more cytotoxic than those from healthy donors when used with the intermediate concentration of IFNs (P < 0.01). To demonstrate the ability of IFN-activated monocytes to kill tumors in vivo, nude mice injected intraperitoneally with OVCAR8 cells were treated 14 days later with a single intraperitoneal dose of monocytes, IFNs, or the combination (Fig. 2G). Given the requirement for direct cell contact between the IFN-activated monocytes for killing the tumor cells (Fig 1D), we chose IP injection of the monocytes over other routes of administration. The combination displayed a 17% increase in overall survival versus over monocytes alone (40.5 days vs. 33.5 days, P < 0.01), and 60% increase over IFNs alone (40.5 days vs. 26 days, P < 0.01). Together, the data presented above demonstrate that IFN activated monocytes kill tumors via direct cell contact through the TRAIL pathway and that they prolong survival in vivo.

The above data suggested the translational potential of IFN-activated monocytes as an antitumor therapy. The intraperitoneal route of administration was chosen because of our observation in Fig. 1D, which monocytes not in direct contact with cancer cells could potentially increase cancer growth. We hypothesized that intraperitoneal administration of the combination of dual IFNs and autologous monocytes, a combination not previously tested, would lead to an antitumor response in patients with ovarian cancer, and that this combination would be safe and well-tolerated.

To determine the safety and tolerability of this combination, we performed a phase I, dose escalation clinical trial in patients with ovarian cancer with platinum-resistant or platinum-refractory disease. The primary objective was to determine the safety of the treatment and the recommended phase II dose (RP2D), or maximum tolerated dose (MTD). Four dose levels were enrolled, alternately escalating monocytes and IFN doses (Fig. 3A). Patients underwent apheresis to collect mononuclear cells, and monocytes were enriched by elutriation (21). On day 1 of each 28-day cycle, patients received intraperitoneal infusion of IFNα (25 or 250 mg) and IFNγ (5 or 50 mg), with or without their autologous monocytes (75 or 750 × 10⁶ cells; Fig. 3A and B). Dose de-escalation was allowed if patients had completed the DLT period (1 cycle) and continued to show clinical benefit. Excess monocytes were cryopreserved for later use and additional monocytes were collected in later cycles as needed. Before each cycle, patients underwent collection of peripheral blood for multiparameter flow cytometry analysis of immune cell populations and cytokine profiling.

Between February 2017 and June 2019, a total of 18 patients were enrolled and received one or more doses of treatment (Table 1), including 6 total patients at the RP2D, dose level 4 (DL4), corresponding to 250 mcg IFNα, 50 mcg IFNγ, and 750 × 10⁶ monocytes. Patients had epithelial ovarian or fallopian tube cancer histologies, with high-grade serous being the most common histology (n = 12), and a variety of aggressive ovarian cancer morphologies represented. All patients had platinum-resistant disease. This was a heavily pretreated population, with a median of five prior therapies received, ranging from 1 to 13 lines of prior therapy. In all, 83% of patients had previously received bevacizumab, including all 6 patients at the RP2D; 39% of patients had received treatment with a PARP inhibitor, and 17% had received treatment with an immune checkpoint inhibitor on a previous clinical trial.

The combination treatment was generally well-tolerated. At the time of data cut-off in June 2019, all 18 patients had discontinued treatment. One patient had a peritoneal catheter, which became dislodged after one cycle and was unable to be replaced, necessitating her removal from the study. An additional patient elected to stop receiving treatment after one cycle to pursue alternative therapy. Two of 16 patients initially assigned to dose levels with monocytes had difficulties tolerating apheresis, due to a combination of apheresis catheter discomfort and low levels of monocytes harvested, which necessitated frequent apheresis sessions. These two patients elected to step down to dose levels without monocytes (one from DL4 to DL3b after two apheresis procedures, one from DL2 to DL1 after four apheresis procedures) to continue receiving treatment.

Safety was evaluated in all patients (n = 18). The most common toxicities of any grade were lymphocyte decrease (13/18 patients, 72.2%), anemia (9/18, 50%), fatigue (8/18, 44.4%), abdominal pain (6/18, 33.3%), nausea (5/18, 27.8%), and fever (4/18, 22%; Table 2; Supplementary Table S3). There were no grade 4 toxicities. The most common grade 3 toxicities were decreased lymphocyte count (6/18, 33.3%) and abdominal pain (2/18, 11.1%; Table 2). There was one grade 3 skin infection, attributed to a peritoneal catheter, which resolved with appropriate antibiotic treatment and removal of the catheter. This patient subsequently underwent surgical implantation of an IP port and resumed treatment with no further complications. For patients with surgically implanted ports, there were no infectious or other port-related complications.

The only dose-limiting toxicity (DLT) occurred in a patient on DL 2 who developed grade 3 anemia and was discontinued for this reason. This dose level was subsequently expanded to 6 patients without an additional occurrence. The RP2D was determined to be the highest dose level tested, which was 250 mcg IFNα, 50 mcg IFNγ, and 750 × 10⁶ monocytes.

Patients continued to be treated until they were unable to tolerate therapy or therapy was terminated due to either clinical or radiologic progression (Fig. 3C). The mean number of cycles per patient administered was 3.2, with a range of 1 to 10 cycles. Four patients received five or more cycles (2 patients on the highest dose level, DL4, and 1 patient each on dose levels DL1 and DL2).

Because safety was the primary endpoint of this phase I clinical trial, RECIST measurable lesions were not a criterion for enrollment on study. In patients with RECIST measurable disease, response was determined as an exploratory endpoint (Fig. 3D). The best response observed was a partial response (PR). Of the 9 patients with RECIST-measurable lesions, 2 had a PR, both with ≥30 percent decrease in target lesion size. Four patients had stable disease, and 3 had progressive disease as best response. In addition, 1 patient with primary peritoneal cancer whose disease burden was not quantifiable by RECIST criteria had a CA-125 response according to GCIG criteria (23), with CA-125 decreasing from 124 U/mL to a nadir of 31.5 U/mL after one cycle of treatment. Notably, this patient remained on treatment for 10 cycles, before treatment was discontinued due to clinical progression. In addition to the 3 patients with radiologic progression, a total of 5 patients were taken off study due to clinical progression. Overall, these results indicate that treatment was safe and well-tolerated, with evidence of clinical activity.

Our preclinical data implicated TRAIL as a primary mediator of antitumor effect of IFN-activated monocytes in preclinical studies. To determine if TRAIL was upregulated in our clinical trial patients’ monocytes, we performed RT-PCR on the IFN-treated monocytes from the infusion product of each individual patient and compared with the same patient's untreated monocytes (Fig 4A). Importantly, TRAIL was consistently upregulated in the IFN-treated product by a median of 178% (P < 0.05). Other markers of IFN-pathway activation, namely STAT1 and STAT2, showed a strong trend of upregulation in every individual patient's IFN-treated monocytes. The data confirm TRAIL production in the phase I trial and suggest that IFN signal response machinery was activated in the monocytes administered to the patients.

We asked whether this intraperitoneal therapy resulted in detectable anti-IFN antibodies in plasma, because prior studies have shown that administration of intravenous IFNs in patients with melanoma resulted in formation of potentially neutralizing antibodies (14, 30). Plasma samples were collected from all 18 patients at baseline and before the last cycle of treatment received (range 2–10 cycles). ELISA was performed for both anti-IFNγ and anti-IFNα antibodies compared with positive control WHO antibodies and negative control sera from healthy age-matched donors. No anti-IFN antibodies were detected either at baseline or after maximal exposure (Supplementary Figs. S5A and S5B).

We next sought to determine if intraperitoneal treatment exerted systemic effects. Activation of monocytes by IFNs leads to the production of multiple proinflammatory cytokines (16). Here, we profiled cytokines in peripheral blood at baseline, 3 to 5 days after administration of cycle 1 of treatment, and before administration of cycle 2 using a 10-plex cytokine array. Interestingly, intraperitoneal treatment with monocytes and IFNs led to a marked increase in multiple cytokines associated with monocyte activation, including an 8.8-fold upregulation of IFNγ itself. We also measured a 60% increase in IL6, a 2.4-fold increase in IL10, and 30% increase in TNFα (Fig. 4B–E). No adverse events were observed that correlated with this increase in IL6. This elevation persisted throughout the cycle, as levels of IL10 and TNFα remained significantly elevated above baseline at the pre-C2 timepoint. The cytokine profile of response to monocytes and IFNs in the peripheral blood is consistent with the preclinical data and suggests priming of the immune system towards an antitumor environment.

To further understand the immunophenotypic effects of this treatment regimen and correlate these with response, we profiled immune cells from peripheral blood from the best responders, that is, patients who remained on therapy for five or more cycles. Peripheral blood mononuclear cells (PBMC) at baseline were analyzed by flow cytometry to compare long-term responders (>5 cycles, LTR) to the remaining patients for approximately 100 immune subsets involved in both innate and adaptive immunity (Supplementary Table S4). Compared with the others, LTRs had approximately one-third as many natural killer (NK, defined as CD3⁻, CD56⁺) and again one-third as many immunosuppressive T-regulatory (T-reg, defined as percent CD25⁺FoxP3⁺CD127lo/CD3⁺CD4⁺) cells at baseline compared with the other patients (P < 0.01 for both). The LTRs also trended toward a lower baseline number of myeloid-derived suppressor cells (MDSC, defined as percent CD33⁺/CD11b⁺HLA⁻DR⁻; Fig. 4F–H). The long-term responders had a 65% lower baseline neutrophil-to-lymphocyte ratio (NLR) in peripheral blood (Fig. 4I, P < 0.05), a finding that has been previously correlated with better antitumor immunity and prognosis in ovarian cancer (19, 31, 32) and other solid tumors. This pattern of cellular phenotypes suggests that LTRs have a discernible pattern of immune cells in their peripheral blood at baseline, potentially predisposing them to a more favorable response to immunotherapy.

We then identified systemic differences over time, which might correlate either with longer-term effects of therapy or with eventual disease progression. Importantly, there was no change in numbers of total monocytes in peripheral blood either by flow cytometry (Fig. 5A) or by manual differential (Supplementary Fig. S6). There was no effect of therapy on monocyte differentiation in peripheral blood at the early time point (3–5 days after cycle 1 administration; data not shown). Notably, all four LTRs initially had very high percentages (14.3–22.4%) of nonclassical monocytes (CD14⁻CD16⁺ monocytes, for full gating see figure legend) at baseline. Upon initiation of treatment, the percentage of nonclassical monocytes subsequently decreased below 8% in each of these participants (Fig. 5B), and there was a corresponding increase in classical monocytes (CD14⁺ CD16⁻ monocytes) over time, reaching >57% (Fig. 5C). Similarly, we noted an increase in MDSCs (defined as CD33⁺/CD11b⁺HLA⁻DR⁻) at the time of disease progression, with 4/4 long-term responders having a marked increase in MDSCs at progression compared with baseline (2.7- to 25-fold increase, Fig. 5D–F).

We present preclinical data and a first-in-human clinical trial characterizing the therapeutic mechanism, safety, and tolerability of the combination of intraperitoneal IFNα, IFNγ, and autologous monocytes in ovarian cancer. This combination of innate immune system directed therapy has not been previously tested. Current immunotherapies predominantly focus on the adaptive immune system, with the intent of enhancing T cell activity against cancer, predominantly by blocking immune checkpoints (33).

In our preclinical experiments, we characterize the mechanism of tumor cell killing by IFN-activated monocytes as being largely dependent on direct contact between TRAIL, a type II membrane protein which is potently upregulated on monocytes by IFN treatment, and the death receptors DR 4/5 on tumor cells. This contact results in caspase-8-dependent apoptosis of the tumor cell via the extrinsic pathway. Previous attempts to activate the TRAIL pathway in cancer therapy have used synthetic TRAIL or agonist antibodies of TRAIL receptors. These were unsuccessful clinically due to short half-life of the synthetic protein or inability of agonist antibodies to form oligomers of the receptors (34, 35). The ability of IFNs to upregulate endogenous TRAIL on monocytes potentially provides a more stable and physiologic means to activate death receptors in cancer cells.

We then demonstrate the therapeutic potential of this cellular anticancer using a cell culture system in which IFN-activated monocytes from patients with ovarian cancer were fully capable of killing ovarian cancer cells, as well as in a in a mouse model of ovarian cancer, in which a single IP injection of human monocytes and IFNs resulted in a survival advantage. Because of the dispersion of cells after injecting tumor cells intraperitoneally and the rapid onset of disease burden in our mouse model, tumor tissue was unable to be collected to assess monocyte infiltration and differentiation in vivo. Future studies will incorporate intraovarian bursal injections of ovarian cancer tumor cells to assay for monocyte infiltration and differentiation. Furthermore, we plan to implement multiple IP injections and various dosing schedules of human monocytes and IFNs in future in vivo models, so that tumor infiltration of monocytes can be quantified.

Our resulting phase I study successfully determined the safety and tolerability of the combination of intraperitoneal pegylated IFNα, IFNγ, plus autologous monocytes in women with platinum-resistant or refractory ovarian cancer. When designing this study, we took into account the preclinical data suggesting the necessity for direct contact between the activated monocytes and the cancer cells, and therefore chose the intraperitoneal administration route. Ovarian cancer predominantly occurs in the peritoneal cavity, with metastases outside the abdomen occurring in less than half of patients, and at a median of 6 years later than abdominal recurrences (36). Intraperitoneal chemotherapy has been successfully delivered to patients and can improve survival when delivered in the adjuvant setting (3–5). In this trial, we used a similar approach to the delivery of the study agents to the peritoneal cavity, in most cases using an indwelling port to administer cells and IFNs, followed by saline flush and physical rotation of the patient. The overall approach can be standardized and administered in an institution with capabilities for both intraperitoneal port placement and elutriation of monocytes from apheresis products.

The treatment regimen was well-tolerated and permitted repeat dosing at the RP2D of 250 mcg IFNα, 50 mcg IFNγ, and 750 × 10⁶ monocytes, thus allowing for this to be a platform on which to build more targeted immunotherapy for women with ovarian cancer. The most common adverse events observed were expected on the basis of previous trials of individual IFNs and included hematologic toxicities as well as abdominal discomfort and nausea. There was not a clear decrease in tolerability or increase in adverse events at the higher dose levels. There were no grade 4 or grade 5 toxicities, and the most common grade 3 toxicities were decreased lymphocyte count and abdominal discomfort. These toxicities are comparable to previous reports of intraperitoneally-delivered cellular or chemotherapies (3–5, 37–40).

The combination displayed preliminary evidence of antitumor activity, with 2 of 9 measurable patients experiencing a partial response, 5 of 9 with stable disease, and 3 of 9 with progressive disease by RECIST criteria. One additional patient with peritoneal lesions not measurable by RECIST had a CA-125 response by GCIG criteria, with a greater than 50% decrease in CA-125. This was a heavily pretreated population with a median of five prior treatment regimens. In this clinical setting, responses are rare (<20%) and durations of response are brief (under 6 months). A larger number of patients will be needed to more accurately quantify the response rate of this regimen, and an expansion cohort is planned at the RP2D. As an indication of both tolerability and therapeutic activity, 4 patients were able to remain on this regimen for five or more cycles, and 2 completed 10 cycles of treatment.

We performed correlative biomarker studies to further understand the immunologic effects of this regimen as well as to provide information for development of future combination therapies. Importantly, the therapy did not result in the formation of detectable anti-IFN antibodies even with up to 10 cycles of treatment. This may be due to the IP delivery of the IFNs, in contrast to prior studies where they were administered intravenously.

Corroborating our preclinical mechanistic data, RT-PCR on the therapeutic product demonstrated upregulation of TRAIL and other IFN-regulated genes. Cytokine array of peripheral blood additionally revealed that multiple cytokines secreted by activated monocytes were increased by therapy, including TNFα, IFNγ, and IL10. This is important to note, given that the IFNs and activated monocytes were administered to the peritoneal cavity, yet retained the ability to influence the systemic immune system.

Our comprehensive analysis of immune subsets in peripheral blood also provides insight into the wide-reaching effects of this therapeutic approach, and its effect on both innate and adaptive immunity. Interestingly, compared with nonresponders, long-term responders to therapy had at baseline a lower percentage of NK cells and T-regs, a lower neutrophil-to-lymphocyte ratio, and a trend toward lower MDSCs. In particular, the lower frequency of T-regs and MDSCs in the patients who had clinical benefit may indicate a more favorable and less immunosuppressive tumor microenvironment at baseline, thus facilitating the therapeutic effect of this innate immune cell-based therapy in those patients. Consistent with this idea, these patients were noted to have a rise in MDSC at the time of progression, suggesting a switch to a more immune suppressive environment. These findings warrant evaluation in future studies as possible predictive biomarkers of response to therapy, and as potential targets to increase the anticancer activity of this combination.

In addition, there was a shift in monocyte differentiation in peripheral blood of long-term responders over time, with an increase in classical monocytes and decrease in nonclassical monocytes, as well as an increase in MDSCs. The changes measured in peripheral blood suggest that the intraperitoneal administration of the IFNs was able to influence the immune system more broadly. IFN treatment may skew differentiation of monocytes toward nonclassical monocytes, which preferentially differentiate into activated macrophages and dendritic cells with cytotoxic and antitumorigenic functions (41–43). In contrast, immunosuppression in the tumor microenvironment or chronic inflammation can induce formation and/or release from the bone marrow of classical monocytes instead, which have many protumorigenic functions including suppression of T-cell function and recruitment of Tregs (43). Because these changes were observed to correlate with eventual clinical progression, it is possible that these represent markers of immunologic tolerance and disease progression, because both MDSCs and classical monocytes can be associated with a protumorigenic microenvironment.

Importantly, the immunologic findings that we describe in peripheral blood may not capture more extensive and complex changes in the tumor microenvironment itself. For example, classical monocytes can become immunosuppressive by differentiating into tumor-associated or M2-like macrophages in the tumor, which in turn have pleiotropic suppressive effects on both innate and adaptive immune cell populations (44), and may additionally directly facilitate the formation of ovarian cancer stem cells in the tumor itself (45, 46). Likewise, MDSCs are detectable in peripheral blood but are generally present in much larger numbers in the tumor microenvironment, to which they are recruited by the tumor secreting multiple cytokines and chemokines (47–49). In our planned expansion cohort, it will be important to determine the effect of monocytes and IFNs on the tumor microenvironment via biopsies and extensive immunophenotyping of tumor and/or ascites. If the higher T regs, NK cells, and MDSCs that we measured in the nonresponding patients are corroborated in the tumor microenvironment in larger numbers of patients, they suggest a role for combination with therapeutic agents targeting MDSCs and/or tumor-associated macrophages to improve outcomes for these women (50, 51).

In summary, these data demonstrate the feasibility of repeated intraperitoneal administration of the combination of 250 mcg pegIFNα-2b, 50 mcg IFNγ1b, and 750 × 10⁶ monocytes. The combination is well tolerated with potential antitumor activity, and we have characterized the TRAIL/DR4/5 dependent mechanism of this therapy. Future studies will further determine the activity of this regimen as well as the immunologic effects on the tumor microenvironment. We will incorporate analysis of immunophenotypes and tumor characteristics such as DR4/DR5 expression at baseline that correlate with favorable response as well as later changes corresponding to clinical progression. Such studies will aid in future explorations of combination therapies. Our data set the stage for further investigation of this IP monocyte and IFN regimen as an innate immunity “scaffold” with the potential to be combined with additional therapies targeting key components of the immune system to achieve the greatest possible antitumor response.

Y. Casablanca reports other support from AstraZeneca outside the submitted work. S.M. Holland reports other support from NIH during the conduct of the study. A.T. Nunes reports personal fees from AstraZeneca and Merck outside the submitted work. K.C. Zoon reports a patent for IFN issued. No disclosures were reported by the other authors.

D.S. Green: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. F. Ning: Formal analysis, investigation, visualization, methodology. A. Duemler: Conceptualization, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. T.G. Myers: Conceptualization, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. K. Trewitt: Data curation, software, formal analysis, investigation, visualization, project administration. I. Ekwede: Data curation, software, formal analysis, investigation, visualization, project administration. A. McCoy: Conceptualization, investigation, project administration, writing–review and editing. N. Houston: Investigation, project administration. J. Lee: Investigation, project administration. S. Lipkowitz: Investigation, project administration. A. Zimmer: Investigation. M. Pavelova: Investigation, project administration. E.N. Villanueva: Conceptualization, investigation, writing-review and editing. L. Smith: Conceptualization, investigation, project administration. A. Blakely: Conceptualization, resources, investigation, writing–review and editing. Y. Casablanca: Conceptualization, resources, investigation, project administration. S.L. Highfill: Conceptualization, resources, investigation, writing-review and editing. D.F. Stroncek: Conceptualization, resources, investigation. N. Collins-Johnson: Conceptualization, resources, investigation. S. Panch: Conceptualization, resources, investigation. J. Procter: Conceptualization, investigation. C. Pham: Conceptualization, investigation, methodology, writing-original draft. S. Korrapati: Investigation, visualization and methodology. S.M. Holland: Conceptualization, investigation. L.B. Rosen: Conceptualization, investigation, methodology, writing-original draft, writing–review and editing. A.T. Nunes: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K.C. Zoon: Conceptualization, investigation, methodology, writing–original draft, writing–review and editing. C.B. Cole: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C.M. Annunziata: Conceptualization, investigation, writing–original draft, project administration, writing–review and editing.

The authors are most grateful to all our patients who participated in this clinical trial, as well as our excellent research nurses and nurse practitioners, clinic volunteers, study coordinators, medical oncology fellows, and support staff. Special thanks to Patricia Steeg, PhD, for her invaluable and expert guidance in formulating this manuscript. This work is dedicated to the memory of Samuel Baron, MD. This research was supported by the Intramural Research Program of the Center for Cancer Research, National Cancer Institute, NIH (ZIA BC 011584); Merck (drug only, Sylatron); and Horizon (drug only, Actimmune). This project was also funded in part with federal funds from the NCI, NIH, under Contract No. 75N910D00024, Task Order No. 75N91019F00129. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.