Phase I trial combining chemokine-targeting with loco-regional chemo-immunotherapy for recurrent, platinum-sensitive ovarian cancer shows induction of CXCR3 ligands and markers of type 1 immunity

Abstract

Background:

Increased prevalence of cytotoxic T lymphocytes (CTL) in the tumor microenvironment (TME) predicts positive outcomes in patients with epithelial ovarian cancer (EOC), while the regulatory Treg cells predict poor outcomes. Guided by the synergistic activity of TLR3 ligands, interferon-α (IFNα) and cyclooxygenase-2 (COX-2) blockers in selectively enhancing CTL-attractants but suppressing Treg-attractants, we tested a novel intraperitoneal (IP) chemo-immunotherapy combination, to assess its tolerability and TME-modulatory impact in patients with recurrent EOC.

Methods:

Twelve patients were enrolled in phase I portion of the trial NCT02432378, and treated with IP cisplatin, IP rintatolimod (dsRNA, TLR3 ligand) and oral celecoxib (COX-2 blocker). Patients in cohorts 2, 3 and 4 also received IP IFNα at 2, 6 and 18 million units (MU), respectively. Primary objectives were to evaluate safety, identify phase 2 recommended dose (P2RD) and characterize changes in the immune TME. Peritoneal resident cells and IP wash fluid were profiled via NanoString and Meso Scale Discovery (MSD) multiplex assay, respectively.

Results:

The P2RD of IFNα was 6 MU. Median progression-free and overall survival were 8.4 and 30 months, respectively. Longitudinal sampling of the peritoneal cavity via IP washes demonstrated local upregulation of interferon-stimulated genes (ISG), including CTL-attracting chemokines (CXCL-9, -10, -11), MHC I/II, perforin and granzymes. These changes were present two days post chemokine modulation and subsided within one week.

Conclusion.

The chemokine-modulating IP-CITC is safe, tolerable, and associated with ISG changes that favor CTL chemoattraction and function. This combination (plus DC vaccine) will be tested in a phase II trial.

INTRODUCTION

Epithelial ovarian cancer (EOC), the most common form of ovarian cancer, is the most aggressive gynecologic cancer (1). Despite aggressive surgery and platinum/taxane chemotherapy, the 5-year survival rate for patients with advanced (Stage III-IV) high grade serous ovarian cancer remains low, highlighting the need for new therapeutic strategies (2).

Immunotherapy, especially immune checkpoint blockade, carries promising activity in several solid tumors, but the efficacy in recurrent EOC has been underwhelming (3). This can result from the immunosuppressive character of the peritoneal tumor microenvironment (TME). High numbers of CCL22-attracted regulatory T cells (Tregs) in EOC tumor tissues and ascites contribute to immune suppression and inversely correlate with survival and immune therapy outcome(4–6) In sharp contrast, the presence of effector-type CD8⁺ cytotoxic T lymphocytes (CTLs) in EOC and increased CD8/Treg ratio predict delayed recurrence and prolonged survival (5,7–9).

Since the effectiveness of immunotherapy can be predicted by the extent of CTL infiltration (10–14), we developed a synergistic chemokine-modulating regimen (CKM) to be used alone (to target spontaneously arising CTLs which uniformly express two chemokine receptors CXCR3⁺ and CCR5⁺ (15,16) or combined with specialized alpha dendritic cell (αDC1) vaccines, which induces high levels of high CXCR3/CCR5-expressing CTLs (16), with the goal of increasing the pool of intratumoral CTLs cells. Our preclinical data demonstrate a very high level of synergy between TLR3 ligands (rintatolimod or poly-I:C), IFNα and COX-2 blockers in selectively inducing the CTL/Th1/NK cell-attracting chemokines (CCL5 and CXCL10) but suppressing Treg/MDSC attractants CXCL12 or CCL22 in human and mouse tumor models (15,17–20). Our preclinical data further show that rintatolimod/IFNα-activated ovarian cancer tumors show strongly enhanced chemotactic activity for αDC1-expanded CXCR3 and CCR5-expressing CTLs but show reduced attraction of Tregs (20).

Ovarian cancer is a surface malignancy with characteristic diffuse abdominal spread with rare distant metastases, supporting the rationale for locoregional (intraperitoneal, IP) delivery of chemotherapy, which has shown improved survival (21,22). Similarly, IP delivery of immune modulators represents an attractive approach, given the potential benefit of directly targeting the TME, with potentially lower risk for toxicity. The purpose of the current study NCT02432378 (HCC 11-128) was to evaluate if a novel IP chemo-immunotherapy combination (CITC) comprising cisplatin, rintatolimod (Ampligen®, a dsRNA acting as TLR3 agonist), IFN-alpha (IFNα) and celecoxib (COX-2 inhibitor) can reprogram the tumor microenvironment (TME) for enhanced attraction of the spontaneously arising and αDC1-induced CTLs.

In the phase I portion of our study, performed in the absence of the αDC1 component, we tested whether cisplatin- induced cell death followed by intense locoregional modulation of innate and adaptive immune pathways is safe and triggers T cells chemoattraction into the TME. Our goal was to determine the safe phase II recommended dose for IP IFNα (P2RD), to be used in a follow-up Phase II trial that will specifically define the immunologic and clinical efficacy of tumor loaded αDC1 vaccine in conjunction with the cisplatin/CKM combination regimen. Our translational studies aimed to characterize CKM-induced changes in key immune effector molecules in the TME and to inform the strategy for treatment timing for the follow-up phase II trial.

MATERIALS AND METHODS

Clinical Studies

Clinical trial design and eligibility criteria

We conducted phase I of a single center, open label phase I/IIa trial (NCT02432378, https://clinicaltrials.gov/) approved by the University of Pittsburgh Institutional Review Board, and the US Food and Drug Administration. Based on the FDA discussion, αDC1 component, considered to be an unlikely source of toxicities, was omitted from the phase I portion of the trial. The studies were conducted according to the Declaration of Helsinki. Written informed consent was obtained from each subject. The trial enrolled patients with platinum sensitive, recurrent EOC of any histology. The trial was run between 2015–2019 at Magee- Womens Hospital of UPMC. Platinum-sensitive recurrence was defined as > 6 months progression-free interval from completion of prior platinum therapy. Clinical response was determined using Gynecologic Cancer Inter Group-Response Evaluation Criteria in Solid Tumors (RECIST) (23).

Patients were required to have measurable peritoneal disease and be good candidates for laparoscopy and IP platinum therapy, as judged by the treating physician. Other eligibility criteria included GOG performance status of 0–1, with no limit on prior lines of therapy.

Sequence of events during chemoimmunotherapy combination (CITC)

All patients in this 4-cohort study received intraperitoneal (IP) cisplatin 50 mg/m² (via 1-hour long infusion in 1,000 cc saline) on day 1 of each three-week cycle (6 cycles total). For all patients in this trial, cycle 1 was chemotherapy alone), while CKM was initiated at cycle 2. Rintatolimod (Ampligen®, provided by AIM Immunotech, Inc) 200 mg was administered IP 1–2 hours after the end of IFNα infusion on day 2 of each cycle, except for cycle 3, when CKM was delayed for 1 week (Fig.1A), to test if delayed timing influences response. Celecoxib (COX-2 inhibitor) 200 mg was given orally, daily (additional 200 mg was administered on days of chemotherapy and CKM infusion). Patients in cohort 1 received the above regimen only (no IFN-α). For patients in the subsequent three cohorts, IFNα (Interferon Alfa - 2b, Intron A, commercially purchased) was administered IP in a dose escalation fashion with 3 dose levels: 2 million units (MU), 6 MU and 18 MU, respectively (Fig. 1B). Reconstituted IFNα was administered IP over 30–60 minutes in up to 100 ml of 0.9% sodium chloride. After infusion, additional saline was administered up to 500–1000 ml. Treatment continued until progression, unacceptable toxicity, patient refusal or any other occurrences that prevented further treatment in the opinion of the attending physician. Tumor cytoreduction surgery, if feasible, was planned to take place 3 weeks (±1 week) after the third or fourth cycle of CITC.

Fig. 1.

Clinical trial diagram and summary of clinical responses. (A) Diagram of patient enrollment and cohort distribution. (B) Clinical trial schema. Additional details regarding the drug regimen and timing of biospecimen collection are included in the Methods section. * Celecoxib (200 mg) was administered daily throughout treatment; additional 200 mg was given on days of chemotherapy and CKM infusion. (C) Swimmer’s plot showing time of objective response (PR- partial response; SD- stable disease; PD- progressive disease) in relationship to duration of treatment and time of treatment cessation (x axis). The nine evaluable patients are individually shown (y axis) and color-coded according to the cohort. (D) Cell densities of IP wash samples collected longitudinally during treatment, from nine evaluable and two non-evaluable (NE) patients. Each patient is labeled with cohort number and patient number in each cohort (for example, patient 1.1 refers to cohort 1, patient 1).

Safety assessment

The dose limiting toxicity (DLT) monitoring period included the first two cycles after starting IFNα. Pre-selected adverse events were categorized as DLTs if they were possibly attributed to the treatment regimen per standard dose escalation design with continuous safety monitoring using Bayesian methods as described in the statistical section.

Endpoints

Primary endpoint was to determine the safety of this multi drug CITC regimen, and to identify the Phase 2 recommended dose (P2RD) for IFNα. Secondary objectives were changes in peritoneal fluid of CD8⁺, CD4⁺, Tregs and myeloid-derived suppressor cells.

Objective response rate (ORR, defined as partial response-PR, or complete response-CR) was assessed per RECIST 1.1 criteria. Disease control rate (DCR) includes partial, complete response or stable disease (SD) per RECIST 1.1 criteria. Stable disease (SD) was assigned if partial response (PR) or complete response (CR) was not observed on treatment and if criteria for progression were not met.

Overall survival was estimated from date of enrollment to death or last follow-up. Patients were followed for 3 years after completing treatment to determine recurrence. Progression free survival (PFS) was calculated from enrollment in the trial to switching to different regimen after disease progression.

The immunologic objectives were to characterize the immune response changes in each patient in the peritoneal cavity and peripheral blood and tumor tissue, with the baseline evaluation performed at the day of laparoscopic tumor biopsy and IP catheter placement and compared to evaluation at the completion of the neoadjuvant part of study treatment (one week after the third cycle of the chemotherapy+/− CKM regimen) in each of the cohorts of this trial.

After patients completed study treatment, follow-up evaluations were conducted every 3 months. Follow-up information was collected by either direct contact or through medical record review.

Biospecimen collection and sample processing

To characterize longitudinal changes in the local and systemic immune response during chemo-immunotherapy, a total of 375 clinical samples were collected at various time points from blood, tumor tissue and peritoneal fluid (Supplem. Fig. 1A). Peritoneal cavity resident cells were acquired via outflowing ascites (whenever apparent) and via IP washes, performed via injection of 50 mL of sterile saline. Following the injection of the saline, the “peritoneal wash” effluent was collected via the catheter. IP wash samples were processed within 5 hours of collection. The collection time points for IP washes and/or ascites fluid were prior to IP port placement at time of laparoscopic surgery, at the post-op visit when the port is given a trial of fluid to test for leaking around the port, before each IP chemotherapy, before each IP CKM therapy, 72 h and eight days after chemotherapy (when peritoneal flushes are typically done to assist in maintaining peritoneal catheter patency).

Cells and supernatants were separated via centrifugation and cryopreserved in freezing medium consisting of 10% DMSO (Sigma) in FBS (R&D Systems), until ready to use. A fraction of each sample was stored in RLT buffer (Qiagen) for RNA extraction. While baseline samples collected on day 1, prior to cisplatin administration in cycles 1 (C1D1) and 2 (C2D1) were available from all 9 evaluable patients, longitudinal sampling with paired sampled collected before/after chemo and before/after CKM during cycles 2 and 3 was possible only in 6 and 5 patients, respectively. This drop was partially due to inability to collect the sample, or inadequately low RNA due to low cellularity. Unlike in cycle 2, when CKM was administered the day following cisplatin (C2D2), infusion of CKM occurred at day 9 post-cisplatin in cycle 3 (C3D9, Fig. 1). By varying the time of CKM relative to cisplatin in cycles 2 and 3 (24 hours versus 8 days interval), we inquired whether timing relative to cisplatin influences the effect of CKM.

Blood samples were collected prior to IP port placement, before each IP chemotherapy, before and one hour post each IP CKM therapy, and eight days after chemotherapy. Blood was collected in sodium heparin tubes and processed following the Ficoll Paque Plus (Cytiva) protocol to separate the plasma and peripheral blood mononuclear cells (PBMCs).

Tumor tissue was obtained during port placement (baseline biopsy) and interval debulking surgeries at Magee-Womens Hospital. Tissues were placed in saline and transported to the lab for processing within 2 hours of collection. Tissue samples were stored in formalin-fixed paraffin embedded (FFPE) blocks and RLT buffer for RNA extraction.

Serial IP chemotherapy cohort

This cohort included 8 patients diagnosed with epithelial ovarian cancer who underwent treatment with IP platinum-based chemotherapy and had serial IP wash samples collected (as described above), throughout the 6- cycle standard of care protocol. Studies were performed according to IRB approved protocol; written informed consent was obtained from all study participants. Median age was 52, range 25–67. Most patients had high grade serous histology. Two patients had two or more prior recurrences. Patients with prior recurrences were previously treated with IV chemotherapy.

Translational studies

NanoString

The nCounter PanCancer Immune Profiling Panel (770 probes) was used to analyze gene expression in IP wash cells, PBMCs, and tumor tissues. RNA was isolated using the RNeasy Mini Kit (Qiagen) and quantified on a NanoDrop 2000 UV-Vis Spectrophotometer (Thermo Scientific). The RNA samples were prepared according to the manufacturer’s protocols for the NanoString nCounter Prep Station PS0089 and NanoString nCounter Digital Analyzer (NanoString Technologies). All samples were run at the University of Pittsburgh Genomics Core.

Meso Scale Discovery (MSD)

Forty-one IP wash and 39 plasma samples were analyzed with the MSD platform, according to the manufacturer’s U-PLEX Custom Immuno-Oncology and Biomarker Group 1 protocol. The U-PLEX analytes included TNF-α, TNF-ß, CXCL10 (IP-10) CXCL11 (I-TAC), CXCL12 (SDF1α), granzyme B, IL-10, and VEGF-A. The R-PLEX tested perforin and CXCL9 (MIG). Plates were run on the MSD plate reader. Total protein in IP wash supernatants and plasma samples was measured using the Pierce BCA Protein Assay Kit (Thermo Scientific).

Immunohistochemistry and QuPath

Immunohistochemistry was performed for CD3 (Agilent, clone F7.2.38, 1:100), CD8 (Agilent, clone C8/144B, 1:100), FOXP3 (Abcam, clone 236A/E7, 1:50), MUC1 (BD Pharmingen, clone HMPV, 1:600) and Ki-67 (Agilent, clone MIB-1, 1:50) using 4 μm formalin-fixed paraffin embedded tissue sections. Antigen retrieval was obtained by boiling the tissues for 20 minutes, with the MUC1 staining using a low pH citrate buffer and the remaining stains using a high pH tris-EDTA buffer. Tissues were blocked with a 0.3% H₂O₂ and a 2% BSA block, each for 20 minutes. All stains utilized the EnVision+ System HRP anti-mouse IgG (Agilent) secondary Ab. EnVision+ System DAB (Agilent) was used for the chromogen.

Slide scanning was performed using an Aperio ScanScope CS automatic whole slide scanner at 40X magnification located at the Digital Pathology Imaging Group, UPMC (Pittsburgh, PA, USA), producing .svs files for each slide scanned. All digital analyses were performed using QuPath [1] v0.2.3 (https://qupath.github.io/), an open-source digital image analysis software platform with built-in trainable machine learning image analysis algorithms (24). The regions of interest (ROIs) within tumor areas were selected manually and included at least 50% tumor cells. Cell detection was conducted using QuPath’s built-in algorithm for ‘Positive cell detection’. Parameters were tuned separately for each marker.

Statistics and bioinformatics

Patient demographic and adverse event frequencies were summarized using descriptive statistics. The study used accelerated titration with a Narayana k-in-a-row escalation rule. The goal was to find a starting dose with at most a 10% probability of incurring a toxicity (defined DLTs) from prespecified list of 8 dose limiting toxicities (DLTs). For the rapid titration component 2 patients were treated and observed for 2 cycles on each of the 3 dose cohorts. If one or more of the mentioned DLTs were observed, there would be a switch to the Narayana k-in-a-row rule with k = 6. Dose escalation to a previously untested dose cohort could not occur until all patients at the previous dose are fully evaluated for 2 cycles.

Gene expression data were normalized using R package NanostringNorm (25). Unless otherwise indicated, p-values < 0.05 were considered significant, since the analysis is exploratory for the small sample size in this phase I trial, except for where multiple comparison is corrected and q-value < 0.05 is considered (q values< 0.1 were set as cut-off). Differentially expressed (DE) genes were identified using the R package DEseq2 method (26). FDR adjusted p values (Benjamini-Hochberg correction for multiple comparisons) were calculated for all DE genes. Ingenuity pathway analysis (IPA) (QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis) used DE genes to identify significant pathways. For the MSD data, logarithm transform was applied for the normalization prior to the identification of differentially expressed (DE) proteins. Significantly changed proteins (p-values < 0.05) were identified using the paired sample t-tests. Correlations between protein expression levels and gene expression levels are calculated via Spearman’s rank correlation coefficient. Correlation coefficients with absolute values greater than 0.7 were identified as significant correlations using previously reported criteria (27). Overlapping genes were visualized using Venn diagrams (Venny 2.1) (28). Gene Set Enrichment Analysis (GSEA) was performed using GSEA software. Briefly, differentially expressed genes from NanoString data were processed to create a rank list of genes. Genes were ranked based on both LogFC and p-value with a formula of (Sign(logFC)*-log10(Pvalue)) in R. The resulting rank file was processed via ‘GSEAPreranked’ option into GSEA software to analyze the data (29).

Data Availability Statement

Some of the raw data for this study (NanoString) were generated at the University of Pittsburgh Genomics Core. Derived data supporting the findings of this study are within the article and its supplementary data files and available from the corresponding author upon request.

RESULTS:

Determination of the safety profile and identification of P2RD for the combination of local cisplatin, rintatolimod and IFNα, with oral celecoxib

The trial schedule and treatment schema are summarized in Fig 1A. Patients’ characteristics are listed in Table 1. Of the 12 patients enrolled in this Phase I trial, 9 (75%) were evaluable for safety, toxicity, and other endpoints. The 3 non-evaluable patients did not complete at least 3 cycles, due to platinum hypersensitivity reactions or port complications. Overall, the regimen was well tolerated, apart from the highest dose of IFNα. Most common toxicities for all grades were anemia (58%), hypomagnesemia (50%), hyponatremia (41.7%), arthralgia (41.7%), and fatigue (41.7%) (Table 2). There was one grade 4 hypomagnesemia (8.3%). Dose limiting toxicities of abdominal pain of grade 3 or more were noted in two patients who received 18MU IFNα (cohort 4). The P2RD for IFNα was determined to be 6 MU every 3 weeks. Median PFS was 8.4 (3–16.4) months. Median OS was 30 (8–66) months. These survival outcome data provide an encouraging early signal. Overall response rate (ORR) was 55.6% and the disease control rate (DCR) was 77.8%, consistent with the expected platinum-sensitive response (30). Among responders, median duration of response was 11.7 (6–16.4) months. Each patient’s response type and length of response until next line of therapy for the respective cohorts are shown in Fig. 1C. Of note, all patients had serous histology except two patients with carcinosarcoma. Two patients had 5 prior lines of therapy whereas the rest had 1–2 prior lines of therapy.

Table 1.

Patient demographics

Number of Patients	12
Age (years), median (range)	65 (52–73)
Histology
Serous	10 (83%)
Carcinosarcoma	2 (17%)
Grade
1	1 (8%)
2	0 (0%)
3	11 (92%)
Recurrences
1	8 (67%)
2	1 (8%)
3	1 (8%)
4	0 (0%)
5	2 (17%)
Prior Lines of Therapy
2	8 (67%)
>2	4 (33%)
Response
Not Evaluable	3 (25%)
Partial Response	5 (42%)
Stable Disease	2 (17%)
Progressive Disease	2 (17%)
Progression Free Survival (months), median (range)	8.4 (3–16.4)
Overall Survival (months), median (range)	30 (8–66)

Table 2.

Adverse events related to cisplatin/CKM intervention.

AE n (%)	All Grades	Grade 3	Grade 4
Anemia	7 (58)	1 (8)	0
Hypomagnesemia	6 (50)	3 (25)	1 (8)
Abdominal Pain	5 (42)	1 (8)	0
Hyponatremia	5 (42)	1 (8)	0
Arthralgia	5 (42)	0	0
Fatigue	5 (42)	0	0
Edema	4 (33)	0	0
Neutropenia	4 (33)	2 (17)	0
Elevated Creatinine	3 (25)	0	0
Neuropathy	3 (25)	0	0
Nausea	3 (25)	0	0
Port malfunction	2 (17)	0	0
Constipation	2 (17)	0	0
Lipase elevation	2 (17)	1 (8)	0
Rash	2 (17)	0	0
Vomiting	2 (17)	0	0
Hypertension	2 (17)	0	0
Urinary Incontinence	2 (17)	0	0
UTI	2 (17)	0	0
Hypocalcemia	2 (17)	0	0
VTE - line associated	1 (8)	1 (8)	0
Rectal Fistula	1 (8)	1 (8)	0
Infusion Reaction	1 (8)	1 (8)	0
Abdominal Infection	1 (8)	1 (8)	0

Chemo-immunotherapy combination triggers robust transcriptomic changes in peritoneal resident cells obtained in longitudinal sampling via IP washes

Longitudinally collected IP wash samples showed increased cellularity two days post- CKM in each treatment cycle, reflecting an early pro-inflammatory reaction in the TME, and subsided within 5 days from treatment infusion (Fig. 1D Four of the five patients with partial response (PR) showed increased cellularity two days post infusion of immune modulators, in at least one cycle (i.e., day 4 in cycles, 2, 4, 5, 6 or day 11 in cycle 3) (Supplem. Fig. 1B). Analyses of DE genes triggered at day 4 of cycle 2 (C2D4, which corresponds to 2 days post CKM and 3 days post-cisplatin) revealed a robust change in n= 397 genes compared to day 1 of cycle 2 (Fig. 2A and Supplem. Table 1A). Similarly, analyses of DE genes at day 11 of cycle 3 (C3D11 which, similarly to C2D4 also corresponds to 2 days post CKM), identified a total of n= 411 genes that were significantly changed when compared to baseline levels for cycle 3 (C3D1) (Fig. 2A and Supplem. Table 1B). In contrast to the robust changes seen in cycles 2 and 3 (combination chemo-immunotherapy), fewer genes (n=62) were altered by the end of cycle 1, which consisted of chemotherapy alone (C2D1 vs C1D1, Fig. 2A and Supplem. Table 1C). Analyses of overlapped genes show that most (59/62, 95%) of the cisplatin-induced changes upon completion of cycle 1 were also found in at least one of the subsequent two treatment cycles, with most genes (41/62, 66%) common for all three (Fig. 2B), suggesting that these 41 genes are primarily induced by cisplatin. Addition of CKM in cycles 2 and 3 triggers a significant number of additional DE genes (n=257) that are overlapping between cycles 2 and 3, yet not overlapping with cycle 1 (chemo alone, Fig. 2B). Virtually all (99%) of DE genes show concurrent directionality (Supplem. Fig. 2). These results demonstrate a consistent gene expression pattern triggered by the CKM at two consecutive treatment cycles, with similar changes occurring if CKM was administered either 24h (cycle 2) or 8 days (cycle 3) post cisplatin.

Fig. 2.

Treatment-induced transcriptomic changes in IP washes. (A) Heatmaps and volcano plots of DE genes identified in IP washes by NanoString: C1D1 vs C2D1 (n=62, top), C2D1 vs C2D4 (n=397, middle), C3D1 vs C3D11 (n=41, bottom). Volcano plots show magnitude of response for the respective heatmaps on the left. Red dots represent individual DE genes with p-values ≤0.05 (y axis cut-off) and magnitudes of log₂ FC > 0.58 (x axis cut-off). Green dots represent genes with above the cut-off for fold change but below the p-value cut-off. Gray dots are genes with non-significant (NS) p-values and log₂ FC below cut-off. Individual DE genes are listed in Supplementary Table 1. (B) Venn diagram for DE genes shown in A. Cycle 1 (C2D1 vs C1D1); Cycle 2 (C2D4 vs C2D1); Cycle 3 (C3D11 vs C3D1). (C) GSEA histograms showing enrichment of DE genes at cycles 2 and 3 (chemo/CKM), in contrast to pattern observed at cycle 1 (chemo alone). The leading edge (most significant genes) are shown as vertical bars accumulated either left or right of the green line histogram, indicating the up- or down-regulated genes, respectively (D). Heatmaps of n=24 interferon stimulated genes (ISG, listed in textbox) after cycle 1 (top) and two days post-CKM in cycle 2 (bottom). (E) Heatmaps of n=26 interferon gamma induced genes (IFNg_IG, listed in textbox) after cycle 1 (top) and two days post-CKM in cycle 2 (bottom). Numbers on columns refer to patient cohort.

As expected, administration of rintatolimod and IFNα triggered upregulation of TLR and IFNα signaling elements in cycles 2 and 3 (chemo-immunotherapy) reversing the negative trend mediated by chemotherapy alone (cycle 1) (Fig. 2C). A similar pattern was also observed for IFNγ response and antigen processing and presentation pathways (Fig. 2C). Using type I interferon stimulated genes (ISGs) from a previously reported gene signature (31), we found that all the 26 ISGs included in our NanoString panel showed significant changes upon the addition of CKM but not after chemotherapy alone (Fig. 2D). Importantly, ISG upregulation is most prominent in patients from cohorts 2–4 (who received IFNα) but not from cohort 1 (no IFNα), suggesting a drug-specific effect. Similarly, addition of CKM triggers significant changes in 24 interferon-γ response genes that are important for response and resistance to immune therapy and are predictive of survival (Fig. 2E) (32). As found with the ISG, these IFNγ response genes show a non-distinct pattern of expression after cycle 1 (chemo alone, Fig. 2E). In contrast, a clear upregulation occurs two days after CKM, and a consistent pattern for both gene sets is similarly observed at cycle 3 (Supplem. Fig. 3A–B). Among the significantly upregulated genes are those encoding for STAT1/STAT2, T lymphotactic chemokines (CXCL10, CXCL16), MHC I, MHCII, TAP1 proteasomal subunits (PSMB8 and PSMB9), and cytolytic effectors (perforin, granzymes) all of which are important for T lymphotaxis and function via TCR engagement with cognate tumor antigens.

To further identify the role of immune modulators, we used as reference a cohort of patients that received only IP cisplatin, as part of standard of care. NanoString analyses of IP resident cells from these patients show that chemotherapy alone does not trigger the upregulation pattern seen for ISG and IFNγ gene sets, following chemo-immunotherapy (Supplem. Fig. 3C–D) Together, these results point to a significant CKM-induced impact on peritoneal resident cells, primarily via upregulation of interferon stimulated genes.

Protein content of IP wash fluid captures treatment-induced increases of IFN-induced immune effector molecules

Using multiplex platform (MSD), we measured in cell-free IP fluid the concentration of 13 proteins, selected to include both markers of anti-tumor immunity (such as IFN- induced CXCR3 ligands with T lymphotactic properties CXCL-9, -10, -11); immune cytotoxic effectors such as Perforin, Granzyme B, TNFα and TNFβ), as well as markers associated with failed anti-tumor responses (CXCL12, TGFβ-1, -2, -3, IL10 and VEGF) (33–36). Compared to baseline for cycle 2 (C2D1), all 11 markers were increased at day 4 (C2D4). However, the increase was reversed, and expression levels at day 9 were lower compared to day 4, suggesting that CKM - induced increases in these markers occur soon after infusion and return to pre-infusion levels within one week (Fig. 3A). A similar increase in all markers was also observed during cycle 3, two days after CKM (Fig. 3B). While the heatmaps (Fig. 3A and B) show the directionality of change for all 11 measured proteins during two consecutive treatment cycles, we note that no changes in immune suppressive CXCL12, TGFβ, and IL10 were observed, although a trend toward lower VEGFA was identified two days post-treatment (Fig. 3C and Supplem. Fig 4A). Of all immune modulatory molecules tested here, CXCL10 was found to be the most abundant in IP wash fluid and, together with CXCL11 and TNFα, shows significant treatment induced increases at both treatment cycles (Fig. 3C). In addition, the levels of CXCL9, granzyme B and perforin were significantly higher two days after CKM at cycle 3 (Fig. 3C). Importantly, the increases in lymphotactic CXCL9, CXCL10 and immune effector perforin are not seen in the chemo alone reference cohort (Supplem Fig. 3E), suggesting that these are unique to the chemo-immunotherapy combination employed in the trial and they may not occur during standard of care.

Fig. 3.

Protein changes in IP washes. (A) Concentration of 11 different analytes (listed in textbox) in IP washes measured with MSD. Heatmaps show upregulation (red) of protein expression at C2D4 compared to C2D1 and downregulation (green) at C2D9 compared to C2D4. (B) Heatmap showing upregulation for all markers, at C3D11 compared to C3D1. (C) MSD results for individual anlaytes at cycles 2 and 3. *** p< 0.001; ** p< 0.01 * p< 0.05 (paired sample, two tailed t-test).

Correlation of each of the markers with all DE genes from each time point, reveals that for CXCL-9, -10, -11, TNFα and perforin there was a significant positive correlation with STAT1, an essential component of interferon signaling (Supplem. Fig. 4B). Of all genes with positive correlation values (ρ ≥0.7, p <0.05; q< 0.05) STAT1 was the top gene for CXCL10, CXCL11 and perforin and second highest for CXCL9. Overall, these findings demonstrate a consistent pattern of increase in locoregional concentration of several key, interferon stimulated chemoattracting and immune effector molecules. While we cannot ascertain the time corresponding to the peak of expression, this data points to an immune hot tumor microenvironment two days post-CKM. Nevertheless, the pro-inflammatory effect seems to diminish in time, reversing to the baseline levels within one week, in line with the increases in overall cellularity in IP wash samples (Fig. 1D).

Tumor immune profile at interval debulking shows partial overlap with IP wash, and points to upregulation of genes encoding for perforin and granzyme B

Tumor tissue was collected from interval surgery in five patients (one each from cohorts 1, 2 and 4 and two from cohort 3). Compared to baseline tumor biopsy, a total of 43 DE genes were identified at surgery (Fig. 4A, Supplem. Table 1D), of which 28 (65%) were upregulated (Fig. 4B). Most of the DE genes identified in tumor tissue (23/43, 53%) were also found to be DE in IP wash, and of these, 78% (18/23) of genes show concordant directionality (Supplem. Fig. 5). Of the commonly downregulated genes, MUC1 and BIRC5 (which encode for mucin 1 and survivin tumor associated antigens, respectively) were significantly lower in both tumor and IP wash, suggesting tumor involution (Supplem. Fig. 5). We also note that unlike the IP wash analyses, that captured “acute” changes in ISG within two days from administration, gene profiling of tumor tissue (collected one month post cycle 3 CKM) did not show differential expression in these genes. Among the upregulated DE genes, most abundantly present were genes encoding for markers of innate immune cells, (CD11b, CD11c, MARCO, CD36) (Fig. 4C) and for chemo-attracting molecules such as CXCL3 and CXCL2, which attract monocytes and polymorphonuclear cells, respectively (Fig. 4D). Although some of the upregulated chemokines also attract T cells (for instance CCL13), no significant changes were seen in CD3, CD8, CD4 genes or in CD3, CD8, FOXP3 tumor infiltrating cell counts, determined by IHC and QuPath automated counting (Fig. 4E and 4F and Supplem. Fig.6). We note that two patients with PR (1.1 and 3.4) had increased T cells infiltration at baseline, above the cut-off of 363 cells/mm², recently proposed to classify immune “hot” tumors (37). The CD8 counts, and CD8/FOXP3 ratio remained unchanged or were only moderately increased in the residual solid tumor fractions of all patients (Supplem. Fig. 6). Nevertheless, similarly to IP wash protein measurements at cycle 3, two genes encoding for cytolytic proteins perforin-1 (PRF1) and granzyme B (GZMB) were significantly upregulated (Fig. 4G), suggesting an increase in intra-tumoral cytolytic activity at interval surgery.

Fig 4.

Treatment-induced changes in tumor tissue. (A) Heatmaps of DE genes at baseline and interval surgery (5 patients, paired samples, n=43 DE genes, p<0.05). From left to right, each column in the heatmap represents individual patients across cohorts 1–4: one each from cohorts 1, 2 and 4 and two from cohort 3. All DE gene names, individual log2FC, p-values and FDR adjusted p-values are listed in Supplem. Table 1. (B) Volcano plot of DE genes. Cut-offs for p-value (y axis) and log_2αFC (x axis) are described in Fig 2. (C) Expression upregulation of genes associated with macrophage biology (p<0.05). (D) Expression upregulation of genes involved in chemotaxis (p<0.05). (E) Quantitative measurements of tumor CD8+ T cells identified by IHC (upper panels) and counted with QuPath (lower panels). (F) Box plots showing counts of CD8 TILs (QuPath) from individual patients at baseline and interval surgery. (G) Upregulation at interval debulking (compared to baseline) of genes encoding for perforin 1 (PRF1) and granzyme B (GZMB) (p<0.05).

DISCUSSION

We performed a first-in-human, phase 1 dose-escalation clinical trial of IP cisplatin plus local rintatolimod/IFNα/celecoxib, the CKM combination synergistically promoting selective CTL attraction in preclinical human and mouse models of EOC (19,20). Our results demonstrate that this four-drug combination is safe and generally well tolerated at low doses of interferon, with little added toxicity, compared to cisplatin alone. Treatment-related AEs were mostly grade 1–2. When present, grade 3 and 4 toxicities were attributed to the known safety profile of IP cisplatin and high dose interferon (38). Previously reported GOG regimen of IP cisplatin and INFα exhibited similar toxicities at comparable dosing (38–41). The overall response rate was consistent with an active regimen in platinum-sensitive disease (30,42) but the long term survival seen in 3 of the 5 responding patients is similar to what we and others have reported in other immune directed therapies, including with IP IL-2 (43). This extended survival despite a modest initial response is not typical of cisplatin alone in this population, may be associated with the favorable chemokine modulation of the TME, and warrants phase II investigation.

We have previously shown that IP chemotherapy in the recurrent platinum- sensitive setting is feasible and efficacious (44). Here, the IP catheter enabled the delivery of immune modulatory agents directly into the TME. Moreover, repeated washes of the IP catheters, employed to maintain their patency, provided us with a unique opportunity to sample the peritoneal cavity at multiple time points during potential tumor involution, and study longitudinal changes of the TME in response to chemotherapy and biologics. Our results demonstrate that CITC-triggered “acute” inflammatory changes in IP wash cells and fluid. Most notably, we noted significant increases in CXCR3 ligands CXCL9 and CXCL10, which mediate the recruitment of tumor targeting CXCR3+ T cells and natural killer (NK) cells into solid cancers (45) and are strong and independent predictors of improved patient survival in EOC (46). Current findings also point to upregulation of IFNγ pathway in IP wash cells and increased concentration of TNFα in IP fluid.

Our treatment regimen included TLR3 ligand rintatolimod (poly-I:C₁₂U), related to the more widely used poly-I:C (47). However, unlike poly I:C, rintatolimod does not activate the RIG-I/MDA5 pathway, thus avoiding NFκB- dependent induction of immune suppressive IDO1, IL-10 and of Treg chemoattracting CCL22 (19,20), none of which we found to be increased in this trial.

The observed upregulation of ISG and IFNγ response genes, combined with increases in TNFα, granzyme B and perforin concentration point to the ability of the CITC to induce type 1 immunity in the TME. Although IFNγ may induce opposing roles in the immune TME, the outcome depends on context – particularly the ability to enhance MHC-I presentation. In line with these observations, we note that antigen processing and MHC-I presentation pathways were upregulated in our dataset, further suggesting the engagement of ISG in the immune cell compartment and highlighting their potential to promote anti-tumor effector mechanisms (48). Nevertheless, despite their cytotoxic properties, IFNγ and TNFα, synergize in the induction of COX-2 (49), the key mediator of prostaglandin E2 (PGE2) synthesis (50). Our previous studies demonstrate that PGE2 coordinates intra-tumoral production of multiple suppressive factors and Treg/MDSC attractants in EOC TME, supporting the rationale for inclusion of COX-2 inhibitor celecoxib (20,51,52). Indeed, despite changes in markers of type 1 immunity, the current study showed no significant changes in COX-2- encoding PTGS2 in IP wash cells and no increases in concentration of MDSC chemoattractant CXCL12 in IP fluid. This suggests that inclusion of a COX-2 inhibitor mitigates PGE2-triggered mechanisms of immune suppression that could otherwise self-limit the effectiveness of rintatolimod/IFNα induced type 1 immunity. Additionally, we observed no treatment induced changes of transcription of genes encoding for some of the major immune checkpoints (including PDL1, PD1, CTLA4), when comparing post- vs pre- treatment IP wash cells in each cycle.

In contrast to the acute changes seen in IP wash cells within two days from CKM infusion, the gene expression profile of tumor tissue, obtained more than one month from cycle 3 CKM, did not reveal the same ISG inducing effect. The dominance of innate immune (mostly monocyte/macrophage related) genes at interval debulking may be reflective of the more “chronic” effects of treatment, leading to tumor involution and the resulting tissue repair reaction. Most tumor-associated macrophages (TAMs) are CD163 + M2 TAMs and correlate negatively with outcome (53). Although our studies could not fully capture changes in monocyte/macrophage phenotypes, we note that treatment included cisplatin and celecoxib, both of which have been reported to cause a shift in TAM function. Chemotherapy skews the TAM populations in HSGOC toward an antitumor phenotype while inhibition of Cox-2/PG-E2 can also decrease M2 polarization (54,55).

As one of the important limitations of our study, we note the small patient cohort size. Moreover, tumor tissue analyses were possible only in a minority of patients, making the sample size inadequate for statistical comparisons, especially for assessment of CD8 TILs. Additionally, confirmation of treatment- induced changes in IP wash immune cell subsets (using live cell assays like flow cytometry) was not consistently feasible in this study, due to the limited number of total cells retrieved via IP washes.

In summary, we report here an acceptable safety profile of a complex, four-drug CITC regimen for recurrent EOC. The observed significant increase in ISGs, T cell chemotactic CXCR3 ligands and markers of type 1 immunity (TNFα, IFNγ genes, granzyme B, perforin) without upregulation of markers of secondary suppression (like CXCL12) are encouraging. We are currently implementing the follow-up Phase II trial, which will combine the CKM component with autologous tumor loaded αDC1 vaccine (19) and the CKM component to expand tumor-specific CXCR3⁺/CCR5⁺ effector-type CTL in order to promote CTL infiltration of the tumor sites and potentially improve the overall outcomes.

Statement of Translational Relevance.

Strategies to enhance the response to therapy in epithelial ovarian cancer (EOC) are greatly needed. In this phase I trial, we established the safety and phase II recommended dose of intraperitoneal (IP) chemo-immunotherapy combination (CITC) as a novel therapeutic venue for recurrent EOC. Patients received IP cisplatin and a chemokine modulation regimen (CKM) that combines rintatolimod (dsRNA acting as TLR3 agonist), IFN-alpha (IFNα) and celecoxib (COX-2 inhibitor). Further, in this study, we present correlative translational data regarding the impact of CKM on loco-regional immunity. NanoString gene expression analyses and cytokine protein measurements in serially collected IP fluid samples reveal significant increases in T cell chemotactic CXCR3 ligands (CXCL-9, 10, 11) and markers of type 1 immunity (TNFα, IFNγ genes, granzyme B, perforin), without upregulation of markers of secondary suppression (CXCL12). The CITC regimen reported here will be combined with a dendritic cell vaccine in a follow-up phase II trial.