IFNα Augments Clinical Efficacy of Regulatory T-cell Depletion with Denileukin Diftitox in Ovarian Cancer

Tumor-associated factors and cells inhibit anticancer immunity and reduce immunotherapy efficacy (1). Inhibitory immune checkpoints are effectively blocked by antibodies against them as immunotherapy (2–4). However, cancer-related immune factors aside from immune checkpoints, including CD4⁺CD25^hiFoxp3⁺ regulatory T cells (Tregs) also subvert antitumor immunity (1, 5–7). Abundant animal data demonstrate Treg depletion improves antitumor immunity and elicits beneficial clinical responses (7). Human trials suggest Treg depletion improves immunity in various cancers (8–10), but clinical effects are generally modest or unreported.

High-grade serous ovarian carcinoma is the most common ovarian cancer histologic type. Because ovarian cancer is immunogenic, immunotherapy could be beneficial (11), but immunotherapy trials in ovarian cancer have generally not shown efficacy (12). Tregs inhibit endogenous antitumor immunity and predict poor ovarian cancer survival (5), suggesting Tregs as a specific target for ovarian cancer immunotherapy. We tested the IL2/diphtheria toxin, denileukin diftitox (DD) as it depletes Tregs in patients with cancer (8, 10, 13). One DD infusion depleted blood Tregs and improved immunity. On the basis of the partial clinical DD response of one patient with ovarian cancer in our phase 0/I trial, we undertook a phase II trial of DD in advanced stage, treatment-refractory ovarian cancer, but failed to show clinical efficacy (14). We thus studied additional approaches.

We found that IFNα improves clinical and immune DD-mediated effects. Although it is common to switch agents for clinical failure, DD achieved desired Treg depletion. On the basis of preclinical observations, we opened a companion phase II trial adding pegylated (IFN)α2a if DD alone failed clinically, rather than stopping DD and switching to another agent. This study characterizes adding pegylated IFNα2a to DD for clinical benefit and mechanisms.

The phase 0/I trial used 3+3 dose escalation to test primary endpoints of DD (Eisai or Ligand Pharmaceuticals) dose, schedule, safety, tolerability, and Treg reduction in patients with histologically proven epithelial carcinomas failing standard of care without curative options enrolled between May 2003 and August 2005 (patient characteristics; Supplementary Table S1). Secondary outcomes included immunity in peripheral blood. Patients were ≥18 years, had measurable disease, good organ function and performance status, and no uncontrolled medical conditions. Exclusion criteria included concomitant steroids or autoimmune disease history. Toxicities were determined by NCI Common Toxicity Criteria, version 2.

In “A Trial of Intravenous Denileukin Diftitox plus Subcutaneous Pegylated IFNα2A in Stage III or IV Ovarian Cancer” (NCT01773889), DD was administered intravenously at 12 μg/kg over 1 hour every 3 or 4 weeks. Pegylated human IFNα2a (Roche) was administered subcutaneously at 180 μg weekly within 1 hour of DD. Premedication included oral acetaminophen 650 mg, diphenhydramine 50 mg, and prochlorperazine 25 mg given 30 minutes before DD without steroids. Three patients were enrolled in 2009, two of whom received treatment before the trial was halted for lack of DD. These patients (103 and 105), both with metastatic epithelial ovarian cancer, had progressed in the phase II trial, “Intravenous Denileukin Diftitox in Stage III or IV Ovarian Cancer” (NCT00880360), received five or 15 cycles of combination therapy, respectively. Patients underwent regular history, physical and laboratory examinations, and toxicity assessments.

Blood was collected into sterile, heparin-containing tubes. Peripheral blood mononuclear cells (PBMC) were isolated following centrifugation over a discontinuous Ficoll-Hypaque density gradient (Sigma-Aldrich) and quantified manually or by Vi-Cell (Beckman-Coulter). In all patient studies, approval was obtained from the Institutional Review Boards and informed consent was obtained.

C57BL/6 (BL6; NCI), syngeneic RAG1^−/−, RAG1^−/−OT-I (OT-I), and RAG1^−/−OT-II (OT-II; Taconic), and IFNAR^−/− mice (gift of Ross Kedl, National Jewish) were housed in microisolator cages and given commercial chow and water ad libitum (15). For survival studies, mice were weighed every other day beginning 28 days after tumor challenge and followed until death, discomfort or ≥30% ascites weight gain. Animals were euthanized by isoflurane inhalation followed by cervical dislocation, and tissues collected aseptically. Animal studies were approved by the institutional animal care and use committee.

ID8, an epithelial ovarian cancer cell line, and ID8-OVA, expressing ovalbumin (OVA; ref. 16), both on the BL6 background, was cultured in RPMI1640 medium plus 10% FBS, 4 mmol/L glutamine, 10 mmol/L (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid and antibiotics at 37°C in a 5% humidified CO₂ atmosphere. A total of 10 million cells in 100 μL PBS were injected intraperitoneally into female mice. Tissues were collected aseptically, mechanically disrupted into single-cell suspensions in MACS buffer and filtered (70 μm) into single-cell suspensions. Erythrocytes were lysed with Red Blood Cell Lysing Buffer.

Cells were prepared and sorted and/or analyzed via flow cytometry as described on LSRII and FACSAria hardware (Becton-Dickinson) using FACSDiva or Flo-Jo software (representative examples; Supplementary Fig. S2; ref. 17). Cells were resuspended in PBS plus 0.05% FBS and 0.5 mmol/L EDTA and quantified by Vi-CELL. Antibodies were added for 30 minutes at 4°C. Cells were washed, resuspended in fixation/permeabilization solution (eBioscience) for 20 minutes at 4°C, washed twice with permeabilization buffer (eBioscience), stained with intracellular antibodies for 30 minutes at 4°C and fixed in 1% paraformaldehyde. For intracellular cytokine staining, cells were stimulated with 100 ng/mL lipopolysaccharide (Sigma) overnight at 37°C, and leukocyte activation cocktail (BD Biosciences) was added for 4 hours at 37°C for staining/fixation as above. For detection of OVA (tumor antigen)-specific cells, SIINFEKL pentamer (Proimmune) was added for 10 minutes at ambient temperature in the dark. Cells were washed, stained for surface markers, incubated with fixation/permeabilization solution, stained with intracellular antibodies, and fixed.

We used validated, commercially available antibodies for mouse cell flow cytometry: from eBioscience (clone): Foxp3 (FJK-16s), anti-CD62 L (MEL-14); from BD Biosciences: anti-CD3 (500A2), anti-CD4 (GK1.5), anti-CD25 (PC61), anti-CD69 (H1.2F3), anti-CD11c (HL3), anti-CD11b (M1/70), anti-MHC class II (2G9), anti-IL2 (C15.6), anti-IFNγ (clone XMG1.2, BD Biosciences); anti-CD8 (clone 5H10, Invitrogen); from BioLegend: anti-CD80 (16-10A1), anti-CD86 (GL-1). For human cell flow cytometry: anti-Foxp3 (236A/E7, eBioscience); from BD Biosciences: anti-CD3 (SK7), anti-CD25 (M-A251), anti-CD127 (HIL-7R-M21); anti-CD4 (OKT4, BioLegend); anti-CD8 (5H10, Invitrogen). In the earliest studies reported here, commercial anti-Foxp3 antibodies were not available, and Treg were identified as CD3⁺CD4⁺CD25⁺ cells. Anti-FoxP3 was added to panels when it became commercially available.

Mouse bone marrow–derived dendritic cells (DCs) and human monocyte–derived DCs were generated as described previously (7).

DD was reconstituted in sterile PBS, aliquoted and frozen at −80°C until use. Human universal IFN (“IFNα”, PBL Assay Science) is an IFNαa1/a2 hybrid protein delivering a pan-IFNα signal in human and mouse cells (18). Two weeks after tumor challenge, mice were treated with intraperitoneal DD 5 μg/mouse/week, IFNα 20,000 units/mouse on the first 4 consecutive days of weekly treatment cycles or both, with DD administered on day 3 of weekly cycles. Anti-IL6 (R&D Systems, 100 μg/mouse) was injected intraperitoneally 1 day before DD and/or IFNα versus PBS control. For in vitro studies, mice underwent one treatment cycle and were euthanized 2 days after DD, unless otherwise specified.

Treg suppression was tested as we reported previously (5, 17, 19). Briefly, CD4⁺ T cells were enriched by RosetteSep CD4⁺ T cell isolation kit (Stemcell Technologies). CD4⁺CD25⁻ responder T cells and CD4⁺CD25^hi Tregs were electronically sorted from these to >98% purity (Supplementary Fig. S1C). Responder CD4⁺CD25⁻ T cells (2 × 10⁵/mL) were stimulated with 2.5 μg/mL anti-human CD3 (clone UCHT1), 1.2 μg/mL anti-human CD28 (clone CD28.2; both Becton-Dickinson) and fresh monocytes (2 × 10⁵/mL) plus Tregs as indicated. T-cell proliferation was assessed 72 hours later by thymidine incorporation. Antigen specificity of cytotoxicity was done as reported previously (5). Briefly, HLA-A2⁺ lymphoblastoid T2 cells (5 × 10⁶/mL, ATCC) were labeled with 10 μmol/L 5,6-carboxyfluorescein diacetate succinimidyl ester (CFSE, Molecular Probes) for 10 minutes, 37°C and pulsed with 5 μg/mL of three HLA-A2-binding Her2 peptides: p369–384 (KIFGSLAFLPESFDGDPA), p688–703 (RRLLQETELVEPLTPS) and p971–984 (ELVSEFSRMARDPQ), or HLA-A2 influenza virus matrix control GILGFVFTL (Multiple Peptide System). 5 × 10⁵ pulsed, labeled T2 cells were incubated with 2 × 10⁵ sorted ascites CD45⁺CD4⁺CD8⁺ T cells an HLA A2⁺ subject plus Tregs as indicated. Cytotoxicity was evaluated by killing of CFSE-labeled Her2 peptide-pulsed T2 cells minus cytotoxicity of matrix control T2 cells using flow cytometry to detect annexin V and 7-AAD. Mouse cell isolation procedures used mouse instead of human reagents, but were otherwise identical.

Treg suppression was calculated as [1 – (T-cell proliferation without Treg − T-cell proliferation with Tregs)/T-cell proliferation without Treg] × 100%. For DC suppression, CD11c⁺ DCs were harvested from spleens, mesenteric lymph nodes (mLN), or peritoneal exudate cells (PEC) from tumor-bearing or naïve wild-type (WT) or IFNAR^−/− mice and purified by EasySep Mouse CD11c Positive Selection (StemCell Technologies, Inc.) according to manufacturer directions. DCs were incubated with CD4⁺ responder T cells and beads ± Tregs (1 DC:10 Tregs), with anti-IL6 (0.2 or 0.5 μg/mL), anti-CD80, anti-CD86 antibodies or isotype controls (BioLegend) as indicated. Recombinant mouse IL6 (R&D Systems) at 0.1 or 1.0 μg/mL was added without DCs as indicated. After incubation at 37°C for 72 hours, CD4⁺ responder T-cell proliferation was quantified by CFSE dilution using flow cytometry, normalized to controls.

To test DC and IFNα effects, human monocyte–derived DCs from PBMCs of healthy female donors as described, or sorted from ascites of patients with ovarian cancer were added to Treg suppression assays (1 DC: 10 Tregs), plus IFNα (5,000 U/mL) (17). Isotype control, anti-IL6 antibody (10 μg/mL, BioLegend) or recombinant human IL6 (0.1 μg/mL, R&D Systems) were added at culture initiation. After incubation for 72 hours at 37°C, CD4⁺ naive T-cell proliferation was quantified by CFSE dilution using flow cytometry, normalized to controls.

A total of 5 × 10⁶ DCs from naïve WT or IFNAR^−/− mouse spleens were adoptively transferred intravenously into ID8 tumor-bearing IFNAR^−/− mice who were treated 1 day later with PBS or IFNα plus DD. DCs or Tregs were harvested, purified as described, and added to Treg suppression assays. After incubation for 72 hours at 37°C, supernatants from in vitro functional assays at experimental conclusion were analyzed for cytokines by Luminex (eBioscience) according manufacturer directions.

OVA-specific CD8⁺ (OT-I) or CD4⁺ (OT-II) T cells were purified from spleens and lymph nodes using EasySep Negative Selection Mouse CD8⁺ or CD4⁺ T Cell Enrichment (Stemcell Technologies), respectively, suspended in PBS at 1 × 10⁷/mL, and 1 × 10⁶/100 μL and transferred intraperitoneally into IFNAR^−/− mice 2 weeks after ID8 challenge.

Naïve IFNAR^−/− female mouse splenocytes were incubated with SIINFEKL 1 μg/mL or no peptide for 1 hour at 37°C, stained with 5.0 μmol/L (hi, peptide loaded cells) or 0.5 μmol/L (lo, control cells) CFSE for 10 minutes at 37 °C. Cells were washed, resuspended in PBS CFSE⁺(hi) and CFSE⁺(lo) cells mixed 1:1 and injected intraperitoneally (10 × 10⁶/100 μL) into tumor-bearing mice 1 day after the last drug treatment. Bone marrow–derived DCs (5 × 10⁶ cells/mouse) were transferred intraperitoneally 1 day before the first IFNα treatment as indicated. Mice were sacrificed approximately 16 hours later and relative change of peritoneal exudate CFSE⁺(hi) and CFSE⁺(lo) cells was analyzed by flow cytometry (e.g., Supplementary Fig. S2D). Specific lysis was calculated as [(%CFSE^lo − %CFSE^hi)/%CFSE^lo] × 100%.

Data are represented as mean and SEM unless otherwise specified. Results are representative experiments unless otherwise stated in the figure legends. Survival was estimated by Kaplan and Meier analysis and statistical significance for survival determined by log-rank test. Significance was defined as P < 0.05 by two-tailed Student t test unless otherwise specified.

In the phase 0/I clinical trial (patient characteristics in Supplementary Table S1), cohorts of three patients received one intravenous DD infusion of 9 or 12 μg/kg. One patient with metastatic pancreatic cancer treated in a separate protocol using weekly DD at 12 μg/kg was included in indicated analyses. There was no meaningful difference in immune outcomes in three patients receiving DD at 9 μg/kg versus four patients who received 12 μg/kg. Thus, pooled data are shown.

Tregs were defined as CD4⁺CD25^hi cells expressing Foxp3 by PCR and flow cytometry (e.g., Supplementary Figs. S1 and S2). Function was confirmed by suppression of T-cell proliferation, IFNγ, and cytotoxicity (e.g., Supplementary Figs. S2 and S3A–S3C). Elevated mean blood CD4⁺CD25^hi T-cell prevalence among CD3⁺CD4⁺ T cells (33.7%) dropped to 27.6% (P = 0.025) 3 to 7 days after one DD infusion (Fig. 1A; Supplementary Figs. S2 and S4). Mean blood CD4⁺CD25^hi T-cell concentration was simultaneously reduced (156/mm³ to 108/mm³, Fig. 1B; Supplementary Fig. S2). CD4⁺CD25^hi T-cell concentration fell in 6 of 7 patients, but not in patient 5 with metastatic OC who received DD at 12 μg/kg.

Mean blood IFNγ⁺CD8⁺ T-cell prevalence among CD3⁺CD8⁺ T cells (21.0%–46.5%) and concentration (137/mm³ to 198/mm³) increased after DD, including in patient 5, in whom DD did not reduce blood CD4⁺CD25^hi Tregs (Fig. 1B–D). Peripheral blood T cell Ki-67 increased, consistent with augmented proliferation (e.g., Fig. 2; Supplementary Fig. S4). Treatments were well tolerated with ≤ grade 2 toxicities (Supplementary Table S2). Patient 3 experienced clinical capillary leak syndrome with asymptomatic 5-pound weight gain 3 days after DD infusion that resolved within 1 week without specific intervention. The first patient in cohort 3, a 42-year-old male with metastatic bladder cancer, tolerated infusion of DD 15 μg/kg without toxicity. Dose escalation was then halted due to Hurricane Katrina and never resumed.

Patient 4, with metastatic ovarian cancer, demonstrated exceptional immunologic DD response with blood IFNγ⁺CD3⁺CD8⁺ T-cell prevalence among all CD3⁺CD8⁺ T cells (21%–37%) and concentration (100/mm³ to 233/mm³) increasing and remaining elevated at least 28 days after treatment (Fig. 1E). One DD infusion suppressed PBMC Treg function 30 days after DD, similar to reduced Treg function in patient 3, 23 days after DD (Supplementary Fig. S3D and S3E). Although this single-dose phase 0/I trial was designed only to assess immunity and safety, we received approval for additional DD for patient 4 with therapeutic intent. Peripheral blood CA-125, assessing OC tumor burden, dropped by approximately 60% 39 days after one DD infusion of 12 μg/kg. She then received six weekly DD infusions at 12 μg/kg, starting 39 days after the initial infusion, effecting another approximately 60% CA-125 reduction to within normal limits (Fig. 1F). Imaging demonstrated a mixed partial clinical response with complete resolution of most bony, lymphatic, and visceral metastases, although an inguinal lymph node increased (Fig. 1G), which recurred clinically four months later. After failing local external beam γ-irradiation therapy, she refused alternative treatments and died 13 months later from sepsis without evidence of tumor outside her left groin by CT of chest, abdomen and pelvis. Her blood CA-125 remained normal.

Although one DD infusion increased IFNγ⁺CD8⁺ blood T cells for at least 1 month in patient 4, CA-125–specific IFNγ⁺CD8⁺ T cells in peripheral blood by EliSpot (20) dropped from 84/100,000 T cells at weekly infusion initiation (39 days after initial infusion) to 28/100,000 T cells 1 week after the end of the 6 weekly infusions. There was no blood to test CA-125–specific T cells at trial initiation. A 72-year-old male with pancreatic cancer (patient 7) was treated in a distinct protocol with weekly DD, 12 μg/kg. After the first two infusions, blood CD3⁺CD8⁺IFNγ⁺ T cells increased as CD4⁺CD25^hi T cells decreased. However, by the fourth weekly infusion, peripheral blood cytokine-expressing T cells dropped precipitously (Fig. 2A and B; Supplementary Fig. S2). CA-19-9, a pancreatic tumor burden marker, in peripheral blood was 24,500 units/mL at therapy initiation and 27,400 units/mL after four weekly infusions. DD was thus discontinued after the fourth infusion. CA-19-9 was 44,000 units/mL 3 weeks after the final infusion, with radiographic evidence for worsening peritoneal carcinomatosis and ascites. These additional weekly DD infusions, and those in patient 4, elicited no new toxicities versus the single infusion.

To test DD immune mechanisms, we used the well-established ID8 ovarian cancer model. ID8 tumor cell challenge into WT mice increased CD4⁺CD25⁺Foxp3⁺ T-cell prevalence among tumor environmental CD45⁺CD3⁺CD4⁺ T cells and numbers in the tumor environment of peritoneal exudate cells (PECs), but not spleen or tumor draining mLNs, consistent with localized disease effects (Supplementary Fig. S2 and S5). DD also depletes Tregs in other mouse tumor models (21, 22). In ID8-bearing mice, one 5 μg DD injection depleted Tregs from spleen, mLN and PEC (Supplementary Fig. S6) and maximally depleted Tregs by 2 days (Supplementary Fig. S7A), consistent with DD effects in other mouse tumors (21, 22). A total of 5 μg/mouse DD given weekly modestly but significantly prolonged survival (median 87 days vs. 77 days with PBS; Fig. 3A). DD increased tumor microenvironmental OVA (tumor)-specific CD8⁺ T-cell prevalence among tumor environmental CD45⁺CD3⁺CD8⁺ T cells without change in cytotoxicity (Fig. 3B and C; Supplementary Fig. S2), consistent with improved antitumor immunity to improve survival (16).

Because DD contains human IL2, it could affect treatment independent of diphtheria toxin or through human IL2. Mice challenged with ID8 were treated with DD, equimolar human or mouse IL2, or PBS control with the same DD dose and schedule as before. Neither human nor mouse IL2 promoted a survival benefit (Supplementary Fig. S7B) consistent with treatment effects through the fusion toxin.

Ovarian cancer phase II DD clinical data and melanoma trials data suggested DD alone is insufficient clinically (8, 10, 14, 23). We hypothesized IFNα could boost Treg depletion efficacy by activating effector T cells, slowing depleted Treg recovery, and/or improving antigen presenting cell functions based on known IFNα effects (24). DD plus IFNα significantly prolonged survival after ID8 challenge (median 108 vs. 82 days for DD, 87 days for IFNα; Fig. 4A). Combination efficacy was abolished in RAG1^−/− mice (Fig. 4B), suggesting adaptive immunity contributions. Untreated ID8 challenged, syngeneic IFNα receptor IFNAR^−/− mouse survived comparable to WT (Fig. 4C), suggesting no significant role for endogenous type I IFNs anti-ID8 immunity, as in some cancers (25).

IFNα did not alter Treg function in spleen or mLN of ID8-bearing mice (Fig. 4D and E). However, combined with DD, IFNα significantly increased Treg suppression in spleen but not mLN, and simultaneously reduced function (not numbers) of tumor microenvironmental Tregs (Fig 4D–F), suggesting compartment-specific effects. IFNα plus DD increased DC prevalence and numbers among tumor microenvironmental CD45⁺ cells in PEC greater than either agent alone (Supplementary Fig. S8A and S8B). DCs from mLN of untreated, ID8 tumor-bearing mice significantly suppressed T-cell proliferation in vitro (Supplementary Fig. S8C) suggesting DC dysfunction as described previously (26). IFNα upregulated DC CD86 (Supplementary Fig. S8D), as reported in noncancer models, whereas DD upregulated DC CD80 (Supplementary Fig. S8E), suggesting differential DC maturation effects (27). In support, DCs from IFNα-treated mice did not suppress T-cell proliferation in vitro whereas DCs from DD-treated mice did. DD did not alter IFNα effects on DC-mediated T-cell proliferation suppression (Supplementary Fig. S8C).

IFNα can alter Treg function indirectly in vitro (28), which we confirm and show that CD11b⁺CD11c⁺ DCs from naïve mouse spleens mediate this effect (Supplementary Fig. S8F). To test in vivo effects, we transferred WT or IFNAR^−/− DCs into ID8-challenged IFNAR^−/− mice and administered DD plus IFNα. IFNAR^−/− DC transfer did not alter Treg functions, whereas WT DC recipients experienced significantly reduced tumor microenvironmental Treg function (Fig. 5A), consistent with indirect IFNα effects on DCs in vivo to reduce Treg suppression.

Naïve DCs, but not Tregs alone, produce detectable IL6 and IL10 in vitro. Coculture of naïve DCs and Tregs increased IL6 and reduced IL10. IFNα tripled IL6 and slightly decreased IL10 and increased IL6 further in DC alone (Fig. 5B), implicating IFNα-induced DC IL6 as a Treg inhibitor. In support, cocultured DC had no impact on Treg suppression, but IFNα significantly reduced Treg suppression with DC coculture, which was reversed by anti-IL6 antibody (Fig. 5C). In further support, recombinant IL6 (Fig. 5B) inhibited Treg suppression without DC (Fig. 5D). DC CD80 and CD86 can alter Treg suppression (29), but neither anti-CD80 nor anti-CD86 antibodies affected IL6-mediated Treg suppression (Fig. 5E). To test in vivo IL6 effects we challenged IFNAR^−/− mice with ID8, transferred WT DCs and treated with DD + IFNα ± anti-IL6. Consistent with in vitro findings, compromised Treg function was reversed in vivo by anti-IL6 (Fig. 5F).

Neither DD nor IFNα alone nor combined altered CD8⁺ T-cell prevalence or numbers among tumor environmental CD45⁺ cells in ID8-challenged mice (Supplementary Fig. S9A and S9B). DD alone had little effect on T-cell activation. In contrast, IFNα alone significantly increased CD8⁺ T-cell activation in PEC, but DD added no further benefit (Supplementary Fig. S2 and S9C). Consistent with activation results, DD had little impact on CD8⁺ T cell IFNγ whereas IFNα alone significantly increased CD8⁺ T cell IFNγ in PEC or mLN (Supplementary Fig. S2 and S9D).

DD increased tumor OVA (antigen)-specific CD8⁺ T cells, whereas IFNα did not (Fig. 6A). Each agent individually, but noncooperatively, boosted antigen-specific T-cell activation (Fig. 6B). To test in vivo antigen-specific cytotoxicity, we transferred OVA-specific CD8⁺ OT-I T cells one day before treatments in ID8-bearing WT mice. DD had little effect on tumor microenvironmental tumor antigen-specific CD8⁺ T-cell cytotoxicity, whereas IFNα significantly improved in vivo tumor antigen-specific cytotoxicity that was not further enhanced by DD (Fig. 6C; Supplementary Fig. S2). These data demonstrate that DD and IFNα mediate distinct CD8⁺ T-cell effects through complementary actions.

To test if IFNα acted directly on T cells in vivo, we transferred OVA-specific CD8⁺ OT-I T cells (with intact IFNα signaling) into ID8-OVA-challenged IFNAR^−/− mice. IFNα significantly boosted tumor environmental antigen-specific cytotoxicity (Fig. 6D; Supplementary Fig. S2), confirming direct IFNα effects on CD8⁺ T cells. Contrasting with CD8⁺ T cells, individual and combination treatments had little effect on CD4⁺ T-cell numbers among tumor environmental CD45⁺ T cells, cytokine production or activation as detected by CD69 (Supplementary Fig. S10A–S10D), suggesting differential IFNα activation of CD4⁺ versus CD8⁺ T cells. Nonetheless, CD4⁺ OVA-specific OT-II T cells significantly boosted tumor antigen-specific cytotoxicity of cotransferred CD8⁺ OT-I cells in ID8-OVA tumor-bearing IFNAR^−/− mice after treatment with IFNα, suggesting some CD4⁺ T-cell contributions to antitumor immunity. No significant effect in PEC was seen (Fig. 6E and F), demonstrating IFNα-induced cooperation between CD4⁺ and CD8⁺ T cells in a compartment-specific manner without a requirement for accessory cell participation (although such additional effects are not excluded). IFNα + DD also significantly increased tumor microenvironmental DC greater than either agent alone (Supplementary Fig. S10A and S10B).

Two patients in whom treatment failed in our phase II trial of single-agent DD were enrolled in a rollover phase II trial of weekly subcutaneous pegylated IFNα2a 180 μg added to failed DD, which was continued at 12 μg/kg intravenously every 3 weeks, given on the same day as pegylated IFNα2a. Adding weekly pegylated IFNα2a to clinically failed DD effected a long-lasting clinical response in patient 103, but she subsequently interrupted treatment for personal reasons for 2 months after cycle 12, after which it was no longer effective (Fig. 7A). Adding weekly pegylated IFNα2a to clinically failed DD stabilized rising CA-125 for 2 additional months and delayed clinical progression for 33 weeks in patient 105 (Fig. 7B).

Both patients on DD plus pegylated IFNα2a experienced grade I fatigue. Patient 103 experienced grade 3 nonautoimmune anemia requiring blood transfusions, which did not occur on DD alone (Supplementary Table S2).

We did not detect consistent effects on numbers or activation of lin⁻CD11c⁺DR⁺ myeloid DC or lin⁻CD11c⁻CD123⁺DR⁺ plasmacytoid DCs in peripheral blood of these two patients, and there were too few DCs for functional studies. There was a trend for pegylated IFNα2a to increase serum IL6 in patient 103 and to decrease blood Tregs better than DD alone (Fig. 7C and D) but low numbers precluded statistical analyses and patient 103 did not have data from 7 days after DD alone for comparison. Insufficient material was available for further studies of cells from either patient. We thus obtained flow-cytometrically sorted lin⁻CD11c⁺DR⁺ DC from ascites of three patients with ovarian cancer not in this trial. IFNα alone and DC alone did not affect in vitro Treg suppressive function, whereas Treg suppressive function was significantly reduced by adding IFNα to DCs (Fig. 7E). We generated monocyte-derived DCs from normal donors and again found that DCs or IFNα alone had no effect on Treg suppression whereas IFNα reduced Treg suppression in the presence of these DC in an IL6-dependent manner, reproduced by recombinant IL6 without DC or IFNα (Fig. 7F).

Tumor antigen-specific immunity is easily demonstrable in ovarian cancer (12), suggesting potential susceptibility to immunotherapy, consistent with the observation that T-cell infiltration in ovarian cancer predicts improved survival (11). Nonetheless, ovarian cancer immune-based therapies, including immune checkpoint blockade, have generally been clinically ineffective (12, 30).

Tregs reduce immunity and survival in ovarian cancer (5, 31) and can impede immunotherapy in distinct cancers (1, 7). Treg reduction to bolster antitumor immunity (7, 32, 33) has not been clinically successful. DD is a fusion protein of human IL2 and diphtheria toxin that kills IL2 receptor-expressing cells (34). DD depletes Tregs in human patients with cancer, can enhance efficacy of a cancer vaccine, and depletes Tregs and improves antitumor immunity and survival in distinct mouse cancer models (13, 21, 22). DD depletes Tregs and provides some clinical benefits in human metastatic melanoma (8, 10). Thus, Treg depletion is a candidate cancer immunotherapy strategy. As we showed that Tregs potently inhibit ovarian cancer–specific immunity and portend reduced ovarian cancer survival (5), Treg depletion to treat ovarian cancer merits consideration.

Our phase 0/I trial showed that one DD infusion reduces functional Tregs in blood of patients with various malignancies, associated with improved immunity, consistent with predicted immunopathogenic potential for human Tregs (1, 5). DD was well tolerated without toxicities seen in DD-treated patients with T-cell malignancies, similar to excellent tolerability in melanoma trials (8, 10, 34). Toxicities could reflect the unique biology of the malignant T cells. Nonetheless, our phase II ovarian cancer trial of DD (14) failed to replicate the significant clinical response or long-term Treg depletion seen in ovarian cancer patient 4 in the phase 0/I trial reported here. As Treg depletion with DD alone is not curative in any cancer thus far, and efficacy could be dampened by depleting antitumor immune cells as we found here with frequent administrations, or from rapid Treg regeneration, we sought easily translated means to improve Treg efficacy for proof of concept.

We identified candidate mechanisms in murine ovarian cancer suggesting utility of IFNα plus Treg depletion with potential for clinical translation. IFNα plus DD significantly reduced tumor microenvironmental Treg function without reducing phenotypic Treg numbers and did not compromise Treg function outside the tumor microenvironment, suggesting tumor environmental-specific mechanisms. Local Treg suppression could also mitigate adverse autoimmune effects. In vitro murine studies identified IFNα-induced IL6 as a mediator of reduced Treg suppression, supported by in vitro human cell studies, the effects of which appear indirect through IL6 from DC or other myeloid cells. Thus, local IL6 induction from DC could improve antitumor immunity by blunting Treg suppression. However, IL6 is generally detrimental to anticancer immunity (35). Therefore, IL6 production would optimally be generated proximate to Tregs. Although our data suggest an indirect IFNα effect on limiting Treg suppression, IFNα could also have direct effects on ovarian cancer Treg that affect distinct functions such as their proliferation (28).

There is considerable interest in how DC dysfunction in ovarian cancer reduces immune surveillance (17, 19, 26). DD can make adoptively transferred DC tolerogenic in patients with melanoma (36). Our data suggest that IFNα inhibits DC-mediated T-cell suppression. IFNα could thus mitigate this deleterious DD effect to improve clinical outcomes, which requires further investigation as do other agents with such potential. We unexpectedly also found DD plus IFNα significantly increased the prevalence and absolute numbers of tumor microenvironmental DC, which could be further therapeutically useful. Capillary leakiness induced by the IL2 moiety of DD combined with known capacity of Tregs to suppress DC activation and numbers in vivo which is known to control myeloid cell populations in vivo (37) could contribute to such an effect, versus altered DC survival or migratory capacity, areas in need of additional investigation.

DD improved antigen-specific immunity, consistent with data from other cancer models (7). IFNα did not increase antigen-specific CD8⁺ T-cell numbers, but increased their effector functions; however, that alone was clinically not significant in this model. In contrast, DD did not boost antigen-specific CD8⁺ T-cell cytotoxicity, but increased antigen-specific CD8⁺ T-cell numbers, complementing IFNα effects. Observed compartment-specific effects could reflect increased local tumor-specific CD8⁺ or CD4⁺ T-cell numbers, IFNα-mediated facilitation of T-cell interactions, or alterations in factors produced by microenvironmental cells, among other considerations that could be exploited in local treatments. We assessed T-cell activation by CD69 expression, which measures acute activation from treatment versus other markers including CD44/CD62 L that report chronic activation. Thus, although we noted important differences in functions and cell–cell cooperation between CD4⁺ and CD8⁺ T cells, other differences remain possible. Defining specific mechanisms for how IFNα versus Treg depletion improves tumor-specific immunity helps identify rational treatment combinations. We postulate that DD improves antitumor T-cell functions indirectly such as through alleviating Treg suppression whereas IFNα acts directly on CD8⁺ T cells to promote antitumor factors such as IFNγ and cytolytic molecules such as granzymes, but could improve effects indirectly by alleviating microenvironmental Treg suppression. Immune checkpoint blockade could be useful in combinations with DD or IFNα by reducing T-cell exhaustion, promoting antitumor functions or reducing Treg suppression, all of which have been seen or postulated as their mechanisms of action.

IFNα alone does not treat ovarian cancer (38, 39) but warrants further study combined with Treg depletion based on our data. IFNα doses we used were tolerable, but because of potential adverse type I IFN effects, additional Treg managing adjuncts are also worth identifying. In this regard, CD25 is highly expressed on activated antitumor T cells and Tregs. It highly binds IL2, helping DD target Tregs. Unfortunately, CD25 also targets antitumor T cells for DD-mediated elimination with repeated dosing as we noted in patients 4 and 7. Antitumor T cells express CD122 in the IL2 intermediate affinity receptor much higher than Tregs. CD122-biased IL2 preferentially activated T cells over Tregs to treat ovarian cancer (40) and could thus combine with IFNα alone or plus Treg depletion.

Treg management in humans is complicated by lack of Treg-specific targets. DD targets IL2 receptor–expressing cells, not just Tregs, which we show here can also deplete tumor-specific T cells. How DD preferentially depletes Tregs is unclear. We reported that anti-CD25 antibody and DD mediate distinct clinical outcomes in different preclinical cancer models despite depleting equivalent numbers of Tregs (12). This observation is consistent with lack of Treg-targeting agent specificity and underscores how mechanistic insights could help improve current Treg mitigation strategies or identify optimal agents.

Two DD melanoma trials failed to demonstrate Treg depletion, although a third metastatic melanoma DD trial showed Tregs depletion and clinical benefit (8, 10, 41, 42). Mechanisms for Treg depletion variability with DD are unclear. We also noted less Treg depletion in our phase II (14) versus this phase 0/I trial. Discrepancies in melanoma trials appear unrelated to prior treatment regimens, including recombinant IL2 (8, 10). Studies of differences in DD doses or schedule are inconclusive. A laboratory-made DD equivalent significantly and durably depleted Tregs in bone marrow in a mouse leukemia model (43), suggesting effects could be compartment specific, which could affect outcomes in specific scenarios. Manufacturing issues (such as those that halted our phase II trial) could affect DD efficacy. DD from an improved manufacturing process, renamed E7777, is in final stages of regaining FDA approval. Thus, additional DD-like approaches could be pursued.

One study suggested Treg depletion will not improve immune checkpoint blockade using anti–PD-1 or anti–CTLA-4, but investigators used Foxp3^DTR mice in which Tregs were essentially entirely depleted (44), a strategy not currently possible in humans. We reported distinct immune outcomes in preclinical cancer models by specific Treg depletion in Foxp3^DTR mice versus DD (7). Understanding related mechanisms could help identify strategies for more effective immunotherapy through Treg mitigation and improve efficacy of existing agents.

We show that adding a second agent to a clinically failing agent is reasonable if the clinically failing agent is achieving its immunological goal. Furthermore, we show that it is not required for individual agents to have significant activity to elicit improved human clinical efficacy in combination, a strategy often seen in animal models, but rarely encountered in human trials. Although our mouse preclinical modeling supports our conclusions regarding human treatment effects, the numbers of patients studied here is relatively small, requiring additional studies.

T.J. Curiel reports nonfinancial support from Eisai during the conduct of the study. No disclosures were reported by the other authors.

S.R. Thibodeaux: Data curation, formal analysis, supervision, validation, investigation, writing–original draft, writing–review and editing. B. Barnett: Data curation, formal analysis, investigation, writing–original draft. S. Pandeswara: Formal analysis, investigation. S.R. Wall: Investigation. V. Hurez: Resources, data curation, formal analysis, supervision, investigation, methodology, project administration. V. Dao: Investigation. L. Sun: Investigation. B.J. Daniel: Resources, data curation, formal analysis, investigation. M.J. Brumlik: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. J. Drerup: Investigation. Á. Padrón: Investigation. T. Whiteside: Data curation, formal analysis, supervision, validation. I. Kryczek: Data curation, formal analysis, validation, investigation, writing–review and editing. W. Zou: Conceptualization, data curation, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. T.J. Curiel: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.

This study was supported by FD003118, CA100425, CA105207, CA054174, CA164122, Fanny E. Rippel Foundation, Voelcker Trust, Holly Beach Public Library Association, STARS, Owens Foundation, the UTHSA Daisy M. Skinner endowment, Louisiana Board of Regents RC/EEP, The Ovarian Cancer Alliance of Greater Cincinnati.

Thanks to Alan Mita and Kevin Hall for referring patients. Thanks to Gabby Rennebeck, Xiuhua Sun, Dakshayni Lomada, Mark Kious, Sara Ludwig, Duane Jeansonne, and Tahiro Shin for technical help.

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