Abstract
To evaluate the safety and feasibility of preoperative locoregional cytokine therapy (IRX-2 regimen) in early-stage breast cancer, and to evaluate for intratumoral and peripheral immunomodulatory activity.
Sixteen patients with stage I–III early-stage breast cancer (any histology type) indicated for surgical lumpectomy or mastectomy were enrolled to receive preoperative locoregional immunotherapy with the IRX-2 cytokine biological (2 mL subcutaneous × 10 days to periareolar skin). The regimen also included single-dose cyclophosphamide (300 mg/m2) on day 1 to deplete T-regulatory cells and oral indomethacin to modulate suppressive myeloid subpopulations. The primary objective was to evaluate feasibility (i.e., receipt of therapy without surgical delays or grade 3/4 treatment–related adverse events). The secondary objective was to evaluate changes in stromal tumor–infiltrating lymphocyte score. The exploratory objective was to identify candidate pharmacodynamic changes for future study using a variety of assays, including flow cytometry, RNA and T-cell receptor DNA sequencing, and multispectral immunofluorescence.
Preoperative locoregional cytokine administration was feasible in 100% (n = 16/16) of subjects and associated with increases in stromal tumor–infiltrating lymphocytes (P < 0.001). Programmed death ligand 1 (CD274) was upregulated at the RNA (P < 0.01) and protein level [by Ventana PD-L1 (SP142) and immunofluorescence]. Other immunomodulatory effects included upregulation of RNA signatures of T-cell activation and recruitment and cyclophosphamide-related peripheral T-regulatory cell depletion.
IRX-2 is safe in early-stage breast cancer. Potentially favorable immunomodulatory changes were observed, supporting further study of IRX-2 in early-stage breast cancer and other malignancies.
In a recent phase III trial, immune-checkpoint inhibition improved clinical outcomes in metastatic triple-negative breast cancer; however, benefit was limited to patients whose tumors were positive for the PD-L1 biomarker. Cytokine-based therapies may facilitate antigen presentation, priming, and PD-L1 upregulation. IRX-2 contains multiple cytokines at physiologic doses, generated by purification of cytokine products from ex vivo–stimulated lymphocytes. Here, we report a locoregional approach for early-stage breast cancer whereby cytokines are injected in the periareolar tissue, which is known to communicate directly with tumor-draining lymph nodes via lymphatics. This pilot study demonstrated that the regimen is well tolerated and associated with favorable changes in the tumor microenvironment, including PD-L1 upregulation as well as expansion of tumor-infiltrating lymphocytes. These findings support further study of locoregional cytokines in early-stage breast cancer, potentially in combination with immune-checkpoint inhibition and chemotherapy for high-risk triple-negative breast cancer.
Introduction
Despite systemic and locoregional therapeutic advances for early-stage breast cancer (ESBC), in 2018 metastatic recurrence was associated with 41,400 fatalities in the United States (1). Immunotherapy represents a promising new therapeutic direction for several reasons. First, the extent of stromal tumor–infiltrating lymphocytes (sTILs) has been established as a favorable prognostic and predictive marker, especially in immunogenic histologic types including triple-negative breast cancer (TNBC; refs. 2, 3). Second, immune-checkpoint blockade is clinically active in metastatic TNBC, with a phase III study demonstrating improved progression-free survival with the addition of atezolizumab to nab-paclitaxel among tumors bearing PD-L1–positive immune cells (4), and preliminary clinical trials of anti-programmed death 1 (PD-1) or its ligand (PD-L1) demonstrating response rates of 5% to 25% as monotherapy (5–7). In early-stage TNBC, the addition of pembrolizumab (anti–PD-1) to anthracycline-based neoadjuvant chemotherapy was associated with a near-tripling of pathologic complete response (pCR) rate in the I-SPY2 clinical trial, and an improvement in pCR with carboplatin-containing neoadjuvant therapy in a phase III trial (KEYNOTE-522, NCT03036488; ref. 8).
Combination immunotherapy approaches offer the opportunity to increase likelihood of clinical benefit and expand benefit to less immunogenic tumor types such as PD-L1–negative TNBC. Here, we report a cytokine-based immunotherapy to enhance antigen presentation, TIL density, and TIL activation in early-stage disease. Cytokine therapy is currently being evaluated in clinical trials in breast cancer; however, most of these trials evaluate individual cytokines with/without anti–PD-1/L1 (9). In light of the complexities of cytokine biology and the potential for mechanistic interactions between cytokines, an underexplored alternative approach is combination cytokine therapy.
The IRX-2 cytokine product is a combination of various cytokines delivered at physiologic doses (IL2, IL1β, IFNγ, TNFα, and others) that replicate endogenous cytokine release following T-cell stimulation. IRX-2 was initially developed as a locoregional therapy for study in head and neck squamous cell carcinomas (HNSCCs), inspired by preceding clinical trials whereby low-dose perilymphatic injection of recombinant IL2 was found to regress tumors and improve both disease-free survival (DFS) and overall survival (OS; 5-year OS 63 v. 55%, P < 0.03; ref. 10). In preclinical models, the physiologically derived cytokine mixture (IRX-2) resulted in tumor regressions greater than that observed with individual cytokine components (11, 12), IRX-2 was later shown to selectively stimulate effector T cells over regulatory T cells, promote the cytolytic functions of natural killer (NK) cells (13, 14), facilitate activation of antigen-presenting cells (15), protect T cells from activation-induced apoptosis (16, 17), and enhance the effect of peptide vaccines (18). In a murine model (SCC7), the efficacy of IRX-2 was enhanced by the addition of anti–PD-L1 therapy in a dose-dependent manner, with an increase in tumor-specific CD8+ cytotoxicity (19). Finally, IRX-2 was shown to be most effective when administered repeatedly for up to 10 doses, leading to the development of a clinical regimen that includes 10 days of perilymphatic mixed cytokine injection (18).
IRX-2 is being clinically evaluated in gastric, hepatocellular, cervical, HNSCC, and breast cancers. Data from phase I and phase IIa trials in HNSCC have been published. In these trials, IRX-2 was coadministered in conjunction with ancillary therapies designed to maximize immune priming, including an oral cyclooxygenase 1/2 inhibitor (indomethacin) and an intravenous dose of low-dose cyclophosphamide (Cy). Cyclooxygenases are known to suppress immunity via production of prostaglandins, which can inhibit Th1-type cytokine production (20) and cytotoxic T-lymphocyte activity (21), but also promote the development of tumor-associated macrophages/myeloid-derived suppressor cells (22–24). Low-dose Cy is known to deplete T-regulatory cells (25), enhance vaccine immune response (26), facilitate immunogenic cell death (27), and modulate the microbiome (28–30). The combination of IRX-2, indomethacin, and Cy was tolerated in a phase I trial in locally recurrent HNSCC with reports of one complete response (CR) and five of eight stable disease at first radiographic evaluation. Mild/moderate indomethacin-related gastrointestinal toxicities were reported (31).
In a subsequent biomarkers-driven phase IIa trial, 27 resectable HNSCC subjects received preoperative IRX-2 with Cy and indomethacin (32, 33). The regimen was safe, and 80% of evaluable subjects experienced stable disease (RECIST1.1), with 4 subjects experiencing 10% to 20% reduction in tumor dimension over the short therapeutic window. Importantly, the therapy was associated with an increase in TILs among 80% of evaluable tumors (20/25, P < 0.001), also with increases in tumor necrosis (P < 0.001; ref. 32). Multispectral immunofluorescence (PerkinElmer) and RNA analysis (NanoString) was used to further characterize immune infiltrates. In these experiments, membrane PD-L1 staining was found to be increased in four of seven evaluated tumors. IRX-2 was also associated with upgregulation in genes expressed by T cells, B cells, as well as NK cells, as well as upregulation of various chemoattractants, including CXCL12, CCL2, CCL14, CCL19, and CCL21 (34). A phase II randomized clinical trial (INSPIRE) is ongoing to evaluate the contributory effects of IRX-2/indomethacin/Cy versus indomethacin/Cy control, and to estimate DFS/OS (NCT02609386).
Here, we present the final results of a phase Ib trial of the IRX-2 regimen in ESBC. Conserved and well-defined anatomic relationships exist between breast tumors, overlying dermal tissue, and tumor-draining axillary lymphatics, with tumor-draining lymph nodes of breast tumors being located principally in the ipsilateral axillary lymph nodes. During routine sentinel lymph node localization, isosulfan blue dye and/or radioactive colloid is injected into the periareolar dermal tissue (35), which communicates with sentinel axillary lymph nodes, allowing for successful localization of sentinel lymph nodes in >95% of cases. These anatomic considerations could be leveraged to deliver cytokines or other immunotherapies via the afferent lymphatics into lymph nodes that communicate directly with the tumor. This therapeutic approach has not been studied extensively in humans. The primary objective of this study is to evaluate the safety and feasibility of this novel treatment during a brief window of opportunity between diagnosis and definitive surgical resection. We also aimed to confirm the known biomarker changes observed in the HNSCC trial. If active, a randomized phase II study would be designed to further develop IRX-2 in ESBC.
Patients and Methods
Study design and participants
Between May 2016 and May 2018, women with biopsy-proven stage I–III newly diagnosed breast cancer planned for definitive surgical resection (either lumpectomy or mastectomy) were considered for enrollment at the Providence Cancer Institute in Portland, OR. Inclusion criteria included: resectable >0.5 cm primary tumor, Karnofsky performance status of ≥70%, adequate organ function, absence of steroid-dependent medical conditions, and absence of prior immunotherapy or other treatments for their breast cancer. Any breast cancer subtype [hormone receptor (HR)–positive/negative; human epidermal growth factor receptor 2 (HER2)–positive/negative] was allowed. HR positivity was defined as ≥1% expression of either estrogen receptor (ER) or progesterone receptor (PR) by immunohistochemistry; HER2 positivity was defined as either 3+ staining by IHC or positivity by fluorescence in situ hybridization per local laboratory guidelines.
Participants received the IRX-2 regimen preoperatively before advancing to definitive surgical resection. Sixteen patients (n = 16) were enrolled in the clinical trial. The study was performed in accordance with ethical principles of the Declaration of Helsinki and the International Conference on Harmonization of Good Practice, and approved by the Providence Health and Services Institutional Review Board. All participants provided written informed consent.
Study agent and intervention
IRX-2 biological is produced from ex vivo phytohemagglutinin (PHA)-based stimulation of pooled donor peripheral blood leukocytes under current good manufacturing procedures (CGMP) by IRX Therapeutics (15). Briefly, pooled buffy coat cells (derived from whole blood donations from FDA-licensed blood centers) were centrifuged to permit concentration of mononuclear cells. They are then cultured with PHA and ciprofloxacin for 2 hours before washing. PHA is a plant glycoprotein that binds to the T-cell receptor (TCR) and results in preferential production of Th1 cytokines. Following 24 hours in culture, cellular elements and PHA were removed by centrifugation and filter sterilization. In compliance with FDA and CGMP standards, cytokine analysis was conducted to ensure consistency across production lots (CV215005AU, CV216012AU, CV216011A1, and CV217005AU). The measured concentration range of IL1β was 300–1,400 pg/mL, IL2 was 4,000–8,000 pg/mL, IFNγ was 1,000–3,800 pg/mL, and TNFα is 1,000–4,300 pg/mL. Additional measurable cytokine constituents included IL6, IL8, granulocyte colony stimulating factor (G-CSF), and granulocyte monocyte colony stimulating factor (GM-CSF). The following cytokines were not detectable: IL3, IL4, IL5, IL7, and IFNα.
The IRX-2 regimen is graphically illustrated in Fig. 1A–C. Subjects received perilymphatic IRX-2 at a dose of 1 mL × 2 subcutaneous intra-periareolar injections daily for 10 days over a 15-day presurgical window period. The 10-day course was selected based upon pharmacodynamic activity in preclinical models (18), to allow for maximum cytokine exposure within a reasonable presurgical window of opportunity, and to be consistent with preceding and ongoing clinical trials of the IRX-2 regimen across other malignancy types (36). Injections were administered along the tumor axis and 90° (alternating daily clockwise/counterclockwise). Consistent with the phase I/II preceding trial, subjects also received a single dose of intravenous Cy (300 mg/m2, day 1), oral indomethacin (25 mg three times daily, days 1–21), and supportive medications (multivitamins, proton pump inhibitor). Definitive surgical resection was scheduled within 12 days of completing IRX-2 injections.
Peripheral blood isolation for flow cytometry
Peripheral blood was collected via venipuncture into sodium heparin vacutainer tubes on day 1 (baseline, prior to cyclophosphamide), day 4 (prior to IRX-2 injection), twice during IRX-2 injections, prior to definitive surgical resection, and at a 30-day postoperative safety visit. Flow cytometry was performed on fresh peripheral blood mononuclear cells (PBMC) as previously described within 24 hours of collection (37). For each test panel, 50 μL of blood was incubated with antibody cocktail for 30 minutes at room temperature. Next, blood was incubated with 1 mL 1× Becton Dickinson (BD) FACS Lysing Solution for 15 minutes at room temperature. Lysed blood was washed twice with 2-mL flow wash buffer (1% BSA, 0.1% sodium azide, 10 U/mL heparin in HBSS), and resuspended in 300-μL buffer for acquisition. All samples were processed on a Becton Dickinson LSRFortessa utilizing an HTS unit. Analysis was performed using FlowJo software (version 10) and GraphPad Prism (version 7.c).
Pathology assessment of TILs
Per standard of care, diagnostic biopsy core tumor specimens were formalin-fixed and paraffin-embedded (FFPE). Five-micrometer sections were then stained for routine hematoxylin and eosin (H&E) TIL analysis. sTILs were scored by two blinded study pathologists, according to the published 2015 San Antonio International TILs Working Group criteria (38). Blinded scores were averaged for final analysis, and concordance rates were evaluated. Tumors and immune cells (ICs) were assessed for PD-L1 expression using the Ventana PD-L1 (SP142) assay and scored independently by two trained study pathologists. To serve as an informal untreated control, FFPE tissue from core biopsies and resections of a cohort of 14 contemporary untreated ESBCs were obtained, and analyzed serially for sTIL score and SP142 PD-L1 status.
To permit exploratory single-cell IC PD-L1 expression analysis, tumor cells and ICs were assessed for PD-L1 expression using a multispectral immunofluorescence (mIF) approach with the PerkinElmer Vectra platform and validated antibodies, using a methodology previously described by PerkinElmer and our group (39). Cytokeratin (CK) was used for tissue segmentation into tumor and stromal compartments. ICs were defined as lymphocytes (CD3+CD8−, CD3+CD8+, CD3+FOXP3) or macrophages (CD163+) using DAPI for nuclear counterstain.
Measurement of tumor tissue T-cell repertoire and RNA tumor tissue expression
For each sample, DNA and RNA were extracted from ten 5-μm slides of tumor-bearing FFPE tissue using a Qiagen all-prep kit. Extracted DNA and RNA were quantified using Quant-iT dsDNA and RNA Assay Kits, respectively (Molecular Probes Life Technologies). DNA was submitted to Adaptive Biotechnologies for quantitative TCR DNA sequencing using previously described methods (40). Clonality was measured as 1 − (entropy)/log2 (number of productive unique sequences), with entropy taking into consideration clonal frequency. When available, paired lymph node specimens were used for DNA/RNA extraction and analyzed. RNA expression of various immune genes and gene signatures was analyzed using the NanoString nCounter platform with the PanCancer Immune Profiling probe set, which characterizes 770 genes from 24 different immune cell types. The nCounter software control (QC) metrics were used utilizing the imaging and binding density as well as positive control linearity and limit of detection parameters. Data were normalized based on positive controls and housekeeping codeset content.
Statistical methods
This was a phase Ib trial to evaluate the primary outcome of safety/tolerability, which was defined as the proportion of evaluable subjects who received at least seven injections of IRX-2 and 10 days of indomethacin and multivitamins with zinc in the absence of grade 3/4 treatment–associated adverse events, treatment-associated surgical delays, or withdrawal of consent. With a sample size of n = 16 evaluable patients, a tolerability boundary of at least n = 12/16 subjects was established to differentiate a null tolerability proportion of 56%, against an alternative tolerability proportion of 81% in a two-sided test of proportions with a type 1 error of 10% and computed power of 83%. In addition to the formal test of tolerability, point estimates of adverse event rates according to the common terminology criteria for adverse events version 4.0 (CTCAE 4.0) were also to be reported. A prespecified, secondary outcome was to evaluate treatment effects on associated stromal TIL infiltration by the 2015 San Antonio TILs Working Group Criteria. A paired t test of differences assuming equal variance was used to determine whether sTILs were significantly increased at a two-sided 5% significance level. Concordance was measured by Spearman rank correlation using Graph Prism software. Exploratory immune biomarkers were to be analyzed graphically and using statistical testing if the data permitted.
Results
Study population
All subjects (n = 16) were evaluable for safety, tolerability, and secondary/exploratory immunologic endpoints. The median age at diagnosis was 55 years (range, 45–78 years). The majority of subjects (n = 11/16, 69%) had HR-positive, HER2-negative disease, whereas four subjects had HER2-positive disease (n = 4/16, 25%) and one subject had TNBC (n = 1/16, 6%). Eight women (n = 8/16, 50%) had nodal involvement on final pathologic review. With a median follow-up of 21.9 months (as of April 1, 2019), zero subjects (n = 0/16, 0%) have experienced locoregional or distant recurrence. Table 1 summarizes the demographic and tumor characteristics of subjects enrolled on the trial.
Patient ID . | Age . | Estrogen receptor % . | Progesterone receptor % . | HER2 status . | Grade . | KI67 (%) . | Tumor size . | Lymph node involvement . | Pretreatment sTIL score . | Posttreatment sTIL score . |
---|---|---|---|---|---|---|---|---|---|---|
IRXB-001 | 78 | 100 | 100 | Negative | 2 | 11 | 1.9 cm | Yes | 8.75 | 12.5 |
IRXB-002 | 58 | 96 | 90 | Negative | 2 | 19 | 2.4 cm | Yes | 13.75 | 20 |
IRXB-003 | 40 | 98 | 100 | Negative | 2 | 17 | 3.0 cm | Yes | 1.75 | 1.75 |
IRXB-004 | 64 | 0 | 0 | Positive | 3 | 75 | 4.2 cm | Yes | 6.25 | 10 |
IRXB-005 | 48 | 98 | 92 | Negative | 1 | 7 | 1.7 cm | Yes | 3.75 | 5 |
IRXB-006 | 62 | 0 | 0 | Positive | 3 | 73 | 2.7 cm | Yes | 18.75 | 18.75 |
IRXB-007 | 45 | 98 | 100 | Negative | 3 | 55 | 3.7 cm | Yes | 18.75 | 20 |
IRXB-008 | 54 | 100 | 42 | Negative | 3 | 50 | 2.1 cm | No | 7.5 | 16.25 |
IRXB-009 | 56 | 100 | 100 | Negative | 2 | 11 | 0.7 cm | No | 1 | 5 |
IRXB-010 | 46 | 100 | 99 | Negative | 2 | 33 | 1.8 cm | No | 11.5 | 18.75 |
IRXB-011 | 52 | 69 | 0 | Positive | 3 | 38 | 1.5 cm | Yes | 6.25 | 5 |
IRXB-012 | 52 | 98 | 100 | Positive | 3 | 12 | 2.1 cm | No | 61.25 | 72.5 |
IRXB-013 | 59 | 100 | 7 | Negative | 3 | 87 | 2.2 cm | No | 31.25 | 55 |
IRXB-014 | 45 | 100 | 100 | Negative | 1 | N/A | 2.2 cm | No | 1 | 1 |
IRXB-015 | 61 | 100 | 50 | Negative | 2 | 30 | 1.9 cm | No | 3.5 | 7 |
IRXB-016 | 66 | 0 | 0 | Negative | 3 | 95 | 1.0 cm | No | 1 | 13.75 |
Patient ID . | Age . | Estrogen receptor % . | Progesterone receptor % . | HER2 status . | Grade . | KI67 (%) . | Tumor size . | Lymph node involvement . | Pretreatment sTIL score . | Posttreatment sTIL score . |
---|---|---|---|---|---|---|---|---|---|---|
IRXB-001 | 78 | 100 | 100 | Negative | 2 | 11 | 1.9 cm | Yes | 8.75 | 12.5 |
IRXB-002 | 58 | 96 | 90 | Negative | 2 | 19 | 2.4 cm | Yes | 13.75 | 20 |
IRXB-003 | 40 | 98 | 100 | Negative | 2 | 17 | 3.0 cm | Yes | 1.75 | 1.75 |
IRXB-004 | 64 | 0 | 0 | Positive | 3 | 75 | 4.2 cm | Yes | 6.25 | 10 |
IRXB-005 | 48 | 98 | 92 | Negative | 1 | 7 | 1.7 cm | Yes | 3.75 | 5 |
IRXB-006 | 62 | 0 | 0 | Positive | 3 | 73 | 2.7 cm | Yes | 18.75 | 18.75 |
IRXB-007 | 45 | 98 | 100 | Negative | 3 | 55 | 3.7 cm | Yes | 18.75 | 20 |
IRXB-008 | 54 | 100 | 42 | Negative | 3 | 50 | 2.1 cm | No | 7.5 | 16.25 |
IRXB-009 | 56 | 100 | 100 | Negative | 2 | 11 | 0.7 cm | No | 1 | 5 |
IRXB-010 | 46 | 100 | 99 | Negative | 2 | 33 | 1.8 cm | No | 11.5 | 18.75 |
IRXB-011 | 52 | 69 | 0 | Positive | 3 | 38 | 1.5 cm | Yes | 6.25 | 5 |
IRXB-012 | 52 | 98 | 100 | Positive | 3 | 12 | 2.1 cm | No | 61.25 | 72.5 |
IRXB-013 | 59 | 100 | 7 | Negative | 3 | 87 | 2.2 cm | No | 31.25 | 55 |
IRXB-014 | 45 | 100 | 100 | Negative | 1 | N/A | 2.2 cm | No | 1 | 1 |
IRXB-015 | 61 | 100 | 50 | Negative | 2 | 30 | 1.9 cm | No | 3.5 | 7 |
IRXB-016 | 66 | 0 | 0 | Negative | 3 | 95 | 1.0 cm | No | 1 | 13.75 |
Abbreviations: HER2, human epidermal growth factor receptor; sTIL, stromal tumor–infiltrating lymphocyte.
Safety and tolerability
The preoperative IRX-2 regimen met the primary endpoint of tolerability. Overall, most subjects (n = 13/16, 81.3%) met all prespecified tolerability criteria, whereas three subjects (3/16, 18.8%) failed to meet criteria owing to early termination of oral indomethacin. Subject 1 discontinued indomethacin at day 4 after experiencing grade 2 nausea and grade 1 emesis which resolved after discontinuation. Subject 3 discontinued oral indomethacin at day 10 after experiencing grade 1 cognitive disturbance, which was thought to be possibly attributed to indomethacin. Subject 4 discontinued oral indomethacin at day 3 after experiencing grade 1 nausea which resolved after discontinuation. All subjects (16/16, 100%) completed the protocol-specified 10 daily injections of IRX-2. No subjects (0/16, 0%) experienced treatment-related grade 3/4 toxicities. One subject (1/16, 6.3%) experienced a grade 3 vasovagal syncope episode, which was unrelated to therapy and attributed to a preexisting condition. No subjects (0/16, 0%) experienced treatment-related surgical delays or unanticipated surgical complications.
Consistent with the known adverse event profile of indomethacin, a substantial proportion of participants experienced grade 1/2 gastrointestinal events, including eight subjects (50%; 95% CI, 25%–75%) with grade 1/2 nausea, three subjects (19%; 95% CI, 4%–46%) with grade 1/2 abdominal bloating, two subjects (13%; 95% CI, 2%–38%) with flatulence, and two subjects (13%; 95% CI, 2%–38%) with emesis. Table 2 summarizes all toxicities occurring in > 15% of subjects and associated attributions.
Toxicity . | Grade 1/2a, any attribution (%) . | Grade 1/2, attributed to cyclophosphamidea (%) . | Grade 1/2, attributed to indomethacina (%) . | Grade 1/2, attributed to IRX-2a (%) . |
---|---|---|---|---|
Nausea | 13 (81%) | 3 (19%) | 8 (50%) | |
Bruising | 8 (50%) | 7 (44%) | ||
Fatigue | 9 (56%) | 5 (31%) | ||
Injection site reaction | 8 (50%) | 8 (50%) | ||
Abdominal cramping/bloating | 4 (25%) | 3 (19%) | ||
Increased ALT/AST/AlkP | 4 (25%) | |||
Anemia | 4 (25%) | 1 (6%) | 1 (6%) | |
Headache | 5 (31%) | 1 (6%) | ||
Hypokalemia | 4 (25%) | |||
Anorexia | 3 (19%) | |||
Diarrhea | 3 (19%) | 1 (6%) | 1 (6%) | |
Flatulence | 3 (19%) | 2 (13%) | ||
Injection site pain | 3 (19%) | 3 (19%) | ||
Vomiting | 3 (19%) | 2 (13%) | 2 (13%) | |
Constipation | 2 (13%) | |||
Dizziness | 2 (13%) | |||
Flu-like symptoms | 2 (13%) | 1 (6%) | ||
Hot flash | 2 (13%) | |||
Hyperglycemia | 2 (13%) | 1 (6%) | ||
Hypoglycemia | 2 (13%) | 1 (6%) | ||
Hyponatremia | 2 (13%) | |||
Nasal congestion/runny nose | 2 (13%) | |||
Other pain | 2 (13%) |
Toxicity . | Grade 1/2a, any attribution (%) . | Grade 1/2, attributed to cyclophosphamidea (%) . | Grade 1/2, attributed to indomethacina (%) . | Grade 1/2, attributed to IRX-2a (%) . |
---|---|---|---|---|
Nausea | 13 (81%) | 3 (19%) | 8 (50%) | |
Bruising | 8 (50%) | 7 (44%) | ||
Fatigue | 9 (56%) | 5 (31%) | ||
Injection site reaction | 8 (50%) | 8 (50%) | ||
Abdominal cramping/bloating | 4 (25%) | 3 (19%) | ||
Increased ALT/AST/AlkP | 4 (25%) | |||
Anemia | 4 (25%) | 1 (6%) | 1 (6%) | |
Headache | 5 (31%) | 1 (6%) | ||
Hypokalemia | 4 (25%) | |||
Anorexia | 3 (19%) | |||
Diarrhea | 3 (19%) | 1 (6%) | 1 (6%) | |
Flatulence | 3 (19%) | 2 (13%) | ||
Injection site pain | 3 (19%) | 3 (19%) | ||
Vomiting | 3 (19%) | 2 (13%) | 2 (13%) | |
Constipation | 2 (13%) | |||
Dizziness | 2 (13%) | |||
Flu-like symptoms | 2 (13%) | 1 (6%) | ||
Hot flash | 2 (13%) | |||
Hyperglycemia | 2 (13%) | 1 (6%) | ||
Hypoglycemia | 2 (13%) | 1 (6%) | ||
Hyponatremia | 2 (13%) | |||
Nasal congestion/runny nose | 2 (13%) | |||
Other pain | 2 (13%) |
Note: Transient treatment-unrelated grade 3 syncope was observed in one subject with prior history of vasovagal syncope.
aDefinitely or likely attributed, assessed by treating subinvestigator.
The IRX-2 periareolar injections were well tolerated, as evidenced by 100% of patients completing the protocol-directed 10-day treatment course. Half (n = 8/16, 50%; 95% CI, 25%–75%) of subjects experienced grade 1 injection reactions, often described as self-limited redness and swelling. Seven subjects (n = 7/16, 44%; 95% CI, 20%–70%) also reported grade 1 bruising at the site of IRX-2 injections. Subjects were given the opportunity to request topical lidocaine cream to minimize injection-associated pain; however, none of the subjects required this intervention.
Assessment of treatment-associated stromal TILs
The validated San Antonio TILs working group guidelines (38) were used to assess for changes in sTIL score associated with therapy. sTIL scores were independently scored by two blinded pathologists. Scores from the pathologists were highly correlated (Spearman r = 0.84, P < 0.0001) with fair interobserver agreement (Cohen linear weighted kappa = 0.53, P < 0.001). Figure 2A and B illustrates averaged pretreatment versus posttreatment sTIL scores. Consistent with previous reports, baselines sTIL scores were numerically lower for ER+ breast cancers relative to HER2+/TNBC breast cancers (average 9% vs. 19%; ref. 3). There was an increase in sTIL score associated with therapy (139%, P = 0.001, Wilcoxon matched-pairs rank test; Fig. 2). In this limited data set, there were no apparent associations of therapy-associated sTIL change with baseline characteristics including tumor histologic type, grade, or tumor proliferation index. For comparison, nonrandomized control cohort blinded sTIL scores from matched biopsy and resection tissue are illustrated in Supplementary Fig. S8 and show relative stability of sTIL score.
Assessment of PD-L1 expression
Serial tissue PD-L1 expression was assessed by chromogenic IHC (Ventana, SP142), mIF, and RNA expression (NanoString). The Ventana chromogenic IHC method characterizes tumors categorically according to the proportion of tumor area infiltrated by PD-L1–expressing ICs. Levels of positivity are: level IC0 (<1% ICs positive for PD-L1), level IC1 (1%–4%), IC2 (5%–9%), and IC3 (≥10%). At baseline in pretreatment specimens, 53% of biopsies were classified as IC0 (n = 8/15) and 47% as IC1 (n = 7/15). IRX-2 was associated with upclassification in 69% of evaluable tumors (n = 2/13 IC0→IC1, n = 2/13 IC0→IC2/3, n = 5/13 IC1→IC2/3), and not changed in 31% of subjects (n = 4/13; Supplementary Table S10b). No subjects were down-classified. Tumor cell–only PD-L1 expression by chromogenic IHC was negative in 100% of baseline specimens (n = 0/14) and 8% of posttreatment specimens (n = 1/13). For comparison, control cohort SP142 scores from matched biopsy and resection tissue are illustrated in Supplementary Fig. S8. In the controls, PD-L1 status was upclassified in 29% of cases (n = 4/14), and not changed in 71% of cases (n = 10/14).
mIF was used to characterize PD-L1 expression [measured in immunofluorescence (IF) units] on the level of individual cells across multiple regions of interest per tumor specimen. This method, while not validated for clinical use, allows for evaluation of treatment-associated changes across various cell phenotypes. Hierarchical mixed-effect regression analysis was used to evaluate treatment effect while adjusting for heterogeneity across regions of interest and patients. By this method, IRX-2 therapy was associated with a 1.8-fold increase in PD-L1 IF on tumor cells (CI 0.92,3.46; P = 0.09), 2.1-fold increase on CD3+ lymphocytes (CI 1.12,3.97; P = 0.02), and 2.26-fold increase on macrophages (CI 1.18, 4.34; P = 0.01).
To recapitulate the Ventana clinical PD-L1 assay using mIF, ICs were classified categorically as positive versus negative using an IF unit threshold of 2.6. This threshold was found to maximize the area under the curve (AUC 0.98) of the receiver operating curve, using blinded pathologist visual classification as the gold standard. By this method, IRX-2 was associated with increases in PD-L1+ ICs in 73% of subjects with matched specimens (n = 11/15), with a mean 5.4-fold increase (range, −0.65 to +70.0) in the number of PD-L1+ cells per region of interest (Table 3).
The NanoString platform was used to estimate average PD-L1 mRNA expression among tumor-bearing FFPE specimens. As illustrated in Fig. 2A, C and D and itemized in Table 3, there was an increase in PD-L1 mRNA expression in 75% of patients (n = 12/16). Median transcript levels increased from pretreatment median of 75/ng RNA to posttreatment median of 150/ng RNA (P = 0.04, Wilcoxon matched-pairs rank test). Changes across individuals are summarized in Table 3. Increases occurred across all histologic subtypes. Figure 2A provides representative images of a subject who experienced robust increases in PD-L1 by both mRNA (threefold change) and by histology.
Genomic and clonality assessment of TILs
The heat maps in Fig. 3 depict log2-transformed fold increases in gene signature panels using the NanoString assay and nSolver software, stratified according to PD-L1 upclassification response using the clinical SP142 assay. Therapy-associated increases in intratumoral checkpoint molecule expression and leukocyte recruitment/cytotoxic T-cell signature genes were observed in most patients (Fig. 3A–C). Similarly, increases in expression profiles of various cell markers were observed, with the most pronounced increases observed in Th1-type, CD8+, and NK cell signatures. Differential gene expression was performed on the single gene level, demonstrating statistically significant increases in a number of genes, including FOS, CXCL2, CXCR4, IL6, IL1R2, EGR1/2, and others (Fig. 3E). Changes in individual genes are also illustrated in Fig. 4.
The ImmunoSEQ platform was used to determine changes in T-cell number (based on the normalized number of TCRs) and clonality (Supplementary Figs. S4–S5). The Morisita index describes similarity of TCRs on a scale from 0 (entirely dissimilar) to 1 (identical), where a Morisita's index of 0.75 is expected from repeated biopsies of the same lesion (41). IRX-2 therapy was associated with Morisita similarity index <0.75 in 53% (8/15) of subjects, suggesting T-cell remodeling and recruitment. Significantly expanded new clones were observed in posttreatment tissue of more than 50% of specimens analyzed (Supplementary Fig. S5b and S5d). Changes in T-cell number and clonality varied across patients. In the single case of TNBC, T-cell clonality increased by 148%, with an increase in T-cell number from 792 to 13,509. Observations were concordant with quantification of CD3+ cells by the mIF method (Supplementary Fig. S5a).
Because treatment included Cy and indomethacin, we evaluated for modulation of T-regulatory and myeloid populations by both RNA expression and mIF. Following treatment there was no change in T-regulatory cell gene signature (mean +0.04 fold, standard deviation 0.09) or macrophages signature (mean +0.001-fold, standard deviation 0.05). By an alternative gene deconvolution method (Cybersort; Fig. 4), Treg levels were very low at baseline and following treatment with no apparent change. Relative macrophage M2 signatures appeared to decrease following therapy, with no apparent change in M1 signatures. By mIF, no significant changes in FOXP3+ Tregs (mean: −0.08, 95% CI: −0.40 to +0.44) or CD163+ macrophages (mean: −0.11; 95% CI, −0.42 to +0.38) were detected on a cohort level (Table 3 shows patient-level changes).
Assessment of lymph nodes
When available, paired lymph node specimens were analyzed for RNA expression and DNA T-cell repertoire. Because pretreatment lymph node biopsy is only customary for clinically positive lymph nodes, only three available pre/post-IRX-2 treatment matched lymph nodes were available for analysis. Results are illustrated in Figs. 3D and 4 and Supplementary Fig. S6. In patient 1, there was a robust increase in nodal PD-L1 expression and gene signatures, concordant with effects seen in the patient's two tumor foci. In patients 7 and 11, effects were modest in both lymph node and tumor, suggesting concordance between tumor and lymph node. Cohort-level mean gene expressions for pretreatment versus posttreatment lymph node and tumor samples are illustrated (Fig. 4A).
By T-cell repertoire analysis, we characterized clonal expansions within the lymph node, and evaluated for concordance with matched tumor (Supplementary Fig. S6). Among the 30 most expanded lymph node T-cell clones, 43% were detectable within the tumor (range, 13%–66%), many of which were concordantly expanding (Supplementary Fig. S6a).
Evaluation of peripheral immune responses
Peripheral blood changes were assessed comprehensively by flow cytometry using T-cell, B-cell, NK cell, and dendritic cell panels. Consistent with previous reports, intravenous cyclophosphamide was associated with a modest and transient depletion of CD3+, CD4+, CD8+, and Tregs (CD4+/CCR4+/CD25hi/CD127lo; refs. 13, 14), which all recovered during the course of therapy (Supplementary Fig. S1). B cells also decreased following cyclophosphamide; however, there were no apparent trends in the number of NK cell, dendritic cell, or MDSCs over time. Consistent with the locoregional nature of IRX-2 injections, there were no apparent changes in frequency and absolute counts in the peripheral blood cells following initiation of the injections. Comprehensive monitoring data are included in Supplementary Figs. S1 to S3.
Discussion
TILs reside within clinically apparent ESBCs yet are prognostic of survival especially in TNBC, highlighting the presence of tumoral and/or stromal factors that impede otherwise-protective antitumor immunity. Furthermore, many ESBCs exhibit sparse TIL infiltrates, suggesting that some tumors are less inherently immunogenic at baseline. Because cytokines serve as an endogenous mechanism of immune cell recruitment and activation (42), exogenous administration may serve to facilitate antigen presentation, effector T-cell recruitment, and activation. Here, we demonstrate the feasibility of perilymphatic cytokine administration during a preoperative therapeutic window of opportunity in ESBC. The accuracy of sentinel lymph node mapping in ESBC provides a compelling rationale for intralymphatic cytokine administration: if cytokines can be safely injected in the same anatomic site of SLN mapping, it may be possible to deliver effective doses of cytokine to the putative site of tumor-antigen presentation in ESBC. This approach could facilitate TIL recruitment and antitumor immune responses while minimizing systemic toxicity.
The study met its primary objective of demonstrating feasibility, with 100% of subjects tolerating the perilymphatic cytokine injections without any related grade 3/4 toxicities or delays in surgical management. Additionally, we provide preliminary evidence that the IRX-2 regimen is associated with increases in the clinically validated sTIL prognostic score, as well as upclassification of clinical PD-L1 status by the SP142 assay. One limitation of the clinical sTIL and SP142 scores is that they depend on semiquantiative visual assessment by pathologists and therefore may have limited resolution to detect changes (43). We used mIF with the PerkinElmer assay to further evaluate the effect of IRX-2 on TILs. By this method, we observed more robust increases in infiltration of lymphocytes, and these data were generally concordant with the H&E sTIL score and RNA cell signatures. By this method, we also observed robust increases in average PD-L1 immunofluorescence intensity and in PD-L1+ immune cell counts that were concordant with the SP142 clinical assay, providing supportive data for subsequent use of this multimodality biomarkers approach in future ESBC clinical trials.
We also observed increases in PD-L1 RNA levels and T-cell signatures by the NanoString RNA assay. Table 3 illustrates the directionality and degrees of change in the various biomarkers according to patient. Findings are consistent with known pleiotropic effects of cytokine therapy: on the one hand, the constituents of IRX-2 (including IL2, IL12, and IFNγ) are known to facilitate T-cell activation and recruitment, but on the other hand, they may upregulate tumor/immune cell PD-L1 expression via JAK-STAT signaling (44). The observed effects on PD-L1 provide a strong rationale for further study of IRX-2 in conjunction with an inhibitor of the PD-1/L1 pathway, which has the potential to enhance the overall effect on TILs and survival of cytotoxic T cells. In this exploratory analysis, we also observed upregulation of gene clusters associated with cytotoxic T cells, leukocyte recruitment, and checkpoints in the majority of patients, which were not always concordant with the above changes in sTILs and PD-L1. In the limited specimens with matched pre/post lymph node biopsy, we saw changes that were concordant with the tumor. One limitation of our DNA/RNA pipeline was that microdissection techniques were not used to specifically isolate viable tumor from surrounding biopsy tissue. This may have reduced the observed effect sizes and concordance of DNA/RNA data with the histologic findings. In future studies, laser microdissection techniques will be used to more precisely characterize changes within ICs residing in viable tumor.
In this single-arm study, we observed phamacodynamic effects similar to findings from the phase IIa HNSCC study, including increases in TILs, chemokine/recruitment signatures, and immune-checkpoint upregulation. The single-arm nature of this trial limited our ability to evaluate the individual contributory effects of IRX-2, indomethacin, and Cy (33, 45, 46). We evaluated for Cy-related T-regulatory cell depletion and indomethacin-related myeloid suppression (47). In our data set, Cy was associated with transient peripheral depletion of both T-regulatory and cytotoxic T cells in blood. The effect of IRX-2 on tumoral Tregs (by NanoString and mIF) was equivocal. However, we observed significant increases in tumoral CD8+ cells and Th1 signatures, suggesting that effector/regulatory ratio could be favorably modulated by IRX-2. By NanoString and mIF, we observed no change in macrophage markers or CD163+ cell concentrations, and no change in myeloid signatures.
Further clinical evaluation is required to test the individual contributions of IRX-2, Cy, and indomethacin, and to ascertain whether the inclusion of these agents justifies the observed increased in toxicities. The ongoing randomized INSPIRE trial in HNSCC will provide additional insight, evaluating the contribution of Cy/indomethacin with or without IRX-2. We have developed a biomarkers-driven multiarm phase II clinical trial in stage II–III TNBC to address these considerations. The trial will evaluate a backbone of standard-of-care neoadjuvant chemotherapy plus anti–PD-1 (pembrolizumab) in stage II–III TNBC, with or without IRX-2 therapy. Subsequent arms will be added to evaluate various modifications of IRX-2, for example, without cyclophosphamide or in combination with other therapies. Importantly, the trial will utilize pCR rate as a surrogate of long-term efficacy, and will also use postinduction biopsy to evaluate the effect of various pembrolizumab/IRX-2 combinations within the tumor microenvironment.
The strengths of this trial are that it demonstrates feasibility of a novel therapeutic approach of administering immunotherapy via intralymphatic injection, which could afford the opportunity to modulate immunologic priming within the tumor-draining lymph nodes and the tumor microenvironment while minimizing systemic toxicities. One limitation of this trial is that it lacked a randomized control; however, in a contemporary untreated control set, we observed relative stability of sTIL and SP142 scores, partially mitigating the concern that our observed treatment effects could be explained solely by biopsy trauma. Second, our trial was enriched for luminal-type tumors, which are thought to be less inherently immunogenic at baseline (as evidenced by their low PD-L1 expression and low sTIL scores). Finally, in the absence of clinical correlates, it is difficult to ascertain whether the observed biomarker changes such as sTIL increase or PD-L1 upclassification may translate to clinical benefit. Virtually all of these limitations have been directly addressed in the design of the follow-up phase II trial, which will provide randomized control, will enroll a more homogeneous and immunogenic tumor population, and will incorporate a clinical outcome for correlation with biomarkers changes.
Conclusion
In this study, we establish the feasibility and safety of locoregional, perilymphatic cytokine therapy in ESBC, administered during a window of opportunity preceding curative-intent therapy. The IRX-2 regimen is associated with increases in stromal TIL score, increases in PD-L1 expression (at the mRNA and protein level), and enhanced RNA signatures of lymphocyte infiltration, cytotoxic T-cell activation, and leukocyte recruitment. These preliminary data suggest that IRX-2 may be immunologically active in ESBC, supporting further study of this regimen. The current study was limited by small sample size, lack of control arm, lack of clinical outcomes, as well as a heterogeneous tumor population. A follow-up randomized phase II trial will address these limitations by evaluating IRX-2 in a more homogeneous patient population (TNBC). This study will also prospectively evaluate the pharmacodynamic effects of IRX-2 versus control, and its effect on pathologic CR rate, a surrogate of disease-free and overall survival in TNBC (48).
Data availability
All data are available from the corresponding authors upon request. Gene-expression raw data that support the findings of this study have been deposited in Gene Expression Omnibus under the accession number GSE114944. TCR sequencing data have been deposited in the ImmuneACCESS database under the accession code DOI: 10.21417/DBP2019CCR. https://clients.adaptivebiotech.com/pub/page-2019-ccr.
Disclosure of Potential Conflicts of Interest
D.S. Page is an employee/paid consultant for Brooklyn Immunotherapeutics, Merck, Bristol-Myers Squibb, Genentech, reports receiving commercial research grants from Brooklyn Immunotherapeutics, Merck, Bristol-Myers Squibb and AstraZeneca, and reports receiving speakers bureau honoraria from Genentech and Novartis. N. Moxon is an employee/paid consultant for Immunomedics, Inc. and Seattle Genetics, Inc. and reports receiving speakers bureau honoraria from Seattle Genetics, Inc. W.L. Redmond reports receiving commercial research grants from Bristol-Myers Squibb, AstraZeneca, Merck, Nektar Therapeutics, Galectin Therapeutics, Aeglea Biotherapeutics, Shimadzu, Tesaro/GSK, MiNA Therapeutics, and Veana Therapeutics, holds ownership interest (including patents) in Galectin Therapeutics, and is an advisory board member/unpaid consultant for Vesselon. M. Shah is an employee/paid consultant for Brooklyn. W.J. Urba reports receiving other remuneration from AstraZeneca. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: D.B. Page, J. Pucilowska, N. Moxon, S.L. Mellinger, M. Shah, W.J. Urba
Development of methodology: D.B. Page, J. Pucilowska, Y. Wu, Z. Sun, D. Waddell, D. Laxague, M. Shah
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.B. Page, J. Pucilowska, K.G. Sanchez, A.K. Conlin, A.K. Acheson, K.S. Perlewitz, J.H. Imatani, S. Aliabadi-Wahle, N. Moxon, S.L. Mellinger, M. Martel, Y. Wu, Z. Sun, W.L. Redmond, V.K. Conrad
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.B. Page, J. Pucilowska, K.G. Sanchez, Y. Wu, W.L. Redmond, V.K. Conrad, V. Rajamanickam, S.-C. Chang, W.J. Urba
Writing, review, and/or revision of the manuscript: D.B. Page, J. Pucilowska, K.G. Sanchez, A.K. Conlin, N. Moxon, S.L. Mellinger, Y. Wu, W.L. Redmond, D. Waddell, M. Shah, S.-C. Chang, W.J. Urba
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.B. Page, A.Y. Seino, Y. Wu, Z. Sun
Study supervision: D.B. Page, J. Pucilowska, N. Moxon, S.L. Mellinger, Y. Wu
Other (patient research advocate): D. Laxague
Acknowledgments
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.