Andecaliximab/GS-5745 Alone and Combined with mFOLFOX6 in Advanced Gastric and Gastroesophageal Junction Adenocarcinoma: Results from a Phase I Study

Matrix metalloproteinases (MMP) comprise a family of at least 23 Zn²⁺-dependent proteases involved in degradation and remodeling of the extracellular matrix and basement membranes. MMPs also play a role in activation or inactivation of growth factors, cytokines, and chemokines, in normal and pathologic biological processes (1). MMP9 is an inducible MMP expressed heterogeneously by tumor epithelia as well as infiltrating macrophages, neutrophils, other inflammatory cells, fibroblastic stroma, and tumor-associated endothelial cells. MMP9 activation can release cytokines, growth factors, and bioactive protein fragments that modulate inflammation, neovascularization, and matrix remodeling (2–4).

Tumor expression of MMP9 is implicated in many protumorigenic processes. During invasion and proliferation processes, MMP9 expression has been associated with loss-of-tumor suppression activity, as well as with gain-of-oncogenic activity. In addition, MMP9 appears to contribute to temporal responses to changes in local tumor environment (3–5). MMP9 expression by infiltrating inflammatory cells in the tumor microenvironment is associated with local protumorigenic immune suppression, blunting T- and natural killer (NK)–cell responses via several potential mechanisms. MMP9 promotes local activation of TGFβ and other immune-suppressive factors, inactivation of chemokines necessary for trafficking of effector T cells, and cleavage of factors that enable tumor cell recognition by NK cells, such as major histocompatibility complex class I polypeptide-related sequence A and other NK group 2, member D ligands (6–9). Elevated MMP9 protein or RNA levels are associated with reduced overall survival in gastric and other cancers (10–12).

Early clinical experience with pan-MMP inhibitors in cancer demonstrated potential efficacy but was associated with dose-limiting musculoskeletal syndrome (13, 14). Andecaliximab is a recombinant chimeric immunoglobulin G₄ monoclonal antibody that demonstrates high affinity and selectivity for MMP9 (15, 16). The mitigation of both immunogenicity and off-target effects may overcome major shortcomings that have hindered the therapeutic success of MMP inhibition in the past. Andecaliximab was engineered without T-cell epitopes to reduce the risk of immunogenicity (17, 18). A murine monoclonal antibody targeting the same MMP9 epitope as andecaliximab significantly reduced primary tumor growth in a murine colorectal cancer xenograft mode. In addition, targeting either stromal or epithelial MMP9 in the same model reduced the incidence of metastases (15).

The negative prognostic effects and patterns of MMP9 expression and the precedents of prior pan-MMP inhibitors provided a rationale for evaluating andecaliximab + chemotherapy in advanced solid tumors, including gastric and gastroesophageal junction (GEJ) adenocarcinoma. We initiated a phase I study (ClinicalTrials.gov Identifier: NCT01803282) in two parts: a monotherapy dose-finding stage (part A), and a combination treatment stage (part B) in patients with selected tumor types. Here, we present data from part A and the gastric/GEJ adenocarcinoma cohort of part B.

This phase I study was divided into two parts: part A, dose-finding monotherapy phase enrolling patients with advanced solid tumors, and part B, andecaliximab + chemotherapy in specific patient expansion cohorts, including patients with advanced pancreatic adenocarcinoma, non–small cell lung cancer, gastric/GEJ adenocarcinoma, colorectal adenocarcinoma, or HER2-negative breast cancer (see Supplementary Fig. S1). Data from cohorts other than gastric/GEJ will be published separately. Planned enrollment included 12 to 48 patients in part A and 15 to 40 patients per cohort in part B. The determination of the final sample size followed the principles of a two-stage design, whereas the decision to enroll 40 patients was contingent on the safety and efficacy signals from the initial 15 patients. Enrollment of 40 patients would be sufficient to assess pharmacokinetics (PKs) and safety, and to give a reasonably precise estimate of efficacy, that is, tumor response. For an estimated overall response rate of 50%, a sample size of 40 patients would allow the 95% confidence interval (CI) to be ±15%. The study was conducted in accordance with the Declaration of Helsinki; local ethics committees/institutional review boards at all participating centers approved the study. All patients provided written informed consent before entering the study.

Key inclusion criteria for all patients included age ≥18 years; Eastern Cooperative Oncology Group Performance Status of ≤1; life expectancy of >3 months; adequate hematologic, hepatic, and coagulation function; serum creatinine ≤1.5 × upper limit of normal; and willingness to follow adequate precautions to prevent pregnancy. For part A, patients were required to have histologically or cytologically confirmed advanced malignant solid tumors refractory to or intolerant of standard therapy, or for which no standard therapy was available. For the gastric/GEJ adenocarcinoma cohort of part B, patients were required to have histologically confirmed, HER2-negative, inoperable advanced gastric/GEJ adenocarcinoma treatment-naïve in the metastatic setting (prior systemic therapy in the nonmetastatic setting was allowed). Key exclusion criteria for all patients included significant comorbid medical conditions that posed a risk to patient safety or limited study participation; pregnancy or lactation in women; untreated central nervous system metastases; and known HIV, HBV, or HCV infection.

Andecaliximab was administered at 200 (cohort 1), 600 (cohort 2), and 1,800 mg (cohort 3) via i.v. infusion over 30 minutes every 2 weeks (q2w) following a 3 + 3 design. If dose-limiting toxicities (DLTs) were experienced by ≥2 patients in a cohort, dose de-escalation would occur to an intermediate dose. The MTD was defined as the highest andecaliximab dose level at which ≤33% of patients experienced a DLT during days 1 to 28 (cycle 1). The following adverse events (AEs) occurring during cycle 1 (possibly related to andecaliximab) were considered DLTs: grade 4 neutropenia [absolute neutrophil count (ANC) < 500/μL] for >7 days; febrile neutropenia (ANC < 1,000/μL with fever >101°F (38.5°C)]; grade 4 thrombocytopenia; grade 3 thrombocytopenia with bleeding; grade 3 or 4 nonhematologic toxicity (excluding rash, nausea, diarrhea, and vomiting if controlled with standard supportive care); treatment delay of ≥14 days due to unresolved toxicity and nonhematologic toxicity of ≥grade 2 that was dose-limiting. Andecaliximab was permanently discontinued for recurrent toxicities meeting the definition of a DLT or for toxicities that did not resolve within 28 days.

The combination with 5-fluorouracil + leucovorin + oxaliplatin (mFOLFOX6) was selected based on U.S. treatment guidelines for patients with advanced HER2-negative gastric/GEJ adenocarcinoma (19). Andecaliximab (800 mg) was administered on days 1 and 15 of each 28-day cycle. mFOLFOX6 was administered immediately following andecaliximab on days 1 and 15 of each 28-day cycle as follows: l-leucovorin (200 mg/m²) or dl-leucovorin (400 mg/m²), oxaliplatin (85 mg/m²), 5-fluorouracil bolus (400 mg/m²), followed by a 46-hour infusion of 5-fluorouracil (2,400 mg/m²).

Patients were treated until unacceptable toxicity, withdrawal of consent, disease progression, or death.

Blood was collected for evaluation of plasma concentrations of andecaliximab at scheduled times following infusion. Plasma drug levels were determined with a validated proprietary electrochemiluminescence immunoassay method. PK parameters were calculated using standard noncompartmental methods with Phoenix WinNonLin software. Data were listed and summarized using descriptive statistics by cohort and treatment.

Total (bound and free) MMP9 and free MMP9 were measured as described in Supplementary Material. Maximal andecaliximab binding to MMP9 was achieved when levels of andecaliximab-free MMP9 were below the limit of detection.

The collagen I cleavage fragment C1M was measured in serial serum samples by immunoassay (Nordic Bioscience A/S, Herlev, Denmark). Interpatient and intrapatient variabilities (measured at two time points prior to drug treatment) were 101% and 39%, respectively, in the gastric cohort.

Efficacy data are available from part B only. Disease burden was evaluated at screening by physical examination and radiographic assessment (contrast-enhanced CT or MRI) and then every 8 weeks. Responses were assessed by investigators per the RECIST version 1.1 criteria (20). Objective response rate (ORR) was defined as the proportion of patients with complete response (CR) or partial response (PR) as best overall response during combination therapy. The Clopper–Pearson method was used to calculate the exact CIs of ORR. Progression-free survival (PFS) was defined as the time interval from the first dose of andecaliximab to the earlier of the first documentation of definitive disease progression or death from any cause, analyzed using Kaplan–Meier methods. A post hoc analysis of efficacy was performed in first-line patients enrolled in the combination cohort, defined as patients who were treatment-naïve or >190 days from prior treatment for localized gastric/GEJ adenocarcinoma. Patients were not followed for overall survival.

Safety assessments were performed prior to each andecaliximab infusion, following the CTCAE version 4.03 criteria (21). Paresthesias/dysesthesias were graded according to the following scale: grade 1, of short duration that resolved and did not interfere with function; grade 2, interfered with function, but not with activities of daily living (ADL); grade 3, presented with pain or with functional impairment that also interfered with ADL; grade 4, persistent and disabling or life-threatening.

Between April and August 2013, 13 patients with advanced solid tumors were enrolled into part A of the study from three sites in the Unites States. All 13 patients were treated and evaluable for the safety analysis (Table 1).

Across the three monotherapy cohorts in part A, the median duration of exposure to andecaliximab was 6.1 (range, 0.1–31.7) weeks with a median of 4 (range, 1–16) doses received. No dose reductions of andecaliximab were required. No DLTs were reported. No patients discontinued monotherapy because of AEs. AEs attributed to andecaliximab (all grades 1–2) were observed in 7 of 13 (53.8%) patients; and the most common AEs were nausea (31%), vomiting, and fatigue (23% each), and diarrhea, asthenia, arthralgia, joint stiffness, and dyspnea (8% each). Grade 3–4 AEs were observed in four patients: aspartate transaminase increase [cohort 1, one patient (8%)]; and acute kidney injury, duodenal obstruction, and pleural effusion [cohort 3, one patient (8%) each]. No grade 3–4 AEs were attributed to andecaliximab (Table 2). No grade 5 AEs were reported. All patients discontinued study treatment due to progression.

The mean plasma concentration versus time profile and the calculated PK parameters of andecaliximab following doses of 200, 600, and 1,800 mg are shown in Fig. 1. At each dose level, maximum plasma concentrations were observed at the end of infusion and increased approximately dose-proportionally, whereas the area under the curve increased more than dose-proportionally. The terminal phase of elimination was faster at the lowest dose and approached parallel elimination at the higher doses. The clearance decreased from 1,172 mL/day in the 200 mg cohort to 377 mL/day in the 1,800 mg cohort. The terminal half-life increased with dose from approximately 2.4 days to 8.3 days. These data are consistent with target-mediated drug disposition having a significant effect on the clearance of andecaliximab at the lower doses.

Between November 2013 and July 2015, 41 patients with gastric/GEJ adenocarcinoma were enrolled into part B of the study from 10 sites in the United States; one patient did not receive study treatment, so 40 patients were treated and followed for safety and efficacy (Table 1).

In the gastric/GEJ adenocarcinoma cohort, the median duration of exposure to andecaliximab was 21.0 (range, 0.1–121) weeks with a median of 10.5 (range, 1–57) doses received. The median duration of exposure and median number of doses of mFOLFOX6 were as follows: 5-fluorouracil 21.3 (range, 0.4–121.4) weeks, 10.5 (range, 1–57) doses; leucovorin 19.4 (range, 0.4–121.4) weeks, 10 (range, 1–57) doses; and oxaliplatin 16.4 (range, 0.4–121.4) weeks, 8 (range, 1–20) doses.

AEs of any grade related to andecaliximab observed in ≥10% of all patients included fatigue (35%), nausea (28%), neutropenia (18%), dysgeusia (15%), decreased platelet count (15%), decreased appetite (13%), diarrhea (13%), thrombocytopenia (10%), and decreased neutrophil count (10%; Table 2). Four grade 5 AEs were observed (all considered unrelated to andecaliximab or mFOLFOX6 by the investigator): abdominal ascites (in the setting of disease progression in the peritoneum), gastrointestinal hemorrhage (classified as disease-related in a patient with stage IV disease at diagnosis who had not undergone resection), septic shock, and pneumothorax (classified as intercurrent illnesses in the setting of pneumonia and bronchopleural fistula following a thoracentesis) in 1 patient each (3%). As of August 31, 2016, 6 patients with gastric/GEJ adenocarcinoma continued to receive study treatment. As of September 22, 2017, one patient remains on study treatment.

Exposure and response data for the gastric/GEJ adenocarcinoma cohort are shown in Supplementary Fig. S2. The ORR for all patients was 47.5% (90% CI, 33.8–61.5) with 3 (7.5%) CRs and 16 (40%) PRs observed, and 50% for first-line patients (Table 3). Although the ORR was approximately 50%, the percent change in tumor size for all patients (Fig. 2A, waterfall plot) demonstrates that all but five patients with measurable disease had a reduction in tumor size. The duration of response was 8.4 months for all patients and 9.3 months for first-line patients. PFS was 7.8 (90% CI, 5.5–13.9) months in all patients and 9.9 (90% CI, 5–13.9) months in first-line patients (Fig. 2B).

Total MMP9 (andecaliximab-bound plus -free) and andecaliximab-free MMP9 in circulation were measured in plasma. The total and free MMP9 concentrations at baseline (prior to andecaliximab infusion) were within 5% (Supplementary Material). In the monotherapy cohorts, free MMP9 was undetectable in all treated subjects except for two subjects in the 200-mg cohort (1 at days 15 and 43 and 1 at day 43), indicating that all circulating MMP9 was bound to andecaliximab at the 600- and 1,800-mg dose levels (Fig. 3A; Supplementary Fig. S3). In the gastric/GEJ adenocarcinoma cohort, free MMP9 was detectable at baseline in all patients, and undetectable in 91% to 100% of patients at postbaseline time points, with little change in total MMP9 (Fig. 3B).

MMP9 cleaves extracellular matrix proteins including collagens; therefore, fragments of collagen, such as C1M, detectable in blood may serve as a marker of MMP activity. Serum C1M was not dose-dependently modulated at day 15 or 43 in the monotherapy cohorts of mixed tumor types (Fig. 3C). However, a significant decrease in C1M was observed at week 4 [median % baseline, 79.0 (Q1 46.5, Q3 112.1), P = 0.001] and maintained over time for the gastric/GEJ adenocarcinoma cohort (Fig. 3D). Patients with the highest concentration of circulating C1M at baseline had the largest decrease in C1M upon treatment (Fig. 3E).

Andecaliximab is a novel, highly selective antibody inhibitor of MMP9. The purpose of this study was to determine the MTD of andecaliximab monotherapy and to assess its efficacy and safety in combination with chemotherapy in patients with advanced solid tumors. In the monotherapy cohort, the plasma concentration versus time profile of andecaliximab displayed typical nonlinear PK at the low dose, 200 mg, and was linear in the higher-dose cohorts, 600 and 1,800 mg. The clearance of andecaliximab decreased as the dose increased. Combined, these data suggest a contribution of target-mediated drug disposition to andecaliximab plasma elimination. As is typical for monoclonal antibodies, the steady-state volume of distribution was low, suggesting limited extravascular distribution. Based on known clearance mechanisms, no drug–drug interaction between andecaliximab and mFOLFOX is expected when andecaliximab is combined with mFOLFOX in the gastric/GEJ adenocarcinoma cohort.

The MTD of andecaliximab was not determined in this study. Andecaliximab monotherapy appeared well tolerated when administered at 200, 600, and 1,800 mg i.v. q2w, and no DLTs were observed at any dose. A dose of 800 mg q2w (or 1,200 mg q3w) was selected for further development as it was expected to achieve plasma concentrations that are in the linear range of the PK profile and to achieve adequate steady trough concentrations (i.e., saturate target-mediated drug disposition). A population PK model for andecaliximab, incorporating nonlinear elimination processes and relevant covariates, was developed and supports evaluation of the 800 mg q2w i.v. regimen (22).

Reduction in the level of circulating andecaliximab-free MMP9 after treatment demonstrated on-target binding of andecaliximab. In the monotherapy cohorts, all circulating MMP9 was bound to andecaliximab after single doses of 600 or 1,800 mg. Maximal binding of MMP9 by andecaliximab was achieved in all patients in the gastric/GEJ adenocarcinoma cohort, supporting further development of the 800 mg q2w dose.

Treating patients who have advanced gastric/GEJ adenocarcinoma with andecaliximab 800 mg + mFOLFOX6 q2w resulted in a median PFS of 7.8 (90% CI, 5.5–13.9) months with ORR of 48%. Among first-line patients (defined as patients who were treatment-naïve or >190 days since previous therapy for localized gastric/GEJ adenocarcinoma), PFS was 9.9 (90% CI, 5–13.9) months, suggesting a potential improvement over conventional chemotherapy in first-line metastatic or unresectable advanced gastric/GEJ adenocarcinoma (23–25).

The most frequently reported toxicities in the monotherapy cohorts were nausea, fatigue, and dyspnea, whereas the most common toxicities of andecaliximab + mFOLFOX6 were nausea, fatigue, diarrhea, and peripheral neuropathy. In contrast to the pan-MMP inhibitor marimastat (14), andecaliximab was not associated with treatment-emergent musculoskeletal syndrome. Andecaliximab + mFOLFOX6 appears well tolerated without new or unexpected safety signals. The safety profile of andecaliximab + mFOLFOX6 appears similar to the previously characterized toxicity profile of mFOLFOX6 in advanced gastric adenocarcinoma, with nausea, vomiting, diarrhea, neuropathy, and neutropenia being commonly observed in previous reports (23–25). Of note, the four on-study deaths were not attributed to study treatment, but to either disease progression or intercurrent illness.

Generation of the collagen cleavage fragment C1M was investigated as a marker of MMP9 activity. C1M did not decrease in the monotherapy cohorts; however, a significant decrease in C1M was sustained from week 4 onward in the gastric/GEJ adenocarcinoma cohort treated with andecaliximab + mFOLFOX6. Patients with the highest baseline level of C1M had the greatest decrease. The discrepancy in C1M reduction between the monotherapy and gastric/GEJ adenocarcinoma cohorts could be due to the impact of mFOLFOX6 on C1M. In addition, because other MMPs cleave collagens (26), the difference could be related to the distribution of MMPs in particular diseases as patients with eight different solid tumor diseases were treated in the monotherapy cohorts.

Based on the safety and efficacy profile in the gastric/GEJ adenocarcinoma combination cohort, a randomized phase III study (27) comparing mFOLFOX6 with and without andecaliximab 800 mg q2w as first-line treatment for HER2-negative advanced gastric/GEJ adenocarcinoma has been initiated (28).

A. Starodub reports receiving speakers bureau honoraria from Bristol-Myers Squibb and Lexicon, and is a consultant/advisory board member for AstraZeneca. Z.A. Wainberg is a consultant/advisory board member for Array, Gilead, and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.A. Shah, M. Gordon, J.D. Maltzman, V. Smith, J.A. Silverman

Development of methodology: M.A. Shah, G. Vosganian, J.A. Silverman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.A. Shah, A. Starodub, S. Sharma, J. Berlin, M. Patel, Z.A. Wainberg, J. Chaves, M. Gordon, K. Windsor, C.B. Brachmann, G. Vosganian, H.-J. Lenz, J.C. Bendell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.A. Shah, S. Sharma, J. Berlin, M. Patel, Z.A. Wainberg, M. Gordon, C.B. Brachmann, X. Huang, G. Vosganian, J.A. Silverman, H.-J. Lenz

Writing, review, and/or revision of the manuscript: M.A. Shah, A. Starodub, S. Sharma, J. Berlin, M. Patel, Z.A. Wainberg, J. Chaves, M. Gordon, C.B. Brachmann, X. Huang, G. Vosganian, J.D. Maltzman, V. Smith, J.A. Silverman, H.-J. Lenz, J.C. Bendell

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.A. Shah, J. Berlin, J. Chaves, X. Huang, G. Vosganian, H.-J. Lenz

Study supervision: M.A. Shah, M. Patel, Z.A. Wainberg, G. Vosganian, J.D. Maltzman

The authors thank the patients for their participation in this study. Biomarker data analysis was provided by Yafeng Zhang, PhD, Gilead Sciences, Inc. Professional medical writing assistance was provided by Beth Sesler, PhD, CMPP, at Impact Communications, New York, NY, and was supported by Gilead Sciences, Inc., Foster City, CA.

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.