Phase I/Ib study

This was a phase I/Ib study of sotigalimab (CD40a) in combination with nivolumab (αPD-1) and cabiralizumab (αCSF1R) in patients with advanced solid tumors whose disease progressed on αPD-(L)1 therapy. The primary objectives of the phase I dose escalation portion were to determine safety of the doublet (cabiralizumab plus sotigalimab) and triplet (doublet plus nivolumab) in patients with advanced melanoma, NSCLC or RCC as well as the RP2D. Sotigalimab was dose escalated from 0.03 to 0.1 to 0.3 mg/kg intravenously (IV) using a 3 + 3 design, in combination with fixed doses of cabiralizumab IV at 4 mg/kg plus or minus nivolumab IV at 240 mg every 2 weeks. The RP2D of the triplet was determined to be sotigalimab 0.3 mg/kg in combination with cabiralizumab 4 mg/kg and nivolumab 240 mg IV every 2 weeks. The phase I results have been published [18]. A phase Ib dose expansion portion was conducted for melanoma patients and is reported here. Patient material from phase Ib was analyzed for pharmacodynamic studies reported here.

Statistical clinical trial design for phase Ib

Eligible patients for the phase Ib dose expansion portion were treated in two disease-specific cohorts, advanced melanoma (reported here, Table 1) or NSCLC. Patients were required to have biopsy-proven disease with radiographic and/or clinical progression on αPD-(L)1 without intervening therapy. Eligibility criteria included age ≥ 18, ECOG performance status 0–1, life expectancy > 6 months, and normal organ function. Patients with melanoma were included irrespective of BRAF status. Any number of prior therapies was allowed, however patients treated with prior αCSF1R and/or CD40a therapies were excluded. A detailed list of eligibility criteria can be found in the study protocol (Related files).

Treatment and assessments for phase Ib

Prophylactic medications administered 30 min before treatment included diphenhydramine, famotidine, ibuprofen and acetaminophen. Nivolumab, cabiralizumab, and sotigalimab were administered sequentially over 30, 30, and 60 min, respectively, with 30 min breaks in between. Treatment was administered every 14 days until disease progression, intolerable toxicity, or consent withdrawal. Radiographic assessments including body CT or PET CT and MRI brain were performed at baseline, every 8 weeks for 4 months, and every 12 weeks thereafter. Treatment beyond progression was allowed if clinical benefit was derived as determined by the treating investigator.

Objectives for phase Ib

The primary objectives were to determine the ORR using RECIST v1.1 to sotigalimab in combination with cabiralizumab and nivolumab in patients with advanced melanoma and to evaluate the safety and tolerability of the regimen. Secondary objectives were to determine PFS and OS and to assess the association of selected biomarker and clinical efficacy measures using pre-treatment and on-treatment tumor biopsies. Exploratory objectives were to identify immune correlates that are associated with clinical response or resistance to the combination.

Statistical methods for phase Ib

Simon’s two-stage design was used. The null hypothesis that the true response rate to a regimen with minimal activity is 10% was tested against a one-sided alternative. In the first stage, 13 patients in each disease cohort were planned for accrual. If there were 1 or fewer responses in these 13 patients, then enrollment in that disease cohort would be stopped. Otherwise, 21 additional patients would be accrued for a total of 34 patients per disease cohort. The null hypothesis would be rejected if 6 or more responses are observed in 34 patients. This design yields a type I error rate of 0.1 and a power of 80% when the true response rate is 25%.

CyTOF

PBMCs were collected at three time points: C1D1, C1D2 and C2D1. CyTOF was performed on frozen PBMCs as described [40]. In brief, cells were barcoded with anti-CD45 antibodies conjugated to unique metal isotopes before pooling samples together to limit batch effects. Cells were stained with Cell-ID Intercalator-103Rh viability marker and a panel of 33 metal-conjugated antibodies (Supplementary Table 2). Cells were spiked with EQ 4 element beads (Fluidigm) and acquired in Helios mass cytometry system (Fluidigm). Bead-based normalization and debarcoding were performed. We processed 37 samples (3 time points from 11 patients and 2 time points from 2 patients). Left-over beads, debris, dead cells, and doublets were eliminated leaving singlet CD45 + cells [41]. Gating was done in Cytobank. The gates of various cell populations are shown in dot plots (Supplementary Fig. 4 and Supplementary Table 3) [41]. Main cell populations and their respective subpopulations were analyzed as percentage of CD45 + CD66b- cells at timepoints C1D1, C1D2 and C2D1 for treatment related changes. Ratios of different cell populations between timepoints C1D1-C2D1 and C1D1-C1D2 were analyzed for changes associated with duration of time on trial (< 200 days on trial vs > 200 days on trial) and best response (PD vs SD/PR). Ratios of mean expression of individual markers within a cell population between timepoints C1D1-C2D1 and C1D1-C1D2 were analyzed for changes associated with duration of time on trial (< 200 days on trial vs > 200 days on trial) and best response (PD vs SD/PR).

Statistical methods for correlative studies from phase Ib

Paired t-tests were used to analyze percentage of cell populations differences between different time points (C1D1, C1D2 and C2D1). Unpaired t-tests were used to analyze ratios of different cell populations or individual markers’ association with duration of time on trial and best response. A P value ≤ 0.05 was considered statistically significant.

In vivo melanoma models

2 × 106 YUMMER1.7 cells or 5 × 105 YUMM1.7 cells (kindly gifted from Dr. Marcus Bosenberg, Yale University; RRID:CVCL_A2AX and RRID:CVCL_JK16) were subcutaneously injected into the left flank of 8–9 week-old C57Bl6 male mice. YUMM1.7 is a BrafV600E /Pten−/−, Cdkn2a−/− cell line. YUMMER1.7 is a UV irradiated derivative of YUMM1.7 carrying a higher number of somatic mutations in addition to the three driver mutations: BrafV600E, Pten−/− and Cdkn2a−/− [19]. Seven days after tumor cell injection, 200 μg αPD1 (Bio X Cell, Clone RMP1-14), 200 μg αCD40 (Bio X Cell, Clone: FGK4.5/ FGK45) and 200 μg or 400 μg αCSF1R (Bristol Myers Squibb, mG1D265A) were injected IP twice a week for a total of five treatments. 200 μg of αCSF1R was the dose initially chosen (“lower dose”). We arbitrarily chose to double the initial dose to study “higher dose”. Control mice received PBS. Tumors were measured with a digital caliper and tumor volume estimated with the ellipsoid volume calculation formula. Endpoint was defined as the time until tumors reached 1000 mm3.

Cytokine/chemokine profiling for murine experiments

Whole blood was collected from control or treated mice 24 h after the first treatment in EDTA tubes. Plasma was analyzed for cytokine/chemokine expression (31-plex Mouse Cytokine / Chemokine Array, cat# MD31, Eve Technologies). Cytokine/chemokine levels (pg/ml) were analyzed in R Studio (version 3.6.2) or GraphPad Prism 9.

Immunohistochemistry of murine tumors

YUMMER1.7 tissues collected at endpoint were fixed in 10% neutral-buffered formalin overnight and embedded in paraffin by the Yale Pathology department Histology core. 5 μm FFPE tumor sections were stained with antibodies towards CD3 (cat# CP215, Biocare Medical), CD8 (clone 4SM15, eBioscience), CD163 (clone M-96, Santa Cruz), and CD68 (cat# ABIN3044428, Antibodies online). Three random intra tumoral regions were selected and positive cells were manually counted. The values in treated groups relative to controls were plotted in GraphPrism 9.

scRNA sequencing of murine tumors

Groups of three mice were treated IP on day 7 post YUMMER1.7 cell injection. On day 8, tumors were minced in RPMI with 2% FBS, incubated with 0.1 mg/ml collagenase and DNase I for 30 min at 37ºC. Cells were filtered through 70 μM filters to obtain single cell suspension, washed with RPMI + 10% FBS and pelleted by centrifugation at 1750 rpm for 5 min and resuspended in RPMI + 20% FBS. For sorting, cells were incubated for 30 min at 4ºC with fluorophore-conjugated antibodies. Samples were sorted into three populations: tumor cells (CD45- cells), myeloid cells (CD45 + CD3-, CD19-, NK1.1- cells) and CD45 + , CD3 + , CD19 + , NK1.1 + (B cells, T-cells and NK cells). Antibodies used for cell sorting were anti-CD45 (clone 30-F11, BD Biosciences), anti-CD3 (clone 17A2, BD Bioscience), anti-CD19 (clone 1D3, BD Biosciences) and anti-NK1.1(clone PK136, BD Biosciences). For live/dead staining, AmCyan Kit (Thermo Fisher Scientific) was used. Sorted cell subsets (BD FACSAriaII) were combined at a ratio of 50% tumor cells, 30% B cells, T-cells and NK cells and 20% myeloid cells. Library preparation was performed for scRNA-seq using the 3′ transcriptome kit (10 × Genomics) according to the manufacturer’s instructions. cDNA libraries were sequenced on a NovaSeq instrument (Illumina) at the Yale Center for Genome Analysis and the aligned reads were mapped to the mouse reference transcriptome (mm9). Digital count matrices were analyzed to identify cell types using R Studio (version 4.1.2) and the package Seurat v.4.0.5. To remove low quality cells following thresholds were applied: > 500 nUMI, > 250 genes, > 0.8 log10GeneperUMI and < 0.2 mitochondrial gene ratio and only genes expressed in 10 or more cells. Data was normalized and integrated using the “sctransform” method. Principal component (PC) scores from the first 40 PCs were used for clustering. For dimension reduction, Uniform Manifold Approximation and Projection (UMAP) was used with a resolution of 0.6. Distinct cell types were generated using the function “FindMarkers to determine genes that are differentially expressed between each cluster. For this, each cluster (cluster 0) was compared to all the remaining clusters (clusters 1–18). This generated a list of genes for each cluster including an average log fold change (positive value means that the gene is more highly expressed in the cluster being compared to all other clusters), pct.1 informs of percentage of cells in which the gene is detected in the first group (e.g. cluster 0), pct.2, informs of the percentage of cells where the gene is detected in the second group (e.g. clusters 1–18), and an adjusted p-value based on Bonferroni correction using all genes in the dataset is generated.

Statistical methods for murine studies

Log-rank statistics were used for survival analysis. Mann–Whitney test was used to analyze difference in expression levels of cytokine/chemokine between treatment groups. Unpaired t-test was used to analyze the difference in positive cells stained for IHC analysis in tumors. Wilcoxon Rank Sum test was used for scRNAseq data to identify distinct cell types. A P value ≤ 0.05 was considered statistically significant.