Abstract
Immune checkpoint blockade (ICB) has become the standard of care for several solid tumors. Multiple combinatorial approaches have been studied to improve therapeutic efficacy. The combination of antiangiogenic agents and ICB has demonstrated efficacy in several cancers. To improve the mechanistic understanding of synergies with these treatment modalities, we performed screens of sera from long-term responding patients treated with ipilimumab and bevacizumab. We discovered a high-titer antibody response against EGF-like repeats and discoidin I–like domains protein 3 (EDIL3) that correlated with favorable clinical outcomes. EDIL3 is an extracellular protein, previously identified as a marker of poor prognosis in various malignancies. Our Tumor Immune Dysfunction and Exclusion analysis predicted that EDIL3 was associated with immune exclusion signatures for cytotoxic immune cell infiltration and nonresponse to ICB. Cancer-associated fibroblasts (CAF) were predicted as the source of EDIL3 in immune exclusion–related cells. Furthermore, The Cancer Genome Atlas Skin Cutaneous Melanoma (TCGA-SKCM) and CheckMate 064 data analyses correlated high levels of EDIL3 with increased pan-fibroblast TGFβ response, enrichment of angiogenic signatures, and induction of epithelial-to-mesenchymal transition. Our in vitro studies validated EDIL3 overexpression and TGFβ regulation in patient-derived CAFs. In pretreatment serum samples from patients, circulating levels of EDIL3 were associated with circulating levels of VEGF, and like VEGF, EDIL3 increased the angiogenic abilities of patient-derived tumor endothelial cells (TEC). Mechanistically, three-dimensional microfluidic cultures and two-dimensional transmigration assays with TEC endorsed EDIL3-mediated disruption of the lymphocyte function-associated antigen-1 (LFA-1)–ICAM-1 interaction as a possible means of T-cell exclusion. We propose EDIL3 as a potential target for improving the transendothelial migration of immune cells and efficacy of ICB therapy.
Introduction
Immune checkpoint blockade (ICB) has become the standard of care for numerous malignancies, including melanoma, for which approximately 50% of patients with metastatic disease receive benefits (1). Combinatorial approaches to improve outcomes and overcome treatment resistance are sought for a sizeable patient population with unmet needs. The immune system is known to have an active interplay with the vascular system, as it is the gateway to the tumor microenvironment (TME). Angiogenic factors have demonstrated immune suppression capabilities, and tumor blood vessels may limit the access of immune effector cells to the TME. Innovative strategies, including the combination of ICB with antiangiogenic treatment, have shown efficacy in a number of tumor types, including renal cell carcinoma, hepatocellular carcinoma, endometrial cancer, non–small cell lung cancer, and melanoma (2). Correlation analyses revealed that VEGF blockade is associated with improved lymphocyte trafficking across the endothelium, as well as improved outcomes when tumor assessment has a myeloid expression signature (3).
To understand the immunologic role of combination therapy and to identify targets associated with favorable outcomes, as well as potential functional relevance, sera from patients who received long-term benefits from therapy with the ICB agent ipilimumab plus the antiangiogenic agent bevacizumab (Ipi-Bev) were screened using the human protein array. Such serologic screens have previously identified targets with functional activity (4–6).
Materials and Methods
Patients
Patients with metastatic melanoma treated in the phase I Ipi-Bev clinical trial (NCT00790010) have been reported previously (4). The collection and use of samples from these patients were in compliance with the Declaration of Helsinki and approved by the institutional review board (IRB) of the Dana-Farber/Harvard Cancer Center, Cancer Therapy Evaluation Program, and Eastern Cooperative Oncology Group-American College of Radiology Imaging Network. Informed written consent was obtained from all patients.
Collection of patients’ plasma and detection of antibodies specific for EGF-like repeats and discoidin I–like domains protein 3
Briefly, heparinized peripheral blood samples were obtained before and after Ipi-Bev treatment from patients with melanoma participating in IRB-approved protocols at the Dana-Farber/Harvard Cancer Center. Blood samples were processed on the same day by standard gradient centrifugation using Ficoll-Paque Plus (Sigma-Aldrich; #GE17-1440-02). The upper phase was transferred into 1.5 mL tubes and centrifuged at 12,000 × g for 10 minutes at 4°C. The resulting plasma was aliquoted and stored at −80°C until use. Studies of humoral responses were primarily based on four cohorts of patients with advanced melanoma who were treated with (i) ipilimumab alone (n = 34), (ii) Ipi-Bev (n = 42), (iii) anti-PD-1 alone (n = 21), or (iv) ipilimumab and nivolumab (Ipi-Nivo; n = 41) as the standard of care or through participation in the Ipi-Bev clinical trial. As per the manufacturer's instructions, ProtoArray Human Protein Microarray v5 (Invitrogen; #25-1011) was used to screen posttreatment plasma samples for antibodies from long-term responders to Ipi-Bev combination therapy. A Z-factor cutoff of 0.4, recommended by the manufacturer, was used to identify potential antibody targets.
ELISA
For the detection of antibodies in plasma, ELISAs were performed as described previously (5). Briefly, 96-well high-binding microtiter plates (Corning; #9018) were coated overnight with 100 ng of purified human recombinant proteins rEDIL3 (R&D; #6046-ED-050) and rMFGE8 (Milk fat globule-EGF factor 8; R&D; #2767-MF-050) in 50 μL of TBS using His-tag and no plasma as negative controls. After washing three times with TBS containing 0.05% Tween 20 (TBST), the plates were blocked with a blocking buffer (Pierce Protein-Free blocking buffer, Thermo Fisher Scientific; #37572) for 2 hours at room temperature. Plasma samples (dilution, 1:500–1:2,000 in blocking buffer) were incubated for 1.5 hours at 4°C. After washing, horseradish peroxidase (HRP)–labeled anti-human IgG (Jackson laboratories; #109-035-088; dilution, 1:4,000) in 1% BSA-TBST was added for 1 hour at room temperature. An ELAST amplification system was used to enhance the sensitivity of the immunoassay (PerkinElmer; #NEP116001EA). The signal was developed with 3,3′,5,5′-tetramethylbenzidine (TMB; Millipore Sigma; #T0440) substrate for 5 minutes. The reaction was stopped with 1N HCl, and the absorption was measured at 450 and 570 nm.
Measurement of secreted EGF-like repeats and discoidin I–like domains protein 3
Secreted EGF-like repeats and discoidin I–like domains protein 3 (EDIL3) levels in patient plasma as well as in conditioned medium from cell culture were determined in duplicate using a human EDIL3 ELISA kit (R&D Systems; #DY6046-05), according to the manufacturer's instructions. A standard curve was constructed for each assay.
Bioinformatics data analysis
The Cancer Genome Atlas (TCGA) datasets available from the GDC portal were used to examine the correlation (Kendall rank correlation) between the gene expression of immunologically important glycoproteins EDIL3 and MFGE8 and Tumor Immune Dysfunction and Exclusion (TIDE) signatures (7). TIDE generated signatures for T-cell dysfunction, T-cell exclusion, and the predicted response to ICB. CTNNB1 (β-catenin) was used as a reference marker for immune exclusion (8). TCGA Skin Cutaneous Melanoma (SKCM) and BMS CheckMate 064 datasets were analyzed for gene expression grouped by TGFβ signaling, pan-fibroblast TGFβ response signature, epithelial-to-mesenchymal transition (EMT), and angiogenesis signatures. Melanoma cases from TCGA-SKCM (n = 469; ref. 9) and BMS CheckMate 064 pretreatment patients (n = 90; ref. 10) with RNA sequencing (RNA-seq) data for EDIL3 expression were analyzed. Gene expression data were log2(TPM+1) transformed. Pathway expression was calculated using the GSVA tool (11) for TGFβ, pan-fibroblast TGFβ response signature (F-TBRS), EMT, and angiogenesis (Angio) gene sets (12). A High-Low designation was assigned for angiogenesis expression using agglomerative clustering by splitting the expression values into two clusters, and the two-sided Mann–Whitney U test was applied for hypothesis testing.
Cell lines
Tumor endothelial cells (TEC) were isolated from a patient with melanoma enrolled in the Ipi-Bev trial posttreatment, following IRB-approved protocols (5). Human umbilical vascular endothelial cells (HUVEC) and Human Dermal Microvascular Endothelial Cells (HDMEC) were purchased from ATCC in 2021. Endothelial cells were maintained in endothelial basal medium (EBM-2; PromoCell; #C-22111) and used within 3–4 passages. Cancer-associated fibroblasts (CAF) were isolated from melanoma tumor samples from 2 patients, as described previously (13). P4-CAF were isolated posttreatment. Normal fibroblasts (NF) were provided as a gift from Kenneth C. Anderson's laboratory at Dana-Farber Cancer Institute (DFCI). The CAFs and NFs were maintained in DMEM (Thermo Fisher Scientific; #10313039) supplemented with 10% FBS (Thermo Fisher Scientific; #10437028) and 1% Pen Strep (Gibco; #15140-122). Jurkat (human T lymphocytes) and THP-1 (human monocyte cell line) cells were obtained from ATCC and maintained in RPMI1640 (Thermo Fisher Scientific; #11875093) medium containing 10% FBS and 1% Pen Strep. Human melanoma cell lines established from patients as described previously (14) and A375 purchased from ATCC in 2022 were maintained in RPMI1640 media supplemented with 10% FBS and 1% Pen Strep. Regulatory T cells from two donors were a gift from Rizwan Romee's laboratory at DFCI and immediately used in assays. H226 were purchased from ATCC and cultured in RPMI1640 medium supplemented with 10% FBS and 1% Pen Strep.
Peripheral blood mononuclear cells (PBMC) were isolated from normal donor blood samples using Ficoll density gradient separation per the IRB-approved protocol. T cells were isolated from PBMCs using a Pan T Cell Isolation Kit (Miltenyi Biotec; #130-096-535). T-cell activation was achieved by adding human monoclonal anti-CD3 (BioLegend; #317326) and anti-CD28 (BioLegend; #302934) at 1 μg/mL of 1:1 ratio in RPMI media supplemented with 10% FBS, 50 μg/mL penicillin, 100 μg/mL streptomycin, and 50–100 IU/mL IL2 (Miltenyi Biotec; #130-097-742). The cell lines were not reauthenticated and cultures were routinely tested for Mycoplasma using the Universal Mycoplasma Detection Kit (ATCC; #30-1012K). Experiments on cell lines were performed within no more than 10 passages.
Immunoblotting
Immunoblot analysis was performed as reported previously (15). Briefly, whole-cell lysates of NF, CAF, or TEC were lysed in 1× lysis buffer (Cell Signaling Technology; #9803) supplemented with protease inhibitor (TargetMol; #C0001) and phosphatase inhibitor (TargetMol; #C0002), and centrifuged at 14,000 rpm for 20 minutes at 4°C. Supernatants were collected and resolved on 4%–12% SDS-PAGE gels (Thermo Fisher Scientific; #NP0321BOX) and transferred onto activated polyvinylidene difluoride (PVDF) membranes. The membranes were blocked in 5% skim milk in PBS and probed with antibodies specific for EDIL3 (Abcam; #190692), pSMAD2 (Cell Signaling Technology; #3108T), SMAD2/3 (Cell Signaling Technology; #8685T), intercellular adhesion molecule 1 (ICAM-1; Abcam; #ab10936) or β-actin (Sigma; #A1978) overnight at 4°C. The following day, the membranes were washed and incubated with the corresponding HRP-conjugated secondary antibodies for an hour at room temperature. Proteins were visualized using an enhanced chemiluminescence kit (Perkin Elmer; #NEL103001EA) and digitally processed using ImageQuant LAS-4000. Bevacizumab was purchased from Selleckchem (#A2006). NF were pretreated with or without LY2109761 (LY, Cayman Chemicals; #15409) followed by TGFβ1 (BioLegend; #781802) treatment for 24 hours. EDIL3 expression in the conditioned medium was examined by human EDIL3 ELISA kit and in the whole-cell lysates by immunoblot analysis.
For the detection of EDIL3-specific antibody in plasma, equal amounts of rEDIL3 protein denatured in 2× SDS-containing Laemmli buffer were resolved by SDS-PAGE and transferred to a PVDF membrane. Each lane of the membrane was cut and immunoblotted with pretreatment and posttreatment plasma samples in a separate container. Plasma samples were diluted in 1% BSA in TBS with TBST at 1:500 dilutions. The membranes were stripped and reprobed with a commercial EDIL3-specific antibody as a loading control.
RNA extraction and qPCR
RNA was extracted from P4-CAF or TEC using an RNeasy Mini Kit (Qiagen; #74104). Reverse transcription for cDNA synthesis was performed from 2 μg total RNA using the SuperScript VILO cDNA Synthesis Kit, according to the manufacturer's instructions (Invitrogen; #11754050). Gene expression levels were analyzed by qPCR using primers EDIL3 (h)-PR (Santa Cruz Biotechnology; #sc-91971-PR). A standard program at ABI 7500: 95°C for 10 minutes, 40 cycles at 95°C for 15 seconds, and 60°C for 1 minute was used. Following Normfinder analyses for the stable reference gene, β-actin (Forward primer: TCCACCTTCCAGCAGATGTG; Reverse primer: GCATTTGCGGTGGACGAT) was chosen. Three technical replicates for each sample were run and analyzed in one qPCR reaction. The relative abundance of mRNAs was calculated with the comparative ΔΔCt method.
For bulk RNA sequencing (RNA-seq), total RNA was isolated as mentioned above from biological duplicates of P4-CAF and NF untreated or treated with TGFβ1 (5 ng/mL; Peprotech; #100-21) for 24 hours. Bulk RNA-seq libraries were generated and sequenced by Novogene using the NovaSeq 6000 platform at PE150 targeting an average of 20M reads/sample. Transcriptome analysis for the RNA-seq was based on the GENCODE v26 (Human genome version GRCh38.p10), and transcripts per million (TPM) data extraction was completed using Salmon v0.8.1 software (16). The gene expression data (TPM) are provided in Supplementary Table S1.
EDIL3 silencing
P4-CAFs or TECs were seeded in 6-well plates at 80% confluency for one day before transfection. The cells were transfected with either control siRNA-A (Santa Cruz Biotechnology; #sc-37007) or siEDIL3 (Santa Cruz Biotechnology; #sc-91971) using Lipofectamine RNAiMAX Transfection Reagent (Invitrogen; #13778030) in Opti-MEM I reduced serum medium (Corning; #31985062), according to the manufacturer's instructions. After 6 hours, the media was changed to the default growth medium, and the cells were cultured overnight. Silencing of EDIL3 was confirmed by qRT-PCR or immunoblot analysis.
Wound healing assay
Wound healing assays were performed as described previously (17). Briefly, TECs were seeded in fibronectin-coated 12-well plates and allowed to grow until they reached confluence. At 90% confluency, cells were starved in 0.4% FBS containing EBM-2 medium for 24 hours followed by wounding across the center of the well with a sterile 200 μL pipette tip; followed by incubation in control, VEGF-A (R&D; # BT-VEGF-050; 20 ng/mL), or EDIL3 (50 ng/mL) supplemented low serum medium. Using a phase-contrast microscope, the leading wound edges were photographed at 0, 24, and 48 hours at the same reference areas for each well. Three independent experiments were conducted, with each treatment in triplicate, with three areas of reference per replicate. The percent wound area was calculated using ImageJ software.
Tube formation assay
Endothelial tube formation assays were performed using the Cultrex In Vitro Angiogenesis Assay (R&D Systems; #3470-096-K) according to the manufacturer's protocol. Briefly, the Cultrex RGF BME vial was thawed on ice overnight before seeding. In a chilled 96-well plate, Cultrex RGF BME was added at 50 μL per well using a chilled 200 μL pipette tip. The plate was centrifuged at 250 × g for 5 minutes at 4°C to allow the basement membrane extract (BME) to spread evenly. To solidify BME, the coated plates were incubated for 30 minutes at 37°C. Meanwhile, TECs starved of serum overnight were collected and diluted in conditioned medium (EBM-2 without growth factors and with 0.4% FBS) as control or supplemented with either VEGF-A (20 ng/mL) or EDIL3 (50 ng/mL). The cells were seeded on BME at a density 30 × 103 per well in triplicate and incubated at 37°C for 12 hours followed by imaging using a Nikon Eclipse TE2000-S microscope at 40× magnification. Quantitative image analyses were performed by using ImageJ version 1.53k with “Angiogenesis Analyzer” plugin.
Surface flow cytometry analysis for lymphocyte function-associated antigen-1 (CD11a)
THP-1 cells grown in suspension in RPMI1640 were preactivated with phorbol 12-myristate 13-acetate (PMA; Millipore Sigma: # P1585) 50 ng/mL for 24 hours and rested overnight. Human Jurkat T cells were grown in log phase in RPMI1640 medium. The cells were counted and maintained at 1–2 × 106 cells/mL by adding fresh RPMI1640 medium containing 50 to 100 IU/mL IL2 every 2 to 3 days. On day 9, the cells were restimulated with the human T-activator CD3/CD28 (Gibco; #11131D). The expanded T cells were used for flow cytometry analysis to verify lymphocyte function-associated antigen-1 (LFA-1) expression. Briefly, PMA-activated THP-1, Jurkat T cells and expanded human T cells were stained with FITC-conjugated CD11a (BioLegend; #350602) or isotype control (BioLegend; #400123). The cells were washed twice and resuspended in the FACS buffer. The cells were acquired on a BD LSRII flow cytometer using the FACSDiva software (BD Biosciences). The data were analyzed using FlowJo_v10.7.1 software (Tree Star, Inc).
Adhesion assay
Adhesion assays were performed as described previously, with some modifications (18, 19). Briefly, TECs were grown on fibronectin-coated 96- or 24-well black-walled plates. After 24 hours of seeding, the cells were treated with BSA or stimulated with TNFα (BioLegend; # 570102; 5 ng/mL) for another 24 hours. Activated T cells were labeled with BCECF (1 μmol/L, Life Technologies; #B1170) for 30 minutes, followed by incubation with the indicated concentrations of rEDIL3 for 1 hour at 37°C in dark or low light conditions. Endothelial monolayers were washed three times with PBS at the end of the treatment. Pretreated BCECF-labeled activated T cells were added to all the wells and allowed to adhere for 45 minutes at 37°C. Fluorescence intensity was measured at 485 nm (excitation) and 530 nm (emission), before (input), and after four washes with warm RPMI1640 medium to remove nonadherent cells using a SpectraMax M3 Multi-Mode Microplate Reader. The adhesion data were plotted as a percentage of the untreated control (100%). Representative images were acquired for each treatment group.
Endothelial cell transmigration assay
For EDIL3 knockdown, TECs were transfected as described above for 24 hours prior to seeding in 24-well transwells. Endothelial cells were grown for 48 hours to form monolayers on fibronectin-coated polycarbonate membrane with a 5.0 μm pore size and 6.5 mm insert diameter in 24-well transwell plates (Costar; #3421). Preactivated T cells labeled with BCECF were pretreated with BSA or rEDIL3 (50 ng/mL) for 1 hour at 37°C. The endothelial monolayers were gently rinsed with warm RPMI1640 media before adding pretreated T cells at 1 × 105 cells/100 μL in each insert. These inserts were transferred to a new 24-well plate with lower chambers containing 0.6 mL of RPMI1640 media supplemented with or without the chemoattractant IP-10 (R&D; # 266-IP-010/CF) and incubated for 2 to 4 hours at 37°C. Transendothelial migration of the T cells was assessed by measuring the fluorescence at 485 (excitation) and 530 nm (emission) using a SpectraMax M3 Multi-Mode Microplate Reader with and without the inserts in the respective wells of the transwell plate.
Three-dimensional tumor-vascular model
The three-dimensional (3D) tumor vascular model was generated as described previously (20). Briefly, 5 × 105 cells/mL H226 cells were plated in ultra-low attachment dishes (Corning; #3471) for 24 hours to form self-assembled aggregates of cells referred to as spheroids. The H226 spheroids, were resuspended with collagen rat tail hydrogel mixture at a final concentration of 2.5 mg/mL was injected into the center gel region of the 3D microfluidic chip (AIM Biotech, #DAX01) as a source of IP-10. After incubation for 30 minutes at 37°C in sterile humidity chambers, all the side walls of one flanked channel (media channel) were coated with 150 μg/mL collagen solution (Corning; #354236) in PBS to allow for better adhesion of TECs to the channel. After 15 minutes, the channels were washed once with medium. To create 3D vessels for the tumor-vascular model or 3D vascular model, 25 μL cell suspension of 3 × 106 cells/mL of patient-derived TECs was injected into the media channel coated with collagen. The microfluidic chip was rotated twice to create a confluent hollow-lumen 3D vessel. To allow the cells to attach to the media–gel interface to form a monolayer, the chip with the media–gel interface was faced down for 15 minutes. A 50 μL cell suspension was reinjected, and the chip was placed upside down to cover the upper part of the 3D vascular channel. After 90 minutes of incubation in a humidity chamber at 37°C, the cell culture medium was gently added to both media channels. The chips were then placed in an incubator to form a confluent monolayer. After vascular formation in the tumor-vascular model or 3D vascular model, CD3+CD8+ T cells (labeled with CellTrace, Thermo Fisher Scientific #C34567) were pretreated for 1 hour with rEDIL3 (50 ng/mL) at 37°C and finally loaded in rEDIL3 supplemented media and added at a 2:1 TEC:T-cell ratio. For the 3D vascular model, IP-10 supplemented medium was added to the fluidic channel placed on the opposite side of the vascular barrier. T-cell migration across blood vessels in the middle chamber was captured at 48 to 72 hours using a Nikon Eclipse 80i fluorescence microscope equipped with a Z-stack (Prior) and a Cool SNAP CCD camera (Roper Scientific). Quantification was performed using the NIS-Elements AR software package.
Statistical analysis
Analyses of clinical responses according to fold changes (FC) in EDIL3 or MFGE8 were based on Wilcoxon rank-sum tests. Categorical analyses of response by FCs were divided at 1.5 using Fisher exact tests. Distributions of overall survival (OS) were summarized using the method of Kaplan–Meier with comparisons based on log-rank tests. Statistical significance was defined as P < 0.05, using two-sided testing. The analysis was performed using SAS9.5 (SAS Institute Inc.). In vitro statistical analyses were performed using GraphPad Prism software, version 9.3.1.
Data availability
The data generated in this study are available in the article and its Supplementary Data, or upon reasonable request from the corresponding author. The gene expression data (TPM) from the bulk RNA-seq are provided in Supplementary Table S1. Use of the singular patient sample, collected prior to 2015, adheres to our institutional guidelines permitting analysis and data sharing for publication. However, these guidelines did not include consenting language for the deposition of genomic data in public databases in line with NIH GDS Policy for post–January 25, 2015. Therefore, we are unable to provide the raw sequencing data, and sharing deidentified expression data, which directly informs our article's analysis.
Results
Ipi-Bev therapy elicited humoral responses to EDIL3 and MFGE8
To study the humoral responses in patients with advanced melanoma responding to Ipi-Bev combination therapy, posttreatment plasma samples from patients with long-term clinical benefits were screened using a protein array (Supplementary Fig. S1A). As per the manufacturer's instructions, proteins with a Z-factor of 0.4 or greater were weighed as potential targets. EDIL3 was identified as a target in this screen with plasma samples of P12 [a patient who achieved complete response (CR)] and P13 [a patient who had partial response (PR)] with Z-factor of 0.92 and 0.89, respectively (Supplementary Fig. S1B). To further assess EDIL3-specific antibody responses following Ipi-Bev treatment, pretreatment and posttreatment plasma samples from 42 patients with advanced melanoma who received Ipi-Bev combination therapy were analyzed by ELISA (Supplementary Fig. S1C–S1F). To determine whether humoral responses against EDIL3 were associated with clinical outcomes, the patients were divided on the basis of EDIL3-specific antibody titer into high and low groups at a cutoff of 50% increase considered clinically significant (i.e., FC ≥ 1.5; Fig. 1A). Six of the 7 responders showed FCs of at least 1.5 in EDIL3-specific antibody titers. Overall, a high number of patients with CR/PR according to the RECIST to Ipi-Bev combination therapy showed a significant EDIL3-specific antibody response, followed by patients with stable disease (SD) and progressive disease (PD; CR/PR = 85.7% vs. SD = 27.3%; PD = 15.4%; P = 0.006; Fig. 1B). The expression and increase in EDIL3-specific antibodies post-treatment was further confirmed by immunoblot assay (Fig. 1C) and ELISA (Supplementary Fig. S1G) with pretreatment and posttreatment plasma of 3 selected responders, P6 (PR), P17 (SD), and P26 (SD), from the Ipi-Bev cohort. Circulating EDIL3 concentrations in pretreatment and posttreatment plasma of Ipi-Bev patients were also measured (Supplementary Fig. S1H). Patients with a humoral response typically displayed an increase in the level of EDIL3-specific antibodies following initial Ipi-Bev treatment (Supplementary Fig. S2A–S2E). The antibody levels either declined after an initial increase, as observed for patients P6 and P12, or remained at the elevated level for patients P13, P17, and P26. P21 with PD status is shown as representative of patients with an inadequate EDIL3-specific antibody response (Supplementary Fig. S2F).
Humoral immune responses elicited to EDIL3 were associated with clinical outcomes in patients with metastatic melanoma receiving ipilimumab plus bevacizumab (Ipi-Bev). A, EDIL3-specific antibody FC measured by ELISA in posttreatment versus pretreatment plasma samples of 42 Ipi-Bev patients. CR/PR (green); SD, stable disease (blue); and PD, progressive disease (red). FC ≥1.5 considered clinically significant. B, Frequencies of EDIL3-specific antibody (FC ≥1.5) by clinical responses. C, Immunoblot analysis of EDIL3-specific Ig expression in pretreatment (Pre) and posttreatment (Post) patient's plasma samples of responders. Polyclonal anti-EDIL3 (Abcam) was used in the first lane. Densitometric quantification analysis for relative band intensities shown. D, Kaplan–Meier survival analyses for patients with FC ≥1.5 or <1.5 for EDIL3-specific antibody to Ipi-Bev treatment. E, Percentage of patients with EDIL3-specific antibody FC ≥1.5 in patient cohorts treated with Ipi-Bev (n = 42), ipilimumab (n = 34), PD-1 blockade (n = 21), and nivolumab-ipilimumab (n = 41) therapy.
Humoral immune responses elicited to EDIL3 were associated with clinical outcomes in patients with metastatic melanoma receiving ipilimumab plus bevacizumab (Ipi-Bev). A, EDIL3-specific antibody FC measured by ELISA in posttreatment versus pretreatment plasma samples of 42 Ipi-Bev patients. CR/PR (green); SD, stable disease (blue); and PD, progressive disease (red). FC ≥1.5 considered clinically significant. B, Frequencies of EDIL3-specific antibody (FC ≥1.5) by clinical responses. C, Immunoblot analysis of EDIL3-specific Ig expression in pretreatment (Pre) and posttreatment (Post) patient's plasma samples of responders. Polyclonal anti-EDIL3 (Abcam) was used in the first lane. Densitometric quantification analysis for relative band intensities shown. D, Kaplan–Meier survival analyses for patients with FC ≥1.5 or <1.5 for EDIL3-specific antibody to Ipi-Bev treatment. E, Percentage of patients with EDIL3-specific antibody FC ≥1.5 in patient cohorts treated with Ipi-Bev (n = 42), ipilimumab (n = 34), PD-1 blockade (n = 21), and nivolumab-ipilimumab (n = 41) therapy.
The EDIL3-specific antibody response was also associated with better OS of patients in the Ipi-Bev cohort (log-rank P = 0.04; Fig. 1D). The median survival of patients with EDIL3-specific antibody FC ≥ 1.5 was 35.1 months [95% confidence interval (CI): 12.7–not reached], whereas that of patients with FC < 1.5 median survival was 16.3 months (95% CI: 12.2–19.5). Humoral responses against EDIL3 observed in Ipi-Bev–treated patients prompted us to look at MFGE8, another angiogenic protein studied previously for potential immunotherapeutic targeting (21). MFGE8 shares structural and functional homology with EDIL3 (22). MFGE8 is a potent soluble angiogenic factor that drives the progression of melanomas. Targeting MFGE8 cooperates with conventional cancer therapies (23). Antibodies against MFGE8 were also detected in a subset of the patients (Supplementary Fig. S3A–S3D). However, the humoral response toward MFGE8 did not correlate with clinical outcomes. MFGE8-specific antibody response was associated with PD (Supplementary Fig. S3E) and no changes in the OS were observed (Supplementary Fig. S3F). To further study the humoral response to EDIL3 as a function of treatment, the pretreatment and posttreatment plasma samples of patients with metastatic melanoma treated with ipilimumab alone (n = 34), PD1 blockade alone (n = 21), or ipilimumab in combination with nivolumab (Ipi-Nivo; n = 41) were also analyzed (Fig. 1E). The Ipi-Bev cohort had the highest percentage of patients (33.3%) with FC ≥ 1.5 for EDIL3-specific antibody titers, when compared with other cohorts that had a comparable number of patients with FC ≥ 1.5: 14.7% for Ipi alone, 14.3% for PD1 blockade, and 12.2% for Ipi-Nivo combination therapy. Thus, Ipi-Bev treatment elicited a humoral response to EDIL3 in patients with advanced melanoma, and this correlated with better clinical outcomes.
EDIL3 and MFGE8 expression correlates with tumor immune exclusion
TIDE analyses, a transcriptome biomarker platform for evaluating ICB response by inferring gene roles in modulating tumor immunity (7), indicated that increased EDIL3 and MFGE8 mRNA expression across tumors correlated with an enhanced CD8+ T-cell exclusion phenotype and not dysfunction (Fig. 2A; Supplementary Fig. S4A). As β-catenin (CTNNB1), is known to promote T-cell exclusion and resistance in melanoma, it was used as a reference gene for TIDE analysis (24, 25). In line with these results, we found that the TIDE scoring model consistently showed a correlation between CTNNB1 expression and immune evasion. Nevertheless, EDIL3 and MFGE8 did not exhibit similar relationships with the TIDE projected dysfunction score, indicating that their influence was limited to exclusion. Furthermore, TIDE analysis revealed that high EDIL3 expression was substantially (P = 7.4e-11) linked with predicted-worse ICB outcomes in patients with melanoma, although MFGE8 and CTNNB1 were not associated (Fig. 2B). Thus, we found EDIL3 expression correlates with CD8+ T-cell exclusion and a predicted nonresponse to ICB therapy in patients with melanoma.
EDIL3 expression is associated with tumor T-cell immune exclusion gene signatures by TIDE analyses. A, Heat maps showing the correlation (Kendall rank) between EDIL3, MFGE8 and CTNNB1 genes expression and TIDE-predicted signatures for CTL exclusion and dysfunction respectively within TCGA tumor types. B,EDIL3, MFGE8, and CTNNB1 log2(TPM+1) expression and TIDE-predicted value for immunotherapy response in TCGA-SKCM dataset with EDIL3 expression significantly higher among predicted nonresponders (P = 7.4e-11, two-sided Mann–Whitney U test). In the box and whisker plots, the box delineates the first and third quartiles and is bisected by the median; whiskers extend to a maximum of 1.5 times the interquartile range.
EDIL3 expression is associated with tumor T-cell immune exclusion gene signatures by TIDE analyses. A, Heat maps showing the correlation (Kendall rank) between EDIL3, MFGE8 and CTNNB1 genes expression and TIDE-predicted signatures for CTL exclusion and dysfunction respectively within TCGA tumor types. B,EDIL3, MFGE8, and CTNNB1 log2(TPM+1) expression and TIDE-predicted value for immunotherapy response in TCGA-SKCM dataset with EDIL3 expression significantly higher among predicted nonresponders (P = 7.4e-11, two-sided Mann–Whitney U test). In the box and whisker plots, the box delineates the first and third quartiles and is bisected by the median; whiskers extend to a maximum of 1.5 times the interquartile range.
EDIL3 upregulation is associated with core biological pathways in melanoma
We next performed analysis of the RNA-seq data from all 469 primary and/or metastatic melanoma samples in TCGA-SKCM dataset (9). In addition, we also analyzed the CheckMate 064 dataset utilizing all 90 pretreatment samples with available RNA-seq data from both study arms (10). With increasing EDIL3 expression in TCGA-SKCM subjects, significant enrichment of TGFβ signaling, pan-fibroblast TGFβ response signature, EMT induction, and angiogenic signatures was observed (Fig. 3A). TCGA-SKCM analysis was contrasted with the CheckMate 064 dataset generated from subjects with metastatic melanoma undergoing an open-label, randomized, phase II study of nivolumab administered sequentially with ipilimumab. The expression profile of EDIL3 along with the gene sets of selected pathways from TCGA-SKCM analysis were visualized as a heat map for 90 samples from the CheckMate 064 dataset. The CheckMate 064 dataset analysis resonated with observations from TCGA-SKCM analysis (Fig. 3B).
EDIL3 expression associated with TGFβ signaling, EMT, and angiogenesis signatures. A and B,EDIL3 expression correlated with core biological pathways. EDIL3 expression ordered from low (left) to high (right). Rows display gene expression (z scores normalized) for genes grouped under pathway expression for TGFβ signaling, TGFβ signaling in fibroblast, EMT and angiogenesis signatures in TCGA-SKCM (n = 469; A) and CheckMate 064 (n = 90; B) databases. C,EDIL3 expression correlated with CAF FAP signature score using TIDE.
EDIL3 expression associated with TGFβ signaling, EMT, and angiogenesis signatures. A and B,EDIL3 expression correlated with core biological pathways. EDIL3 expression ordered from low (left) to high (right). Rows display gene expression (z scores normalized) for genes grouped under pathway expression for TGFβ signaling, TGFβ signaling in fibroblast, EMT and angiogenesis signatures in TCGA-SKCM (n = 469; A) and CheckMate 064 (n = 90; B) databases. C,EDIL3 expression correlated with CAF FAP signature score using TIDE.
The gene query tool on the TIDE platform was used to evaluate EDIL3 and MFGE8 expression in immunosuppressive cell types that drive T-cell exclusion from the TME. Among the cell types promoting T-cell exclusion, CAFs showed high EDIL3 expression levels, followed by tumor-associated macrophages (TAM) of the M2-like phenotype. However, myeloid-derived suppressor cells (MDSC) were negatively associated with EDIL3 expression. Expression of MFGE8 was also found to be strongly associated with CAFs and weakly associated with TAM of the M2-like phenotype followed by MDSCs (Fig. 3C). These analyses indicated that immunosuppressive CAFs are an alternate source of EDIL3 associated with TGFβ signaling, EMT activation, and angiogenesis in melanoma tumors.
EDIL3 is overexpressed in CAFs isolated from melanoma patient's tumors and linked with TGFβ signaling
Because TIDE gene query analysis suggested EDIL3 expression in CAFs, we conducted studies on fibroblasts isolated from melanoma tumors. The expression of the CAF markers FAP and LRRC15 was confirmed by bulk RNA-seq (Supplementary Fig. S4B and S4C). Compared with NFs, patient-derived CAFs (P4-CAF and CAF2) secreted abundant levels of EDIL3 in the medium, as detected by ELISA [P4-CAF (P < 0.0001); CAF2 (P < 0.0001)] (Fig. 4A). TGFβ treatment of NF activated downstream SMAD signaling. Upregulation of EDIL3 protein expression in whole-cell lysates and secretion in the conditioned medium in a TGFβ concentration–dependent manner was detected by immunoblot assay and ELISA, respectively (Fig. 4B and C). Inhibition of TGFβ signaling using LY2109761 led to the downregulation of TGFβ-induced EDIL3 expression in both cellular and extracellular compartments, along with the inhibition of Smad2 activation, establishing the regulation of EDIL3 expression via TGFβ. Bulk RNA-seq expression data analysis also validated the overexpression of EDIL3 and MFGE8 in P4-CAF versus NF and suggested only EDIL3 induction under TGFβ regulation (Supplementary Fig. S4D). The role of EDIL3 in TGFβ-induced EMT in P4-CAF was assessed. Silencing EDIL3 in P4-CAF led to the downregulation of the key EMT marker transgelin (TGLN) at the basal level and significantly abrogated TGFβ-induced expression of TGLN and smooth muscle α-actin (ACTA2; Fig. 4D, E, and F). β-actin was employed as a reference gene because it was found to be the most stable housekeeping gene examined by Normfinder (Supplementary Fig. S4F). Next, the effect of VEGF blockade on P4-CAF–derived EDIL3 expression was assessed. Immunoblot analysis of whole-cell lysates of P4-CAF treated with VEGF-A or/and bevacizumab were performed. EDIL3 expression was downregulated by bevacizumab alone and in combination with VEGF-A (Fig. 4G). Altogether, as suggested by the TIDE analyses, in CAFs from patients with melanoma, EDIL3 is overexpressed and subject to TGFβ regulation, which in turn controls the associated EMT.
EDIL3 is abundantly expressed in CAFs and is upregulated by TGFβ. A, Detection by ELISA of secreted EDIL3 in conditioned medium from NF and patient-derived CAFs (P4-CAF and CAF2). B and C, NF pretreated with or without LY2109761 (LY) followed by TGFβ treatment for 24 hours. EDIL3 expression was examined by B, ELISA of conditioned medium and C, immunoblot analyses of whole-cell lysates. D, qRT-PCR analysis of EDIL3 silencing by siRNA in P4-CAF. EDIL3 mediated regulation of TGFβ target genes Transgelin (TAGLN;E) and α-SMA (ACTA2;F). G, Immunoblot analysis of EDIL3 expression in P4-CAFs treated with VEGF-A (20 ng/mL) or/and bevacizumab (25 μg/mL) for 24 hours. β-actin was used as a loading control. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as **, P < 0.01; ****, P < 0.0001.
EDIL3 is abundantly expressed in CAFs and is upregulated by TGFβ. A, Detection by ELISA of secreted EDIL3 in conditioned medium from NF and patient-derived CAFs (P4-CAF and CAF2). B and C, NF pretreated with or without LY2109761 (LY) followed by TGFβ treatment for 24 hours. EDIL3 expression was examined by B, ELISA of conditioned medium and C, immunoblot analyses of whole-cell lysates. D, qRT-PCR analysis of EDIL3 silencing by siRNA in P4-CAF. EDIL3 mediated regulation of TGFβ target genes Transgelin (TAGLN;E) and α-SMA (ACTA2;F). G, Immunoblot analysis of EDIL3 expression in P4-CAFs treated with VEGF-A (20 ng/mL) or/and bevacizumab (25 μg/mL) for 24 hours. β-actin was used as a loading control. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as **, P < 0.01; ****, P < 0.0001.
EDIL3 correlates with angiogenesis and its signatures in TCGA-SKCM and CheckMate 064 databases
To assess the potential angiogenic roles of EDIL3, pretreatment plasma from 39 Ipi-Bev–treated patients was analyzed for the interdependence of circulating levels of VEGF and EDIL3. The Spearman rank correlation was 0.44 (P = 0.005), suggesting a moderate and positive relationship, that is, high pretreatment EDIL3 levels were associated with high pretreatment VEGF levels (Fig. 5A).
Circulating EDIL3 expression correlates with serum VEGF levels and angiogenesis signatures in TCGA-SKCM and CheckMate 064 databases. A, Spearman rank correlation analysis of pretreatment circulating serum levels of EDIL3 versus VEGF-A in Ipi-Bev–treated patients with melanoma (n = 39). B,EDIL3 log2(TPM+1) transformed expression is significantly associated with high angiogenic signatures (Angio) in TCGA-SKCM and CheckMate 064 datasets (two-sided Mann–Whitney U test). C and D, EDIL3 (50 ng/mL) promoted migration of patient-derived endothelial cells similar to VEGF-A (20 ng/mL) assessed by wound healing assay. E–G, EDIL3 promoted tube formation ability of patient-derived endothelial cells similar to VEGF-A. All results are presented as means ± SD and represent three independent experiments. Statistical significance indicated by P values or as *, P < 0.05.
Circulating EDIL3 expression correlates with serum VEGF levels and angiogenesis signatures in TCGA-SKCM and CheckMate 064 databases. A, Spearman rank correlation analysis of pretreatment circulating serum levels of EDIL3 versus VEGF-A in Ipi-Bev–treated patients with melanoma (n = 39). B,EDIL3 log2(TPM+1) transformed expression is significantly associated with high angiogenic signatures (Angio) in TCGA-SKCM and CheckMate 064 datasets (two-sided Mann–Whitney U test). C and D, EDIL3 (50 ng/mL) promoted migration of patient-derived endothelial cells similar to VEGF-A (20 ng/mL) assessed by wound healing assay. E–G, EDIL3 promoted tube formation ability of patient-derived endothelial cells similar to VEGF-A. All results are presented as means ± SD and represent three independent experiments. Statistical significance indicated by P values or as *, P < 0.05.
On the basis of the angiogenetic signature panel (12) consisting of TEK, CDH5, SOX17, and SOX18, 469 subjects from TCGA-SKCM dataset were divided into high (n = 83) and low (n = 386) angiogenesis groups by agglomerative clustering with two clusters of the angiogenesis pathway expression. Consistent with our correlation observation above, the expression of EDIL3 was significantly upregulated (P = 1.3e-07) in patients with a high angiogenic phenotype compared with that in patients with a low angiogenic phenotype. The CheckMate 064 dataset analysis with 90 patients divided into 22 high and 68 low angiogenesis groups based on angiogenic signature also showed a significant relationship (P = 0.0002), that is, EDIL3 was resolutely associated with high angiogenesis in patients with advanced melanoma (Fig. 5B).
EDIL3 expression was observed to be high in endothelial cells with higher angiogenic potential by qPCR (HDMEC, TECs, and HUVEC; Supplementary Fig. S5A). Further angiogenic functional evaluation showed that rEDIL3 promoted the migration of TECs compared with the untreated control, comparable with VEGF-A, at 24 and 48 hours (Fig. 5C and D). rEDIL3 also promoted the development of more capillary-like structures, which were denser than the untreated control and similar to those induced by VEGF-A treatment (Fig. 5E). A significant increase in the number of branches (mean mesh size, P < 0.05) and mesh index (P < 0.05), similar to VEGF-A, was observed with rEDIL3 treatment compared with control TECs (Fig. 5F and G). Our findings suggest that EDIL3 is a positive regulator of pathologic angiogenesis in the melanoma TME.
EDIL3 blocks immune–endothelial cell adhesion and inhibits T-cell transmigration
The adhesion of immune cells is a prerequisite for transendothelial migration, and the interaction of LFA-1 (also known as CD11a) on immune cells and ICAM-1 on endothelial cells is one of the crucial steps. EDIL3 antagonizes adhesion dependent on LFA-1/ICAM-1 (26). To validate the functional role of EDIL3 in immune exclusion in the TME, activated immune cells, that is, THP1 cells, Jurkat T cells, and primary T cells, were screened for the expression of LFA-1 by flow cytometric analysis (Fig. 6A). To study EDIL3-mediated immune cell exclusion, an adhesion assay was performed using patient-derived TECs and CD3/CD28-activated primary T cells (Fig. 6B). rEDIL3 pretreatment of activated T cells inhibited their adhesion to unstimulated TEC monolayers compared with untreated activated T cells (P < 0.01). TNFα upregulated ICAM-1 expression in TECs (Supplementary Fig. S5B), leading to an increase in the adhesion of untreated T cells by 22% compared with that of unstimulated TEC monolayers. However, rEDIL3 pretreatment dose-dependently inhibited the adhesion of activated T cells on stimulated TEC monolayers as compared with the TNFα treated control at 50 and 200 ng/mL showing close significant inhibition (P < 0.01; Fig. 6C). Activated T cells bound to immobilized rEDIL3 with a higher affinity than rICAM-1 (P < 0.05; Supplementary Fig. S5C). Activated T cells pretreated with a combination of rEDIL3 and LFA-1–specific antibody showed 54% (P < 0.01) inhibition of adhesion, as compared with 43% (P < 0.01) inhibition when treated with rEDIL3 or LFA-1–specific antibody alone on TNFα-treated monolayers of TEC as control (Fig. 6D; Supplementary Fig. S5D). Thus, disruption of the LFA-1 and ICAM-1 interaction is one of the underlying mechanisms of EDIL3-mediated inhibition of adhesion of T cells to TECs.
EDIL3 blocks immune–endothelial cell adhesion and inhibits T-cell migration. A, Expression of LFA-1(CD11a) on immune cells by flow cytometry. B, Schematic representation of immune–endothelial cell adhesion assay showing EDIL3 antagonizes LFA-1/ICAM-1–dependent adhesion. Graphics created with Biorender. C, rEDIL3 (10, 50, and 200 ng/mL) inhibits adhesion of activated T cells to TNFα-activated tumor endothelial cell's monolayer. D, rEDIL3 (50 ng/mL) blocks the immune–endothelial cell adhesion by interfering with the LFA-1 and ICAM-1 interaction. E,EDIL3 silencing in endothelial cells potentiated T-cell transendothelial migration. F, rEDIL3 partially blocked T-cell transendothelial migration induced by IP-10. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as *, P < 0.05; **, P < 0.01.
EDIL3 blocks immune–endothelial cell adhesion and inhibits T-cell migration. A, Expression of LFA-1(CD11a) on immune cells by flow cytometry. B, Schematic representation of immune–endothelial cell adhesion assay showing EDIL3 antagonizes LFA-1/ICAM-1–dependent adhesion. Graphics created with Biorender. C, rEDIL3 (10, 50, and 200 ng/mL) inhibits adhesion of activated T cells to TNFα-activated tumor endothelial cell's monolayer. D, rEDIL3 (50 ng/mL) blocks the immune–endothelial cell adhesion by interfering with the LFA-1 and ICAM-1 interaction. E,EDIL3 silencing in endothelial cells potentiated T-cell transendothelial migration. F, rEDIL3 partially blocked T-cell transendothelial migration induced by IP-10. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as *, P < 0.05; **, P < 0.01.
Next, we determined whether EDIL3 affected the transendothelial migration of preactivated T cells across the TEC monolayer. A positive regulator of the chemotaxis chemokine IP-10, known to stimulate T-cell migration, was used as a chemoattractant. Migration of activated T cells across TEC monolayers grown on fibronectin-coated microporous membranes of transwells with and without IP-10 in the lower chamber was examined using the Boyden chamber assay. IP-10 enhanced the migration of activated T cells when added to the lower chamber compared with the control. Silencing EDIL3 in TECs increased the transmigration of T cells compared with the control by 37% (P < 0.05). The IP-10–induced transmigration was also potentiated following EDIL3 silencing by 21% (P < 0.05; Fig. 6E; Supplementary Fig. S5E). rEDIL3 pretreatment of T cells decreased IP-10–induced transmigration of T cells across the endothelial monolayer by 23% (P < 0.05; Fig. 6F). We also screened melanoma cell lines and regulatory T cells from two donors for EDIL3 expression. Heterogenous expression of EDIL3 in melanoma cells lines with K008 showing highest expression was observed. Whereas none of the regulatory T cells expressed EDIL3 (Supplementary Fig. S5F).
The transmigration of CD8+ T cells in the presence of rEDIL3 was further studied using microfluidic 3D cocultures. Two models of coculture were used to study the transmigration of CD8+ T cells across the endothelium as a model of the vascular barrier. In the first model, the 3D vascular model, IP-10 was added to the last chamber and a large 3D vessel was perfused with rEDIL3 pretreated labeled CD8+ T cells and allowed to transmigrate over 48 hours. While IP-10 induced transmigration by 2.3-fold (P < 0.01), the addition of rEDIL3 pretreated CD8+ T cells and rEDIL3 to the medium decreased the overall IP-10–induced migration (P < 0.01; Fig. 7A). In the second model, the tumor vascular model, extracellular matrix containing tumor spheroids was added to the middle chamber as a source of IP-10 to establish a chemotactic gradient to facilitate CD8+ T-cell transmigration. rEDIL3 pretreatment of CD8+ T cells significantly (P < 0.0001) downregulated their transmigration across the vascular barrier (Fig. 7B).
EDIL3 blocks transendothelial CD8+ T-cell migration in microfluidic 3D cocultures. Representative images and quantification of migrated CD8+ T-cell numbers in adjacent extracellular matrix (region of interest) after 48 hours of coculture with, vasculature from patient-derived endothelial cells in the presence of the chemoattractant IP-10 (A) or tumor spheroids as the source of IP-10 with vasculature from patient-derived endothelial cells (B). rEDIL3 (50 ng/mL) pretreatment of CD8+ T cells blocked their transendothelial migration. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as **, P < 0.01; ****, P < 0.0001. C, Model of effects for anti-EDIL3 humoral responses in patients with advanced cancer: Tumor-derived EDIL3 mediates T-cell exclusion in the TME. EDIL3 binds to LFA-1 on T cells and prevents their adherence via ICAM-1 on TECs, thus impeding immune cell extravasation into the TME. ICB may augment a humoral response against EDIL3 accompanied by activated tumor endothelium and increased CD8+ T-cell infiltration. Patients with melanoma with anti-EDIL3 humoral immune responses demonstrated improved therapeutic response to treatment. In addition, immunosuppressive CAFs were identified as an alternate source for EDIL3 in the TME. Graphics created with Biorender.
EDIL3 blocks transendothelial CD8+ T-cell migration in microfluidic 3D cocultures. Representative images and quantification of migrated CD8+ T-cell numbers in adjacent extracellular matrix (region of interest) after 48 hours of coculture with, vasculature from patient-derived endothelial cells in the presence of the chemoattractant IP-10 (A) or tumor spheroids as the source of IP-10 with vasculature from patient-derived endothelial cells (B). rEDIL3 (50 ng/mL) pretreatment of CD8+ T cells blocked their transendothelial migration. All results are presented as means ± SD and represent three independent experiments. Statistical significance was determined by t test and indicated by P values or as **, P < 0.01; ****, P < 0.0001. C, Model of effects for anti-EDIL3 humoral responses in patients with advanced cancer: Tumor-derived EDIL3 mediates T-cell exclusion in the TME. EDIL3 binds to LFA-1 on T cells and prevents their adherence via ICAM-1 on TECs, thus impeding immune cell extravasation into the TME. ICB may augment a humoral response against EDIL3 accompanied by activated tumor endothelium and increased CD8+ T-cell infiltration. Patients with melanoma with anti-EDIL3 humoral immune responses demonstrated improved therapeutic response to treatment. In addition, immunosuppressive CAFs were identified as an alternate source for EDIL3 in the TME. Graphics created with Biorender.
Overall, these data show that one of the mechanisms by which EDIL3 prevents CD8+ T-cell transmigration across the vascular barrier and causes exclusion is by interfering with the interaction of LFA-1 and ICAM-1 on TECs, which is essential for adhesion.
Discussion
EDIL3, also known as DEL-1 (developmental endothelial locus 1) was identified as the top target in our serologic screen. EDIL3 is a proangiogenic factor associated with tumor progression and poor prognosis in multiple malignancies (27). Multiple studies have reported that EDIL3 overexpression in the TME drives tumor progression (28–30) by promoting EMT (31, 32), anoikis (29, 33), angiogenesis (32), and metastasis (34, 35). However, relationships between EDIL3 and immune exclusion, EMT, and angiogenesis, as well as its prognostic significance in patients with cancer, and potential as a therapeutic target are currently under investigated. The current study investigated the predictive and prognostic value of humoral responses to EDIL3 in ICB therapy as well as the immune regulation mediated by EDIL3 in the TME. The immune modulatory effects of MFGE8 have previously been studied (21). Given the substantial sequence homology and similarity of functional properties of MFGE8 and EDIL3, it was intriguing to compare their underlying immunologic processes (36). They are homologous glycoproteins possessing an evolutionarily conserved Arg-Gly-Asp (RGD) motif. Both are implicated in the regulation of leukocyte recruitment and inflammation (26, 37). The cross-presentation of immunogenic antigens is known to be strengthened by systemic targeting of MFGE8 (37). The combination of Ipi-Bev therapy elicited a humoral immune response towards EDIL3, which was associated with clinical benefit in patients with advanced melanoma. Interestingly, humoral responses to MFGE8 were also observed but, unlike the responses to EDIL3, they were not associated with clinical benefit. MFGE8 also showed no predictive value for ICB in our TIDE analysis. Although overexpressed in CAFs, MFGE8 was not regulated by TGFβ suggesting a different mechanism of action to EDIL3. Among the screened cohorts, EDIL3-specific antibody responses were most robust in patients treated with a combination of Ipi-Bev compared with Ipi alone, anti-PD1 alone, or the combination of Ipi-Nivo. Bevacizumab in combination therapies is reported to increase CD8+ T-cell infiltration (38) and downregulate the TGFβ signature (39, 40), whereas EDIL3 was found to be under TGFβ control, facilitating CD8+ T-cell exclusion. EDIL3 could be an indirect target of bevacizumab and may serve as a biomarker of response toward Ipi-Bev combination therapy, although this requires further investigation. Although it was also associated with ICB alone, the contribution of the EDIL3 humoral responses to the outcomes of particular treatments may differ. In addition, the relevant roles of circulating EDIL3 and EDIL3-specific antibodies may be considered and utilized in prognostic and predictive evaluations.
Our TIDE correlation analysis delineated the positive association between EDIL3 expression and gene signatures of immune evasion through CD8+ T-cell exclusion in large melanoma patient cohorts. High EDIL3 expression in SKCM subjects was associated with TIDE prediction of nonresponse to ICB therapy, consistent with previous reports of EDIL3 as a poor prognostic marker in multiple malignancies (29, 30). Our in vitro transendothelial migration studies demonstrated rEDIL3-mediated CD8+ T-cell exclusion at the vascular–immune interface. EDIL3 acts as an endogenous inhibitor of LFA-1–dependent leukocyte recruitment for inflammatory diseases (26). Thus, our observations may be important for understanding the robust CD8+ T-cell infiltration seen in patients who benefit from Ipi-Bev treatment while developing humoral immunity to EDIL3. Recently, EDIL3 has been acknowledged as an important player in T-cell immunity by promoting regulatory T-cell responses under inflammatory conditions by upregulating FOXP3 expression through αvβ3 integrin signaling (41). We also observed that the capacity of TECs for migration and tube formation was promoted by EDIL3. This might be the result of signaling between the RGD motif of EDIL3 and αvβ3 on endothelial cells that causes endothelial attachment, a requirement for migration, prevents endothelial cell death, and thus promotes angiogenesis (42). Additional studies are warranted to delineate the relative functional roles of LFA-1 and αvβ3 integrin signaling with EDIL3 in the TME.
Among the cell types known to mediate T-cell exclusion, CAFs were found to be strongly associated with EDIL3 expression. Very few studies have explored the role of EDIL3 in fibroblasts, emphasizing current findings. Colorectal carcinoma liquid biopsy analysis traced EDIL3 as a plasma-derived exosomal cargo protein exclusively associated with patient-derived CAFs (43). CAF-secreted EDIL3 has been proposed as a prognostic signature associated with clinical outcomes, tumor progression, and genetic alterations in breast cancer (44). In line with our TCGA analysis, TGFβ induced the expression of EDIL3 and its silencing inhibited TGFβ1-induced EMT phenotype markers in CAFs. This suggests an EDIL3-mediated molecular feedback loop in the regulation of TGFβ-induced EMT in CAFs. Mariathasan and colleagues reported that TGFβ signaling in fibroblasts is associated with a lack of response toward ICB, particularly in tumors showing CD8+ T-cell exclusion.
In conclusion, our data indicate that EDIL3 produced by CAFs may support immune evasion through CD8+ T-cell exclusion and the regulation of TGFβ-induced EMT in melanoma tumors. EDIL3 could be one of the underlying targets of bevacizumab, effectuating an excluded to an inflamed phenotype observed in responders to treatment (Fig. 7C). Hence, EDIL3 is not only identified by our findings as a useful mechanistic insight into potential synergistic effects but also as a therapeutic target.
Authors' Disclosures
S. Tabasum reports other support from Bristol Myers Squibb during the conduct of the study; in addition, S. Tabasum has a patent for provisional patent application pending. D. Thapa is currently an employee of Agenus. This research was conducted while D. Thapa was a research fellow at DFCI in the absence of any relationships or influence of Agenus. J.L. Weirather reports other support from Bristol Myers Squibb during the conduct of the study. D.A. Barbie reports other support from Xsphera Biosciences; grants from Bristol Myers Squibb and Daiichi Sankyo; and personal fees from Qiagen outside the submitted work. F.S. Hodi reports grants from NIH, nonfinancial support from Bristol Myers Squibb and Genentech during the conduct of the study; grants and personal fees from Bristol Myers Squibb; personal fees from Merck, Novartis, Surface, Compass, Apricity, Bicara, Pieris Pharmaceutical, Checkpoint Therapeutics, Genentech, Bioentre, Gossamer, Iovance, Catalym, Immunocore, Amgen, Kairos, Rheos, Zumutor, Corner Therapeutics, Puretech, Curis, and AstraZeneca outside the submitted work; in addition, F.S. Hodi has a patent for “Methods for treating MICA-related disorders” (#20100111973) pending, licensed, and with royalties paid, a patent for “Tumor antigens and uses thereof” (#7250291) issued, a patent for “Angiopoiten-2 biomarkers predictive of anti-immune checkpoint response” (#20170248603) pending, a patent for “Compositions and methods for identification, assessment, prevention, and treatment of melanoma using PD-L1 isoforms” (#20160340407) pending, a patent for “Therapeutic peptides” (#20160046716) pending, a patent for “Methods of using pembrolizumab and trebananib” pending, a patent for “Vaccine compositions and methods for restoring NKG2D pathway function against cancers” (#10279021) pending, licensed, and with royalties paid, a patent for “Antibodies that bind to MHC class I polypeptide-related sequence” (10106611) pending, licensed, and with royalties paid, a patent for “Antibodies that bind to MHC class I polypeptide-related sequence” (#10106611), a patent for “Anti-galectin antibody biomarkers predictive of anti-immune checkpoint and anti-angiogenesis responses” (#20170343552) pending, and a patent for “Antibodies against EDIL3 and methods of use thereof” pending. No disclosures were reported by the other authors.
Authors' Contributions
S. Tabasum: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. D. Thapa: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–review and editing. A. Giobbie-Hurder: Data curation, formal analysis, investigation, methodology, writing–review and editing. J.L. Weirather: Data curation, formal analysis, investigation, methodology. M. Campisi: Formal analysis, validation, investigation, methodology, writing–review and editing. P.J. Schol: Formal analysis, validation, investigation. X. Li: Resources, data curation, methodology, writing–review and editing. J. Li: Resources, data curation, writing–review and editing. C.H. Yoon: Resources. M.P. Manos: Resources, data curation, project administration. D.A. Barbie: Resources, supervision, validation. F.S. Hodi: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, project administration, writing–review and editing.
Acknowledgments
This work was supported by the E. Michael Egan Melanoma Research Fund, Jane and Brian Crowley Melanoma Research Fund, Vanessa and Anthony Beyer Fund for Melanoma Research, Malcolm and Emily MacNaught Fund for Melanoma Research, and Christin Holbrook Harding Fund for Melanoma Research at the Dana-Farber Cancer Institute.
We appreciate Xinqi Wu's contributions in establishing screening methods and scientific discussions throughout the project. The results shown in Fig. 2 are in whole or part based upon data generated by TCGA Research Network: https://www.cancer.gov/tcga.
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).