Targeting the tumor cell-intrinsic ITGB2 axis inhibits melanoma progression

Abstract

Background

ITGB2 is a critical integrin mediator of immune cell activation and trafficking. Its expression has been claimed as exclusive to hematopoietic cells. Consequently, the significance of cancer cell-intrinsic ITGB2 in solid tumor progression and therapy has not been rigorously evaluated.

Methods

We leveraged single-cell and bulk RNA sequencing, real-time quantitative PCR, multiplex immunofluorescence, flow cytometry, immunoblotting, and intercellular adhesion molecule (ICAM)-1-dependent adhesion and proliferation assays to uncover melanoma cell-intrinsic ITGB2 functional expression, association with clinical tumor progression, activation, protumorigenic signaling, adhesive and proliferative functions utilizing patient melanoma biospecimens, established human and murine melanoma lines. In vivo tumorigenicity studies in immunocompromised NOD/SCID interleukin-2 receptor γ chain null (NSG), immunocompetent wildtype, and Icam1 knockout (KO) C57BL/6 mice were performed to dissect melanoma-ITGB2 downstream pathway activity and functions in tumor growth and metastasis. The cancer cell-intrinsic ITGB2 axis was targeted using CRISPR/Cas9-based Itgb2 KO, blocking ITGB2 antibodies, ITGB2-activating CD44 crosslinking, and pharmacologic inhibition of ITGB2-dependent Wnt signaling using LGK974, zamaporvint, and FDA-approved pyrvinium pamoate repurposed for cancer therapy.

Results

This work demonstrates nonhematopoietic expression and protumorigenic functions of ITGB2 intrinsic to melanoma cells. Tumor cell-ITGB2 mediated adhesion to ICAM-1, promoted cancer progression in preclinical melanoma models, was enriched in clinical metastatic versus primary melanomas or benign nevi, and predicted sentinel lymph node metastasis in patients with primary disease. Consistently, inhibition of melanoma cell-intrinsic ITGB2 using blocking antibodies or Itgb2 gene KO potently suppressed ICAM-1-mediated melanoma cell adhesion, tumor growth, and metastatic dissemination. Melanoma cell-ITGB2:ICAM-1 interaction activated downstream Wnt signaling, the pharmacologic inhibition of which suppressed melanoma-ITGB2-mediated tumorigenesis.

Conclusions

This work overturns the longstanding paradigm that ITGB2 is restricted to leukocytes by discovering a tumor cell-intrinsic ITGB2:ICAM-1:Wnt protumorigenic axis as a bona fide cancer therapeutic target in melanoma.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12943-025-02527-z.

Background

The integrin superfamily of adhesion molecules is a prominent regulator of diverse immune cell functions, including activation, proliferation and homing [1]. Of the 24 known integrin (ITG) αβ heterodimeric receptors [2], several have also been implicated in cancer progression, including of melanoma [3]. However, their inhibition proved largely ineffective in thwarting tumor outgrowth owing to functional redundancy and nonselective expression across multiple cell types [2]. In contrast to these integrins, ITGB2 (β2) heterodimers stand apart, having evolved as specialized, non-redundant homing receptors essential for hematopoietic cell proliferation and trafficking to sites of inflammation, including into tumor tissue [1]. ITGB2-mediated homing involves interaction with intercellular adhesion molecule (ICAM)−1 [2]. Inhibition of ITGB2 family members alone on immune cells is sufficient to fully abrogate activation and trafficking [1, 4], thereby underscoring the preeminence of the ITGB2:ICAM-1 axis above all other adhesive interactions in regulating immune cell functions. This is exemplified in patients with leukocyte adhesion deficiency-1 (LAD-1), a genetic disease characterized by ITGB2-inactivating mutations, which fully abrogate immune cell activation and homing resulting in life-threatening vulnerability to infection [1].

The ITGB2 invariant chain pairs with four distinct α subunits, ITGAD (αD, CD11d), ITGAL (αL, CD11a, LFA-1), ITGAM (αM, CD11b, Mac-1), or ITGAX (αX, CD11c), of which ITGAL functionally predominates in T cell-driven immune conditions [2]. To date, it has been widely claimed that all four ITGB2 heterodimers are exclusively expressed only by leukocytes and thus absent from other cell types [5]. However, this view has recently been challenged by a few reports of tumor cell-intrinsic ITGB2 expression in epithelial cancers, albeit with incompletely defined functions in tumorigenesis [6–8]. Here we report nonhematopoietic ITGB2 gene and protein expression by melanoma cells. Crosslinking of the CD44 hyaluronic acid receptor known to regulate ITGAL/ITGB2 functional activation on leukocytes [9] also induced ITGAL/ITGB2 expression and conformational activation on melanoma cells. Cancer cell-ITGB2 promoted in vitro adhesion to ICAM-1, in vivo tumorigenesis in human and murine melanoma models, was enriched in clinical metastatic versus primary melanomas or benign nevi, and correlated with sentinel lymph node (SLN) metastasis in patients with primary melanomas.

Inhibition of melanoma cell-intrinsic ITGB2 using blocking antibodies (abs) or CRISPR/Cas9-mediated Itgb2 gene knockout (KO) abrogated melanoma cell adhesion to ICAM-1 and in vivo tumor growth in immunocompromised nonobese diabetic (NOD)-severe combined immunodeficiency (SCID) interleukin 2 receptor γ chain-null (NSG) mice. In immunocompetent C57BL/6 mice, both ITGB2 blockade or host-Icam1 deficiency significantly attenuated spontaneous dissemination of murine B16-F10 and YUMM5.2 tumors. Melanoma cell-intrinsic Itgb2 KO also inhibited primary tumor growth in wildtype, but not Icam1^−/− null hosts. Unbiased RNA-sequencing (seq) analyses uncovered the Wnt pathway as a downstream target of melanoma cell-ITGB2:ICAM-1. Indeed, pharmacologic Wnt antagonism suppressed melanoma-ITGB2-mediated adhesion and tumorigenesis. Our work identifies a tumor cell-intrinsic ITGB2:ICAM-1:Wnt axis as a growth-accelerating mechanism and bona fide cancer therapeutic target in melanoma. These results overturn the longstanding paradigm that ITGB2 is exclusive to leukocytes and uncover a ‘leukocyte mimicry’ mechanism [10] exploited by solid tumor cells to promote cancer progression.

Methods

Cell culture

Authenticated, mycoplasma-free human A2058 (Cat# CRL-3601, RRID: CVCL_1059), A375 (Cat# CRL-1619, RRID: CVCL_0132), and MDA-MB-435S (Cat# HTB-129, RRID: CVCL_0622) as well as murine B16-F10 (Cat# CRL-6475, RRID: CVCL_0159), YUMM1.7 (Cat# CRL-3362, RRID: CVCL_JK16), YUMM3.3 (Cat# CRL-3365, RRID: CVCL_JK36), YUMM4.1 (Cat# CRL-3366, RRID: CVCL_JK38), and YUMM5.2 (Cat# CRL-3367, RRID: CVCL_JK43) melanoma cell lines were obtained from the American Type Culture Collection (ATCC, Gaithersburg, MD). Human melanoma LOX-IMVI cells (RRID: CVCL_1381) were from MilliporeSigma (Burlington, MA), C8161 cells (RRID: CVCL_6813) were from Dr. Mary Hendrix (Children’s Memorial Research Center, Chicago, IL), and FEMX cells (RRID: CVCL_A011) were from Dr. Udo Schumacher (University Hospital Hamburg-Eppendorf, Hamburg, Germany). The human HSB-2 (Cat# CCL-120.1, RRID: CVCL_1859) and murine EL-4 (Cat# TIB-39, RRID: CVCL_0255) T lymphoblastic lines as well as human umbilical vein (HUVEC, Cat# CRL-1730, RRID: CVCL_2959) and murine C166 (Cat# CRL-2581, RRID: CVCL_6581) endothelial cells were obtained from ATCC. All cell lines were used at low passage, < 70% confluency, and were cultured in DMEM or RPMI-1640 medium (Life Technologies, Carlsbad, CA) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS, MilliporeSigma) and 1% (v/v) penicillin/streptomycin (Life Technologies) as described [11]. HUVEC were cultured in fibronectin-coated (MilliporeSigma, 20 µg/ml) flasks and maintained in EGM-2 BulletKit Endothelial Cell Growth Medium (Lonza, Basel, Switzerland).

Clinical specimens

All studies involving human specimens were approved by Institutional Review Boards of Mass General Brigham, under protocol numbers 2022P002062, 2022P000827, and 2013P001014; the University of Zurich, Switzerland, under biobank regulation protocol number: PB.2018_00194; and the University of Bonn, Germany, under protocol number 187/16. Written informed consent was obtained from all subjects. All studies were conducted in accordance with the Declaration of Helsinki. Single-cell suspensions were generated from human melanoma specimens by collagenase digestion, as described [12].

Antibodies and biologic reagents

The following abs and reagents were used for flow cytometric (FC) analysis: PE-conjugated anti-human ITGB2 ab (clone MEM-48, Thermo Fisher Scientific, Waltham, MA, Cat# A15766, RRID: AB_2534546) or PE-conjugated mouse IgG1 isotype control ab (clone MOPC-21, Thermo Fisher Scientific, Cat# MA1-10415, RRID: AB_2536783), APC-conjugated anti-human ITGB2 ab (clone MEM-48, Thermo Fisher Scientific, Cat# MA1-19456, RRID: AB_1071389) or APC-conjugated mouse IgG1 isotype control ab (clone MOPC-21, Thermo Fisher Scientific, Cat# MA5-18093, RRID: AB_2539476), PE-conjugated anti-human ITGB2 ab (clone MEM-148, Novus Biologicals, Centennial, CO, Cat# NB500-480PE, RRID: AB_3191646) or PE-conjugated mouse IgG1 isotype ab (clone MG1, Novus Biologicals, Cat# NBP1-97005PE, RRID: AB_3242285), PE-conjugated anti-human ITGB2 ab (clone KIM-127, Leinco Technologies, St. Louis, MO, Cat# C552) or PE-conjugated mouse IgG1 isotype ab (clone HKSP, Leinco Technologies, Cat# I-103), APC-conjugated anti-human ITGAL ab (clone HI111, BioLegend, San Diego, CA, Cat# 301212, RRID: AB_314150) or APC-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400120, RRID: AB_2888687), Pacific Blue-conjugated anti-human ICAM-1 ab (clone HCD54, BioLegend, Cat# 322716, RRID: AB_893384) or Pacific Blue-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400131, RRID: AB_2923473), BV570-conjugated anti-human CD3 ab (clone UCHT1, BioLegend, Cat# 300436, RRID: AB_2562124) or BV570-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400160, RRID: AB_10900440), AF700-conjugated anti-human CD19 ab (clone SJ25C1, BioLegend, Cat# 363034, RRID: AB_2616936) or AF700-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400144, RRID: AB_2923250), PerCP-conjugated anti-human/mouse ITGAM ab (clone M1/70, BioLegend, Cat# 101230, RRID: AB_2129375) or PerCP-conjugated rat IgG2b isotype ab (clone RTK4530, BioLegend, Cat# 400630, RRID: AB_893676), SB645-conjugated anti-human CD31 ab (clone WM-59, Thermo Fisher Scientific, Cat# 64–0319−42, RRID: AB_2717093) or SB645-conjugated mouse IgG1 isotype ab (clone P3.6.2.8.1, Thermo Fisher Scientific, Cat# 64–4714−82, RRID: AB_470111), PE-conjugated anti-human CD140a ab (clone 16A1, BioLegend, Cat# 323506, RRID: AB_2268113) or PE-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400113, RRID: AB_326435), BV711-conjugated anti-human CD56 ab (clone 5.1H11, BioLegend, Cat# 362542, RRID: AB_2565920) or BV711-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400168), PE/Dazzle594-conjugated anti-human CD45 ab (clone HI30, BioLegend, Cat# 304052, RRID: AB_2563567) or PE/Dazzle594-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400176, RRID: AB_2923261), BV785- conjugated anti-human ITGAX ab (clone 3.9, BioLegend, Cat# 301644, RRID: AB_2565778) or BV785-conjugated mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400170, RRID: AB_11219601), FITC-conjugated anti-mouse ITGB2 ab (clone M18/2, BioLegend, Cat# 101405, RRID: AB_312814) or FITC-conjugated rat IgG2a isotype ab (clone RTK2758, BioLegend, Cat# 400506, RRID: AB_2736919), AF647-conjugated anti-mouse ITGB2 ab (clone M18/2, BioLegend, Cat# 101414, RRID: AB_2265032) or AF647-conjugated rat IgG2a isotype ab (clone RTK2758, BioLegend, Cat# 400526, RRID: AB_2864284), PE/Cy7-conjugated anti-mouse ITGAL ab (clone M17/4, BioLegend, Cat# 101122, RRID: AB_2562781) or PE/Cy7-conjugated rat IgG2a isotype ab (clone RTK2758, BioLegend, Cat# 400522, RRID: AB_326542), PE-conjugated anti-mouse ICAM1 ab (clone YN1/1.7.4, BioLegend, Cat# 116108, RRID: AB_313699) or PE-conjugated rat IgG2b isotype ab (clone RTK4530, BioLegend, Cat# 400608, RRID: AB_326552), anti-human/mouse CD44 ab (clone 5F12, Thermo Fisher Scientific, Cat# MA5-12394, RRID: AB_10984604) or mouse IgG1 isotype ab (clone PPV-06, Thermo Fisher Scientific, Cat# MA1-10405, RRID: AB_2536773) and goat anti-mouse IgG Fc ab (Thermo Fisher Scientific, Cat# 31170, RRID: AB_228290), anti-human/mouse CD44 ab (clone IM7, Thermo Fisher Scientific, Cat# 14–0441−82, RRID: AB_467246) or rat IgG2b isotype control ab (clone eB149/10H5, Thermo Fisher Scientific, Cat# 14–4031−82, RRID: AB_470099) and goat anti-Rat IgG Fc ab (Thermo Fisher Scientific, Cat# 31226, RRID: AB_228348).

The following abs were used for immunoblotting: unconjugated rabbit anti-human ITGB2 ab (clone D4N5Z, Cell Signaling Technology, Danvers, MA, Cat# 73663, RRID: AB_2799842), unconjugated rabbit anti-mouse ITGB2 ab (clone E9O7W, Cell Signaling Technology, Cat# 72607, RRID: AB_3083057), unconjugated rabbit anti-mouse non-phospho (non-p, active) β-catenin (Ser33/37/Thr41) ab (clone D13A1, Cell Signaling Technology, Cat# 8814, RRID: AB_11127203), unconjugated rabbit anti-mouse LEF-1 ab (clone C12A5, Cell Signaling Technology, Cat# 2230, RRID: AB_823558), unconjugated rabbit anti-mouse p-VANGL2 (Thr78, Ser79, Ser82) ab (clone 7H7C2, Thermo Fisher Scientific, Cat# MA5-38241, RRID: AB_2898157), unconjugated rabbit anti-mouse VANGL2 ab (Thermo Fisher Scientific, Cat# 21492-1-AP, RRID: AB_11182263), horseradish peroxidase (HRP)-linked anti-rabbit IgG (Cell Signaling Technology, Cat# 7074, RRID: AB_2099233), HRP-linked anti-β-actin ab (clone D6A8, Cell Signaling Technology, Cat# 12620, RRID: AB_2797972).

The following abs and reagents were used for cell adhesion, proliferation, or in vivo tumorigenicity studies: recombinant human ICAM-1 (Abcam, Waltham, MA, Cat# ab82125), recombinant mouse ICAM-1 (Abcam, Cat# ab277758), Ultra-LEAF anti-human ITGB2 ab (clone TS1/18, BioLegend, Cat# 302116, RRID: AB_2561483) or Ultra-LEAF mouse IgG1 isotype ab (clone MOPC-21, BioLegend, Cat# 400166, RRID: AB_2927801), Ultra-LEAF anti-mouse ITGB2 ab (clone M18/2, BioLegend, Cat# 101421, RRID: AB_312810) or Ultra-LEAF rat IgG2a isotype ab (clone RTK2758, BioLegend, Cat# 400573, RRID: AB_11147167), anti-mouse ITGB2 ab (clone GAME-46, BD Biosciences, Franklin Lakes, NJ, Cat# BDB557440, RRID: AB_396703) or mouse IgG1 κ (clone R3-34, BD Biosciences, Cat# BDB553922, RRID: AB_479672), pyrvinium pamoate (MedChemExpress, Monmouth Junction, NJ, Cat# HY-A0293), LGK974 (Selleck Chemicals, Houston, TX, Cat# S7143), zamaporvint (Selleck Chemicals, Cat# E1486), and vemurafenib (Selleck Chemicals, Cat# S1267).

The following abs and reagents were used for immunohistochemistry and immunofluorescence: unconjugated mouse anti-human ITGB2 ab (clone MEM-48, Novus Biologicals, Cat# NB500-379, RRID: AB_10000712), Dako REAL Detection System, Alkaline Phosphatase/RED (Agilent Dako, Santa Clara, CA, Cat# K5005), Biotin-conjugated goat anti-mouse IgG (Thermo Fisher Scientific, Cat# 31800, RRID: AB_228305), AF546-conjugated goat anti-mouse IgG1 (Thermo Fisher Scientific, Cat# A-21123, RRID: AB_2535765), and AF488-conjugated goat anti-mouse IgG1 (Thermo Fisher Scientific, Cat# A-21121, RRID: AB_2535764), unconjugated rabbit anti-human SOX10 (clone EPR4007, Abcam, Cat# ab155279, RRID: AB_2650603), unconjugated rabbit anti-human CD3 (clone SP162, Abcam, Cat# ab135372, RRID: AB_2884903), unconjugated rabbit anti-human CD31 (clone EPR3094, Abcam, Cat# ab76533, RRID: AB_1523298), unconjugated rabbit anti-human ICAM-1 (MilliporeSigma, Cat# SAB5700809, RRID: AB_3669069) and Cy3-conjugated goat anti-rabbit IgG (Thermo Fisher Scientific, Cat# A10520, RRID: AB_10563288) or AF488-conjugated goat anti-rabbit IgG (Thermo Fisher Scientific, Cat# A-11008, RRID: AB_143165), unconjugated mouse anti-human SOX10 (clone 1D2C8, Proteintech, Rosemont, IL, Cat#66786-1-Ig, RRID: AB_2882131) and AF647-conjugated goat anti-mouse IgG2a (Thermo Fisher Scientific, Cat# A-21241, RRID: AB_2535810), unconjugated mouse anti-human PU.1 (clone G148-74, BD Biosciences, Cat# 554268, RRID: AB_395335) and AF488-conjugated goat anti-mouse IgG2a (Thermo Fisher Scientific, Cat# A-21131, RRID: AB_2535771).

Immunofluorescence and immunohistochemistry

Multicolor immunofluorescence labeling for ITGB2, SOX-10, CD3, CD31, PU.1, and/or ICAM-1 with DAPI was performed on formalin-fixed, paraffin-embedded (FFPE) tumor biospecimens obtained from melanoma patients, as described [11–13]. Briefly, FFPE sections were baked for 30 min at 56 °C, followed by xylene deparaffinization for 10 min, and rehydrated by successive 2 min bath incubations of 100%, 100%, 95%, and 75% (v/v) ethanol. Sections were submerged in Target Antigen Retrieval solution, pH 6.0 (Agilent Dako), and then pressure cooked in a decloaking chamber (Biocare Medical, Pacheco, CA) at 110 °C for 45 min for antigen retrieval. Sections were subsequently blocked with 10% (v/v) goat, avidin-, and biotin-blocking sera (Vector Laboratories, Newark, CA), and finally incubated with mouse anti-human ITGB2 ab (1:100 or 1:200, Novus Biologicals) at 4 °C overnight. Sections were washed thrice with tris-buffered saline (TBS)-Tween 20 (T, 0.1%) for 5 min, incubated for 1 h at RT with biotin-conjugated goat anti-mouse IgG (1:200, Thermo Fisher Scientific), washed thrice, followed by incubation at RT for 1 h with strepatividin-Cy5 (1:50, Thermo Fisher Scientific) or AF546-conjugated goat anti-mouse IgG1 (1:500, Thermo Fisher Scientific). After blocking with 10% goat serum, sections were incubated with rabbit anti-human SOX-10 ab (1:500, Abcam), rabbit anti-human CD3 ab (1:100, Abcam), rabbit anti-human CD31 ab (1:100, Abcam), mouse anti-human PU.1 ab (1:100, BD Biosciences), or rabbit anti-human ICAM-1 ab (1:200, MilliporeSigma) at 4 °C overnight. After blocking with 0.1% (v/v) Sudan Black B solution (MilliporeSigma) and rinsing with TBS-T, sections were incubated for 1 h at RT with Cy3-conjugated goat anti-rabbit IgG ab (1:2000, Thermo Fisher Scientific), AF647-conjugated goat anti-mouse IgG2a (1:2000, Thermo Fisher Scientific), AF488-conjugated goat anti-mouse IgG2a (1:1000, Thermo Fisher Scientific), or AF488-conjugated goat anti-rabbit IgG (1:1000, Thermo Fisher Scientific) and mounted with ProLong Glass Antifade Mountant with NucBlue Stain (Thermo Fisher Scientific) or DAPI (Thermo Fisher Scientific). Immunofluorescence was analyzed on an EVOS FL Auto 2 imaging system (Thermo Fisher Scientific). Immunohistochemistry of ITGB2 was performed as above, using the Dako REAL Detection System, Alkaline Phosphatase/RED (Agilent Dako) according to the manufacturer’s recommendations after overnight incubation with the anti-human ITGB2 ab (1:100 dilution, Novus Biologicals).

Immunofluorescence staining of a tissue microarray (TMA) was used to examine association of ITGB2 expression with melanoma progression. ITGB2 staining of patient primary melanomas (n = 107, University of Bonn) was used to assess association with clinically confirmed sentinel lymph node metastases (J.L.). The TMA [14] was constructed from 2 mm replicate cores of patient benign common nevi (n = 7), primary melanomas (n = 24), and metastatic melanomas (n = 13). Patients were diagnosed in either 2013 (nevi samples) or between 1997 and 2007 (melanoma samples) and tissues subsequently stored as FFPE blocks at BWH. All tissue sections were stained with hematoxylin and eosin and clinical diagnoses independently reviewed and confirmed by two dermatopathologists (G.F.M. and C.G.L.) prior to analyses. Multicolor immunofluorescence was performed as previously described [11–14] using mouse anti-human ITGB2 ab (1:200, Novus Biologicals), mouse anti-human SOX10 ab (1:500, Proteintech), and DAPI. Sections were then stained with a secondary ab mixture of AF488-conjugated goat anti-mouse IgG1 (1:1000, Thermo Fisher Scientific) and AF647-conjugated goat anti-mouse IgG2a (1:2000, Thermo Fisher Scientific). Appropriate isotype-matched ab and tissue controls were employed. Duplicate images encompassing each intact TMA core or patient primary melanoma specimen were acquired with a 20x objective. Mean frequency of ITGB2⁺SOX10⁺ versus total melanoma cells positive for nuclear SOX10 was determined for replicate core sets by manual counting in ImageJ (NIH, Bethesda, MD) by investigators blinded to sample allocation.

RNA extraction and RT-qPCR

Total RNA was extracted from cell lines and tumor grafts using the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany, Cat# 74134) or mouse lung tissue using the RNeasy Fibrous Tissue Mini Kit (Qiagen, Cat# 74704) according to the manufacturer’s instructions, as described [11, 12, 15]. For RT-qPCR analyses, RNA was subsequently converted to cDNA using the SuperScript VILO cDNA synthesis kit (Thermo Fisher Scientific, Cat# 11754050). Samples from in vitro lines or murine tissue biospecimens were assayed in triplicate or duplicate at minimum, respectively, using the Fast SYBR Green Master Mix (Thermo Fisher Scientific) or TaqMan (Thermo Fisher Scientific) primer sets, as shown in Supplementary Tables S1-3, on a QuantStudio 3 or QuantStudio 5 Real-Time PCR system (Applied Biosystems, Waltham, MA). Thermal cycling was carried out at 95 °C for 20 s, followed by 40 cycles at 95 °C for 1 s, and then at respective annealing temperatures as shown in Supplementary Tables S1-3, followed by melt-curve analysis, as appropriate. In the case of human 18s rRNA amplification, thermal cycling was carried out at 94 °C for 2 min, followed by 40 cycles at 94 °C for 15 s, 60 °C for 20 s and 68 °C for 1 min, followed by melt-curve validation, as described [12]. Thermal cycling of TaqMan RT-qPCR was performed according to the manufacturer’s standard protocol. Data was normalized to human 18s rRNA, human or murine β-actin (ACTB/Actb) housekeeping genes. Relative transcript levels were calculated utilizing the delta-delta Ct method [11, 12, 15, 16]. Samples with threshold cycle (Ct) numbers greater than 35 or the water negative control, with more than one melt curve, or with TaqMan housekeeping threshold cycles exceeding 25 were considered as not detected (nd).

Genomic DNA extraction and qPCR

Genomic DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen, Cat# 69504), according to the manufacturer’s protocol, as described. Genomic DNA (0.5 µg) and TaqMan qPCR primers specific for EGFP (Supplementary Table S3, Thermo Fisher Scientific) were combined, and qPCR was performed on a QuantStudio 5 Real-Time PCR system (Applied Biosystems) to quantify melanoma lung metastases normalized to murine Actb level. PCR conditions were as recommended by the manufacturer’s protocol.

Flow cytometry

ITGB2 heterodimer and ICAM-1 surface protein expression by human and murine melanoma lines, human HSB-2 and murine EL-4 T lymphoblastic lines, HUVEC and murine C166 endothelial cells, and cell suspensions derived from melanoma patient specimens were analyzed by multi-color FC, as described [11, 12, 15, 16]. Nonviable cells were excluded by staining with the Zombie NIR Fixable Viability Kit (1:1000, BioLegend, Cat# 423106) for 10 min at RT in the dark, as per the manufacturer’s protocol. Nonspecific ab binding was blocked by Human TruStain FcX (Fc Receptor Blocking Solution, BioLegend, Cat# 422302) for 10 min at RT or TruStain FcX PLUS (anti-mouse CD16/32) Antibody (BioLegend, Cat# 156604) for 10 min at 4 °C, as described [13, 15]. Cells were then stained with fluorochrome-conjugated abs (5–10 µg/mL) in FC buffer, phosphate-buffered saline (PBS + 2% (v/v) FBS), for 30 min at 4 °C, followed by washing, as described [11–13, 15, 16]. All flow cytometry assays involved isotype-matched control abs as well as exclusion of dead cells and cell doublets. Fluorescence emissions were recorded on a FacsCanto (BD Biosciences) or an Aurora Spectral Analyzer (Cytek, Fremont, CA), and data was analyzed using FlowJo version 10.8.1 (TreeStar, Ashland, OR).

CD44 crosslinking

Regulation of ITGB2 surface protein expression and activation by CD44 crosslinking was analyzed by flow cytometry as above. Briefly, human and murine melanoma cells were first incubated with anti-CD44 ab (10 µg/ml, clones 5F12 or IM7, Thermo Fisher Scientific) or respective isotype control abs in FC buffer for 1 h at 37 °C. Cells were then washed, and CD44 was crosslinked via incubation with respective goat anti-mouse or goat anti-rat IgG Fc abs (10 µg/ml, Thermo Fisher Scientific) in FC buffer for 45 min at 37 °C. Cells were then washed two times in FC buffer and then subjected to surface protein staining of ITGB2 and its α pairing subunits, as above.

Immunoblotting

Melanoma cells (wildtype, Itgb2 KO, or Cas9 control) were lysed in ice-cold RIPA buffer (Thermo Fisher Scientific) supplemented with Protease/Phosphatase Inhibitor Cocktail (Cell Signaling Technology, Cat# 5872) and vortexed for 30 min at 4 °C, as described [11, 12]. Protein concentrations were measured using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific), according to the manufacturer’s recommendations. Lysates were resolved by SDS-PAGE gel electrophoresis and transferred to Immobilon-P PVDF membranes (MilliporeSigma). Membranes were blocked in TBS-T containing 5% (w/v) non-fat dry milk (Thermo Fisher Scientific) for at least 1 h at RT. Membranes were incubated overnight at 4 °C with an anti-mouse ITGB2 ab (clone E9O7W, 1:1000, Cell Signaling Technology) or an anti-human ITGB2 ab (clone D4N5Z, 1:1000, Cell Signaling Technology). Blots were then washed thrice in TBS-T and incubated with HRP-conjugated secondary ab (1:2000, Cell Signaling Technology) for 1 h at RT. Antigens were visualized by the Lumi-Light Western Blotting Substrate (MilliporeSigma) on HyBlot CL Autoradiography Films (Thomas Scientific, Swedesboro, NJ) via a Kodak Min-R mammography processor (Kodak, Rochester, NY), as described [11–13, 16]. In some cases, membranes were stripped with ReBlot Plus Strong Antibody Stripping Solution (MilliporeSigma) according to the manufacturer’s instructions. Membranes were subsequently blocked and incubated overnight at 4 °C with an HRP-conjugated anti-β-actin ab (clone D6A8, 1:1000, Cell Signaling Technology) in TBS-T + 5% (w/v) non-fat dry milk and subsequently developed as above.

For experiments involving Wnt inhibitors, YUMM5.2 cells were plated in 60 mm cell culture dishes containing RPMI-1640 medium with 10% (v/v) heat-inactivated FBS and 1% (v/v) penicillin/streptomycin, serum-starved overnight at 37 °C, 5% CO₂ in RPMI-1640 medium devoid of FBS and containing 1% (v/v) penicillin/streptomycin. Cells were subsequently treated with pyrvinium pamoate (6.5 µM), LGK974 (20 µM), zamaporvint (300 nM), or vehicle control for 20 min–12 h. For analysis of Wnt effector expression in YUMM5.2 wildtype cells treated with pyrvinium pamoate, LGK974 or zamaporvint versus respective vehicle controls as well as in YUMM5.2 Itgb2 KO versus Cas9 control cells, lysates were prepared and resolved by SDS-PAGE gel electrophoresis as above, and transferred using a Trans-Blot Turbo Transfer System (Bio-Rad Laboratories, Hercules, CA) according to the manufacturer’s recommendations. Membranes were blocked in TBS-T containing 5% (w/v) non-fat dry milk (Thermo Fisher Scientific) for at least 1 h at RT. Membranes were incubated overnight at 4 °C with unconjugated rabbit anti-mouse non-p (active) β-catenin (Ser33/37/Thr41) ab (clone D13A1, 1:1000, Cell Signaling Technology), unconjugated rabbit anti-mouse LEF-1 ab (clone C12A5, 1:1000, Cell Signaling Technology), or unconjugated rabbit anti-mouse p-VANGL2 (Thr78, Ser79, Ser82) ab (clone 7H7C2, 1:500, Thermo Fisher Scientific). Blots were then washed thrice in TBS-T and incubated with HRP-conjugated secondary ab (1:1000, Cell Signaling Technology) for 1 h at RT. Antigens were visualized by the SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific) using a Fusion FX Imaging System (Vilber Lourmat, Collégien, France). In some cases, membranes were stripped with Restore™ PLUS Western Blot Stripping Buffer (Thermo Fisher Scientific) according to the manufacturer’s instructions or run separately when stripping interfered with detection. Membranes were subsequently blocked and incubated overnight at 4 °C with HRP-conjugated anti-β-actin ab (clone D6A8, 1:1000, Cell Signaling Technology) or unconjugated rabbit anti-mouse VANGL2 ab (1:500, Thermo Fisher Scientific) followed by incubation with HRP-conjugated secondary ab (1:2000, Cell Signaling Technology) for 1 h at RT, as above, in TBS-T + 5% (w/v) non-fat dry milk. Blots were subsequently developed as described above.

Generation of stable nuclear GFP-expressing B16-F10 and YUMM5.2 melanoma variants

To generate B16-F10 and YUMM5.2 melanoma variants stably expressing nuclear green fluorescent protein (GFP), we fused a nuclear localization signal with GFP in the MSCV-N-Flag-HA-GFP vector (Addgene, Watertown, MA, plasmid #41034, RRID: Addgene_41034). This vector was packaged into retroviral particles via HEK-293EBNA cells (ATCC, Cat# CRL-10852, RRID: CVCL_6974) co-transfected using Lipofectamine 2000 (Thermo Fisher Scientific) with the viral packaging plasmids [11], GagPolDS and pN8e-VSV-G, according to the manufacturer’s protocol. Viral supernatants were harvested 48–72 h after transfection and filtered as described [12]. Cells were stably transduced and selected with puromycin, followed by multiple rounds of cell sorting to >95% purity for GFP expression on either FACS Aria II (BD Biosciences) or MoFlo Astrios EQ (Beckman Coulter, Brea, CA) cell sorting machines at the Beth Israel Deaconess Medical Center FC Core (Boston, MA).

Generation of stable Itgb2 knockout B16-F10 and YUMM5.2 melanoma variants

To create B16-F10 and YUMM5.2 Itgb2 KO variants, we used the CRISPR/Cas9 gene editing platform, as described [15]. Briefly, we first generated stable Cas9-expressing B16-F10 and YUMM5.2 monogenetic melanoma variants by lentiviral transduction. Specifically, Lenti-X 293 T packaging cells (Takara Bio, San Jose, CA, Cat# 632180, RRID: CVCL_4401) were co-transfected with the lentiCas9-enhanced GFP (EGFP) expression vector (Addgene, plasmid #63592, RRID: Addgene_63592) along with psPAX2 (Addgene, plasmid #12260, RRID: Addgene_12260) and pMD2.G (Addgene, plasmid #12259, RRID: Addgene_12259) packaging vectors using Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s protocol. Viral supernatants were harvested 48–72 hrs post transfection, centrifuged at 20,000 rpm for 6 hrs, and filtered (0.45 µm). Lentiviral particles were added to B16-F10 and YUMM5.2 cells in the presence of 10 µg/ml polybrene (MilliporeSigma). EGFP-positive cells were single cell-sorted into flat-bottom 96-well plates (Thermo Fisher Scientific) using a MoFlo Astrios EQ (Beckman Coulter) cell sorter at the Beth Israel Deaconess Medical Center FC Core (Boston, MA). Multiple single-cell clones were expanded, and EGFP-positivity of >95% was confirmed using an Aurora Spectral Analyzer (Cytek). To validate that Cas9-expressing B16-F10 and YUMM5.2 variants exhibited in vivo tumor growth kinetics representative of respective wildtype cell lines, multiple Cas9-EGFP-expressing single-cell clones and wildtype control cells were injected into the flanks of C57BL/6 wildtype mice. A Cas9-expressing variant exhibiting wildtype tumor growth kinetics was then transduced with a plasmid encoding two guide (g)RNAs, TGCCCCCGTTGGTCGAACTC and GTGACTTTCCGGCGGGCCAA, targeting murine Itgb2 and transiently expressing a blue fluorescent protein (BFP) selection marker (VectorBuilder, Chicago, IL). Cells positive for BFP and EGFP were single-cell sorted into flat-bottom 96-well plates (Thermo Fisher Scientific), as above, 48 hrs after transduction. Colonies were selected, expanded, and screened for loss of Itgb2 gene and ITGB2 protein expression by real-time RT-qPCR, flow cytometry, and immunoblotting, as described [15]. To validate chromosomal disruption of the Itgb2 gene, genomic DNA was extracted from candidate clones (DNeasy Blood & Tissue Kit, Qiagen), and the Itgb2 locus was PCR-amplified (Platinum Taq DNA Polymerase, Thermo Fisher Scientific) with primers comprising the Itgb2 gRNA binding sites [15]. Thermal cycling was carried out in the presence of 1.5 mM MgCl₂ at 94 °C for 2 min, followed by 35 cycles at 94 °C for 30 sec, 62 °C for 30 sec, and 68 °C for 20sec, and a final extension for 10 min at 68°C. The primer sequences used were: forward, 5’-CCTGACCTCAGATCCCTCCT-3’ and reverse, 5’-GGTAATTGCCATGCGGGTTC-3’. The 328 bp PCR products reflecting individual alleles were gel-purified, ligated into the pCR2.1 vector (TA Cloning Kit, Thermo Fisher Scientific), and transformed into One Shot TOP10 Chemically Competent E. coli (Thermo Fisher Scientific), as described [15]. Plasmids were purified using the ZymoPURE Plasmid Miniprep Kit (Zymo Research, Irvine, CA), sequenced (GENEWIZ/Azenta Life Sciences, Burlington, MA) [15], and homozygous KO of Itgb2 confirmed.

ICAM-1 adhesion assays

To assess ICAM-1-binding of melanoma cells, we performed cell adhesion assays using 96-well polystyrene plates with white sides and clear bottoms (Thermo Fisher Scientific), as described [17], with some protocol modifications. Specifically, wells were coated in quadruplicate with 25 µg/ml recombinant human ICAM-1 (Abcam, Cat# ab82125) or 50 µg/ml mouse ICAM-1 (Abcam, Cat# ab277758) in PBS (Life Technologies), centrifuged at 2,000 rpm for 5 min, and incubated 2 h at 37 °C. Subsequently, wells were blocked with 1% (w/v) BSA or 1% (v/v) FBS in PBS for 30 min at 37 °C. Wildtype, Itgb2 KO, or Cas9 control cells were suspended in adhesion buffer (RPMI 1640, 0.2% BSA). For ITGB2 ab inhibition studies, wildtype melanoma cells were preincubated for 30 min on ice at 4 °C with 20 µg/ml blocking anti-human ITGB2 ab (clone TS1/18, BioLegend) or blocking anti-mouse ITGB2 ab (clone M18/2, BioLegend or GAME-46, BD Biosciences) versus respective isotype-matched control ab. In some cases, cells were preincubated in parallel with 50 mM EDTA (Thermo Fisher Scientific) in PBS (without Ca²⁺/Mg²⁺). For Wnt inhibition assays, wildtype murine cells were first preincubated in adhesion buffer containing ITGB2 or isotype control abs as above followed by addition of pyrvinium pamoate (6.5 µM), LGK974 (20 µM), zamaporvint (10 µM), or vehicle control for 30 min on ice at 4°C. In the case of Itgb2 KO and Cas9 control cells, Wnt inhibitors or vehicle control were preincubated with cells as above. Melanoma (4–5 × 10⁴/well) or T lymphoblastic (1.5 × 10⁵/well) cell suspensions (100 µl/well) were added to 96-well plates. Cells were allowed to adhere for 45 min at 37 °C. Wells were washed three times with PBS (containing Ca²⁺/Mg²⁺, Thermo Fisher Scientific) using a squirt bottle. Cell adhesion was quantified using the Promega CellTiter-Glo2.0 Assay (Promega, Madison, WI, Cat# G9242), according to the manufacturer’s protocol on a Promega GloMAX 96 Microplate Luminometer (Promega), as described [16]. Recorded values represent cell adhesion relative to total input cells.

Melanoma 3D proliferation assays

Melanoma tumor sphere growth of Itgb2 KO versus Cas9 control cell variants was examined by the Promega CellTiter-Glo2.0 Cell Viability Assay (Promega), according to the manufacturer’s instructions, using a Promega GloMAX 96 Microplate Luminometer (Promega), as above. To that end, Anti-Adherence Rinsing Solution (Stemcell Technologies, Cambridge, MA) was added (60 µl/well) to 96-well microplates, as above, centrifuged immediately at 2,000 rpm for 5 min, wells aspirated, and washed with RPMI-1640. Cells were suspended (8 × 10³ per 100 µl) in growth media containing RPMI 1640, 1% (w/v) methylcellulose, 1% (v/v) FBS, and 1% (v/v) penicillin/streptomycin, filtered through Cell Strainer Snap Caps (0.35 μm, Corning, Corning, NY), and seeded (100 µl) in wells, as described [11, 12]. In some instances, cells in culture were treated with vemurafenib (20 µM) or vehicle control. Cells were cultured 48 h in an incubator at 37 °C, 5% CO₂.

Mouse strains and in vivo tumorigenicity studies

Wildtype C57BL/6 (RRID: IMSR_JAX:000664), Icam1^tm1Jcgr KO C57BL/6 mice [18] (RRID: IMSR_JAX:002867), and immunodeficient nonobese diabetic/severe combined immunodeficiency (NOD/SCID) IL-2Rγ^−/− KO (NSG, RRID: IMSR_ JAX:005557) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed at the BWH animal facility as described [11, 12, 16]. Both female and male mice, at least 6 weeks of age and matched by gender and age between experimental groups, were used for this study and treated in accordance with the National Institutes of Animal Healthcare Guidelines under the BWH Institutional Animal Care and Use Committee (IACUC)-approved experimental protocol 2016N000112. Human C8161 and MDA-MB-435S or murine B16-F10 and YUMM5.2 melanoma lines were injected subcutaneously into the flanks of recipient mice at 1 × 10⁶ cells/inoculum for C8161 or MDA-MB-435S cells in NSG mice, 1 × 10⁵ cells/inoculum for B16-F10 in wildtype C57BL/6, Icam1 KO C57BL/6, and NSG mice, and 1 × 10⁶ cells/inoculum for YUMM5.2 in wildtype C57BL/6, Icam1 KO C57BL/6, and NSG mice as described [11, 12, 16]. For ITGB2 ab targeting studies, mice were injected intraperitoneally with 200 µg blocking Ultra-LEAF anti-human ITGB2 ab (clone TS1/18, BioLegend) or blocking Ultra-LEAF anti-mouse ITGB2 ab (clone M18/2, BioLegend) versus respective isotype-matched control ab every three days starting on the day of tumor cell inoculation for the duration of the experiment. For Wnt pathway inhibition studies, mice were injected intraperitoneally with pyrvinium pamoate (0.3 mg/kg [19]), or vehicle control every day for the duration of the experiments. Alternatively, mice were fed submaximal doses of either LGK974 or zamaporvint (3 mg/kg body weight/day [20, 21]), or vehicle control incorporated into rodent chow (Research Diets, Inc., New Brunswick, NJ). Tumor volumes were measured at least once per week by determination of tumor volume (TV) according to the established formula [TV (mm³) = π/6 × 0.5 × length × (width)²] until either the experimental endpoint or disease state or excessive tumor burden required protocol-stipulated euthanasia as described [12, 16]. Maximal tumor size/burden permitted by the BWH IACUC is 2 centimeters (cm) at the widest point. These limits were not exceeded in any experimental instance herein. Tumor ulceration or moribund condition also require euthanasia. These additional IACUC criteria were also met for all tumorigenicity experiments [12, 16].

Total mRNA sequencing

For total mRNA sequencing (RNA-seq) of melanoma tissues, freshly harvested tumor samples were snap-frozen, temporarily stored at −80 °C, and sent to GENEWIZ/Azenta for further processing. Briefly, RNA extraction, RNA library preparations, and sequencing reactions were performed at GENEWIZ/Azenta. Total RNA was extracted from frozen tissue samples using the RNeasy Plus Mini Kit (Qiagen, Cat# 74134) following the manufacturer’s instructions. RNA samples were quantified using Qubit 2.0 Fluorometer (Thermo Fisher Scientific), and RNA integrity was measured using the RNA Screen Tape on Agilent 2200 TapeStation (Agilent Technologies, Santa Clara, CA). The RNA sequencing libraries were then prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA) according to the manufacturer’s recommendations. To that end, mRNAs were initially enriched with Oligod(T) beads, fragmented for 15 min at 94 °C, and second strand cDNAs were subsequently synthesized. Next, cDNA fragments were end-repaired and adenylated at 3’ends, and universal adapters were ligated to cDNA fragments, followed by index addition and library enrichment by PCR with limited cycles. The sequencing library was validated on the Agilent TapeStation (Agilent Technologies) and quantified using the Qubit 2.0 Fluorometer (Thermo Fisher Scientific) and quantitative PCR (KAPA Biosystems, Wilmington, MA). The sequencing libraries were multiplexed and clustered on 3 lanes of a flowcell. After clustering, the flowcell was loaded on the Illumina HiSeq 4000 (Illumina, San Diego, CA) according to manufacturer’s instructions. The samples were sequenced using a 2 × 150 Pair-End configuration.

Processing and analysis of total mRNA sequencing data

Reads were processed to counts through the RNA-seq pipeline implemented in bcbio-nextgen v1.2.9-8eb78b7 (https://bcbio-nextgen.readthedocs.org). Quality control of raw reads was performed using FastQC v0.11.9 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), followed by library generation, sequencing, and further analysis. Reads were aligned to the Mus musculus genome version mm10 ucsc-201,112 using STAR v2.6.1d [22]. Alignments were examined for uniformity of coverage, rRNA content, genomic context of alignments (e.g. alignments in known transcripts and introns), complexity and other quality using a combination of FastQC v0.11.9, Qualimap v2.2.2d [23], MultiQC v1.11 (https://github.com/ewels/MultiQC), and custom tools. Transcripts Per Million (TPM) measurements per isoform were generated by quasialignment using Salmon v1.6.0, given that quantitating at the isoform level has been shown to produce more accurate results at the gene level. Differential gene expression was determined by employing counts per gene estimated by tximport using the Salmon TPM measurements with DESeq2 v1.38.3 [24]. QC and differential expression analyses were performed using the R statistical software (v.4.2.2, https://www.R-project.org/). Mouse strains (i.e. wildtype C57BL/6, Icam1 KO C57BL/6, and NSG) were analyzed separately for differential expression between Cas9 control vs. Itgb2 KO tumors. To reduce background noise, log2-fold changes were shrunken using the apeglm method [25]. We identified differentially expressed transcripts as those with a shrunken log2-fold change >0.5 and an adjusted P-value after correction for false discovery rates of < 5%. Genes which were differentially regulated in Cas9 control vs. Itgb2 KO tumors in NSG and wildtype but not Icam1 KO C57BL/6 mice, were identified and subjected to protein-protein interaction analysis (STRING, https://string-db.org). An unbiased, full network confidence analysis with an interaction score of 0.371 was performed. Heatmaps and volcano plots were generated based on all nodes, which were connected to the main network using pheatmap (https://CRAN.R-project.org/package=pheatmap) and EnhancedVolcano R software packages (https://github.com/kevinblighe/EnhancedVolcano). The STRING interaction network represents the main clusters based on a confidence analysis with an adjusted interaction score of 0.54.

Analysis of public single cell RNA data

We downloaded the Seurat object from Pozniak et al. [26] and used the Seurat dotplot function to visualize both the percent of cells expressing ITGB2 and ICAM1 and the expression level. We also examined expression of these genes in cells that were ABCB5⁺ or ABCB5⁻ as well as NGFR⁺ or NGFR⁻. Because ABCB5⁺ and NGFR⁺ cells tended to be more deeply sequenced, we normalized the data before plotting. We used the number of genes expressed by a cell as a proxy for sequencing depth and established quartiles from ABCB5⁻ and NGFR⁻ cells. We then randomly subseted our ABCB5⁺ and NGFR⁺ populations to include equal numbers of cells in each quartile and plotted expression of ITGB2 and ICAM1 using the Seurat dotplot function. We downloaded the data from Goyal et al. [27] and created values for both the canonical and non-canonical Wnt pathways by summing the expression of all genes in the canonical Wnt signaling pathway Gene Omnibus (GO) term and in the non-canonical Wnt signaling pathway GO term. We graphed expression of these pathways along with expression of ITGB2 and ICAM1 using the Seurat dotplot function. Additional analysis of ITGB2, ICAM1, and CD44 gene expression by patient melanoma, T cells, and endothelial cells was performed using the Tirosh et al. [28] dataset.

Statistical analysis

Statistical differences between groups were determined using the Student’s t test, the nonparametric Mann-Whitney, paired Wilcoxon test, or Fisher’s Exact Test (comparison of two experimental groups), one-way ANOVA with Dunnett’s post-test or Kruskal-Wallis with Dunn’s post-test (comparison of three or more experimental groups), or repeated-measures two-way ANOVA or mixed model, followed by Šídák’s multiple comparisons correction (time-dependent statistical differences, i.e. tumor growth). A value of P < 0.05 was considered statistically significant. All statistical analyses were performed using the PRISM 10.1.1 software (GraphPad, San Diego, CA).

Results

Melanoma cells express ITGB2 and its level and activation state are regulated by CD44

Analysis of a single-cell (sc) RNA-seq dataset [28] demonstrated human ITGB2 gene (ITGB2) expression by melanoma cells in 13 of 14 patient lesions examined (fig. S1A), at ranges overlapping ITGB2 levels in T cell positive controls and exceeding those in endothelial cells (ECs, Fig. 1A). Flow cytometric (FC) analysis confirmed surface ITGB2 protein expression by 22.5 ± 13.5% (mean ± SEM) of melanoma cells in five of five patients (Fig. 1B) versus 99.4 ± 0.1% and 13.7 ± 5.8% by patient tumor-infiltrating T cells and ECs, respectively (Fig. 1B, fig. S1B). Mean ITGB2⁺SOX10⁺ melanoma cell frequency was significantly higher in metastatic (10.4 ± 2.9%, n = 13 patients) versus primary melanomas (4.5 ± 1.6%, n = 24 patients) or benign nevi (0.7 ± 0.5%, n = 7 patients), as determined by immunofluorescence staining of a clinical melanocytic tissue microarray (TMA [14], Fig. 1C). Immunostaining of additional patient primary melanomas (n = 105) for ITGB2 revealed cancer cell-ITGB2 positivity of < 2% in 40 patients, 2–25% in 36 patients and >25% in 29 patients, of which the latter two groups showed greater incidence of sentinel lymph node (SLN) metastases compared to the cohort with low to absent cancer cell-ITGB2 frequency (Fig. 1D). Multiplex immunofluorescence was used to discriminate ITGB2 expression by nuclear SOX10⁺ melanoma cells from CD3⁺ T cells, CD31⁺ endothelium, or PU.1⁺ macrophages in clinical tumor biospecimens (Fig. 1E).

Fig. 1.

Melanoma cell-intrinsic ITGB2 expression and activation by CD44 (A) Single-cell (sc) RNA-seq analysis of human ITGB2 gene (ITGB2) expression by patient melanoma (MM) cells versus tumor-infiltrating T cells or endothelial cells (ECs), as depicted by violin plots (median, bold white line; top and bottom quartiles, thin white lines) overlayed with dots representing respective single cells (B) Percentages (mean,) of human ITGB2 surface protein expression by patient MM cells, T cells, and ECs (n = 5 patients) as determined by flow cytometry (C) Mean ITGB2⁺SOX10⁺ frequency (%) in benign nevi (n = 7 patients), primary melanomas (n = 24 patients), and metastatic melanomas (n = 13 patients) as determined by multicolor immunofluorescence staining of a patient melanocytic tissue microarray (TMA). Kruskal-Wallis multiple comparisons test was used to assess statistical significance (D) Incidence (%) of patient sentinel lymph node (SLN) metastases versus respective primary melanoma biospecimen cohorts (n = 105) of increasing cancer cell-ITGB2 positivity, 0–2% (n = 40), 2–25% (n = 36), >25% (n = 29), as determined by immunostaining. Frequencies of ITGB2-positive (black bars) and ITGB2-negative (white bars) melanoma cells within each cohort are shown. Fisher’s exact test was performed to determine statistical significance (E) Representative multiplex immunofluorescence staining of a patient primary melanoma biopsy for co-expression of ITGB2 (red, all panels) and the melanocytic marker, nuclear SOX-10 (green, first panel), pan T cell marker, CD3 (green, second panel), vascular endothelial marker, CD31 (green, third panel), or macrophage marker, PU.1 (green, fourth panel). Nuclei were counterstained with DAPI (blue). Size bars, 50 μm (F and G), Representative immunoblots of ITGB2 protein expression by (F) human melanoma lines, A2058, A375, C8161, FEMX, LOX-IMVI, MDA-MB-435S, and control HSB-2 T lymphoblastic leukemia cells and HUVEC endothelial cells, and (G) murine melanoma lines, B16-F10, YUMM1.7, YUMM3.3, YUMM4.1, YUMM5.2, and control EL-4 T cell lymphoma cells and C166 endothelial cells (H and I) Effect of CD44 ab-mediated crosslinking (black bars) versus isotype control ab treatment (white bars) on ITGB2 surface protein expression level (mean fluorescence intensity, MFI, ± SEM) by (H) human and (I) murine melanoma lines and respective cell controls (gray bars) as above, based on FC analysis (J and K) Effect of CD44 ab crosslinking as in (H and I) on the activation state of human melanoma cell-ITGB2 as determined by FC (MFI ± SEM) using the activation-sensitive ITGB2 antibody clones (J) KIM-127 and (K) MEM-148. Results are representative of at least n = 3 independent experiments. *, p < 0.05; **, p < 0.01; NS, not significant. See also figs. S1, S2, and S3.

Expression analysis by scRNA-seq of the four ITGB2 heterodimer pairing partners [2] in melanoma patient tissue showed substantial gene expression of ITGAL by tumor cells and lower to negligible levels of the ITGAD, ITGAM, and ITGAX subunits (fig.S1C). Consistently, FC analysis detected significant surface ITGAL basal protein expression (6.6 ± 3.6%), but only negligible levels of ITGAM (2.4 ± 1.4%) or ITGAX (1.1 ± 1.0%) on clinical melanoma cells (n = 5 patients, fig. S1D). Reverse transcription quantitative polymerase chain reaction (RT-qPCR) confirmed ITGB2 (fig. S2A), ITGAD, ITGAL, ITGAM, and ITGAX expression (fig. S2B) by established human melanoma lines, A2058, A375, C8161, FEMX, LOX-IMVI, and MDA-MB-435S, as well as HSB-2 T lymphoblastic leukemia positive control cells and human umbilical vein endothelial cells (HUVEC). RT-qPCR also detected murine Itgb2, Itgad, Itgal, Itgam, and Itgax gene transcripts in B16-F10, YUMM1.7, YUMM3.3, YUMM4.1, and YUMM5.2 melanoma lines, as well as EL-4 T cell lymphoma positive and C166 endothelial control cells (fig. S2, C and D). Immunoblot analyses further validated ITGB2 protein expression by the above human melanoma lines and HSB-2 cells, while HUVEC expressed only low levels (Fig. 1F). Similarly, murine melanoma lines and EL-4, but not C166 cells expressed ITGB2 (Fig. 1G).

Antibody (ab)-mediated crosslinking of the hyaluronic acid receptor, CD44, on melanoma cells markedly elevated baseline ITGB2 surface protein levels across all human (Fig. 1H, fig. S2E) and murine lines examined (Fig. 1I, fig. S2F), including in some cases > 70-fold. Similarly, CD44 crosslinking induced expression of the predominant cancer cell-ITGB2 pairing subunit, ITGAL, but not ITGAM or ITGAX, on both human (fig. S2G) and murine melanoma lines (fig. S2H). CD44 bridging also induced structural change of melanoma-ITGB2 to an open, more active state as recognized by the conformation-sensitive abs, KIM-127 (Fig. 1 J) and MEM-148 (Fig. 1 K). Notably, CD44 was highly expressed by patient melanoma cells at levels significantly exceeding those observed in tumor-infiltrating T cells and ECs (fig. S3A), supporting its role in melanoma cell-ITGAL/ITGB2 regulation. Together, these results establish expression of the ITGB2 heterodimer by melanoma cells and its induction and activation by CD44.

Antibody-mediated blockade of melanoma cell-ITGB2 inhibits ICAM-1-dependent adhesion, tumor growth, and metastasis

To dissect cancer cell-intrinsic ITGB2 functions in tumorigenesis, we first performed in vitro melanoma cell adhesion assays involving its major ligand [2], ICAM-1. Highly metastatic human C8161 and MDA-MB-435S melanoma and HSB-2 positive control cells firmly adhered to immobilized human ICAM-1 at levels significantly exceeding negative coating controls (Fig. 2A). ICAM-1-dependent adhesion was significantly inhibited by treatment with ITGB2 blocking ab or ethylenediaminetetraacetic acid (EDTA, Fig. 2A), an established antagonist of integrin activity [29]. Consistently, B16-F10 and YUMM5.2 melanoma and EL-4 cells firmly bound murine ICAM-1, which was significantly reversed by ITGB2 ab blockade or EDTA treatment (Fig. 2B). Administration of a human-specific ITGB2 neutralizing ab significantly attenuated human C8161 and MDA-MB-435S tumor xenograft growth in T cell null NSG mice (Fig. 2C). Treatment of NSG mice grafted with murine B16-F10 or YUMM5.2 cells with anti-mouse ITGB2 blocking ab also markedly suppressed tumor growth (Fig. 2D). Consistent with the pronounced tumor-inhibitory effect of ITGB2 antagonism in NSG mice, both melanoma-intrinsic ITGB2/Itgb2 and ICAM1/Icam1 gene levels were between 4-fold and >1000-fold elevated in tumor tissue in vivo versus respective in vitro cultures (fig. S3, B and C). On the other hand, treatment of T cell-competent C57BL/6 mice with ITGB2 blocking ab exacerbated growth of B16-F10 and had no significant effect on YUMM5.2 tumors versus isotype-matched controls (Fig. 3A), consistent with marked suppression of respective T cell infiltration into both B16-F10 and YUMM5.2 grafts by the ITGB2 ab (Fig. 3B) and counteraction of its melanoma-directed antitumor effect in NSG mice (Fig. 2D). Strikingly, however, ITGB2 blockade inhibited pulmonary metastasis formation 8-fold and >270-fold, respectively, of subcutaneously implanted GFP-labeled B16-F10 and YUMM5.2 tumors of comparable volume, as determined by qPCR of genomic GFP in lung tissue (Fig. 3C). To test host-ICAM-1 dependency of melanoma-ITGB2-driven growth and metastasis, B16-F10 and YUMM5.2 cells were grafted to Icam1^−/− KO C57BL/6 mice and treated with ITGB2 neutralizing or isotype control abs (Fig. 3D). As expected, host- Icam1 deficiency reduced tumor-infiltrating T cell frequencies (Fig. 3E), a scenario resembling T cell-deficient NSG models responsive to melanoma-ITGB2 inhibition (Fig. 2D). However, ITGB2 blockade did not substantially affect tumor growth in Icam1 null mice (Fig. 3D), consistent with the reliance of melanoma cell-ITGB2 on host-ICAM-1 in tumorigenesis. Nevertheless, host-Icam1 KO increased growth of both B16-F10 and YUMM5.2 tumors (Fig. 3D) relative to ICAM-1-expressing wildtype mice (Fig. 3A), in agreement with elevated tumoral T cell abundance in the latter model (Fig. 3E). Strikingly, despite harboring larger primary tumors, pulmonary dissemination was suppressed in host Icam1 KO versus ICAM-1-expressing wildtype mice, particularly for YUMM5.2 tumors (Fig. 3F), consistent with potent melanoma-ITGB2:host-Icam1 pro-metastatic activity (Fig. 3C). These results identify a melanoma cell-intrinsic ITGB2:ICAM-1 circuit in tumor growth and metastasis.

Fig. 2.

Antibody-based blockade of melanoma cell-intrinsic ITGB2 inhibits ICAM-1-dependent adhesion and growth (A and B) Relative in vitro adhesion (mean ± SEM) to immobilized ICAM-1 versus negative coating control of (A) human melanoma C8161 and MDA-MB-435S or positive control HSB-2 cells and (B) murine melanoma B16-F10 and YUMM5.2 or positive control EL-4 cells, either untreated (respective left panels) or treated with ITGB2 blocking ab or EDTA pan-integrin antagonist versus isotype control ab (respective right panels). (C and D) Tumor growth kinetics in vivo (mean ± SEM) of (C) human C8161 and MDA-MB-435S cells in NSG mice treated with human-specific ITGB2 blocking ab versus isotype control ab or (D) murine B16-F10 and YUMM5.2 cells in NSG mice treated with anti-murine ITGB2 blocking versus isotype control ab. Results in panels (A and B) are representative of and/or pooled from at least n = 3 independent experiments. The unpaired Student’s t test was used to statistically compare two groups and one-way ANOVA with Dunnett’s post-test for comparison of three groups. Panels (C and D) involved n = 5–20 mice per respective treatment group. Repeated-measures two-way ANOVA or mixed model followed by Šídák’s multiple comparisons correction were used to assess statistical differences in tumor growth. *, p < 0.05; **, p < 0.01; ***, p < 0.001. See also Figs. 3 and 4, and 5, fig. S3

Fig. 3.

Antibody-based ITGB2 blockade or host Icam1 deficiency inhibit melanoma metastasis (A to C) Effect of anti-murine ITGB2 blocking ab versus isotype control ab on tumorigenesis of B16-F10 and YUMM5.2 cells in wildtype (WT) C57BL/6 mice. (A) Tumor growth kinetics (mean ± SEM), (B) relative intratumoral T cell levels, and (C) relative lung metastasis of GFP-expressing melanoma cells were determined by qPCR-based quantitation of genomic Cd3 or GFP in tumor and lung tissue, respectively. (B) Primer specificity for Cd3 was validated using positive control murine T cells and negative control B16-F10 and YUMM5.2 cells. (C) Specificity of GFP primers was authenticated using positive control GFP-expressing B16-F10 and YUMM5.2 cells and negative control lungs obtained from WT mice without tumors. (D to F) Effect of anti-murine ITGB2 blocking ab versus isotype control ab on tumorigenesis of B16-F10 and YUMM5.2 cells in Icam1^−/− C57BL/6 mice. (D) Tumor growth kinetics (mean ± SEM), (E) intratumoral T cell levels, and (F) lung metastasis in Icam1-deficient mice were determined by qPCR analysis using positive and negative cell and sample controls, as above. Panels (A and D) involved n = 16–20 mice per respective treatment group. Results in panels (B, C, E, and F) are representative of and/or pooled from at least n = 3 independent experiments. Tumor control groups in panels B and E, C and F are identical, respectively. Repeated-measures two-way ANOVA or mixed model followed by Šídák’s multiple comparisons correction were used to assess statistical differences in tumor growth in panels (A and D). Data in (B, C, E, and F) were statistically compared using the unpaired Student’s t test. *, p < 0.05; NS, not significant; nd, not detected. See also Figs. 4 and 6, fig. S3

CRISPR/Cas9-based genetic knockout of melanoma cell-intrinsic Itgb2 suppresses adhesion to ICAM-1 and resultant tumor growth

To assess ITGB2 functions in tumorigenesis directly on cancer cells, we generated stable Itgb2 gene KO B16-F10 and YUMM5.2 melanoma cells using CRISPR/Cas9. Homozygous loss of Itgb2 gene and ITGB2 protein was validated by genomic sequencing (not shown), RT-qPCR and immunoblotting (Fig. 4A). Melanoma cell-intrinsic Itgb2 KO significantly inhibited both adhesion to immobilized ICAM-1 (Fig. 4B) and in vitro three-dimensional culture growth of B16-F10 and YUMM5.2 cells (Fig. 4C). In agreement with the observed inhibition of tumor cell growth in vitro and its suppression in vivo in NSG mice via ab-mediated ITGB2 blockade (Fig. 2, C and D), melanoma-specific Itgb2 KO robustly antagonized B16-F10 and YUMM5.2 tumor advancement in both NSG (Fig. 4D) and C57BL/6 hosts (Fig. 4E). In contrast, YUMM5.2 Itgb2 KO tumor volumes did not significantly differ from respective controls in Icam1^−/− null C57BL/6 mice (Fig. 4F), further implicating ICAM-1 dependency of protumorigenic melanoma-ITGB2 activity. On the other hand, Itgb2 KO in B16-F10 cells significantly reduced tumor growth compared to controls, even in Icam1^−/− C57BL/6 mice (Fig. 4F), suggesting tumor cell-intrinsic ITGB2 involvement in melanoma progression that is at least partly independent of host-ICAM-1. Indeed, murine melanoma cells intrinsically expressed ICAM-1 in vitro (fig. S4A), which was induced to even greater levels in vivo (fig. S3C). Of note, tumor cell-Icam1 was substantially more upregulated in B16-F10 versus YUMM5.2 tumor grafts in Icam1^−/− KO compared to wildtype hosts (Fig. 4G). This result is suggestive of greater contribution of cancer cell-ICAM-1 versus host-ICAM-1 in B16-F10 relative to YUMM5.2 growth. It thus potentially explains the lesser reliance of ITGB2-mediated B16-F10 tumorigenesis on host-ICAM-1 (Fig. 4F). In support of such divergence in tumor-intrinsic ICAM-1 dependency, human melanoma lines showed marked variation in ICAM1 gene and ICAM-1 protein expression, ranging from negligible levels to amounts exceeding those in T cell lineage and positive control EC types (fig. S4B). This wide spectrum of ICAM1 gene (Fig. 4H and fig. S4C) and ICAM-1 protein (Fig. 4I and fig. S4D) expression level was similarly observed in patient melanoma cells. Multicolor immunofluorescence analyses of patient melanoma tissue revealed a tendency for ITGB2⁺ melanoma cells to reside in proximity to ICAM-1⁺ cells (Fig. 4J). Moreover, ICAM-1 showed significantly increased expression on ITGB2⁺ versus ITGB2⁻ patient melanoma cells as determined by multicolor FC (fig. S4E). In total, these findings substantiate a significant functional role for tumor cell-intrinsic ITGB2 in ICAM-1-dependent tumorigenesis and provide a rationale for therapeutic targeting of melanoma-ITGB2:ICAM-1 interactions.

Fig. 4.

CRISPR/Cas9-based genetic knockout of melanoma cell-intrinsic Itgb2 suppresses adhesion to ICAM-1 and resultant tumor growth (A) Validation of CRISPR/Cas9-mediated stable KO of Itgb2 gene and ITGB2 protein in B16-F10 and YUMM5.2 melanoma cells as determined by RT-qPCR (left panel) and immunoblotting (right panel). (B to F) Itgb2 KO versus respective Cas9 control B16-F10 and YUMM5.2 tumor cell relative (B) in vitro adhesion (mean ± SEM) to immobilized ICAM-1, with or without negative control EDTA treatment, (C) in vitro growth (mean ± SEM) as determined by CellTiter-Glo-based luminescence analysis, and (D to F) in vivo tumor growth kinetics (mean ± SEM) in (D) NSG mice, (E) C57BL/6 mice, and (F) Icam1^−/− C57BL/6 mice. (G) Relative Icam1 gene expression in B16-F10 and YUMM5.2 tumors from C57BL/6 mice (black bars) versus Icam1^−/− C57BL/6 mice (white bars), with positive control murine T cells and C166 endothelial cells shown (gray bars). (H) scRNA-seq analysis of human ICAM1 gene expression in patient melanoma (MM) cells, tumor-infiltrating T cells, and endothelial cells (ECs) as depicted by violin plots (median, bold white line; top and bottom quartiles, thin white lines) overlayed with dots representing respective single cells. (I) Percentages (mean) of human ICAM-1 surface protein expression by patient MM cells, T cells, and ECs (n = 5 patients) as determined by FC. (J) Multiplex immunofluorescence staining of a representative (n = 4 patients) clinical melanoma biospecimen for expression of the melanocytic marker, nuclear SOX-10 (red, first panel), ITGB2 (yellow, second panel), and ICAM-1 (green, third panel). The merged image is also shown (fourth panel). Nuclei were counterstained with DAPI (blue). Size bars, 50 μm. Results in panels (A, B, C, and G) are representative of and/or pooled from at least n = 3 independent experiments. The unpaired Student’s t test was used to statistically compare two groups and one-way ANOVA with Dunnett’s post-test for comparison of three groups. Panels (D to F) involved n = 10 mice per respective melanoma cell variant. Repeated-measures two-way ANOVA was used to assess statistical differences in tumor growth. **, p < 0.01; ***, p < 0.001; NS, not significant; nd, not detected. See also Figs. 2 and 4, figs. S3 and S4

Melanoma cell-ITGB2:ICAM-1 interaction stimulates downstream Wnt pathway activation, the pharmacologic inhibition of which suppresses ITGB2-mediated tumorigenesis

To systematically dissect melanoma cell-ITGB2 downstream effector pathways involved in ICAM-1-dependent tumorigenesis, we performed unbiased RNA-seq analyses of Itgb2 KO versus control YUMM5.2 tumors grown in NSG, wildtype or Icam1^−/− null C57BL/6 mice. We identified 151 differentially expressed genes (DEGs) among Itgb2 KO versus control tumors altered consistently in both NSG and wildtype C57BL/6, but not Icam1^−/− C57BL/6 hosts (Fig. 5A; the 51 most DEGs are shown). STRING protein association network analysis [30] identified three major interconnected clusters consisting of 22 genes strongly associated with melanoma-ITGB2 downstream pathway activity. These included Wnt signaling effectors, integrin and extracellular matrix (ECM) molecules, and growth factor network components. Notably, the Wnt mediators [31, 32], Wnt5a, Wnt5b, Notum, and Dkk2, exhibited the strongest interaction scores (Fig. 5B), consistent with the established association of Wnt signaling with ICAM-1 expression in cancer cells [33]. Because Wnt signaling is implicated in melanoma phenotype switching [34] and resistance to BRAF inhibition [35, 36], we examined melanoma cell-ITGB2 and ICAM1 association with proliferative versus neural crest-like invasive and BRAF resistance states. Analysis of published scRNA-seq datasets [26, 27] revealed similar ITGB2 and ICAM1 expression across melanoma phenotypes, with ITGB2 enriched in proliferative (mitotic) over invasive (neural crest-like) tumor cells (fig. S5A), and in pro-proliferative [37] ABCB5⁺ but not pro-invasive [26] NGFR⁺ subsets versus respective negative cohorts (fig. S5B, C). Both Wnt pathway and ITGB2:ICAM1 genes showed increased expression among BRAF treatment resistant versus naïve melanoma cells (fig. S5D, E) and Itgb2 KO enhanced sensitivity to the BRAF tumor growth inhibitor, vemurafenib, compared to controls (fig. S5F).

Fig. 5.

The melanoma cell-ITGB2:ICAM-1 axis stimulates downstream Wnt pathway activation, the inhibition of which suppresses cancer cell:ICAM-1 adhesion (A) Heatmaps of differentially expressed genes (DEGs) exhibiting pathway interconnectivity (n = 51) in Itgb2 KO versus control YUMM5.2 tumors and which showed consistent trends in both NSG (left panel) and wildtype (WT) C57BL/6 mice (middle panel), but not in Icam1^−/− C57BL/6 hosts (right panel), as determined by RNA-seq analysis. (B) Protein-protein interaction and cluster map (STRING) of 22 of the 51 DEGs described in (A) exhibiting the strongest interaction scores. Respective network clusters (gray ovals) and relative strengths of direct protein-protein interactions (stronger, wider lines; weaker, thinner lines) as well as indirect associations (dashed lines) are shown. Proteins without any designated cluster associations were omitted. The paired Wilcoxon test was used to assess statistical significance. (C) Magnitude of difference in expression of each Wnt pathway DEG in Itgb2 KO versus Cas9 control melanomas (log₂ fold change) as in (A) and identified in the Gene Ontology Biological Process (GOBP) database. Wnt signaling effectors were grouped into activating (Frat2, Kpna1, Wnt5a, Wnt5b) versus inhibitory (Dkk2, Igfbp4, Kank1, Notum) cohorts. Medians are represented by horizontal bars in box and whiskers plots. (D) Validation by RT-qPCR (fold change) of Wnt effector DEGs as in (C) using independent Itgb2 KO versus Cas9 control YUMM5.2 tumor biospecimens from NSG, WT, or Icam1^−/− C57BL/6 mice. Medians are represented by horizontal bars in box and whiskers plots. (E) Representative immunoblots of canonical Wnt mediators, active (non-p) β-catenin and LEF-1, and ACTB loading control (left), and non-canonical Wnt effector, p-VANGL2, and respective total controls (right) in Itgb2 KO versus Cas9 control YUMM5.2 melanoma cells. (F) Representatie immunoblots of Wnt signaling mediators as in (E) of YUMM5.2 melanoma cells treated with the Wnt inhibitors, pyrvinium pamoate, LGK974, or zamaporvint, versus vehicle control. (G and H) Relative in vitro adhesion (mean ± SEM) to immobilized ICAM-1 as determined by CellTiter-Glo-based luminescence analysis of (G) Itgb2 KO versus Cas9 control YUMM5.2 variants and (H) anti-murine ITGB2 blocking ab versus isotype control ab treated YUMM5.2 wildtype cells, in the combined presence or absence of pyrvinium pamoate, LGK974, zamaporvint, or vehicle control. The paired Student’s t test was used to assess statistical significance. Panels (A, B, C, and D) are representative of n = 2–6 tumors per variant group in each respective animal host. Results in (E, F, G, and H) are representative of and/or pooled from at least n = 2–7 independent experiments each. *, p < 0.05; **, p < 0.01; ***, p < 0.001; NS, not significant. See also Figs. 2 and 3, and 6, figs. S5 and S6

Focused analysis of individual Icam1-dependent DEGs (Fig. 5A) identified in the Gene Ontology Biological Process (GOBP) database as canonical or non-canonical Wnt signaling pathway members, revealed induction of both activating (Frat2, Kpna1, Wnt5a, Wnt5b) and inhibitory (Dkk2, Igfbp4, Kank1, Notum) Wnt effectors [38] in Itgb2 KO versus Cas9 control YUMM5.2 tumors in both NSG and wildtype C57BL/6 but not Icam1^−/− hosts (fig. S6A). Conspicuously, endogenous Wnt inhibitors predominated over activators (Fig. 5C), suggestive of overall Wnt pathway suppression from Itgb2 KO. In agreement, RT-qPCR analyses of a separate Itgb2 KO versus Cas9 control in vivo cohort independently confirmed enrichment of the above Wnt genes (fig. S6B) as well as preferential expression of inhibitory over activating Wnt effectors (Fig. 5D). Consistently, in Itgb2 KO versus Cas9 control YUMM5.2 cells, immunoblotting revealed decreased expression of the canonical Wnt effectors, active (non-p) β-catenin and LEF-1, with lesser reduction of the non-canonical Wnt mediator, p-VANGL2 (Fig. 5E).

To corroborate Wnt pathway involvement in melanoma cell-ITGB2:ICAM-1 functions, we employed three independent pharmacologic Wnt signaling antagonists, including FDA-approved pyrvinium pamoate, repurposed for cancer therapy [19, 39], as well as LGK974 and zamaporvint [20, 21]. All three Wnt inhibitors not only suppressed canonical and, to a lesser extent, non-canonical Wnt signaling (Fig. 5F) but also abrogated ITGB2:ICAM-1-specific adhesion of melanoma cells (Fig. 5G). Similar to the effect of ITGB2 ab blockade, pyrvinium, LGK974, or zamaporvint treatment also significantly inhibited ICAM-1 adhesion of YUMM5.2 cells treated with isotype control ab (Fig. 5H). Combined incubation with pyrvinium or zamaporvint and ITGB2 neutralizing ab synergistically inhibited ICAM-1 binding (Fig. 5H). Consistently, Itgb2 KO or administration of pyrvinium, LGK974, or zamaporvint significantly suppressed YUMM5.2 tumor growth in NSG and C57BL/6 hosts compared to controls (Fig. 6A). Similarly, pyrvinium, LGK974, or zamaporvint suppressed tumorigenesis of isotype control ab-treated wildtype YUMM5.2 tumors and synergized with ITGB2 ab blockade in thwarting tumor outgrowth in T cell-deficient NSG, but not T cell-competent C57BL/6 mice (Fig. 6B). Strikingly, the tumor growth-suppressive efficacy of all three Wnt inhibitors, like that of upstream Itgb2 receptor KO or blocking ab treatment, was fully abrogated in Icam1^−/− KO C57BL/6 mice (Fig. 6A, B). Together, these findings uncover a melanoma cell-intrinsic ITGB2:ICAM-1:Wnt protumorigenic signaling circuit and highlight its translational potential for tumor cell-directed antagonism of ITGB2-dependent cancer progression.

Fig. 6.

Wnt antagonism suppresses ITGB2:ICAM-1-dependent melanoma growth in vivo (A and B) Tumor growth kinetics (mean ± SEM) of (A) Itgb2 KO versus Cas9 control YUMM5.2 variant cells or (B) YUMM5.2 wildtype cells treated with anti-murine ITGB2 blocking ab versus isotype control ab, with or without concurrent administration of the Wnt inhibitors, pyrvinium pamoate, LGK974, zamaporvint, as well as vehicle control in NSG (left panel), wildtype (WT) C57BL/6 (middle panel), or Icam1^−/− C57BL/6 mice (right panel). Because tumorigenicity experiments evaluating LGK974 and zamaporvint effects were conducted concurrently, vehicle control groups for both drugs are identical. Panels (A and B) involved n = 6–10 mice per respective treatment group. Repeated-measures two-way ANOVA or mixed model followed by Šídák’s multiple comparisons correction were used to assess statistical differences in tumor growth in panels. *, p < 0.05; **, p < 0.01; ***, p < 0.001; NS, not significant. See also Fig. 5

Discussion

ITGB2 cell-adhesive molecules are critical for immune cell activation, proliferation and trafficking [40]. Since their discovery, expression of ITGB2 heterodimers has been claimed to be exclusive to cells of the hematopoietic lineage, including adaptive and innate immune cells as well as hematologic cancer cells [5, 41, 42]. More recently, there have been a few reports finding ITGB2 expression in some epithelial cancer types [6–8], albeit with incompletely defined functions in tumor progression. Using several independent methodologies, our study now reveals functional ITGB2 expression intrinsic to melanoma cells in patient tumor biospecimens and established human and murine cancer cell lines. Percentages of ITGB2⁺SOX10⁺ cells increased from benign nevi to primary melanomas to metastatic disease in clinical biospecimens. Moreover, in patient primary melanomas, ITGB2⁺ cancer cell frequency correlated with SLN metastasis, implicating tumor cell-intrinsic ITGB2 involvement in neoplastic advancement.

Crosslinking of the HA receptor [43], CD44, on melanoma cells induced expression and functional activation of ITGAL/ITGB2 as recognized by conformation-sensitive abs, thus mimicking CD44-dependent regulation of ITGAL/ITGB2 in leukocytes [9]. The other α pairing subunits, ITGAD, ITGAM, or ITGAX, were expressed at lower levels in patient melanoma cells and not induced by CD44 crosslinking. Hence, modulation of melanoma cell-ITGB2 heterodimer composition, level, and activation is likely to involve additional factors and layers of complexity beyond CD44. Such regulator(s) might further vary by identity and concentration across in vitro culture or in vivo conditions to control, or even select for, frequencies of ITGB2-expressing melanoma subpopulations. Similar to leukocyte-ITGB2 [2], melanoma cell-ITGB2 bound ICAM-1. This interaction was inhibited via ab-based ITGB2 blockade, EDTA integrin antagonism, or Itgb2 KO. Together, these findings uncover a new ‘hematopoietic mimicry’ mechanism [10] by which melanoma cells coopt leukocyte ITGB2 processes to facilitate tumor progression.

Indeed, ab-mediated ITGB2 blockade and CRISPR/Cas9-based tumor cell-Itgb2 KO significantly suppressed growth of human and murine melanomas in immunocompromised NSG mice, paralleling the known anti-proliferative effect of ITGB2 inhibition in T cells [44]. Similarly, in immunocompetent C57BL/6 mice, Itgb2 KO also inhibited tumorigenesis. In this model, administration of ITGB2 blocking ab did not attenuate growth, consistent with its reduction of tumor-infiltrating T cells. Nevertheless, ITGB2 blockade significantly reduced spontaneous pulmonary metastasis formation compared to isotype control treatment in C57BL/6 mice with comparable primary tumor volumes, underscoring dual functions of melanoma cell-ITGB2 in growth and metastatic dissemination. In this T cell-competent setting, inhibition of T cell-ITGB2 thus counteracted tumor growth suppression from melanoma cell-ITGB2 blockade, but not metastasis formation. Clearly, antagonists with selectivity towards cancer cell-ITGB2 would avoid undesired tumor growth-promoting effects from T cell-ITGB2 blockade. Hence, our data of pronounced growth inhibition via melanoma-Itgb2 KO provides proof-of-principle for the feasibility of such tumor cell-directed ITGB2 targeting approaches. A recent publication leveraged another cancer cell-ITGB2 inhibitory strategy involving chimeric antigen receptor (CAR) T cells targeting a leukemia-specific ITGB2 conformation to eliminate tumor cells while sparing normal hematopoietic cells [42]. Similar structural epitopes could be exploited to selectively target active forms of melanoma cell-intrinsic ITGB2. The drawback of cell type-nonspecific clinical ITGB2 inhibitors evaluated previously for the treatment of inflammatory and autoimmune disorders [2, 5], including erlizumab (rhuMAb) and rovelizumab (LeukArrest) [45], is highlighted by their withdrawal. Despite high drug potency and intended disruption of immune cell-ITGB2 functions, a small subset of patients developed progressive multifocal leuko-encephalopathy due to inability to clear the opportunistic JC polyomavirus [45]. Nevertheless, these clinical studies highlight ITGB2 as a feasible target and urge development of cancer cell-specific ITGB2 inhibitors.

The promise of targeting cancer-intrinsic ITGB2 is further bolstered by the fact that statins, a family of FDA-approved small molecule inhibitors with newly uncovered ITGB2:ICAM-1 neutralizing activity [46] also show antitumor properties across multiple malignancies [47]. Though originally developed to inhibit cholesterol biosynthesis [47], statins have recently been found to directly bind ITGB2 and block its interaction with ICAM-1 [46]. In light of our findings, the observed antitumor activity of statins [46, 47] might thus relate to their direct antagonism of tumor cell-intrinsic ITGB2:ICAM-1 interactions, therefore warranting further investigation in the cancer context.

Our data demonstrating reduced frequency of tumor-infiltrating T cells in syngeneic Icam1^−/− null versus wildtype C57BL/6 mice is consistent with the established role of endothelial ICAM-1 in ITGB2-dependent T cell homing [40]. Compared to wildtype mice, host-Icam1 deficiency also mirrors ITGB2 inhibition in its significant suppression of spontaneous metastasis formation in mice with size-matched B16-F10 and YUMM5.2 primary tumors. These results thus identify ICAM-1 as a crucial mediator not only of T cell but also tumor cell trafficking. Host-Icam1 loss-of-function additionally reversed growth inhibition from melanoma-Itgb2 KO in YUMM5.2 tumor-bearing mice. These results corroborate important roles of host-Icam1 in ITGB2-dependent tumorigenesis. On the other hand, growth of Itgb2 KO B16-F10 melanomas did not significantly differ between Icam1 null and wildtype hosts, suggesting that cancer cell-expressed ICAM-1 might compensate for host-Icam1 deficiency in some cases. In fact, melanoma cells, like leukocytes, intrinsically express ICAM-1 at varying levels as reported herein and by others [48]. Indeed, tumoral Icam1 expression was substantially higher in B16-F10 versus YUMM5.2 grafts in Icam1-deficient hosts, pointing to significant involvement of both host and tumor-intrinsic ICAM-1 in ITGB2-mediated growth. Because additional ICAM family members are known to also bind ITGB2 [40], melanoma-ITGB2 interaction with other ICAMs might also modulate tumorigenesis. Nonetheless, our findings identify a pivotal role for cancer cell-ITGB2:ICAM-1 interactions in melanoma progression.

The ITGB2:ICAM-1 axis modulated downstream Wnt pathway activity [31, 32, 49], as determined by unbiased RNA-seq, RT-qPCR, and immunoblotting analyses. Melanoma-Itgb2 KO significantly induced tumor cell-intrinsic gene expression of multiple Wnt pathway members in NSG and wildtype but not Icam1^−/− null C57BL/6 mice. Nonetheless, endogenous inhibitory Wnt mediators showed greater elevation over activating Wnt effectors, indicative of collective pathway suppression from Itgb2:Icam1 axis deficiency. Indeed, immunoblotting confirmed inhibition of established canonical Wnt effector proteins in Itgb2 KO versus Cas9 control cells, whereas non-canonical signaling showed no apparent differences. These results are consistent with enrichment of canonical Wnt signaling in patients with high tumoral ICAM-1 levels [33]. They are also in agreement with our findings of preferential Wnt, ITGB2 and ICAM1 gene expression in BRAF inhibitor resistant versus treatment naïve patient melanoma cells [27], illuminating a tumor-intrinsic ITGB2:ICAM-1:Wnt pathway. Because our unbiased RNA-seq and STRING analyses uncovered additional gene clusters other than Wnt downstream of melanoma-Itgb2:Icam1, including growth factor networks, ITGB2 might also regulate alternative targetable mediators, thus rationalizing further investigation. Nevertheless, pharmacological Wnt antagonism using clinical grade LGK974 [20], zamaporvint [21], or FDA-approved [39] pyrvinium [19] potently suppressed ITGB2:ICAM-1-mediated adhesion and melanoma growth in NSG and wildtype C57BL/6 mice, but not Icam1^−/− null hosts. These findings identify Wnt as a critical melanoma-ITGB2:ICAM-1 downstream effector pathway and underscore the therapeutic promise of targeting multiple vulnerabilities of the tumor-intrinsic ITGB2 signaling cascade.

Conclusions

Our work uncovers a melanoma cell-intrinsic ITGB2:ICAM-1:Wnt protumorigenic axis. Inhibition of melanoma cell-ITGB2:ICAM-1 interactions or downstream Wnt signaling potently suppressed tumorigenesis in preclinical models, thereby illuminating the tumoral ITGB2 pathway as a bona fide cancer therapeutic target in melanoma and potentially other ITGB2-expressing malignancies [6–8]. Our findings further rationalize the development of next generation [42, 50] cancer cell-specific ITGB2 antagonists that do not recognize leukocyte-ITGB2. Such tumor-directed ITGB2 inhibitors would suppress tumorigenesis while also avoiding unwanted adverse immune consequences from conventional cell type non-selective ITGB2 blockade [5, 45].

Abbreviations

Ct

Threshold cycle

EDTA

Ethylenediaminetetraacetic acid

FC

Flow cytometry

FBS

Fetal bovine serum

GEO

Gene Expression Omnibus

HRP

Horseradish peroxidase

IACUC

Institutional Animal Care and Use Committee

ICAM-1

Intercellular adhesion molecule 1

Icam1^-/-

Intercellular adhesion molecule 1 knockout

ITGAD

Integrin αD, CD11d

ITGAL

Integrin αL, CD11a

ITGAM

Integrin αM, CD11b

ITGAX

Integrin αX, CD11c

ITGB2

Integrin β2, CD18

KO

Knockout

MFI

Mean fluorescence intensity

Nd

Not detected

NSG

Nonobese diabetic/severe combined immunodeficiency (NOD/SCID) IL-2Rγ(-/-) knockout

PBS

Phosphate-buffered saline

RT

Room temperature

RT-qPCR

Reverse transcription quantitative polymerase chain reaction

SEM

Standard error of the mean

Seq

Sequencing

TBS

Tris-buffered saline

TBS-T

Tris-buffered saline-Tween 20 (0.1%)

WT

Wildtype

Authors’ contributions

E.R., A.B., S.R.B., and T.S. conceptualized the study. E.R., L.M., A.B., P.S., C.M., A.K., A.I.K., N.S., Z.K., J.R., M.S., E.Z., S.X., K.M., J.H., J.B.W., N.L., S.R.B., and T.S. carried out experimental work. E.R., L.M., A.B., P.S., C.M., A.K., A.I.K., N.S., Z.K., J.R., M.S., E.Z., S.X., K.M., J.H., E.B., M.P.L., J.M.M.G., R.D., J.L., C.G.L., G.F.M., E.L.B., S.H.S., T.S.K., N.N.R., N.L., S.R.B., and T.S. analyzed data or provided clinical samples. E.R., L.M., A.B., S.R.B., and T.S. wrote the paper. All authors discussed the results and commented on the manuscript.

Funding

This work was supported by a Research Grant from the Dermatology Foundation, Milstein Research Scholar Award from the American Skin Association, Fund to Sustain Research Excellence from the Brigham Research Institute, Brigham and Women’s Hospital, NIH/NCI grants R01CA190838 (to T.S.), R01CA247957 and R01CA258637 (to S.R.B. and T.S.). Partial support was provided by a Walter Benjamin Scholarship from the German Research Foundation (DFG 470684420) and a POTENTIAL Clinician Scientist Fellowship from the Else-Kroener-Fresenius Foundation (EFKS, both to E.R.), a Foreign Exchange Fellowship from the German National Merit Foundation and a Carl Duisberg Fellowship from the Bayer Foundation (to A.B.), a Dermatology Fellowship Award from the Melanoma Research Alliance (MRA, to M.S.), a Klaus-Wolff Fellowship from the Austrian Society of Dermatology and Venereology (OeGDV) and a Marietta Blau-Grant from Austria’s Agency for Education and Internationalization (OeAD, to J.H.), the BioBank Core Facility of the University Hospital Bonn, Germany (to J.L.), and a Medical Student Grant from the American Skin Association (ASA, to N.L.).

Data availability

The raw RNA-seq data reported in this study are publicly available in Gene Expression Omnibus (GEO) at GSE261845. The single-cell RNA-seq data analyzed in this study were obtained from GEO at GSE72056 or GSE233766, or from https://doi.org/10.48804/GSAXBN. Code used for this manuscript has been made available in the Github repository under the following link: https://github.com/hbc/HBC04855_Schatton_mouse_RNAseq. Methodological details on parameters used are available in the Methods section of this manuscript. Source data are publicly available within the Harvard Dataverse at 10.7910/DVN/5LFKTB. Correspondence and requests for materials should be addressed to Dr. Tobias Schatton or Dr. Steven R. Barthel.

Declarations

Ethics approval and consent to participate

All studies involving human specimens were approved by Institutional Review Boards of Mass General Brigham, the University of Zurich, Switzerland, and the University of Bonn, Germany. All studies involving animal experiments were approved by BWH Institutional Animal Care and Use Committee (IACUC).

Consent for publication

Not applicable.

Competing interests

R.D. is a consultant for Novartis, Merck Sharp & Dohme (MSD), Bristol-Myers Squibb (BMS), Roche, Amgen, Takeda, Pierre Fabre, Sun Pharma, Sanofi, Catalym, Second Genome, Regeneron, Alligator, T3 Pharma, MaxiVAX SA, Pfizer, Simcere, and touchIME, not related to the submitted work. All other authors declare that no competing interests exist.