Kaplan–Meier curves for relapse-free survival and overall survival were calculated for all subtypes of invasive breast carcinoma and for the luminal B subtype. Datasets were derived from Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/gds) and The European Genome-phenome Archive (EGA) (https://www.ebi.ac.uk/ega/) and were analyzed by a Kaplan–Meier plotter (23). The cutoff for high and low groups was set by using the auto-select best cutoff. All other settings were set to default. Luminal B patient analysis was based on the 2013 StGallen criteria (24). P-values were calculated using the log-rank test (Mantel–Cox test) and by the Cox proportional hazard ratio using 95% confidence interval.

Immunoblotting was performed as described previously (25). The antiestrogen receptor-α antibody (Clone F-10, Cat. SC-8002) was used to detect the receptor in E0771 cells. Michigan Cancer Foundation-7 (MCF-7) cells were used as a positive control. Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) was used as a loading control.

Murine breast adenocarcinoma cells (E0771) were purchased from CH3 Biosystems (Stock Keeping Unit (SKU): 94A001) and grown in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1-mM sodium pyruvate, 10-mM HEPES, and 1% penicillin–streptomycin–amphotericin B solution. G-CSF-overexpressing B16-F10 cells (26) were a generous gift of Dr. Asaph Spiegel from the Robert A. Weinberg lab at the Whitehead Institute for Biomedical Research, Cambridge, Massachusetts, and were grown in DMEM supplemented with 10% FBS and 1% penicillin–streptomycin–amphotericin B solution. 4T1 cells were obtained from American Type Culture Collection (ATCC) and grown in RPMI 1640 supplemented with 10% FBS and 1% penicillin–streptomycin–amphotericin B solution. Cell lines were passaged up to 10 times and were periodically validated as negative for mycoplasma. Cell cultures were kept at 37°C in a humidified incubator in the presence of 5% CO₂. Murine CD8 effector T cells were generated from T cells isolated from the spleens of OT-I mice. Spleen-derived lymphocytes were stimulated in vitro with SIINFEKL-loaded syngeneic DCs and expanded in Interleukin-2 (IL-2)-rich media as described (27).

Clustered regularly interspaced short palindromic repeats (CRISPR) guides were chosen using several design tools, including: the Massachusetts Institute of Technology (MIT) CRISPR design tool (28) and Single guide RNA (sgRNA) Designer, Rule set 2 (29), in the Benchling implementations (www.benchling.com), SSC (30), and sgRNAscorer (31), in their websites. E0771 cells were transfected by electroporation in Opti-Minimum Essential Media (MEM) media with the Cas9 enzyme (Cat. 1081058, Integrated DNA Technologies (IDT)), two guide RNA oligoes, and tracer RNA (Cat. 1073189, IDT). The sequences of the guide RNA oligos were GCGCGAGGTTTCCCGGAAAG (Guide 1) and AGGGATCACAACGGTGACCA (Guide 2). Cells were incubated for 72 h after transfection and were stained and sorted by fluorescence-activated cell sorting (FACS) for ICAM-1-positive and -negative expressing cells.

For the analysis of cell surface molecule expression, E0771 cells were stained in an FACS buffer [Phosphate-buffered saline (PBS) supplemented with 2% bovine serum albumin (BSA) and 5-mM Ethylenediaminetetraacetic acid (EDTA)] with monoclonal antibodies specific to ICAM-1 (Allophycocyanin (APC) conjugated, clone YN/1.7.4, Cat. 116119, Biolegend); ICAM-2 (Alexa 647 conjugated, clone 3C4 (MIC2/4), Cat. 105611, Biolegend); CD47 (APC conjugated, clone miap301, Cat. 127513, Biolegend); Integrin β1 (APC conjugated, clone HMB1-1, Cat. 102215 Biolegend); CXCR4 (PE conjugated, clone 1.276F12, Cat. 146505, Biolegend); MHC-I (H2K-b) (APC-conjugated, Clone- AF6-88.5, Cat. 116517, Biolegend), and to SIINFEKL bound to H2K-b [Clone 25-D1.16 (32), Cat. 12-5743-82, Thermo Fisher]. Background stainings were determined with matched fluorescence-labeled isotype control monoclonal antibodies (mAbs). P-selectin ligand staining was performed with P-selectin fusion protein as previously described (33). BM neutrophils were stained with an anti-Ly6G antibody (APC/Cy7-conjugated, clone 1A8, Cat. 127623, Biolegend). Cell surface staining was analyzed in a CytoFLEX flow cytometer (Beckman Coulter).

E0771 cells were used as a syngeneic model for orthotopic breast cancer growth, spontaneous metastasis, and experimental metastasis in wild-type C57BL/6 mice (21, 22). Mice were maintained in a pathogen-free facility, and all animal procedures were approved by the Animal Care and Use Committee of the Weizmann Institute of Science. Orthotopic E0771 breast cancer lesions were generated by inoculating the mammary fat pad of 8–12-week-old recipient female mice with 1 × 10³ E0771 cells (suspended in 50 μl of Matrigel^® Matrix diluted 2-fold in PBS). The tumor size was assessed throughout the duration of the experiment by the caliper measurements (34) of length (L) and width (W), and the tumor volume (V) was calculated using the formula: V = (L × W × W)/2. For spontaneous metastasis, 3 × 10⁵ tumor cells (suspended in Matrigel^®, 1:1 in 50 μl of PBS solution) were implanted in the mammary fat pad of recipient female mice. Approximately 2–3 weeks later, when the tumor diameter reached approximately 1.5 cm, tumors were surgically removed. Then, 2–3 weeks later, mice were sacrificed by administration of sodium pentobarbital (200 mg/kg), and lungs were harvested and prepared for LSM or FACS analysis as described in other sections.

Control and ICAM-1 knockout (KO) E0771 cells were seeded at 20,000 cells per well in a 6-well tissue culture plate. To determine their growth, cells were trypsinized, suspended, and counted by flow cytometer. To determine etoposide-induced growth arrest, E0771 cells were plated at 10,000 cells per well in a 96-well tissue culture plate. Approximately 1 day later, etoposide was added to the culture media at various concentrations. Then, 48 h later, the number of cells in each well was determined by flow cytometry and growth inhibition was calculated as

A total of 10³ control or ICAM-1 KO E0771 cells (each suspended in Matrigel, 1:1 in PBS) were implanted in the mammary fat pad and 10 days later, tumors were harvested and fixed O.N. in 4% PFA/PBS solution. Paraffin embedding and haematoxylin and eosin stain (H&E) staining of 5-µm-thin sections were performed in the histology core unit of the Weizmann Institute of Science. Sections were stained according to standard protocols for ICAM-1 IHC (Cat 550547, Biolegend) and for 4′,6-diamidino-2-phenylindole (DAPI). Microscopy-recorded sections were digitalized using a Panoramic SCAN II (3DHISTECH) and analyzed using CaseViewer software (3DHISTECH).

A total of 10³ or 3 × 10⁵ control or ICAM-1 KO E0771 cells were implanted in the mammary fat pad as described in previous sections. Approximately 10 days later, tumors were harvested, minced and incubated in RPMI-1640 containing collagenase A (1.5 mg/ml) and DNase I (20 µg/ml) at 37°C for 45 min (35). Tumor cell suspensions were transferred through a 100-µm cell strainer and centrifuged at 0.2 × g for 5 min at 4°C. Red blood cells (RBCs) were subsequently lysed with and an RBC lysis buffer (Sigma-Aldrich, Rehovot, Israel, Cat. R7757). The remaining cells were resuspended in an ice-cold FACS buffer (PBS with 1% BSA, 0.1% sodium azide and 5-mM EDTA), filtered through a 100-µm strainer, stained according to standard FAC protocols with the cocktails of mAbs to specific cellular markers, and analyzed using a CytoFLEX flow cytometer. In parallel experiments, tumor-draining lymph nodes and spleen were harvested and cell suspensions were prepared, stained and analyzed as above.

10³ control or ICAM-1 KO E0771 cells (suspended in Matrigel, 1:1 in PBS) were implanted in the mammary fat pad of FOXP3-Diphtheria toxin receptor-Green fluorescent protein (DTR-GFP) mice (36). From day 5 to day 13, a diphtheria toxin (DTx) (50-µg/kg body weight, Cat. D0564, Sigma-Aldrich) was injected Intraperitoneal (i.p) every 2 days. The E0771 tumor size was measured on days 11, 13, and 15. Tumor-draining lymph nodes were harvested on day 15, single-cell suspensions were prepared, and their T regulatory cell (Treg) content was determined by the flow cytometry of GFP-positive cells.

Control and ICAM-1 KO E0771 cells were stably transduced with an OVA-mCherry-encoding construct cloned in the pHR OVA/p2a/mCherry-CaaX lentiviral vector (Addgene, cat. 113030). Cells expressing the mCherry reporter were sorted and their expression of the OVA SIINFEKL peptide complexed with cell surface MHC-I (H-2Kb) was measured by 25-D1.16 mAb staining (32)). Control or ICAM-1 KO was seeded in a 96-well plate and cultured O.N. OT-1 effector CTLs were labeled with Carboxyfluorescein succinimidyl ester (CFSE) and overlaid on the tumor cells at different effector to target ratios. One day later, the number of viable E0771 cancer cells was determined by FACS. For in vivo killing analysis, 10³ control or ICAM-1 KO E0771 cells (suspended in Matrigel, 1:1 in PBS) were implanted in the mammary fat pad. Approximately 3 days later, 10⁷ effector OT-I CTLs were injected intravenous (i.v) and the tumor size was measured 12, 14, and 16 days later.

1 × 10⁵ control or ICAM-1 KO E0771 cells were seeded on a μ-SlideVI0.4 ibiTreat (ibidi) previously coated with 5 μg/ml fibronectin (Cat. #F1141, Sigma). A day later, the ibidi slide was connected to a microfluidic flow system. OT-1 CTLs were prepared as described above and perfused in binding medium (Hank’s balanced-salt solution containing 2-mg/ml BSA and 10-mM HEPES pH 7.4, supplemented with 1-mM CaCl₂ and 1-mM MgCl₂) into the chamber and allowed to settle on the different E0771 cell monolayers for 4 min in the absence of shear flow. Adherent CTLs were subsequently subjected to progressively increasing shear stresses (started at 5 dyn/cm² and increased to 30 dyn/cm² with 5-s-long intervals as described (37)). Real time imaging was performed utilizing a phase contrast IX83 Inverted Microscope (Olympus) and CTLs that remained firmly adherent to the E0771 monolayers at the end of the detachment assay were counted using cellSens Dimension v1.16 software (Olympus). The fractions of adherent CTLs out of originally settled CTLs were determined in nine fields of view (664 × 664 μm; 324 × 324 nm/pixel).

Primary tumors were generated by orthotopic injections of 0.5 × 10⁶ E0771 cells into the mammary fat pad. Neutrophils were purified 3 weeks later as described before (38). In brief, whole blood was collected by cardiac puncture using a heparinized (Sigma) syringe. The blood was subjected to a discontinuous Histopaque (Sigma) gradient (1.077 and 1.119), and normal-density neutrophils were collected as described (15). Luciferase-containing E0771 cells (10,000/well) were plated in RPMI supplemented with 2% FBS. Approximately 4 h later, purified neutrophils (100,000/well) were added to the plated tumor cells and cocultured overnight with the medium. Following overnight incubation, luciferase activity was measured using a Tecan F200 microplate luminescence reader as described (15).

Orthotopic 4T1 breast cancer lesions were generated in 8–10-week-old NSG females by injection of 0.5 × 10⁶ 4T1-GFP cells into the mammary fat pad as described (39). A blood draw for CTC analysis was performed 3 weeks after implantation through cardiac puncture and processed immediately on a Parsortix microfluidic device (39). Cancer cell-white blood cell (WBC) conjugates were detected by GFP and anti-CD45 staining as described (39).

For spontaneous metastasis, 3 × 10⁵ tumor cells (suspended in Matrigel^®, 1:1 in 50 μl of PBS solution) were implanted in the mammary fat pad of recipient female mice. CD31-positive vessels were labeled 15 min before mice were sacrificed by an intravenous injection of 6 µg of Alexa-647 conjugated anti-CD31 mAb (Cat. 102515, clone MEC13.3). Immediately after the sacrifice, mice were transcardially perfused with PBS and the lungs were inflated via the trachea with low-gelling agarose (Cat. A9045, Sigma-Aldrich), fixed with paraformaldehyde (4% in PBS) for 2 h, dehydrated, and cleared using ethyl cinnamate as described (40, 41). Cleared lung lobes were imaged in an Ultramicroscope II (LaVision BioTec) operated by the ImspectorPro software (LaVision BioTec, Bielefeld, Germany) as described (41). For experimental metastasis, 1 × 10⁴ E0771 cells labeled with 10-µM (5-(and-6)-(((4-chloromethyl)benzoyl)amino)tetramethylrhodamine) (CMTMR) (Cat. C2927, Thermo Fisher Scientific) for 30 min were washed and injected in the retro-orbital sinus of recipient mice. CD31-positive vessels were labeled as described above and VCAM-1 positive vessels were labeled by intravenous injection of 5 µg of an Alexa-647 conjugated anti-VCAM-1 mAb (Cat. BD561612, clone 429 (MVCAM.A) BD). For nuclear staining, CMTMR-labeled E0771 cells were labeled with REDDOT1 (Cat. BTM-40060-T, NBT).

The three-dimensional rendering of LSM was performed via Imaris software (Oxford Instruments, Abingdon, UK). The surfaces of RFP or CMTMR-labeled tumor cells were created using volume (ranging from 1,000 to 200,000 μm⁴) and intensity (max of red fluorescent channel) as defining features to unequivocally separate them from back-ground signals. Each cell was individually segmented and its distance was measured with respect to the CD31-labeled blood vessels. The cell positions of intravascular were determined by this spatial analysis. For the neutrophil location, spots of Ly6G-647-labeled neutrophils were created based on the relative intensity. The distance between the centroids of tumor cells and neutrophils was determined using distance transformation. The distance between tumor cells and vessels was determined using the shortest distance calculation. For tumor nuclear staining, nuclei surfaces were reconstituted using intensity plots.

CMTMR-labeled E0771 cells were resuspended in PBS and injected into the retro-orbital sinus of recipient mice. Mice were sacrificed and transcardially perfused with PBS, and the lungs were extracted, minced, and incubated in RPMI 1640 containing collagenase type 4 (1.5 mg/ml) and Dnase I (20 µg/ml) at 37°C for 40 min. Lung cell suspensions were transferred through a 1,000-µm cell strainer and centrifuged at 0.2 × g or 5 min at 4°C. RBCs were subsequently lysed, and cells were resuspended in an ice-cold FACS buffer, filtered through a 70-µm strainer and analyzed using a CytoFLEX flow cytometer. The CMTMR label was readily distinguishable from the background fluorescence of lung cells.

A total of 10⁴ E0771 cells were injected into the retro-orbital of recipient mice (41) and mice were sacrificed 14 days later, transcardially perfused with PBS, and the lungs were extracted and stored in 4% PFA for 24 h followed by incubation in 1% PFA for additional 24 h. The paraffin embedding and H&E staining of 5-µm-thin sections were performed by the histology core unit of the Weizmann Institute of Science. Sections were digitalized using a Panoramic SCAN II (3DHISTECH) and analyzed using CaseViewer software (3DHISTECH).

A total of 10 (3) WT E0771 cells (in 50 μl of Matrigel^® Matrix diluted 2-fold in PBS) were implanted in the mammary fat pad of C57BL/6 female donor mice. Approximately 14 days after tumor implantation, the spleens of the tumor-bearing donor mice or of naïve mice or of tumor-bearing mice that underwent neutrophil or CD8+ T-cell depletion (by the i.p. injection of the anti Ly6G or anti CD8 antibody 2 days before spleen isolation) were harvested. Notably, the neutrophil content in spleens isolated from WT and ICAM-1 null E0771 tumor-bearing mice was similar. Single-cell suspensions were prepared, and either whole splenocytes or splenocytes depleted of endogenous neutrophils or CD8 T cells were injected i.v. into recipient syngeneic mice into which 10⁴ control or ICAM-1 KO E0771 cells had been injected i.v. a day before. 13 days later, the lungs of the E0771 and splenocyte-injected recipient mice were harvested and processed for histological analysis as described above.

Data in graphs are represented as mean or mean ± standard error of the mean (SEM). Student’s two-tailed unpaired t-test or the Mann–Whitney two-tailed U test was used to determine the significance of the difference between the means of two groups. One- or two-way ANOVA tests were used to compare the means among three or more independent groups/categories. Significance was set to p < 0.05. The statistical details of experiments can be found in the figure legends.

ICAM-1 is expressed by human and murine breast cancer cells (9). The analysis of breast cancer patient data (23) suggests that the expression of ICAM-1 on breast tumor cells is associated with longer relapse-free periods and a higher overall survival of breast cancer patients (Supplementary Figure 1A). The link between elevated ICAM-1 expression in breast cancer cells and the favorable outcome of breast cancer patients prompted us to assess the direct in vivo contribution of ICAM-1 expressed by breast cancer cells to breast tumorigenesis and lung metastasis in syngeneic immunocompetent mice hosts using E0771 as our breast cancer cell model.

We first validated that the E0771 cells used throughout our study are HER2+ and express ER-α+ and can therefore be considered as luminal B breast cancer cells (Supplementary Figure 1B). We next verified that the relapse-free survival of luminal B cancer patients is significantly prolonged by an elevated ICAM-1 expression (Supplementary Figure 1C). To study the role of ICAM-1 in breast tumor progression and metastasis in a murine model, we deleted it from E0771 cells using CRISPR/Cas9 targeting (Supplementary Figure 2). Two RNA guides, 20 bp each, were designed to target a 178-bp segment that includes the initiator methionine in the first exon, common to all forms of the gene. By this strategy, the surface expression of the protein was abolished (Supplementary Figure 2A). The Cas9-targeted cells were sorted by FACs for either ICAM-1-negative cells (herein ICAM-1 KO E0771) or ICAM-1-positive cells expressing normal ICAM-1 levels [herein control E0771 cells (Supplementary Figure 2B)]. The deletion of the 17-bp segment was also verified by PCR (Supplementary Figure 2C). The characterization of other canonical adhesion molecules expressed by this breast cancer line revealed high expression levels of β1 integrins and low expression levels of the low affinity LFA-1 ligand ICAM-2 with a negligible expression of β2 integrins and P-selectin ligand implicated in cancer–cell platelet interactions (42) (Supplementary Figure 3). E0771 cells lack surface CXCR4, a widely expressed chemokine receptor involved in breast cancer survival and metastasis (43), but retains high surface levels of the “don’t eat me signal” CD47 (44), implicated in the escape of CTC from systemic macrophage and neutrophil-mediated phagocytosis (44) (Supplementary Figure 3). Importantly, the expression levels of these surface molecules were not affected by ICAM-1 deletion (Supplementary Figure 3), confirming the high selectivity of our CRISPR-targeted deletion strategy.

We next tested if ICAM-1 expression by E0771 cells affects the intrinsic growth of these breast cancer cells. To that end, 2 × 10⁴ control or ICAM-1 KO E0771 cells were seeded in 6-well plates and their proliferation was assessed. The early growth of both cell types was identical, but when cells approached confluency, ICAM-1 null cells grew slightly faster (Figure 1A). Knocking out ICAM-1 in E0771 did not affect the susceptibility of these cells to apoptotic signals induced by the potent DNA damage and cell cycle arrest inducer etoposide (Figure 1B).

Tumor growth in mice can be influenced by numerous paracrine signals. The common practice of cancer cell implantation involves a large number of cells that results in massive cell death and local inflammation that could mask any intrinsic or extrinsic effects resulting from ICAM-1 deletion. For this reason, we implanted a very low number of control and ICAM-1 KO E0771 cells embedded in Matrigel (45) in the mammary fat pad of syngeneic C57BL/6 female mice (i.e., 1 × 10³ cells) and followed tumor growth 13, 17, and 21 days after implantation (Figure 1C). As both control and ICAM-1 KO E0771 cells gave rise to similar tumors at all tested time points, we concluded that ICAM-1 expression by E0771 is dispensable for primary tumor growth in vivo. Notably, the implanted control E0771 cells maintained their surface ICAM-1 expression in this in vivo tumor growth model (Figure 1D).

Since all major types of hematopoietic CD45+ cells express the ICAM-1 binding LFA-1 and Mac-1 integrins (46–48), we next tested if the composition of these cells in the E0771 tumor microenvironment (TME) is affected by the absence of E0771-expressed ICAM-1 (Figure 1E). Consistent with the normal in vivo growth of ICAM-1 KO E0771 inside the mammary fat pad, the number of DCs, B cells, and TAMs recovered from orthotopic E0771 tumors was not affected by ICAM-1 deficiency in the implanted E0771 cells (Figure 1E). Importantly, the composition of naïve T cells and the accumulation of endogenous effector CD4+ and CD8+ T cells in the TME, as well as their expression of the immune checkpoint protein PD-1, were not significantly affected by the loss of E0771 ICAM-1 (Supplementary Figure 4). Interestingly, however, the accumulation of DCs and B cells, but not of CD8+ T cells, in the tumor-draining LNs (TdLNs) surrounding these tumors, was slightly reduced (Figure 1F) while their numbers in the spleens were unaffected (Figure 1G). These results indicate that breast cancer–expressed ICAM-1 may facilitate the egress of tumor-associated DCs from the tumor via lymphatics to nearby LNs and may consequently also enhance B-cell entry into these LNs.

Tregs in the TME repress the ability of endogenous tumor-specific CTLs to eradicate primary tumors (49). To assess if this cytolytic activity of endogenous CTLs triggered by the elimination of Tregs requires the presence of ICAM-1 on E0771 tumor cells, we implanted primary E0771 tumors in syngeneic FOXP3-DTR-GFP mice that express a diphtheria toxin receptor selectively on FOXP3-positive Treg cells (50) and used a DTx to effectively deplete Tregs from the tumor-implanted mice (Figures 2A , B). When control or ICAM-1 KO E0771 cells were similarly implanted in the mammary fat pad of FOXP3-DTR-GFP recipients subjected to Treg depletion, both control and ICAM-1 KO E0771 tumors were eradicated to similar extents (Figure 2C). Thus, the presence of ICAM-1 on E0771 cells is not required for the CTL-mediated eradication of these cells in vivo triggered by Treg depletion. To further substantiate these results, we established another model of primary tumor killing by endogenous T cells. Instead of Treg depletion, we installed an E0771 tumor in a distal mammary fat pad (i.e., left mammary fat pad) 10 days before either control or ICAM-1 KO E0771 cells were implanted in the right mammary fat pad (Figure 2D). Notably, the presence of a distal primary E0771 tumor but not of an irrelevant distal tumor (e.g., B16 melanoma (16)) totally eliminated the growth of the second primary E0771 tumor but did so irrespectively of ICAM-1 expression by the second primary tumor (Figure 2E). Thus, the antitumor immune response experienced by E0771 cells inside a primary breast tumor elicited by E0771 cells implanted in a distal site does not depend on ICAM-1 expression on the tumor target. Notably, this anti-E0771 tumor killing by a distal E0771 tumor was FcR independent, and therefore ADCC or ADCP independent (51), since the distal E0771 tumor also efficiently eliminated the growth of the second primary E0771 tumor in FcR KO recipient mice (13) (Figure 2F). Taken together, these results suggest that the eradication of a primary E0771 breast tumor elicited by either Treg depletion or by a tumor vaccine is not facilitated by ICAM-1 expression on the breast cancer cell targets.

To further dissect the involvement of tumor ICAM-1 in an antitumor immune response elicited by antitumor CTLs, we next tested if monoclonal effector CTLs specific for the OVA neoantigen introduced into E0771 cells can eliminate these cancer cells in vitro and in vivo in an ICAM-1-facilitated manner. To this end, we used OT-1 CTLs specific for the immunodominant ovalbumin (OVA)-derived peptide SIINFEKL presented by the MHC-I H-2Kb haplotype (27) abundantly expressed by both control and ICAM-1 KO E0771 cells (Supplementary Figure 3). Control and ICAM-1 KO E0771 cells were stably transfected with OVA, and the expression of SIINFEKL-bound MHC-1 (pMHC) complexes was analyzed with a specific antibody that recognizes the SIINFEKL/H-2Kb complex (52). Importantly, both control and ICAM-1 KO E0771 cells expressed comparable MHC-I (Supplementary Figure 3) and a similar content of SIINFEKL/H-2Kb complexes (Figure 3A). The adhesive function of ICAM-1 on E0771 cells was also verified since the monolayers of ICAM-1 KO E0771 cells devoid of cognate antigen supported weaker adhesions of in vitro-generated OT-I CTLs compared to control ICAM-1 expressing E0771 monolayers (Figure 3B). Furthermore, ICAM-1 KO OVA-expressing E0771 cells were slightly less efficiently killed by OT-I CTLs in vitro (Figure 3C).

We next tested if i.v.-injected effector OT-I CTLs can reach the mammary fat pad implanted with either control or ICAM-1 KO OVA- expressing E0771 cells and kill these target cells in vivo (Figure 3D). The primary breast tumors of either control or ICAM-1 KO OVA-expressing E0771 cells were implanted in syngeneic mice; 3 days later, OT-I CTLs were i.v.-injected into these mice, and the tumor size was monitored from day 12 to day 16 postinitial tumor implantation. The growth of the control OVA-expressing tumor was completely blocked by injected OT-I CTLs (Figure 3E). Consistently with our in vitro observations, the elimination of ICAM-1 KO OVA-expressing E0771 tumors by i.v.- injected OT-I CTLs was comparable to that of control OVA-expressing E0771 tumors (Figure 3E). Thus, the presence of ICAM-1 on OVA-expressing E0771 cells is fully dispensable for their in vivo elimination by adoptively transferred Ag-specific monoclonal CTLs (i.e., OT-I) as well as by endogenous polyclonal CTLs (Figures 2C –E).

Advanced breast tumors consisting of cancer cells orthotopically implanted in syngeneic immunocompetent mice can give rise to spontaneous metastasis at various target organs. Since ICAM-1 deletion did not affect primary E0771 breast tumor growth, we next used a model of spontaneous metastasis established for E0771 (53) to test if the presence of ICAM-1 on orthotopically implanted E0771 cells affects their lung metastasis potential. To that end, we implanted large primary orthotopic tumors in recipient syngeneic mice, resected the tumors, and followed individual metastatic E0771 cells that have disseminated into the circulation either prior to or after resection and accumulated in the lungs. To accurately determine the number of accumulated E0771 cells and their precise location, we introduced into our control and ICAM-1 KO E0771 cell lines a red fluorescent protein (RFP) reporter. Importantly, as observed with small E0771 breast tumors (Figure 3A), the growth of the large tumors (Supplementary Figure 5A) and the main leukocyte composition of their TME (Supplementary Figure 5B) were unaffected by E0771 ICAM-1 expression. Nevertheless, we could not detect the circulating RFP+ tumor cells (CTCs) of either control or ICAM-1 KO E0771 cell tumors in the circulation following the resection of the respective primary breast tumors (data not shown). In contrast, another well-studied breast cancer line that metastasizes to lungs, 4T1, gave rise to spontaneous CTC generation in immunodeficient NSG mice as reported (39) (Supplementary Figure 6B), and all CTCs recovered from these mice expressed high levels of both ICAM-1 and the myeloid-specific chemokines we also found in E0771 cells (Supplementary Figures 3, 6, and Supplementary Table 1). Nevertheless, 4T1 and E0771 varied considerably with respect to the expression of another key integrin ligand, the VLA-4 ligand VCAM-1. Whereas E0771 expressed only negligible levels of VCAM-1, 4T1 cells expressed significant levels of this VLA-4 ligand (Supplementary Figures 6). Interestingly, although we could not detect circulating RFP+ E0771 CTCs in the systemic circulation of E0771 tumor-bearing C57BL/6 mice, these cells were readily detected in the lungs of all mice implanted with large primary breast tumors (Figures 4A, B). Nevertheless, and in agreement with a recent study that suggests that E0771 cells do not metastasize to bones (53), we could not detect E0771 metastasis to this organ or to the liver (data not shown). Notably, similar numbers of both control and KO RFP–expressing E0771 cells were recovered in the total lung suspensions of recipient mice 14 days after their orthotopic implantation (Figures 4A , B). At this time point, similar numbers of control and ICAM-1 KO RFP-expressing E0771 cells were also detected by the three-dimensional (3D) imaging of whole lung lobes using the light sheet microscopy (LSM) of lipid-cleared lungs (Figures 4C , D and Movie 1). These results suggested that although E0771 CTCs could not be detected in our tumor implanted mice, similar numbers of control and ICAM-1 KO E0771 were most likely released from their respective primary tumors and accumulated in the lungs. Interestingly, the vast majority of the RFP+ E0771 cells detected in the lungs at this early time point were intravascular irrespective of their ICAM-1 expression (Figure 4E and Movie 2). These data collectively suggested that E0771 cells are released from primary breast tumors, enter the circulation, and subsequently get entrapped in the pulmonary vasculature independently of their ICAM-1 expression.

Next, we followed the fate of these disseminated RFP+ E0771 cells recovered in recipient lungs 2 weeks after resection of the primary breast tumor (Figure 5A). Strikingly, at this time point, whereas the number of recovered control E0771 cells was reduced 15-fold from 2,150 ± 480 cells (Figure 4B, left bar) to 145 ± 80 cells (Figure 5B, left bar, p<0.0004), the number of ICAM-1 KO E0771 cells recovered from similar lung cell suspensions increased 20-fold, from 1,750 ± 600 (Figure 4B, left bar) to 35,000 ± 17,650 cells (Figure 5B, right bar, p<0.25). These results suggest that although the initial accumulation of ICAM-1 KO RFP+ E0771 cells inside the lung was normal, ICAM-1-deficient RFP+ E0771 cells were protected and readily expanded inside the lungs. We next repeated this line of experiments with non-RFP labeled E0771 cells and their ICAM-1 KO E0771 counterparts. Notably, and consistent with the FACS based results, resected ICAM-1 KO E0771 tumors gave rise to large metastatic nodules detected on the surface of the lungs while control E0771 did not (Figure 5C).

To further investigate the basis for the superior retention and survival of ICAM-1 KO RFP+ E0771 cells inside the recipient lungs, we next analyzed the distribution of these cells by the LSM of whole lung lobes. This analysis indicated that while control RFP+ E0771 cells remained singular (Figure 5D, left image), a fraction of ICAM-1 KO RFP+ E0771 cells gave rise to both oligoclusters and large macrocellular clusters (Figure 5D, right image and the two magnified rectangles, and Movie 3). Consistent with the FACS results (Figure 5B), both the total number of detectable RFP foci (i.e., spots of any dimension) and the fraction of macrocellular assemblies within these foci were significantly higher for ICAM-1 KO RFP-E0771 cells than for control RFP+ E0771 cells (Figure 5E). Strikingly, the LSM images indicated that the majority of ICAM-1 KO E0771 that survived and expanded in the lungs as singular cells or clusters remained inside the pulmonary vasculature, giving rise to intravascular metastasis (Figure 5F and Movie 4). These results collectively suggest that following the initial entry and accumulation of both control and ICAM-1 KO RFP+ E0771 cells inside the lung vasculature, ICAM-1-deficient E0771 cells, unlike their control counterparts, remain protected and can readily expand inside the pulmonary vasculature.

To further dissect the potential role of ICAM-1 in the suppression of the lung metastasis of intravascularly entrapped E0771 cells observed in the spontaneous metastasis model (Figure 5), we took a complementary approach, based on an experimental metastasis model of i.v.-injected E0771 cells. This model bypasses the steps involved in CTC dissemination from the primary tumor (i.e., intravasation into the circulation) but enables to directly investigate in defined time points and in a synchronized manner the potential communication of these cells with leukocytes circulating through the lung vasculature. Imaging control E0771 cells by LSM shortly after their i.v. injection into syngeneic C57BL/6 recipients indicated that all the cells reaching the lungs became entrapped inside CD31+ vessels primarily as singular cells (Figures 6A , B, and Movie 5) and could be detected inside these vessels even 14 days later (Figure 6B and Movie 6). Notably, we could not find any injected E0771 cells entering the lungs inside peribronchial blood vessels (Supplementary Figure 7 and Movie 7), readily detected by their constitutive expression of the integrin ligand VCAM-1 (54). These findings indicated a preference of E0771 cells to arrest inside pulmonary capillaries. Furthermore, consistent with the spontaneous E0771 lung metastasis model, when the lungs of recipient mice injected with either control or ICAM-1 KO E0771 cells were harvested 14 days after i.v. injection and analyzed by H&E histology (Figure 6C), the number of metastatic lesions generated by ICAM-1 KO E0771 cells significantly exceeded those generated by control E0771 cells (Figure 6D), and metastatic ICAM-1 KO cell clusters could be detected inside rather than outside lung capillaries at this time point (Figure 6E).

Neutrophils recovered from breast tumor–bearing mice [termed tumor-entrained neutrophils (TENs)] readily eliminate lung-entrapped breast cancer cells like 4T1 cells (15). To assess the effects of similar circulating TENs on the metastatic potential of either control or ICAM-1 KO E0771 cells in our experimental system, we designed a protocol for the i.v. injection of control E0771 cells to syngeneic recipient mice followed by a transfer of spleen cell suspensions from control E0771 breast tumor–bearing donor mice to the recipient mice (Figure 7A). The transfer of a whole spleen was designed to avoid purifying TENs generated in the donor mice (Figure 7A). As reported, the donor mice implanted with orthotopic E0771 tumors exhibited massive neutrophilia and splenomegaly (15), but their content of inflammatory monocytes was unaltered (Supplementary Figure 8). Notably, the ability of lung vessel–entrapped E0771 breast cancer cells to generate metastatic lesions 2 weeks after injection was essentially eliminated by the donor splenocytes (Figure 7B, first vs. second bar). We next repeated this procedure, but instead of transferring intact donor splenocytes from donor E0771 tumor- bearing mice, we transferred the donor spleen cell suspensions of tumor-bearing donor mice that have been in vivo depleted of their neutrophils (by at least 90%) or CD8 T cells (by at least 95%, Supplementary Figure 9). Notably, while neutrophil-depleted splenocytes isolated from E0771 tumor-bearing donor mice could not eliminate the metastatic lesions of the lung vessel–entrapped E0771, CD8 T-cell-depleted splenocytes isolated from the donor mice normally eliminated all the metastatic lesions (Figure 7B, first vs. third or fourth bars, Supplementary Figure 9). Thus, neutrophils derived from the spleens of E0771 tumor-bearing donor mice but not CTLs derived from similar donor spleens can readily eliminate the metastatic lesions of control E0771 cells entrapped inside lung vessels. Furthermore, the ability of these donor spleen neutrophils to eliminate E0771 lung metastasis was entirely dependent on the presence of ICAM-1 on the breast cancer cells (Figure 7B, second vs. fifth bar). Thus, neutrophils accumulated in the spleens of tumor-bearing donor mice can readily enter the lung vasculature and use ICAM-1 on E0771 cells entrapped inside this vasculature for the elimination of the breast cancer targets. Consistent with these in vivo results, TENs isolated from E0771 breast tumor-bearing mice could also kill control E0771 cells more readily than their ICAM-1-deficient E0771 cell counterparts in vitro (Figure 7C). Collectively, our results suggest that TENs use ICAM-1 expression on E0771 cells entrapped inside the lung vasculature to eliminate intravascular metastatic breast cancer cells before they can expand into metastatic lesions.