The Immunosuppressive Microenvironment in BRCA1-IRIS–Overexpressing TNBC Tumors Is Induced by Bidirectional Interaction with Tumor-Associated Macrophages

Evading immune system destruction is an effective strategy for cancer progression. This tolerogenic environment induced by factors released from tumor cells expands protumor myeloid cells, such as macrophages (1). Tumor-associated macrophages (TAM) can be polarized to antitumor, M1 macrophages that express metabolic enzymes, such as iNOS, cytokines, such as IL12, chemokines, such as CXCL10, and transcription factors, such as IRF5 (2), or protumor, M2 macrophages that express metabolic enzymes, such as ARG1, cytokines, such as IL10, chemokines, such as CCL17, and transcription factors, such as IRF4 (1, 2).

GM-CSF stimulates proliferation and survival of macrophages, neutrophils, eosinophils, dendritic cells (DC), and microglia (3). GM-CSF also affects mature hemopoietic cells through cytokines production, antigen presentation, and phagocytosis (4, 5). In mice and human, increased circulated GM-CSF level indicates late-stage tumors and/or tumor progression (6). At low level, GM-CSF exerts an antitumor effect by activating DC within the tumor, while at high level, GM-CSF activity is mainly immunosuppressive, enriching tumor microenvironment of M2 TAMs (7).

Increased expression of the CD47 on cancer cell transduces a “do not eat me” inhibitory signal through activating signal regulatory protein alpha (SIRPα) on macrophages (8). Preclinical models showed anti-CD47 therapies stimulate phagocytosis of cancer cells, in vitro, and antitumor immune responses, in vivo (8). In contrast, calreticulin expression at the cancer cell surface promotes phagocytosis of cancer cells by engaging the LRP1 receptor on macrophages (9).

The immunoinhibitory receptor, PD-1, is predominantly expressed on activated effector T cells (10), or antigen-specific T cells chronically exposed to the antigen (11). PD-1 ligands, PD-L1 are expressed on hematopoietic and cancer cells, and PD-L2 is expressed on macrophages and DCs (12). PD-L1 on aggressive tumor cells engagement of PD-1 on T cells leads to T-cell exhaustion (i.e., anergy) including inhibition of proliferation and cytokines, for example, IL2 and IFNγ production, and apoptosis, resulting in tumor cells evasion of recognition/destruction by the immune system (13, 14). Anti-PD-1 or anti-PD-L1 is now a successful cancer immunotherapy that stimulates the adaptive immune system to attack cancer cells (14).

BRCA1-IRIS (i.e., IRIS, in-frame reading of intron 11 splice variant) is an oncogene produced by the alternative usage of the BRCA1 locus (15). IRIS expression is high in breast cancers, especially triple-negative breast cancers (TNBC; ref. 16). Deliberate IRIS overexpression (IRISOE) in normal human mammary epithelial (HME) cells or luminal A/ER⁺ cells converts them into genuine TNBC cells expressing basal, epithelial-to-mesenchymal (EMT), and stemness markers, and lacking ERα and BRCA1 proteins expression, in vitro and in vivo (17). HME/IRISOE cells formed genuine TNBCs in SCID mice (16, 18), including the presence of a necrotic/hypoxic/inflamed cores within them (16, 19). We recently named this core “the aggressiveness niche” (19), and showed experimental evidence that IRISOE TNBC metastatic precursors develop within this aggressiveness niche (18, 20, 21).

Here, we show that GM-CSF secretion from IRISOE cells activates STAT5, NF-κB, and ERK signaling in TAMs to enhance their proliferation, recruitment, survival, M2 polarization, and expression/secretion of TGFβ1. Inhibiting GM-CSF signaling attenuated TAMs recruitment and M2 polarization and decreased the immunosuppressive ability of IRISOE cells, leading to significantly reduced aggressiveness and regression of IRISOE tumors through an activated adaptive immune response.

Mammary cell lines used here were from the ATCC and maintained as described previously (18). The doxycycline-inducible IRISOE cell lines (IRISOE1–5) generation and maintenance were described earlier (15). Orthotopic injection of these cell lines in SCID mice and maintaining mice on doxycycline-supplemented drinking water led to primary (1°) orthotopic IRISOE mammary tumors. Few of these tumors used to develop 1° orthotopic IRISOE breast cancer cell lines “IRIS291, IRIS292, and IRIS293” were used in the experiments described herein (16, 18). These cell lines were maintained in doxycycline-supplemented RPMI1640 medium containing 10% FBS. IRIS291 and IRIS293 clones expressing red fluorescent protein (RFP) were developed by a viral infection of a mCherry-expressing cDNA and antibiotic selection. Human THP1s from ATCC were maintained in PRMI/10% FBS in our laboratory. Authentication of all commercial and in-house cell lines was done using short tandem repeat profiling and also tested for Mycoplasma contamination.

Mouse monoclonal (mono) anti-human IRIS and rabbit polyclonal (poly) anti-mouse Iris antibodies were developed in our laboratory. Mouse anti-human GM-CSF (R&D Systems, catalog no. 3209), Rabbit (Rb) poly-anti-mouse GM-CSF (ab9741), mouse mono-anti-human HIF1α (Novus Biological, NB100–105), Rb poly-anti-human p-p65/NF-κB (Cell Signaling Technology, catalog no. 3033), Rb poly-anti-human γ-tubulin (Abcam, ab11321), Rb mono-anti-human Actin (Cell Signaling Technology, 13E5), Rat mono-anti-human GM-CSF Rα (R&D Systems, catalog no. 698423), mouse mono-anti-human and -mouse TGF-β1 (R&D Systems, catalog no. 9016), Rb poly-anti-human TβRII (Y⁴²⁴, sc-17007-R), Rb poly-anti-human cyclin D1 (Thermo Fisher Scientific, catalog no. RB-010-P0), Rb poly-anti-human p-SMAD2 (S⁴⁶⁵, S⁴⁶⁷, Thermo Fisher Scientific, 44–244G), Rb mono-anti-human p-SMAD3 (S⁴²³, S⁴²⁵, Cell Signaling Technology, C25A9), mouse mono-anti-mouse and -human TGFβ1 (Bio-Cell, 704719J1), FITC-conjugated rat mono-anti-human- and mouse-CD11b and isotype control (Cell Signaling Technology, catalog nos. 24442 and 56722, respectively), F4/80 Rat mono-anti-mouse F4/80 (Abcam, ab6640), PE-conjugated mono-mouse anti-human CD206 and isotype control (Invitrogen, 12–2069, P3.6.2.8.1, respectively), Rat mono-anti-mouse CD25 (Abcam, ab210333), Rb mono-anti-human and -mouse FOXP3 (Abcam, ab215206), Rat mono-anti-mouse CD8 (BioLegend, 100702), Rb poly- anti-human Calreticulin (R&D Systems, catalog no. 681233), Rb mono-anti-human PD-L1 (Abcam, ab251611).

Human recombinant GM-CSF was obtained from R&D Systems (7954-GM), reconstituted in sterile PBS to a stock concentration of 100 μg/mL, and used at 50 ng/mL. Human recombinant EGF was obtained from Cell Signaling Technology (catalog no. 8916), reconstituted in sterile PBS to a stock solution of 100 μg/mL, and used at 10 ng/mL. Human recombinant TGFβ1 was obtained from R&D Systems (7754-BH), reconstituted in 4 mmol/L HCl to a stock concentration of 100 μg/mL, and used at 5 ng/mL. Vehicles for these recombinant proteins were PBS and/or 0.2 μmol/L HCl.

The NF-κB inhibitor JSH-23 from Sigma (catalog no. J4455, 10 μmol/L working concentration), the STAT5 inhibitor IQDMA from Abcam (ab141192, 10 μmol/L working concentration), the NF-κB inhibitor Bay 11–7082 from Sigma (B5556; 10 μmol/L working concentration), the ERK inhibitor PD98059 from Calbiochem (catalog no. 51300; 10 μmol/L working concentration), the PI3′K inhibitor LY294002 from Calbiochem (catalog no. 440202; 10 μmol/L working concentration), the SMAD3 inhibitor SIS3 from Calbiochem (catalog no. 566405; 10 μmol/L working concentration), and the TβRI/II inhibitor LY2109761 from Santa Cruz Biotechnology (700874-71-1; 5 μmol/L working concentration).

Conditioned media (CM) from equal numbers from IRISOE cell lines (IRIS1–IRIS5) grown in the absence (i.e., naïve HME) or the presence (i.e., HME/IRIS) of doxycycline plated in serum-free medium for 20 hours used to screen for cytokine, chemokine, and growth factors levels on antibody arrays (RayBio). Assay performed according to the manufacturer's instructions and as described previously (22) repeated on three different preparations (i.e., n = 3 for each cell line in the presence or absence of doxycycline).

Naïve HME, IRIS291, and IRIS293 cells seeded at a density of 3 × 10⁵ cells/well in a 6-well plate for 16–18 hours transiently transfected with siLuc siRNA, siHIF1α (using two different siRNA), siGM-CSF, siTGF-β1, siSMAD5, sip65, siERK, or siIRIS using Xfect Transfection Reagent (Clontech) according to the manufacturer's instructions. After 48 hours, media was changed, and cells exposed to normoxia (20% O₂) or hypoxia [1% O₂, using hypoxia chamber (STEMCELL Technologies, catalog no. 27310)] for an additional 24 hours, at which time, CM collected for ELISA analysis and total proteins for Western blot. At least n = 3 for each assay was performed.

Briefly, CM from normoxic or hypoxic (24 hours) naïve HME, IRIS291, IRIS292, or IRIS293 directly analyzed for secreted factors using ELISA or receptors expression using membrane preparations. The CM following certain siRNA transfection or hypoxia was added onto THP1 cells. Secretion from THP1 was assessed using ELISA 24 hours later, and receptor expression on the surface was determined using Western blot analysis of membrane extracts. At all steps, equal numbers of each cell type were seeded to avoid discrepancies due to cell number variations. At various steps in this protocol, specific NeuAb and/or drugs were added, as indicated in the Results section.

Cocultures CM or mice sera diluted in carbonate coating buffer (pH 9.6) was used to coat 96-well ELISA plates overnight at 4°C. Plates were then washed three times with phosphate-buffered saline - 0.05% Tween-20 (PBS-T) and blocked with 2% BSA for 1 hour at room temperature were incubated with primary antibody diluted in blocking solution for 2 hours at room temperature. After washing primary antibodies, horseradish peroxidase (HRP)-conjugated secondary antibodies were added for 1 hour at room temperature, and the reaction was determined using Western Lightning Plus-ECL (PerkinElmer) as a substrate. All experiments were done in triplicates, performed three separate times, and shown as mean ± SD.

Boyden chambers (BD Biosciences) of 8-μm pore size (for migration) or 0.4-μm pore size (for secretome) analysis were used to layer IRISOE cells in the bottom chamber with or without neutralizing antibodies and the THP1 cells in the transwell inserts. THP1s migrated to the lower face of these inserts and were counted and plotted 24–72 hours later. Cocultures were performed under normoxic or hypoxic conditions, in triplicates three separate times.

Western blot analysis was performed as described previously (22). Protein lysates were prepared from cell sonication (in PBS containing protease and phosphatase inhibitor, Thermo Fisher Scientific) or membrane fraction. Protein concentrations were estimated using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) and 25 μg of proteins was denatured in NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) were resolved on NuPAGE Gels (Thermo Fisher Scientific). Proteins were electro-transferred to polyvinylidene difluoride membranes that were blocked with 5% dry milk for 1 hour, washed five times (10 minutes/each) with PBS-T. Incubation with primary antibody was always done overnight at 4°C when membranes were again washed five times (10 min/each) with PBS-T and exposed to suitable HRP-conjugated secondary antibodies for 1 hour at room temperature. Washed membranes were developed using Western Lightning Plus-ECL as a substrate. Tubulin or actin was used as an internal loading control. All Western blots were done at least three separate times.

IHC analysis as described previously (22), was done on 4-μm thick paraffin-embedded sections of tumor tissue excised from IrisOE syngeneic mammary tumors generated in BALB/c mice. After deparaffinization and rehydration, antigen retrieval for Iris was performed using pepsin (10 μmol/L) for 20 minutes at 37°C, whereas for all other antigens by boiling the slides in citrate buffer (pH 6.0) for 10 minutes in a microwave. Cooled slides were washed in PBS (3 × 15 minutes each), incubated in 3% hydrogen peroxide (H₂O₂) for 10 minutes to block endogenous peroxidase activity unless fluorescence analyses were performed. Following washing, slides were blocked with 10% normal goat serum for 1 hour at room temperature and probed with primary antibodies overnight at 4°C in a moist chamber. Slides were then incubated with suitable HRP, or Alexa Fluor–conjugated secondary antibodies for 1 hour at room temperature (depending on the analysis). After washing, HRP-conjugated secondary antibody–treated slides were developed using Vector DAB Substrate Kit (Vector Laboratories) and counterstained with Meyer hematoxylin (Thermo Fisher Scientific) for 2 minutes, washed, dehydrated, and mounted with Permount (Thermo Fisher Scientific). Alternatively, slides that were stained with Alexa Fluor–conjugated secondary antibody were counterstained and mounted with VECTASHIELD mounting medium for fluorescence with DAPI (Vector Laboratories) and were imaged under the microscope. All experiments were done on at least five tumors from phenotypes or treatments.

One-hundred nanogram of total RNA was processed for qRT-PCR using iScript One-Step RT-PCR Kit with SYBR Green (Bio-Rad) using EGFR mRNA, forward: 5′-CCAGGACCCCCACAGCACTGCAGTGGGCAA-3′, reverse: 5′-GTGGGTCTAAGAGCTAATGCGGGCATGGCT-3′; CK5 mRNA, forward: 5′-GCGGTTCCTGGAGCAGCAGAACAAGGTTCT-3′, reverse: 5′-CTGAGGTGTCAGAGACATGCGTCTGCATCT-3′; CK17 mRNA, forward: 5′-CTGGCTGCTGATGACTTCCGCACCAAGTTT-3′, reverse: 5′-CGCAGTAGCGGTTCTCTGTCTCCGCCAGGT-3′; CDH2 mRNA, forward: 5′-ACAGTGGCCACCTACAAAGG-3′, reverse: 5′-CCGAGATGGGGTTGATAATG-3′; Twist mRNA: forward: 5′-GGAGTCCGCAGTCTTACGAG-3′, reverse: 5′-TCTGGAGGACCTGGTAGAGG-3′; SnailI mRNA, forward: 5′-CCTCCCTGTCAGATGAGGAC-3′, reverse: 5′-CCAGGCTGAGGTATTCCTTG-3′; Sox2 mRNA, forward: 5′-TTCATCGACGAGGCTAAGCGGCTG-3′, reverse: 5′-AGCTGCCGTTGCTCCAGCCGTTCA-3′; Oct4 mRNA, forward: 5′-ACATGTGTAAGCTGCGGCC-3′, reverse: 5′-GTTGTGCATAGTCGCTGCTTG-3′; and Nanog mRNA, forward: 5′-ATGCCTCACACGGAGACTGT-3′, reverse: 5′-AGGGCTGTCCTGAATAAGCA-3′. The following primers for β-actin mRNA (internal control), forward: 5′-ACAGAGCCTCGCCTTTGC-3′, reverse: 5′-GCGGCGATATCATCATCC-3′ were used to normalize with expression and data were correlated to the calibrator sample.

The qPCR reactions were performed in a final volume of 25 μL. The mean C_t was calculated for each sample and used to determine the ΔC_t for this sample as follows: ΔC_T = C_t for the gene of interest − C_t of the internal control gene (β-actin). Then the ΔΔC_T was calculated as follows: ΔΔC_T = [(C_t for the gene of interest − C_t of the internal control gene, β-actin) for sample A − (C_t for the gene of interest − C_t of the internal control gene (β-actin) for sample B], where sample B is the calibrator. For the statistical analysis, the ΔΔC_T and not the raw C_t data were used (23). All were done in triplicates, performed three separate times.

Single-cell suspensions from the in vivo–treated tumors were processed for FACS analysis. One million cells were stained with specific fluorescently labeled antibodies on ice for 1 hour. Cells were then washed three times with FACS buffer (1% BSA in PBS) by centrifugation at 2,000 rpm at 4°C for 10 minutes, and further incubated with FITC-conjugated secondary antibody for 30 minutes in ice. Cells were washed and then analyzed for surface staining on Gallios Flow Cytometer (Beckman Coulter). For sorting, Moflo XDP Cell Sorter (Beckman Coulter) was used. The data were analyzed with Kaluza Flow Cytometry Analysis Software v 1.2. Each FACS was performed three separate times in triplicates with identical results.

All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Mississippi Medical Center and in accordance with the NIH guidelines. Athymic or BALB/c (6–8 weeks old) female mice (numbers are indicated in Results) were injected in the second left thoracic mammary fat pad with orthotopic 1° IRISOE mammary tumor cell lines: IRIS291 or IRIS293 (also RFP-expressing variants), or the human IRISOE/TNBC cell lines, MDA-MB-231 (hereafter, MDA231) or MDA-MB-468 (hereafter, MDA468) cell lines (17, 19–21) expressing shCtrl or shIRIS, or the mouse TNBC cell line, 4T1 expressing shCtrl or shIris1 or shIris2, admixed or not with human THP1-macrophages (at 10:1 ratio). After IRISOE tumors reached a tumor volume indicated in figures and text, randomized mice were divided into groups that were injected intratumorally with vehicle or IRISpep four times every third day for 2 weeks. Tumor formation or death was monitored and recorded, as indicated in text and figures. A digital caliper was used to measure tumors every third day, and volume was calculated using the formula: volume = (length × width²)/2. Endpoints were death or sacrifice of mice and a collection of tumors and peripheral blood. Excised tumors were divided into several portions, flash-frozen to generate DNAs, RNAs, and proteins, and prepare single cells as described in (24) for FACS analysis as indicated, or paraffin embedding, sectioning (at 4 μm), and IHC staining as described above.

Kaplan–Meier plotter was used to delineate the association between IRIS gene expression and disease-free survival (DFS) or overall survival (OS). Within each cohort, high expresser and low expresser patients were analyzed and compared for their OS and DFS, respectively. Breast tissue microarrays comprised of normal, DCIS, and invasive and metastatic samples were purchased from US Biomax, Inc. IHC protocols were described earlier (16). A semiquantitative scoring system was used to identify the percentage of tumor cells showing positive staining (25). Scoring represents the overall stain intensity and percentage of cancer cells stained in four high-magnification fields of each sample. Average overall staining intensity (26) was valued as percentage of cell stained/field: zero (<1% staining) was considered very weakly stained, 1 (1%–10% staining) was considered weakly stained, 2 (10%–25% staining) was considered medium stained, 3 (25%–50% staining) was considered strongly stained, and 4 (>50% staining) was considered very strongly stained. The positive staining scoring method is subjective, and artifacts such as high background or variable stain deposition can skew the results, and the scores for the two categories remain as separate functions and cannot be combined for analysis and comparison (27).

The study also involves a cohort of 96 patients with breast cancer with primary invasive ductal carcinomas recently diagnosed and treated at the NCI (Bethesda, MD) and Cairo University, (Cairo, Egypt) between September 2009 and October 2012, were included in the study as recently reported by us (28). None of the patients showed metastasis at the time of the initial diagnosis. Expression of estrogen receptor, progesterone receptor, and HER2/neu was assessed in all tumor samples. On the basis of this analysis 43 of the patients were negative for all three markers and thus considered TNBCs (mean age of 51.91 ± 12.34 SD, range: 30–78 years) and 53 showed expression of some/all of the markers and thus were considered non-TNBCs (mean age of 52.77 ± 12.13 SD, range: 27–81 years). Twenty normal breast tissue samples obtained from reduction mammoplasty (mean age 35 ± 13.94 SD; range, 22–64 years) were included as controls in the study. World Health Organization classification of breast tumors was used to grade the tumors, and American Joint Committee on Cancer's Staging Manual (7th edition) was used to stage the tumors (29). All participants signed written informed consent before enrollment in the study that was approved by the Institutional Review Board of the NCI (Cairo, Egypt), according to the 2011 Helsinki Declaration, CIOMS, Belmont Report, and U.S. Common Rule. Patients enrolled in the study were ≥18 years old, and none of the patients were pregnant or breast feeding, and had an active second malignancy; were involved in another clinical trial were excluded from the study. The median follow-up period was 33 months.

Statistical analysis was performed using unpaired, two-tailed Student t test. In all figures, data represents the mean from at least three separate biological repeats done in at least triplicates each ± SD; *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

IHC analysis of paraffin-embedded breast cancer TMAs (n = 511, Biomax.us) with a monoclonal anti-IRIS antibody (16) showed low expression in normal/near cancer tissues (n = 66, Supplementary Fig. S1A₁), increased in the DCIS tumors (n = 180, Supplementary Fig. S1A₂), increased further in invasive tumors (n = 100, Supplementary Fig. S1A₃) to reach the highest level in metastatic tumors (n = 165, Supplementary Fig. S1A₄).

IHC staining of another TMA (n = 326, all subtypes, Hawaiian SEER collection) with anti-IRIS, -AKT1, -AKT2, or -p-AKT specific antibodies (16) showed 84% (274/326) of the tumors were IRISOE (i.e., express ≥2-fold over normal tissues). Among those, 64% (176/274, P = 0.012), 65% (180/274, P = 0.006), 68% (188/274, P = 0.030) stained positive for AKT1, AKT2, and p-AKT, respectively (Supplementary Fig. S1B, left; ref. 16). In a TNBC sub-cohort (n = 72), 88% (63/72) were IRISOE tumors (16). Spearman correlation coefficient (r) analysis showed r = 0.748 (P = 0.00043) for IRIS and AKT1 or AKT2 overexpression, and r = 0.834 (P = 0.0038) for IRIS and p-AKT overexpression (Supplementary Fig. S1B, right).

We reassessed an Egyptian breast cancer cohort (n = 96) that contained 45% patients with TNBC and 55% patients with non-TNBC (28). Among the TNBC group, 65% (28/43), while 28% (15/53) among the non-TNBCs group were IRISOE patients (28). Only in the TNBC group, DFS (P = 0.05) and OS (P = 0.04) were significantly different between IRISOE and IRIS^low patients (compare right with left, Supplementary Fig. S1C). Together, these data suggest that in TNBCs, the increase in IRIS and AKT1/2 expression/activation correlates with enhanced breast cancer aggressiveness, and reduced DFS and OS.

Reanalysis of our recent antibody array screening for IRISOE cells secretome showed that compared with HME cells, CM of IRISOE cells contained >18-fold higher GM-CSF (CSF2, Fig. 1A), while same levels CSF3 (Fig. 1A). Recently, we proposed that the necrotic/hypoxic/inflamed core within an IRISOE TNBC tumor represents an “aggressiveness niche” (19), where IRISOE TNBC metastatic precursors are developed through bidirectional interactions between IRISOE TNBC cells and activated stromal cells, for example, MSCs and TAMs (20, 21).

To evaluate the aggressiveness niche microenvironment role in GM-CSF production and secretion, siLuc- or siHIF1α–transfected HME, IRIS291, and IRIS293 cells (48 hours) were exposed to normoxia (20% O₂) or hypoxia (1% O₂, for 24 hours). First, as we recently reported even under normoxic conditions, IRISOE cells maintain high levels of HIF1α (Fig. 1B). Second, ELISA analysis showed that normoxic siLuc-transfected HME cells secrete low levels of GM-CSF (Fig. 1C). Hypoxia and HIF1α silencing did not alter that (Fig. 1C). In contrast, normoxic siLuc-transfected IRIS291 or IRIS293 cells secreted approximately 5-fold higher GM-CSF that increased under hypoxia to 15- to 20-fold (Fig. 1C). HIF1α silencing (Fig. 1B) in IRISOE cells decreased GM-CSF secretion in normoxic and hypoxic cells (Fig. 1C). It will be essential to assess the role of HIF2α (30) in this process in the near future.

Inflammation activates NF-κB to promote the expression of cytokines (31). Compared with vehicle-treated HME, IRIS291, and IRIS293 cells, those treated with 10 μmol/L JSH-23 (inhibits NF-κB transcriptional activity; ref. 32) showed significantly lower p-p65/NF-κB level (Fig. 1D). This led to significantly decreased GM-CSF secretion from IRIS291 and IRIS293 cells according to ELISA analysis (Fig. 1E).

Phorbol 12-myristate-13-acetate (PMA) treatment triggers differentiation of monocytic THP1 cells into none-polarized (M0) macrophages. We layered PMA-untreated or PMA-treated THP1s in 8-μm inserts of Boyden chambers containing HME, IRIS291, or IRIS293 cells CM in the bottom well (media changed daily). Compared with HME CM, IRIS291 or IRIS293 cells CM recruited 400%–600% more PMA-untreated THP1s (72 hours later) that was blocked by GM-CSF neutralizing antibody (NeuAb, Fig. 1F). Moreover, HME CM recruited very low number of PMA-treated THP1s (24 hours later) that was not affected by GM-CSF NeuAb (compare Fig. 1G with H). In contrast, IRIS291 CM recruited significantly higher numbers of PMA-treated THP1 (compare Fig. 1I with G) that was blocked by GM-CSF NeuAb (compare Fig. 1J with I), Similarly, IRIS293 CM recruited significantly higher numbers of PMA-treated THP1 (compare Fig. 1K with G) that was also blocked by GM-CSF NeuAb (compare Fig. 1L with K). Quantitative analysis of these effects is presented in (Fig. 1M). Interestingly, compared with HME cells, IRISOE cells expressed approximately 3-fold higher of the GM-CSF receptor, CSF2Rα (Fig. 1N). Compared with M1 polarized THP1s, M2 polarized expressed >15-fold higher of CSF2Rα (Fig. 1N).

Taken together, these data suggest that: (i) in the aggressiveness niche, IRISOE stabilizes HIF1α (even under normoxia) (21, 33) and activates NF-κB (17, 22) to trigger GM-CSF expression/secretion from TNBC cells; and (ii) because CSF2Rα is the only high-affinity receptor, GM-CSF recruits M2 more efficiently than M1-polarized macrophages to IRISOE cells (Fig. 1O).

ELISA analysis of CM from siLuc- or siGM-CSF–transfected HME, IRIS291, or IRIS293 cells (48 hours) exposed to normoxia or hypoxia (additional 24 hours) showed that untransfected and untreated THP1s or those exposed to normoxic siLuc- or siCSF2-transfected HME cells CM secrete low levels of TGFβ1 (Fig. 2A). Hypoxic siLuc-transfected HME cells CM triggered a 2-fold higher TGFβ1 secretion from THP1s that was blocked by GM-CSF silencing (Fig. 2A). In contrast, normoxic siLuc-transfected IRIS291 or IRIS293 cells CM induced approximately 5-fold higher TGFβ1 secretion from THP1s that was increased to >10-fold by the hypoxic CM (Fig. 2A). In both situations, GM-CSF silencing blocked the ability of the CM to induce TGFβ1 secretion from THP1s (Fig. 2A).

THP1 cells silenced from SMAD5, p65/NF-κB, or ERK (48 hours) were exposed to 50 ng/mL of recombinant GM-CSF for 24 hours, or they were exposed to 50 ng/mL of recombinant GM-CSF in the presence of the vehicle; the STAT5 inhibitor, IQDMA; the NF-κB inhibitor, Bay 11–7082; or the ERK inhibitor, PD98059 (24 hours). ELISA analysis showed recombinant GM-CSF induced >2-fold increase in TGFβ1 secretion by THP1s (vehicle, Fig. 2B). Decreasing expression or activity of SMAD5, p65/NF-κB, or ERK (Fig. 2B, inset) blockade recombinant GM-CSF–induced TGFβ1 secretion from THP1s (Fig. 2B).

Compared with HME cells, IRIS291 and IRIS293 express 10- to 20-fold higher of TβRII^Y424 (constitutively active by autophosphorylation on tumor cells; ref. 34) correlated with a 5- to 10-fold higher cyclin D1 expression (Fig. 2C). We transfected the TNBC cell line MDA-MB-468 that overexpresses IRIS (hereafter MDA468; ref. 17) with siLuc or siIRIS (72 hours, Fig. 2D). In parallel, we grew HME and IRIS293 cells (24 hours). IRIS-silencing decreased TβRII^Y424 expression >70%, and cyclin D1 expression >60%, despite increasing p-SMAD2^S465,S467 level by approximately 3-fold (Fig. 2D). Conversely, IRISOE triggered approximately 9-fold increase in TβRII^Y424 and 5-fold increase in cyclin D1 expression, despite decreasing p-SMAD2^S465,S467 level by >90% (Fig. 2D).

MDA468/shCtrl, MDA468/shIRIS, HME, or IRIS293 cells were all treated with vehicles, 10 ng/mL recombinant EGF, or 5 ng/mL recombinant TGFβ1 (24 hours). Nuclear/chromatin extracts were probed for IRIS and p-SMAD2^S465,S467 levels. Recombinant EGF treatment did not affect IRIS expression in any of the four cell lines (Fig. 2E), while decreased p-SMAD2^S465,S467 level by approximately 60% in MDA468/shCtrl, and by approximately 70% in IRIS293 cells (Fig. 2E). Similarly, recombinant TGFβ1 treatment had no effect on IRIS expression in any of the four cell lines (Fig. 2E), decreased p-SMAD2^S465,S467 level by approximately 60% in MDA468/shCtrl cells and by approximately 50% in IRIS293 cells, and increased it by 2-fold in MDA468/shIRIS cells and HME cells (Fig. 2E). Importantly, IRIS-silencing increased p-SMAD2^S465,S467 level approximately 4-fold, approximately 8-fold, and approximately 12-fold under vehicle, recombinant EGF, or recombinant TGF-β1 treatments, respectively, in MDA468 cells (Fig. 2E), and overexpression decreased it by 50% under vehicle, recombinant EGF, or recombinant TGF-β1 treatment in HME cells (Fig. 2E). Identical results were obtained using MDA231 and IRIS291 cells.

In total extracts, compared with vehicle, treatment with 5 ng/mL of recombinant TGFβ1 (48 hours) had no effect on IRIS or AKT^T308/S473 levels, while induced 10-fold increase in p-SMAD3^S423/S425 level (Fig. 2F, black) in HME cells. Compared with vehicle, treatment with 5 ng/mL of recombinant TGFβ1 (48 hours) had no effect on IRIS level, increased p-AKT^T308/S473 level by 5-fold, while decreased p-SMAD3^S423/S425 level by 50% in MDA231 and MDA468 cells (Fig. 2F, black). Moreover, compared with HME cells, the effect of recombinant TGFβ1 (48 hours) translates to no effect on IRIS level, approximately 15-fold increase in p-AKT^T308/S473 level, and approximately 95% decrease in p-SMAD3^S423/S425 level in MDA231 and MDA468 cells (Fig. 2F, red).

IRIS291 and IRIS293 cells starved from serum (48 hours) were exposed to 5 ng/mL of recombinant TGFβ1 plus 10 μmol/L of different inhibitors (Fig. 2G). In chromatin fraction, under basal condition, IRIS291 and IRIS293 showed low p-SMAD3^S423,S425 level, and stimulation with recombinant TGFβ1 increased that by approximately 2-fold in both cell lines (Fig. 2G). As expected, SMAD3 inhibitor (SMAD3i) decreased recombinant TGFβ1-induced p-SMAD3^S423,S425 by >3-fold in both cell lines (from 2- to 0.6-fold, Fig. 2G). In contrast, inhibiting TβRI/II signaling using LY2109761 or PI3′K/AKT signaling using LY294002 increased p-SMAD3^S423,S425 level by >5-fold and >3-fold in both cell lines, respectively (Fig. 2G).

We treated MDA468 cells in 2D cultures (24 hours) or in 3D-Matrigel (7 days, media/factor/drugs changed every 3rd day) with none or 5 ng/mL recombinant TGFβ1 plus vehicle, or 2 μmol/L of IRIS inhibitory peptide (hereafter IRISpep; refs. 17, 18, 20–22), 10 μmol/L of erlotinib, LY294002, or SMAD3i (24 hours). In the 3D-Matrigel cultures, recombinant TGFβ1 induced IRIS291 cells to form larger, unorganized, and invasive acini (compare Fig. 2H, 1 and 2; ref. 17). Interestingly, while SMAD3i had no effect (Fig. 2H, 3), IRISpep (Fig. 2H, 4), erlotinib (Fig. 2H, 5), and LY294002 (Fig. 2H, 6) all blocked these recombinant TGFβ1-induced effects. This was also consistent with the real -time qRT-PCR on total RNAs from the 2D cultures that showed recombinant TGFβ1 increased expression of basal; EGFR, CK5, CD17, and EMT; CDH2, Twist, SnailI, and stemness; and Sox2, Oct4, and Nanog markers (Fig. 2H, 7). IRISpep, erlotinib, and LY294002 blocked recombinant TGFβ1-induced expression of these proteins, whereas SMAD3i had no effect (Fig. 2H, 7).

Taken together, these data suggest that: (i) GM-CSF secreted from IRISOE cells triggers TGFβ1 production and secretion from TAMs by activating CSF2Rα-induced SMDA5, p65/NF-κB, and/or ERK signaling; and (ii) in combination with (Supplementary Fig. S1B) and our previous studies (16, 17, 18, 20–22), IRISOE-induced EGFR and TβRII expression allows EGF (secreted by IRISOE cells; ref. 28) and TGFβ1 [secreted by TAMs (this study)] to redirect canonical-SMAD signaling that promotes growth suppression, apoptosis, and genomic stability (34) into noncanonical-AKT signaling that promotes aggressiveness in IRISOE cells (Fig. 2I).

To establish the role of TAMs in promoting IRISOE tumor aggressiveness, in vivo, we orthotopically injected 2 × 10⁶ RFP-IRIS291 or RFP-IRIS293 cells alone or admixed with 2 × 10⁵ THP1s (n = 10/each, Fig. 3A). Compared with alone, THP1 coinjection increased IRIS291 tumor size from 0.9 ± 0.2 cm³ to 1.8 ± 0.05 cm³ (2-fold increase), and IRIS293 tumor size from 0.7 ± 0.1 cm³ to 1.7 ± 0.08 cm³ (∼3-fold increase) within 10 weeks (Fig. 3B). Ex vivo analysis showed increased lung (Fig. 3C, top arrowheads) and brain (Fig. 3C, bottom arrowheads) metastases in mice coinjected with THP1s. THP1 coinjection also led to the death of 7 (IRIS291) and 6 (IRIS293) of the 10 mice (Fig. 3D).

To further establish this, 4 × 10⁶ MDA231 cells or MDA468 cells expressing shCtrl or shIRIS (Fig. 3E, inset) were injected in female athymic mice (see Fig. 3E). Forty days later, tumors and peripheral blood were collected from all mice (Fig. 3E). At day 40, shIRIS-expressing tumors were approximately 70% (MDA231, 0.38 ± 0.07 cm³ vs. 1.2 ± 0.15 cm³; Fig. 3F) and approximately 90% (MDA468, 0.15 ± 0.03 cm³ vs. 1.3 ± 0.07 cm³; Fig. 3G) smaller than shCtrl-expressing tumors. In addition, only MDA231/shCtrl and MDA468/shCtrl tumors led to the death of 50% of the mice within the 40 days (Fig. 3G, left arrows).

Moreover, unpermeabilized single-cell preparations generated from these tumors were colabeled with anti-mouse CD11b (monocytes/macrophage marker) and CSF2R antibodies. We gated forward scatter (as a relative size of the cells) and side scatter (as a measure of the granularity and complexity of the cells) to remove debris and conjugates, and normalized to isotype controls (in all subsequent experiments as well). This analysis revealed that compared with controls, CD11b⁻ cells increased in MDA231/shIRIS tumors (73.8% ± 3.7% vs. 56.2% ± 4.7%, P = 0.0002) and MDA468/shIRIS tumors (74.3% ± 4.5% vs. 55.6% ± 2.7%, P > 0.00005), while CD11b⁺ cells decreased in MDA231/shIRIS tumors (26.2% ± 3.7% vs. 43.8% ± 4.7%, P = 0.0002) and MDA468/shIRIS tumors (25.7% ± 4.5% vs. 44.4% ± 2.7%, P > 0.00005; Fig. 3H). Importantly, compared with control tumors, CD11b⁻/CSF2R⁺ cells increased in MDA231/shIRIS tumors (13.3% ± 1.1% vs. 7.8% ± 0.6%, P < 0.000001) and MDA468/shIRIS tumors (13.9% ± 1.7% vs. 6.9% ± 0.6%, P < 0.000001) and CD11b⁺/CSF2R⁺ cells decreased in MDA231/shIRIS tumors (13.6% ± 2.8% vs. 24.1% ± 2.4%, P = 0.0002) and MDA468/shIRIS tumors (13.2% ± 2.9% vs. 23.8% ± 2.4%, P > 0.0003, Fig. 3H).

Real-time qRT-PCR analysis performed on total RNA isolated from these tumors showed that compared with MDA231/shCtrl tumors, MDA231/shIRIS tumors contained 8-, 11-, and 5-fold higher HLA-DR, IL12, and iNOS (M1 markers), respectively, while >80% reduced expression of CD206, CCL17, and Arg1 (M2 markers, Fig. 3I, yellow). Similarly, compared with MDA468/shCtrl tumors, MDA468/shIRIS tumors contained 6-, 10- and 9-fold higher of HLA-DR, IL-12, and iNOS, respectively, while >90% reduction in CD206, CCL17, and ARG1 levels (Fig. 3I, red). Finally, ELISA analysis showed that compared with naïve mice (n = 40, Fig. 3J, top/white), MDA231/shCtrl or MDA468/shCtrl tumor–bearing mice (n = 10/each) showed 12- to 15-fold higher, while MDA231/shIRIS or MDA468/shIRIS tumors–bearing mice (n = 10/each) showed only 3- to 5-fold higher circulating GM-CSF and TGFβ1 (>60% decrease, Fig. 3J).

Taken together, these data suggest that: (i) the bidirectional interaction with TAMs enhances IRISOE cells' aggressiveness, in vivo; and (ii) IRIS silencing decreases TNBC tumors size, circulating GM-CSF, and TGFβ1 levels, and the number of CSF2R⁺ macrophages recruited into these tumors (Fig. 3K).

The BALB/c TNBC tumor cell line, 4T1 (35), overexpress Iris (mouse IRIS homolog; ref. 36) when compared with normal mouse HC11 cell line. We developed two short hairpin RNAs, shIris1 and shIris2 that target Iris intron 11 part and decrease Iris expression in 4T1 cells by >95% (Fig. 4A).

We injected 1 × 10⁵ 4T1/shCtrl, 4T1/shIris1, or 4T1/shIris2 cells (n = 8/each) in female BALB/c mice (Fig. 4B). Upper limits were approximately 0.5 cm³ tumor volume or 40 days. The 4T1/shCtrl cells formed 0.58 ± 0.1 cm³ tumors by day 20, while 4T1/shIris1 cells 0.13 ± 0.09 cm³ (>75% reduction, P < 0.000001), and 4T1/shIris2 cells 0.07 ± 0.06 cm³ (>85% reduction, P < 0.000001) tumors by day 40 (Fig. 4C).

We injected 5 × 10⁵ 4T1/shCtrl cells or 4T1/shIris1 cells (n = 10/each) in female BALB/c mice (Fig. 4B). Upper limits were approximately 1.0 cm³ tumor volume or 30 days. The 4T1/shCtrl cells formed 1.3 ± 0.1 cm³ tumors by day 20, while 4T1/shIris1 cells 0.59 ± 0.2 cm³ tumors (>55% reduction, P = 0.000002) by day 30 (Fig. 4D). ELISA analysis on sera isolated from these mice showed that compared with naïve mice, 4T1/shCtrl tumor–bearing mice circulation contained 10.1 ± 1.1- and 13.3 ± 0.8-fold higher GM-CSF and TGFβ1, respectively, while 3.3 ± 1.0- and 3.9 ± 0.6-fold, respectively (>60% decrease, Fig. 4E) in 4T1/shIris1 tumor–bearing mice.

To evaluate the tumor-infiltrating myeloid compartment within these tumors, unpermeabilized single-cell preparations labeled with anti-mouse FITC-CD11b and PE-F4/80 or PE-CSF2R antibodies were FACS analyzed. The 4T1/shCtrl tumors contained approximately 50% CD11b⁻ and approximately 50% CD11b⁺ fractions (Fig. 4F), increased to approximately 75% and approximately 25%, respectively, in 4T1/shIris1 tumors (Fig. 4F). Compared with 4T1/shCtrl tumors, 4T1/shIris1 tumors showed increase in M1 population (CD11b⁻/F4/80⁺ cells, 27.5% ± 2.6% vs. 53.7% ± 2.9%, P = 0.00004), and decrease in M2 population (CD11b⁺/F4/80⁺ cells, 46.2% ± 3.4% vs. 22.9% ± 1.9%, P = 0.00009), and myeloid-derived suppressor cell population (CD11b⁺/F4/80⁻ cell; ref. 37; 2.9 ± 0.5 vs. 1.6% ± 1.9%, P = 0.00271; Fig. 4F and G).

In addition, compared with 4T1/shCtrl tumors, in 4T1/shIris1 tumors the CD11b⁻/CSF2R⁻ (48.7% ± 1.5% vs. 62.8% ± 2.7%, P = 7.6 × 10⁻⁶), and CD11b⁻/CSF2R⁺ (5.1% ± 0.5% vs. 8.9% ± 0.6%, P = 4.0 × 10⁻⁶) populations increased (Fig. 4H and I). In contrast, the CD11b⁺/CSF2R⁻ (19.4% ± 0.6% vs. 12.9% ± 2.3%, P = 0.00031), and CD11b⁺/CSF2R⁺ (26.7% ± 1.8% vs. 15.3% ± 4.1%, P = 0.00044) populations decreased (Fig. 4H and I).

Indeed, paraffin-embedded sections from these tumors IHC stained for myeloid cells' markers (Fig. 4B) confirmed the lack of CSF2R-expressing M2 CD206⁺-TAMs within 4T1/shIris tumors (compare Fig. 5A with B). In fact, these analyses also showed that 4T1/shCtrl tumors contained 30.8 ± 7.3 F4/80⁺ cells/high power field (HPF) and 7.5 ± 3.1 CD206⁺ cells/HPF, while 4T1/shIris1 tumors contained 18.7 ± 4.9 F4/80⁺ cells/HPF (>40% decrease, P < 0.000001) and 1.1 ± 1.1 CD206⁺ cells/HPF (>80% decrease, P < 0.000001; Fig. 5C and D). Moreover, 4T1/shCtrl tumors contained 25.5 ± 5.5 CD25⁺ cells/HPF and 31.3 ± 4.6 FOXP3⁺ cells/HPF, while 4T1/shIris1 tumors contained 9.8 ± 3.5 CD25⁺ cells/HPF (>60% decrease, P < 0.000001) and 15.3 ± 2.9 FOXP3⁺ cells/HPF (>50% decrease, P < 0.000001; Fig. 5E and F). Finally, according to IHC (Fig. 5G) and FACS analysis performed on single-cell preparations from these tumors (Fig. 5H), 4T1/shCtrl tumors contained 25.7% ± 2.7% CD8⁺ and showed almost complete absence of PD-1⁺ cells, while 4T1/shIris1 tumors contained 38.1 ± 4.2 CD8⁺ cells (∼50% increase, P = 0.0001) and massive number of PD-1⁺ cells (Fig. 5G and H).

Taken together, these data suggest that: (i) like IRIS, depleting Iris also reduces TNBC cells ability to form tumors; and (ii) the bidirectional interaction between IrisOE cells and TAMs through GM-CSF and TGFβ1 triggers immunosuppressive microenvironment within TNBC tumors, including high levels of CSF2R-expressing M2 F4/80⁺/CD206⁺ TAMs and CD25⁺/FOXP3⁺ regulatory T cells, while lower levels of CD8⁺/PD-1⁺ cytotoxic T cells (CTL, Fig. 5I).

We injected 1 × 10⁶ of 4T1/shCtrl or 4T1/shIris2 cells (n = 10/cells) in female BALB/c mice (Fig. 4B). The limit was death from the disease within 40 days. Tumors, adrenal glands, brains, hearts, kidneys, livers, lungs, and spleens were collected from each anesthetized mouse, paraffin-embedded on a single block, sectioned at 4 μm, and hematoxylin and eosin–stained (Fig. 4B). As expected, 4T1/shCtrl tumors were larger. They showed evidence of aggressive TNBC cell morphology, including poorly differentiated epithelial cells with large hyperchromatic nuclei and relatively small cytoplasm (Fig. 6A, arrowheads), mesenchymal cells (Fig. 6A, arrows), and a large number of mitotic cells (Fig. 6A, asterisks). These mice showed relatively low area of internal necrosis throughout the tumor (Fig 6B, two-heads arrows). In contrast, 4T1/shIris2 tumors were small in size, contained differentiated epithelial cells only (Fig. 6C, arrowheads), and lacked almost completely mitotic cells (Fig. 6C). Moreover, compared with 4T1/shCtrl tumors, 4T1/shIris tumors exhibited extensive necrosis throughout the tumor (Fig 6B and D, two-heads arrows) accompanied by an increase in leukocytes infiltration (Fig. 6D, arrows).

In addition, compared with lungs from 4T1/shCtrl cell–injected mice, the lung of 4T1/shIris cell–injected mice were free from metastasis (compare Fig. 6E and F with G and H). Similarly, compared with livers from 4T1/shCtrl cell–injected mice, the livers of 4T1/shIris cell–injected mice were free from metastasis (compare Fig. 6I and J with K and L). Finally, compared with kidneys from 4T1/shCtrl cell–injected mice, the kidneys of 4T1/shIris cell–injected mice were free from metastasis (compare Fig. 6M and N with O and P). Interestingly, in the lungs (Fig. 6F) and kidneys (Fig. 6M), metastasis was found within or close to afferent vessels and in most cases, appeared infiltrative, whereas in the liver were more localized, often appearing spherical (Fig. 6I and J). More importantly, all mice injected with the 1 × 10⁶ 4T1/shCtrl cells (n = 10) died within 25 days (Fig. 6Q), while only 3 (n = 10) mice injected with 1 × 10⁶ 4T1/shIris2 cells died by day 40 (Fig. 6Q, P = 0.0001).

Taken together, these data suggest that IRISOE/IrisOE generates immunosuppressive microenvironment within TNBC tumors that is favorable for the generation of deadly metastatic precursors.

We injected 4 × 10⁶ IRIS291 (n = 20) and IRIS293 (n = 20) cells in female athymic mice (Fig. 3A). By day 20, all mice developed approximately 0.5 cm³ tumors (Fig. 7A). Each group was randomized and intratumorally injected with scrambled peptide (Fig. 7A, black arrows, n = 10/cell line) or the inhibitory IRIS peptide (IRISpep; refs. 17, 22; Fig. 7A, red arrows, n = 10/cell line) every 3rd day, a total of four times. By day 30, scrambled-injected IRIS291 2° orthotopic mammary tumors were 1.46 ± 0.11 cm³, and IRIS293 tumors were 1.49 ± 0.11 cm³, while IRISpep-injected IRIS291 tumors were 0.31 ± 0.08 cm³ (>75% regression, P = 0.0002) and IRIS293 tumors were 0.34 ± 0.14 cm³ (>75% regression, P = 0.004; Fig. 7A).

Unpermeabilized single-cell preparations from each tumor were labeled with FITC anti-CD11b and PE anti-CSF2R antibodies. FACS analysis showed scrambled-injected IRIS291 tumors contained 49.7% ± 5.1%, and IRIS293 tumors 47.4% ± 2.1% of CD11b⁻/CSF2R⁻ cells (Fig. 7B), while IRISpep-injected IRIS291 tumors 57.1% ± 5.4% (∼20% increased, P = 0.05), and IRIS293 tumors 60.2% ± 2.1% (>25% increase, P < 0.0001, Fig. 7B). Correspondingly, scrambled-injected IRIS291 tumors contained 22.5% ± 2.2%, and IRIS293 tumors 23.7% ± 1.8% of CD11b⁺/CSF2R⁺ cells (Fig. 7B), while IRISpep-injected IRIS291 tumors 13.8% ± 3.2% (40% decrease, P = 0.0009), and IRIS293 tumors 12.7% ± 1.2% (50% decreased, P < 0.0001, Fig. 7B).

In addition, ELISA analysis on sera isolated from these mice showed that compared with naïve mice, inactivating IRIS reduced the increase in circulating levels of GM-CSF by 74% and TGFβ1 by 65% in IRIS291 tumor–bearing mice, and by 70% and 65%, respectively, in IRIS293 tumor–bearing mice (Fig. 7C).

FACS analysis showed parental HME cell line contained approximately 50% stem-like/CD44⁺CD24⁻ cells, while IRISOE 1° tumors >95% (Fig. 7D; ref. 18), the majority of which are also CD47⁺ (Fig. 7E). Within this population in scrambled-injected 2° IRIS293 tumors 3.8% ± 2.5% were CD44⁻/CD47⁻ cells, and 94.7% ± 2.0% CD44⁺/CD47⁺ cells, while in IRISpep-injected tumors 58.1% ± 5.8% were CD44⁻/CD47⁻ cells (>15-fold increase, P = 0.00011), and 40.0% ± 5.2% CD44⁺/CD47⁺ cells (∼60% decrease, P = 0.00007, Fig. 7F and G). Similarly, in scrambled-injected 2° IRIS291 tumors 2.2% ± 0.5% were CD44⁻/CD47⁻ cells, and 96.6% ± 1.4% CD44⁺/CD47⁺ cells (Fig. 7G), while in IRISpep-injected tumors 56.8% ± 7.1% were CD44⁻/CD47⁻ cells (>25-fold increase, P = 0.00018), and 42.6% ± 6.9% were CD44⁺/CD47⁺ cells (>55% decrease, P = 0.00018, Fig. 7G).

Finally, these stem-like/CD44⁺CD24⁻ cells isolated from the 2° IRIS291 and IRIS293 tumors and 1° IRIS292 tumors expressed 80%–90% lower calreticulin and 9- to 10-fold higher PD-L1 on their surface compared with stem-like/CD44⁺CD24⁻ isolated from HME cell line (Fig. 7H).

Taken together, these data suggest that: (i) IRISOE increases CD47 (don't-eat-me initiator); (ii) decreases calreticulin (eat-me initiator); and (iii) increases PD-L1 expression on TNBC tumor cells, which promotes immunosuppression and TNBC deadly metastasis.

Tumor microenvironment plays a crucial role in promoting cancer metastasis. In this study, we showed that within the aggressiveness niche (19) in TNBC tumors, IRISOE cells secrete high levels of GM-CSF. At a low level, GM-CSF exerts an antitumor function by marking cancer cells for elimination by the immune system, or monocyte/macrophage phagocytosis (38). In contrast, high levels of GM-CSF expression exert a protumor function, for example, by enhancing MMPs expression (38). An increased level of GM-CSF in serum indicates poor prognosis in many cancers. In breast cancer, GM-CSF has a crucial role in the establishment of the premetastatic niche in the lungs (7, 39). Increased serum GM-CSF levels in breast cancer–bearing mice promote differentiation of bone marrow–derived cells into neutrophils (tumor-associated neutrophils, TAN) that accumulate in the lungs to promote the metastatic niche (39). Interestingly, our syngeneic model showed that an increase in lung metastasis correlated with high levels of circulating GM-CSF. It remains to be established whether this is accompanied by an increased in the number of TANs in the lungs of these mice.

GM-CSF silencing in IRISOE completely abolished TGFβ1 secretion by THP1s, suggesting bidirectional interaction with macrophages in the form of GM-CSF/TGFβ1. Like GM-CSF, TGFβ1 inhibition or overexpression results in carcinogenesis (40). TGFβ1 binding to TβRII recruits TβRI and activates it. During canonical signaling, activated TβRI/II phosphorylates SMAD2/3 that binds to SMAD4, translocates into the nucleus, and activates or repressors transcription. TGFβ1/TβRI/II can also signal through noncanonical pathways, including AKT (41). The fact that IRISOE blocked SMAD2/3 phosphorylation by TGFβ1 treatment in TNBC cells suggests that IRISOE blocks TGFβ1 canonical signaling and channels it toward noncanonical AKT signaling, perhaps to enhance stemness/EMT/aggressiveness in TNBC cells. Although we do not yet know the mechanism involved, IRISOE may upregulate one or more of the SMAD2/3 antagonistic pathways (42). We are evaluating this possibility. However, our data present another possible scenario. IRISOE upregulation of an autocrine EGF/EGFR signaling in TNBC cells (17, 22) could block TGFβ1/TβRI/II canonical signaling (43) and enhances noncanonical signaling to increase TNBC cells aggressiveness (Fig. 2I).

By inhibiting the “eat-me” signal and activating the “don't eat-me” signal in macrophages, IRISOE effectively renders macrophages within the tumors into nonphagocytic/cytokine-producing factories (Fig. 7I). How does IRISOE promote CD47 and suppress calreticulin expression in TNBC cells, remains to be seen? However, HIF1α, NF-κB/p65, and AKT are all known inducers of CD47 and at the same time, are upregulated in IRISOE in TNBC cells (16, 17, 18–22). Calreticulin is an endoplasmic reticulum resident that facilitates the folding of MHC class I molecules. Suppressing its expression would also influence antigen presentation to cytotoxic T cells within the tumor.

A significant obstacle toward an effective TNBC tumor vaccine is the inefficient stimulation of immune cells present within the tumor (44). Reversing this immunosuppression requires activation of CTLs, and suppression of any immune inhibitory elements within the tumor, for example, M2 macrophages (45). In fact, in TNBC tumors, chemotherapies induce the coordinated transcriptional induction of CD47 and PD-L1 mRNAs and proteins expression, leading to T-cell anergy or death (46). Interestingly, we recently showed that chemotherapies upregulate IRIS expression (17, 22). Moreover, during cancer progression, the continuous antigenic stimulation results in T-cell exhaustion (47). PD-1 expression is a common feature of exhausted CD8 T cells, and blockade of PD-1 restores T-cell function in vivo (48). Besides, the PD-L1 blockade could lead to a proinflammatory macrophage phenotype, for example, production of higher levels of TNFα and IL12 (49). Our data combined implicate inhibiting IRIS expression or activity (with a specific drug, ongoing) could eliminate the IRISOE/CD47⁺/PD-L1⁺/stem-like cell populations within TNBCs involved in the tumors immune evasion, chemo resistance, and metastases. Alternatively, blocking PD-1 or PD-L1 signaling in IRISOE TNBC tumors could rescue exhausted CTLs, and increase the size, proliferation, and survival of macrophages within the tumor by upregulating expression of the MHCII and the costimulatory molecule CD86 (49) and elicit an endogenous and synergistic innate and adaptive immune response against these IRISOE TNBC tumors.

No potential conflicts of interest were disclosed.

The funding body had no role in the design of the study and collection, analysis, interpretation of data, and writing the article.

Conception and design: W.M. ElShamy

Development of methodology: B.T. Paul, W.M. ElShamy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Sami, W.M. ElShamy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.A. Koziol, W.M. ElShamy

Writing, review, and/or revision of the manuscript: J.A. Koziol, W.M. ElShamy

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.M. ElShamy

Study supervision: W.M. ElShamy

Other (performed experiments): E. Sami, W.M. ElShamy

W.M. ElShamy is Dr. Lawrence & Mrs. Bo Hing Chan Tseu, American Cancer Society Research Scholar. This research is supported by an NCI grant R01 CA194447.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.