NR2F1 Is a Barrier to Dissemination of Early-Stage Breast Cancer Cells

Accumulating evidence in experimental models and human specimens across different cancers supports that early in cancer evolution, cancer cells disseminate to different anatomical sites (1–11). For instance, patients with breast ductal carcinoma in situ (DCIS) lesions, defined by pathologists as noninvasive, carry early disseminated cancer cells (eDCC) in target organs (5). Interestingly, a small but significant percentage of women with DCIS who never progressed to an in-breast invasive disease died from a distant recurrence (8, 12), suggesting that eDCCs can fuel the formation of metastatic events (6, 7). However, the mechanisms by which early cancer cells (ECC) acquire motile and invasive features that allow dissemination and target organ colonization are not clearly understood.

We previously showed that early activation of the rat Erbb2 (erb-b2 receptor tyrosine kinase) oncogene in mammary epithelial cells in young 14- to 18-week–old MMTV-HER2 females induced dissemination by turning off p38α/β signaling and promoting a wingless-type MMTV integration site family, member 1 (Wnt)-dependent partial epithelial–mesenchymal transition (EMT) program (7). Furthermore, we showed that upregulation of p38α/β signaling in dormant epithelial cancer cells [e.g., head and neck squamous cell carcinoma (HNSCC)] could restore the expression of a member of the steroid hormone receptor family, nuclear receptor subfamily 2, group F, member 1 (NR2F1; refs. 13, 14). NR2F1 levels were downregulated in early lesions and primary tumors from the MMTV-HER2 mouse model and also in a HER2-independent MMTV-Myc model when compared to wild-type (WT) siblings (13). It is unclear whether the loss of NR2F1 in the case of HER2⁺ early lesions is linked to tumor suppression, early dissemination, or both.

Here, we show that in murine and human early mammary cancer lesions, HER2 signaling reduced NR2F1 expression in a p38α-dependent manner. We also demonstrate that NR2F1 downregulation in HER2⁺ ECCs resulted in disorganized laminin 5, reduced E-cadherin expression, activation of WNT-dependent β-catenin signaling pathway as well as gain of a hybrid luminal/basal phenotype. Importantly, NR2F1 depletion in HER2⁺ ECCs also induces a partial EMT program. We also show using high-resolution intravital two-photon microscopy that loss of NR2F1 in MMTV-HER2 ECCs favors in vivo motility and invasion without increasing cancer cell proliferation. Finally, we report an enrichment of premalignant cells with inverse correlation between NR2F1 and paired related homeobox 1 (PRRX1) expression levels (NR2F1^LOW/PRRX1^HIGH) in DCIS samples. We conclude that early in cancer evolution, NR2F1 serves as a barrier to dissemination of ECCs by repressing a partial EMT and a hybrid luminal/basal program triggered by HER2 signaling. These studies provide novel insight into early changes in cancer progression that may result in early metastatic seeding.

Animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Icahn School of Medicine at Mount Sinai (New York, NY) protocol IACUC-2017–0162 and PROTO202100033. C57BL/6J MKK3^−/−/MKK6^+/− and MMTV-HER2-CFP and MMTV-HER2 mice were previously described (7, 15). FVB/N- Tg(MMTVneu)202Mul/J, nude (FOXN1^nu, RRID:IMSR_JAX:002376), WT C57BL/6J (RRID:IMSR_JAX:000664), and FVB/NJ (RRID:IMSR_JAX:001800) animals were purchased from Jackson Laboratories. MMTV-rtTA-TetO-Neu mice (16, 17) were donated by Dr. Emily Bernstein (Icahn School of Medicine at Mount Sinai). Human MDA-MB-261 (RRID:CVCL_0179) and BT474 cell (RRID:CVCL_5593) lines were donated by Dr. Mihaela Skobe at Icahn School of Medicine at Mount Sinai and used between passages P3–P8. Mycoplasma testing (ATCC; #30–1012K) was performed when cells were received on January 2021. SUM225 human cell line (RRID:CVCL_5593) was donated by F. Behbod at University of Kansas Medical Center (Kansas City, Kansas) and used between passages P99–P106. Mycoplasma testing was performed before shipping at Dr. Behbod lab, with date September, 2020.

For generation of MMTV-HER2 ECCs cultures, we freshly isolated mammary epithelial cells from MMTV-HER2 females of 14 to 18 weeks of age that did not have any visible small lesions or palpable tumors among all five pairs of mammary glands as we have previously established in this model (7). MMTV-HER2 mice were euthanized using CO2 inhalation at 14 to 18 weeks of age. Whole mammary glands were minced and digested in Collagenase/BSA at 37°C for 45 to 60 minutes. Red blood cell lysis buffer was used to remove blood cells and cells were then plated for 10 to 15 minutes in DMEM + 10% FBS in 100-mm dishes at 37°C for fibroblast removal. Cells were then incubated in 1 mmol/L PBS-EDTA for 15 minutes at 37°C and passed through a 25-gauge needle. Cell suspensions were filtered through a 70-μm filter before counting. For generation of mammosphere cultures, isolated cells were seeded in 6-well ultra-low adhesion plates (Corning #3471) at a density of 1 × 10⁶ cells per well in 1.5 mL mammosphere medium [DMEM/F12 (Gibco #11320–082), 1:50 B27 Supplement (Gibco #17504–044), EGF (10 ng/mL; Peprotech #AF-100- 15-A), 1:100 Pen/Strep (Invitrogen #15070–06)]. The following day another 0.5 mL of mammosphere medium was added to each well. Mammosphere cultures were kept for 7 to 10 days at 37°C before using them for experiments. For the generation of acini, 50,000 isolated cells were seeded on a layer of growth factor-reduced Matrigel (Corning #344230) in assay medium with 2% of Matrigel (Corning #356231). Assay medium [DMEM/F12 (Gibco #11320–082), 2% horse serum (Invitrogen #16050–122), 0.5 μg/mL hydrocortisone (Sigma-Aldrich #H-0888), 100 ng/mL cholera toxin (Sigma-Aldrich #C-8052), 10 μg/mL insulin (Thermo Fisher Scientific #12585–014), 1:100 Pen/Strep (Invitrogen #15070–06)] was supplemented with 5 ng/mL EGF (Peprotech #AF- 100–15-A) immediately before seeding cells. For generation of organoids, 100,000 freshly isolated mammary epithelial cells were seeded in a 6-well low attachment plate in 2 mL per well of organoid media [DMEM/F12 (Gibco #11320–082), 5% FBS (Gibco #10438–026), 20 ng/mL basic fibroblast growth factor (bFGF; R&D Systems #233-FB-10), 10 ng/mL EGF (Peprotech #AF-100–15-A), 2.5 μmol/L Rock inhibitor (EMD Millipore #420220), 4 ug/mL heparin (Sigma-Aldrich #H3393)] with 3% Matrigel. After every 3 to 4 days 0.5 mL fresh media with Matrigel was added until used for experimental purposes. Cells were passaged for up to 15 times. Mammary epithelial cells from 11 weeks old bitransgenic MMTV-rtTA/TetO-NeuNT females (that never received doxycycline while alive) were freshly isolated, grown in organoid media, and induced in vitro with DOX for 48 hours.

After 5 to 7 days in culture, acini were transfected with siRNA against mouse NR2F1 gene (Sigma-Aldrich #SASI_Mm01_00118737 and #SASI_Mm_00118738), siRNA against human Nr2f1 gene (Sigma-Aldrich #SASI_Hs01_00095428), control siRNA (Thermo Fisher Scientific #AM4611), mouse Mapk14 gene (Sigma-Aldrich #SASI_Mm01_00020743), mouse Mapk11 gene (Sigma-Aldrich #SASI_Mm01_00044863). Acini were also transduced with short hairpin RNA (shRNA) against mouse Nr2f1 gene (Origene #TL516691C and #TL516691D), or control shRNA (Origene #TR30021, noneffective scrambled shRNA). Acini were fixed 48 hours after transfection/transduction with 4% paraformaldehyde (PFA) and immunofluorescence (IF) assays were performed using the indicated antibodies. To measure mRNA changes, acini were transfected with the indicated siRNA and 48 hours later RNA was extracted from cultures using 1 mL TRIzol (Invitrogen #15596018), followed by RNA extraction. In some cases, cultures were treated for 24 hours with 1 μmol/L lapatinib (SelleckChem #S2111), 200 ng/mL Wnt4 (R&D Systems #475-WN-005/CF), 5 μmol/L SB203580 (SelleckChem #S1076), 2 μmol/L GSK2830371 (SelleckChem #S7573), or vehicle and RNA was extracted. In some cases, mammospheres growing in suspension conditions were transfected with siRNA against Nr2f1 or control for 48 hours and then RNA was extracted. For evaluation of invasion in response to NR2F1 expression, ECCs were transduced with lentiviral particles carrying an NR2F1 responsive promoter driving RFP expression. The lentiviral vector used, [pLV(Exp)-Puro-COUP-TF cis element>mRFP1], was custom-designed by VectorBuilder (vector ID: VB170830- 1011esp) and packaged in-house. Transduced cells were maintained in organoid cultures for up to 15 passages. Transduced acini were evaluated by confocal microscopy using a Leica SP5 confocal microscope and the LAS X Leica imaging software. For generation of stably MMTV-HER2 ECCs with NR2F1 depletion, freshly isolated MMTV-HER2 ECCs grown as mammospheres for 7 days were seeded in 2D and transduced with shNr2f1 (catalog no. TL516691D) or shControl (catalog no. TR30021, noneffective scrambled shRNA). Transduced cells were selected with 1 μg/mL puromycin (Alpha Aesar #J67236) for 3 days. Then, stable MMTV-HER2 ECCs were cultured in organoid media, in presence of puromycin for the first week, and then for not more than 15 passages to avoid changes in phenotype and cellular properties. For mammosphere formation assays, stable shControl and shNr2f1 MMTV-HER2 ECCs grown in organoids were seeded as single cells in 96-well plates and the number of spheres per well were quantified over time.

MMTV-HER2 ECCs grown as organoids were transiently transfected with 1 ug of pcDNA or pcDNA carrying the mouse Nr2f1 open reading frame plasmids (13). Then, 100,000 cells were seeded into the receiving chamber of transwell inserts with 8-μm pores (Corning #353097) in 200 μL of serum-free medium. On the bottom chamber, in 24-well plates, 500 μL of complete medium were added to establish a chemoattractant gradient. After 16 hours, cells on the receiving chambers (nonmigrated) were removed with a cotton swab. Migrated cells (on the opposite side of the membrane) were fixed with 4% PFA, stained with DAPI, and the number of cells in ten random 20X fields were counted using inverted fluorescent microscope. Technical triplicates were included in all experiments.

Three-dimensional (3D) acini cultures were time-lapse imaged using an inverted microscope Olympus IX-70 with Live Cell enclosed chamber. We recorded four positions per condition in parallel for 2 hours with a 10-minute interval. Temperature was maintained at 37°C and CO₂ at 5% throughout the imaging session. Movies were made by using either Metamorph (Molecular Devices) or the open source imaging software ImageJ (RRID:SCR_003070; ref. 18). This was done using the services of the Microscopy CoRE at Icahn School of Medicine at Mount Sinai. This research was supported in part by the Tisch Cancer Institute at Mount Sinai P30 CA196521–Cancer Center Support Grant. Protrusion number and their length-over-width (L/W) ratio were quantified. In L/W ratio plots, data points represent the mean ± SD of measurements taken by two independent operators. Invasive phenotype (a single elongated cell or cell file columns invading the Matrigel) or noninvasive phenotype was determined.

Single-cell suspension from organoids or mammosphere cultures were spun down at 230 g for 4 minutes and then 100,000 cells were suspended in 150 μL PBS with 1 mmol/L calcium and 0.5 mmol/L magnesium (PBS++). Matrigel was then added in a 1:1 ratio. Cells were injected into one inguinal (#4) mammary gland fat pad using a 27-gauge needle. Injection sites were monitored for development of tumors weekly for 2 months and final tumor incidence was calculated. Mice injected with primary tumor–derived cells were sacrificed when tumors reached 500 mm³ according to IACUC regulations.

For 2-photon intravital microscopy of mammary glands, GFP-tagged stable MMTV-HER2 ECCs expressing either shControl or shNR2F1 lentiviruses were orthotopically injected in FOXN1^nu athymic mice (Jackson Laboratory). Mice were anesthetized using ketamine/xylazine and a skin flap surgery performed exposing the fourth and fifth mammary glands as previously described (19). Intravital imaging was performed using an Olympus FV1000 MPE two-laser multiphoton microscope. An Infra-red tunable laser was used at 880-nm excitation for imaging of GFP-tagged cells. Imaging was done with a 25 × 1.05NA (XLPL25XWMP2, Olympus) water immersion objective. For each mouse, fields were imaged at 5-μm z steps to a depth of approximately 50 μm, with each stack taken approximately every 2 minutes for 1 hour. Average number of movies per animal = 2.2 to 3.8. Images were reconstructed and analyzed in ImageJ (18) utilizing the custom written ImageJ plugin ROI_Tracker (20). Any residual x-y drift not eliminated by the fixturing window was removed with postprocessing using the StackReg plugin (https://doi.org/10.1109/83.650848) for ImageJ. All procedures were conducted in accordance with the NIH regulations and approved by the Icahn School of Medicine at Mount Sinai animal use committee. Multiphoton microscopy was performed in the Microscopy CoRE at the Icahn School of Medicine at Mount Sinai, supported with funding from NIH Shared Instrumentation Grant (1S10RR026639).

Three-dimensional cultures were fixed with 4% PFA for 20 minutes at room temperature. Staining was performed as previously described for MMTV-HER2 3D cultures (7). Briefly, cells were permeabilized using 0.1% Triton X-100 for 20 minutes and blocked using 1x IF wash buffer (components) + 10% normal goat serum (NGS; Gibco #PCN5000) for 1 hour. Primary antibodies were all used at 1:100 concentration unless specified otherwise. Antibodies used were: E-cadherin (BD Biosciences #610181; RRID:AB_397580); β-catenin (BD Biosciences 610153; RRID:AB_397554); laminin 5 (Progen #10765; RRID:AB_2909803); CK14 (Biolegend #905301; RRID:AB_2565048); 1:25 NR2F1 (Abcam #ab181137; RRID:AB_2890250); PRRX1 (Novus Biologicals #NBP2–13816; RRID:AB_2909804); TWIST (Sigma-Aldrich #ABD29; RRID:AB_10807559); 1:50 HER2 (R&D Systems #OP15L; RRID:AB_2099415); PH3 (Cell Signaling Technology #9701S; RRID:AB_331535); SNAIL (Novus Biologicals #NBP1–80022; RRID:AB_11039370); CK18 (Progen #GP-CK-18; RRID:AB_2909805); PanCK (eBioscience #42–9003–80; RRID:AB_10804753); vimentin (R&D Systems #MAB2105; RRID:AB_2241653). The following secondary antibodies were used at 1:1000 dilution: Alexa Fluor-488 goat anti-mouse (Invitrogen #A-11001; RRID:AB_2534069) and Alexa Fluor-568 goat anti-rabbit (Invitrogen #A-11036; RRID:AB_10563566). F-Actin was detected using Alexa Fluor 568 Phalloidin (Invitrogen #A-12380). Chambers were removed from slides and wells were fixed and mounted with ProLong Gold Antifade reagent with DAPI (Invitrogen #P36931). Acini were imaged in 3D using a Leica SP5 confocal microscope and the LAS X Leica imaging software. Acini stained with anti-SNAIL antibody was considered positive when containing more than three z-stacks with more than one cell positive for nuclear SNAIL each. Acini stained with anti-VIMENTIN antibody is considered positive when containing more than three z-stacks with one cell positive. Acini stained with anti–E-cadherin antibody is considered positive when containing more than three z-stacks with more than three positive cells each. Acini stained with anti–β-catenin antibody is considered positive when containing more than five positive cells with β-catenin localized in the membrane. For all experiments in acini cultures, MMTV-HER2 ECCs from two to nine animal donors were used, and acini quantified ranged from 10 to 58 siControl (average = 30) and 15 to 52 siNR2F1 (average = 28) and data expressed per acinus or percentage of acini per condition. For experiments in acini cultures derived from organoids, at least two independent experiments were carried out, with 48 acini quantified by red fluorescent protein (RFP) expression.

Mammary gland section imaging was done using Leica DMi8 fluorescence microscope using LAS X Leica software. For IF staining of paraffin-embedded tissue: sections were routinely deparaffinized and rehydrated. Antigen retrieval was done in 10 mmol/L sodium citrate buffer (pH 6) to later permeabilize using 0.1% Triton x-100 in PBS. Blocking was done using 3% BSA in PBS with 1.5% NGS or normal donkey serum (Sigma-Aldrich #D9663; RRID:AB_2810235). Primary antibodies at 1:100 dilution unless otherwise specified–were assessed overnight at 4°C and secondary antibodies at 1:1000 dilution were assessed for 1 hour at room temperature. The following primary antibodies were used: β-catenin (Cell Signaling Technology #8480S; RRID:AB_11127855 or BD Biosciences #610153, RRID:AB_397554); CK14 (Biolegend #905301; RRID:AB_2565048); 1:50 HER2 (Abcam, #ab2428; RRID:AB_303063 or R&D Systems #OP15L; RRID:AB_2099415); PH3 (Cell Signaling Technology #9701S; RRID:AB_331535); TWIST1 (EMD Millipore # ABD29; RRID:AB_10807559); 1:25 NR2F1 (Abcam #ab181137; RRID:AB_2890250). The following secondary antibodies were used: Alexa Fluor-488 goat anti-mouse, Alexa Fluor-568 goat anti-rabbit, Alexa Fluor-488 donkey anti-goat, Alexa Fluor-647 donkey anti-rabbit (Invitrogen #A-31573; RRID:AB_2536183). Slides were mounted using Prolong Gold Antifade mounting media with DAPI (Molecular Probes #P36931).

All image analysis was performed in ImageJ. Circularity was calculated as Circularity = 4pi(area/perimeter^2, with value of 1.0 indicating a perfect circle.

For sorting of luminal cells, freshly isolated MMTV-HER2-CFP ECCs in single-cell suspension were incubated for 30 minutes on ice with anti-mouse Fc Block CD16/32 antibody (Biolegend #101319; RRID:AB_1574973) in FACS buffer (1% BSA 2 mmol/L EDTA in PBS) to avoid nonspecific antibody binding. Cells were then washed in FACS buffer and stained with fluorophore conjugated antibodies in FACS buffer. Cells were then washed in PBS and stained with Zombie NIR (Biolegend #423105) for 30 minutes on ice to determine live and dead (L/D) cell populations. Cell sorting of luminal (Lin^–CD24^HICD29⁺) cells was done over the live single-cell population on an IMIL3 Sorter (BD Biosciences) from the Flow Cytometry Core at Icahn School of Medicine at Mount Sinai. Antibodies used were: anti-CD45-PE (Miltenyi #130–117–498; RRID:AB_2727965); CD31-PE (Miltenyi #130–119–653; RRID:AB_2751780); CD140a-PE (Miltenyi#130–102–502; RRID:AB_2660635); CD24-VioBlue (Miltenyi #130–102–734; RRID:AB_2656579); CD29-PE-Vio770 (Miltenyi #130–105–125; RRID:AB_2660697). Cells were collected and seeded as described before for acinar cultures. For flow cytometry analysis after knockdown of NR2F1, acini cultured in Matrigel were collected in 2 mmol/L EDTA in PBS and incubated at 4˚C for 1 hour. Acini in suspension were trypsinized for 5 minutes and single-cell suspension immediately processed for viability staining with Zombie NIR. Cells were then fixed in 4% PFA and permeabilized using Perm/Wash buffer (BD Biosciences #554723; RRID:AB_2869011). Antibodies used were: cytokeratin 5-PE (Biorbyt #orb498435; RRID:AB_2909799); cytokeratin 14-CF640 (Biotium #BNC400532; RRID:AB_2909800); cytokeratin 18-CF488A (Biotium #BNC880041; RRID:AB_2909801); and cytokeratin 18-CF488A (Biotium #BNC880799; RRID:AB_2909802). Cells were then incubated with fluorescently labeled antibodies for intracellular detection of antigens. Cells were analyzed using an LSR Analyzer (BD Biosciences) from the Flow Cytometry Core at Icahn School of Medicine at Mount Sinai. Cells were gated in the following order: FSC/SSC > FCS-W/FSC-A > SSC-W/SSC-A > Lin-(CD45, CD31, and CD140a) > L/D (Zombie NIR-) > CFP⁺ > CD29/CD24 plots. Luminal cells sorted were cultured in matrigel to form acini and transfected with siRNAs as described previously at day 7 in culture and collected 48 hours after. Knockdown validation for NR2F1 was done by IF and quantification of the MFI in the nuclear area (as RNA collected was limited by low cell number availability). Cell analyzed = 212 for siControl; 311 for siNR2F1.

Samples were collected in RIPA buffer and centrifuged at 4°C, 18,000 g to clarify lysate. Protein concentration was determined by Pierce BCA Protein Assay Kit (Thermo Fisher Scientific #23225) and a standard BSA curve. Samples were then boiled for 5 minutes at 95°C in sample buffer (0.04M Tris-HCL pH 6.8, 1%SDS, 1% β-mercaptoethanol, and 10% glycerol). SDS- PAGE 10% gels were run in Running Buffer (25 mmol/L Tris, 190 mmol/L glycine, 0.1% SDS) and transferred to polyvinylidene difluoride (PVDF) membranes in Transfer Buffer (25 mmol/L Tris base, 190 mmol/L glycine, 20% methanol). Membranes were then blocked in 5% milk in Tris-buffered saline with 0.1% Tween20 detergent (TBST). Primary antibodies were left overnight at 4°C. Following washing with TBST buffer horseradish peroxidase (HRP)-conjugated secondary antibodies were left at room temperature for 1 hour. Western blot development was done using Amersham ECL Western Blot Detection (GE #RPN 2106) and GE ImageQuant LAS 4010. Primary antibodies used were: p38 1:500 (Cell Signaling Technology #9212; RRID:AB_330713); p-p38 (Thr180/Tyr182) 1:1000 (Cell Signaling Technology #4511; RRID:AB_2139682), p38α 1:5000 (BD Biosciences # 612168; RRID:AB_399539), GAPDH 1:1000 (Cell Signaling Technology #5174; RRID:AB_10622025); NR2F1 1:500: (Abcam #ab181137; RRID:AB_2890250), ribosomal protein S6 (pS6) 1:1000 (Cell Signaling Technology #5364; RRID:AB_10694233), tubulin 1:1000 (Abcam #15568; RRID:AB_2210952), Total β-catenin 1:1000 (Cell Signaling Technology #8480; RRID:AB_11127855), Lamin B1 1:1000 (Cell Signaling Technology #12586; RRID:AB_2650517), ATF2 1:1000 (Cell Signaling Technology #35031; RRID:AB_2799069), p-ATF2/ATF7 1:1000 (Cell Signaling Technology #24329; RRID:AB_2909807), p-HER2/ErbB2 1:1000 (Cell Signaling Technology #2243; RRID:AB_490899). Secondary antibodies used were peroxidase horse anti-mouse IgG (1:5000; Vector #PI-2000; RRID:AB_2336177) or anti-rabbit IgG (1:5000’; Vector #PI-1000; RRID:AB_2336198).

RNA was extracted from 2D or 3D cell cultures using TRIzol following provider's recommendation. For early mammary gland tissue RNA was extracted using Qiagen's RNeasy Lipid Tissue Midi Kit (Qiagen #74804). RNA (1–2 μg) was retrotranscribed into cDNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific # K1621). qRT-PCR was performed using PerfeCTa SYBR green (Quantabio #P/N 84069) in Bio-Rad thermocycler. Gapdh was used as reference gene control for all plates. All reactions were carried out with technical triplicates. Mouse qPCR primers were: Gapdh forward primer 5′-AACTTTGGCATTGTGGAAGGGCTC-3′; Gapdh reverse primer 5′-TGGAAGAGTGGGAGTTGCTGTTGA-3′; Twist1 forward primer 5′-AACTGGCCTGCAAAATCATA-3′; Twist1 reverse primer 5′-ACACCGGATCTATTTGCATT-3′; zinc finger E-box binding homeobox 1 (Zeb1) forward primer 5-ACCCCTTCAAGAACCGCTTT-3′; Zeb1 reverse primer 5-CAATTGGCCACCACTGCTAA-3′; Nr2f1 forward primer 5′-CCAATACTGCCGCCTCAA-3′; Nr2f1 reverse primer 5′-GGTTGGAGGCATTCTTCCTC-3′; Prrx1 forward primer 5′-GCACAAGCAGACGAAAGTGT-3′; Prrx1 reverse primer 5′-CTGGTCATTGTCCTGCTGAG-3′; Snai1 forward primer 5′-GGCGGAAGCCCAACTATAGC-3′, Snai1 reverse primer 5′-AGGGCTGCTGGAAGGTGAA-3′; Vim forward primer 5′-AGGAGGCCGAAAGCACCCTGC-3′, Vim reverse primer 5′-CCGTTCAAGGTCAAGACGTGCCA-3′; Cdh1 forward primer 5′- ACAGACCCCACGACCAATGA-3′, Cdh1 reverse primer 5′-CCTCGTTCTCCACTCTCACA-3′; Axin2 forward primer 5′-CAAAAGCCACCCAAAGGCTC-3′, Axin2 reverse primer 5′-TGCATTCCGTTTTGGCAAGG-3′, P38alpha (Mapk14) forward primer 5′-TGACCCTTATGACCAGTCCTTT-3′; P38alpha (Mapk14) reverse primer 5′-GTCAGGCTCTTCCACTCATCTAT-3′; P38beta (Mapk11) forward primer 5′-GCGGGATTCTACCGGCAAG-3′; P38beta (Mapk11) reverse primer 5′-GAGCAGACTGAGCCGTAGG-3′. Human QPCR primers were: PRRX1 forward primer 5′-CTGATGCTTTTGTGCGAGAA-3′; PRRX1 reverse primer 5′-ACTTGGCTCTTCGGTTCTGA-3′; ZEB1 forward primer 5′-GATGATGAATGCGAGTCAGATGC-3′; ZEB1 reverse primer 5′-ACAGCAGTGTCTTGTTGTTGT-3′; TWIST1 forward primer 5′-TGCAGCTATGTGGCTCACGA-3′; TWIST1 reverse primer 5′-ACAATGACATCTAGGTCTCCGGC-3′; CDH1 forward primer 5′-CGAGAGCTACACGTTCACGG-3′; CDH1 reverse primer 5′-GGGTGTCGAGGGAAAAATAGG-3′. Rat HER2 expression in the MMTV-rtTA-TetO-Neu cells was determined by Taqman assays against rat Her2 (Rn00566561_m1-FAM; Thermo Fisher Scientific) for rat Erbb2 and Gapdh (Mm99999915_g1-FAM; Thermo Fisher Scientific) and expressed as fold change over unspecific levels from unstimulated samples.

Paraffin-embedded sections from DCIS and invasive breast cancer patient tumors were obtained from the Cancer Biorepository at Icahn School of Medicine at Mount Sinai. Samples were deidentified and obtained with Institutional Review Board (IRB) approval, which indicated that this work does not meet the definition of human subject research according to the 45 Code of Federal Regulations (CFR) 46 and the Office of Human Subject Research. IF and IHC analysis was done using samples from 21 patients with DCIS. Benign adjacent (BA) and tumor areas from patients with DCIS were identified by a pathologist in adjacent sections stained with hematoxylin and eosin. Paraffin-embedded sections from DCIS patient samples injected in animals [mouse mammary intraductal (MIND) model; ref. 21] were obtained from F. Behbod. Samples were deidentified and obtained with IRB approval. The arbitrary threshold for NR2F1 levels (negative, high, and low) was established by comparing distribution plots of the mean fluorescence intensity (MFI) for NR2F1 and IgG for all benign tissues and DCIS samples. We followed a similar approach using distribution plots of the MFI for PRRX1 and IgG in all samples to establish PRRX1 levels (negative, high, and low). NR2F1^LOW/PRRX1^HIGH ratio was determined over HER2⁻ (n = 6) and HER2⁺ (n = 7) DCIS samples (including DCIS sections from MIND model; ref. 21), and benign adjacent tissues (n = 6). MFI values were quantified from the DAPI⁺ (nuclear) areas of all panCK⁺ cells per field of view (FOV; >8,000 cells). Percentage of cells with NR2F1^LOW/PRRX1^HIGH staining was calculated per samples and the frequency of samples positive and negative for this staining was determined per group and plotted (right panel).

Sections were routinely deparaffinized and rehydrated before staining procedures. Sections were treated with stabilized 3% hydrogen peroxide solution for 30 minutes to block endogenous peroxidase activity, incubated in a water bath immersed in sodium citrate buffer (pH 6.0) for 40 minutes at 97°C for antigen retrieval, and incubated with 3% BSA in PBS for 1 hour to block nonspecific binding, washing thoroughly in PBS between steps. HER2 IHC was performed with anti–c-ErbB2/c-Neu (Ab-3) Mouse mAb (3B5) 1/200 (Sigma-Aldrich t# OP15) overnight at 4°C. After washing, secondary horse anti-mouse IgG antibody (H + L), peroxidase (Vector #PI-2000) was added diluted 1/200 for 1 hour at room temperature. Signal was developed for 10 minutes in DAB solution (Vector #SK-4100 prepared according to manufacturer's instructions), sections were then counterstained for 1 minute with Harris hematoxylin, rehydrated, and mounted. Negative controls using mouse IgG showed no immunoreactivity. Evaluation of immunostainings was performed by a trained pathologist and scored according to American Society of Clinical Oncology/College of American Pathologists guidelines (22): negative for 0 (no membrane staining) and 1+ (faint or barely perceptible incomplete membrane staining); equivocal for 2+ (moderate circumferential staining in >10% of tumor cells or strong circumferential membranous staining in ≤10% of tumor cells) and positive for 3+ (strong circumferential membranous staining in >10% of tumor cells).

NR2F1 IHC was performed with 1:100 NR2F1 (Abcam #ab41858) following the same protocol. Total nuclear NR2F1 was detected and quantified.

Statistical analysis was done using Prism Software (RRID:SCR_002798). Differences were considered significant if P values were <0.05. For the majority of in vitro experiments, two- tailed Student t test was performed unless otherwise specified. For mouse experiments and human samples two-tailed Mann–Whitney test was used. Sample size chosen was done empirically. To determine the frequency of samples carrying NR2F1^LOW/PRRX1^HIGH ratio between benign adjacent and DCIS samples, Fisher exact test was used. Research reported in this publication was supported in part by the NCI Cancer Center Support Grant P30CA196521–07 awarded to the Tisch Cancer Institute of the Icahn School of Medicine at Mount Sinai and used the Biostatistics Shared Resource Facility. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Data were generated by the authors and included in the article.

Previous studies have shown that p38α/β signaling was enriched in a dormancy model of HNSCC cells and positively regulated NR2F1 levels (13, 14). As the phosphorylation of both p38α and p38β isoforms and NR2F1 expression (mRNA and protein) were reduced in early lesions of MMTV-HER2 mice when compared with normal mammary glands from FvB female siblings (7, 13, 23), we aimed to determine whether downregulation of NR2F1 in HER2⁺ ECCs was functionally linked to a reduction in p38 α/β activity. To this end, we treated young MMTV-HER2 females with a p38α/β inhibitor (SB203580), a time when this kinase is active in the mammary epithelium (7). p38α/β inhibition further downregulated Nr2f1 mRNA and protein levels in mammary glands when compared with vehicle group (Fig. 1A). In vitro, p38α/β inhibition did not change the levels of Nr2f2, another member of the NR2F-family, although suppression of E-cadherin and induction of twist basic helix-loop-helix transcription factor 1 (TWIST1) expression was observed, as reported (Supplementary Fig. S1A; ref. 7). Wang and colleagues (24) have recently showed that inhibition of the WT p53 induced phosphatase (WIP1) increased p-p38 levels in MMTV-HER2 ECCs and reduced dissemination of ECCs (24). We therefore determined whether treatment of MMTV-HER2 ECCs with WIP1 inhibitor (GSK2830371) could modulate Nr2f1 levels. We observed increased levels of p-p38 and Nr2f1 (Supplementary Fig. S1B), but no changes in the levels of Nr2f2 (Supplementary Fig. S1B). We also analyzed the regulation of NR2F1 by p38 in the genetically advanced MDA-MD-361 and BT474 HER2⁺ tumor cell lines and again observed upregulation of p-p38 levels upon treatment with WIP1 inhibitor in the absence of changes in NR2F1 or Nr2f2 expression (Supplementary Fig. S1C). These results support the idea that context-specific regulation occurs at different stages of HER2-driven breast tumor progression. Moreover, MKK3^–/–/MKK6^+/– (mitogen-activated protein kinase kinase 3/6) mice, with targeted disruptions of the Map2k3 and Map2k6 genes, both of which are upstream activators of p38, also showed reduced NR2F1 protein levels compared with syngeneic WT mice (Supplementary Fig. 1D), supporting the idea that even in the normal mammary gland, MKK3/6-p38α/β signaling regulates expression of NR2F1. To determine which p38α or p38β isoforms regulated NR2F1 expression in MMTV-HER2 ECCs, we used siRNAs directed against Mapk14 (p38α) and Mapk11 (p38β) isoforms and measured NR2F1 levels. While we observed downregulation of NR2F1 in cells transfected with sip38α but not sip38β (Supplementary Fig. S1E and S1F), in contrast, Nr2f2 was not regulated by either p38 isoform (Supplementary Fig. S1E and S1F). Together, these results suggest that p38α regulates NR2F1 expression but not Nr2f2 in MMTV-HER2 ECCs.

Next, we decided to determine whether HER2 negatively regulated NR2F1 in ECCs. First, we evaluated whether HER2⁺ ECCs had low levels of NR2F1 expression using IF staining. Our results showed that approximately 90% of HER2⁺ mammary cells expressed low/negative levels (based on fluorescence intensity threshold) of NR2F1 (Fig. 1B). Inhibition of HER2 (and HER1) with lapatinib upregulated Nr2f1 levels in MMTV-HER2 ECCs (Fig. 1C) but not in normal FvB mammary epithelial cells (Supplementary Fig. S1G). Moreover, addition of doxycycline to mammary epithelial cell cultures isolated from unstimulated MMTV-rtTA/TetO-Neu (HER2 inducible) females induced Erbb2 rat oncogene expression and caused downregulation of Nr2f1 when compared with vehicle-treated controls (Fig. 1D). Nr2f1 levels were also upregulated in lapatinib-treated human HER2⁺ SUM225 DCIS cells originally established from a patient with DCIS (25) when compared with control cells (Supplementary Fig. S1H). However, we did not observe changes in NR2F1 or NR2F2 levels after depletion of HER2 oncogene in genetically advanced human HER2⁺ tumor cell lines (Supplementary Fig. S1I), supporting our previous data (Supplementary Fig. S1C), which demonstrated that genetically evolved MDA-MD-361 and BT474 cells do not share the same upstream signaling pathway that controls Nr2f1 expression.

To determine whether HER2 was inhibiting Nr2f1 expression via p38, we treated MMTV-HER2 ECCs with SB203580 and TAK-715, a p38 inhibitor that is 28-fold more selective for p38α than p38β and does not inhibit p38γ/δ (26), in presence or absence of lapatinib (Fig. 1E). Lapatinib-induced Nr2f1 reexpression was blocked in the presence of SB203580 or TAK-715 (Fig. 1E, Supplementary Fig. S1J). Similar results were obtained in human HER2⁺ SUM225 DCIS cells (Supplementary Fig. S1K). Overall, these studies suggest that HER2 blocks NR2F1 expression via p38 signaling in human and murine premalignant models.

We showed that WNT4 was a member of a progesterone-driven signature that mediated dissemination of ECCs (6). Importantly, we observed downregulation of NR2F1 expression upon treatment of ECCs with WNT4 (Fig. 1F and G). We conclude that the activation of HER2 and WNT4 signaling causes reduced expression of NR2F1 in HER2⁺ ECCs.

We next examined the expression of NR2F1 in human DCIS using IF and observed significant reduction in the percentage of nuclear NR2F1⁺ cells in DCIS when compared with benign adjacent lesions (Fig. 1H; Supplementary Fig. S2A–S2C). We then correlated NR2F1 levels in DCIS with HER2 status (Supplementary Fig. S2) and observed that the percentage of NR2F1⁻ cells and the percentage of NR2F1^HIGH cells remarkably increased and decreased, respectively in HER2⁺ DCIS samples (P < 0.0001) when compared with benign adjacent lesions (Fig. 1I; Supplementary Fig. S2A–S2C). It is, however, interesting that HER2- DCIS samples were also mostly low/negative for NR2F1 expression when compared with benign adjacent lesions (P < 0.0001 for NR2F1 negative population; P < 0.01 for NR2F1^LOW population; Fig. 1I; Supplementary Fig. S2A–S2C). This was not altogether surprising as other oncogenic signals such as MYC have been reported to reduce NR2F1 protein levels in early lessions from MMTV-Myc females (13). These results suggest that alternative mechanisms, such as MYC oncogene or progesterone-induced paracrine signals as shown in Fig. 1F and G, could be involved in NR2F1 downregulation and that the loss of NR2F1 is by itself a crucial step in the early steps of cancer progression.

NR2F1 has been linked to growth suppression, lineage commitment, and differentiation (13, 27, 28). However, some studies have shown a role in promoting cell migration and axogenesis (29, 30). Furthermore, because reduced p38α/β signaling in HER2⁺ ECCs caused accelerated early dissemination (7, 24), and our data (Fig. 1A and E; and Supplementary Fig. S1B, S1D, S1E, S1J, and S1K) showed that p38 regulated NR2F1, we hypothesized that low NR2F1 levels will correlate with an invasive phenotype in MMTV-HER2 ECCs. To test this, MMTV-HER2 ECC acini were transduced with a lentiviral vector carrying an RFP-reporter for NR2F1 activity and quantified by confocal microscopy. The results showed that approximately 70% of ECCs that invaded outwards from the acini had low NR2F1 activity (Fig. 2A) while we observed consistently intra-acinar NR2F1-reporter activity.

Next, we used time-lapse microscopy to track and quantify cell dissemination into 3D Matrigel using siRNA and shRNA against Nr2f1 in MMTV-HER2 ECCs acini. NR2F1 depletion was efficient at the protein and mRNA levels (Supplementary Fig. S3A–S3D), did not change Nr2f2 levels (Supplementary Fig. S3D) and did not induce apoptosis (Supplementary Fig. S3E). The results of this study showed that while individual cells moving on the surface of siControl acini largely remain rounded, NR2F1 downregulation induced cell file columns invading the Matrigel (Fig. 2B–D; Supplementary Data ssm1 and ssm2 for siControl and Supplementary Data ssm3 and ssm4 for siNR2F1). Analysis of the L/W ratio of the protrusions (one cell or cell file columns) showed higher elongation towards the Matrigel in the siNR2F1 group when compared with control group (Fig. 2B and E; Supplementary Fig. S3F and S3G). The number of protrusions per acini did not change between siControl versus siNr2f1 group (Supplementary Fig. S3H). However, we observed higher percentage of protrusions with an invasive phenotype (a single elongated cell or cell file columns invading the Matrigel) after Nr2f1 knockdown compared with the control group (Fig. 2C, D, and F; Supplementary Fig. S3I). Moreover, overexpression of Nr2f1 (Supplementary Fig. S3J) in MMTV-HER2 ECCs significantly decreased their migratory capacity, as shown by migration assay in transwell inserts (Fig. 2G).

Next, we quantified how downregulation of NR2F1 expression affects the invasion in vivo. We performed intravital imaging through mammary fat pad windows (Fig. 2H; ref. 19) of GFP-tagged MMTV-HER2 ECCs stably expressing either shControl or shRNA targeting Nr2f1 (Supplementary Fig. S3K). Intravital time-lapse imaging revealed that GFP⁺ cells were viable and organized in acinar-like structures (representing ∼60%) or duct-like structures (∼40%) at the same proportions in shControl versus shNr2f1 injected ECCs (Supplementary Fig. S3L). Two-photon microscopy analysis showed that NR2F1 downregulation significantly increased the dissemination of ECCs from acini into the surrounding stroma when compared with shControl ECCs (Fig. 2I and J; Supplementary Data ssm5 for shControl and Supplementary Data ssm6 for shNR2F1). Analysis of the acini morphology showed less circularity and fewer cells per area in shNr2f1 acini than shControl acini, suggesting increased motility (Supplementary Fig. S3M).

To reinforce the link between HER2 and NR2F1 in ECCs, we tested the effect of lapatinib in ECCs that had lost NR2F1 expression using intravital imaging. Mice injected with shControl or shNr2f1 MMTV-HER2 ECCs were treated with either DMSO or lapatinib for 48 hours followed by intravital imaging through mammary fat pad windows. The efficacy of lapatinib was verified by in situ staining of mammary gland plugs with a pS6 (ribosomal protein S6) antibody (Supplementary Fig. S3N). Treatment with lapatinib inhibited invasion of ECCs isolated from mice injected with the control shRNA but not those injected with shNr2f1, arguing that lapatinib was not able to restore NR2F1 expression under these conditions (Fig. 2K). Percentage of cleaved-caspase 3 (CC3)⁺ ECCs in DMSO- and lapatinib-treated mice showed no differences (Supplementary Fig. S3O). These results imply that the motility of HER2⁺ ECCs is NR2F1-dependent and that NR2F1 acts as a barrier for dissemination downstream of HER2.

We previously reported that MMTV-HER2 ECCs gained expression of the EMT gene TWIST1 and reduced the expression of E-cadherin (7). We hypothesized that if NR2F1 limits early dissemination, it might regulate components of the EMT program. In support of our hypothesis, we found that depletion of Nr2f1 decreased E-cadherin levels and junctions in MMTV-HER2 ECCs acini and 2D at the mRNA and protein levels, suggesting destabilization (Fig. 3,A and B; Supplementary Fig. S4A). Furthermore, the EMT master regulators TWIST1, PRRX1, and Zeb1 were upregulated in shNR2F1 MMTV-HER2 ECCs acini and in situ staining from mammary gland tissues when compared with shControl group (Fig. 3A, C, and D; Supplementary Fig. S4B, Supplementary Fig. S3K). Depletion of NR2F1 mRNA levels in HER2⁺ SUM225 DCIS cells decreased E-cadherin levels and upregulated TWIST1 and ZEB1 mRNA levels without affecting NR2F2 levels (Supplementary Fig. S4C). The mRNA levels of PRRX1 were too low to be detected by qPCR in SUM225 cells. We also observed upregulation of Zeb1 and Prrx1 levels in SB203580-treated MMTV-HER2 ECCs acini when compared with DMSO-treated acini (Supplementary Fig. S4D). Nr2f1 knockdown reduced E-cadherin in FvB normal acini; however, it was not accompanied by transcriptional changes in the expression of EMT master regulators (Supplementary Fig. S4E). These results suggest that these mechanisms are specific to oncogenic HER2-regulated early progression stages. The protein levels of SNAIL (snail family zinc finger 1) and VIMENTIN (the latter of which was very low) in MMTV-HER2 ECCs upon Nr2f1 knockdown were not affected (Supplementary Fig. S4F and S4G). These findings support the notion that only certain components of the EMT program are controlled by NR2F1.

To determine whether low levels of NR2F1 and high levels of EMT genes are inversely correlated, we stained benign adjacent tissues and DCIS samples using anti-NR2F1 and anti-PRRX1 antibodies (Fig. 3E; see Materials and Methods). We analyzed paraffin-embedded DCIS sections and sections from the MIND model, which were freshly isolated patient-derived DCIS cells were injected in athymic mice (31). These studies revealed that the NR2F1^LOW/PRRX1^HIGH ratio was more frequently detected in DCIS (HER2⁺ and HER2⁻) samples than in benign adjacent tissues (Fig. 3E).

Next, we investigated whether NR2F1 controls β-catenin localization as this Wnt-regulated signal was associated with early dissemination in our prior studies (7). IF staining using a total β-catenin antibody showed less membranous localization in NR2F1-depleted MMTV-HER2 ECC acini and mammary glands than in controls (Fig. 3F; Supplementary Fig. S4H), similar to what we had observed when p38 was inhibited (7). Unfortunately, due to the limited sensitivity of the technique (7), IF detection of nuclear β-catenin was not possible. To address this issue, we transfected MMTV-HER2 ECC acini with either siControl or siNr2f1#2 (which targets a different sequence in the Nr2f1 gene than that used in previous experiments) and subjected nuclear fractions to Western blot analysis. Increased accumulation of nuclear β-catenin after Nr2f1 depletion was observed when compared with control cells (Supplementary Fig. S4I), suggesting that disappearance of membrane β-catenin upon Nr2f1 depletion correlates with β-catenin nuclear accumulation. To determine whether NR2F1 regulated β-catenin localization in a WNT-dependent manner, we transfected MMTV-HER2 ECC acini with either siControl or siNr2f1#2 (Supplementary Fig. S4J) and treated MMTV-HER2 ECC acini with either vehicle or DKK1, a WNT inhibitor. As expected, siNR2F1#2 decreased membrane localization of β-catenin localization when compared with siControl conditions (Fig. 3G), which is consistent with our previous studies (Fig. 3F; Supplementary Fig. S4H). The addition of DKK1 did not affect membrane β-catenin localization in siControl ECC acini; however, it was able to block the decrease in membrane β-catenin localization upon NR2F1 depletion (Fig. 3G). This result suggests that NR2F1 blocks β-catenin nuclear localization downstream of WNT signaling. Moreover, unlike what we observed with p38 α/β inhibition, the mRNA levels of Axin2, a transcriptional target of Wnt/β-catenin that creates a negative feedback loop to silence the signaling pathway, did not increase upon NR2F1 knockdown (Supplementary Fig. S4K). Overall, these results suggest that NR2F1 knockdown leads to translocation of nuclear β-catenin from the membrane to the nucleus via WNT-regulated pathways.

It has been previously shown that NR2F1 expression correlates with a more luminal phenotype in mammary epithelial cells (32) and acquisition of a basal phenotype has been linked to dissemination, metastasis, and poor survival (33–35). Our studies revealed that knockdown of Nr2f1 resulted in an increase in the percentage of CK14⁺ invading cells in MMTV-HER2 ECCs acini (Fig. 3H) and 2D (Supplementary Fig. S5A), which is consistent with increased levels of CK14 expression in acini-like structures measured after injection of shNR2F1 MMTV-HER2 ECCs into the fat pad of nude mice (Supplementary Fig. S5B). Interestingly, depletion of NR2F1 in MMTV-HER2 ECC acini (Supplementary Fig. S5C) did not change the percentages of CK5⁺, CK18⁺, or CK19⁺ cells detected by flow cytometry (Fig. 3I) and slightly increased Krt5 mRNA levels (Supplementary Fig. S5D). Furthermore, NR2F1 downregulation slightly reduced GATA binding protein 3 (Gata3) mRNA levels in MMTV-HER2 ECCs (Supplementary Fig. S5D). Moreover, no changes in the percentages of viable sorted out luminal MMTV-HER2 ECCs or changes in the percentages of CK18⁺ cells were observed as a function of NR2F1 expression (Supplementary Fig. S5E–SG), which is in agreement with the results from bulk primary cultures shown in Fig. 3I. These results suggest that NR2F1 depletion does not selectively induce luminal cell death and favors the acquisition of a hybrid luminal/basal phenotype in ECCs.

Finally, we observed that NR2F1 depletion disrupted laminin 5 deposition in MMTV-HER2 ECCs acini (Fig. 3J), supporting that downregulation of NR2F1 allows ECCs to invade into interstitial space. This result was also observed when MMTV-HER2 ECCs acini were treated with the p38 inhibitor SB203580 (7).

Based on these results we argue that depletion of NR2F1 in ECCs allows them to acquire a partial EMT program and a basal/luminal hybrid phenotype, which are required for early dissemination.

To assess the tumor-initiating ability of ECCs after NR2F1 knockdown, we injected either shControl or shNr2f1 MMTV-HER2 ECCs (Supplementary Fig. S3K) in the mammary fat pads of nude mice and monitored tumor outgrowth. Mice injected with shControl MMTV-HER2 ECCs did not develop tumors over the 2-month monitoring period (Fig. 4A), which is in agreement with our previous work showing that MMTV-HER2 ECCs are largely nontumorigenic in orthotopic sites (7). Interestingly, mice injected with shNr2f1 MMTV-HER2 ECCs did not develop tumors (Fig. 4A). By contrast, we observed tumor take in all mice injected with late stage MMTV-HER2 tumor cells as previously shown (Fig. 4A; ref. 7). shNr2f1 MMTV-HER2 ECCs injected into the fat pad had the same percentage of cells that stained positive for the proliferative marker phospho-histone 3 (PH3) when compared with shControl cells (Fig. 4B). The same result was observed in MMTV-HER2 3D acini (Supplementary Fig. S5H). These results suggest that downregulation of NR2F1 do not induce primary tumor outgrowth in animals. To assess the in vitro stemness capacity of MMTV-HER2 ECCs after Nr2f1 knockdown, we transduced MMTV-HER2 ECCs with shControl or shNr2f1 lentiviruses and measured number of mammospheres. We observed an increased number of spheres upon Nr2f1 depletion when compared with shControl group (Supplementary Fig. S5I), leading us to conclude that Nr2f1 knockdown induces proliferation and stemness in in vitro mammospheres assays. However, in in vivo experiments, the dissemination of ECCs upon NR2F1 knockdown dominates over the proliferation capacity. We therefore did not observe tumor formation upon Nr2f1 knockdown (Fig. 4A) and similar levels of PH3⁺ cells were detected in shC and shNr2f1 ECCs injected in the fat pad of nude mice (Fig. 4B).

To strengthen our hypothesis concerning the role of NR2F1 in spontaneous dissemination to the lungs, we harvested the lungs of the above-mentioned mice after a 10- to 16-day period postinjection to measure the number of infiltrating eDCCs (Fig. 4C and D). The number of single lung eDCCs was higher in shNr2f1 group when compared with shControl group (Fig. 4D), supporting the notion that NR2F1 represses systemic dissemination to distant organs such as the lung. At this time point, there was no difference in the percentage of proliferative eDCCs measured by PH3 staining (Fig. 4E), which ruled out the possibility that the increased number of eDCCs seen in shNr2f1 group was due to the capacity of eDCCs to start proliferating immediately. It is worth mentioning that no metastases were found in either group, as metastases take longer to develop due to a prolonged dormancy phase (7).

Overall, our results suggest that during early stages of cancer progression, NR2F1 primarily functions to limit an EMT- and a hybrid luminal/basal-like dissemination program activated by HER2 signaling, which is linked to motility and invasion processes rather than a “growth-limiting” program (Fig. 5).

Our study expands our understanding of how an early-established HER2-dependent program primarily deregulates motility and invasion to favor dissemination of HER2⁺ ECCs to secondary organs before activating oncogene addictive growth (Fig. 5).

We report that NR2F1 is positively regulated by p38α signaling and repressed by HER2 and WNT4 pathways in human and murine HER2⁺ ECCs but not in genetically evolved HER2⁺ tumor cell lines. Recent papers that built on our previous work on the HER2-p38 axis (7) have reported a HER2-dependent degradation of Tpl2 (MAPK8) and that inhibition of Tpl2 enhanced early dissemination in the MMTV-HER2 mouse model (36). Whether NR2F1 is transcriptionally or posttrancriptionally regulated by Tpl2 activity will be the focus of future work.

Supporting the idea of a linear signaling pathway that has p38α upstream of NR2F1 in HER2⁺ ECCs is the fact that a WIP1 inhibitor, which has been shown to increase p-p38 and E-cadherin levels and prevent early dissemination in the MMTV-HER2 mouse model (24) also increased NR2F1 expression (Supplementary Fig. S1B). Moreover, NR2F1 blocked the expression of Zeb1 and TWIST1, known negative regulators of E-cadherin (37, 38). Genomic data analysis revealed two peaks for the NR2F1 binding in the ZEB1 promoter in GM12878 cells (ENCSR514VYD data set, ENCODE). Moreover, Rada-Iglesias and and colleagues showed the binding of NR2F1 to Twist1 promoter using NR2F1 chromatin immunoprecipitation (ChIP) assays in neural crest cells (28). Future work is needed to validate the binding of NR2F1 in ZEB1 and TWIST1 promoter regions in HER2⁺ ECCs by ChIP analysis. Here, we showed that NR2F1 blocked β-catenin nuclear localization downstream of WNT (Fig. 3G and Supplementary Fig. S4I). Previous studies in MMTV-HER2 ECCs have reported that inhibition of MK2 (MAPK2), a target of p38, led to decreased levels of phosphorylated heat shock protein 2 (HSP27), resulting in its ability to activate β-catenin (24). Thus, future work will focus on understanding whether NR2F1 regulates HSP27 phosphorylation to induce β-catenin activity.

We also observed that NR2F1 depletion in ECCs increased PRRX1 expression. PRRX1 is particularly interesting, as PRRX1a and PRRX1b isoforms have been involved in tumor progression and they play important roles in metastatic outgrowth and dissemination, respectively (39). Whether NR2F1 can modulate the alternative splicing of PRRX1 in the mammary epithelium remains to be elucidated.

Simultaneous overexpression of TWIST1 and downregulation of E-cadherin in normal mammary epithelial cells has been shown to increase cell motility in cells expressing the basal CK14 and luminal CK8 markers (40). Our data show the same hybrid luminal/basal phenotype acquired by premalignant lesions upon the depletion of NR2F1 that may facilitate dissemination, survival at ectopic sites, drug resistance, and tumor-initiating potential (41).

NR2F1 regulates dormancy in a variety of tumor types (13). NR2F1 expression is relatively low in breast, HNSCC and prostate primary tumors. However, once cancer cells disseminate to secondary organs as DCCs they regain NR2F1 expression (13, 42). We also showed that NR2F1, downstream of p38 signaling, regulated dormancy of late DCCs in advanced tumor models (13) and that detection of NR2F1 in DCCs from breast and prostate cancer patient's bone marrow aspirates correlated with good prognosis (43, 44). Moreover, we showed that the dormancy program of eDCCs was p38-independent (7). These results indicate that different dormancy programs are present in early versus late DCCs. Whether NR2F1 expression is regained in eDCCs and whether it plays a role in dormancy of eDCCs remains to be determined.

The downregulation of NR2F1 levels was linked to the deletion of a susceptibility locus in a rat-human comparative genetic analysis (45). Thus, determining these susceptibility loci may predict patients at risk for early spread, especially for those patients with family history. In this study, we have corroborated the detection of an inverse correlation between NR2F1 and PRRX1 expression in human HER2⁺ and HER2- DCIS samples. The clinical application of this result needs to be further evaluated by determining the association of NR2F1^LOW/PRRX1^HIGH ratio with the clinical detection of eDCCs and whether it could be used to predict the risk of relapses in HER2⁺ and HER2⁻ patients. Furthermore, our intravital imaging results suggest that treatment of patients carrying HER2⁺ premalignant lesions may benefit from lapatinib treatments if NR2F1 levels are able to be restored by lapatinib.

J.J. Bravo-Cordero reports grants from NCI (grant nos. R01CA244780 and P30-CA196521) and grants from Susan G. Komen/CCR18547848 during the conduct of the study. F. Behbod reports grants from Cancer Research UK and KWF Kankerbestrijding (reference no. C38317/A24043), grants from NIH (grant nos. NIH-R00 CA127462, NIH-R01CA207445, and NIH-R21CA226567); KU Cancer Center Support Grant (grant no. P30 CA168524); grants from The Kansas Institute for Precision Medicine - COBRE (grant no. P20 GM130423; AKG) during the conduct of the study; and Pilot Grant: NCATS Frontiers-CTSA grant from NCATS awarded to the University of Kansas for Frontiers (reference no. UL1TR002366). M.S. Sosa reports grants, personal fees, and other support from Susan G. Komen CCR17483357, K22 NIH K22CA201054, Melanoma Research Foundation, Melanoma Research Alliance, Schneider-Lesser Foundation Fellow Award, CSBC Pilot Project-Sage Bionetworks, Breast Cancer Alliance; grants and other support from P30CA196521-07; grants and other support from 1S10RR026639 during the conduct of the study; and nonfinancial support from HiberCell LLC outside the submitted work. No disclosures were reported by the other authors.

C. Rodriguez-Tirado: Data curation, formal analysis, investigation, methodology, writing–review and editing. N. Kale: Formal analysis, investigation, methodology. M.J. Carlini: Formal analysis, investigation, methodology, writing–review and editing. N. Shrivastava: Formal analysis, investigation, methodology. A.A. Rodrigues: Formal analysis. B.D. Khalil: Formal analysis, validation. J.J. Bravo-Cordero: Formal analysis. Y. Hong: Resources. M. Alexander: Validation. J. Ji: Data curation. F. Behbod: Resources. M.S. Sosa: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

The authors thank Dr. Roger Davis for MKK3^–/–/MKK6^+/– samples and his valuable feedback. They also thank Dr. Skobe for providing the BT474 and MDA-MB-361 cells. Grant support: grant number CCR17483357 (to M.S Sosa); grant number MRF-CDA (to M.S Sosa); grant number K22CA201054 (to M.S. Sosa); CSBC Pilot Project-Sage Bionetworks (to M.S. Sosa); Schneider-Lesser Foundation Fellow Award (to M.S. Sosa); MRA (to M.S. Sosa); BCA (to M.S. Sosa); grant number CCR18547848 (to J.J. Bravo-Cordero); grant number R01CA244780 (to J.J. Bravo-Cordero); grant number K22CA196750 (J.J. Bravo-Cordero); grant number P30-CA196521 (to J.J. Bravo-Cordero); grant number NIH-R00 CA127462 (to F. Behbod); grant number NIH-R01CA207445 (to F. Behbod); grant number NIH-R21CA226567 (to F. Behbod); grant number P30 CA168524 (to F. Behbod); grant number P20 GM130423 (to F, Behbod); grant number UL1TR002366 (to F. Behbod); grant number C38317/A24043 (to F. Behbod); grant number P30CA196521–07 (Icahn School of Medicine at Mount Sinai, ISMMS); grant number 1S10RR026639 (ISMMS).

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.