Tumor cell Jagged-1 promotes regional lymphatic metastasis and predicts recurrence in node-positive breast cancer Authors: Benjamin Gordon, Bhairavi Swaminathan, Seock Won Youn, Natalia Anna Obacz, Rahul Vadakath, Pamela Teneqexhi, Isabel Alvarez-Lopez, Marta Rezola, Sara Manzano, Zuhara Telletxea, Ziqiao Xu, Zhengjia Chen, Ekrem Emrah Er, L. A. Naiche, Maria M. Caffarel, Jan Kitajewski Published: 2026 ABSTRACT BackgroundLymph node metastasis marks a clinically and biologically important transition in breast cancer, reflecting more aggressive disease and increased risk of recurrence. Expression of the Notch ligand JAG1 is linked to poor breast cancer outcomes, but its role in lymphatic dissemination is unclear.MethodsTumor microarray analysis with multiplex staining was performed on synchronous breast and lymph node metastases from node-positive patients to quantify JAG1 expression and correlate findings with clinical outcomes. Orthotopic implantation of human and murine breast cancer cells was used to assess the impact of JAG1 expression on lymphovascular invasion and lymph node metastasis in vivo. The contribution of VEGFR3 signaling to JAG1-mediated lymphatic dissemination was evaluated using a soluble VEGFR3 antagonist. The cell-specific response of lymphatic endothelial cells to JAG1 was assessed by transendothelial migration assays, barrier examination, and sequencing.ResultsMammary tumor cells expressing JAG1 demonstrated increased lymphovascular and lymph node metastasis in mouse models. Inhibition with a soluble VEGFR3 antagonist attenuated JAG1-dependent lymph node metastasis. Exposure to JAG1 disrupted lymphatic endothelial barrier integrity, promoted tumor cell migration across cultured endothelium, and induced pro-inflammatory and pro-adhesion changes in lymphatic endothelial cells. Tumor microarray analyses revealed enrichment of JAG1 in metastatic tumor cells within lymph nodes relative to matched primary tumors. High JAG1 expression in lymph node metastases was significantly associated with increased risk of recurrence.ConclusionsThese findings support a role for tumor cell JAG1 promoting lymphatic dissemination and lymph node metastasis in breast cancer through effects on lymphatic endothelial signaling and barrier function. Enrichment of JAG1 in lymph node metastases and its association with recurrence further support its potential utility as a prognostic biomarker and therapeutic target in node-positive breast cancer.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13058-026-02331-7. Introduction Lymph node metastasis serves as a sentinel event in breast cancer, marking a prognostically distinct stage that alters clinical management and reflects a transition to biologically more aggressive disease [1, 2]. In animal models, nodal spread contributes to disease progression, and mechanisms have been identified that promote lymphangiogenesis, lymphatic invasion and lymph node colonization [3–5]. In preclinical models, invasion of tumor cells into lymphatic vessels and lymphogenous metastasis can be inhibited by targeting receptors on lymphatic endothelial cells (LECs), indicating that LECs are active participants in tumor cell dissemination rather than passive conduits [6–8]. More recently, transcriptionally distinct LEC subpopulations involved in locoregional spread have been described in normal physiology and oncologic contexts [9, 10]. How breast tumor cells engage LECs to activate pro-metastatic signaling pathways, however, remains unclear. Notch signaling regulates cancer cell behavior, the tumor vascular microenvironment [11], and tumor immunity across multiple malignancies [12]. Expression of the Notch ligand JAG1 in breast tumor cells has long been associated with poor prognosis in patients [13–15]. In retrospective studies, higher tumor JAG1 expression correlates with lymphovascular invasion (LVI) and breast cancer patient lymph node status [13–17], raising the possibility that tumor expression of JAG1 actively contributes to lymphatic dissemination. Prior work has shown that Notch ligands, including JAG1, expressed on cancer cells can activate Notch signaling in other cells in the tumor microenvironment, including neighboring blood endothelial cells (BECs) [11, 18–25]. JAG1/Notch signaling is integral to lymphatic vascular development, sprouting, and adult lymphatic function [24–27]. We and others have established that Notch signaling in LECs has distinct roles from that in BECs, and in LECs Notch signaling both upregulates and provides feedback to VEGFR3 (FLT4), a master regulator of the lymphatic vasculature [24–26, 28]. These data suggest that JAG1 on breast cancer cells may interact with LECs to drive lymphatic changes that promote tumor dissemination. We tested whether JAG1 expression in breast tumor cells drives disease by promoting lymph node metastasis. Clinically, through examination of synchronous breast and node samples of node-positive breast cancer patients, we show that tumor cells that invade lymph nodes have higher levels of JAG1 expression compared to the originating breast tumor. We demonstrate that JAG1 expression on breast tumor cells promotes progression through the lymphogenous metastasis cascade: from the tumor to the mammary collecting lymphatic vessels, to the subcapsular sinus of the lymph node, and finally overt nodal metastasis. JAG1 promotion of lymph node metastasis was inhibited by a soluble antagonist of the lymphatic endothelial receptor VEGFR3, suggesting that this phenotype is mediated by LECs. We identify JAG1-regulated interactions between tumor cells and LECs which promote tumor transendothelial migration and endothelial inflammation. Sentinel lymph nodes (sLNs) are routinely sampled during breast cancer staging and provide an opportunity to study molecular features of metastasizing disease. We determined that JAG1 expression levels in the node, but not the tumor, can predict disease recurrence post-lumpectomy, demonstrating that enrichment of JAG1 in lymph nodes may be clinically meaningful and further suggesting that molecular examination of tumor characteristics in sLNs may be warranted in patients with lymph node metastasis. Results Tumor-cell JAG1 expression is selectively enriched in lymph node metastases compared to matched primary breast tumors Although prognostic markers are typically assessed in primary tumors, metastatic lymph nodes represent an underexplored reservoir of biologically selected tumor cell states [29]. The majority of pathologically node-positive (pN+) patients have breast tumors with high JAG1 [13, 15], and we therefore hypothesized that the tumor levels of JAG1 would be concordantly high or even enriched in the lymph node. To examine JAG1 expression in primary breast tumor samples and lymph node metastases, we evaluated a tissue microarray (TMA) consisting of 284 pN+ patients with estrogen receptor positive (ER+) breast cancer with synchronous TMA cores from breast and lymph nodes (Supplementary Tables S1 and S2). We used multiplex OPAL-based staining to detect JAG1 expression and identify tumor cells (PanCK-positive), lymphatic endothelium (LYVE1, PDPN, CD31), and immune cells (CD45) (Fig. 1A). We then employed Halo High Plex artificial intelligence to quantify tumor cell-specific JAG1 expression levels (Fig. 1B). In paired analysis of synchronous breast primary tumor and lymph node tumor cells, tumor-derived JAG1 was significantly enriched in the patient lymph nodes both in percent positivity and H-score, with a mean increase in H-score of 14.89 (95%CI, 10.42–19.35, Paired Student’s t-test) in the lymph nodes (Fig. 1C and D). Per-patient analysis of breast versus lymph node showed a significant JAG1 H-score enrichment in lymph nodes in just under 70% of patients (SlopeLN /Breast= 1.174, Fig. 1E−F). Thus, JAG1 expression is significantly enriched in breast tumor cells that have metastasized to lymph nodes compared to matched primary tumors. Fig. 1Tumor cell JAG1 expression is increased in metastatic lymph nodes in comparison to synchronous primary breast tumors. A Multiplex OPAL-based staining of PanCK (tumor), JAG1, LYVE1 (lymphatic endothelium), PDPN (lymphatic endothelium), CD31 (pan-endothelial), and CD45 (immune cells) in a tumor microarray of 284 hormone receptor positive breast cancer patients containing paired samples from both breast and lymph node. Scale bars, 500 and 100 μm, respectively. Top left panel shows overall core, other panels show indicated detail area. B Representative images of breast (top) and lymph node (bottom) demonstrating JAG1 staining. Arrowheads point to tumor cells. Scale bar, 100 μm. C Paired comparison of percent positivity for JAG1 expression in PAN-CK positive cells in breast and synchronous lymph nodes. N = 284. Paired student’s t-test. D Paired comparison of JAG1 H-score in PanCK positive regions in breast and synchronous lymph nodes. N = 284. Paired student’s t-test. E Regression analysis between JAG1 H-score in breast and lymph nodes demonstrating JAG1 enrichment in lymph nodes. Pink area denotes the 95% confidence interval of slope while all values in the blue area represent those in which H-score(Lymph Node) > H-score(Breast), with slope > 1. F Normalized JAG1 H-Score per patient, indicating fold change between breast and node. Paired student’s t-test. G (top panel) Representative images of core staining showing PDPN and LYVE1 overlap, indicating lymphatic vessels, and CD31 + LYVE1neg vessels, indicating blood vessels (top panel). Blood vessels also express JAG1. (bottom panel) HALO AI markup of PDPN+LYVE1 + lymphatic vessels overlaid with tumor (PanCK) markup to identify tumoral and stromal lymphatics. This also captures an example of LVI. Scale bar, 100 μm. H Linear regression analysis between %area of lymphatic vessels in breast cores (stromal and tumoral ROIs) and JAG1 H-score in breast. I Linear regression analysis between %area of lymphatic vessels in lymph node cores (stromal and tumoral ROIs) and JAG1 H-score in lymph node. J JAG1 H-score in breast cores in patients that were scored as exhibiting LVI (LVI+) or no LVI (LVI-) in original pathology reports (see Table 1). Unpaired student’s t-test Tumor cell JAG1 expression is increased in metastatic lymph nodes in comparison to synchronous primary breast tumors. A Multiplex OPAL-based staining of PanCK (tumor), JAG1, LYVE1 (lymphatic endothelium), PDPN (lymphatic endothelium), CD31 (pan-endothelial), and CD45 (immune cells) in a tumor microarray of 284 hormone receptor positive breast cancer patients containing paired samples from both breast and lymph node. Scale bars, 500 and 100 μm, respectively. Top left panel shows overall core, other panels show indicated detail area. B Representative images of breast (top) and lymph node (bottom) demonstrating JAG1 staining. Arrowheads point to tumor cells. Scale bar, 100 μm. C Paired comparison of percent positivity for JAG1 expression in PAN-CK positive cells in breast and synchronous lymph nodes. N = 284. Paired student’s t-test. D Paired comparison of JAG1 H-score in PanCK positive regions in breast and synchronous lymph nodes. N = 284. Paired student’s t-test. E Regression analysis between JAG1 H-score in breast and lymph nodes demonstrating JAG1 enrichment in lymph nodes. Pink area denotes the 95% confidence interval of slope while all values in the blue area represent those in which H-score(Lymph Node) > H-score(Breast), with slope > 1. F Normalized JAG1 H-Score per patient, indicating fold change between breast and node. Paired student’s t-test. G (top panel) Representative images of core staining showing PDPN and LYVE1 overlap, indicating lymphatic vessels, and CD31 + LYVE1neg vessels, indicating blood vessels (top panel). Blood vessels also express JAG1. (bottom panel) HALO AI markup of PDPN+LYVE1 + lymphatic vessels overlaid with tumor (PanCK) markup to identify tumoral and stromal lymphatics. This also captures an example of LVI. Scale bar, 100 μm. H Linear regression analysis between %area of lymphatic vessels in breast cores (stromal and tumoral ROIs) and JAG1 H-score in breast. I Linear regression analysis between %area of lymphatic vessels in lymph node cores (stromal and tumoral ROIs) and JAG1 H-score in lymph node. J JAG1 H-score in breast cores in patients that were scored as exhibiting LVI (LVI+) or no LVI (LVI-) in original pathology reports (see Table 1). Unpaired student’s t-test Table 1Clinicopathologic characteristics of node-positive patients treated with breast-conserving surgeryCharacteristicN = 1051 Hospital HUD51 (49%)ONK54 (51%) Age 57 (48, 70); 58.5 ± 13.4Unknown83 Grade 114 (13%)258 (55%)329 (28%)Unknown4 (4%) Estrogen receptor Positive104 (99%)Negative0 (0%)Unknown1 (1%) Progesterone receptor Positive91 (87%)Negative13 (12%)Unknown1 (1%) HER2 Score = 068 (65%)Score = 125 (24%)Score = 25 (5%)Score = 36 (5%)Unknown1 (1%) Menopause Pre-menopause33 (31%)Post-menopause72 (69%)Unknown0 (0%) Surgery No-mastectomy (lumpectomy)105 (100%)Mastectomy0 (0%)Unknown0 (0%) Hormone therapy Tamoxifen (Tmx)31 (30%)Aromatase inhibitor (AI)33 (31%)Tmx-> AI39 (37%)None/unknown2 (2%) XRT Breast conserving RT101 (95%)No RT3 (3%)Chest-wall RT1 (1%)Unknown1 (1%) Adjuvant chemotherapy No22 (21%)Yes78 (68%)Unknown1 (1%) Tumor size (mm) 22 (16.5, 25); 22.3 ± 7.5Unknown0 # of LN Mets 3.0 (1.0, 6.0); 4.5 ± 4.5Unknown1 LVI Yes28 (27%)No68 (65%)Unknown9 (8%)5 Year survival88% (84%, 93%)5 Year recurrence free75% (69%, 82%)1 n (%); Median (IQR); Mean ± SD Table 1Clinicopathologic characteristics of node-positive patients treated with breast-conserving surgeryCharacteristicN = 1051 Hospital HUD51 (49%)ONK54 (51%) Age 57 (48, 70); 58.5 ± 13.4Unknown83 Grade 114 (13%)258 (55%)329 (28%)Unknown4 (4%) Estrogen receptor Positive104 (99%)Negative0 (0%)Unknown1 (1%) Progesterone receptor Positive91 (87%)Negative13 (12%)Unknown1 (1%) HER2 Score = 068 (65%)Score = 125 (24%)Score = 25 (5%)Score = 36 (5%)Unknown1 (1%) Menopause Pre-menopause33 (31%)Post-menopause72 (69%)Unknown0 (0%) Surgery No-mastectomy (lumpectomy)105 (100%)Mastectomy0 (0%)Unknown0 (0%) Hormone therapy Tamoxifen (Tmx)31 (30%)Aromatase inhibitor (AI)33 (31%)Tmx-> AI39 (37%)None/unknown2 (2%) XRT Breast conserving RT101 (95%)No RT3 (3%)Chest-wall RT1 (1%)Unknown1 (1%) Adjuvant chemotherapy No22 (21%)Yes78 (68%)Unknown1 (1%) Tumor size (mm) 22 (16.5, 25); 22.3 ± 7.5Unknown0 # of LN Mets 3.0 (1.0, 6.0); 4.5 ± 4.5Unknown1 LVI Yes28 (27%)No68 (65%)Unknown9 (8%)5 Year survival88% (84%, 93%)5 Year recurrence free75% (69%, 82%)1 n (%); Median (IQR); Mean ± SD Clinicopathologic characteristics of node-positive patients treated with breast-conserving surgery 1 n (%); Median (IQR); Mean ± SD We confirmed the findings of JAG1 expression using an independent 96-sample, 48 patient TMA (Reveal Biosciences), with samples representing a range of breast cancer subtypes including both lymph node negative (pN0) and pN+ patients (Supplementary Figure S1A and B). Consistent with previous reports [13, 15], breast cancer tumors from patients with affected lymph nodes (pN1-3) show increased JAG1 positivity compared to cases without lymph node metastasis (pN0) (Supplementary Figure S1C–D). The percentage of JAG1 expressing tumor cells and JAG1 H-score was significantly higher in lymph nodes than matched primary tumors, confirming the results in ER+ breast cancer patients in a more extensive range of breast cancer subtypes (Supplementary Figure S1E–G). To assess the translational potential of JAG1 detection modalities via traditional immunohistochemical techniques with manual scoring, we compared our OPAL multiplex staining with clinically applicable IHC staining quantified by a pathologist with clinical experience (Supplementary Figure S2A). Scoring extremes, i.e. ‘absent’ versus ‘moderate/strong’ IHC intensity scoring, significantly segregated with OPAL multiplex H-scores (Supplementary Figure S2B). However, there was large variability, likely due to lower sensitivity of manual scoring compared to AI-assisted OPAL multiplex staining incorporating Pan-CK+ assisted identification of tumor cells. We assessed correlations between tumor JAG1 and tumor lymphangiogenesis using the OPAL multiplex staining approach, which allowed visualization of the lymphatic vessels. We identified PDPN+ LYVE1 + lymphatic vessels within the tumor and surrounding tumor microenvironment and we defined areas of LVI, characterized by tumor cells nested fully within lymphatic vessels (Fig. 1G). In breast cores, the tumor JAG1 H-score did not significantly correlate with the area of lymphatic vessels or lymphovascular invasion (Fig. 1H and J). In the lymphatic cores, lymphatic area in the stroma did not correlate with JAG1 H-score, but within tumoral regions, there was a small but significant inverse correlation between lymphatic area and JAG1 H-score (Fig. 1I). These results suggest that tumoral JAG1 expression does not significantly affect tumor-associated lymphangiogenesis. However, a large fraction of tumor lymphangiogenesis occurs in the peritumoral region [30], while TMA cores generally target tumor masses, and our analysis may not have included tissue samples fully reflective of tumor lymphangiogenesis. Taken together, our data show that in multiple contexts, the tumor cells that invade the lymph node have higher JAG1 expression compared to the JAG1 levels of primary breast tumors. Expression of JAG1 by breast and mammary tumor cells promotes lymphovascular invasion in a mouse model of breast cancer The presence of cancer cells that have invaded the lymphatic vasculature of mammary tissue correlates with poor patient outcomes [31, 32]. The enrichment of JAG1 in cancer cells that have invaded tumor-draining lymph nodes suggested that JAG1 actively promotes entry of breast cancer cells into the lymphatic system. To investigate JAG1 in lymphovascular invasion, we used orthotopic implantation of MDA-MB-231-D3H2LN (MDA231-LN) [33], a lymph node-tropic derivative of the widely used human MDA-MB-231 cell line which expresses high levels of JAG1 (Supplementary Figure S3). To examine tumor growth and dissemination of MDA231-LN cells with and without JAG1, we used CRISPR/Cas9 to target JAG1 in MDA231-LN cells. We identified subclones with homozygous frameshift mutations that eliminated production of JAG1 protein (JAG1KO Clones 1–5) (Supplementary Figure S4). Independent JAG1KO clones had no consistent changes in growth rates in vitro or differences in the levels of Notch-related genes (Supplementary Figure S5). Subclones of tumor cell lines can exhibit phenotypic and transcriptional characteristics that are distinct from the more heterogenous parental line (Supplementary Figure S5) [34, 35]. Therefore, to directly compare behavior of JAG1-expressing and non-expressing cells within individual clonal lines, we introduced either a rescue vector expressing JAG1 (JAG1LV) or an empty lentivirus vector (VectorLV) into JAG1KO clones (Fig. 2A), thus creating matched MDA231-LN cell lines with and without JAG1. To aid in tracking dissemination, we transduced MDA231-LN cells with an RFP expression construct. Fig. 2Tumor JAG1 expression promotes lymphovascular invasion (LVI) of human and mouse breast cancer cells. A JAG1 expression was induced in human MDA231-LN JAG1KO Clone 1 line (top, also see Supplementary Figure S2E) or in murine 66cl4 mammary cancer cell line (bottom) via lentiviral transduction of a human JAG1 expression vector (JAG1LV) or empty vector (VectorLV) control. All images for each cell line come from a single Western blot and uncropped blots are available in supplemental materials. B Diagram of experimental mouse, indicating location of RFP+ tumor implantation in the fourth mammary gland and locations of lymph nodes analyzed by lymphangiography. Representative wholemount image of exposed subiliac area in a tumor bearing mouse with subiliac node (LN) and tumor (T) labeled. line Scale bar, 1500 μm. C JAG1LV and VectorLV transduced MDA231-LN and 66cl4 cells were orthotopically implanted into the fourth mammary fatpad of NCG (NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl) and BALB/c female mice, respectively, and permitted to grow to indicated final tumor volume or weight. Final tumor size at harvest and placement showed no difference between groups. Unpaired student’s t-test. D MDA231-LN lymphovascular invasion (LVI) was assessed in NCG using wholemount microscopy of ipsilateral abdominal and axillary fatpads. Representative images (top) and quantification (bottom). Scale bar, 1000 μm. LVI was quantified as the percentage of FITC-positive area that showed RFP signal, normalized to control (yellow arrowheads = LVI; white arrowheads= lymphatics without LVI). Each data point represents individual mouse. Unpaired t-test. E 66cl4 tumor cell LVI assessed in BALB/c mice in sections of the tissue surrounding the subiliac lymph node. Scale bar, 100 μm (representative images, top), or 20 μm (detail indicated by square). Representative images (top) and quantification (bottom). LVI was quantified as the percentage of FITC-positive area that showed RFP signal, normalized to control. Each data point represents an individual mouse. Unpaired t-test Tumor JAG1 expression promotes lymphovascular invasion (LVI) of human and mouse breast cancer cells. A JAG1 expression was induced in human MDA231-LN JAG1KO Clone 1 line (top, also see Supplementary Figure S2E) or in murine 66cl4 mammary cancer cell line (bottom) via lentiviral transduction of a human JAG1 expression vector (JAG1LV) or empty vector (VectorLV) control. All images for each cell line come from a single Western blot and uncropped blots are available in supplemental materials. B Diagram of experimental mouse, indicating location of RFP+ tumor implantation in the fourth mammary gland and locations of lymph nodes analyzed by lymphangiography. Representative wholemount image of exposed subiliac area in a tumor bearing mouse with subiliac node (LN) and tumor (T) labeled. line Scale bar, 1500 μm. C JAG1LV and VectorLV transduced MDA231-LN and 66cl4 cells were orthotopically implanted into the fourth mammary fatpad of NCG (NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl) and BALB/c female mice, respectively, and permitted to grow to indicated final tumor volume or weight. Final tumor size at harvest and placement showed no difference between groups. Unpaired student’s t-test. D MDA231-LN lymphovascular invasion (LVI) was assessed in NCG using wholemount microscopy of ipsilateral abdominal and axillary fatpads. Representative images (top) and quantification (bottom). Scale bar, 1000 μm. LVI was quantified as the percentage of FITC-positive area that showed RFP signal, normalized to control (yellow arrowheads = LVI; white arrowheads= lymphatics without LVI). Each data point represents individual mouse. Unpaired t-test. E 66cl4 tumor cell LVI assessed in BALB/c mice in sections of the tissue surrounding the subiliac lymph node. Scale bar, 100 μm (representative images, top), or 20 μm (detail indicated by square). Representative images (top) and quantification (bottom). LVI was quantified as the percentage of FITC-positive area that showed RFP signal, normalized to control. Each data point represents an individual mouse. Unpaired t-test We implanted 2.0 × 105 MDA231-LN Clone 1 VectorLV and JAG1LV cells into the fourth mammary fat pad of NCG (Non-Obese Diabetic Prkdcem26Cd52Il2rgem26Cd22/NjuCrl) immunodeficient mice (Charles River, Fig. 2B). Tumors from VectorLV and JAG1LV cells were allowed to grow for 28 days, at which point tumors were injected with Dextran-FITC to initiate lymphatic mapping (Fig. 2B). At this point, there were no significant differences in tumor mass (Fig. 2C). We employed lymphangiography to track sequential steps of lymphogenous metastasis, from lymphatic capillaries, to collecting LECs, to the subcapsular sinus where tumor cells first invade, and then further into the lymph node (Supplementary Figure S6). For lymphangiography, we injected 2000 kDa Dextran-FITC, which is preferentially taken up by the lymphatic vessels [36, 37], within the mammary tumor. We confirmed that 2000 kDa Dextran co-localized to LYVE1 + and PDPN+ lymphatic vessels (Supplementary Fig. 11). This approach recapitulates human lymphatic mapping as performed during surgical approaches employed during node removal in the axilla (usually lymphoscintigraphy and/or methylene blue). In immunocompetent mice, we traced to the subiliac lymph node found at the intersection of the subcutaneous abdominal veins (also often called the inguinal lymph node), which is the primary tumor draining lymph node (tdLN) in our model, and additionally examined the axillary/brachial lymph nodes in the thoracic fat pads, which are further downstream of this lymphatic flow (Fig. 2B) [6, 38–41]. In immunodeficient NCG mice, which have underdeveloped lymph nodes [42], we traced lymphatic vessels in corresponding anatomical regions. The subiliac region and thoracic fat pads ipsilateral to the implanted tumor were examined via whole-mount fluorescence imaging, and we quantified the mean RFP (tumor cell) fluorescence that mapped to regions of FITC+ lymphatics (Fig. 2D and Supplementary Figure S6). We confirmed that overlapping RFP signal represented lymphovascular invasion by staining thick sections with tumor marker PanCK (Supplementary Figure S6C). We found that NCG mice with JAG1LV tumors had significantly greater disseminated tumor cells in the lymphatic vasculature surrounding the ipsilateral subiliac, brachial, and axillary lymph node regions than VectorLV controls (Fig. 2D). To study the role of JAG1 in lymphovascular invasion in immunocompetent mice, we used BALB/c syngeneic 66cl4 mouse mammary tumor cells, which are derived from the same spontaneous mammary tumor as 4T1 cells and which exhibit lymphogenous metastasis [43]. We transduced 66cl4 cells with an RFP expression vector to facilitate detection. 66cl4 cells show low levels of endogenous JAG1 expression (Supplementary Figure S7). JAG1-expressing and JAG1-low lines were created with JAG1LV and VectorLV vectors, respectively, without prior knockout of JAG1 (Fig. 2A). 2.0 × 106 cells were injected into the fourth mammary fat pad of BALB/c mice, allowed to grow for 7 days (approximately 40 mm3), and then mice were subjected to lymphangiography as described above (Fig. 2B). Tumor size at harvest and the distance between tumor and subiliac lymph node was consistent between groups (Fig. 2C). The subiliac lymph nodes were cut into thick 30 μm cryosections, sections were stained with anti-RFP and the RFP positive area was quantitated in the dextran-mapped lymphatic vasculature surrounding the subiliac lymph node (Fig. 2E). Higher levels of JAG1 expression by tumor cells was associated with significantly increased lymphovascular invasion of 66cl4 cells in the region surrounding the subiliac lymph node (Fig. 2E). Increased lymph node metastasis can be associated with or secondary to an increase in tumor lymphangiogenesis [6, 32, 44, 45]. Our examination of human samples suggested that tumoral JAG1 expression does not associate with increased tumor lymphangiogenesis (Fig. 1H). To experimentally determine whether the observed increase in lymphovascular tumor cell invasion of JAG1 expressing cells is due to increased intra- or peri-tumoral lymphangiogenesis, we examined the density of lymphatic vasculature in or near implanted tumors. In 66cl4 JAG1LV and VectorLV tumors, we quantitated the amount of lymphatic vasculature using overlapping expression of two established LEC markers, Lyve1 and Vegfr3. No significant difference in Lyve1 + Vegfr3+ area was observed between JAG1-expressing tumors versus control tumors, supporting a model where JAG1 does not increase lymphangiogenesis (Supplementary Figure S8A–B). We also examined lymphatic function of intratumoral lymphatics by injecting 2000 kDa Dextran-FITC into tumors and quantitating the uptake and transport of fluorescent marker through the lymphatic system [6, 41]. We did not see an increase in the peri- and intra-lymphatic function in MDA231-LN tumors expressing JAG1 (Supplementary Figure S8C–D), suggesting that the JAG1-induced increase in lymphovascular invasion is not due to alterations in the amount or function of tumor lymphatic vessels in these models. JAG1 expression is not associated with hematogenous circulating tumor cells (CTCs) levels Metastatic cells may reach the lymph node via lymphogenous spread or by hematogenous dissemination of circulating tumor cells (CTCs). Previous studies have shown that high JAG1 expression in CTCs is associated with shorter survival times in patients with breast cancer treated for bone metastasis, but such analysis does not speak to the total amount of CTCs present [15]. To examine whether tumoral JAG1 expression affected the numbers of CTCs in the mouse models we used, 150 µl of blood was collected retro-orbitally from tumor-bearing mice of consistent tumor size (Fig. 2C). After RBC lysis, RFP+ CTCs were detected via flow cytometry (Supplementary Figure S8E). CTCs were detectable from both MDA231-LN and 66cl4 tumors, but JAG1 expression did not affect the number of CTCs relative to the total blood cell count (Supplementary Figure S8E). These findings suggest that JAG1 did not play a role in the number of CTCs found in hematogenous circulation. JAG1 promotes lymph node metastasis in a syngeneic mouse model of breast cancer Having discovered that JAG1 expression increases tumor invasion into lymphatic vessels, we assessed lymph node metastasis by evaluating tumor cells that have disseminated into the tumor-draining subiliac lymph node (Supplementary Figure S6). 2.0 × 106 RFP+ JAG1LV and VectorLV 66cl4 tumor cells were implanted in the fourth mammary fat pad of BALB/c mice and allowed to grow for one week, at which time, lymphatic mapping was performed using tumors of equal size for analysis (Fig. 3A and B). We examined the subiliac lymph node with tumor marker RFP and tumor/epithelial marker PanCK to determine the total area of metastatic lesion (high magnification, Fig. 3C and Supplementary Fig. 6C; low magnification overview, Fig. 3D). We found that JAG1 expression in 66cl4 tumor cells significantly increased the area of lymph node metastasis in the tdLN by ~ 2.68 fold (Fig. 3E). Fig. 3Tumor JAG1 expression promotes VEGFR3-dependent 66cl4 lymph node metastasis in BALB/c mice. A JAG1LV and VectorLV transduced 66cl4 cells were orthotopically implanted into the fourth mammary fatpad of BALB/c female mice. The subiliac node was harvested seven days post-implantation. B Tumor volumes at endpoint when lymph nodes were harvested in A. C High powered confocal images of medullary sinus of lymph node in tumor free or tumor-bearing mouse. Medullary sinus representative location represented in VEGFR3-stained image on left in white box. Scale bar, 1000 μm. D Representative fluorescence images of subiliac lymph nodes from mice harboring 66cl4 JAG1LV and VectorLV tumors of comparable size. Dotted white line outlines the lymph node. Scale bar, 100 μm. E Percent area of lymph node area staining positive for metastatic markers PanCK and RFP (area of metastasis), with each dot representing the average of separate lymph node. Unpaired t-test. F Percent area of lymph node area staining positive for metastatic markers PanCK and RFP in MAZ51 treated mice, with each dot representing the average of separate lymph node. One way ANOVA. G Percent area of lymph node area staining positive for metastatic markers PanCK and RFP with sFLT4 mediated Vegf-c blockade, with each dot representing the average of separate lymph node. One way ANOVA. H Representative images of peritumoral dextran drainage, quantified by area in I. One way ANOVA. Scale bar, 100 μm. J Perinodal dextran and intranodal dextran collection (within dotted lines) at endpoint, quantified in K. One way ANOVA. Scale bar, 100 μm Tumor JAG1 expression promotes VEGFR3-dependent 66cl4 lymph node metastasis in BALB/c mice. A JAG1LV and VectorLV transduced 66cl4 cells were orthotopically implanted into the fourth mammary fatpad of BALB/c female mice. The subiliac node was harvested seven days post-implantation. B Tumor volumes at endpoint when lymph nodes were harvested in A. C High powered confocal images of medullary sinus of lymph node in tumor free or tumor-bearing mouse. Medullary sinus representative location represented in VEGFR3-stained image on left in white box. Scale bar, 1000 μm. D Representative fluorescence images of subiliac lymph nodes from mice harboring 66cl4 JAG1LV and VectorLV tumors of comparable size. Dotted white line outlines the lymph node. Scale bar, 100 μm. E Percent area of lymph node area staining positive for metastatic markers PanCK and RFP (area of metastasis), with each dot representing the average of separate lymph node. Unpaired t-test. F Percent area of lymph node area staining positive for metastatic markers PanCK and RFP in MAZ51 treated mice, with each dot representing the average of separate lymph node. One way ANOVA. G Percent area of lymph node area staining positive for metastatic markers PanCK and RFP with sFLT4 mediated Vegf-c blockade, with each dot representing the average of separate lymph node. One way ANOVA. H Representative images of peritumoral dextran drainage, quantified by area in I. One way ANOVA. Scale bar, 100 μm. J Perinodal dextran and intranodal dextran collection (within dotted lines) at endpoint, quantified in K. One way ANOVA. Scale bar, 100 μm In humans, metastasis is frequently observed after the primary tumor has been removed. We analyzed mice four weeks after surgical resection of the primary tumor (Supplementary Figure S9). Surgical resection was conducted at a uniform tumor size, but at these later time points, differences in the growth rate of the primary tumor became evident, which may confound analysis. When we examined NCG-immunosuppressed mice implanted with MDA231-LN JAG1LV and VectorLV tumor cells four weeks after resection, we observed very large metastatic outgrowths in the lymph node regions, reaching up to 5% of the mouse body weight, which were not significantly different in size between JAG1LV and VectorLV tumors (Supplementary Figure S9A−D). We surmise that the super-physiological metastatic outgrowth observed is due to aggressive tumor growth in immunosuppressed mice. When we examined BALB/c mice with 66cl4 JAG1LV or VectorLV tumors four weeks after surgical resection, JAG1 expression correlated with larger ipsilateral subiliac, axillary, and total lymph node volumes (Supplementary Figure S9E−G). It is unclear whether this difference is a result of greater early regional dissemination, accelerated tumor division, or nodal immune expansion. We confirmed that increase in lymph node size was at least in part due to presence of tumor cell infiltration using PanCK staining (Supplementary Figure S9G). Production of a Vegfr3 inhibitor, sFLT4-Fc, attenuates JAG1-mediated lymph node metastasis Activation of Notch signaling upregulates endothelial expression of Vegfr3 [25–28, 46], a critical regulator of lymphatic specification, growth, and tissue fluid balance [8, 45, 47, 48]. To determine whether JAG1-induced lymphogenous spread depends on Vegfr3 lymphatic endothelial signaling in tumor associated lymphatics or lymph node endothelium, we examined metastasis of 66cl4 JAG1LV and VectorLV control cells in the presence and absence of Vegfr3 inhibition. To inhibit VEGF signaling pharmacologically, we used MAZ51, a tyrosine kinase inhibitor that potently inhibits Vegfr3, but may weakly interfere with other tyrosine kinase receptors [49, 50]. We implanted 66cl4 JAG1LV and VectorLV tumors as previously described in BALB/c mice, and treated with MAZ51 (10 mg/kg) or DMSO vehicle daily for 5 days (Supplementary Figure S10A). There were no significant differences in tumor mass at endpoint (Supplementary Figure S10B). We observed an increase in lymph node metastasis in 66cl4 JAG1LV bearing mice relative to control as expected, and MAZ51 did not significantly affect the level of lymph node metastasis in either group despite a small cohort of mice showed a strong response to MAZ51 in the VectorLV group (Fig. 3F). We pursued an alternate Vegfr3 inhibitory approach by providing a secreted Vegfr3 inhibitor directly from tumor cells. We constructed a Vegfr3 “decoy” consisting of the signal peptide and extracellular domain of Vegfr3 (sFlt4) fused to an Fc domain (referred to as sFlt4-Fc), which is expected to bind and sequester Vegfr3 ligands, primarily VEGF-C (Supplementary Figure S10C−E). We generated 66cl4 cells that expressed and secreted sFlt4-Fc (Supplementary Figure S10D). We then conducted a seven-day tdLN metastasis experiment with 66cl4 cells and evaluated the effect of sFLT4-Fc on JAG1-mediated nodal metastasis. No changes in primary tumor growth within JAG1LV and VectorLV groups were seen (Supplementary Figure S10F). The sFlt4-Fc VEGFR3 inhibitor did not affect metastasis of VectorLV cells, but significantly attenuated the amount of metastatic lesions formed by JAG1LV cells (Fig. 3G). These data indicate that the JAG1-driven increase in metastasis requires Vegfr3 activity. It is not clear why sFlt4-Fc had a stronger effect than MAZ51, but it is possible that local secretion of sFlt4-Fc produced higher effective concentrations of Vegfr3 inhibition or that sFlt4-Fc sequestration of VEGF-C also affected other lymphatic VEGF receptors [51]. To assess potential effects of sFLT4-Fc on lymphatic endothelium, we evaluated whether tumor cell secretion of sFLT4-Fc would attenuate tumor cell migration across cultured LEC monolayers. We did not observe a significant decrease of tumor transendothelial migration as a result of sFLT4 -Fc (Supplementary Figure S10G). Using lymphatic mapping techniques, we quantified lymphatic drainage in the peritumoral region (Fig. 3H–I), in the fat pad lymphatic vessels draining to the tdLN (Supplementary Figure S11A), and in the tdLN (Fig. 3J–K), all of which were unaffected by tumor JAG1 expression or treatment with soluble sFLT4-Fc. We confirmed the specificity of peritumoral, fat-pad, and lymph node dextran localization to the lymphatic vessels using lymphatic specific markers (Supplementary Figure S11B−F). Thus, JAG1-mediated lymph node metastasis likely involves a tumor-lymphatic interaction that does not influence lymphangiogenesis or lymphatic drainage of the tumor in the 66cl4 model. Tumor cell JAG1 promotes transit of tumor cells across lymphatic endothelium A key step facilitating lymphatic invasion is transit of tumor cells across lymphatic endothelium to enter or exit the lymphatic system. We utilized an in vitro system with cultured lymphatic endothelial cell monolayers to assess the ability of JAG1 expressing and non-expressing tumor cells to undergo lymphatic-specific transendothelial migration (TEM). Human dermal lymphatic endothelial cells (HDLEC) were grown to confluence on transwell inserts. JAG1LV and VectorLV cells from two independent RFP+ MDA231-LN JAG1KO clones created by two separate sgRNAs (Clones 1 and 3, respectively, Supplementary Figure S4B) were seeded on top of the HDLEC monolayer and allowed to migrate for six hours to the basal transwell compartment (Fig. 4A). The use of Clone 1 and 3 thus allowed evaluation of cell lines with JAG1 knockout generated by different sgRNAs and potential off-target effects. Fig. 4Tumor-derived JAG1 promotes transendothelial migration across lymphatic monolayers. A Schematic of transendothelial migration experiment consisting of RFP+ MDA231-LN cells migrating across monolayers of HDLEC using 3 μm transwells. B Representative images of VectorLV versus JAG1LV in MDA231-LN Jag1KO Clone 1 transendothelial migration across HDLEC monolayers. Scale bar, 50 μm. C-D Transendothelial migration across HDLEC monolayers with VectorLV versus JAG1LV in MDA231-LN Jag1KO Clone 1 and MDA231-LN Jag1KO Clone 3 cell lines. Unpaired t-test. E MDA231-LN migration when no HDLEC monolayer is present. Unpaired t-test. F Transendothelial migration across HDLEC monolayers in transwells coated with IgG-Fc or Jag1-Fc fibronectin matrix with JAG1KO MDA231-LN Jag1KO Clone 1. Unpaired t-test. G Confocal images of VE-Cadherin immunostained HDLEC monolayers either on IgG-Fc or Jag1-Fc fibronectin matrix. Scale bar, 20 μm. H Quantification of monolayer dissociation in Panel G after IgG-Fc or JAG1-Fc treatment of HDLECs. Unpaired t-test. Scale bar, 20 μm Tumor-derived JAG1 promotes transendothelial migration across lymphatic monolayers. A Schematic of transendothelial migration experiment consisting of RFP+ MDA231-LN cells migrating across monolayers of HDLEC using 3 μm transwells. B Representative images of VectorLV versus JAG1LV in MDA231-LN Jag1KO Clone 1 transendothelial migration across HDLEC monolayers. Scale bar, 50 μm. C-D Transendothelial migration across HDLEC monolayers with VectorLV versus JAG1LV in MDA231-LN Jag1KO Clone 1 and MDA231-LN Jag1KO Clone 3 cell lines. Unpaired t-test. E MDA231-LN migration when no HDLEC monolayer is present. Unpaired t-test. F Transendothelial migration across HDLEC monolayers in transwells coated with IgG-Fc or Jag1-Fc fibronectin matrix with JAG1KO MDA231-LN Jag1KO Clone 1. Unpaired t-test. G Confocal images of VE-Cadherin immunostained HDLEC monolayers either on IgG-Fc or Jag1-Fc fibronectin matrix. Scale bar, 20 μm. H Quantification of monolayer dissociation in Panel G after IgG-Fc or JAG1-Fc treatment of HDLECs. Unpaired t-test. Scale bar, 20 μm MDA231-LN Clone 1 JAG1LV and Clone 3 JAG1LV cells demonstrated increased migration across HDLEC monolayers compared to their paired VectorLV cells (Fig. 4B–D). In the absence of the HDLEC monolayer, MDA231-LN JAG1LV cells did not exhibit increased migration compared to control VectorLV cells (Fig. 4E), indicating that the increased migratory phenotype of JAG1-expressing cells is due to interactions with LEC and not a general increase in migratory activity. To model JAG1 juxtacrine signaling to lymphatic endothelial cells, we coated the transwell with a JAG1-Fibronectin matrix before seeding HDLECs. JAG1-coated transwells caused an increase in the transendothelial migration of MDA231-LN JAG1KO Clone 1 cells compared to IgG-control coated transwells (Fig. 4F). Using the same coating strategy, but this time on optically clear wells, we assessed the HDLEC monolayer integrity after JAG1 exposure. Compared to HDLEC monolayers treated with IgG-control, those treated with JAG1 had greater areas of endothelial dissociation (breaks in VE-Cadherin positive cell junctions, Fig. 4G–H). To directly assess the overall effect of JAG1 on LEC junctions, we have examined VEcadherin staining at LEC borders in vitro. On fibronectin-coated surfaces, LECs form linear zipper-like junctions typical of lymphatic collecting ducts, where VEcadherin is localized to a continuous thin band at the border between cells [52, 53]. Disruption of endothelial junctions causes membrane ruffling and overlapping between cells, resulting in an apparent increase in the width of VEcadherin staining. Overall VE-Cadherin staining intensity did not appear different between groups. We observed no difference in the width of VEcadherin-labeled LEC junctions in JAG1-coated wells. JAG1-induced LEC disassociation may therefore reflect changes in cell-substrate adhesion, monolayer stability, or other post-transcriptional mechanisms (Supplementary Figure S12A). JAG1 promotes lymphatic inflammatory gene expression and transendothelial migration (TEM) Having established that tumor-cell JAG1 enhances lymphatic endothelial barrier transit in an LEC-dependent manner (Fig. 4B–D), we next asked whether JAG1 signaling can directly reprogram lymphatic endothelium toward a pro-permeability, pro-inflammatory state that could facilitate invasion. To test this, we examined transcriptional responses of HDLECs following controlled juxtacrine exposure to immobilized JAG1 ligand in the absence of tumor cells. We examined the effects of juxtacrine exposure of lymphatic endothelial cells to JAG1 on endothelial gene expression. We employed an in vitro tethered ligand assay (TLA) [54] using the ectodomain of human JAG1 fused to the Fc domain and attached to fibronectin-coated dishes, which were then covered with a monolayer of HDLECs. RNA was isolated and sequenced from HDLECs seeded on JAG1-Fc or control-Fc plates. Transcriptional profiles of LECs exposed to JAG1 for six hours were enriched in inflammatory and cytokine genes, including NF-κB, TLR-related, and cytokine signaling pathways (Fig. 5A-B). In these upregulated genes, CCL2, CXCL2, and multiple other secreted proteins have well-established roles in disrupting the endothelial barrier and/or promoting tumor cell entry [55–58]. Changes induced by JAG1 in key inflammatory targets were confirmed by RT-PCR (Fig. 5C). Fig. 5JAG1 promotes lymphendothelial inflammatory signaling. A Heatmap of top-15 differentially expressed genes in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. B GSEA Reactome pathway analysis in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. C qRT-PCR based quantification of relative mRNA expression to confirm RNAseq hits in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. D in silico analysis of the differentially expressed gene set upregulated by JAG1 (DGE) mapped en bloc to human lymph node subcapsular sinus. fLEC=subcapsular floor LEC, Marco-LEC=MARCO+ medullary sinus LECs, Ptx3-LEC=Ptx3 + medullary sinus LECs, Valve=LECs expressing lymphatic valve markers. cLEC, cLECs1, cLECs2, and cLECs3 represent subcapsular ceiling LEC subtypes, with cLECs1-s3 representing subset found in breast cancer patients (data set from Xiang et al. 2020 Front Cardiovasc Med). E Relative mRNA expression of RND1 in JAG1-stimulated HDLECs with Compound E treatment. Unpaired t-test JAG1 promotes lymphendothelial inflammatory signaling. A Heatmap of top-15 differentially expressed genes in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. B GSEA Reactome pathway analysis in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. C qRT-PCR based quantification of relative mRNA expression to confirm RNAseq hits in HDLECs either stimulated by recombinant hJAG1-Fc or hIgG-Fc in tethered-ligand assay. D in silico analysis of the differentially expressed gene set upregulated by JAG1 (DGE) mapped en bloc to human lymph node subcapsular sinus. fLEC=subcapsular floor LEC, Marco-LEC=MARCO+ medullary sinus LECs, Ptx3-LEC=Ptx3 + medullary sinus LECs, Valve=LECs expressing lymphatic valve markers. cLEC, cLECs1, cLECs2, and cLECs3 represent subcapsular ceiling LEC subtypes, with cLECs1-s3 representing subset found in breast cancer patients (data set from Xiang et al. 2020 Front Cardiovasc Med). E Relative mRNA expression of RND1 in JAG1-stimulated HDLECs with Compound E treatment. Unpaired t-test We evaluated the JAG1-TLA differentially expressed gene set (DGE) en bloc against a previously published scRNA-seq of the lymphatic endothelium of lymph nodes [10]. We found that genes enriched by JAG1 (JAG1-induced DGE) are enriched in lymphatic endothelial cells of the subcapsular floor LECs, a specialized immunomodulatory population that form the first point of contact for antigens and tumor cells entering the lymph node [9], and in cLECs2 subcluster, a subcluster found in human breast cancer patients that shows high expression of inflammatory markers [10] (Fig. 5D, see fLEC and cLECs2). Of the genes upregulated by JAG1, RND1 is a well-documented direct target of canonical Notch signaling [54]. Treatment with the gamma-secretase inhibitor (GSI) Compound E significantly attenuated JAG1-mediated upregulation of RND1 in HDLECs after six hours of ligand and drug combined treatment (Fig. 5E). We examined mRNA expression of junctional protein genes, and determined that the majority of junctional proteins show no differences in expression in LECs stimulated by JAG1 (Supplementary Fig. S12B–C). However, JAG1 stimulation significantly upregulates both CLDN5, which is thought to stabilize LEC junctions [59, 60], and inflammatory adhesion proteins SELE and ICAM1, which are VEGFR targets and associated with endothelial activation, leukocyte recruitment, and altered barrier function in inflammatory contexts [61–64]. These alterations suggest that JAG1 could promote both inflammatory barrier disruption and compensatory barrier strengthening. Collectively, these data show that JAG1 signaling in LECs upregulates a broad set of inflammatory, adhesion, and secreted molecules that disrupt lymphatic junctions and facilitate tumor cell entry. JAG1-regulated tumor genes overlap with genes associated with nodal organotropism To examine the tumor-intrinsic changes induced by JAG1, we conducted transcriptional profiling of our tumor cell lines. Unbiased transcriptomic comparison between the parental MDA-MB-231 (MDA231-P) line and MDA231-LN, which had been selectively passaged for lymph node tropism, demonstrates increased tyrosine-kinase signaling (Supplementary Figure S13A) and upregulation of multiple Notch genes including JAG1/2, NRARP, NOTCH3, RBPJ, and MAML3 (Supplementary Figure S13B). To assess the consequences of JAG1 expression within the MDA231-LN nodal tropic line, we sequenced RNA from MDA231-LN JAG1KO Clone 1 VectorLV and JAG1LV (Supplementary Figure S13D). JAG1 expression upregulated several Notch and lymphangiogenic signatures [24] such as JAG1, Notch3, FLT4, HES1, ITGA9, UNC5B, SOX18, many of which echoed the genes upregulated during passage selection for nodal tropism (Supplementary Figure S13C−E). Comparative analysis showed that of the 2866 differentially regulated genes significantly upregulated by JAG1, 26.8% overlapped with those enriched in MDA231-LN compared to parental. Comparison with a curated list of genes associated with lymphangiogenesis [46] showed that many of these genes were upregulated by JAG1 expression (Supplementary Figure S13E). To determine whether transcriptomic responses to JAG1 translated to increased secretion of pro-lymphangiogenic factors and whether these factors were upregulated in both MDA231-LN and 66cl4 cells, we analyzed the composition of conditioned medium of 66cl4 JAG1LV and VectorLV using a cytokine-detection array (Supplementary Figure S13F). We observed that secretion of Wisp1, Cxcl1, Igfbp6, Cx3cl1, and Il-6 was higher in JAG1 expressing tumor cells compared to controls, which correspond to genes identified in RNAseq analysis (complete list of secreted proteins in Supplementary documents). IL-6 is known to induce lymphatic vessel hyperpermeability [65, 66], while Cx3Cl1 and CCL2 induce hyperpermeability in BECs, which may extend to LECs [55, 56]. Cxcl1 promotes transendothelial migration of immune and tumor cells [57, 58]. Collectively, these data suggest that breast cancer expression of JAG1 likely increases the capacity of cancer cells to invade lymphatic vessels via tumor-intrinsic mechanisms in addition to direct JAG1 signaling to LECs. Tumoral JAG1 in lymph nodes predicts disease recurrence Sentinel lymph node biopsy is essential in determining if a patient is pN+. However, adjuvant treatment plans are primarily based on the histological and genomic analysis of the primary tumor despite emerging evidence of tumor-node discordance [11, 13]. Recent clinical trials show that most patients with early breast cancer (cT1-2) do not significantly benefit from complete axillary dissection upfront [67–71], but post-operative risk assessment for disease progression remains an active area of study. We aimed to explore tumor JAG1 as a clinically useful marker in the lymph node that could help further risk-stratify pN+ patients undergoing breast conserving therapy with sentinel lymph node biopsy. We found a discordance in JAG1 protein levels in tumor cells in breast versus node, as established in Fig. 1. We further examined tumor cells within the lymph node specifically in pN+ patients with breast-conserving lumpectomies, but not total mastectomies, to determine if JAG1 expression levels predict clinical characteristics in these patients (Table 1, N = 105). In patients with breast-conserving lumpectomies from the ER + pN+ 284-patient TMA and in patients from the commercial TMA with a range of histological characteristics, tumor cell JAG1 expression levels in the lymph node did not correlate with tumor grade, size, HER2 status, treatment modalities, or histological variables, suggesting that JAG1 expression in disseminated tumor cells is an independent variable (Fig. 6A–C and Supplementary Figure S14). Fig. 6JAG1 expression in lymph node metastatic tumor cells can predict patient disease recurrence in pathologically node positive (pN+) breast cancer patients. A Forest plot demonstrating odds ratios for patient characteristics with lymph node JAG1hi set as risk exposure. JAG1-high and JAG1-low were stratified by upper and low median tumor JAG1 H-scores, respectively. B Tumor sizes were not significantly different in patients with lymph node JAG1-high and JAG1-low. Unpaired t-test. C The amount of lymph node metastasis was not significantly different in patients with lymph node JAG1-high and JAG1-low. Unpaired t-test. D pN+ patients with JAG1-high lymph node tumor cells experienced greater disease recurrence than patients with JAG1-low nodal tumor cells. Log-rank test. E Proportion of patients with regional and distant metastasis in lymph node JAG1-high and JAG1-low groups. Fisher’s exact test. F Time to recurrence is better predicted by JAG1 expression in lymph node samples than from tumor samples coming from breast. G Model depicting role of tumor-derived JAG1 in lymphogenous metastasis JAG1 expression in lymph node metastatic tumor cells can predict patient disease recurrence in pathologically node positive (pN+) breast cancer patients. A Forest plot demonstrating odds ratios for patient characteristics with lymph node JAG1hi set as risk exposure. JAG1-high and JAG1-low were stratified by upper and low median tumor JAG1 H-scores, respectively. B Tumor sizes were not significantly different in patients with lymph node JAG1-high and JAG1-low. Unpaired t-test. C The amount of lymph node metastasis was not significantly different in patients with lymph node JAG1-high and JAG1-low. Unpaired t-test. D pN+ patients with JAG1-high lymph node tumor cells experienced greater disease recurrence than patients with JAG1-low nodal tumor cells. Log-rank test. E Proportion of patients with regional and distant metastasis in lymph node JAG1-high and JAG1-low groups. Fisher’s exact test. F Time to recurrence is better predicted by JAG1 expression in lymph node samples than from tumor samples coming from breast. G Model depicting role of tumor-derived JAG1 in lymphogenous metastasis In patients with breast-conserving lumpectomies, JAG1 H-scores above the median (LN H-scorehi) in tumor cells within the lymph node significantly predicted patient recurrence free survival (RFS) (Fig. 6D). LN H-scorehi patients had greater amounts of both regional and distant metastasis (Fig. 6E). Patients with lymph node JAG1 H-scorelow had a 5-year RFS rate of 87%, and patients with lymph node JAG1 H-scorehi had 5-year RFS rate of 70% (Fig. 6F). JAG1 tumor cell percent positivity in the lymph node also conferred borderline predictive value. Conversely, JAG1 in primary breast tumor samples did not significantly predict RFS either by H-score or by JAG1 tumor cell percent positivity (Fig. 6F). When modeled as a continuous variable, increasing nodal JAG1 expression demonstrated a consistent association with higher recurrence risk (HR 1.37 per standard deviation increase, 95% CI 0.94–1.97; p = 0.097). JAG1 expression in breast tumors can predict lymph node metastasis [14, 15] (Supplementary Figure S1C–D), but our data set already includes pN + as inclusion criteria. Thus, within a data set of pN+ patients with baseline JAG1 positivity, further granular analysis of JAG1 expressed in lymph node cancer cells significantly predicts recurrence free survival (Fig. 6G). Discussion Lymph node metastasis represents a clinically meaningful transition in breast cancer, signaling both regional spread and increased risk of distant relapse [1, 2, 72, 73]. While modern management has moved beyond the strict Halstedian view that nodal disease is solely a sequential surgical problem [67–70, 74], the biology of lymph node metastasis remains highly relevant to recurrence risk and systemic progression. As central sites of antigen presentation and immune coordination, lymph nodes may also shape tumor–host interactions once metastatic colonization occurs [3, 5]. In this study, we identify tumor-cell JAG1 as an important promoter of lymphogenous dissemination in breast cancer. Multiplex analysis of matched primary tumors and lymph node metastases demonstrated enrichment of JAG1 in nodal tumor cells and JAG1 increased lymphovascular invasion and subsequent lymph node metastasis across orthotopic human and syngeneic mouse models. Our experimental data suggest that one mechanism by which tumor-cell JAG1 may facilitate lymph node metastasis is through altered interactions with the lymphatic endothelium. Suppression of VEGFR signaling, likely targeting endothelial VEGF receptors, was sufficient to block the effects of JAG1 expression in tumor cells. JAG1-expressing tumor cells demonstrated enhanced migration across cultured lymphatic endothelial monolayers, consistent with their increased ability to undergo lymphovascular invasion [72]. These effects appeared to be mediated by changes in lymphatic endothelium, as LEC exposure to JAG1 was sufficient to enhance tumor cell transmigration and LEC barrier disruption. Exposure to JAG1 caused cultured LECs to shift to a more proinflammatory state resembling subcapsular LECs from breast cancer patients and increased expression of tumor cell adhesion molecules. These results suggest that JAG1 expression creates permissive signaling interactions that allow tumor cells to enter lymphatic vessels during early metastatic spread. Notably, JAG1 did not measurably increase lymphangiogenesis in either mouse or clinical samples, suggesting that effects on endothelial permeability, remodeling, contractility, or tumor–endothelial crosstalk may be more relevant than expansion of lymphatic vessel density. As a clinical indicator, our data showed that higher JAG1 expression in tumor cells invading the lymph node is associated with reduced recurrence-free survival among node-positive patients. Together, these findings support a functional role for JAG1 as a driver of regional metastatic progression and suggest that JAG1 expression within nodal metastases may provide clinically meaningful prognostic information. These data warrant further study specifically of JAG1 as a biomarker, and more generally of the clinical utility of molecular examination of resected lymph nodes. Recent studies have emphasized that lymphatic endothelial cells are active participants in tumor progression rather than passive conduits for metastatic spread [6, 9, 75]. Lymphatic endothelial cells regulate permeability, chemokine gradients, immune trafficking, and stromal remodeling within both peripheral lymphatics and the lymph node subcapsular sinus—and in turn, are remodeled by the metastatic tumor cells [75]. Several transcriptional programs induced by JAG1 in cultured lymphatic endothelial cells in our study overlap with pathways linked to inflammatory activation and endothelial remodeling. Ccl2, Cxcl2, Bmp2, Serpine1, Pim1, Rgs16, Junb, Cdkn1a, Sqtsm1, and Jag1 are all lymphatic endothelial genes that are known previously to increase in LECs in metastatic settings and are also increased by JAG1 ligand presentation in our studies [9]. While these data do not establish a singular downstream mediator, they are consistent with a model in which tumor-cell JAG1 conditions lymphatic endothelium toward a state that favors tumor cell entry and transit. The relationship between JAG1 and VEGFR signaling warrants cautious interpretation. In our models, soluble sFLT4-Fc attenuated JAG1-associated lymph node metastasis, whereas small-molecule VEGFR3 inhibition either did not reproduce this effect or reproduced it incompletely. These findings suggest the suppression of the JAG1 phenotype by VEGFR inhibition may not be explained by VEGFR3 kinase signaling alone. Because soluble VEGFR3 antagonism can alter ligand availability within broader VEGF-C/VEGF-D signaling networks, our data more support involvement of VEGFR-associated lymphatic signaling, but do not definitively establish a VEGFR3-specific mechanism. Genetic ablation of VEGFR3 from lymphatic endothelium in the tumor microenvironment would be highly informative in establishing receptor and cell type specificity. An important limitation of the present study is that it does not fully distinguish lymphatic endothelial responses to JAG1 presented by tumor cells from tumor-intrinsic changes inducing by JAG1 signaling within tumor cells. In addition to the tumor-independent JAG1-induced changes observed in LEC, JAG1 expression altered expression and secretion of a range of pro-lymphangiogenic and pro-inflammatory proteins in tumor cells. Both tumor cell intrinsic and tumor-LEC juxtacrine signaling may therefore contribute to the metastatic phenotype observed here. Future studies separating tumor-autonomous, endothelial-autonomous, and immune microenvironmental effects of JAG1 will be important. In a comprehensive analysis of JAG1 expression in primary breast cancer and matched lymph node samples derived from 105 patients receiving breast conserving therapy, whose outcomes were followed over sixteen years, we found that JAG1 expression was enriched within nodal tumor cells relative to paired primary tumors. This enrichment may reflect selective dissemination of pre-existing JAG1-high subclones, adaptive upregulation within the lymph node microenvironment, or both. Higher JAG1 expression in lymph node metastases—but not primary tumors—was associated with worse recurrence-free survival. These findings raise the possibility that molecular characterization of nodal metastases may provide prognostic information not captured by analysis of the primary tumor alone [29, 76–78]. This concept is gaining traction in the literature: for example, next generation sequencing of synchronous node-breast pairs demonstrates NOTCH3amp, SNV mutational discordance in breast cancer nodal metastases in patients with extensive nodal disease [79]. However, our preliminary testing of IHC-based JAG1 evaluation by a clinical pathologist suggested that semi-quantitative IHC analysis does not yet have the precision to stratify patient populations. Validation in modern multi-institutional cohorts with standardized staining and scoring methods will be necessary to define whether JAG1 can be incorporated into risk stratification for node-positive disease. Our findings provide evidence that JAG1 contributes to lymphatic dissemination and identifies a subset of node-positive patients at higher risk of recurrence. Further study of JAG1-dependent signaling within tumor cells, lymphatic endothelium, and the lymph node microenvironment may help refine prognostic assessment and uncover new therapeutic opportunities in regional metastatic breast cancer. Methods Sex as a biological variable Our study exclusively examined female mice because the disease modeled is more represented in females. Mice All mouse experiments were performed under approved UIC IACUC protocols. Female NCG mice (NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl) were purchased from Charles River Laboratories and were used between 6 and 8 weeks of age for cancer cell injections. BALB/c (BALB/cAnNCr) mating pairs were purchased from Charles River Laboratories and bred in the pathogen-free Biological Resources Lab at the UIC Hospital. 6–8 week old female mice were used for experimentation. All mice were kept on 14 h light/10 h dark cycle and were provided with feed and water ad libitum. UIC has an Animal Welfare Assurance Number A346.01 with the Office of Laboratory Animal Welfare, NIH. In-depth mouse experimental details are provided in the Supplementary Material. Opal-dye based multiplexing tumor microarray analysis All patients whose samples were included in the TMAs have given written informed consent for their archival tissue to be used for scientific research in accordance with the Declaration of Helsinki, and the TMA construction was approved by the responsible local Ethical Committees. Commercial 96-sample TMA (Reveal Biosciences) contained core FFPE sections of 48 patient’s primary tumors and synchronous lymph nodes in 36 of the cases. De-identified clinical and histological patient information was provided by the Reveal Biosciences. Our proprietary TMA included 284 synchronous samples of early ER + BC and affected lymph nodes from cases operated in Donostia University and Onkologikoa Hospitals from 2000 to 2014. The mean duration of follow-up was 99.88 months, standard deviation 37.05, and the median duration of follow-up 105.6 months. Complete histopathological information, date, and cause of death, as well as the date of local and/or distant relapse were available for all the patients. Both TMAs were processed with multiplex staining with Opal 6-Plex Detection Kits (Akoya Biosciences, antibody descriptions in Supplementary Material). The slides were then scanned at 20x magnification an Akoya Biosciences Vectra 3. The raw images were unmixed in the inform, stitched in Indica Labs HALO, and segmented into TMA spots. The HALO AI MiniNet Classifier was used to create an epithelial mask based on either E-cadherin or PanCK channels, depending on the TMA. The ROI’s were then confirmed for accuracy and manually segmented when required. JAG1 immunoreactivity in tumor and lymph node samples was evaluated by the semiquantitative method of H-score (or “histo” score) [80, 81]. In the H-score, staining intensity is classified on a scale from 0 to 3 (0 = negative, 1 = weak, 2 = moderate, 3 = strong), and this value is then multiplied by the percentage of stained cells to generate a score ranging from 0 to 300. H-score was calculated with equation H-score = (1 × %CellsWeak Staining) + (2 × %CellsModerate Staining) + (3 × %CellsStrong Staining). Figure diagrams Diagrams were generated using BioRender online software (BioRender.com). Statistical analysis The data are represented using box plots overlaid with individual data points. Each box plot includes key statistics such as maximum and minimum values, lower quartile (Q1), upper quartile (Q3), and the median. Unless specified otherwise, each data point in the mouse experiments corresponds to an individual mouse, tumor, or tissue sample. Error bars in the qRT-PCR relative mRNA expression bar graphs denote the standard error of the mean. Throughout the experiments, statistical analysis involved the use of unpaired Student’s t-tests, paired Student’s t-tests, one-way ANOVA, and two-way ANOVA, as indicated in the figure legends. To assess whether JAG1 expression was elevated compared to tumor tissue, a paired t-test was performed. For the analysis of Recurrence-Free Survival (RFS) based on clinical factors, Kaplan-Meier methods were used, and group comparisons were conducted using the Log-rank test. Multivariate analysis to construct a predictive model for patient survival employed the Cox proportional hazard model. The backward selection method, with a selection threshold of 0.1 was used, with the JAG1 variable included as a fixed factor in the model. Separate models were built using H-Score in lymph nodes, H-Score in breast tumors, JAG% in lymph nodes, and JAG1% in breast tumors as predictive factors. Significance was determined with a threshold of P < 0.05. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 - uncropped Western blots Supplementary Material 2 - uncropped Western blots