Podoplanin+ tumor lymphatics are rate limiting for breast cancer metastasis

Metastatic dissemination employs both the blood and lymphatic vascular systems. Solid tumors dynamically remodel and generate both vessel types during cancer progression. Lymphatic vessel invasion and cancer cells in the tumor-draining lymph nodes (LNs) are prognostic markers for breast cancer metastasis and patient outcome, and tumor-induced lymphangiogenesis likely influences metastasis. Deregulated tumor tissue fluid homeostasis and immune trafficking associated with tumor lymphangiogenesis may contribute to metastatic spreading; however, the precise functional characterization of lymphatic endothelial cells (LECs) in tumors is challenged by the lack of specific reagents to decipher their rate-limiting role in metastasis. Therefore, we generated novel transgenic mice (PDPN promoter-driven Cre recombinase transgene [PDPN-Cre] and PDPN promoter-driven thymidine kinase transgene [PDPN-tk]) that allow for the identification and genetically controlled depletion of proliferating podoplanin (Pdpn)-expressing LECs. We demonstrate that suppression of lymphangiogenesis is successfully achieved in lymphangioma lesions induced in the PDPN-tk mice. In multiple metastatic breast cancer mouse models, we identified distinct roles for LECs in primary and metastatic tumors. Our findings support the functional contribution of primary tumor lymphangiogenesis in controlling metastasis to axillary LNs and lung parenchyma. Reduced lymphatic vessel density enhanced primary tumor lymphedema and increased the frequency of intratumoral macrophages but was not associated with a significant impact on primary tumor growth despite a marked reduction in metastatic dissemination. Our findings identify the rate-limiting contribution of the breast tumor lymphatic vessels for lung metastasis.


Introduction
Metastasis is responsible for 90% of deaths of breast cancer patients [1,2]. The contribution of both cancer cells and stromal cells (such as fibroblasts, endothelial cells, pericytes, and immune cells) is important for cancer development and metastasis, including in breast cancer [2][3][4][5].
The lymphatic system, consisting of lymphatic vessels and lymphoid organs, is an essential regulator of tissue fluid homeostasis, immune cell trafficking, and immunological surveillance [6][7][8]. Lymphatic vessels can be divided into several subtypes: initial lymphatic with incomplete basement membrane and no pericyte/smooth-muscle-cell coverage; transitional precollecting lymphatics; and larger collecting lymphatics with a complete basement membrane and smooth muscle investment. In the context of cancers, these different types of lymphatic vessels can be actively regulated by tumor-derived growth factors [8,9]. Lymphangiogenesis, the formation of new lymphatic vessels, has been associated with metastasis of solid tumors to lymph nodes (LNs) and distant organs [8][9][10][11][12][13]. Recent studies demonstrate that lymphatic vessels undergo dynamic remodeling, including lymphangiogenesis and lymphatic enlargement, which facilitates tumor metastasis [14][15][16]. Furthermore, two recent studies further confirmed the dissemination of cancer cells from LN to distant organs through LN blood vessels in tumor-bearing mice [17,18]. Previous studies using various transgenic mouse models employing vascular endothelial growth factor C/D (VEGF-C/D) overexpression or VEGF-C/D trap suggested a potential role for lymphangiogenesis in cancer progression [19][20][21][22][23][24]. Given that VEGF-C/D can also target nonlymphatic processes, we aim at establishing new mouse models that can specifically target lymphangiogenesis via genetic depletion of proliferating lymphatic endothelial cells (LECs).
Lymphatic vessel markers include Prox1 [25], the lymphatic vessel endothelial hyaluronan receptor-1 (LYVE1) [26], podoplanin (PDPN), and VEGF receptor 3 (VEGFR3) [27]. PDPN, a 43-kDa membrane protein, is present in podocytes [28] and is one of the most widely employed markers of LECs [29,30]. To functionally evaluate the specific role of lymphatic vessels in cancer progression and metastasis, we generated novel transgenic mice that express the herpes simplex virus (HSV) thymidine kinase (tk) under the control of the PDPN gene promoter (PDPN-tk mice). Upon ganciclovir (GCV) administration to PDPN-tk mice, PDPNpositive cells that also express tk will convert GCV into a nucleoside analog that irreversibly arrests DNA replication, resulting over time in the depletion of proliferating PDPN-expressing LECs. Here, we demonstrate that the depletion of proliferating PDPN-expressing LECs significantly inhibits lymphangiogenesis in mammary tumors, resulting in decreased distant metastasis without an impact on primary tumor growth.

Generation and characterization of the PDPN-Cre and PDPN-tk mice
The PDPN-tk mouse model (BALB/c background) was generated using a 4-kb PDPN promoter sequence cloned and ligated to HSV viral tk sequence using the topoisomerase I-activated pCR2.1-TOPO (pCR2.1-TOPO) vector. The PDPN-Cre mouse model was generated using the same PDPN promoter sequence cloned and ligated to Cre recombinase sequence (Fig 1A). The final constructs were confirmed by DNA sequencing. To examine the specificity of the PDPN promoter, we generated the PDPN-Cre; LoxP-Stop-LoxP (LSL)-yellow fluorescent protein (YFP) transgenic mice (BALB/c background) to lineage trace the PDPN + cells. The YFP expression colocalized with LYVE1-or PDPN-positive lymphatic endothelium in normal organs (S1 Fig). Additionally, the eyes of PDPN-Cre; LSL-YFP mice exhibited YFP/ green fluorescent protein (GFP) expression (Fig 1B), consistent with previous observation using Prox1-GFP transgenic mice [31]. Primary LECs were isolated as previously documented [32] from incomplete Freund's adjuvant (IFA)-induced benign mouse lymphangioma. Briefly, LECs were isolated from hyperplastic lymphatic vessels, cultured, and expanded (S2A Our previous study identified that blood-vascular endothelial-cell-specific deletion of β1 integrin (Tie2-Cre; β1 integrin (Int) loxP/loxP mice) resulted in embryonic lethality due to severe vascular defects [33], while others demonstrated that blood-vascular endothelial-cell-specific deletion of transforming growth factor (TGF) β type II receptor (cadherin 5 promoter-driven tamoxifen-inducible Cre recombinase transgene (Cdh5-CreERT2); TGFBRII loxP/loxP mice) also resulted in embryonic lethality [34]. In contrast, PDPN-Cre; β1 Int loxP/loxP and PDPN-Cre; TGFBRII loxP/loxP mice were born in the expected Mendelian ratio without any noticeable abnormality/defect (S3A and S3B Fig), supporting the specificity of the PDPN-Cre transgenic in targeting gene deletion in lymphatic vessels and not blood vessels.
To evaluate the efficacy of the PDPN-tk transgene, PDPN + LECs separated from lymphangioma tissues of PDPN-tk or wild-type (WT) mice were cultured (S4A Fig) and treated with increasing concentrations of GCV (Fig 1C). A dose-dependent depletion of LECs (derived from PDPN-tk mice but not WT mice) was observed, reaching 56% of LEC depletion at an exposure of 50 μM GCV. In addition, in vivo administration of GCV to PDPN-tk mice (daily 50 mg/kg body weight) inhibited the formation of IFA-induced lymphangioma when compared to control (WT) mice (Fig 1D). As previously documented [32], IFA-induced benign mouse lymphangioma formed white solid masses on the abdominal surface of the diaphragm and on the surface (under the Glisson's capsule) of the liver (Fig 1D). The hyperplastic LECs forming these masses present with enlarged lumens that are distinct from adipose tissue (S4B Fig). LEC lumen formation within lymphangioma tissue was specifically impaired in the PDPN-tk mice when compared to control mice (Fig 1E), indicative of the depletion of hyperplastic LECs. A significant and specific decrease in proliferating LECs is recorded in lymphangioma of PDPN-tk mice compared to control mice (Fig 1F and 1G). Notably, lymphangioma formation was not altered in control mice, including WT mice (with or without GCV treatment) and non-GCV-treated PDPN-tk mice. The depletion of proliferating LECs resulted in a decrease in the size of lumen structures of lymphangioma, and this was accompanied with a modest increase in α-smooth muscle actin (αSMA)-expressing myofibroblasts in these benign lesions (S4C Fig). However, these myofibroblasts did not appear to play a role in IFA-induced lymphangioma because the formation of these lesions was not impaired in αSMA-tk transgenic mice (depletion of proliferating myofibroblasts that exhibit αSMA expression [35], S4D  Fig). These results underscore the specificity of PDPN-tk mice and support that LEC proliferation, but not myofibroblast proliferation, is essential for the formation of IFA-induced lymphangioma.

Lymphatic suppression in PDPN-tk mice with GCV treatment
Matrigel plug assay was conducted to determine the functional role of PDPN + LECs in lymphangiogenesis. Growth-factor-reduced matrigel supplemented with VEGF-C induced robust lymphatic vessel formation as well as blood vessel formation after subcutaneous implantation (400 μL matrigel per plug; one plug per mouse). In contrast with WT + GCV mice, PDPN-tk + GCV mice exhibited reduced lymphangiogenesis (Fig 2A) and LEC proliferation (S5A Fig) in the matrigel plugs, while the angiogenesis response, measured by cluster of differentiation (CD) 31 immunolabeling, was unaffected (Fig 2A). Decreased lymphatic vessel density in matrigel plugs of PDPN-tk mice was also confirmed by immunohistochemical assessment of lymphatic markers, PDPN (Fig 2B) and LYVE1 (S5B Fig). The PDPN-expressing cells within the matrigel plugs were predominantly co-immunolabeled with the LEC marker LYVE1 but did not express the cancer-associated fibroblast marker αSMA (S5C Fig). These results support that the cell population targeted by the PDPN-tk transgene in the aforementioned matrigel plug assays comprises of LECs and not fibroblasts.
Additionally, we examined the LN, intestine, and kidney of WT + GCV mice and PDPN-tk + GCV mice bearing the VEGF-C enriched matrigel plug. Although specific depletion of PDPN-expressing LECs was observed in the plug with active lymphangiogenesis (Fig 2A), no changes were noted for PDPN immunolabeling in these normal, unaffected tissues (S6A Fig), supporting that our genetic strategy only targets proliferating PDPN-expressing cells.

Depletion of LECs inhibits lung metastasis but not primary mammary tumor growth
Tumor lymphangiogenesis was examined in orthotopic 4T1 mammary tumors established in either PDPN-tk or WT female mice (all treated with GCV). Tumor tissues were scanned for lymphatic vessels and blood vessels. GCV-treated PDPN-tk mice revealed significantly suppressed lymphangiogenesis in both tumor center and tumor margin/periphery, defined as 100 μm from the tumor edge [36,37], while angiogenesis was not significantly altered (Fig 2C,  S6B Fig, and S7A Fig). Despite a significant suppression of lymphangiogenesis, the growth of orthotopic 4T1 mammary tumors in GCV-treated PDPN-tk mice was unchanged when compared with GCV-treated WT control mice (Fig 3A). Interestingly, despite the unchanged primary tumor growth, PDPN-tk-GCV mice with tumors exhibited significantly fewer surface metastatic lung nodules (Fig 3B) and histologically identified lung metastases (Fig 3C) when compared to WT mice. No tumor-infiltrated axillary or inguinal LN was observed in PDPNtk-GCV mice (0 out of 9 mice), whereas WT mice in the control group occasionally presented with axillary and/or inguinal LN metastases (2 out of 8 mice) (Fig 3A).

Fig 2. Depletion of proliferating LECs inhibits lymphangiogenesis. (A and B)
Lymphangiogenesis and/or angiogenesis in subcutaneously implanted matrigel plugs (growth-factor reduced, supplied with VEGF-C or PBS) from PDPN-tk or WT mice (n = 5 mice per group). Lymphatic vessel density and blood vessel density was examined by PDPN and CD31 immunofluorescence staining (A). Overall lymphatic vessels density was also evaluated by IHC staining for PDPN (B). Scale bars (A-B), 100 μm. (C) Orthotopic 4T1 mammary tumors in PDPN-tk or WT mice (n = 5 female mice per group) examined for lymphatic vessel density (PDPN staining) and blood vessel density (CD31 staining) in both the margin and intratumoral regions of tumors. Scale bar, 200 μm. GCV treatment was conducted as daily intraperitoneal injections of 50 mg/kg body weight of GCV. The white arrows point to the PDPN-expressing lymphatic vessels. These data are also depicted in S6B Fig. Data are represented as mean ± SEM. Significance is determined using an unpaired two-tailed Student t test ( � p < 0.05, �� p < 0.01, ��� p < 0.001). The underlying data can be found in S1 Data. CD, cluster of differentiation; GCV, ganciclovir; IHC, immunohistochemistry; LEC, lymphatic endothelial cell; NS, not significant; PDPN, podoplanin; tk, thymidine kinase; VEGF-C, vascular endothelial growth factor C; WT, wild type We also employed the mouse mammary tumor virus-polyoma middle tumor antigen (MMTV-PyMT) model, in which spontaneous mammary carcinomas and lung metastasis develop, to examine the impact of lymphatic/LEC depletion on cancer progression. MMTV-PyMT mice were bred with PDPN-tk mice to generate the MMTV-PyMT; PDPN-tk mice, as well as the MMTV-PyMT; WT littermate control mice. Female mice were monitored for tumor growth. The growth of MMTV-PyMT tumors was not significantly altered in GCVtreated MMTV-PyMT; PDPN-tk mice when compared to MMTV-PyMT; WT mice ( Fig 4A). Decreased lymphatic vessel density in MMTV-PyMT; PDPN-tk tumors was confirmed by immunohistochemical staining for LYVE1 (S7B Fig). The MMTV-PyMT; PDPN-tk mice exhibited increased incidence of cystic tumors, possibly resulting from enhanced lymphedema, when compared to control mice (Fig 4A and 4B). Previous studies have established that impaired lymphatic function can result in the accumulation of macromolecular proteins (such as albumin) because of compromised lymphatic drainage [38][39][40]. The increased level of lymphedema in primary tumor tissues of MMTV-PyMT; PDPN-tk mice was confirmed by albumin immunohistochemistry (Fig 4B). The total number of surface metastatic lung nodules, histologically identified lung metastatic lesions, and axillary LN metastasis was significantly reduced in MMTV-PyMT; PDPN-tk mice when compared to MMTV-PyMT; WT mice (Fig 4C and 4D). Taken together, these results support that suppression of lymphangiogenesis in primary mammary tumors did not impact their growth but limited their metastatic dissemination. Previous studies indicated that PDPN may also be expressed by cancer-associated fibroblasts [41][42][43] or macrophages [44]. Our analyses revealed that PDPN-expressing cells did not coexpress the breast-tumor-associated fibroblast marker αSMA but predominantly coexpressed the LEC-associated marker VEGFR3 (S7C Fig) and weakly coexpressed or failed to coexpress the vascular marker CD31 (S7D Fig), consistent with previous observations [19]. We also noted that PDPN + cells did not show colocalization with the macrophage marker CD68 in 4T1 tumors (WT mice), although close contact between CD68 + macrophages and PDPN + /LYVE1 + lymphatic vessels could be occasionally observed (S7E Fig).
Further, The Cancer Genome Atlas (TCGA) data set of 844 patients with invasive breast carcinoma (RNA sequencing version 2 analysis [RNA Seq V2] normalized gene expression with RSEM output [RSEM]) revealed a correlation between PDPN mRNA level and LN metastasis, showing higher levels of PDPN mRNA (PDPN mRNA expression normalized to Gapdh) associated with more LN metastasis (Fig 5A). These results were consistent with previous reports regarding the correlation between PDPN level (as examined by immunohistochemistry) and LN metastasis in breast cancer patients [45,46]. We also found marginally decreased occurrence of metastasis in distant organs (such as bone and lung) in PDPN-low patients compared to PDPN-high patients (S8A Fig). However, the number of cases with known distant organ metastasis was too low to offer conclusive evidence regarding the correlation between PDPN level and occurrence of distant metastases.

Discussion
Lymphangiogenesis, the formation of new lymphatic vessels, is associated with the progression of solid tumors [8][9][10][11][12]. It is known that lymphatics in the tumors are related to distant metastasis and contribute to immune surveillance and tissue fluid homeostasis. In this study, we performed experiments to determine the functional contribution of lymphatic vessels in lung metastasis associated with breast cancer. To achieve this goal, we generated two new transgenic mouse strains that allowed for the selective depletion of proliferating PDPN-positive LECs (PDPN-tk mice) and for the fate mapping/lineage tracing of PDPN-positive LECs (PDPN-Cre mice). The inhibition of lymphangiogenesis employing PDPN-tk mice supports that breastcancer-associated LN and lung metastasis is in part relying upon dissemination of cancer cells via lymphatic vessels. Interestingly, two recent studies highlighted that the dissemination route of cancer cells from LN to distant organs employs LN blood vessels in tumor-bearing mice [17,18].
In our studies, the vascular density in the mammary tumors was unchanged upon depletion of PDPN + cells. Lineage tracing experiments employing the PDPN-Cre mice showed that PDPN + cells are associated with lymphatic vessels but not the blood vessels. Interestingly, the growth of primary mammary tumors was not markedly altered when lymphangiogenesis was inhibited. These observations are also in alignment with a previous study showing that lymphangiogenesis induced by VEGF-C overexpression facilitates tumor metastasis without contributing to any growth advantage of primary tumor cells [19]. Various cancer types have distinct preferences in metastatic routes (such as a hematogenous route or a lymphatic route), yet the underlying mechanisms of such phenomena are still poorly understood. A recent study demonstrated the hematogenous route for ovarian cancer metastasis [50] in contrast to a The correlation between PDPN-relative mRNA level and LN metastasis in 844 breast cancer patients with available data for both RNA Seq V2 RSEM and examined LN metastasis count (identified by H&E) from TCGA breast invasive carcinoma data set. PDPN level is presented as relative mRNA expression normalized to the Gapdh housekeeping gene. A table presenting the number of LN metastasis positive cases out of total cases is also shown (χ 2 analysis). (B) Percentages of CD45 + , CD3 + , CD4 + , and CD8 + cells in 4T1 orthotopic mammary tumors of PDPN-tk or WT mice (n = 9 mice per group). (C) Percentages of CD11b + , CD11b + Gr1 − (Ly6C − Ly6G − ), and CD11b + Gr1 + cells in 4T1 orthotopic mammary tumors of PDPN-tk or WT mice (n = 9 mice per group). (D) Percentages of CD11b + Ly6G + , CD11b + Ly6C + , CD11c + cells, and ratio of CD4 + FoxP3 − Teffs to CD4 + FoxP3 + (Treg) in 4T1 orthotopic mammary tumors of PDPN-tk or WT mice (n = 9 mice per group). Data are represented as mean ± SEM. Significance is determined using an unpaired two-tailed Student t test ( � p < 0.05). The underlying data can be found in S1 Data. CD, cluster of differentiation; GCV, ganciclovir; H&E, hematoxylin and eosin; LN, lymph node; PDPN, podoplanin; RNA Seq V2, RNA Sequencing Version 2 analysis; RSEM, normalized gene expression with RSEM output; TCGA, The Cancer Genome atlas; Teff, effector T cell; tk, thymidine kinase; Treg, regulatory T cell; WT, wild type. peritoneal circulation-facilitated spread as previously proposed. Notably, depletion of lymphatic vessels did not alter the vascular density or lead to suppression of tumor growth but resulted in intratumor lymphedema due to potential imbalance in tissue fluid homeostasis. The new mouse models described herein may prove helpful for future studies related to breast-cancer-associated lymphedema, a substantial clinical problem observed in breast cancer patients.
Although our results support that the newly generated PDPN-tk transgenic mice enable the specific targeting of LECs in various models of lymphangiogenesis, including tumor lymphangiogenesis, it remains possible that immunolabeling for PDPN could be observed in other stromal cells in the tumor microenvironment, including cancer-associated fibroblasts, as noted in human breast cancer tissues [41][42][43]. The prognostic value of PDPN-expressing mesenchymal cells in the tumor microenvironment remains to be further studied.
The lymphatic system can regulate immune cell trafficking and tissue fluid homeostasis, yet our results indicated that suppression of tumor lymphangiogenesis did not significantly alter tumor immune infiltration. This may reflect a cancer-type-specific observation since it was reported in melanomas of mice lacking dermal lymphatic vessels that lymphatics were critical in establishing tumor-associated inflammation and immunity [24]. The percentage of intratumoral CD11b + Gr1 − macrophages, however, was significantly elevated with PDPN + LEC depletion. This may reflect a host response to compensate decreased lymphangiogenesis, in particular since macrophages play a role in regulating lymphangiogenesis and releasing lymphangiogenic factors [51][52][53][54]. Increased numbers of CD11b + Gr1 − macrophages in mammary tumors with PDPN + LEC depletion is also consistent with previous reports on increased macrophage infiltration as a hallmark of lymphedema [48,49] and could support a potential role of these cells in metastasis, albeit further study is still needed. Intriguingly, our results suggest that the decreased metastatic burden associated with suppressed lymphangiogenesis may be independent of a lymphocytic polarization in the primary tumor microenvironment.

Ethics statement
Mice were euthanized using CO 2 inhalation. All mice were maintained under standard housing conditions at the MD Anderson Cancer Center (MDACC) animal facility and the Beth Israel Deaconess Medical Center (BIDMC) animal facility, and all animal procedures were approved by the MDACC Institutional Animal Care and Use Committee and the BIDMC Institutional Animal Care and Use Committee (IACUC number: 1033).

Mice
The PDPN-tk mouse strain was generated by cloning and ligating the 4-kb PDPN promoter sequence to HSV viral tk sequence using the pCR2.1-TOPO vector (Invitrogen, Carlsbad, CA, USA). A similar approach was used to generate the PDPN-Cre mouse strain. Both transgenic mice were generated by the Transgenic Mouse Core Facility at Harvard Medical School. The mice were backcrossed (over 20 generations) and maintained on the BALB/c genetic background. Primers for PDPN-tk genotyping PCR are PDPN-forward 5 0 -ACCGGAGACATAA ATGCCGA-3 0 and TK-reverse 5 0 -AGCACCCGCCAGTAAGTC-3 0 . Primers for PDPN-Cre genotyping PCR are PDPN-forward 5 0 -ACCGGAGACATAAATGCCGA-3 0 and Cre-reverse 5 0 -CGCCGCATAACCAGTGAAAC-3 0 . αSMA-tk mice were generated and characterized in our previous study [55]. TGFBRII flox mice were kindly provided by H. Moses, Vanderbilt University [56]. β1 integrin flox mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Investigators were not blinded to group allocation but were blinded for the histological assessment of phenotypic outcome. No randomization method was used, and no animal was excluded from the analysis. The experimental endpoint is defined as when tumor burden reaches 1,500 mm 3 or 1.5 cm in diameter (whichever comes first). For the evaluation of surface lung nodules in mouse mammary tumor models, all surfaces of all of the lobes were ascertained for the presence of surface lung nodules. For the microscopic evaluation of lung metastases in mouse mammary tumor models, we counted the number of nodules observed on a single H&E-stained cross section of the lungs.

Matrigel plug assay
Matrigel plug assay was conducted to determine the functional role of PDPN + LECs in lymphangiogenesis. Growth-factor-reduced matrigel (Corning, Corning, NY, USA) supplemented with VEGF-C (1 μg /400 μL matrigel) was subcutaneously implanted in WT and PDPN-tk mice (400 μL matrigel per plug; one plug per mouse; n = 5 mice per group). . For all immunohistochemical stainings, sections were incubated with biotinylated secondary antibody and then streptavidin-HRP (Vector Labs, Burlingame, CA, USA). Counterstaining with hematoxylin was conducted, and DAB positivity was examined in randomly selected visual fields. For all immunofluorescence stainings, sections were incubated with fluorescent-labeled secondary antibodies according to the primary antibody usage. For the YFP staining of tissue samples from PDPN-Cre; LSL-YFP mice, optimized protocols for tissue collection and immunofluorescence staining were used in order to minimize the autofluorescence in the skin and intestine sections. These optimized protocols include: conducting PBS perfusion before collecting the organs from mice; using Sudan Black B (Sigma-Aldrich, St. Louis, MO, USA) incubation on the sections before the staining [57]; blocking with 4% cold water fish gel before primary antibody incubation; and decreased secondary antibody concentration. Staining for αSMA was performed with Mouse-on-Mouse (MOM) kit (Vector Laboratories) following the manufacturer's instructions. The images of at least 3 random visual fields for each sample section were quantified for positive area using NIH ImageJ analysis software (albumin, CD31, Ki67, LYVE1, or PDPN). Quantified values for multiple visual fields were averaged to produce a single value for each animal, which was then averaged again to represent the mean bar for the group in each graph.

Mouse lymphatic tissue induced by IFA
Either WT or PDPN-tk mice (3 months old, female) were intraperitoneally injected twice (day 1 and day 14) with IFA (200 μL, 1:1 mixed with PBS) to induce the formation of mouse hyperplastic lymphatic tissue (lymphangioma), as previously described [32]. The lymphangioma confluence was quantified as the percentage coverage by lymphangioma area among the total area of the diaphragm, as quantified by ImageJ software. The average diameter of lumen structures within lymphangioma tissues was calculated by measuring 10-20 randomly selected lumens within microscopic (40×) images of H&E-stained tissue slides using ImageJ software. Mouse lymphangioma tissue, formed by hyperplastic lymphatic vessels on the diaphragm and liver of mice in response to IFA treatment, was collected on day 21 and digested with 1 mg/mL collagenase solution (collagenase I:collagenase II = 1:1) at 37˚C for 30 min. Cell suspension was filtered and purified for LECs using anti-PDPN antibody (Abcam, ab11936) and Magnetic Dynabeads (Thermo Fisher Scientific, Waltham, MA, USA). Isolated primary mouse LECs were cultured in endothelial cell growth medium (Lonza, Basel, Switzerland). For the detection of YFP-positive LECs in IFA-induced lymphangioma by flow cytometry, mouse lymphangioma tissue was collected, prepared as a single-cell suspension according to the same protocol above, and examined for YFP fluorescence signal by flow cytometry.

Cell viability assay
For in vitro treatment of GCV, LECs from WT or PDPN-tk mice were isolated using the same method listed above, cultured, treated with 0, 5, or 50 μM GCV for 48 h, and then examined for cell viability (measured as the absorbance at 450 nm by a microplate reader) using the Cell Counting Kit-8 (Dojindo Molecular Technologies, Kumamoto, Japan). Results of cell viability were expressed as percentage of viable cell counts using the control vehicle-treated group as the reference.

4T1 orthotopic mammary tumor model
Either WT or PDPN-tk female mice, around 3 months old, were used for orthotopic implantation of 4T1 mammary epithelial cancer cells. 4T1 Cells were from American Type Culture Collection (ATCC) and cultured in DMEM with 10% FBS and 100 U/mL penicillin-streptomycin. Cells were examined monthly to ensure a negative result for mycoplasma test. Mice were anesthetized with ketamine/xylazine, the skin near the mammary gland was incised, and 4T1 cancer cells were injected into the mammary glands (in total, 1 × 10 6 cells per mouse; 5 × 10 5 cells at each side, for both left and right sides), as previously described [4].  3 . Tumor volumes were measured twice per week using digital calipers and calculated using the equation length × width 2 × 0.52. Mice were sacrificed when tumor volume reached approximately 1,500 mm 3 or 1.5 cm in diameter (whichever came first). A tumor was counted as a cystic tumor when it formed prominent fluid-filled cyst with a volume greater than 100 mm 3 . Tumors and other organs, including the lungs, were collected as previously described [58].

TCGA data set analysis
The mRNA data (RNA Seq V2 RSEM) and clinical data of 844 patients with invasive breast carcinoma from the TCGA data set were obtained using the cBioPortal for Cancer Genomics (http://www.cbioportal.org/) [59]. χ 2 analyses, using SPSS statistical software, were performed comparing LN metastatic frequency between PDPN-High and PDPN-Low groups of patients. The metastasis occurrence at distant organs of these patients was also analyzed based on the detailed clinical information from TCGA data set.

Statistics
Statistical analyses of flow cytometry and immunostaining quantifications were performed with unpaired, two-tailed t test, one-way ANOVA with Tukey's multiple comparison test, or Fisher's exact test with GraphPad Prism (GraphPad Software, San Diego, CA, USA). A p value < 0.05 was considered statistically significant. Error bars represent SEM when multiple visual fields were averaged to produce a single value for each animal, which was then averaged again to represent the mean bar for the group in each graph. Lymphangiogenesis and/or angiogenesis in subcutaneously implanted matrigel plugs (growthfactor reduced, supplied with VEGF-C or PBS) from PDPN-tk or WT mice (n = 5 mice per group). The proliferation status of LECs was examined by Ki67 and PDPN immunofluorescence staining (A). Overall lymphatic vessel density was also evaluated by IHC staining for LYVE1 (B). Scale bars (A-B), 100 μm. The cell populations within the matrigel plugs were further examined by PDPN, VEGFR3, and αSMA immunofluorescence staining (C). Scale bars (C), 25 μm. Data are represented as mean ± SEM. Significance is determined using an unpaired two-tailed Student t test ( � p < 0.05, ��� p < 0.001). The underlying data can be found in S1 Data. αSMA, α-smooth muscle actin; IHC, immunohistochemistry; Ki67, cell proliferation antigen Ki-67; LEC, lymphatic endothelial cell; LYVE1, lymphatic vessel endothelial hyaluronan receptor-1; PDPN, podoplanin; tk, thymidine kinase; VEGF-C, vascular endothelial growth factor C; VEGFR3, vascular endothelial growth factor receptor 3; WT, wild type.