A Role for Bone Morphogenetic Protein-4 in Lymph Node Vascular Remodeling and Primary Tumor Growth

Running title: BMP-4 in vascular remodeling and tumor growth Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Abstract Lymph node metastasis, an early and prognostically important event in the progression of many human cancers, is associated with expression of vascular endothelial growth factor-D (VEGF-D). Changes to lymph node vasculature that occur during malignant progression may create a metastatic niche capable of attracting and supporting tumor cells. In this study, we sought to characterize molecules expressed in lymph node endothelium that could represent therapeutic or prognostic targets. Differential mRNA expression profiling of endothelial cells from lymph nodes that drained metastatic or non-metastatic primary tumors revealed genes associated with tumor progression, in particular bone morphogenetic protein-4 (BMP-4). Metastasis driven by VEGF-D was associated with reduced BMP-4 expression in high endothelial venules, where BMP-4 loss could remodel the typical high-walled phenotype to thin-walled vessels. VEGF-D expression was sufficient to suppress proliferation of the more typical BMP-4-expressing high endothelial venules in favor of remodeled vessels, and mechanistic studies indicated that VEGFR-2 contributed to high endothelial venule proliferation and remodeling. BMP-4 could regulate high endothelial venule phenotype and cellular function, thereby determining morphology and proliferation responses. Notably, therapeutic administration of BMP-4 suppressed primary tumor growth, acting both at the level of tumor cells and tumor stromal cells. Together, our results show that VEGF-D-driven metastasis induces vascular remodeling in lymph nodes. Further, they implicate BMP-4 as a negative regulator of this process, suggesting its potential utility as a prognostic marker or anti-tumor agent. Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.


Introduction
Lymphatic dissemination is considered to be an early and crucial route of metastasis for many cancers (1,2).Blindending lymphatic capillaries drain fluid, cells, and macromolecules from tissue interstitium into a hierarchy of vessels punctuated by lymph nodes (LN), which provide immunologic surveillance for a particular lymphatic drainage basin (3).The presence of metastatic tumor cells in the "sentinel" LN draining a tumor site is a key factor in disease management: substantial clinical data indicates adverse prognostic significance of tumor-positive LNs for many tumor types (4,5).However, a clear understanding of the mechanistic role of LNs in tumor progression is still lacking.
VEGF-D and VEGF-C are important inducers of the growth and differentiation of blood vessels and lymphatics.When overexpressed in experimental tumors these growth factors elicit angiogenesis and lymphangiogenesis, and are furthermore associated with increased metastasis to LNs and distant organs (1).VEGF-D and VEGF-C expression is also associated with metastasis to LNs in many human cancers, and is independently associated with poor prognosis (6).Recently, it has emerged that modulation of lymphatics and blood vesselsincluding high endothelial venules (HEV), vessels specialized for leukocyte trafficking (7,8)-also occurs in draining LNs of some tumors (9,10).Such alterations can precede the arrival of metastatic cells (7,(11)(12)(13), and members of the VEGF family have been implicated in these changes (12)(13)(14)(15).The importance of alterations to LN endothelium is highlighted by studies of human breast cancer: lymphangiogenesis or angiogenesis within metastatic tumor deposits in sentinel LNs was found to be associated with, and sometimes independently predictive of, distant metastasis or survival (9,16,17).
Here, we sought to characterize changes to the vasculature within tumor-draining LNs, to identify molecules with prognostic or therapeutic potential.We compared the molecular profiles of enriched endothelial cell (EC) populations from LNs draining nonmetastatic tumors with those from LNs draining metastatic (VEGF-D-overexpressing) tumors.BMP-4 was downregulated in the HEVs of LNs draining metastatic tumors.This observation was linked with the remodeling of HEVs induced by VEGF-D-driven metastasis, thus implicating BMP-4 as a regulator of HEV morphology and cell function.Furthermore, therapeutically applied BMP-4 protein inhibited primary tumor growth.This study indicates that VEGF-D's prometastatic activity includes remodeling of specialized LN endothelium, and identifies new roles for BMP-4 in cancer and vascular biology.

Materials and Methods
Lists of antibodies, primers and detailed protocols are contained in the Supplementary Methods section.

Metastatic and nonmetastatic xenograft tumor models
293 EBNA-1 tumor cell lines stably expressing full-length human VEGF-D (293-VEGF-D), human VEGF-C (293-VEGF-C), or vector alone (293-Apex) were established in SCID/NOD mice as described (18).293 EBNA-1 cells were a gift from Kari Alitalo, University of Helsinki, Finland.Regular growth and morphology of transfected cell lines was monitored routinely and growth factor expression verified by Western blot prior to each experiment.LNs were analyzed within the timeframe that metastasis typically occurs in this model; that is, 2 to 4 weeks postimplantation.All animal experiments were carried out with the approval of the Institutional Animal Ethics Committee.

Enrichment of LN EC populations
Draining LNs of metastatic or nonmetastatic tumors pooled from 1 to 5 mice were enzymatically digested, then tumor cells and leukocytes were depleted using immunomagnetic selection (Miltenyi Biotec) for class I HLA and CD16/CD32.The remaining cells were cultured in EGM-2 MV media (Lonza) before enrichment for ECs by selection for podoplanin (19).See Supplementary Fig. S1 for detailed procedure.

Microarray analysis
Duplicate samples of LN EC total RNA (RNeasy Plus kit, Qiagen) were applied to Affymetrix expression arrays (430 2.0; Australian Genome Research Facility).Raw intensity data were analyzed using GeneChip Operating Software (Affymetrix), and profiles compared via Robust Multiarray Analysis and linear modeling using AffylmGUI software (20).Microarray data are deposited in NCBI's Gene Expression Omnibus; series accession number GSE31123 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc¼GSE31123).

Human LNs
Breast cancer-associated LNs with or without histologically identifiable metastases (n ¼ 7 patients, 22 LNs), or control nontumor-associated LNs collected during cardiac surgery (n ¼ 3 patients), were obtained as a pilot study.Access to deidentified tissue (formalin fixed, paraffin embedded) was provided by the Pathology Department, Royal Melbourne Hospital, with permission from the Melbourne Health Human Research Ethics Committee.

Immunostaining and image quantitation
For BMP-4/MECA-79 quantitation, 2 to 3 sections of each tumor-draining LN ($6 per group) were immunostained (18).For HEV morphometry, the luminal and basal edges of HEVs were traced using Metamorph Premier (Molecular Devices), to determine lumen area, average vessel wall width and endothelial area using Integrated Morphometry Analysis parameters (journal available on request).HEVs with 50% or more of their circumference staining for BMP-4 were designated BMP-4 high ; or otherwise BMP-4 low .Data were analyzed according to a linear mixed model (Supplementary Methods).

Treatment of ear-draining LNs with recombinant VEGF-D
One microgram of purified VEGF-D dimers (0.05 mg/mL; Vegenics Ltd.) in PBS, or PBS alone as control, was injected intradermally into the ears of SCID/NOD mice for 3 consecutive days.On day 4, BrdU (Invitrogen) was injected intraperitoneally, and ear-draining (superficial parotid) LNs were harvested 2 days later.

Treatment of tumors with neutralizing antibodies
Mice bearing metastatic (VEGF-D overexpressing) tumors received 3 times weekly intraperitoneal injections of neutralizing antibodies (800 mg) to VEGF receptor-2 (VEGFR-2; DC101; Imclone) or VEGF-D (VD1; ref. 21), or PBS.For analysis of HEVs, sections of LNs draining nonmetastatic and antibodytreated metastatic tumors were used from one experiment.LNs of PBS-treated metastatic tumors where HEVs were not obscured by tumor infiltration were included from an identically conducted experiment as a control.

BMP-4 therapeutic model
Tumor-bearing mice were injected intraperitoneally from day 1, 3 times weekly, with 1.4 mg of human BMP-4 (R&D Systems) in 200 mL PBS with 0.652 mg/mL BSA, or a vehicle control of PBS with 0.32 mmol/L HCl and 1 mg/mL BSA, until day 12 or experiment termination.Serum was sampled 60 minutes posttreatment and BMP-4 quantified by ELISA (R&D).

Statistical analysis
Data were compared using a 2-tailed Student t test, or Fisher's exact test for comparison of proportions.Graphed data represent mean AE SE unless specified otherwise.

Enrichment of endothelial cells from tumor-draining LNs
A model of VEGF-D-driven tumor metastasis to regional LNs was used to examine molecular changes in LN endothelium during metastasis (Fig. 1A).Overexpression of VEGF-D in 293-EBNA-1 tumor cells drives metastasis to local LNs within 2 to 4 weeks of implantation in approximately 80% of cases.Vector-transfected tumor cells (no VEGF-D) served as a nonmetastatic control (18).Podoplanin (19) was used as a highly expressed, protease-resistant selection marker to derive cell populations enriched for lymphatic ECs and related EC types, which may respond to VEGF-D (Fig. 1B).Microarray analysis revealed expression of EC-characteristic genes, including VEGFR-2, neuropilin-1 and neuropilin-2, endothelial nitric oxide synthase, CD34 and TIE-2; while desmin and calponin-1, found in fibroblastic lineages, and chondroitin sulfate proteoglycan 4 (NG-2 antigen), characteristic of pericytes, were absent.These findings confirmed that the podoplanin þve cells were enriched for ECs.The LN ECs heterogeneously expressed ICAM-1 and endoglin, markers of endothelial activation in inflammation and angiogenesis, respectively (Fig. 1C; Supplementary Methods).

Identification of endothelial-expressed genes modulated during metastasis to LNs
LN ECs from metastatic and nonmetastatic tumor models were compared by microarray (Fig. 2A).Of the top 10 differentially expressed genes (ranked by adjusted P value), 9 were downregulated in LNs draining metastatic tumors compared with their nonmetastatic counterparts, and all 10 showed more than 2-fold difference in expression (Table 1; Fig. 2B).Candidates were subsequently selected for further analysis based on relevance to endothelial and cancer biology.qRT-PCR validated the downregulation of Bmp4, Unc5c, Cfh, Emcn, and Gpr39 in ECs from LNs draining metastatic tumors, and the upregulation of Nova1 (Fig. 2C).Bmp4 showed the greatest abundance and more than 5-fold difference in expression, and was thus selected for further investigation.

Localization of BMP-4 protein in HEVs and differential expression in metastasis
Immunohistochemistry showed that BMP-4 protein was localized to HEVs (Fig. 3A), confirmed by costaining for the specific MECA-79 epitope (22).BMP-4 protein was present in a subset of HEVs in LNs draining both nonmetastatic and metastatic tumors (Fig. 3A), and in LNs from nontumorbearing SCID/NOD and immunocompetent mice (Fig. 3A, data not shown).HEVs did not endogenously express podoplanin (Supplementary Fig. S2), suggesting podoplanin probably became upregulated in HEV ECs during the brief culturing between extraction from LN and purification for microarray analysis (23,24).Although MECA-79 stained the surface of HEV ECs, BMP-4 seemed primarily in the cytoplasm (Fig. 3A inset), implying that HEV ECs express BMP-4 protein.No other sites of BMP-4 localization were observed in the LN or primary tumor.This supported the conclusion that HEV ECs are the main source of BMP-4 mRNA and protein in LNs.
Quantitation of staining revealed that HEV-expressed BMP-4 was significantly reduced (by $50%), in LNs draining metastatic versus nonmetastatic tumors (P < 0.001; Fig. 3B and C).This illustrated a shift from predominately BMP-4 high to predominately BMP-4 low HEVs in LNs draining nonmetastatic versus metastatic tumors respectively; however, both LN types contained some BMP-4 high and some BMP-4 low HEVs (Fig. 3B  and C).Therefore, the downregulation of BMP-4 mRNA was reflected at the protein level in vivo.

BMP-4 loss marks HEV remodeling in cancer
We examined LNs for evidence of tumor-induced HEV remodeling (7), and explored whether VEGF-D or BMP-4 was associated with this process (Fig. 4A).In LNs draining nonmetastatic tumors, BMP-4 high HEVs had significantly smaller lumen areas than BMP-4 low HEVs (P ¼ 0.0017; Fig. 4B).In LNs draining metastatic tumors, however, the BMP-4 high HEVs were more dilated than in the nonmetastatic context (P ¼ 0.028; Fig. 4B).Significantly, BMP-4 high HEVs had thicker vessel walls than BMP-4 low HEVs in all LNs (P < 0.001; Fig. 4B), suggesting that BMP-4 expression was closely linked with HEV morphology.Although the remaining BMP-4 high HEVs in LNs draining metastatic tumors largely retained their greater vessel wall width, there was a strong trend suggesting reduced width compared with those in LNs draining nonmetastatic tumors, indicating that VEGF-D-driven metastasis could affect the endothelial width of BMP-4 high HEVs (P ¼ 0.064; Fig. 4B).We also observed remodeled HEVs in a pilot study of human breast cancer-associated LNs with or without histologically identifiable metastasis (Fig. 4E), confirming its occurrence in human disease (7).
We next investigated whether HEV remodeling involved EC proliferation.Interestingly, BMP-4 high HEV ECs in LNs draining metastatic tumors had a significantly lower proliferation rate than those from the nonmetastatic model (P ¼ 0.026; Fig. 4C).Furthermore, BMP-4 low HEV ECs in LNs draining metastatic tumors had a significantly higher proliferation rate than the BMP-4 high HEV ECs (P ¼ 0.015; Fig. 4C).These results indicated that the proliferation response of HEV ECs to tumor-secreted VEGF-D may be modulated by BMP-4; another way in which VEGF-D-driven metastasis may induce remodeling of HEV characteristics via reduction of BMP-4 expression.

The role of VEGF-D and VEGFR-2 in HEV remodeling
To determine whether HEVs could respond directly to tumor-secreted human VEGF-D, we examined VEGFR-2 and VEGFR-3 expression in LNs.VEGFR-2 was expressed on most HEVs, blood vessel capillaries and lymphatics (Fig. 4D).VEGFR-3 was strongly expressed on lymphatics, but was essentially absent from HEVs.Thus HEVs are capable of responding to VEGFR-2 ligands.
In vivo approaches were utilized to investigate the specific pathways controlling HEV remodeling.Injection of VEGF-D into the mouse ear mimics tumor-secreted growth factor draining to regional LNs.After 3 days of VEGF-D treatment, proliferation of BMP-4 high HEV ECs was decreased (P ¼ 0.034; Fig. 5A), suggesting VEGF-D was responsible for the effect observed in tumor-draining LNs (Fig. 4C), and that suppression of proliferation in BMP-4 high HEVs by VEGF-D may occur early in the metastatic process.Alteration of lumen area, vessel wall width and BMP-4 expression may require a longer stimulation period as neither was affected in this experiment (Fig. 5A and B); however, BMP-4 high HEVs again exhibited significantly thicker vessel walls (Fig. 5A).

Effects of exogenous BMP-4 on tumor progression
As this study was designed to identify and analyze molecular targets with prognostic and/or therapeutic potential, we established a therapeutic model to determine the effects of exogenously-administered BMP-4.Activity and stability of recombinant human BMP-4 were verified by bioassay (Supplementary Fig. S3A; Supplementary Methods).As shown in Fig. 6A, BMP-4 inhibited the exponential growth of VEGF-D-overexpressing primary tumors by approximately 50% (day 20, P ¼ 0.056; day 22, P ¼ 0.036; day 24, P ¼ 0.080).In addition, similar tumors overexpressing VEGF-C were reduced in size by approximately 56% by BMP-4 treatment (day 15, P ¼ 0.067; day 18, P ¼ 0.021; day 23, P ¼ 0.026).BMP-4 could thus inhibit tumor growth driven by 2 different lymphangiogenic/angiogenic growth factors.ELISA results confirmed that injected BMP-4 reached the systemic circulation at approximately 1,200 pg/mL in serum after 60 minutes (Fig. 6B).Interestingly, under the conditions and timecourse of these experiments the BMP-4 treatment did not seem to affect metastasis to LNs or HEV morphology (Fig. 6C and data not shown).Analysis of HEVs did reveal a trend suggesting that in metastasis-positive LNs draining VEGF-D-overexpressing tumors, more BMP-4 high HEVs were observed under BMP-4 treatment than for the control (mean AE SE: BMP-4, 40.9 AE 10.1; vehicle, 29.5 AE 10.0; n ¼ 5, P ¼ 0.16).Furthermore, BMP-4 high HEVs again exhibited thicker vessel walls than BMP-4 low HEVs in both treatment conditions (P < 0.001; Supplementary Fig. S3B), confirming the importance of endogenous BMP-4 expression.

Mechanisms of BMP-4-induced tumor growth suppression
To clarify the mechanism by which BMP-4 suppressed primary tumor growth, we first examined the distribution of its receptors.BMPs bind a heterodimeric complex of type I and type II receptors (25).Immunohistochemistry for BMPR-II revealed broad expression on multiple cell types including tumor cells, stroma, and endothelium of large blood vessels (Fig. 6D).Microarray analysis indicated that the VEGF-Doverexpressing tumor cells expressed BMPR2, as well as BMPRIA and ACTR1A, but not BMPR1B (Supplementary Table S2), whereas immunocytochemistry confirmed expression of BMPR-IA and BMPR-II protein on tumor cells and tumorderived fibroblasts (Supplementary Fig. S4A; Supplementary Methods).Interestingly, Western blotting revealed that BMPR-II protein was more abundant in BMP-4-treated than controltreated VEGF-D-overexpressing metastatic tumors (P ¼ 0.048; Fig. 6E and Supplementary Methods), potentially representing a feedback loop that could contribute to tumor suppression.

Discussion
Changes to the blood or lymphatic vasculature in tumordraining LNs have prognostic significance in cancer (9,16,17,26), and may facilitate metastasis (11)(12)(13).Understanding the mechanisms and functional consequences of these alterations will be critical in determining the overall role of LN metastasis in tumor progression, and could advance prognostication and treatment for cancer patients.Here, we have identified molecules involved in the remodeling of HEVs in tumor-draining LNs, and an additional role for BMP-4 in suppressing primary tumor growth.
Microarray analysis of enriched LN EC populations revealed differential expression of several genes with significance to endothelial and tumor biology.Analysis of isolated EC subtypes has enabled identification of important functional molecules (ref.27 and manuscript submitted).Although our isolation strategy utilized podoplanin, commonly used to distinguish lymphatic endothelium, immunohistochemical validation revealed BMP-4 to be differentially expressed in HEVs, a specialized venous endothelial type that did not express podoplanin in vivo.Subsequent to observations that blood vascular ECs cocultured with lymphatic ECs could spontaneously acquire expression of lymphatic-characteristic molecules including podoplanin (23), it has been shown that substantial plasticity exists between arterial, venous, and lymphatic EC lineages, controlled by specific transcription factors and reflecting their common embryonic origin (24).Our observations provide further confirmation of this plasticity and relatedness.Another similar study used microarray analysis of isolated lymphatic ECs from primary tumors, which were briefly cultured, to identify novel markers with prognostic significance (28).Our study advances upon this by examining the endothelium of tumor-draining LNs.
The morphologic changes we observed to be associated with VEGF-D-driven metastasis and BMP-4 reduction-that is,   remodeling of the normally "high"-walled HEVs into flat walled, more dilated vessels with altered proliferation responseswere consistent with those observed in mouse models and human breast cancer (7).Other investigators observed suppression of the HEV-expressed lymphotactic chemokine CCL21 and reduced lymphocyte recruitment in tumor-draining LNs (29).Such physical and molecular features of HEVs endothelium are integral to their role in trafficking leukocytes into the LN to facilitate immune responses (8).Although these investigators analyzed total HEVs, we identified HEV subtypes (BMP-4 high and BMP low ) which can respond differentially to tumor-associated stimuli.Although the functional significance of HEV height is poorly understood, flattening of HEV ECs seems to reduce leukocyte transmigration rates (30).Lower branching-order HEVs were observed to support lower rates of lymphocyte adhesion than higher-order HEVs (29); interestingly, in our studies lower-order HEVs tended to have flatter endothelium and lower BMP-4 expression than higher-order HEVs.It is possible that HEV remodeling may echo homeostatic differences in the morphologic, molecular, and functional characteristics of different branching-order HEVs.Ultimately, tumor-induced HEV remodeling could assist in generating a metastatic niche (31): proliferating, dilated blood vessels derived from remodeled HEVs could enrich the nutrient and oxygen supply to a LN, whereas impaired immune function would promote tumor cell survival.The proximity of remodeled HEVs and lymphatic vessels could provide a shortcut for metastatic cells into the blood vasculature and thus systemic circulation (31,32).
Our study provides an important contribution to understanding the molecular mechanisms driving tumor-induced HEV remodeling (Fig. 5E).The effects of BMP-4 and VEGF-Ddriven metastasis on HEV vessel wall width were strongly evident, whereas differences in lumen area and proliferation were more dynamic and may be sensitive to other factors.The differing impacts of VEGFR-2 and VEGF-D blockade suggest involvement of another VEGFR-2 ligand.Several studies have implicated VEGF-A in stimulating HEV growth and remodeling in immune responses (33,34); thus endogenous VEGF-A could contribute to VEGFR-2-mediated HEV dilation in tumor-draining LNs.In addition, VEGF-A might be involved in the differential proliferative response of BMP-4 high and BMP-4 low HEVs to VEGF-D.BMP-4 can increase expression and phosphorylation of VEGFR-2 in ECs, thus enhancing responsiveness to autocrine or paracrine VEGF-A (35).BMP-4 itself could signal to ECs in an autocrine manner (36), and might upregulate VEGF-A expression by LN stromal cells (34,37), thus potentiating a VEGF-A/VEGFR-2 signaling loop.VEGF-D may then inhibit proliferation of BMP-4 high HEV ECs in the tumor context by competing with VEGF-A for binding to  As a member of the TGF-b superfamily of multipotent cytokines, the role of BMP-4 in tumor progression can be complex and highly context specific (25,39).We showed that while endogenously expressed BMP-4 regulates HEVs, exogenous BMP-4 can restrict primary tumor growth.BMP-4 is also known to induce apoptosis of other tumor cell lines (40,41) and microvascular ECs (42), although in other studies proangiogenic responses were observed, possibly due to potentiation of VEGF-A/VEGFR-2 signaling (35).Our data suggest that lymphatic ECs may respond to BMP-4 in a similar way.An increase in proliferation of tumor-derived fibroblasts stimulated with BMP-4 in vitro is intriguing considering that cancer-associated fibroblasts are commonly implicated in promoting tumorigenesis (43).The upregulation of BMPR-II expression in BMP-4treated tumors recapitulates a similar observation in Xenopus embryos indicating that Bmpr2 is a target gene of BMP-4 signaling (44).Expression of several other regulators of BMP-4 signaling is also induced by BMP-4, raising the possibility that blockade of relevant signaling inhibitors might enhance the efficacy of BMP-4 treatment.Previous in vivo studies have described antitumorigenic effects of BMP-4 for  several tumor types (40,(45)(46)(47)-as well as protumorigenic effects for some-but thus far only one other study, using a model of glioblastoma multiforme, has demonstrated an antitumor effect of therapeutically administered recombinant BMP-4 (48).Although the authors identified a prodifferentiation effect on tumor stem cells, we noted that VEGF-D is highly expressed in glioblastoma multiforme (49).Our study adds weight to the potential of BMP-4 as an antitumor agent by showing that it can inhibit tumor growth driven by 2 different lymphangiogenic/angiogenic factors through action on both tumor cells and stroma.
The context-specific nature of BMP-4 signaling does compel careful tuning of BMP-4 targeting and dosage to ensure a robust antitumor effect.A more constant dosage of BMP-4, or a delivery system more targeted to the LN, may help clarify whether therapeutically administered BMP-4 can reverse HEV remodeling or inhibit metastasis.Nevertheless, reduction of BMP-4 expression in HEVs is an important early molecular indicator of remodeling, as it precedes loss of MECA-79 upon incorporation into the vasculature of the tumor deposit (7).Clinical studies will establish whether BMP-4 may represent a convenient surrogate marker of HEV remodeling in cancer.Furthermore, BMP-4 or HEV remodeling may serve as indicators of systemic or distant effects of prometastatic tumorderived factors such as VEGF-D, and provide prognostic information relevant to metastasis, treatment response, or patient outcome.Our data further highlight the need to better under-stand the functional and prognostic significance of the LN, and in particular its vasculature, to cancer metastasis, as well as the potential of BMP-4 as a multipotent antitumor agent.

Figure 1 .
Figure 1. Isolation of ECs from tumor-draining LNs.A, schematic of approach to investigate differentially expressed genes in enriched ECs from LNs draining metastatic or nonmetastatic tumors.B, immunomagnetic selection for podoplanin-enriched populations of LN ECs, as confirmed by flow cytometry.Gray line, isotype control; percentages represent proportions within podoplanin þve gate (isotype control proportion subtracted).C, enriched EC populations from LNs draining metastatic tumors were analyzed for ICAM-1 (green) and endoglin expression by immunofluorescence or flow cytometry.

Figure 2 .
Figure 2. Identification of differentially expressed genes in LN ECs.A, ECs from LNs draining metastatic or nonmetastatic tumors (labeled nonmetastatic or metastatic LN EC) were compared by microarray.B, a volcano plot of log odds of differential expression against fold change illustrates significantly differentially expressed genes.C, for selected genes, differential expression was validated by qRT-PCR.Shown are 2 representative examples (1 and 2) of pairwise comparisons.Data are mean AE SD of triplicate reactions.Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001.

Figure 6 .
Figure 6.Therapeutic administration of BMP-4.A, BMP-4 or vehicle control was administered to mice from day 1 until day 12 or experiment termination, and tumor volume measured (n ¼ 9-11).B, detection of BMP-4 in serum by ELISA (n ¼ 3).C, LNs were scored histologically positive or negative for metastatic cells.D, immunohistochemistry detecting BMPR-II expression on multiple tumor cell types including blood vessels, inset.E, Western blot detecting BMPR-II in cultured tumor cells and metastatic (VEGF-D) tumor lysates, and densitometric quantitation of expression (n ¼ 3; full-length blot, Supplementary Fig. S5).
M.G.Achenand S.A. Stacker: commercial research grant, Imclone; ownership interest, Circadian Technologies; consultant/advisory board, Vegenics.R. Shayan: ownership interest, Circadian Technologies.The other authors disclosed no potential conflicts of interest.