Bone Morphogenetic Protein 2 Signaling Negatively Modulates Lymphatic Development in Vertebrate Embryos

Abstract

Rationale

The emergence of lymphatic endothelial cells (LECs) seems to be highly regulated during development. Although several factors that promote the differentiation of LECs in embryonic development have been identified, those that negatively regulate this process are largely unknown.

Objective

Our aim was to delineate the role of bone morphogenetic protein (BMP) 2 signaling in lymphatic development.

Methods and Results

BMP2 signaling negatively regulates the formation of LECs. Developing LECs lack any detectable BMP signaling activity in both zebrafish and mouse embryos, and excess BMP2 signaling in zebrafish embryos and mouse embryonic stem cell–derived embryoid bodies substantially decrease the emergence of LECs. Mechanistically, BMP2 signaling induces expression of miR-31 and miR-181a in a SMAD-dependent mechanism, which in turn results in attenuated expression of prospero homeobox protein 1 during development.

Conclusions

Our data identify BMP2 as a key negative regulator for the emergence of the lymphatic lineage during vertebrate development.

The lymphatic system is essential for maintenance of interstitial fluid balance and uptake of lipids, and it provides the major conduit for immune cells.^1,2 The dysfunction of the lymphatic system leads to tissue fluid accumulation and edema, a condition commonly known as lymphedema, illustrating the critical function of the lymphatic system in fluid homeostasis. For instance, inherent mutations can predispose patients to primary lymphedema.^3–5 Similarly, removal of the lymph node during cancer treatment⁶ or parasite infection, such as lymphatic filariasis,⁷ can result in secondary lymphedema. In addition to its role in fluid homeostasis, the lymphatic system plays a critical role in the onset and progression of diverse pathological conditions, including tumor metastasis and obesity.^8,9 Although several genes required for lymphatic development have been identified, the generation of novel and effective therapies to improve patient outcomes awaits a more comprehensive understanding of the coordination among endogenous agonistic and antagonistic factors governing lymphatic fate.

During development, lymphatic vessels emerge at approximately 3 days postfertilization (dpf) in zebrafish, embryonic day (E) 10 in mice, and embryonic week 6 to 7 in humans, and only after the formation of blood vessels and initiation of blood circulation.^10–13 Lineage tracing in mice embryos and in vivo time lapse analyses in zebrafish demonstrated that the lymphatic vessels sprout from the veins, as proposed by Sabin.^12,14,15 In zebrafish, lymphatic endothelial cells (LECs) originate from the presumptive parachordal vessel and migrate out to form the thoracic duct (TD) between the dorsal aorta and cardinal vein (CV).¹² At 5 dpf, an elaborate and extensive lymphatic network is present in zebrafish.^12,16,17 In mice, a subset of venous endothelial cells (ECs) in the CV becomes committed to the lymphatic lineage and begins to express Sox18 and Prox1, which subsequently migrate out of the CV after the gradient formed by vascular endothelial growth factor-C (VEGF-C).¹⁸ At least 2 distinct mechanisms have been suggested to explain the initial segregation of LECs from blood ECs and formation of the jugular lymph sacs, the first lymphatic organ to emerge during development.^19,20 Similarly, additional lymph sacs are formed along the anterior–posterior axis, which contribute to the primary capillary plexuses.²¹ At approximately E15.5, lymphatic vessels undergo morphogenetic changes to form mature lymphatic structures, such as capillaries, precollectors, and collecting vessels.²² Although several key factors promoting lymphatic development have been identified, including PROX1 (prospero homeobox protein 1),¹⁰ SOX18,²³ VEGF-C,^24,25 and VEGF receptor 3,^5,26,27 which seem to have evolutionarily conserved function, those that inhibit the differentiation of lymphatic vessels remain largely unknown.

Bone morphogenetic protein (BMP) molecules, which are members of the transforming growth factor-β (TGF-β) superfamily, are a potential candidate signaling pathway that may negatively regulate lymphatic development. During development, diverse components of BMP signaling, including BMP2, influence vascular development in vertebrates,^28–32 and dysregulation of BMP signaling through genetic mutations has been implicated in diverse human pathological conditions, including hereditary hemorraghic telangiectasia and primary pulmonary arterial hypertension.^33–35 Mechanistically, BMP ligands transduce their activity via various signaling cascades. On ligand binding, activated type I BMP receptors phosphorylate Smad transcription factors SMAD1, SMAD5, and SMAD8, which along with the receptor-independent SMAD protein, SMAD4, translocate to the nucleus and regulate the expression of the target genes, such as the inhibitor of DNA-binding (Id) gene family.³⁶ In addition to the SMAD-dependent pathway, BMP ligands can transduce signaling via MAP kinase pathways.³⁷ During vascular development, BMP signaling seems to possess a context-dependent function to regulate diverse aspects of EC behaviors. For instance, BMP2 functions as a venous-specific proangiogenic cue in zebrafish³² and is essential for the stalk cell fate in the mouse hindbrain plexus,³¹ whereas BMP9 seems to function as a circulating vascular quiescence factor.^38–40

In this article, we present compelling evidence indicating a novel role of BMP2 signaling as a negative modulator for the lymphatic fate. We show that BMP2 signaling inhibits the differentiation of LECs in zebrafish embryos and mouse embryoid bodies (EBs). Mechanistically, we demonstrate that BMP2 signaling promotes the expression of miRNAs, including miR-31 and miR-181a, to negatively regulate PROX1 expression, in a SMAD-dependent but not ERK1/2-dependent manner. Although several factors have been described to negatively modulate lymphangiogenesis postnatally, including endostatin, TGF-β, and IFN-γ,^41–46 these data, to our knowledge, are the first in vivo demonstrations of a negative modulator for lymphatic fate during vertebrate development.

Methods

An expanded Material and Methods section can be found in the Online Data Supplement.

Zebrafish Husbandry, Heatshock Conditions, Drug Treatment, and MO Injection

Zebrafish (Danio rerio) embryos were raised as previously described.⁴⁷ The following transgenic fish lines were used: Tg(fli1a:EGFP)^y1;⁴⁸ Tg(fli1a.ep:DsRedEx)^um13;49 Tg(kdrl:GFP)^s843;⁵⁰ Tg(hsp70l:bmp2b)^fr13;51 Tg(Bre:nmCherry)^nc24 (this study); and TgBAC(prox1:KalT4-UAS: uncTagRFP)^nim5 (this study). Microinjections of morpholino anti-sense oligonucleotide (MO) were performed as previously described.³⁰ Conditions for heatshock or drug treatments and sequence of MOs can be found in the Online Data Supplement. Detailed description of the construction of Tg(Bre:nmCherry)^nc24 and TgBAC(prox1:KalT4-UAS: uncTagRFP)^nim5 transgenic lines can be found in the Online Data Supplement.

Cell Culture and Embryoid Body Differentiation

HMVEC-dLy (Cat#CC-2810; Lonza) were grown on 0.1% gelatin-coated plates with EGM2-MV (Lonza) media and used at passages 3 to 7. HMVEC-dLy were stimulated with 50 ng/mL BMP2 (Sigma H4791) or 5 ug/mL actinomycin D (Sigma A1410). siRNA silencing was performed with SMAD4 (SI00076020; Qiagen) and Control Allstar siRNAs (Qiagen) and Lipofectamine 2000 (Invitrogen).

Embryonic stem cells were cultured on irradiated mouse embryonic fibroblast feeder cells in Dulbecco modified Eagle medium (Invitrogen) supplemented with 15% knockout serum replacement (Invitrogen), leukemia inhibitory factor-conditioned media, 0.1 mmol/L nonessential amino acid (MEM-NEAA; Invitrogen), 2 mmol/L L-glutamine (Invitrogen), 1 mmol/L sodium pyruvate (Invitrogen), 1% penicillin and streptomycin (Invitrogen), and 0.1 mmol/L 2-mercaptoethanol (Sigma). When embryonic stem cells reached 70% to 80% confluency, cultures were dissociated with 0.25% Trypsin-EDTA (Invitrogen) and then transferred to gelatin-coated tissue culture dishes to adapt to feeder-free conditions. During this adaptation stage, embryonic stem cells were grown in Iscove modified Dulbecco medium (Invitrogen) supplemented with 15% knockout serum replacement, leukemia inhibitory factor-conditioned medium, 2 mmol/L L-glutamine, 1% penicillin and streptomycin, and 0.1 mmol/L 1-thioglycerol (Sigma). After 2 days in feeder-free culture, mouse embryonic stem cell colonies were dissociated, and hanging drops containing 540 embryonic stem cells in 15 μL differentiation medium were prepared. EBs formed within 2 days of hanging drop cultures in differentiation medium containing Iscove modified Dulbecco medium, 15% fetal bovine serum (Gemini), 50 μg/mL ascorbic acid (Sigma), and 2 mmol/L L-glutamine. To differentiate EBs, 2 days postcoitum (dpc) EBs were plated on gelatin-coated 8-well chamber slides (Laboratory-Tek) or 12-well dishes at a density of 6.25 EB/cm² in complete differentiation media without leukemia inhibitory factor. To manipulate EBs, BMP2, VEGF-C (2176-VC; R&D System), Noggin (6057-NG; R&D System), and DMH1 (D8946; Sigma) were used. At 5 and 7 dpc, EBs were treated with 0.1% bovine serum albumin or growth factors (200 ng/mL VEGF-C, 100 ng/mL BMP2, 250 ng/mL Noggin, 0.1 Limol/L DMH1). At 8 dpc, EBs were analyzed by immunocytochemistry or flow cytometry with CD31 and LYVE1 antibodies. Immunohistochemistry samples were imaged by confocal microscopy (Leica SP5), equally processed by Adobe Photoshop, and quantified by Image J (National Institutes of Health). For quantification, images were acquired at 2 to 3 random fields of the periphery of ≥4 EBs per condition per experiment. Each experiment was performed ≥3 times. The ratio of lymphatic vessel area to total vascular area was determined by differential immunofluorescence stains (LYVE1/CD31).

Zebrafish Imaging and Reconstruction

At 4 dpf, Tg(fli1a.ep:DsRedEx)^um13; Tg(kdrl:GFP)^s843 or Tg(kdrl:GFP)^s843; TgBAC(prox1:KalT4-UAS:uncTagRFP)^nim5 double-transgenic embryos were dechorionated, anesthetized with tricaine, and embedded in 1% agarose/phosphate-buffered saline on a glass-bottom Petri dish (MatTek). Trunk regions immediately dorsal to the yolk extension were imaged with a Nikon T–E Eclipse confocal microscope with Hamamatsu camera (C9100-50). Volocity software (Perkin Elmer) was used to generate 3-dimensional optical projections.

Quantification of Lymphatic Development in Zebrafish Embryos

To quantify lymphatic defects in developing transgenic embryos, the percentage of TD formation was quantified in 10 consecutive segments in Tg(fli1a.ep:DsRedEx)^um13;Tg(kdrl:EGFP)^s843 or Tg(kdrl:GFP)^s84;TgBAC(prox1:KalT4-UAS:uncTagRFP)^nim5 double-transgenic embryos. At least 10 embryos per genotype/condition were analyzed from ≥3 independent clutches.

Statistical Analyses

All data were analyzed using the 2-tailed Student t test assuming unequal variance. P<0.05 was considered significant.

Results

BMP2 Signaling Inhibits Zebrafish Lymphatic Development

We have previously demonstrated that Bmp2b (an ortholog for mammalian BMP2) signaling components are enriched in the caudal vein plexus and surrounding mesenchyme and are required for venous angiogenesis in the developing zebrafish.^30,32 Interestingly, Bmp2b seems to increase the number of venous ECs without affecting cell proliferation.^30,32 Considering that LECs have a venous origin during development, it is tempting to speculate that an excessive level of Bmp2b signaling may increase the number of venous ECs at the expense of presumptive LECs.

To examine this possibility, we first overexpressed Bmp2b in zebrafish embryos with Tg(fli1a:DsRed);Tg(kdrl:EGFP) double transgenic background and examined the effects on lymphatic development. Because kdrl (kinase insert domain receptor-like) is expressed only in BECs, whereas fli1a (friend leukemia integration 1a) is expressed in both LECs and BECs, emerging LECs can be easily distinguished (Online Figure I).^12,17 We found that Bmp2b-overexpressing embryos completely lack any detectable lymphatic structures (Figure 1A and 1B and Online Figure I). For instance, although wild-type embryos have a clearly separated TD between dorsal aorta and caudal vein, it is largely absent in embryos with excessive Bmp2b expression (Figure 1A and 1B). To further confirm the lymphatic defects in Bmp2b-overexpressing embryos, we used another transgenic line, TgBAC(prox1a:KalT4-UAS:uncTagRFP)^nim5, hereafter referred to as Tg(prox1a:RFP). In Tg(prox1a:RFP), strong RFP expression can be detected in developing lens, heart, liver, neural tube, and somites, in addition to lymphatic vessels, which is reminiscent of the previously reported expression of endogenous prox1a⁵² (Online Figure II).

Figure 1. Bmp2b signaling may function as an antilymphangiogenic cue in zebrafish embryos.

A, Transverse section taken from the trunk region of 4 days postfertilization (dpf) wild-type and Tg(hsp70l:bmp2b)^fr13 embryos in Tg(kdrl:EGFP)^s843;Tg(fli1a.ep:DsRedEx)^um13 background. Overexpression of Bmp2b abrogates thoracic duct (TD; arrowhead in wild-type) formation. B, Quantification of percent TD formation measured over 10 segments within the trunk region. Compared with wild-type embryos (n=50), the formation of TD is substantially reduced in embryos with an elevated level of Bmp2b activity by heatshock treatment at either 26 hours postfertilization (hpf; n=39) or 50 hpf (n=36). C, Lateral view of wild-type and Bmp2b-overexpressing embryos at 4 dpf. TD (arrow) is absent in Tg(hsp70l:bmp2b)^fr13 embryo. Embryos shown have TgBAC(prox1:KalT4-UAS:uncTagRFP)^nim5;Tg(kdrl:GFP)^s843 background. D, Quantification of percent TD formation in wild-type and Tg(hsp70l:bmp2b)^fr13 embryos measured over 10 segments (n=36 for wild-type embryos, and n=32 for Tg(hsp70l:bmp2b)^fr13 embryos). E, Number of ventral sprouts from the TD (*) increased in 1 μm DMH1-treated embryos compared with DMSO-treated control embryos. F, Quantification of the numbers of ventral sprouts from the TD in control and DMH1-treated embryos (n=23 for DMSO-treated control embryos and n=25 for DMH1-treated embryos). Compared with control group, DMH1-treated embryos had a significant increase in the number of ventral sprouts (P<0.001). G, Quantification of flow-sorted prox1a:RFP⁺, fli1:EGFP⁺ lymphatic endothelial cells relative to all fli1:EGFP⁺ ECs in DMSO-treated or DMH1-treated embryos (N=3). CV indicates cardinal vein; and DA, dorsal aorta. Scale bars: 25 μm (A), 50 μm (C), and 100 μm (E).

In wild-type embryos, consistent with previous reports,¹² prox1:RFP⁺ and fi1a:EGFP⁺ cells first emerged along the horizontal myoseptum as a string of individual LECs approximately at 50 hours postfertilization. Subsequently, RFP⁺ LECs migrated dorsally and ventrally along arterial intersegmental vessels to form lymphatic structures (Online Figure IIB and IIC). On Bmp2b overexpression, we found that RFP expression within lymphatic vessels was selectively diminished, whereas the expression was maintained in other organs, indicating that Bmp2b may regulate the expression of prox1a in a tissue-specific manner (Online Figure II). Consequently, at 60 hours postfertilization, LECs were completely absent in Bmp2b-overexpressing embryos, which led to failure of forming the TD at 4 dpf as observed in Tg(fli1a:RFP);Tg(kdrl:EGFP) double-transgenic background (Figure 1C and 1D). Conversely, inhibition of Bmp2b signaling seems to promote lymphatic growth. Treatment with DMH1, a potent chemical inhibitor for BMP type I receptors, induced ectopic lymphangiogenesis and led to precocious and excessive ventral sprouting of the TD (Figure 1E and 1F) and increased the number of LECs in an otherwise phenotypically wild-type zebrafish embryos (Figure 1G), further supporting the idea that Bmp2b signaling may negatively regulate lymphatic development in zebrafish.

SMAD, but Not ERK1/2, Mediates Antilymphangiogenic Effects of BMP2 Signaling

Because Bmp signaling activates 2 main downstream effectors, SMADs and ERK1/2, in zebrafish development,^53,54 we next examined which downstream effector mediates the effects of Bmp2b signaling during lymphatic development. Inhibition of Erk1/2 activity by U0126 did not alleviate the lymphatic defects in embryos with excessive Bmp2b activity. However, inhibition of SMAD activity by addition of DMH1 substantially restored the lymphatic defects in Tg(hsp70l:bmp2b)^fr13 embryos (Figure 2A and 2B), indicating that SMAD activity is essential to mediate Bmp2b signaling during lymphatic development.

Figure 2. Antilymphangiogenic effects of Bmp2b signaling are mediated by SMAD proteins.

A, Four days postfertilization Tg(hsp70l:bmp2b)^fr13;TgBAC(prox1:KalT4-UAS:uncTagRFP)^nim5 embryos treated with DMSO, 1 μm DMH1 (Smad1/5/8 inhibitor), or 10 μm U0126 (Erk1/2 inhibitor). Arrow points the TD in DMH1-treated embryos. Asterisks mark lack of TD. B, Quantification of percent TD formation in DMSO, DMH1, or U0126-treated Tg(hsp70l:bmp2b)^fr13 embryos measured over 10 segments (n=30 for DMSO treatment, 37 for DMH1 treatment, and 32 for U0126 treatment). The percentage of Tg(hsp70l:bmp2b)^fr13 embryos with lymphatic structure development increased significantly between DMSO- and DMH1-treated embryos (25% and 94%, respectively; P<001), whereas U0126-treated embryos were unaffected (21%). CV indicates cardinal vein; DA, dorsal aorta; and TD, thoracic duct. Scale bar, 50 μm.

To examine the effect of BMP2 signaling on mammalian lymphatic specification, we used a modified mouse EB lymphatic differentiation model (Figure 3A). We found that differentiation of LECs spontaneously occurs within EBs as previously reported (Figure 3B).^55,56 In this model, VEGF-C treatment starting at 5 days postplating, which is equivalent to E8.5 when the endothelial lineage emerges,⁵⁷ substantially increased the number of LECs (Figure 3B), consistent with previous reports.^55,58 The CD31⁺/LYVE1⁺ cells derived from mouse EBs also express podoplanin, indicating that these cells are fully differentiated LECs (Figure 3C). To examine whether BMP2 inhibited lymphangiogenesis in this model, we cotreated EBs with BMP2 and VEGF-C. Cotreatment with BMP2 completely abrogated VEGF-C–mediated LEC induction in the EB periphery and substantially decreased the number of CD31⁺/LYVE1⁺ LECs (Figure 3D and 3E). The inhibitory effect of BMP2 stimulation was completely blocked when EBs were treated with DMH1, further corroborating that BMP2-mediated inhibition of lymphangiogenesis requires SMAD activity. Moreover, we found that inhibiting BMP signaling through the addition of the endogenous BMP inhibitor Noggin to mouse EBs resulted in increased LEC specification, further illustrating that BMP2 signaling may function to inhibit lymphatic development (Online Figure III). Taken together, these data suggest that BMP2 signaling is a highly conserved negative modulator of lymphatic development throughout vertebrates, and its antilymphangiogenic effect is mediated by SMAD activity.

Figure 3. BMP2 attenuates lymphatic endothelial cell (LEC) differentiation from cultured mouse embryoid bodies (EBs).

A, Mouse EBs were prepared by hanging drop and then grown in 2-dimensional culture to induce lymphatic differentiation. At EB day 8, the expression of Vegfr2 and Prox1 increased together (arrow), indicating that LECs emerge within the EBs. B, Differentiation of lymphatic endothelial cells (CD31⁺/LYVE1⁺) in mouse EBs was increased in the presence of VEGF-C. C, Differentiated lymphatic endothelial cells within mouse EBs express hallmark of lymphatic lineage–specific genes, including LYVE1 (red) and Podoplanin (blue). D, Representative micrographs of the periphery of EBs on treatment with VEGF-C, BMP2, VEGF-C and BMP2, or with VEGF-C, BMP2, and DMH1. E, Quantification of (D) measuring the ratio of lymphatic vessel area (LYVE1⁺/CD31⁺) to total vasculature (CD31⁺) at the EB periphery (N=3; 4–6 EBs analyzed per condition per experiment). Scale bars: 50 μm (B), 50 μm (C), and 100 μm (D).

Reduced Level of BMP Signaling Activity in LECs

Because overexpression of Bmp2b adversely affected lymphatic development, we hypothesized that Bmp2b signaling may need to be suppressed in developing LECs. To further examine whether developing LECs lack BMP2 signaling activity, we generated a transgenic zebrafish, Tg(Bre:nmCherry)^nc24, that expresses nuclear mCherry under the regulation of minimal BMP response element (BRE) from mouse Id2 gene⁵⁹ (Online Figure IV). We reasoned that Id2 promoter would more faithfully represent Bmp2b signaling in the zebrafish vasculature because we previously found that Id2 is the most highly induced Id family member in Tg(hsp70l:bmp2b)^fr13 after heatshock induction.³²

In developing zebrafish, mCherry expression faithfully recapitulated the endogenous Bmp2b activity (Online Figure IV) and overlapped with previously reported BMP signaling activity during development.⁶⁰ For instance, mCherry colocalized with p-SMAD-1/5/8 (Online Figure IV). In addition, mCherry expression was elevated in Bmp2b-overexpressing embryos and was decreased in embryos with reduced Bmp2b signaling (Online Figure IV), suggesting that BRE:mCherry expression can be used as a surrogate measure for Bmp2b signaling activity. During development, mCherry expression was widely detected at gastrula stages. At later stages, mCherry could be widely detected in diverse tissues and cell types, including all fli1a⁺ BECs within the arteries and veins (Online Figure IV). Unexpectedly, mCherry expression was completely absent in LECs at 3 dpf and the TD at 4 dpf zebrafish embryos (Figure 4A–4C), suggesting the lack of Bmp2b signaling activity in developing LECs.

Figure 4. Different levels of BMP signaling activity in blood endothelial cells (BECs) and lymphatic endothelial cells (LECs).

A, Emerging parachordal lymphangioblasts (red arrows), the lympahtic progenitors in zebrafish, lack BRE:mCherry expression at 60 hours postfertilization, whereas endothelial cells (ECs) within intersegmental vessels (yellow arrowheads) are strongly positive for BRE:mCherry. B, LECs are largely devoid of BMP signaling activity. Arrows point to mCherry⁻/EGFP⁺ LECs, and arrowheads point to mCherry⁺/EGFP⁺ BECs. C, Quantification of the numbers of EGFP⁺ and EGFP⁺/mCherry⁺ ECs within the dorsal aorta, cardinal vein, and thoracic duct (TD; N=3; 10 embryos per experiment). D, Transverse section of embryonic day (E) 11.5 mouse embryos stained with PROX1 and P-SMAD1/5/8 antibodies. Two sections adjacent to each other are shown. p-SMAD1/5/8 staining is largely absent in developing LS but is strongly shown (arrowheads) in cardinal vein (CV). E, Transverse section of E10.5 and E11.5 Tg(BRE:EGFP) mouse embryos through the jugular region. PROX1 (red) and PECAM (blue) are visualized by antibody staining. Areas containing the wall of CV are shown. Arrows point to LECs within the vein wall, and arrowheads point to delaminated LECs. F, Quantification of the numbers of EGFP⁺ and EGFP⁺/PROX1⁺ ECs in CV and thoracic duct (TD; N=4). DA indicates dorsal aorta; ISV, intersegmental vessel; LS, lymph sac; and PL, parachordal lymphangioblast. Scale bars: 50 μm (A), 50 μm (B), 20 μm (C), and 20 μm (E).

To examine whether BMP signaling activity within emerging LECs is similarly modulated during mammalian development, we first examined the status of Bmp2b signaling within PROX1⁺ LECs in mouse embryos. At E11.5, pSMAD1/5/8 staining can be detected in the majority of ECs within the developing CV. In contrast, in nearby LECs, pSMAD1/5/8 staining was largely absent, supporting our idea that BMP2 signaling activity is attenuated in LECs during development (Figure 4D). Similarly, in Tg(BRE:EGFP)³⁶ transgenic mouse embryos, where BMP signaling activity can be measured by the expression level of EGFP, we found that EGFP expression was similarly reduced in LECs compared with BECs within CVs (Figure 4E and 4F). Taken together, our data allude to an intriguing possibility that BMP2 signaling may need to be suppressed to allow the emergence of LECs. This notion is also consistent with previous findings that demonstrated the lack of BMP2 transcript in LECs.⁶¹

BMP2 Signaling Negatively Regulates PROX1 Transcripts in a SMAD-Dependent Manner

To uncover the molecular mechanisms responsible for BMP2-mediated inhibition of lymphangiogenesis, we isolated fli1a:EGFP⁺ ECs from wild-type and Bmp2b-overexpressing zebrafish embryos in Tg(fli1a:EGFP)^y1 background and examined the expression levels of endothelial-specific transcripts critical for active angiogenesis and lymphangiogenesis (Figure 5A). Among these markers, only the expression of prox1a, the zebrafish ortholog of mammalian Prox1,^12,62 was significantly reduced, suggesting that Bmp2b signaling may attenuate lymphatic development by antagonizing prox1a expression during early stages of lymphangiogenesis in zebrafish. Similarly, BMP2 stimulation significantly reduced basal PROX1 expression level in fully differentiated human dermal microvascular LECs (hLECs), demonstrating evolutionary conserved effects of BMP2 signaling on PROX1 activity (Figure 5B).

Figure 5. Bmp2b signaling selectively represses the expression of prox1a in zebrafish embryos.

A, Expression of prox1a mRNA substantially decreased on Bmp2b stimulation. Quantitative reverse-transcription polymerase chain reaction on selected markers for lymphatic endothelial cells (LECs) and blood endothelial cells (ECs) from fli1a:EGFP⁺ ECs isolated from 72 hours postfertilization (hpf) wild-type and Tg(hsp70l:bmp2b)^fr13 embryos after heatshock at 50 hpf. B, Expression level of PROX1was substantially downregulated on BMP2 stimulation in hLECs starting at 60 minutes (N=4). C, Inhibition of SMAD4 relieves the suppression of PROX1 by BMP2 signaling.

To determine the downstream effectors that mediate the inhibitory effects of BMP2 on the level of PROX1 expression, we examined the function of SMAD and ERK, 2 main effectors of BMP2, in this process. Consistent with our finding that Smad inhibition restored Bmp2b-induced lymphatic defects in zebrafish embryos (Figure 2), siRNA knockdown of SMAD4 in hLECs prevented PROX1 downregulation by BMP2 (Figure 5C). Considering that lymphatic development is highly sensitive to small fuctuations in the level of Prox1 expression,^63,64 a reduction in Prox1 expression by BMP2 signaling is likely to substantially impede lymphatic development.

miR-31 and miR-181a Function Downstream of BMP2 Signaling to Modulate PROX1 Expression

Considering that the expression of sox18, which is known to induce transcription of prox1a during lymphatic development,^23,65 was largely unchanged (Figure 5A), it is possible that Bmp2b signaling may modulate the level of prox1a transcript at the post transcriptional level. Because BMP signaling can influence the processing of miRNAs,⁶⁶ it is tempting to speculate that the antilymphangiogenic effect of BMP2 signaling may be mediated by a miRNA-dependent mechanism. To explore this possibility, changes in miRNA expression on BMP2 stimulation in hLECs were evaluated using miRNA polymerase chain reaction arrays. An elevated level of BMP2 signaling significantly altered the expression level of multiple miRNAs (Figure 6A).

Figure 6. BMP2/Bmp2b signaling induces expression of miR-31 and miR-181a in zebrafish embryos and human lymphatic endothelial cells (hLECs).

A, miRNA polymerase chain reaction array expression analysis of cultured HDMVEC-Ly (hLECs) cells after treatment with 50 ng/mL BMP2 for 1 hour. Several miRNAs with significant expression changes were found, including miR-31 and miR-181a (N=4). Only miRNAs whose expression level exhibited statistically significant changes (P<0.05) in response to BMP2 stimulation are shown. B, Time course of miR-31 and miR-181a expression from fli1a:EGFP⁺ ECs isolated at indicated times from wild-type and Tg(hsp70l:bmp2b)^fr13 embryos after heatshock at 26 hours postfertilization (N=4). C, siRNA knockdown of SMAD4 in LECs prevented BMP2-mediated induction of miR-31 and miR-181a. Additionally, miR-31 and miR-181a expression levels after BMP2 treatment in the SMAD4 knockdown were decreased compared with vehicle treatment, suggesting potential negative feedback regulation (N=4). D, Pretreatment of hLECs with 5 μg/ mL actinomycin D to block transcription completely inhibited BMP2-induced expression of miR-31 and miR-181a, suggesting that BMP2-induced upregulation of miR-31 and miR-181a may require transcriptional activation of these miRNAs (N=4).

Interestingly, 2 of the upregulated miRNAs, miR-31 and miR-181a, have been previously reported to bind the Prox1 3′-UTR, the master transcription factor to initiate and maintain lymphatic fate, leading to transcript degradation and translational inhibition and subsequent maintenance of BEC identity.^67,68 Similarly, ECs isolated from Bmp2b-overexpressing zebrafish embryos revealed increased transcript levels of miR-31 and miR-181a (Figure 6B). In zebrafish and mice, we found that both miR-31 and miR-181a are expressed in the BECs during lymphatic specification (Online Figure V), suggesting the role of these miRNAs in mediating antilymphangiogenic effects of BMP2 signaling. Consistent with previous reports that miR-31 and miR-181a post transcriptionally suppress PROX1 mRNA level in humans and mice,^67,68 we found functionally conserved miR-31-binding and miR-181a-binding sites in zebrafish prox1a 3′-UTR (Online Figure V). To further test that zebrafish miR-31 and miR-181a could target the zebrafish prox1a 3′-UTR, a luciferase reporter containing the full-length of zebrafish prox1a 3′-UTR was generated. We found that both miR-31 and miR-181a mimics recognized the full-length zebrafish prox1a 3′-UTR, leading to a significant decrease in luciferase activity (Online Figure VI). In addition, injection of miRNA mimics to single-cell-stage zebrafish embryos resulted in a drastic decrease in the level of prox1a expression (Online Figure VI), further supporting our idea that miR-31 and miR-181a may mediate Bmp2b signaling and attenuate prox1a expression. Similar to prox1a, other predicted targets of miR-31 and miR-181a, such as rhot2 (in case of miR-31) and ptpn2a (in case of miR-181a), were similarly downregulated in embryos overexpressing Bmp2b (Online Figure VI).

To further investigate how BMP2 signaling increases the level of these miRNAs, we first examined the role of SMAD activity. Knockdown of SMAD4 by siRNA completely abrogated the BMP2-mediated upregulation of miR-31 and miR-181a (Figure 6C). Unexpectedly, we observed that miR-31 and miR-181a expression levels after BMP2 treatment in SMAD4 knockdown were decreased compared with vehicle treatment, suggesting the presence of potential negative feedback regulation. In addition, treatment with actinomycin D to inhibit transcription completely abolished the effects of BMP2 signaling on the levels of miR-31 and miR-181a in hLECs (Figure 6D). Therefore, it is likely that BMP2 signaling promotes transcription of miR-31 and miR-181a in a SMAD-dependent manner.

Next, we examined whether inhibition of miR-31/miR-181a would abrogate antilymphangiogenic effects of Bmp2b signaling and, therefore, at least partially restore the lymphatic defects caused by excessive Bmp2b signaling. We found that blocking the activity of either miR-31 or miR-181a by MOs (validated in Online Figure VII) alleviated lymphatic defects in Bmp2b-overexpressing zebrafish embryos (Figure 7A). A significant portion of MO-injected embryos contained at least partially formed TD (≈70% of embryos), and a sizable portion of embryos generated complete TD (between 10% and 25%; Figure 7A and 7B). Therefore, miR-31 and miR-181a are likely to be essential to mediate antilymphangiogenic effects of Bmp2b/BMP2 signaling during lymphatic development.

Figure 7. BMP2 signaling represses the expression of PROX1 via miRNA-dependent mechanism.

A, Confocal projection of Tg(hsp70l:bmp2b)^fr13 embryos injected with 5 ng control, miR-31, or miR-181a MOs, followed by heatshock at 26 hours postfertilization. Inhibition of miR-31 or miR-181a can partially rescue the lymphatic defects induced by Bmp2b overexpression. Arrows point rescued thoracic duct (TD) in MO-injected embryos. B, Quantification of percent TD formation in control, miR-31, or miR-181a MO-injected wild-type and Tg(hsp70l:bmp2b)^fr13 embryos measured over 10 segments (n>40 for all conditions). TD formation was largely unaffected by MO knockdown of miR-31 or miR-181a in wild-type embryos. However, MO-mediated knockdown of miR-31 or miR-181a in the Tg(hsp70l:bmp2b)^fr13 background resulted in 50% to 70% of embryos forming at least a partial TD and between 20% to 25% forming a complete TD compared with 25% and 0% in control and MO-injected, respectively (P<0.001 for both groups). Scale bar, 50 μm. C, During development, active BMP2 signaling promotes the expression of miR-31/miR-181a in blood endothelial cells (blue) and, therefore, aids BECs to maintain their fate as venous endothelial cells. However, in presumptive lymphatic endothelial cells (green), the activity of BMP2 signaling is attenuated by unknown mechanism, thereby releasing miRNA-mediated repression of PROX1. CV indicates cardinal vein; and DA, dorsal aorta.

Discussion

Our data present compelling evidence indicating that BMP signaling functions as a negative regulator for lymphatic fate during vertebrate development. We found that lymphatic progenitors are largely devoid of BMP signaling activity. Because it has been reported that several Bmp ligands are expressed in ECs or their vicinity during zebrafish development, including Bmp2b and Bmp4,⁵² it remains to be identified which specific Bmp ligands function as the major regulator for the lymphatic fate.

In our study, we found that BMP2 signaling strongly inhibits the expression of PROX1, the master regulator for the lymphatic fate, in a SMAD-dependent and miRNA-dependent manner. Taken together, our data suggest that BMP2 signaling may modulate fate specification within ECs, similar to its role in other cell types, including osteoblasts, neurons, cardiomyocytes, and hepatocytes.^69–72 Combined with our previous reports on the role of BMP2 signaling in promoting venous angiogenesis,^30,32 our results illustrate the context-dependent role of BMP2 signaling during vascular development.

During lymphatic development, miRNAs seem to be important mediators for antilymphangiogenic effects of BMP2 signaling. Using miRNA polymerase chain reaction arrays, we identified that several miRNAs are upregulated by BMP2 signaling, including miR-31 and miR-181a, in a SMAD-dependent manner (Figure 6C). Considering that actinomycin D treatment completely inhibited the BMP2-induced upregulation of miR-31 and miR-181a expression, it is likely that active BMP2 signaling influences transcription of these miRNAs. However, we could not identify consensus SMAD-binding element within the promoter region of miR-31 and miR-181a loci. It is possible that BMP2/SMAD-dependent transcriptional activation of miR-31 and miR-181a is mediated by novel and yet-unidentified variants of SMAD-binding elements. Alternatively, activated SMAD may bind to the unknown enhancer elements that influence the transcription of miR-31 and miR-181a because SMADs can bind to enhancer located outside of promoter region.⁷³ Further analyses of the promoter region of miRNAs that are upregulated by BMP2 stimulation are warranted to delineate precise molecular mechanisms on SMAD-mediated transcriptional regulation of these miRNAs. In Bmp2b-overexpressing zebrafish embryos, we found that expression of prox1a:RFP is selectively down-regulated in developing LECs without affecting the expression in other tissues. The tissue-specific enrichment of miR-31 and miR-181a, which seem to be essential to mediate anti-lymphangiogenic effects of Bmp2b signaling in BECs, may contribute to creating the tissue-specific response of Bmp2b overexpression in developing zebrafish embryos.

Because each miRNA can bind a wide variety of target mRNAs,⁷⁴ additional targets of miR-31 and miR-181a may augment the antilymphatic effects of BMP2 signaling in coordination with PROX1. Although we cannot formally exclude this possibility, we propose that PROX1 is likely to be the primary targets for these miRNAs during lymphatic development. First, PROX1 is the master transcription factor for the lymphatic fate. By regulating PROX1 expression, which is autoregulatory and continuously required to maintain the lymphatic fate, a modest decrease of its expression in early lymphangiogenesis can be amplified to manifest severe lymphatic defects.⁶³ Second, it is the only lymphatic marker whose expression is downregulated in response to BMP2 signaling. Other lymphatic markers, including LYVE1 and VEGFR3, were not altered on BMP2 stimulation in hLECs or zebrafish embryos.

It is interesting that members of let-7 family are among the miRNAs that are downregulated by excessive BMP2 signaling (Figure 6A), considering a previous report that stated that let-7 negatively regulates expression of TGF-βR1 to decrease TGF-β signaling,⁷⁵ which is known to inhibit lymphatic re-generation^43,44 Therefore, downregulation of let-7 miRNA by BMP2 signaling is likely to elevate the level of TGF-β signaling, which in turn further inhibits lymphatic development and provides an additional mechanism for BMP2 signaling to modulate lymphatic development, independent of miR-31/181a and PROX1-mediated regulation.

Our analyses suggest a strong negative correlation between BMP2 signaling activity and the formation of LECs during development. Compared with BECs, the level of BMP2 signaling within LECs seems to be substantially attenuated. Because diverse modifiers of BMP signaling are expressed within ECs,⁷⁶ it is tempting to speculate that endogenous antagonist of BMP2 may be preferentially localized or expressed near the presumptive LECs to allow them to adopt the lymphatic fate (Figure 7C). Although it is not clear how activity of BMP2 signaling is attenuated in LECs, or how other BMP ligands may play an additional role in this process, our model is consistent with previous finding that LECs do not express high level of Bmp2 transcript.⁶¹ Considering that developmental cues that negatively regulate the emergence of LECs are poorly understood, our findings in this report could help us to formulate more comprehensive ideas regarding lymphatic development and to construct a theoretical framework to develop therapeutic manipulation of lymphangiogenesis in development and diseases.

Novelty and Significance.

What Is Known?

• Lymphatic dysfunction can result in debilitating lymphedema.

• Lymphatic vasculature arises from the preexisting venous system driven by prospero homeobox protein 1 (PROX1) expression and is promoted through several growth factors, including vascular endothelial cell growth factor-C (VEGF-C).

• Bone morphogenetic protein (BMP) signaling is a critical regulator of angiogenesis, mural cell recruitment, and cardiomyocyte specification.

What New Information Does This Article Contribute?

• BMP2 signaling negatively impacts specification of lymphatic endothelial cells in zebrafish and mouse embryonic stem cells.

• BMP2 signaling decreases the level of PROX1 transcript via SMAD-dependent and miRNA-mediated mechanisms.

The lymphatic vascular system mirrors the blood vascular system and performs several essential physiological functions, including removing excess fluids from peripheral tissues and serving as the main conduit for trafficking immune cells in the body. Subsequently, when the lymphatic system fails to function properly becaue of genetic defects, disease, or after surgical lymph node removal for cancer treatment, patients are at risk for development of lymphedema or disturbed immune responses. Additionally, in cancer metastasis, the lymphatic system is often used by cancerous cells to spread to distal tissues. Therefore, in both lymphedema and cancer, developing new therapies that can modulate lymphatic vessel growth and function are imperative to improve patient outcomes. However, to date, no pharmacological therapies are available to alleviate symptoms associated with lymphatic dysfunction. In this study, we have identified BMP2 signaling as a novel negative regulator of lymphatic specification during vertebrate development. BMP2 signaling inhibits the expression of PROX1, required to retain lymphatic identity, by inducing the expression of miRNAs targeting PROX1 transcripts. Manipulation of BMP2 signaling alone or in combination with other lymphangiogenic factors may provide more effective therapeutic avenues to regulate lymphatic vessel growth and function.

Nonstandard Abbreviations and Acronyms

BEC

blood endothelial cell

BMP

bone morphogenetic protein

CV

cardinal vein

EB

embryoid body

fli1a

friend leukemia integration 1a

kdrl

kinase insert domain receptor like

LEC

lymphatic endothelial cell

MO

morpholino anti-sense oligonucleotide

PROX1

prospero homeobox protein 1

TD

thoracic duct

VEGF-C

vascular endothelial growth factor-C