Control of retinoid levels by CYP26B1 is important for lymphaticvascular development in the mouse embryo

Josephine Bowles a, Genevieve Secker b, Christelle Nguyen a, Jan Kazenwadel b, Vy Truong a,Emmanuelle Frampton a, Cameron Curtis a, Renae Skoczylas a, Tara-Lynn Davidson a,Naoyuki Miura c, Young-Kwon Hong d, Peter Koopman a, Natasha L. Harvey b,e,Mathias François a,n

a Institute for Molecular Bioscience, The University of Queensland, Brisbane QLD 4072, Australiab Division of Haematology, Centre for Cancer Biology, SA Pathology, Adelaide SA 5000, Australiac Department of Biochemistry, Hamamatsu University School of Medicine, 431-3192 Hamamatsu, Japand University of Southern California, Norris Comprehensive Cancer Center, 1450 Biggy St. NRT6501 M/C9601Los Angeles, CA 90033, USAe School of Medicine, University of Adelaide, Adelaide, Australia

a r t i c l e i n f o

Article history:Received 27 July 2013Received in revised form3 December 2013Accepted 9 December 2013Available online 19 December 2013

Keywords:Prox1MorphogenesisLymphangiogenesisLymphatic endothelial cellsCyp26b1Retinoic acid

a b s t r a c t

During embryogenesis, lymphatic endothelial progenitor cells first arise from a subset of blood vascularendothelial cells in the dorsolateral aspects of the cardinal veins. The molecular cues responsible fordefining the regionalisation of such a discrete pool of progenitors remain uncharacterised. Here weidentify a novel function for CYP26B1, an enzyme known to play a role in tissue morphogenesis by fine-tuning retinoic acid (RA) concentration, in regulating lymphangiogenesis. Cyp26b1-null mice, in whichRA levels are elevated, exhibited an increased number of lymphatic endothelial progenitor cells in thecardinal veins, together with hyperplastic, blood filled lymph sacs and hyperplastic dermal lymphaticvessels. Conversely, mice over-expressing Cyp26b1 had hypoplastic lymph sacs and lymphatic vessels.Our data suggest that RA clearance by CYP26B1 in the vicinity of lymphatic endothelial progenitor cells isimportant for determining the position and size of the progenitor pool specified. Our studies identify agenetic pathway that underpins the architecture of the developing lymphatics and define CYP26B1 as anovel modulator of lymphatic vascular patterning.

& 2013 Elsevier Inc. All rights reserved.

Introduction

The lymphatic vasculature is a highly specialized circulatorysystem that maintains fluid homeostasis, facilitates immune celltrafficking and mediates lipid absorption from the digestive tract(Alitalo et al., 2005). Similar to the blood vasculature, the lympha-tic vasculature is a highly branched network comprised of initiallymphatic vessels, pre-collector and collecting vessels whichfunction co-ordinately to return interstitial fluid and protein tothe bloodstream (Tammela and Alitalo, 2010). To date, the identityand cellular sources of signals that direct morphogenesis of thelymphatic vasculature are poorly understood.

Lymphangiogenesis initiates at around 9.0 days post coitum (dpc)in the mouse embryo, in a subpopulation of endothelial cells locatedin the anterior region of the cardinal veins. Recent identification ofkey transcription factors that drive lymphatic endothelial cell fate

specification has increased our understanding of the early steps inlymphatic vascular development (lymphangiogenesis) at a molecularlevel (Wigle et al., 2002; Wigle and Oliver, 1999; Yang et al., 2012).The specification of lymphatic endothelial cell fate is dependent onthe transcription factors SOX18 and COUP-TFII (Francois et al., 2008;Srinivasan et al., 2010) that directly activate transcription of thehomeobox transcription factor gene Prox1. PROX1 is both necessaryand sufficient to re-specify blood endothelial cells into lymphaticendothelial cells (Hong et al., 2002; Petrova et al., 2002; Wigle et al.,2002). The expression of Prox1 in venous endothelial cells defines apool of lymphatic endothelial progenitor cells that ultimately giverise to the initial lymphatic vascular plexus and lymph sacs via acombination of migrating streams of lymphatic endothelial cells(LECs) (Francois et al., 2012; Hagerling et al., 2013; Yang et al.,2012) and ballooning of small clusters of progenitors (Francois et al.,2012). Expression of Sox18 and Prox1 is only activated in endothelialcells on the dorsolateral side of the cardinal veins, but the identity ofthe inducing or repressing factor(s) that ensure this asymmetricdistribution of lymphatic endothelial cell precursors is not known(see for review Francois et al., 2011; Oliver and Srinivasan, 2010).

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/developmentalbiology

Developmental Biology

0012-1606/$ - see front matter & 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.ydbio.2013.12.008

n Corresponding author. Fax: þ61 7 3346 2101.E-mail address: m.francois@imb.uq.edu.au (M. François).

Developmental Biology 386 (2014) 25–33

Gradients of retinoic acid (RA) play pleiotropic roles duringdevelopment and are essential to control somitogenesis and left/right symmetry of vertebrate embryos (Niederreither and Dolle,2008). In tissues where RA plays an instructive role, the magnitudeof RA signal received by cells is under the control of the coordinatedexpression of RA-synthesising enzymes (retinaldehyde dehydro-genases; RALDH1, -2 and -3) and RA-catabolising enzymes (cyto-chrome P450 enzymes; CYP26A1, -26B1 and -26C1). Previousstudies have shown that loss of Cyp26b1 function is lethal andresults in severe malformations affecting limb, craniofacial andepidermal development (Yashiro et al., 2004). The first link betweenRA signalling and the lymphoid system was made by Mebius andcolleagues, who demonstrated that RA signalling plays a key role inthe initiation of lymph node development via the induction ofCXCL13 expression (van de Pavert et al., 2009). Given that lymphnodes develop in close spatial association with the developinglymphatic vasculature, it raises the possibility that the RA axismay also play a role in patterning the lymphatic vasculature.

Recent work investigating the effects of RA on mouse embryoidbodies and cultured human LECs in vitro has shown that RAinduces the expression of lymphatic endothelial cell markers inembryoid body vascular differentiation assays and promotes theproliferation and migration of hLECs (Choi et al., 2012; Marinoet al., 2011). In addition, the exposure to high RA levels duringdevelopment has been shown to promote lymphangiogenesis inmouse and Xenopus embryos (Marino et al., 2011) and in vivo in amouse model of secondary lymphoedema (Choi et al., 2012). Thesedata suggest that RA signalling is likely to play an important role indevelopmental lymphangiogenesis in vivo, though the dependenceon this signalling axis and the genetic mechanisms involvedremain to be investigated.

In the present study, we reveal a novel function for Cyp26b1 inmodulating initial morphogenetic events important for lymphaticvascular development in the mouse embryo. Detailed character-isation of Cyp26b1 expression combined with analysis of the RAreporter transgenic mouse model RARE-LacZ, revealed that LECprogenitors in the cardinal veins are exposed to relatively lowlevels of RA during embryogenesis, due to the actions of CYP26B1.Loss of function of Cyp26b1, resulting in higher levels of RA,resulted in embryonic phenotypes characterised by generalisededema and blood-filled lymphatic vessels. Excessive RA signallingresulted in an increased population of lymphatic endothelialprogenitor cells specified in the embryonic cardinal veins and,concomitantly, enlarged lymph sacs and a hyperplastic lymphaticvascular plexus. Conversely, Cyp26b1 over-expression in a trans-genic mouse model resulted in a reduced pool of lymphaticendothelial progenitor cells and hypoplastic jugular lymph sacs,revealing that RA levels need to be tightly regulated in vivo todirect normal development of the lymphatic vasculature.

Materials and methods

Mouse strains

Mouse embryos were obtained from inter-crosses of Cyp26b1þ /�

mice (Yashiro et al., 2004) and collected at different time points.Staging was performed by somite count or, for later-stage embryos,days post detection of a vaginal plug. PROX1-EGFP BAC transgenicmice (Tg(Prox1-EGFP)221Gsat/Mmcd) were purchased from theMutant Mouse Regional Resource Centre, The Jackson Laboratory(Choi et al., 2011). To test the effect of overexpressing theRA-degrading enzyme CYP26B1 in embryos, we generated asingle-copy doxycycline-inducible Cyp26b1-overexpressing trans-genic line using a method previously described (Beard et al.,2006). Briefly, a minimal CMV promoter driving the coding region

of mouse Cyp26b1 was introduced downstream of the Col1a1 locusin KH2 ES cells by frt/FLPe recombinase-mediated site-specificintegration. The transgene is inducible by doxycycline due thepresence of the M2rtTA transactivator driven from the Rosa26 locus.Studs (C57Bl/6) harbouring the Cyp26b1 transgene and homozygousfor the Rosa26-M2rtTA locus were mated with wildtype C57Bl/6females and the pregnant females were orally dosed with doxycy-cline (2 mg/ml together with 10 mg/ml of sucrose in water bottles)beginning at 8.5 dpc. Litters were collected at 14.5 dpc and con-tained approximately half wild-type and half transgenic progeny.Cyp26b1 was strongly up-regulated by doxycycline treatment asexpected. Double transgenic mice Cyp261� /�:Prox1-gfp embryoswere obtained by backcrossing the Cyp26b1þ /� mouse into theProx1-gfp background. All procedures involving animals conformedto institutional guidelines (University of Queensland and SA Pathol-ogy/CHN Animal Ethics Committees).

Optical projection tomography

Stained embryos were embedded in warm 1% low-melting-point agarose and left until set, adjusting the orientation asnecessary. Set agarose blocks were glued to aluminium mounts.Specimens were then dehydrated in 50% methanol for 18 h with3 graduated changes of methanol to 100%, and then clearedovernight in benzyl alcohol:benzyl benzoate mixed at a ratio of1:2. Once clear, samples were imaged in a Bioptonics 3001 OPTscanner (Bioptonics, UK). Images were acquired at 0.91 intervalsand reconstructed. Stacks were rendered for presentation usingImaris software.

Immunofluorescence

Embryos were dissected and fixed overnight in 4% PFA andcryosections were cut as previously described (Wilhelm et al.,2005). Primary antibodies in blocking solution (100 mM Maleicacid, 10% horse serum, 1% DMSO in PBS-Triton-X 0.1%) were addedand incubated for 24 h at 4 1C. Samples were washed for 6 h inwashing solution (100 mM Maleic acid, 1% DMSO in PBS-Triton-X0.1%) and incubated for 16 h with secondary antibodies in blockingsolution. Whole mount immunostaining of 10.0–11.5 dpc embryosfor confocal microscopy analysis was performed as follows: usingsharpened tungsten needles, embryos were bisected along thelongitudinal axis through the midline. Samples were washed twicein PBS/Triton-X-100 0.3%, incubated for at least 2 h in blockingsolution (PBS/Triton X-100 0.3%/1% bovine serum albumin) andthen overnight at 4 1C in primary antibody diluted in blockingsolution. Samples were washed five times for 1 h each in PBS/Triton X-100 0.3% at room temperature and then incubated over-night at 4 1C with secondary antibodies in blocking solution.Samples were washed for 5 h in PBS/Triton X-100 0.3% at roomtemperature, then overnight at 4 1C and fixed for 30 min (4%formaldehyde in PBS). For visualization, embryos were dehydratedin a graded series of methanol and cleared in benzyl alcohol:benzyl benzoate overnight. Images were captured using a ZeissLSM 700 confocal microscope (Zeiss Laboratories). Images werecompiled using ZEN (blue edition) version 1.0 (Carl Zeiss) andAdobe Photoshop CS5 version 12.0 (Adobe) software.

Antibodies

Antibodies were used in the following dilutions: rabbit poly-clonal anti-mouse PROX1 (Angiobio), 1:1000; chicken polyclonalanti-GFP (Sapphire Bioscience); 1:250, rabbit polyclonal anti-mouse LYVE-1 (Angiobio), rat anti-mouse CD31 (BD Pharmingen)1:500; rat monoclonal anti-PECAM antibody (BD Pharmingen),1:200; hamster polyclonal anti-Podoplanin (Fitzgerald Industries,

J. Bowles et al. / Developmental Biology 386 (2014) 25–3326

Concord. MA), 1:500; hamster polyclonal anti-ICAM antibody(CD54) (Millipore, Australia) 1:200; goat polyclonal anti-Neuropi-lin-2, goat polyclonal anti-mouse VE-Cadherin (R&D Systems)1:500; rat monoclonal anti-endomucin (Santa Cruz BiotechnologyInc) 1:500; rabbit polyclonal anti-CYP26B1 (Protein Tech™, USA)1:100; rat monoclonal anti-mouse Foxc2 1:1000 (Furumoto et al.,1999). Secondary antibodies anti-rat IgGAlexa 555, anti-rat IgGA-lexa 488, anti-rabbit IgGAlexa 488, anti-hamster IgGAlexa 488,anti-rat IgGAlexa 647, anti-mouse IgGAlexa 594, anti-goat Alexa647, anti-goat Alexa555, anti-rabbit IgGAlexa 555 and anti-rabbitIgGAlexa 647 (Molecular Probes) were used at 1:200 or 1:500 forwhole mount immunostaining.

Histology

Embryos for histological examination were fixed in 4% paraf-ormaldehyde at 4 1C overnight and embedded in paraffin or OCT.Blocks were then sectioned (6–10 mm) and stained with haema-toxylin and eosin.

Proliferation assay

Proliferation assays were performed using the CellTiter 96s

AQueous One Solution Cell Proliferation Assay reagent (Promega),as per manufacturer0s instructions. All experiments were per-formed in 96 well plates with 4–6 replicates. Briefly, mouseembryonic LEC were freshly isolated as previously described(Kazenwadel et al., 2012), diluted to 1�105 cells ml�1 in EGM-2 MV (Lonza) and 0.1 ml of cells was added to each well. Plateswere cultured at 37 1C/5% CO2 for 16–18 h. Cells were thenincubated in EGM-2MVþ0.1% DMSO v/v or EGM-2MV containingRA (10 μM) (Sigma) and 0.1% DMSO v/v for 48 h. Cell proliferationwas measured following the addition of CellTiter 96s AQueousOne Solution Cell Proliferation Assay reagent (0.02 ml) to each welland incubation for a further 4 h. Absorbance was measured at490 nm on a FLUOstar OPTIMA microplate reader (BMG LABTECH).

Quantitative RT-PCR (qRT-PCR) analysis

Duplicate assays were carried out on an ABI Prism 7500Sequence Detector System and the mean relative level of expres-sion and associated SEM were calculated. Taqman gene expressionsets were as follows: Mm00446973_m1 (Tbp, endogenous control,encoding TATA box binding protein), Mm00558507_m1 (Cyp26b1)and Mm01319677_m1 (RAR-AR13).

Results

The venous endothelium and lymphatic endothelial progenitor cellsexpress different RA metabolic enzymes

Two recent studies (Choi et al., 2012; Marino et al., 2011)suggested that all-trans and 9-cis RA could influence the earlystages of lymphatic vascular development in both mouse andXenopus embryos, prompting us to investigate in detail theexpression pattern and function of key enzymes that regulatethe synthesis and degradation of RA with respect to the earlystages of lymphangiogenesis. Using in situ hybridization andimmunofluorescence to investigate the expression of Cyp26b1(RA catabolic enzyme) on transverse sections of 11.5 dpc wild typeand lymphatic reporter Prox1-gfp embryos, we found that theCyp26b1 gene was expressed in a polarised fashion on the dorsalside of the cardinal veins where PROX1-positive cells are specified(Fig. 1A, black arrows and 1B, white arrows) suggesting that RA isbeing degraded specifically in that region. This was further

validated on coronal sections; PROX1-positive LEC precursors inthe venous wall were also positive for CYP26B1 (Fig. 2A).

To further characterise the potential levels of RA in the cardinalvein region of the embryo we also explored the expression patternof Aldh1a2 (encoding a major RA anabolic enzyme), by in situhybridisation. In contrast to Cyp26b1, Aldh1a2was observed on theventral aspect of the cardinal veins, opposite the location of LECprogenitors (Wigle and Oliver, 1999) (Fig. 1C, red arrows). We alsotook advantage of a transgenic reporter mouse which harbours aretinoic acid responsive element driving the beta-galactosidasegene (RARE-LacZ), enabling RA responsive tissues to be visualised(Rossant et al., 1991). In accord with the expression of RA synthesisand catabolism enzymes, beta-galactosidase activity was detectedin the Aldh1a2-positive region suggesting that RA signalling isactive in the ventro-medial side of the cardinal vein (Fig. 1D, blackarrows), but not in the region where PROX1-positive cells arise(Fig. 1D asterisks). Further, immunostaining for PROX1, ENDOMU-CIN and beta-galactosidase on 11.5 dpc RARE-LacZ transgenicembryos demonstrated that LEC progenitor cells are not respon-sive to RA signalling (Supplemental Fig. 1).

Finally, characterisation using double in situ and immunofluor-escence analysis validated the expression of Cyp26b1 (black) in thevicinity of PROX1- positive (red) lymphatic endothelial cell pro-genitors in coronal sections of 11.5 dpc wild type embryos (Fig. 2B,white arrows). Immunohistochemistry for PROX1 (Fig. 2C, black)or LYVE1 (Fig. 2D, brown) on coronal sections of 11.5 dpc RARE-LacZ embryos further confirmed the absence of RA signalling(blue, framed box) in the vicinity of PROX1-positive lymphaticendothelial progenitor cells (bracket).

Taken together, these observations suggest that during devel-opment in vivo, the regions of the cardinal veins in which PROX1expression is induced are exposed to low levels of RA signalling,whereas the non-lymphangiogenic parts of the venous endothe-lium are exposed to higher RA concentrations.

Proper regionalisation of PROX1 positive lymphatic endothelialprogenitor cells in the cardinal veins depends on CYP26B1 activity

Our observation that the dorso-lateral and ventral aspects ofthe cardinal veins are exposed to low and high RA concentrationsrespectively, prompted us to further explore the role of thissignalling pathway in regulating the size of the PROX1-positivelymphatic endothelial progenitor cell pool specified in the veins.Here we used the Cyp26b1 knock-out mouse model, consideredthe gold standard to study interference of RA metabolism duringdevelopment (Bowles et al., 2006; Yashiro et al., 2004).

We first investigated the distribution of lymphatic endothelialprogenitor cell markers PROX1 and LYVE1 in the walls of thecardinal veins in the presence of excessive RA levels in Cyp26b1�/�

mouse embryos (Yashiro et al., 2004). Coronal sections of 11.5 dpcembryos stained with PROX1 (red) or LYVE1 (red) showed a clearorganisation of PROX1-positive lymphatic endothelial progenitorcells along the antero-posterior axis of the vein in wild-typeembryos (Fig. 3A and D, brackets indicate LEC progenitors). Bycontrast, in Cyp26b1� /� animals, the distribution of PROX1þ andLYVE1þ LEC progenitor cells was expanded to the ventral andmedial aspects of the cardinal veins (Fig. 3B and E, white arrows).These results suggest that elevated levels of RA, prior to 11.5 dpc,disrupt the regionalisation of LEC progenitors.

To further investigate lymphatic endothelial progenitor cell spe-cification in Cyp26b1� /� embryos, we backcrossed the Prox1-gfpreporter mouse (Choi et al., 2011) onto the Cyp26� /� background,generating double transgenic mice in which we could visualize GFPþ

LECs in a setting of excessive RA levels. Characterisation of GFPþ LECprogenitor distribution in this setting (Fig. 3C, white arrows) recapi-tulated the abnormal distribution of endogenous PROX1 in the

J. Bowles et al. / Developmental Biology 386 (2014) 25–33 27

Cyp26b1� /� background (Fig. 3B) and confirmed the numerical andspatial expansion of the PROX1-positive LEC progenitor pool inCyp26b1� /� mice. PROX1 expression was also assessed on transversesections of Cyp26b1� /� embryos, revealing that the distribution of

LEC progenitors in the cardinal veins was expanded on the ventro-medial side. Immunofluorescence for PROX1 (red) and VEGFR3 (blue)(Supplemental Fig. 2A and B,) or LYVE1 (red), podoplanin (green) andthe pan-endothelial cell marker PECAM (blue) (Supplemental. Fig. 2C

Fig. 2. LEC progenitors in the veins are exposed to low RA levels. (A) Immunofluorescence for CYP26B1 on coronal sections of 11.5 dpc Prox1-gfp transgenic embryosrevealed the expression of CYP26B1 enzyme (blue) in GFP-positive LEC precursors in the cardinal vein and developing lymph sacs. (B). Double in situ hybridisation andimmunofluorescence for Cyp26b1 and Prox1 revealed the likely presence of the RA-degrading enzyme near LEC progenitors. (C and D) Beta-galactosidase staining of coronalsection of RARE-LacZ reporter embryo at 11.5 dpc revealed that lymphangiogenic territories in the cardinal vein (Prox1þ and LYVEþ) are devoid of RA signal (bracket, LECprog.). By contrast the medial side of the vein close to the dorsal aorta shows active RA signalling (framed box). DA, dorsal aorta; CV, cardinal vein; LS, lymph sac. Scale barsA–C 50 μm.

Fig. 1. Expression of genes encoding an RA-degrading enzyme, Cyp26b1, and an RA-synthesising enzyme, Aldh1a2, is polarized in the cardinal vein. (A) In situhybridisation on transverse sections of 11.5 dpc embryo revealed the expression of Cyp26b1 mRNA on the dorso-lateral side of the cardinal vein (arrows).(B) Immunofluorescence for CYP26B1 on transverse sections of 11.5 dpc Prox1-gfp transgenic embryos demonstrates co-expression of CYP26B1 (arrow) in GFP-positiveLEC precursors. (C) Aldh1a2 is expressed in a polarised manner in the ventral side of the cardinal vein opposed to lymphangiogenic territories (red arrows). (D) RARE-LacZreporter staining further confirmed the lack of RA signal in the lateral side of the cardinal vein (asterisks) and the presence of active RA signalling on the ventro-medial side(black arrows). DA, dorsal aorta; CV, cardinal vein; LS, lymph sac; H, heart. Scale bars, A, B and C 100 μm and D 150 μm.

J. Bowles et al. / Developmental Biology 386 (2014) 25–3328

and D, asterisks) validated the ectopic distribution of LEC precursorsthroughout both dorso-lateral and ventro-medial aspects of thecardinal veins in Cyp26b1� /� embryos. Conversely, the size of thedeveloping cardinal veins did not appear to be dramatically affectedby the change of RA concentration in vivo: the overall morphologyand size of the veins was unaffected in Cyp26b1� /� embryos(Supplemental Fig. 3).

In order to determine whether this change in regionalisation ofPROX1þ LEC in Cyp26b1� /� embryos is a result of RA acting topromote LEC progenitor cell specification or proliferation, weinvestigated the number of PROX1-positive precursor cells inembryonic cardinal veins at 11.5 dpc, prior to any edema. Analysisrevealed that Cyp26b1-null embryos displayed an increase inthe pool of venous LEC precursors (Supplemental Fig. 2E and F).Further, we examined the effects of RA on primary lymphaticendothelial cells isolated from embryonic mouse skin. Exposureof cells to 10 μM RA significantly promoted LEC proliferation(Supplemental Fig. 2G).

No changes in levels of Prox1 mRNA were observed in eitherprimary embryonic LECs or blood endothelial cells (BECs) treatedwith RA (data not shown), suggesting that the effects of excess RA

on lymphatic vascular morphogenesis are likely due to thepromotion of LEC proliferation, shortly after their specification.

Loss of Cyp26b1 function promotes the development of hyperplastic,blood filled deep and superficial lymphatics due to defective lympho-venous valve patterning

To assess whether RA acts as a molecular cue that influencesmorphogenesis of the primitive lymphatic vascular network, weinvestigated lymphatic endothelial cell sprouting from the cardinalveins, together with lymph sac formation, in Cyp26b1� /� mice bywhole mount immunostaining (11.5 dpc and 13.5 dpc, Figs. 3and 4) and staining of transverse sections through the jugularlymph sacs (14.5 dpc) (Fig. 4). At 11.5 dpc, Cyp26b1�/� embryoshad enlarged lymph sacs compared to wild type embryos (Fig. 3F–K, dotted lines). Double transgenic Prox1-gfp::Cyp26b1 mice werealso used to analyze lymph sac development in vivo. Opticalprojection tomography (OPT) and 3D rendering analyses revealedthat lymph sac volumes were enlarged in Cyp26b1 mutantembryos compared to control littermates (Fig. 4A and B, bluesignal and Supplemental movies 1 and 2).

Fig. 3. Loss of Cyp26b1 function triggers disruption of the polarised distribution of LECs in the cardinal vein. (A and B) Immunofluorescence on coronal sections of11.5 dpc embryos revealed that the domain of PROX1-positive, Podoplanin-negative LEC progenitor cells in the cardinal veins is expanded in Cyp26b1 mutants (whitearrows). (C) Double transgenic animals Prox1-gfp::Cyp26b1 also displayed a lack of LEC progenitor organisation as observed on coronal sections of 11.5 dpc embryos. (D andE) Immunofluorescence for lymphatic specific marker LYVE1 and PODOPLANIN on coronal section further validated the aberrant distribution of LEC progenitors in the wallsof the cardinal veins. (F–K) Whole mount immunofluorescence on 11.5 dpc bisected embryos using lymphatic-specific markers LYVE1 (blue) and PROX1 (red) revealedenlarged lymph sacs in Cyp26b1�/� embryos (I–K) compared to wild type embryos (F–H). PDP, PODOPLANIN; CV, cardinal vein; LS, lymph sac; LEC prog., lymphaticendothelial cell progenitor; DA, dorsal aorta. Scale bars, A–E 100 μm; F–K 250 μm.

J. Bowles et al. / Developmental Biology 386 (2014) 25–33 29

Supplementary material related to this article can be foundonline at http://dx.doi.org/10.1016/j.ydbio.2013.12.008.

At 13.5 dpc, Cyp26b1�/� embryos appeared haemorrhagic: anaccumulation of blood was apparent in the region of the devel-oping jugular lymph sacs (Fig. 4C and D white arrow). At 14.5 dpc

and 16.5 dpc, Cyp26b1�/� embryos exhibited pronounced subcu-taneous edema and their superficial lymphatics appeared to befilled with blood (Fig. 4E and F and Supplemental Fig. 2). Thisobservation was confirmed at 14.5 dpc by histological analysis oftransverse sections through the jugular region of wild type andmutant embryos (Fig. 4G and H). Molecular characterisation usinglymphatic specific markers PROX1 (data not shown) or LYVE1(Fig. 4I and J) with Podoplanin and the pan-endothelial cell markerPECAM, revealed that significant enlargement of the developinglymph sacs persisted in Cyp26b1� /� embryos at this develop-mental stage. Quantitation of lymph sac diameter in wild type andmutant embryos confirmed that the jugular lymph sacs wereenlarged as a result of Cyp26b1 loss of function (Fig. 4K).

Analysis of the skin of Prox1-gfp::Cyp26b1� /� embryos at14.5 dpc revealed striking differences in the morphology of thedermal lymphatic vasculature compared to that of their wild typelittermates (Fig. 5A and B). Lymphatic vessels in mutant embryoswere increased in diameter and were substantially less branched.By contrast, the blood vasculature, stained with antibodies toendomucin (Fig. 5A and B, red) did not display any patterningdefects, suggesting that RA acts selectively on the lymphatic vesselnetwork at the developmental stages characterised.

With a view to identifying the cause of blood-filled lymphatics,we analysed development of the lymphovenous valves that areresponsible for proper separation of the blood and lymphaticvascular networks. During development, these valves separatethe jugular and subclavian veins from the jugular lymph sacsand are characterised by high levels of PROX1 and FOXC2 expres-sion (Srinivasan and Oliver, 2011) (Fig. 5 C–F, box). In Cyp26b1� /�

embryos, the lymphovenous valves were expanded in size andexhibited elevated FOXC2 levels (Fig. 5G J, arrows). These defectslikely underlie the phenotype of blood-filled lymphatics inCyp26b1�/� embryos.

Gain of Cyp26b1 function disrupts proper patterningof lymphatic vessels

Having shown that excess RA is detrimental to the orderlydistribution and number of lymphatic endothelial progenitors andto the patterning of lymphatic vessels, the next step was todetermine the effect of reduced RA dosage in vivo. To that end,we generated a transgenic mouse that can be induced to over-express Cyp26b1 under the control of a CMV promoter. To confirmelevated Cyp26b1 expression, we performed qRT-PCR analysis onembryonic tissue dissected from the jugular area where lymphsacs are known to develop (Fig. 6A) and analysed expression levelsof RAR-beta, as an indirect readout for RA concentration (Fig. 6B):

Fig. 4. Loss of Cyp26b1 function induces a severe lymphatic vascular pheno-type. (A and B) OPT processing and 3-D rendering of bisected Prox1-gfp embryosafter whole mount immunofluorescence for GFP (blue) and Endomucin (white)showed that lymph sacs of Cyp26b1� /� embryos are enlarged compared to controllittermates. The first six inter-somitic vessels are indicated in red, the bluerendering is outlining the total surface area of lymph sacs in wild type and mutantembryos. (C and D) Gross morphology observation revealed edema and bloodvascular leakage into the jugular lymph sacs of Cyp26b1� /� embryos at 13.5 dpc(white arrowhead). (E and F) At 14.5 dpc, developing lymphatics are blood-filledand histology analysis (G and H) confirmed the presence of red blood cells in thelymph sacs. (I-J) Immunofluorescence analysis for lymphatic specific markersLYVE1 (red) and PODOPLANIN (green) and the pan-endothelial cell marker PECAMfurther validated the presence of enlarged lymph sacs in the Cyp26b1�/� embryosat 14.5 dpc. In some cases cells single positive for PODOPLANIN were observed onthe medial side of the vein, this cell population corresponds to undifferentiatedneurons in the neural tube. (K) Quantitation of lymph sac diameter comparing wildtype control and Cyp26b1� /� embryos at 13.5 dpc. Mutant embryos showed asignificant enlargement of jugular lymph sacs. CA, carotid artery; JV, jugular vein;LS, lymph sac. Scale bars A and B 250 μm, C–F 500 μm and G–J 200 μm. Statisticalanalysis no0.05 (p¼0.0285), non-paired t-test, (n¼4 for each genotype, 3 serialsections per animal).

J. Bowles et al. / Developmental Biology 386 (2014) 25–3330

as expected, this established RA target was down-regulated inCyp26b1-transgenic embryos. Following Cyp26b1 over-expressionat 8.5 dpc, transgenic embryos exhibited substantially smallerjugular lymph sacs than their wild type counterparts at 14.5 dpc(Fig.6C–F), suggesting that too little RA is also deleterious forlymphatic vascular development. Taken together, these resultsshow that a finely-tuned dosage of RA is critical for propermorphogenesis of the initial lymphatic vascular network in vivo.

Discussion

The molecular basis for the earliest step in lymphatic develop-ment, the polarised distribution of lymphatic endothelial progeni-tor cells in the embryonic cardinal veins, has remained a mystery.Here we present evidence that the well-studied developmentalmorphogen, RA, is asymmetrically available to endothelial cells ofthe cardinal veins, with high levels present on the ventral side andlower levels on the dorsal side, where LEC progenitors arespecified. The asymmetry is set up by the polarised expressionaround the vein of genes encoding an RA-synthesising enzyme,ALDH1a2, and an RA-degrading enzyme, CYP26B1. By examiningCyp26b1� /� and Cyp26b1-overexpressing embryos, we reveal thattightly regulated RA levels are essential for normal lymphaticdevelopment.

Key findings of this study are: (1) By 11.5 dpc, the genes thatencode enzymes involved in RA catabolism and anabolism areexpressed in defined domains of the embryonic cardinal veins;(2) In the setting of elevated RA levels (Cyp26b1�/�), the size ofthe PROX1-positive lymphatic endothelial progenitor cell pool inthe embryonic cardinal veins is increased as early as 11.5 dpc;(3) Hyperplastic, blood filled lymph sacs and lymphatic vessels area feature of Cyp26b1�/� embryos and conversely, lymph sacs arehypoplastic in embryos that express high levels of Cyp26b1, inwhich RA levels are diminished; (4) Elevated RA levels triggerlymphovenous valve malformations and (5) Hyperplasia of lymphsacs and lymphatic vessels in Cyp26b1� /� embryos occurs, at leastin part, due to the promotion of LEC proliferation by RA. This work

has identified a novel molecular cue that is essential to the propermorphogenesis of the lymphatic vascular tree.

Lymphangiogenesis and RA metabolism

A role for RA in lymphangiogenesis has recently been suggestedby two studies showing that exposure of developing embryos tosupra-physiological doses of RA increased the number of PROX1-positive LEC precursor cells in mouse and Xenopus embryos(Marino et al., 2011) and that RA promoted lymphatic vascularregeneration in a mouse model of lymphedema (Choi et al., 2012).Our findings that disruption of the gene encoding the CYP26B1enzyme, resulting in elevated RA levels, leads to an increase in sizeof the LEC progenitor pool, as well as in hyperplasia of lymph sacsand dermal lymphatic vessels in Cyp26b1�/� mice, is in agreementwith these studies. Our work also suggests that the phenotype ofblood filled jugular lymph sacs and dermal lymphatic vesselscharacteristic of Cyp26b1�/� embryos is due to the defectivedevelopment of lymphovenous valves (Srinivasan and Oliver,2011). Whether this is the result of a role for RA in valvespecification, or due to elevated numbers of PROX1-positive cellsin the embryonic veins, remains to be investigated.

The molecular role of RA signalling

Recent work has shown that the nuclear orphan receptorCOUP-TFII is able to directly drive Prox1 transcription to inducelymphatic endothelial cell differentiation (Srinivasan et al., 2010).Although an endogenous ligand for COUP-TFII remains to bedetermined, a previous study has identified COUP-TFII as a low-affinity RA-activated receptor (Kruse et al., 2008). Taken together,these observations raise the possibility that RA might modifyCOUP-TFII activity, either directly or indirectly, via interaction withnuclear RA receptors, which might in turn modulate the level ofProx1 transcription and the number of LEC progenitor cells that arespecified. In keeping with the idea of direct action of RA onendothelial cells of the cardinal vein, others have recently demon-strated that BECs of the cardinal vein express the RAR-α receptor(Marino et al., 2011). This concept is in line with previous findings

Fig. 5. Loss of Cyp26b1 function in vivo and in vitro induces impaired remodelling of superficial lymphatics and hyper-proliferation of LECs. (A and B) Whole mountimmunofluorescence of 14.5 dpc embryonic skin; utilisation of the pan-endothelial cell marker ENDOMUCIN (red) and the lymphatic specific marker neuropilin-2 (NRP2,green) and GFP (blue) revealed that lymphatic vessels of the Cyp26b1� /� embryos fail to assemble properly. By contrast, the blood vasculature remained unaffected. (C–J)Immunofluorescent immunostaining of lymphovenous valve markers PROX1 (red) and FOXC2 (green) with the pan-endothelial cell marker VE-cadherin (blue) on coronalsections of 13.5 dpc wild type and Cyp26b1� /� embryos demonstrated larger valve leaflets in the mutant mice (arrows). CV, cardinal vein, JLS, jugular lymph sac. Scale bars Aand B 200 μm, C–J 100 μm.

J. Bowles et al. / Developmental Biology 386 (2014) 25–33 31

from Lai and colleagues (Lai et al., 2003) showing that RA regulatesendothelial cell proliferation during vasculogenesis.

Genetic pathways that instruct morphogenesis of the lymphaticvascular tree remain mostly unexposed. To date only a handfulof growth molecules such as vascular endothelial growth factor-c(VEGF-C), angiopoeitin-1 or adrenomedullin have been shown asessential to pattern the expanding lymphatic vasculature (Fritz-Sixet al., 2008; Karkkainen et al., 2004; Tammela et al., 2005). Thediscovery of a signalling pathway that provides an asymmetricmorphogenetic cue to endothelial cells of the cardinal vein is there-fore significant and should help us to eventually understand howlymphatic vessels initially assemble. Our work demonstrates thatphysiologically-available RA must be present within a critical range in

order for lymphangiogenesis to begin normally and contributes newgenetic models that can be used to complement existing genetic toolsused in studying the birth of the lymphatic vasculature.

Funding

Australian Research Council; Cancer Council Queensland;National Health and Medical Research Council (NHMRC) of Aus-tralia. NH is a National Heart Foundation Career DevelopmentFellow. PK is a University of Queensland Vice-Chancellor0s SeniorResearch Fellow. MF is a Career Developmental Fellow ofthe NHMRC.

Fig. 6. Gain of Cyp26b1 function results in atrophic jugular lymph sacs. Immunofluorescence for the lymphatic markers PROX1 or LYVE1 (red), podoplanin (green) andthe pan-endothelial cell marker PECAM on transverse sections of 14.5 dpc wild type and Cyp26b1 over-expressing (CypOE) embryos. Embryos over-expressing Cyp26b1enzyme exhibit smaller lymph sacs. (E and F) Quantitative RT-PCR analysis of Cyp26b1 and RARβ (a RA target gene) in jugular tissue dissected out from wild type and CypOEembryos showed that transgenic animals over-express Cyp26b1 mRNA by 16-fold and that this is associated with a 3-fold lower expression of RARA-Rd. Statistical analysis:unpaired Student0s t-test, for CypOE n¼2 and for wild type n¼4, bars indicate the meanþSEM, nnpo0.01, nnnnpo0.0001. Scale bars 200 μm. JV, jugular vein; LS, lymph sac.

J. Bowles et al. / Developmental Biology 386 (2014) 25–3332

Acknowledgements

We thank Prof Hiroshi Hamada for kindly providing theCyp26b1� /� mouse model and Kelly Betterman for technicalassistance. Confocal microscopy was performed at the AustralianCancer Research Foundation Dynamic Imaging Centre for CancerBiology and at the Detmold Imaging Facility.

Appendix A. Supplementary materials

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.ydbio.2013.12.008.

References

Alitalo, K., Tammela, T., Petrova, T.V., 2005. Lymphangiogenesis in development andhuman disease. Nature 438, 946–953.

Beard, C., Hochedlinger, K., Plath, K., Wutz, A., Jaenisch, R., 2006. Efficient method togenerate single-copy transgenic mice by site-specific integration in embryonicstem cells. Genesis 44, 23–28.

Bowles, J., Knight, D., Smith, C., Wilhelm, D., Richman, J., Mamiya, S., Yashiro, K.,Chawengsaksophak, K., Wilson, M.J., Rossant, J., Hamada, H., Koopman, P., 2006.Retinoid signaling determines germ cell fate in mice. Science 312, 596–600.

Choi, I., Chung, H.K., Ramu, S., Lee, H.N., Kim, K.E., Lee, S., Yoo, J., Choi, D., Lee, Y.S.,Aguilar, B., Hong, Y.K., 2011. Visualization of lymphatic vessels by Prox1-promoter directed GFP reporter in a bacterial artificial chromosome-basedtransgenic mouse. Blood 117, 362–365.

Choi, I., Lee, S., Kyoung Chung, H., Suk Lee, Y., Eui Kim, K., Choi, D., Park, E.K.,Yang, D., Ecoiffier, T., Monahan, J., Chen, W., Aguilar, B., Lee, H.N., Yoo, J., Koh, C.J., Chen, L., Wong, A.K., Hong, Y.K., 2012. 9-cis retinoic acid promoteslymphangiogenesis and enhances lymphatic vessel regeneration: therapeuticimplications of 9-cis retinoic acid for secondary lymphedema. Circulation 125,872–882.

Francois, M., Caprini, A., Hosking, B., Orsenigo, F., Wilhelm, D., Browne, C.,Paavonen, K., Karnezis, T., Shayan, R., Downes, M., Davidson, T., Tutt, D., Cheah,K.S., Stacker, S.A., Muscat, G.E., Achen, M.G., Dejana, E., Koopman, P., 2008.Sox18 induces development of the lymphatic vasculature in mice. Nature 456,643–647.

Francois, M., Harvey, N.L., Hogan, B.M., 2011. The transcriptional control oflymphatic vascular development. Physiology (Bethesda) 26, 146–155.

Francois, M., Short, K., Secker, G.A., Combes, A., Schwarz, Q., Davidson, T.L., Smyth, I.,Hong, Y.K., Harvey, N.L., Koopman, P., 2012. Segmental territories along thecardinal veins generate lymph sacs via a ballooning mechanism duringembryonic lymphangiogenesis in mice. Dev Biol. 364, 89–98.

Fritz-Six, K.L., Dunworth, W.P., Li, M., Caron, K.M., 2008. Adrenomedullin signalingis necessary for murine lymphatic vascular development. J. Clin. Invest. 118,40–50.

Furumoto, T.A., Miura, N., Akasaka, T., Mizutani-Koseki, Y., Sudo, H., Fukuda, K.,Maekawa, M., Yuasa, S., Fu, Y., Moriya, H., Taniguchi, M., Imai, K., Dahl, E.,Balling, R., Pavlova, M., Gossler, A., Koseki, H., 1999. Notochord-dependentexpression of MFH1 and PAX1 cooperates to maintain the proliferation ofsclerotome cells during the vertebral column development. Dev. Biol. 210,15–29.

Hagerling, R., Pollmann, C., Andreas, M., Schmidt, C., Nurmi, H., Adams, R.H., Alitalo, K.,Andresen, V., Schulte-Merker, S., Kiefer, F., 2013. A novel multistep mechanism forinitial lymphangiogenesis in mouse embryos based on ultramicroscopy. EMBO J.32, 629–644.

Hong, Y.K., Harvey, N., Noh, Y.H., Schacht, V., Hirakawa, S., Detmar, M., Oliver, G.,2002. Prox1 is a master control gene in the program specifying lymphaticendothelial cell fate. Dev. Dyn. 225, 351–357.

Karkkainen, M.J., Haiko, P., Sainio, K., Partanen, J., Taipale, J., Petrova, T.V., Jeltsch, M.,Jackson, D.G., Talikka, M., Rauvala, H., Betsholtz, C., Alitalo, K., 2004. Vascularendothelial growth factor C is required for sprouting of the first lymphaticvessels from embryonic veins. Nat. Immunol. 5, 74–80.

Kazenwadel, J., Secker, G.A., Betterman, K.L., Harvey, N.L., 2012. In vitro assays usingprimary embryonic mouse lymphatic endothelial cells uncover key roles forFGFR1 signalling in lymphangiogenesis. PLoS One 7, e40497.

Kruse, S.W., Suino-Powell, K., Zhou, X.E., Kretschman, J.E., Reynolds, R., Vonrhein, C.,Xu, Y., Wang, L., Tsai, S.Y., Tsai, M.J., Xu, H.E., 2008. Identification of COUP-TFIIorphan nuclear receptor as a retinoic acid-activated receptor. PLoS Biol. 6, e227.

Lai, L., Bohnsack, B.L., Niederreither, K., Hirschi, K.K., 2003. Retinoic acid regulatesendothelial cell proliferation during vasculogenesis. Development 130,6465–6474.

Marino, D., Dabouras, V., Brandli, A.W., Detmar, M., 2011. A role for all-trans-retinoic acid in the early steps of lymphatic vasculature development. J. Vasc.Res. 48, 236–251.

Niederreither, K., Dolle, P., 2008. Retinoic acid in development: towards anintegrated view. Nat. Rev. Genet. 9, 541–553.

Oliver, G., Srinivasan, R.S., 2010. Endothelial cell plasticity: how to become andremain a lymphatic endothelial cell. Development 137, 363–372.

Petrova, T.V., Makinen, T., Makela, T.P., Saarela, J., Virtanen, I., Ferrell, R.E., Finegold,D.N., Kerjaschki, D., Yla-Herttuala, S., Alitalo, K., 2002. Lymphatic endothelialreprogramming of vascular endothelial cells by the Prox-1 homeobox tran-scription factor. EMBO J. 21, 4593–4599.

Rossant, J., Zirngibl, R., Cado, D., Shago, M., Giguere, V., 1991. Expression of aretinoic acid response element-hsplacZ transgene defines specific domains oftranscriptional activity during mouse embryogenesis. Genes Dev. 5, 1333–1344.

Srinivasan, R.S., Geng, X., Yang, Y., Wang, Y., Mukatira, S., Studer, M., Porto, M.P.,Lagutin, O., Oliver, G., 2010. The nuclear hormone receptor Coup-TFII is requiredfor the initiation and early maintenance of Prox1 expression in lymphaticendothelial cells. Genes Dev. 24, 696–707.

Srinivasan, R.S., Oliver, G., 2011. Prox1 dosage controls the number of lymphaticendothelial cell progenitors and the formation of the lymphovenous valves.Genes Dev. 25, 2187–2197.

Tammela, T., Alitalo, K., 2010. Lymphangiogenesis: molecular mechanisms andfuture promise. Cell 140, 460–476.

Tammela, T., Saaristo, A., Lohela, M., Morisada, T., Tornberg, J., Norrmen, C., Oike, Y.,Pajusola, K., Thurston, G., Suda, T., Yla-Herttuala, S., Alitalo, K., 2005.Angiopoietin-1 promotes lymphatic sprouting and hyperplasia. Blood 105,4642–4648.

van de Pavert, S.A., Olivier, B.J., Goverse, G., Vondenhoff, M.F., Greuter, M., Beke, P.,Kusser, K., Hopken, U.E., Lipp, M., Niederreither, K., Blomhoff, R., Sitnik, K.,Agace, W.W., Randall, T.D., de Jonge, W.J., Mebius, R.E., 2009. ChemokineCXCL13 is essential for lymph node initiation and is induced by retinoic acidand neuronal stimulation. Nat. Immunol. 10, 1193–1199.

Wigle, J.T., Harvey, N., Detmar, M., Lagutina, I., Grosveld, G., Gunn, M.D., Jackson, D.G.,Oliver, G., 2002. An essential role for Prox1 in the induction of the lymphaticendothelial cell phenotype. Embo. J. 21, 1505–1513.

Wigle, J.T., Oliver, G., 1999. Prox1 function is required for the development of themurine lymphatic system. Cell 98, 769–778.

Wilhelm, D., Martinson, F., Bradford, S., Wilson, M.J., Combes, A.N., Beverdam, A.,Bowles, J., Mizusaki, H., Koopman, P., 2005. Sertoli cell differentiation is inducedboth cell-autonomously and through prostaglandin signaling during mamma-lian sex determination. Dev. Biol. 287, 111–124.

Yang, Y., Garcia-Verdugo, J.M., Soriano-Navarro, M., Srinivasan, R.S., Scallan, J.P.,Singh, M.K., Epstein, J.A., Oliver, G., 2012. Lymphatic endothelial progenitorsbud from the cardinal vein and intersomitic vessels in mammalian embryos.Blood 120, 2340–2348.

Yashiro, K., Zhao, X., Uehara, M., Yamashita, K., Nishijima, M., Nishino, J., Saijoh, Y.,Sakai, Y., Hamada, H., 2004. Regulation of retinoic acid distribution is requiredfor proximodistal patterning and outgrowth of the developing mouse limb.Dev. Cell 6, 411–422.

J. Bowles et al. / Developmental Biology 386 (2014) 25–33 33