PREGNANCY

ELABELA deficiency promotespreeclampsia and cardiovascularmalformations in miceLena Ho,1* Marie van Dijk,2† Sam Tan Jian Chye,1† Daniel M. Messerschmidt,3

Serene C. Chng,1 Sheena Ong,1 Ling Ka Yi,3 Souad Boussata,2 Grace Hui-Yi Goh,1

Gijs B. Afink,2 Chin Yan Lim,1 N. Ray Dunn,1 Davor Solter,1

Barbara B. Knowles,1 Bruno Reversade1,2,3,4,5*

Preeclampsia (PE) is a gestational hypertensive syndrome affecting between 5 and 8% of allpregnancies. Although PE is the leading cause of fetal and maternal morbidity and mortality,its molecular etiology is still unclear. Here, we show that ELABELA (ELA), an endogenousligand of the apelin receptor (APLNR, or APJ), is a circulating hormone secreted by theplacenta. Elabela but not Apelin knockout pregnant mice exhibit PE-like symptoms, includingproteinuria and elevated blood pressure due to defective placental angiogenesis. In mice,infusion of exogenous ELA normalizes hypertension, proteinuria, and birth weight. ELA,which is abundant in human placentas, increases the invasiveness of trophoblast-like cells,suggesting that it enhances placental development to prevent PE.The ELA-APLNR signaling axismay offer a new paradigm for the treatment of common pregnancy-related complications,including PE.

The placenta is a mammalian-specific organand a critical source of factors responsiblefor remodeling the maternal cardiovascularsystem to accommodate the needs of thegrowing fetus. Defects in placentation often

result in intrauterine growth restriction (IUGR)for the fetus and gestational complications suchas preeclampsia (PE) for the mother. PE affects5 to 8% of all pregnancies and remains theleading cause of fetal and maternal morbidityand mortality. Current challenges in PE includeearly detection and the availability of effectivedrugs that do not adversely affect fetal develop-ment. ELABELA encodes an endogenous ligandfor the apelin receptor (APLNR, or APJ). It is firstdetected in preimplantation human blastocystsand controls the self-renewal of embryonic stemcells (1). In the adult, its expression is restrictedto a few tissues, including two endocrine organs,the kidneys and the placenta (1). In rodents, theonset of Ela expression coincides with zygotictranscription (fig. S1A), peaks at the blastocyststage, and is similarly restricted in the adult. Inlower vertebrates, Ela is required for proper en-doderm development, and Ela-deficient zebrafishhave profound cardiac malformations resulting

from impaired migration of cardiac progenitors(2, 3). Zebrafish lacking both Ela and Apelin(Apln), the alternate ligand for Aplnr, have defectsin vasculogenesis owing to impaired migrationof angioblasts to the midline (4). At present, themolecular effects of ELA signaling downstreamof APLNR are unknown, and its involvement inmammalian development and physiology hasnot been addressed.To delineate the contribution of ELA to mam-

malian development, we generated Ela knockout(ElaD/D) mice using homologous recombinationto delete exon 3 encoding the mature ELA pep-tide (Fig. 1, A and B, and fig. S1, B and C). Thisstrategy did not result in nonsense-mediateddecay of the ElaD mRNA (Fig. 1C) and presum-ably preserves the potential noncoding func-tions of the Ela transcript (5). Only half of theexpected ElaD/D mice from heterozygous inter-crosses were obtained at weaning (Fig. 1D) (P <0.001, chi-square test with df = 1). Notably, thisreduced recessive Mendelian inheritance was evenmore pronounced for ElaD/D embryos carried byElaD/D mothers (67%) than by ElaD/+ mothers(51%) (Fig. 1D). This apparent maternal contri-bution is not due to Ela mRNA being depositedin the oocyte, because the onset of Ela transcrip-tion is strictly zygotic (fig. S1A). Rather, we surmisethat ELA might be provided by the maternalcirculation or uterine environment. At embry-onic day 10.5 (E10.5), ElaD/D embryos can begrouped into three classes: 48.9% were pheno-typically normal (class 1), 8.5% were delayed witha hypovascular yolk sac (class 2), and 42.6% hadavascular yolk sacs and severe embryonic vascu-lar malformations (class 3) (fig. S1D) that are sim-ilar to those previously reported for Apj knockouts(Fig. 1, E to J). In affected ElaD/D embryos, vascu-logenesis appears to be initiated, as evidenced by

the presence of a CD31/Pecam+ endothelial plexus,which subsequently fails to undergo remodelingand angiogenic sprouting to form organized vitel-line vessels, dorsal aorta, outflow tract, and inter-somitic vessels (Fig. 1, K to S, and fig. S1, E to J).The heart tube is poorly looped, with reducedsmooth actin muscle (SMA) staining (Fig. 1, Qto S), and the most severely affected embryos(class 3) have pericardial edema (fig. S1, K and L).These cardiac defects are consistent with the firstpostgastrulation expression of Ela in the primi-tive foregut overlying the developing heart tube(Fig. 1, T and U) (6). Surprisingly, Ela is not de-tected in endothelial precursors of the yolk sac(Fig. 1W), whereas Apj expression is ubiquitousin embryonic, allantoic, and yolk sac mesoderm,which gives rise to endothelial cells (Fig. 1, Vand X). The expression patterns of Ela and Apjsuggest that the observed cardiac defects arepartly due to insufficient blood flow to stimulateangiogenesis. Outside of the developing hearttube, Ela is first detected in the chorionic tropho-blast of the developing placenta (Fig. 1U andfig. S1, M and N) and is robustly up-regulatedafter allantoic fusion (Fig. 2A), becoming re-stricted to syncytiotrophoblasts (STs) at E10.5(Fig. 2, C and C′). Accordingly, ELA protein isdetected by immunohistochemistry in wild-type(WT) STs but not in ElaD/D placentas (Fig. 2,E and F). ELA-positive STs are juxtaposed toApj-expressing fetal endothelial cells (Fig. 2,B, D, and D′). Hence, ELA may signal to APJ-expressing cells in a paracrine manner but mayalso be circulating systemically because the cho-rioallantoic placenta is perfused by maternal andfetal blood. Indeed, endogenous ELA is detectedby enzyme-linked immunosorbent assay (ELISA)in the serum of pregnant females, peaking atmidgestation, but not in nonpregnant mice (Fig.2G). Systemic ELA in a pregnant mother is contri-buted both maternally and embryonically (Fig.2H), the former reflecting secretion from thematernal endometrial stroma and kidneys (fig.S2, A to C) and the latter from embryonicallyderived STs (Fig. 2C). We therefore conclude thatELA is a pregnancy-associated hormone secretedby the developing conceptus and placenta.ElaD/D placentas from affected embryos have

thin labyrinths (Fig. 2, I and J, and fig. S2, Dand E) with poor vascularization (Fig. 2, K andL), increased apoptosis (fig. S2, F and G), andreduced proliferation (fig. S2, H and I). ElaD/D

placentas from unaffected (class 1) or mildlyaffected (class 2) embryos, which are interme-diately vascularized, nonetheless exhibit delayedST differentiation, as indicated by reduced alka-line phosphatase and syncytin-1 staining at E10.5(Fig. 2, M and N, and fig. S2, J and K). Althoughsuch placentas eventually develop, allowing em-bryo survival, the labyrinth of mutant versus WTplacentas remains thinner until the end of ges-tation (fig. S2E).To understand the pathogenesis of Ela defi-

ciency causing placental dysfunction, we isolatedplacentas denuded of maternal decidua fromWT and ElaD/D conceptuses (Fig. 3A). We choseto carry out the analysis by E9.5 to avoid the

RESEARCH

Ho et al., Science 357, 707–713 (2017) 18 August 2017 1 of 7

1Institute of Medical Biology, A*STAR, 8A Biomedical Grove,Immunos, Singapore 138648. 2Reproductive BiologyLaboratory, Obstetrics and Gynaecology, Academic MedicalCenter (AMC), Meibergdreef 9, 1105 AZ Amsterdam-Zuidoost, Netherlands. 3Institute of Molecular and CellBiology, A*STAR, 61 Biopolis Drive, Proteos, Singapore138673. 4National University of Singapore, Department ofPaediatrics, 1E Kent Ridge Road, Singapore 119228. 5MedicalGenetics Department, Koç University School of Medicine,34010 Istanbul, Turkey.*Corresponding author. Email: lena.ho@reversade.com (L.H.);bruno@reversade.com (B.R.) †These authors contributed equallyto this work.

on March 28, 2021

http://science.sciencem

ag.org/D

ownloaded from

confounding transcriptional changes broughtabout by major cardiovascular anomalies seenat E10.5. ElaD/D placentas were categorized intoclass 1 or class 3 based on the gross morphologyof the corresponding embryos (fig. S3A). RNAsequencing (RNA-seq) and principal componentsanalysis revealed that both class 1 and 3 ElaD/D

placentas clustered closer to each other andaway from WT placentas (fig. S3B). Becauseclass 1 placentas are grossly indistinguishablefromWT counterparts, these results indicate thatthe observed transcriptional changes are dueto ELA deficiency rather morphological defectsalready present at the time of specimen collec-tion. Gene set enrichment analysis (GSEA) (7)

revealed that class 1 and 3 ElaD/D placentas havea gene signature indicative of an elevated hypox-ic response (Fig. 3B; fig. S3, C and D; and tableS1). Consistent with this observation, ElaD/D

placentas have high levels of stabilized Hif1a(fig. S3, E and F) and decreased levels of prolyl-hydroxylated Hif1a (fig. S3, G and H), which istargeted for degradation under normoxia (8).Concurrently, and possibly as part of the ele-vated hypoxic response, Ela deficiency results inan up-regulation of pro-angiogenic genes, evenin class 1 placentas that are bereft of discerniblevascular defects (Fig. 3C).A close examination of differentially regulated

genes revealed a dramatic enrichment in genes

and pathways defining endothelial tip cells (Fig. 3,D and E) (9). Tip cells form the leading edge ofsprouting endothelial cells and migrate in re-sponse to pro-angiogenic signals (10). Functioningin the same way as axonal growth cones, tip cellsextend filopodia to determine the direction of theangiogenic sprout, whereas trailing stalk cellsproliferate to enable lumenogenesis and exten-sion of the vascular sprout (11, 12). Gene on-tology analysis confirmed functional hallmarksof tip cell identity such as vascular endothelialgrowth factor (VEGF) and semaphorin signaling,hormone secretion, axonogenesis, and filopo-dia extension (Fig. 3D). Quantitative polymerasechain reaction (qPCR) analysis validated the

Ho et al., Science 357, 707–713 (2017) 18 August 2017 2 of 7

Fig. 1. Zygotic deletion of Ela causesmidgestation lethality due to cardiovasculardefects and phenocopies loss of Apj.(A) Exon 3 of murine Ela was flanked withloxp sites and excised with cre recombinaseto generate the ElaD allele lacking theELA mature peptide (MP) coding region.(B) Schematic of cDNA from WT and ElaD

alleles. SP, signal peptide. (C) Semi-qPCR ofEla locus from genomic DNA (gDNA) andcDNA. Primer locations are indicated in(B). (D) Distribution of genotypes at E10.5and at weaning from intercrosses andElaD/D (mother) x Ela+/D (father) crosses.%P, penetrance; L, number of litters. Datawere tested using a chi-square test with1 degree of freedom for significant deviationfrom the expected distribution. (E to G) AtE10.5, Ela+/D embryos are indistinguishablefrom WT, whereas 43% (n = 17 of 39) ofElaD/D embryos and 14% (n = 3 of 22) ofApjD/D embryos display cardiovasculardefects along with IUGR. Scale bars, 1 mm.(H to J) At E10.5, Ela+/D yolk sacs havenormal vitelline vessels, whereas affectedElaD/D and ApjD/D embryos have avascular yolksacs with a ruffled appearance. Scale bars,1 mm. (K to M) CD31 staining of Ela+/D, ElaD/D,and ApjD/D yolk sacs reveals poorly maturedvasculature in mutant embryos. Scale bars,50 mm. (N to P) CD31 staining of Ela+/D, ElaD/D,and ApjD/D head vasculature at E10.5.Scale bars, 300 mm. (Q to S) CD31 (green)and SMA (red) staining of Ela+/D, ElaD/D,and ApjD/D hearts. Scale bars, 300 mm.(T) In situ hybridization of Ela at E8, showingmRNA localization in the region overlyingthe developing heart tube (ht) and chordalneural hinge (cnh). Scale bar, 200 mm.(U and V) RNAScope of Ela and Apj in E8embryo within its decidua showing expressionin the primitive foregut (fg) and hindgut(hg) endoderm. Arrowheads indicate thestart of Ela expression in the chorionictrophoblast. Scale bars, 100 mm. (W and X)RNAScope of Ela and Apj in E8 yolk saclayers adhering to underlying decidua.en, endoderm; me, mesoderm. Scale bars,40 mm.

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

up-regulation of tip cell markers and angiogenicgenes, including Vegfa, Apln, Plgf, Esm1, Igfbp3,Flt4, and Adm (Fig. 3F). This was confirmedby direct immunostaining against endogenous

Esm1, a specific tip cell marker (9, 13, 14), whichwas significantly up-regulated in both numberand intensity in ElaD/D placental sections, in-dicating ectopic tip cell differentiation (Fig. 3,

G to I). Using a Apln-LacZ knockin reporter(15) as an alternate marker of tip cell identity,we found that there are indeed more Apln-positive tip cells in ElaD/D labyrinths, with an

Ho et al., Science 357, 707–713 (2017) 18 August 2017 3 of 7

Fig. 2. ELA is a preg-nancy hormonerequired for placentalangiogenesis. (A andB) At E9, Elais expressed in the cho-rionic plate (cp) of thechorioallantoic placenta,and its receptor Apj isexpressed in fetal allan-toic endothelial cells. d,decidua. Scale bars,1 mm. (C and D) AtE10.5, Ela expression inthe placenta labyrinth(lb) is restricted to STs,whereas Apj expressionis restricted to endo-thelial cells adjacent toSTs. Scale bars, 100 mm.(C′ and D′) Highermagnification showingEla expression in STssurrounding maternalblood spaces (mbs) andApj expression in endo-thelial cells (EC) liningfetal blood spaces (fbs).Scale bars, 200 mm.(E and F) ELA can bedetected by immuno-histochemistry using anELA-specific antibody (aC) in WT E10.5 labyrinthin cells lining bloodspaces (arrowheads) butnot in ElaD/D placentas.(G) ELISA detects circu-lating ELA in maternalserum harvested at indi-cated gestational days(GD). n = number of miceassayed at each gesta-tional time point. NP,nonpregnant. Error barsindicate SEM of threeindependent experiments.Data were tested withone-way analysis of vari-ance (ANOVA) (redasterisk) and with two-sample Student’s t test(black asterisks).(H) ELISA of GD 10.5maternal serum harvested from WTor ElaD/D mothers mated with the WTor ElaD/D fathers, indicating a maternal and zygotic origin of circulating ELA duringpregnancy. n = number of mice assayed at each gestational time point. Error bars, SEM of three independent experiments. Data were tested usingone-way ANOVA. In (G) and (H), ELA detected in maternal zygotic knockout is attributed to assay background. (I and J) Hematoxylin and eosin staining ofEla+/D and ElaD/D E10.5 placentas showing poor invasion and angiogenesis of ElaD/D placentas. Scale bars, 250 mm. al, allantois. (K and L) CD31/Pecam-1 stainingof E10.5 Ela+/D and ElaD/D showing a paucity of fetal endothelial cells in the labyrinth. Scale bars, 100 mm. (M and N) Alpp (placenta alkalinephosphatase) staining of E10.5 Ela+/D and ElaD/D placentas showing lack of trophoblasts in the labyrinth. Scale bars, 50 mm. *P < 0.05, **P < 0.01from indicated tests of significance.

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

Ho et al., Science 357, 707–713 (2017) 18 August 2017 4 of 7

Fig. 3. Loss of Ela causes hypoxic response and up-regulation of apro-angiogenic program. (A) Schematic of RNA-seq experiment of E9.5WT versus ElaD/D labyrinths. (B and C) GSEA analysis of ElaD/D (classes1 and 3) showing an up-regulation of hypoxic response and pro-angiogenicgenes in ElaD/D labyrinths, even in morphologically normal class 1 placentas.(D) Gene ontology analysis of genes up-regulated in ElaD/D labyrinths. Inred are pathways enriched in tip cells. P values are derived from a binomialdistribution with Bonferroni correction. (E) GSEA detects an up-regulationof endothelial tip cell genes in class 1 ElaD/D labyrinths. (F) qPCR validationof tip cell–enriched and angiogenic genes in ElaD/D labyrinths (n = 6 WT;n = 6 ElaD/D). Error bars indicate SEM of two independent experiments.

(G) Esm1 immunofluorescence on E9.5 placenta vibratome sections takenfrom medial planes containing the maternal central canal. Dotted line marksthe position of the transitional zone. al, allantois; lb, labyrinth; d, decidua.Scale bars, 40 mm. (H) Number of Esm1+ cells per section (n = 6 WT; n =7 ElaD/D; each section represents a distinct placenta). (I) Seventy-fifthpercentile integrated density of Esm1+ cells in each placental samplequantified in (H). Data presented as arbitrary units (A.U.). (J) b-galactosidase(LacZ transgene in AplnD allele) staining of Ela+/+;Apln+/D and ElaD/D;Apln+/D

placentas indicate increased Apln expression in ElaD/D labyrinths. Scalebars, 300 mm. Data are depicted as mean ± SEM. *P < 0.05, **P < 0.01,***P < 0.001 of two-sample Student’s t test.

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

overall stunted architecture characterized bylittle or no extension and branching of angio-genic sprouts (Fig. 3J and fig. S3, I and J). Thisfinding suggests that the absence of ELA causesan expansion of tip versus stalk cells, which is

expected to impair lumenogenesis and sproutextension, such as in mice haplo-insufficientfor the Notch ligand Dll4 (16, 17). Such an im-balance in tip cell identity is consistent withproliferative perturbations seen in ElaD/D laby-

rinths, which are hypoproliferative (fig. S2, Hand I), and ElaD/D yolk sacs, which are, converse-ly, hyperproliferative (fig. S3, K to O), with anoverall negative effect on angiogenesis. Our re-sults suggest that the ELA actively suppresses

Ho et al., Science 357, 707–713 (2017) 18 August 2017 5 of 7

Fig. 4. Endogenous ELA prevents preeclampsia (PE), and exogenousELA administration rescues PE symptoms in Ela-deficient mice.(A) Urine protein/creatinine ratios fromGD15pregnantmothers (♀) of indicatedgenotype mated with fathers (♂) of indicated genotypes. Each dot representsan individual mouse. Error bars indicate SEM. (B) Repeated tail-cuff systolicblood pressure measurements of WTmothers (n = 7, mated to WT fathers)and ElaD/Dmothers (n = 5, mated to ElaD/D fathers) at the indicated gestationalage. Dotted line indicates day of parturition. Error bars indicate SEM.Two-way ANOVA analysis detected a significant interaction between timeand genotype, F(7,49) = 2.074; P = 0.0413; i.e., ElaD/D females developedsignificantly higher systolic BP compared to the controls as pregnancyprogressed. Asterisks indicate significance of two-sample unpaired t test betweenWTand ElaD/Don the indicated GD. (C) BP readings from (B) calculated in theform of delta BP (BP of indicated GD minus baseline nonpregnant BP of thesamemother). Each dot representsBPof onemouse averagedover 20 readings;error bars indicate SEM.Two-way ANOVA test detected a statistically significantdifference in mean delta BP between WTand ElaD/Dmice; F(1,16) = 11.28, P =0.0100. Asterisks indicate significance of two-sample unpaired t test. (D) Weightof pups at E18.5 collected by caesarean section. Each dot represents one pup.Error bars indicate SEM. (E) Urine protein/creatinine ratios of WTmothers

(mated to WT fathers) (black squares) and ElaD/Dmothers (mated to ElaD/D

fathers) (red circles) implanted at GD 7 with infusion pumps containing eitherphosphate-buffered saline (PBS) (closed symbols) or synthetic ELA peptide(open symbols) measured at GD 15. Each dot represents one mouse; error barsindicate SEM. (F) Systolic delta BP measurements of subjects in (E)measured at GD 14, 16, and 18. Each dot represents one mouse; error barsindicate SEM.Two-way ANOVA test detected a statistically significant differencein mean delta BP between ElaD/D + PBS and ElaD/D + ELAmice; F(1,16) = 6.938,P = 0.0300. Asterisks indicate significance of two-sample unpaired t test.(G) Immunohistochemistryof ELAwithaCbiotinylatedELA-specific antibodyonhuman first trimester (8 + 3 weeks) placental formalin-fixed paraffin-embeddedsections. Scale bars, 200 mm. (H) Transwell invasion assay using Jarchoriocarcinoma cells cultured in the presence of increasing concentrations ofsynthetic ELA peptide. Each dot represents the mean of three wells; errorbars indicate SEM of three independent experiments. (I) Workingmodel: mouseELA, produced by STcells, signals to APJ expressed on fetal endothelial cells(ECs) to facilitate normal placental angiogenesis. ELA also enters the maternalcirculation, where it acts systemically to prevent symptoms of preeclampsiaduring pregnancy.Tb, trophoblast. Data are depicted as mean ± SEM. *P < 0.05,**P < 0.01, ***P < 0.001 of two-sample Student’s t test, unless otherwise stated.

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

the expression of tip cell genes such as Esm1and Apln.Defects in genes required for placental de-

velopment and angiogenesis frequently lead topreeclampsia in mice (18, 19). In light of theplacental defects seen in ElaD/D conceptuses,including a prominent gene signature of in-creased inflammatory response (fig. S4A), andincreased expression of Esm1 (20, 21) and Adm(22), which have been linked to PE in humans,we hypothesized that ElaD/D mothers might ex-hibit symptoms of PE. We thus assessed ElaD/D

pregnant mice for signs of proteinuria and hyper-tension, two diagnostic hallmarks of PE. Indeed,by decreasing the number of WT Ela alleles infetuses and their pregnant mothers, the urineprotein/creatinine ratio at gestational day (GD)15 increased dramatically, indicating an inversecorrelation between endogenous ELA levels andthe severity of proteinuria (Fig. 4A). At the endof pregnancy, histology and transmission elec-tron microscopy of kidney glomerular sectionsrevealed signs of endotheliosis in ElaD/D preg-nant mothers (fig. S4, B to G), a unique renalpathology of women suffering from PE (23, 24).Glomeruli from ElaD/D pregnant mothers wereswollen and had occluded capillaries, with evi-dence of protein and vesicular deposition onendothelial cells, absence of proper endothelialfenestration, and coagulation of red blood cellsin capillary lumens (fig. S4, B to G). Podocytes,on the other hand, appeared normal. Proteinuriawas not observed in nonpregnant ElaD/D mice(fig. S4H), indicating that the renal pathology isunique to pregnancy. Next, we employed a tail-cuff method to measure systolic blood pressure(BP) throughout pregnancy, after training themice for a minimum of 3 days before mating.Although there were no significant differencesin the nonpregnant baseline BP betweenWTandElaD/D mice, pregnant ElaD/D mothers (mated toElaD/D fathers) had significantly higher systolic BP,which returned to normal levels postparturition(Fig. 4B), and delta BP (pregnant BP minusbaseline BP) on GD 16 and 18 (Fig. 4C). In addi-tion, ElaD/D pups from ElaD/D mothers collectedby caesarean section at E18.5 were significantlylower in weight compared with WT pups fromWTmothers (Fig. 4D). This is reminiscent of IUGR,which frequently accompanies PE and placentalinsufficiency. Our findings indicate that ElaD/D

mice suffer from PE and suggest that ELA isnecessary for regulating maternal cardiovascularhomeostasis to prevent gestational hypertension.To determinewhether the loss of ELA is upstreamof well-established biomarkers of human PE, wemeasured both maternal plasma and placentalmRNA levels of sFlt1 (24), Vegf (25), and Plgf(19, 26). At late gestation, ElaD/D placentas haveincreased levels of sFlt1, Vegfa, and PlgfmRNA(fig. S4I), although these transcriptional changesdid not translate into significantly elevated plas-ma levels of the respective proteins (fig. S4, Jto L). Altogether, these data indicate that ElaD/D

mice are not developing PE symptoms simplydue to a decrease in the Plgf/sFlt1 ratio but sug-gest that ELA acts independently of, and possibly

earlier than, these angiogenic factors in the patho-genesis of PE.It is noteworthy that overexpression of Apln

(Fig. 3F), which is the alternate ligand for APJ, isnot sufficient to rescue ElaD/D placentas, suggest-ing that these ligands elicit different signalingoutcomes. Indeed, unlike ElaD/D mothers, AplnD/D

mothers do not develop hypertension (fig. S5A)and in fact have significantly lower levels of pro-teinuria (fig. S5B). Similarly, AplnD/D placentasdo not aberrantly up-regulate endothelial tipcell markers Esm1 and Igfbp3, as seen in ElaD/D

placentas (fig. S5C). To further investigate thebiological differences between APLN and ELA,we treated APJ-expressing primary allantoic cul-tures from somite stage embryos with equal con-centrations of ELA, APLN, or both (fig. S5D).Wefound that ELA and APLN elicited opposite ef-fects on the expression of Esm1 and several hy-poxic response genes (fig. S5, D to E′). Lastly, wefound that ELA could directly repress the expres-sion of Apln in these allantoic explants (fig. S5F),raising the possibility that Apln derepression inthe absence of ELA drives excessive and patho-genic tip cell differentiation. Indeed, we find thatin a litter ofElaD/D null embryos, themost severelyaffected embryos are the ones expressing thehighest levels ofApln (fig. S5G).Moreover, the twoligands display distinct spatiotemporal expressionwhere Ela is highly concentrated and restrictedto the developing heart, caudal neural tube andtrophoblasts, whereas Apln is diffusely expres-sed and widespread in all embryonic and extra-embryonic tissues (fig. S5, H to I′). Altogether, ourdata demonstrate that Ela and Apln are biolog-ically distinct and elicit opposing effects on pla-cental angiogenesis and symptoms of PE.Because ELA appears to act as a systemic

hormone during pregnancy, we asked whetheradministration of synthetic ELAduringpregnancymay alleviate PE symptoms. ELA infusion did notaffect BP and proteinuria parameters in pregnantWTmice (Fig. 4F), nor did it have noticeable sideeffects on embryogenesis, as measured by fetalweight, morphology, and subsequent postnataldevelopment. We were able to normalize protein-uria (Fig. 4E) and BP (Fig. 4F) in ElaD/D pregnantmothers infused with recombinant ELA from GD7.5 onwards, which is consistent with our modelthatELAacts as a systemichormone. Furthermore,infusion of ELA rescued the weight of ElaD/D

fetuses (Fig. 4D) and glomerular endotheliosis inpregnant ElaD/D mothers, as assayed by periodicacid–Schiff and a-fibrinogen staining (fig. S6).In humans, we find that ELA is predominantly

expressed in villous cytotrophoblasts and STs offirst-trimester placental tissue (8 + 3 weeks) (Fig.4G) and term placentas (fig. S4, M and N). In PE,extravillous trophoblast invasion and subsequentspiral artery remodeling are frequently incom-plete, leading to shallow and defective placentation(27). We therefore hypothesized that, in humans,ELA might contribute to trophoblast invasion.Indeed, addition of ELA to trophoblast-like JARchoriocarcinoma cells significantly increased theirinvasiveness in transwell invasion assays (Fig. 4H)(28). These data suggest that ELA, secreted from

the ST layer, has a paracrine effect on trophoblastsdifferentiating into invasive extravillous tropho-blasts. ELAactivity potentiates invasion andmightenhance subsequent spiral artery remodeling toprevent the development of PE during humanpregnancy.Pregnancy is a unique physiological state as-

sociated with increased cardiovascular demandand burden. Many processes work in concert toimpart cardiovascular homeostasis in the preg-nant female, although to date these are largelyunknown. In the mouse, we propose that ELAproduced by placental trophoblasts functionsin two ways to prevent gestational hypertension(Fig. 4I). First, ELA exerts paracrine effects onfetal endothelial cells, where it curbs inappropriatedifferentiation of endothelial tip cells. This enablesnormal branching angiogenesis and the formationof an adequate labyrinth network required forproper perfusion. Second, ELA enters the mater-nal circulation to regulate cardiorenal function.We speculate that the latter role might have adirect effect on the maternal endothelium (e.g.,through the stimulation of vasodilatory mecha-nisms such as nitric oxide production) (29) or byregulating diuresis (30). Although the PE-protectiveeffects are presumably achieved through APJsignaling in endothelial cells, we do not rule outa possible contribution from as-yet-unidentifiedELA receptors (1). ELA deficiency in the mouseleads to classical PE symptoms together withgross abnormalities in placental development.Similarly, ELA is expressed by trophoblasts inthe chorionic villi of human placentas and po-tentiates trophoblast invasion in vitro. We spec-ulate that in humans ELA might contribute toplacentation by stimulating trophoblast migra-tion and invasion, in addition to direct effects onthematernal endothelium, although these remainto be investigated.In conclusion, we propose that ELA is a cir-

culating hormone produced by the mammalianplacenta to ensure cardiovascular integrity of bothmother and fetus during pregnancy. Our resultsraise the possibility that transient enforcementof the ELABELA-nergic axis might therefore bebeneficial for conditions that displayhypertension,such as, but not limited to, preeclampsia.

REFERENCES AND NOTES

1. L. Ho et al., Cell Stem Cell 17, 435–447 (2015).2. S. C. Chng, L. Ho, J. Tian, B. Reversade,Dev. Cell 27, 672–680 (2013).3. A. Pauli et al., Science 343, 1248636 (2014).4. C. S. Helker et al., eLife 4, e06726 (2015).5. M. Li et al., Cell Stem Cell 16, 669–683 (2015).6. A. S. Hassan, J. Hou, W. Wei, P. A. Hoodless, Gene Expr.

Patterns 10, 127–134 (2010).7. A. Subramanian et al., Proc. Natl. Acad. Sci. U.S.A. 102,

15545–15550 (2005).8. T. Hagen, C. T. Taylor, F. Lam, S. Moncada, Science 302,

1975–1978 (2003).9. R. del Toro et al., Blood 116, 4025–4033 (2010).10. H. M. Eilken, R. H. Adams, Curr. Opin. Cell Biol. 22, 617–625 (2010).11. P. Carmeliet, Nat. Med. 9, 653–660 (2003).12. R. H. Adams, A. Eichmann, Cold Spring Harb. Perspect. Biol. 2,

a001875 (2010).13. S. F. Rocha et al., Circ. Res. 115, 581–590 (2014).14. G. A. Strasser, J. S. Kaminker, M. Tessier-Lavigne, Blood 115,

5102–5110 (2010).15. D. N. Charo et al., Am. J. Physiol. Heart Circ. Physiol. 297,

H1904–H1913 (2009).

Ho et al., Science 357, 707–713 (2017) 18 August 2017 6 of 7

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

16. S. Suchting et al., Proc. Natl. Acad. Sci. U.S.A. 104, 3225–3230(2007).

17. A. Duarte et al., Genes Dev. 18, 2474–2478 (2004).18. J. Singh, A. Ahmed, G. Girardi, Hypertension 58, 716–724

(2011).19. S. Venkatesha et al., Nat. Med. 12, 642–649 (2006).20. X. Chang et al., Int. J. Clin. Exp. Pathol. 8, 14733–14740 (2015).21. H. Adekola et al., J. Matern. Fetal Neonatal Med. 28, 1621–1632

(2015).22. A. Al-Ghafra, N. M. Gude, S. P. Brennecke, R. G. King, Mol.

Hum. Reprod. 12, 181–186 (2006).23. I. E. Stillman, S. A. Karumanchi, J. Am. Soc. Nephrol. 18,

2281–2284 (2007).24. S. E. Maynard et al., J. Clin. Invest. 111, 649–658 (2003).25. R. Hayman, J. Brockelsby, L. Kenny, P. Baker, J. Soc. Gynecol.

Investig. 6, 3–10 (1999).26. A. Reuvekamp, F. V. Velsing-Aarts, I. E. Poulina, J. J. Capello,

A. J. Duits, Br. J. Obstet. Gynaecol. 106, 1019–1022 (1999).27. S. J. Fisher, Am. J. Obstet. Gynecol. 213, S115–S122 (2015).28. S. Yagel, R. S. Parhar, J. J. Jeffrey, P. K. Lala, J. Cell. Physiol.

136, 455–462 (1988).

29. C. A. Schreiber, S. J. Holditch, A. Generous, Y. Ikeda, Curr. GeneTher. 16, 349–360 (2016).

30. C. Deng, H. Chen, N. Yang, Y. Feng, A. J. Hsueh, J. Biol. Chem.290, 18261–18268 (2015).

ACKNOWLEDGMENTS

We thank T. Quertermous (Stanford University, USA) for sharingthe Apln and Apj knockout mice; T. Veenboer (AMC, Netherlands)for assistance with human placenta samples; D. Kalicharan fortechnical assistance with electron microscopy; members of theAdvanced Molecular Pathology Laboratory, Institute of Molecularand Cell Biology, for assistance with histopathology; the Institute ofMedical Biology Microscopy Unit for assistance with imaging; andthe Genome Institute of Singapore sequencing facility forassistance with RNA-seq. RNA-seq data are deposited in theNational Center for Biotechnology Information (NCBI) SequenceRead Archive under accession number PRJNA391361. The datareported in this manuscript are tabulated in the main paper and inthe supplementary materials. The authors acknowledge financialsupport from the A*STAR Joint Council Organization (YoungResearcher Collaborative Grant) and a Strategic Positioning Fund

on Genetic Orphan Diseases from the Biomedical Research Council,A*STAR, Singapore. M.v.D. and S.B. are supported by a FerringResearch Institute Innovation Grant. B.R. is a fellow of the BrancoWeissFoundation, a recipient of the A*STAR Investigatorship, an EMBOYoung Investigator, and an AAA fellow from AMC/VUmc. The authorsdeclare no competing financial interests. B.R. and L.H. are inventors onpatent application 10201605841S submitted by A*STAR, Singapore,which covers the use of ELABELA for the diagnosis and treatment ofpreeclampsia. The ELABELA-deficient mice and custom antibody areavailable from B.R. under a material transfer agreement with A*STAR,Singapore.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/357/6352/707/suppl/DC1Materials and MethodsFigs. S1 to S6Table S1References (31–36)

27 December 2016; accepted 21 June 2017Published online 29 June 201710.1126/science.aam6607

Ho et al., Science 357, 707–713 (2017) 18 August 2017 7 of 7

RESEARCH | REPORTon M

arch 28, 2021

http://science.sciencemag.org/

Dow

nloaded from

ELABELA deficiency promotes preeclampsia and cardiovascular malformations in mice

ReversadeBoussata, Grace Hui-Yi Goh, Gijs B. Afink, Chin Yan Lim, N. Ray Dunn, Davor Solter, Barbara B. Knowles and Bruno Lena Ho, Marie van Dijk, Sam Tan Jian Chye, Daniel M. Messerschmidt, Serene C. Chng, Sheena Ong, Ling Ka Yi, Souad

originally published online June 29, 2017DOI: 10.1126/science.aam6607 (6352), 707-713.357Science 

, this issue p. 707; see also p. 643ScienceELABELA displayed defective heart development, and full-term pups had low birth weights.signs of preeclampsia, including high blood pressure and elevated urine protein. A proportion of embryos lackingin preeclampsia (see the Perspective by Wirka and Quertermous). ELABELA-deficient pregnant mice showed clinical

identified ELABELA as a hormone produced by the placenta whose levels are loweret al.informative animal models. Ho be life-threatening to the mother and baby. Research leading to effective treatments has been hampered by a lack of

Preeclampsia, a dangerous pregnancy disorder marked by high blood pressure, can lead to premature birth andModeling a pregnancy disorder

ARTICLE TOOLS http://science.sciencemag.org/content/357/6352/707

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2017/06/28/science.aam6607.DC1

CONTENTRELATED

http://stm.sciencemag.org/content/scitransmed/7/290/290ra88.fullhttp://stm.sciencemag.org/content/scitransmed/7/319/319fs51.fullhttp://stm.sciencemag.org/content/scitransmed/7/319/319ra204.fullhttp://stm.sciencemag.org/content/scitransmed/8/364/364ra154.fullhttp://stm.sciencemag.org/content/scitransmed/6/245/245ra92.fullhttp://science.sciencemag.org/content/sci/357/6352/643.full

REFERENCES

http://science.sciencemag.org/content/357/6352/707#BIBLThis article cites 35 articles, 11 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2017 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on March 28, 2021

http://science.sciencem

ag.org/D

ownloaded from