eo

r

a

e

Abstract

9Uterine spiral arteries play a vital role in supplying nutrients to the placenta and fetus, and for this purpose they are remodelled into highlydilated vessels by the action of invading trophoblast (physiological change). Knowledge of the mechanisms of these changes is relevant for a bet-ter understanding of pre-eclampsia and other pregnancy complications which show incomplete spiral artery remodelling. Controversies stillabound concerning different steps in these physiological changes, and several of these disagreements are highlighted in this review, therebysuggesting directions for further research. First, a better definition of the degree of decidua- versus trophoblast-associated remodelling mayhelp to devise a more adequate terminology. Other contestable issues are the vascular plugging and its relation with oxygen, trophoblast invasionfrom the outside or the inside of the vessels (intravasation versus extravasation), the impact of haemodynamics on endovascular migration, thereplacement of arterial components by trophoblast, maternal tissue repair mechanisms and the role of uterine natural killer (NK) cells. Several ofthese features may be disturbed in complicated pregnancies, including the early decidua-associated vascular remodelling, vascular pluggingand haemodynamics. The hyperinflammatory condition of pre-eclampsia may be responsible for vasculopathies such as acute atherosis,although the overall impact of such lesions on placental function is far from clear. Several features of the human placental bed are mirroredby processes in other species with haemochorial placentation, and studying such models may help to illuminate poorly understood aspectsof human placentation. 2005 Elsevier Ltd. All rights reserved.

Keywords: Spiral arteries; Trophoblast invasion; Vascular remodelling; Human; Pre-eclampsia; Decidualization; Uterine NK cells

1. Introduction

The interrelationship between maternal and fetal circula-tions enabling physiological exchange within the placenta isstill not fully understood. Probably the most significant break-through was achieved in the 18th century when William andJohn Hunter demonstrated by injection experiments thatmaternal and fetal vascular structures remain separate, thusrefuting the previously held opinion of a continuous mother-to-fetus circulation. This might have been a chance observation,since injecting blood vessels with coloured wax was becoming

time. Most probably it was the younger brother John, who ac-tually did the dissections, who quickly grasped the physiolog-ical implications of this observation, which in later years led toa quarrel concerning priority [1,2]. Nevertheless, it was Wil-liam who illustrated for the first time spiral arteries in his fa-mous Anatomia uteri humani gravidi tabulis illustrata e Theanatomy of the human gravid uterus exhibited in figures(1774) [3], in a series of superb engravings based on carefuldissections of several pregnant uteri. The spiral arteries werethen described as arteriae convolutae/convoluted arterieswhich are passing between the womb and placenta. TheseThe Uterine Spiral ArteriFacts and C

R. Pijnenborg*, L. Ve

Department of Obstetrics & Gynaecology, Universitair Ziekenhuis G

Accepted 20 D

Placenta 27 (2006)a routine procedure for anatomical demonstrations at that

* Corresponding author. Tel.: 32 16 346190; fax: 32 16 344205.E-mail address: robert.pijnenborg@uz.kuleuven.ac.be (R. Pijnenborg).

0143-4004/$ - see front matter 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.placenta.2005.12.006s In Human Pregnancy:ntroversies

cruysse, M. Hanssens

sthuisberg, Katholieke Universiteit Leuven, B3000 Leuven, Belgium

cember 2005

39e958vessels were most clearly visible in drawings of the insideof the uterus after separation of the placenta, which we brokethrough in separating these two parts. This short passage re-veals their ignorance about the factual arrangement of mater-nal blood supply to the placenta. In spite of the ingenuity of

this point is considered to form the demarcation line betweenradial and spiral arteries. Meanwhile observations on placentalbed biopsies have confirmed the deep myometrial origin ofbasal arteries, which nourish both the inner myometrium andbasal endometrium. Therefore the designation of the inner my-ometrial arteries as spiral arteries has subsequently been fol-lowed. A precise terminology e although it may look likea rather insignificant issue e is important regarding the depthover which pregnancy-associated changes in spiral arteries arereported to occur.

Studies on the endometrial spiral arteries of the non-pregnant uterus indicated the remarkable sensitivity of thesevessels to stimuli by hormones or growth factors, while thebasal arteries are thought to be more stable non-hormone re-sponsive structures [4,8,9]. The convoluted course of the spiralarteries obviously results from arterial growth exceeding theincrease in endometrial thickness during the cycle and duringpregnancy. There can be no doubt that this spiral shape has

Fig. 1. Vascular anatomy of the pregnant uterus according to (A) Ramsey and Dlocation of the branching points of basal arteries from the main arteries, Ramsey

confirmation of deep basal artery branching points, spiral arteries were recogncoiling of endometrial arteries is related to decidualization ofthe endometrium which in primates is started off spontane-ously during the luteal phase of each cycle. In non-menstruat-ing animals such as rats, mice and hamsters, spiral outgrowthof uteroplacental arteries definitely occurs during pregnancyand is also associated with the decidualization process, whichin these animals does not occur spontaneously during their es-trous cycle, but is induced by blastocyst implantation [11].

2. Basic histological observations on spiral arteriesduring pregnancy: a historical overview

The earliest histological observations on uterine spiral ar-teries in human pregnancy were reported by German investiga-tors in the last third of the 19th century. This early researchwas summarized and extensively discussed by Grosser, andhis monograph should be consulted for detailed references[12]. He quoted Friedlander (1870) as the first to have

onner [8] and (B) Brosens et al. [7]. Because of initial uncertainties about thethe Hunter brothers, they could not have envisioned at thattime how the uteroplacental circulation is established by theaction of invading cells from the placenta. Since this reviewdeals a lot with placentaleuterine interactions, it may be rele-vant that it was also William Hunter who introduced the termdecidua for the thickened inner lining of the pregnant uteruswhich is discarded at parturition. The term is etymologicallysignificant since the Latin word decidere means to falloff, which also applies to the expulsion of the decidualizinguterine mucosa at menstruation.

After these pioneering observations it took more than a cen-tury to clarify the detailed anatomy of the uterine vasculature[4e6]. The matter is basically settled now, although some con-fusion arose concerning terminology, i.e. whether the arteriesin the inner myometrium should be considered as spiral [7]or radial arteries [8] (Fig. 1). Part of the argument was dueto a disagreement about the level where the basal arteriesare branching off in the decidua or in the myometrium, as

haemodynamical repercussions. John Hunter already notedthat The intention of these spiral turns would appear to bethat of diminishing the force of the circulation as it approachesthe spongy substance of the placenta. where quick motion ofthe blood is not wanted (quoted by Ramsey [6]). Their pecu-liar shape would not only lead to a progressive decrease inblood pressure along their length, but would also dampenthe pulse, which is important for maintaining a steady bloodflow to the intervillous space of the placenta. Ramsey foundthat the coiling of the spiral arteries still increased in earlypregnancy, but that uncoiling started around midpregnancy[6,10]. The reserve length of the convoluted vessels obviouslyallows easy stretching during pregnancy-associated uterinegrowth.

Reynolds [4] made the interesting point that coiled or spiralarteries are not seen in the endometria of most non-pregnantanimals, and therefore he speculated that their peculiar shapein primates may be related to menstruation. In fact, increased

940 R. Pijnenborg et al. / Placenta 27 (2006) 939e958considered the whole myometrial segments as radial arteries (A). Following later

ized as having myometrial as well as decidual segments (B).

described endovascular cells in maternal arteries at midpreg-nancy, without making any statement as to their possible ori-gin. Most investigators then thought that these cells werederived from the decidua or endothelium, but Grosser anda few others (Aschoff; Hinselman; Meyer; Schickele; allquoted by Grosser [12]) favoured a trophoblastic origin(Fig. 2). Furthermore, they made the association between in-vading trophoblastic cells and fibrinoid changes in vesselwalls, thought to result from vascular degeneration by fermen-tative action of a trophoblast-secreted choriotrypsin. Theyconceived that in this way maternal tissues are digested and re-placed by fetal cells, sometimes leading to complete pluggingof the vessels. Also breakdown of the elastica and the appear-ance of intimal cushions were described, although it was ques-tioned whether the latter alterations were related to thepregnancy rather than to maternal age. Finally Grosser madethe astute remark that the described vascular changes were re-stricted to the Placentarstelle (presumably the placental in-sertion site, i.e. the placental bed), and were extended to thesuperficial myometrium. It is therefore clear that at the begin-ning of the 20th century several features of spiral artery mod-ification had already been recognized by several authors, yetmost observations were anecdotal, performed on occasionalspecimens. In order to elucidate the step-by-step alterationsin spiral artery conversion in relation to trophoblast invasion,series of well-dated pregnant uteri with the placenta in situhad to be studied in a systematic fashion. Famous classic stud-ies of human placentation, which included observations onspiral artery transformation, were those by Boyd and Hamilton[13e16] and Harris and Ramsey [10,17].

Hamilton and Boyd [16] described the presence of intravas-cular cells in spiral arteries, for which they favoured a cytotro-phoblastic shell origin based on histological continuity, ratherthan their being derived from surrounding maternal or fetalcells. They noted that these cells may form intravascular tro-phoblastic plugs which in their view could conceivably

R. Pijnenborg et al. / PlaFig. 2. Endovascular cells (ET) in spiral arteries of early pregnancy, as illus-

trated by Grosser [12].damp down the arterial pressure, but added that they com-pletely obliterate the vascular lumen only in regions where thecirculation has been passed by a new perforation of a spiralartery proximal to an already established opening. Theyalso described associated degenerative changes in the vesselwalls, but did not seem to consider that the endovascular cellsthemselves might become embedded into the wall. This is themore surprising since they noted their disappearance at the endof pregnancy, seemingly without considering a possible wayhow they might have been lost [9,14]. These authors did notconsider interactions between spiral arteries and interstitialtrophoblast, which they described as being overall multinu-clear. Indeed, they were not able to distinguish morpho-logically interstitial cytotrophoblast from decidual cells,immunohistochemical techniques not yet being available.Elizabeth Ramseys group, on the other hand, described howsome of the syncytial streamers (multinuclear interstitialtrophoblast), come into contact with the outer walls of spiralarteries, without finding evidence that these cells enter the ar-terial lumina. They thought that these perivascular tropho-blasts induced degenerative changes in the arterial walls[17], foreshadowing later findings of a correlation between in-terstitial trophoblast concentrations and degenerative changesin spiral arteries preceding endovascular trophoblast invasion[18]. Harris and Ramsey also noted the appearance of largestellate cells within the vessel walls, but could not ascertaintheir trophoblastic origin. Apart from this interstitial to peri-vascular invasive pathway, like Hamilton and Boyd [15,16],they observed intra-arterial migration by cytotrophoblast thatpresumably originated directly from the cytotrophoblasticshell, proceeding as far as the myometrium. Their ultimatefate was unknown, and they even wondered whether intravas-cular cells might ultimately be washed out into the intervillousspace. They considered that accumulating endovascular cellsmight cause cessation of circulation which, according to theirhistological observations, would be overcome by the forma-tion of multiple openings of spiral arteries into the intervillousspace. Ramsey et al. did consider a bidirectional invasion ofthe spiral arteries from the inside as well as from the outsideof the vessels, the latter illustrated by suggestive microphoto-graphs of perivascular trophoblast in contact with the endothe-lium, seemingly pushing this layer towards the lumen. Suchhistological appearances may of course reflect the tendencyof endothelium to grow over anything placed within a vessellumen [19]. The contribution of endovascular cells to arterialwall changes was supported by Ramseys observations on sim-ilar vascular changes in rhesus monkeys, where interstitial tro-phoblast invasion is uncommon [10].

Both groups (Boyd and Hamilton, Ramsey and coworkers)regularly referred to animal work in their publications to illus-trate the probable trophoblastic nature of cells whose intravas-cular migratory behaviour was regarded as an oddity. Asa support to the reality of this observation, Billingtons 1966paper was frequently quoted, in which he described a similarinvasive behaviour of hamster trophoblast towards the arteries

941centa 27 (2006) 939e958after transplantation of ectoplacental cone tissue into the testis[20]. Also Orsinis study on the golden hamster [21] was often

mentioned, because it revealed a very intensive intravascularinvasion by trophoblast. The same author developed a unilater-ally pregnant deciduoma model, allowing direct comparisonbetween invaded and noninvaded decidua within the same an-imal, thus confirming the trophoblastic nature of endovascularcells in uterine spiral arteries at implantation sites [22].

The aforementioned studies on spiral arteries and their rolein placental development and function were performed froma mainly academical point of view. A change of perspectivewas provided by the publication of a first report in 1953 ondefective placental perfusion in hypertensive pregnancies byBrowne and Veall [23]. This paper suddenly revealed a possi-ble clinical impact of such studies, and stimulated a search forvasculopathies in spiral arteries which might be responsiblefor disturbed placental flow in pathological conditions. There-fore it was necessary to devise a technique for obtainingappropriate tissue samples to be used as a clinical tool. In1956 the first placental bed biopsy was collected during a cae-sarean section by Dixon and Robertson, who subsequentlyjoined forces with Renaer and Brosens who had independentlybeen alerted by the same Browne and Veall paper [24]. Under-standing the histology of this unfamiliar term pregnancy mate-rial was the first challenge, and even biopsies from normalpregnancies presented unexpected difficulties. Instead of thefamiliar spiral arteries known from endometrial tissue sam-ples, big calibre vessel-like profiles were observed, lackingrecognizable vascular smooth muscle, but showing insteadan amorphous fibrinoid wall with embedded pale-lookingcells. Referring to the comprehensive papers by Hamiltonand Boyd and Harris and Ramsey, which were mainly con-cerned with early stages of placentation, did not providemuch guidance at that time. By studying serial sections ofwhole pregnant uteri with placentas in situ Brosens et al. [7]proved that these dilated structures are indeed spiral arterieswhich, by having lost their smooth muscle and elastica, hadbecome physiologically changed to accommodate theincreasing maternal blood supply to the placenta. They alsohypothesized that the fibrinoid-embedded cells are trophoblastic,derived from the endovascular cells which had been observedin the lumina of the spiral arteries by previous investigators.

This pioneering groundwork was followed by the discoverythat in pre-eclamptic women the normal physiological changesof spiral arteries are restricted to the decidual segments [25]. Itwas then felt that a better knowledge of the successive steps ofthe normal vascular transformation would be essential forunderstanding the defects in complicated pregnancies. As a re-sult, several lines of study were started. Electronmicroscopicalobservations supported the endovascular origin of the embed-ded mural cells in spiral arteries by describing successivesteps of endothelial penetration, fibrinoid secretion and muralincorporation [19,26]. A separate study on complete hysterec-tomy specimens of early pregnancy with placentas in situtraced the successive steps in the physiological change of spi-ral arteries [18,27,28]. One of the questions asked was in howfar trophoblast-associated physiological changes were pre-

942 R. Pijnenborg et al. / Placeded by structural alterations in spiral arteries, as previouslyobserved in pregnant golden hamsters [29,30]. By applyingOrsinis unilaterally pregnant deciduoma model, early vascularchanges had been detected in this animal which were closelyassociated with the decidualization process, thus providingthe inspiration for investigating this feature in the human.Brettner [31] had already reported the swelling of vascularsmooth muscle in the decidualized human endometrium. InDixons 8e18 weeks hysterectomy collection early vascularalterations leading to media disruption were also observed inmyometrial spiral arteries as a preliminary step preceding en-dovascular trophoblast invasion. While some of the early vas-cular changes (vacuolation) also occurred outside the placentalbed and were therefore considered to be associated with de-cidualization, the outstanding vascular smooth muscle degen-eration preceding endovascular invasion was found to becorrelated with the presence of interstitial trophoblast aroundthe vessels (Fig. 3) [18]. In the latter paper it was also hypoth-esized that the early loss of smooth muscle coherence leads tovascular distension, which might provide a rheotropic triggerfor endovascular migration.

During the last decennia technical developments have sig-nificantly widened the research possibilities in this excitingfield. A major advance was the introduction of cytokeratinimmunohistochemistry [32e34], by which the trophoblasticnature of the endovascular and intramural cells could at lastbe definitely proven. In addition, the development of a plethoraof immunohistochemical tools for detecting various markermolecules, cell adhesion molecules, receptors, enzymes, cyto-kines and growth factors, allowed more precise studies of thecomplex interrelationships between trophoblast and maternaltissue components. Another important development was theimprovement of methods for maintaining trophoblastic cellsand placental tissue fragments in culture. With regards to spi-ral arteries, the latest development of arterial explant culturespromises to be another valuable tool for analysing the succes-sive steps in uteroplacental vascular remodelling and function[35]. At the clinical side, improvement of Doppler techniquesallowed studies of flow patterns in the uterus and placenta asa functional in vivo correlate to vascular remodelling in nor-mal as well as complicated pregnancies. Placental bed studiesmaking use of all these different technologies will be furtherdiscussed in the following sections of this review.

3. Trophoblast invasion and vascular remodelling:questions and controversies

3.1. Decidualization of spiral arteries andtrophoblast-independent changes

The aforementioned concept of decidual changes in spi-ral arteries as a necessary prelude to trophoblast invasion, at-tractive as it might be, still needs further substantiation. Inhow far early vascular decidualization is a conditio sine quanon for the later trophoblast-associated vascular remodellingis still not known.

Several years after the first vascular decidualization studies

centa 27 (2006) 939e958[18,29,31], the idea of trophoblast-independent changes of spi-ral arteries was revived by Craven et al. [36]. Unfortunately, in

their early presentations the changes observed in uterinedecidual tissue of women with ectopic pregnancies were referredto as physiologic changes [37], thereby infusing the scien-tific community with the idea that pregnancy-associated vas-cular remodelling of spiral arteries does not need trophoblastat all. As a response King and Loke [38] looked for physiolog-ical changes, as previously defined by Brosens [7], in a similarseries of uterine deciduas of ectopic pregnancies and, asexpected, refuted Cravens hypothesis. The importance of us-ing a precise terminology is the principal lesson of the Cravenincident. Indeed, terms such as physiological change, arte-rial conversion, vascular remodelling are probably too

specific terminology, for which we tentatively propose tro-phoblast-associated remodelling to replace the old term ofphysiological change, and decidua-associated remodellingfor referring to the early vascular changes which precede thetrophoblast-associated remodelling. However, since theearly remodelling partly depends on the presence of interstitialtrophoblast, an interstitial trophoblast-associated remodel-ling should be interposed, preceding the endovasculartrophoblast-associated remodelling (Fig. 3). Unfortunatelysuch rather wordy circumscriptions do not really offer apalatable terminology. Moreover, these terms may suggesta well-defined causal relationship which is by no means fully

Fig. 3. Diagram showing the different steps in uterine spiral artery remodelling, starting from the non-pregnant condition. The earliest stage in vascular remodelling

(stage 1) consists of endothelial vacuolation and some swelling in individual muscle cells. Invasion of stromal and perivascular tissues by interstitial trophoblast is

associated with further disorganization of the vascular smooth muscle layer (stage 2). Only then endovascular trophoblast appears (stage 3). Trophoblast becomes

embedded intramurally within a fibrinoid layer, which replaces the original vascular smooth muscle (stage 4). Finally re-endothelialization occurs, which may be

accompanied by the appearance of subintimal cushions containing a-actin immunopositive myointimal cells (stage 5).R. Pijnenborg et al. / Placgeneral and vague for designating the very specific changesin placental bed spiral arteries. There is a need for a more943enta 27 (2006) 939e958established. Another difficulty of these terms is that they stillrefer to a complex of morphological changes, some of which

care ill-defined such as medial disorganization. Identificationof molecular markers for the different steps of early vascularchanges would be helpful for working out better definitions,but so far no such molecule has been identified.

The concept of decidualization of myometrial spiral ar-teries, i.e. the idea that also trophoblast-independent vascularchanges in the myometrium are associated with the decidual-ization process, was recently taken up again by Brosens et al.[39], who considered its possible relevance to pregnancycomplications and even infertility. A cautionary note shouldbe added concerning the application of the term deciduali-zation to myometrial tissue because of the etymologicalroots of the term decidua (see Section 1). The term de-cidualization, as suggested from its first use, indeed refersto the irreversible shedding of tissue and cell death, whilethe decidual changes in the myometrium may bereversible.

3.2. Flow interruption by endovascular plugging

In our historical overview we mentioned the first consider-ations of haemodynamic consequences of endovascular plugs,but also how possible flow defects might be partially overcomeby the formation of multiple openings [16,17]. Because of theoblique course of spiral arteries with respect to the placentalfloor, especially in lateral parts of the placental bed, progres-sive decidual erosion by the basal plate may indeed generateadditional openings. In our previous studies we observedthat such openings into the intervillous space may appear half-way down a decidual segment of a spiral artery, with the moredistal portion eventually ending up within a necrotic mass ofdecidua (Fig. 4) [27,40]. In how far complete plugging ofthe most distal portions is related to this decidual necrosis isnot known.

Fig. 4. Diagram based on histological sections of a 12 week placental attach-

ment site, showing successive cross-sections of a decidual spiral artery, open-

ing into the intervillous space through a gap created halfway through the

decidual segment. The more distal portions of the artery are located within

944 R. Pijnenborg et al. / Plaa necrotic mass of decidua (reproduced from Ref. [40], with kind permission

of the editor of Trophoblast Research).The physiological importance of plugging was deducedfrom the ultrasonographic and hysteroscopic observations ofHustin and Schaaps [41], showing absence of maternal redcells in the intervillous space of early pregnancy. The earlierhistological observations on vascular plugging fitted nicelywith the newer evidence that early placental developmenttakes place in a low oxygen environment [42]. While thisidea was received with a healthy scepticism at the beginning[43e45], different lines of evidence e physiological, bio-chemical, in vitro studies e clearly support a low flow situa-tion in early pregnancy [46]. A revisit to the Boydcollection revealed the presence of channels in the cytotropho-blastic shell from 8 weeks onwards, whereas various degreesof plugging persisted during the following weeks, and onlyfrom the 14th week clear patency of spiral artery outlets intothe intervillous space was observed [47]. Since the extent ofplugging tends to be more complete in centrally located spiralarteries, the same investigators proposed that maternal flowstarts at the periphery rather than at the centre, and there ismorphological as well as ultrasound evidence to support thisview [48]. It is not known in how far the oblique course of spi-ral arteries and the formation of additional openings could berelated to the onset of flow at the periphery. In contrast, from17 weeks onwards Doppler observations on individual spiralarteries indicated higher flow in central than in peripheralspiral arteries, reflecting dissolution of the plugs and morecomplete spiral artery remodelling at the centre than at thesides [49].

Genbacev et al. [50] were the first to establish a relationshipbetween oxygen levels and trophoblast differentiation, in thesense that low oxygen favours proliferation, while high oxy-gen levels are required for differentiation into the invasivephenotype. The in vitro experiments by Caniggia et al. [51]pointed to the same direction, and were extended by the find-ing that the transcription factor HIF-1a plays a principal reg-ulatory role. Although providing a neat cellular model forexplaining trophoblast invasion there are, in our opinion, stillgaps in the overall story. Trophoblastic plugging of spiral ar-teries is considered to be the mechanism for creating a low ox-ygen environment [41], but then the question arises how thistrophoblast gets into this plugging position in the first place.Is endovascular plugging to be regarded as a real invasive phe-nomenon (which would require high oxygen according to theexperimentalists), or does it merely occur by filling up emptyspaces by proliferating trophoblast? During the 7e13 weeksperiod under low oxygen, extravillous trophoblast, whichwould include the endovascular plugs, never shows prolifera-tion [52]. The expression of invasion markers (integrins, met-alloproteinases) in endovascular plugs e if such markerswould persist in that situation e would be an indication foran invasive phenotype. In spite of extensive studies of suchmarkers in extravillous trophoblast, there is a remarkablelack of reported information concerning their expression in en-dovascular plugs. Only in the rhesus monkey immunopositivestaining for MMP-1 in the trophoblast of a plugged artery pro-

enta 27 (2006) 939e958vided unmistaken evidence for invasive potential [53]. Thiswould indicate that the endovascular trophoblasts have

acquired the invasive phenotype from the beginning, notwith-standing the presumably hypoxic condition of early pregnancy.To avoid a vicious circle, we might postulate that arterial plug-ging as a result of trophoblast invasion from the cytotropho-blastic shell, indeed occurs in a high oxygen environment inaccordance to Caniggias experimental findings. The shell isformed during the first 2e3 weeks, well preceding the 8weeks period of the earliest reported intrauterine oxygenmeasurements [42,46]. In this connection one may refer tothe observed hyperaemia of implantation sites in the human[54] as well as in laboratory rodents, where endothelial leak-age provides the earliest sign of implantation [55]. Blastocystimplantation may therefore be associated with (temporary)high oxygen, allowing quick invasion by trophoblast, followedby a period of relatively low oxygen because of the partialsealing of supplying arteries which would protect the develop-ing early embryo against oxidative stress.

It is interesting to consider the occurrence of vascular plug-ging in laboratory animals. The only rodent species in whichvascular plugging has been well documented is the goldenhamster [29]. In this case the plugging is effected by a gi-ant-cell type of trophoblast, preceding deeper endovascularmigration 1 or 2 days later. This is not the case in the mouse,where vascular invasion in this species seems to proceed viaa perivascular, rather than an endovascular, pathway [56e58]. In the rat, in which arterial invasion clearly follows an in-traluminal pathway, complete plugging of small side branchesof spiral arteries may occasionally be seen (unpublished obser-vations by Pijnenborg and Vercruysse). In none of these ani-mals endovascular plugging occurs to the same extent as isseen in the human, and possible events of oxidative stressmust therefore be met in a different way. Moreover, one shouldalso take into account that the chorioallantoic placenta is fullyfunctioning only relatively late in these three rodent species,and that early embryonic development depends mainly onthe yolk sac. The plugging of the spiral arteries in the hamsterfor instance occurs during this yolk-sac dependent period. In-teraction between invading trophoblast and maternal vascula-ture in these animals shows similarities with the human onlyduring the last third of their pregnancy when haemochorialplacentation is fully established. Vascular invasion then takesplace in the period after organogenesis, when the developingembryo is no longer so sensitive to oxidative stress.

3.3. The intravasation versus extravasation controversy

The relationship between interstitial and endovascular tro-phoblast has been a topic of debate and disagreement formany years. In a recent review Kaufmann et al. [59] discussedthe possible origin of endovascular trophoblast, favouringa process of intravasation, i.e. inward movement into thevessel of (interstitial) trophoblast from the outside. Althoughit is conceivable that intravasation indeed occurs in the super-ficial decidua close to the placenta, it can be questioned

R. Pijnenborg et al. / Placwhether the same applies to the deeper layers of the placentalbed, particularly to the myometrium.In early pregnancy spiral arteries become surrounded byinterstitial trophoblast, resulting in marked perivascular clus-tering in the superficial decidua [27,60,61]. Our own observa-tions were performed on 8e18 weeks complete hysterectomyspecimens processed in toto, allowing to evaluate gradientsof changes from decidua to myometrium. We found that vol-ume density of interstitial trophoblast diminishes with depth,while perivascular clusters of interstitial trophoblast seem tofan out away from the vessels in the deeper regions[11,60]. We think it highly probable that in the most superfi-cial decidua at least some of the endovascular trophoblastsare indeed derived from the numerous interstitial trophoblasticcells via intravasation. This is the more likely since in theupper decidual areas the arteries have a poorly defined muscu-lar wall lacking a clear elastica, in contrast to the deeper, es-pecially myometrial, tissue layers [62]. However, deeperdown in the myometrium increasing scarcity of perivascularinterstitial trophoblast may lower the opportunity for suchcells to invade a vessel wall, which at the same time has be-come more muscular compared to the decidua. Hence ourview that the intravascular presence of trophoblast in myome-trial spiral arteries is more likely the result of intraluminal mi-gration from the decidual segments after the disintegration ofthe endovascular plugs (Fig. 5).

Considering the intramural trophoblasts, their presence inthe vessel wall could theoretically be explained by different

Fig. 5. Diagram illustrating the interstitial invasion gradient from decidua to

myometrium, indicating their possible contribution to endovascular tropho-

blasts in spiral arteries. Deeper in the placental bed, interstitial trophoblast

are less concentrated around the spiral arteries, and therefore there is theoret-

945enta 27 (2006) 939e958ically a lower chance of perivascular interstitial trophoblast to penetrate the

spiral artery wall.

eintravasation or extravasation scenarios (Fig. 6). The simplestpossible invasion route would be either direct extravasation(Fig. 6A) or direct intravasation (Fig. 6B). The latter is defi-nitely too simplistic since it does not account for the presenceand subsequent disappearance of intraluminal trophoblast. Onecould speculate that the primary intravasation step consists ofa complete crossing of the vessel wall to reach an endovascu-lar position, followed by intramural incorporation intofibrinoid, possibly after intraluminal migration (Fig. 6C). Al-ternatively, intravasation of interstitial trophoblast may indeedresult in direct intramural incorporation into fibrinoid, but thelatter should then not be regarded as a permanent burial sitefor trophoblast. After being released into the lumen, the tro-phoblast may proceed its intraluminal migration until its finalincorporation into the wall of a deeper vascular segment(Fig. 6D). The theoretical possibility even exists that intramu-ral trophoblast may escape back into the interstitial tissue(Fig. 6E). It is virtually impossible to prove which of thesestories is correct using in vivo collected tissue samples, sincedynamic events e especially when a shuttling to and fro of

Fig. 6. Diagram of different possible scenarios of intramural invasion of a spiral

artery by intravasation or extravasation by interstitial (blue) or endovascular

(green) trophoblast. Arrows indicate the direction of cell migration. (A) Shows

direct extravasation of endovascular trophoblast into the vessel wall, while (B)

indicates direct intravasation of interstitial trophoblast into the wall. Intramural

trophoblast is indicated by its incorporation within fibrinoid (pink). (C) and

(D) indicate two different migratory intravasation pathways. (C) Shows the

theoretical possibility of direct migration of interstitial trophoblast through

the vessel wall, becoming temporarily endovascular and being subsequently

embedded into the vessel wall by fibrinoid deposition. (D) Shows the theoret-

ical possibility of interstitial trophoblast becoming intramural by fibrinoid de-

position, being subsequently released from the fibrinoid to become

endovascular followed by later re-embedding. (E) Shows the theoretical pos-

sibility of interstitial release of previously fibrinoid-embedded endovascular

trophoblast.

946 R. Pijnenborg et al. / Placcells would occur e are notoriously difficult to reconstructfrom static histological preparations. Explant cultures ofisolated spiral arteries cocultured with purified extravilloustrophoblast seem to indicate that trophoblast from outsidethe artery may cross the vessel wall and end up in the lumen[35], and it would be interesting to look for fibrinoid formationin such experimental conditions. In our opinion, a combinationof scenarios A and C of Fig. 6 provides the most likely story,but the impact of scenario C must depend on the local densityof interstitial trophoblast. Considering the correlation betweenthis density and the depth within the placental bed, we stillthink that endovascular e i.e. intraluminal e migration isthe principal route of vascular invasion, which accounts forobservations of continuous streaks of endovascular cellswithin myometrial spiral arteries showing almost no interstitialtrophoblast in their immediate surroundings [18].

Different stages of incorporation of endovascular tropho-blast into the vessel wall have been described by De Wolfand colleagues [26] and similar observations have been re-ported for the rhesus monkey [63]. Although in this species in-terstitial trophoblast is largely absent and trophoblast invasionis mainly endovascular, proceeding at the luminal side of thespiral arteries as far as the decidualemyometrial junction, in-tramural trophoblast embedded into fibrinoid also appears.Similarly, in human spiral artery explants dissected fromnon-placental bed areas and cocultured with purified tropho-blast which is placed into the vascular lumen, incorporationof trophoblast into the vessel wall was observed coincidingwith endothelial cell apoptosis [64]. In how far endovascularand perivascular (interstitial) trophoblast represent separatesubtypes contributing to intramural trophoblast, requires fur-ther investigation for which the expression of specific markerproteins may be helpful. The only exclusive endovascular tro-phoblast marker established so far is the adhesion moleculeNCAM [65]. A recent study indicated the presence of bothNCAM-positive and negative intramural trophoblast in the re-modelled spiral arterial walls, indicating a dual endovascularand interstitial origin [66], and this supports Ramseys earlierideas which were summarized in our historical overview[10,17].

3.4. Possible mechanisms of endovascular migration

Endovascular trophoblast migrates retrogradely within thelumina of spiral arteries, but little is known about mechanismsexplaining their movement against the blood stream. The factthat endovascular invasion only occurs within arteries andnever into veins, suggests a possible contribution of oxygencontent or haemodynamics. Although it is now well knownthat oxygen is important for differentiation and invasiveness[50,51], the possible role of gradients of oxygen levels in di-rected cell migration through the vascular environment hasbeen suggested but not proven.

The physical force of the blood stream has been consideredas a possible trigger to induce trophoblast migration [67].Cells may respond to flow intensities and directions by reor-ganizing their cytoskeleton and/or motile apparatus. Exposure

nta 27 (2006) 939e958of cells to fluid streams may lead to changes in microfilamentpolarity and alterations in membrane recycling, both processes

being involved in cell motility [68]. Such changes may be trig-gered by cell deformation and activation of a mechanotrans-duction pathway [69]. Especially microfilament polarizationchanges, as observed in other cell types exposed to flow, mightbe involved in a directed e rheotactic e motile behaviourof endovascular trophoblast. Recent work on isolated rhesusmonkey trophoblast grown on the inner surface of capillarytubes did show that shear stress, applied by increasing flowin the tubes, leads to trophoblast migration. One could argue,however, that the observed event was not really an active mi-gration, but rather a passive displacement in the direction ofthe fluid stream. Interestingly, a retrograde movement againstflow in this experimental setup was observed, but only whenthe trophoblast was cocultured with endothelial cells [70,71].The same authors also found that shear stress induces b1 integ-rin expression in the trophoblast, indicating an involvement ofcell adhesion molecules in flow associated migration.

The possible triggering effect of increasing blood flow ontrophoblast migration is supported by some observationsin vivo. In the golden hamster it was observed that intraplacen-tal maternal arterial channels, formed in continuity with spiralarteries, appear before the retrograde migration of endovascu-lar trophoblast into the corresponding arteries is started [29].The appearance of the intraplacental channels may be a resultof increased uterine blood flow, which is known to be understeroid (estrogen) control. Indeed, in ovariectomized pregnanthamsters treated with progesterone only, intraplacental arterialchannels did not form in association with all spiral arteries,and trophoblast invasion did not occur in vessels which didnot communicate with such channels [30].

It may also be significant that in the human deep endovas-cular migration into the myometrial spiral arteries occurs onlyafter the alleged onset of placental circulation, reported to oc-cur around 12 weeks [41,46,47]. The decidua- or interstitialtrophoblast-induced vascular remodelling is supposed to causeearly vascular distension accommodating higher flow, whichmay provide the necessary trigger for subsequent endovasculartrophoblast migration [18]. A similar function has been as-cribed to perivascular trophoblast in the guinea pig. It wasshown by immunohistochemistry that this trophoblast containsthe NOS enzyme, and it can be expected that nitric oxide hasa dilatory action on maternal blood vessels, preceding intra-vascular migration of trophoblast [72]. In the human, however,the presence of NOS in extravillous trophoblast is controver-sial, both eNOS and iNOS isoforms reported to be absent[73] and present [74,75] in interstitial trophoblast, includingthe perivascular cells along the blood vessels in the placentalbed. The investigations in the guinea pig have not been re-peated by other research groups.

3.5. Decidual and myometrial invasion: the two-wavehypothesis

Robertson et al. [76] hinted for the first time that in the hu-man, decidual and myometrial migration within spiral arteries

R. Pijnenborg et al. / Placis separated in time, the latter occurring between 12 and 16weeks. In the hysterectomy collection studied by Pijnenborget al. [18] the endovascular invasion of the myometrial spiral ar-teries beyond the superficial myometrium started at 14 weeks,indicating at least a 4 week difference between their appearancein decidual spiral arteries and deeper myometrial arteries, re-spectively (Fig. 7). The separate wave concept had partlybeen inspired by studies on the golden hamster, where two sep-arate endovascular migration waves occur into two differentvascular systems, the first into the circumferential arteries whichare associated with the yolk-sac placenta, and the second, laterin pregnancy, into the central spiral arteries related to the chorio-allantoic placenta [29]. This twowaves terminologywas thenapplied to the human, with the understanding that here they oc-cur within the same vascular system. It could never be estab-lished whether this so-called second wave was the extensionof invasion by endovascular cells that were lying dormantin decidual spiral arteries, or if there was really a new wave oftrophoblast leaving the cytotrophoblastic cell columns or theremnants of the cytotrophoblastic shell.

The two-wave hypothesis has been criticized lately by Rob-son and colleagues, following extensive observations in a largeseries of early pregnancy placental bed biopsies, collected bythe newly established punch biopsy technique under ultra-sound guidance [77,78]. Their results supported a progressiveendovascular migration from decidual into myometrial arteriesrather than showing a distinct second wave. A difficulty in bi-opsy studies is of course that the depth of invasion is difficultto evaluate because of the random orientation of biopsy spec-imens. The original confirmation of the two-wave idea camefrom studies on complete hysterectomy specimens sectionedin toto allowing easy evaluation of the depth of endovascu-lar invasion.

Lyall [79] rightly pointed out the difficulty to ascribea physiological significance to the phenomenon and to con-ceive cellular and molecular mechanisms controlling such a bi-phasic invasive process. This difficulty might eventually beresolved when the haemodynamical factor is brought into con-sideration. It could be conceived that, as uterine blood flowprogressively increases [80], a threshold value may be reachedaround the 12 weeks period, allowing more substantial utero-placental flow, facilitated by dilatation of myometrial arterieswhich have undergone interstitial trophoblast-associatedremodelling, which is accompanied by rising oxygen[41,46,47]. Theoretically, the crossing of this haemodynamicalthreshold may trigger a second wave of rheotactic migra-tion of endovascular trophoblast, possibly via activation ofa still undefined mechanotransduction signalling pathway[69]. For a complete understanding of the actual cellularmechanisms involved, more insight will be needed in the bio-mechanical properties of cells in general [81] and of tropho-blast in particular.

3.6. Intramural incorporation and the fibrinoid question

Following their migration within the arterial lumina thetrophoblastic cells ultimately invade the vessel wall. During

947enta 27 (2006) 939e958the last two decennia cellular mechanisms of trophoblastinvasion have been intensively investigated, initially focussing

on mechanisms of attachment and extracellular matrix break-down [82], but this research quickly expanded to include reg-ulatory factors such as cytokines, growth factors and oxygen[50,51,83e86]. Most studies of invasion mechanisms weremainly applied to the interstitial invasion of stromal tissues,but it is very likely that similar mechanisms contribute toendovascular trophoblast incorporation into the vessel walls.Indeed intraluminal trophoblast has been shown to containmetalloproteinases in both the human [87] and the rhesusmonkey [53]. Such matrix-degrading enzymes may be func-tionally important for the penetration of the subendothelialbasement membranes and other (maternal) matrix compo-nents. It is still unclear in how far the endothelial layer ispenetrated by trophoblast or undergoes spontaneous (?)regression. Although a recent report described apoptosis-induction in endothelial cells by cocultured extravillous tro-phoblast in explanted spiral artery segments [64], it shouldbe noted that this trophoblast may have been of an interstitial(sub)type, not representative for endovascular trophoblast.

A possible contributory factor to the burial of trophoblastinto the vessel wall is the fibrinoid which is characteristicallydeposited in physiologically changed spiral arteries (Figs. 3and 6). In their electronmicroscopical study of the vascularincorporation process De Wolf et al. [26] noted the precipita-tion of fibrinoid material around the endovascular trophoblast,

precipitation of PAS-positive fibrinoid material can be ob-served in between intraluminal cells prior to their intramuralincorporation [88]. Sustained release of matrix material in thevessel wall may be responsible for the spidery shape of in-tramural trophoblast. For a long time we assumed that fibri-noid deposition was a hallmark of endovascular trophoblast,but Bulmer (personal communication, reported in Ref. [89])pointed out that also interstitial trophoblast occasionallyshows fibrinoid deposition and thereby acquires a spideryphenotype.

In spite of extensive studies of the production of matrixproteins by trophoblast in vivo as well as in vitro there are,so far, very few data about the exact nature of the fibrinoidwithin the converted spiral arteries. The matrix-type fibri-noid of the basal plate described by Frank et al. [90,91]may be a good candidate for the intramural fibrinoid of spiralarteries, but their and most other published observations ontrophoblast-associated extracellular matrix mainly relate tointerstitial trophoblast. The fact that interstitial and endovas-cular fibrinoids may be different, was highlighted by immu-nolocalization studies of fibronectin in term placental bedbiopsies, showing extensive immunopositive deposits sur-rounding interstitial trophoblast in contrast to the absenceof staining in the vascular fibrinoid with embedded intramuraltrophoblast [92]. Also laminin was absent from spiral artery

Fig. 7. Diagram illustrating the two wave concept of endovascular trophoblast migration into the decidual segments of spiral arteries in the first trimester (left) and

into the myometrial segments in the early second trimester, i.e. from 14 weeks onwards (right). Red arrow: direction of blood flow; black arrow: direction of

endovascular trophoblast migration.948 R. Pijnenborg et al. / Plawhich they thought to be secretory products of this tropho-blast [19]. Also at the light microscopical level intercellularcenta 27 (2006) 939e958fibrinoid at term, but present around interstitial trophoblasticgiant cells (Pijnenborg and Vercruysse, unpublished results).

These staining patterns contrast with reported data on rhesusmacaque spiral arteries, which do show pericellular lamininand also collagen IV deposits associated with endovasculartrophoblast [93]. Also in the baboon laminin, but not fibro-nectin, has been shown to be a component of the spiral arteryfibrinoid, while both proteins are present in the fibrinoid ma-trix of the cytotrophoblastic shell [94]. Further work on spiralartery fibrinoid is clearly necessary, the more since it alsomay highlight differences in interstitial versus intramuraltrophoblast.

3.7. The endothelial mimicry controversy and maternalvascular repair

The endothelial mimicry hypothesis logically followedfrom the ill-founded idea that endovascular invasion resultsin permanent replacement of the endothelial lining by tropho-blast. Indeed, in such context it would not be illogical to con-sider a possible acquisition of endothelial function togetherwith expression of endothelium-related molecules by cells re-placing the original endothelium. This concept was launchedfor the first time by Zhou et al. [95,96], and especially theirreport on the alleged absence of endothelial molecules inpre-eclamptic trophoblast drew a lot of attention. The endothe-lial molecules investigated were PECAM-1 (CD31), VE-cad-herin and VCAM-1. As far as PECAM-1 is concerned, itsexpression by endovascular trophoblast could not be con-firmed in late [97] (Fig. 8) and early pregnancy [98]. LaterFACS analysis studies, however, confirmed a high-levelexpression of VE-cadherin in isolated trophoblast from earlypregnancies [99]. By applying blocking antibodies in tropho-blasteendothelium coculture experiments, the latter investiga-tors obtained evidence that VE-cadherin, and in a lesser degree

Fig. 8. Cytokeratin (red)/CD31 (blue) double staining of a spiral artery with

physiological change, illustrating endothelial repair (End) after intramural in-

corporation of trophoblast (IMT). The insert is a high power picture showing

R. Pijnenborg et al. / Plaa trophoblastic cell very close to the lumen, covered by a very thin layer of

endothelium.PECAM-1, could play a role in the transendothelial migrationof trophoblast, at least within this in vitro system.

In the discussion of their paper Zhou et al. [96] took theirobservation of trophoblastic expression of endothelial mole-cules as an indication that trophoblast started to behave asan endothelium for the remainder of the pregnancy. Thisidea opposes the finding that after mural incorporation of en-dovascular trophoblast, fresh maternal endothelium overgrowsthe invaded vessel walls (Fig. 3, stage 5). This was clearlydemonstrated by applying double immunostaining for cytoker-atin and PECAM-1 on third-trimester placental bed biopsies.In all cases the inner lining of the spiral arteries was PE-CAM-1 positive but cytokeratin negative, positive cytokeratinimmunostaining being restricted to the mural trophoblast(Fig. 8), which refutes Zhous hypothesis unless e applyinga reasoning ad absurdum e the endovascular trophoblastnot only acquires endothelial markers but also loses its tropho-blastic marker molecules including cytokeratin! Further evi-dence of a maternal repair process after trophoblast invasionof the spiral arteries is provided by the frequent observationof subintimal thickening, resulting in cushions of connectivetissue with myointimal cells interposed between the restoredendothelium and the fibrinoid layer surrounding the mural tro-phoblast. Especially observations on partially invaded spiralarteries illustrate that such intimal thickenings typically over-lie streaks of intramural fibrinoid with embedded trophoblast(Fig. 9). The a-actin-immunopositive myointimal cells mayprobably act as precursor cells for medial smooth muscle re-pair after pregnancy.

3.8. The role of uterine NK cells in spiral arteryremodelling

A characteristic feature of first trimester decidua is the pres-ence of granulated leukocytic cells, referred to in the past asendometrial granulocytes or large granulated lymphocytes(LGLs), now known to be a specific uterine subpopulationof CD56 immunopositive natural killer cells. Since increasednumbers of uterine NK cells are found in association with ex-travillous trophoblast in the placental bed [61], a role in regu-lating trophoblast invasion had been suspected for many years.These morphological findings have been increasingly sup-ported by studies of specific molecular interactions betweenuNK receptors and the HLA-antigens on extravillous tropho-blast [100]. However, a possible relationship between uNKcells and vascular trophoblast invasion is less straightforward.A perivascular concentration of uNK cells has been reported[101], but a possible involvement with spiral artery remodel-ling e directly or indirectly via interactions between uNKcells and invading endovascular or perivascular (interstitial)trophoblast e has not yet been established [102].

Turning to experimental animals there is good evidence, es-pecially in mice, for a direct involvement of uNK cells withvascular remodelling, as it was shown that the spiral arterysmooth muscle layer remains intact in NK-deficient mice, in

949centa 27 (2006) 939e958contrast to normal control animals [103,104]. This remodel-ling occurs in the decidua as well as in the mesometrial

triangle [105], formerly called the metrial gland. Because ofthe heavy infiltration by uNK cells throughout the whole trian-gle tissue in the mouse, this area has recently been referred toas the mesometrial lymphoid aggregate of pregnancy [106].Differently from the mouse, in the golden hamster [29] andthe rat [107] uNK cells show a striking distribution as out-standing perivascular cuffs around spiral arteries. However,the mesometrial triangle area in these two species is also in-vaded by endovascular and interstitial trophoblast [29,107].It is obvious that the relationship between uNK cells and tro-phoblast invasion in these species needs clarification. A possi-bly relevant observation was made by Ain et al. [108], whofound a reciprocal relationship between the presence of uterineNK cells surrounding the spiral arteries and the perivascularinvasion by interstitial trophoblast in normal rat pregnancy.

Two possible actions by uNK cells are emerging from theseanimal data: (1) a direct action on vascular smooth muscle re-modelling, which is probably related to the decidualization

human, one should realize that in our own species trophoblastinvasion is extended deeply into the inner myometrium. Al-though the possibility of non-trophoblast related changes ofspiral arteries in this compartment has been considered [39],uNK cells occur in substantial numbers only in the deciduabut not in the inner myometrium [28], while in the aforemen-tioned rodent species substantial infiltration of the mesome-trial triangle by uNK cells occurs during a process clearlyrelated to decidualization [105]. It must be kept in mind thatthe presence of uNK cells, in human as well as in rodents, isnot limited to areas of trophoblast invasion. Indeed, in themouse the mesometrial triangle is not invaded by trophoblast[58]. Also the fact that deciduomata, induced in pseudopreg-nant or unilaterally pregnant animals, are infiltrated by highnumbers of uNK cells in the absence of trophoblast, is a cleardemonstration of their association with the decidualizationprocess itself. It should be added that in the human uNK cellsalso make their appearance in the decidua vera, albeit in lesser

Fig. 9. Partial physiological change in parallel sections stained (A) for cytokeratin/HE, (B) with PAS to show the fibrinoid, (C) for a-actin and (D) with acid orcein

to show the elastica. (A) Shows intramural trophoblast (IMT) partially covered by an intimal cushion (Int). (B) Illustrates intramural trophoblast (IMT) incorpo-

ration in fibrinoid (Fib). (C) Shows a-actin-immunopositive vascular smooth muscle (VSM) at the noninvaded part of the vessel, as well as a-actin-immunopositive

myointimal cells in the intima (Int) overlying the unstained fibrinoid (Fib). In (D) the elastica (El) is obvious in the noninvaded parts of the vessel, ending abruptly

at the beginning of the fibrinoid zone (compare with B and C).950 R. Pijnenborg et al. / Plaprocess, and (2) a possible controlling action on trophoblastinvasion. While trying to extrapolate these concepts to thecenta 27 (2006) 939e958numbers than in the placental bed [61]. Further research is ob-viously needed to establish any firm relationship between the

cperivascular uterine NK cells, interstitial and/or endovasculartrophoblast invasion and vascular remodelling in human andlaboratory rodents.

4. Defects in spiral artery remodelling in pregnancycomplications

As mentioned before, it was the demonstration of impairedmaternal placental perfusion in hypertensive pregnancies [23]that triggered the search for vasculopathies in the placentalbed as possible causes for placental insufficiency and fetalgrowth restriction [24]. In 1972, Brosens et al. [25] reportedfor the first time that, in contrast to normal pregnancies, spiralarteries with normal histological arterial features, i.e. with-out physiological change, could be easily found in the my-ometrial part of the placental bed of pre-eclamptic women.According to these authors decidual spiral arteries showednormal physiological changes, but a later study reported theregular absence of such changes in this area as well [109].Since in their previous study Brosens et al. had hypothesizedthat the intramural cells of physiologically changed spiral ar-teries are derived from invasive trophoblast [7], it was logicalto propose that the restricted physiological change in pre-eclampsia would be the consequence of defective trophoblastinvasion. An important repercussion of this defect wouldthen be that noninvaded and thus unmodified vessels retaina narrow lumen and an intact muscular wall, which thereforemight hinder a normal blood supply to the placenta. Moreover,some of the noninvaded vessels may develop acute atherosis,a lesion characterized by necrosis and lipophage infiltrationof the vessel wall [110], and it was shown that such vesselscorrespond to infarcted areas in the placenta [111].

The initial observations on impaired vascular conversion inpre-eclampsia have subsequently been repeatedly confirmed[112e120]. But while originally the absence of physiologicalchanges in myometrial spiral arteries was regarded as a patho-gnomonic feature of pre-eclampsia, it has since then beenrecognized that defective arterial remodelling also occurs innon-hypertensive cases of intrauterine growth restriction,non-proteinuric gestational hypertension and chronic hyper-tension, and exceptionally even in an entirely normal preg-nancy. The original black-and-white picture had therefore tobe replaced by the concept of a spectrum of aberrations occur-ring in different categories of pregnancy complications. Never-theless, going back to the first reports of impaired placentalperfusion in hypertensive pregnancies [23], the original aimof the first generation of placental bed investigators has beenamply fulfilled. With increasing sophistication of ultrasoundtechnologies, correlation studies between flow measurementsand placental bed features have provided additional evidencefor a direct link between spiral artery histopathology and pla-cental perfusion [121e127]. Also Doppler observations onindividual spiral arteries have confirmed this picture [49,128].

As to the detailedmechanisms of invasion defects and/or fail-ures in vascular remodelling, all the aforementioned facts and

R. Pijnenborg et al. / Placontroversies should be taken into consideration. Unfortunatelyin most of the cases we can only speculate on the causes ofplacental bed defects. Because the basic placentation defectsnecessarily occur in early pregnancy, even if early placentalbed samples from abortion cases were routinely available, itwould not be possible to relate any aberrations in invasion or ar-terial change to the (unknown) pregnancy outcome.

4.1. Possible decidualization defects

It was suggested recently that the decidualization process,and especially the extension of decidua-associated changesto the inner myometrium, might be deficient in women des-tined to develop pre-eclampsia during their pregnancy [39].Both decidua-associated and interstitial trophoblast-associatedvascular remodelling have been considered as essential pri-mary steps for allowing subsequent endovascular invasion[18], and absence of the former would therefore necessarilylead to inhibition of the latter. Such defects should involve par-ticularly the myometrial segments of the spiral arteries, sinceit is at this level that the major placental bed aberrations occur.In subfertility cases with possible decidualization defects,where pregnancy is eventually established after IVF treatment,the incidence of pregnancy complications such as pre-eclamp-sia would then be expected to increase and this seems to besupported by the facts [129]. Unfortunately there are fewhard data to support the hypothesis of impaired decidualiza-tion in subfertility cases.

Another observation negating a possible link between re-stricted trophoblast invasion and impaired decidualizationwas made in the Dixon collection of complete hysterectomyspecimens. One case out of eight post-15 weeks gestationalage showed absence of endovascular trophoblast in myome-trial segments of spiral arteries, at a period of time when inva-sion should have extended deep in the myometrium. In thiscase there was no indication of a poor decidualization ofthe spiral artery walls [18]. However, as mentioned before,a proper definition of vascular decidualization remains prob-lematic without the availability of suitable molecular markers.Vascular smooth muscle cells in myometrial spiral arteriespossess progesterone receptors, a feature shared by decidualtissue [130], but it is not known whether there are expressionchanges in subfertility cases or in pre-eclampsia.

4.2. Possible plugging or de-plugging defects

Plugging of arterial outlets at the basal plate is a characteristicof early pregnancy, and dissolution of the plugs, presumably as-sociated with the disappearance of intercellular junctions, couldbe considered as necessary for the beginning of invasion. Theo-retically, impaired dissociation of the plugs could be the cause ofdefective endovascular invasion, but there is no direct evidenceto support this view. However, Khong et al. [109,131] made thecurious observation that a proportion of third-trimester basalplate specimens obtained fromwomenwith pre-eclampsia or in-trauterine growth restriction showed intraluminal trophoblast inthe basal plate spiral artery segments. They considered this as

951enta 27 (2006) 939e958a result of a compensatory new wave of endovascular tropho-blast invasion, as a teleological response to convert

inadequately remodelled spiral arteries [132]. Alternatively, andprobably more realistic, such third-trimester endovascular tro-phoblasts in the outlets of the spiral arteries may represent rem-nants of incompletely dissolved plugs of early pregnancy. Itwould beworthwhile to investigate such third-trimester residualplugs for the presence of intercellular junctions. Once againa possible relationship may be considered between the begin-ning of endovascular migration after dissolution of the plugsand haemodynamics. It cannot be excluded that an inadequaterise in early uterine flow may fail to provide a timely triggerfor de-plugging, migration and subsequent intramural invasion.Restriction of the normal increase in blood flow must lead topersisting low oxygen and elevated HIF-1a expression, thuspreventing the trophoblast to differentiate to the invasive pheno-type [51]. Then, of course, the final question to be asked is bywhat mechanism the normal flow increase might be impairedin pre-eclamptic women. Both systemic and local defects (inad-equate early vascular remodelling and/or defective decidualiza-tion) should be considered.

4.3. The extravasationeintravasation controversyrevisited

In their studies of the placental bed in pre-eclampsia, Ger-retsen et al. [113] published striking pictures of spiral arterieswithout physiological change surrounded by dense clusters ofmultinucleated giant cells. Such giant cells had been consid-ered as the result of fusion of interstitially invading tropho-blast, and it had always been assumed that giant cells,representing the final stage of trophoblast differentiation,show diminished invasive activity [28]. Hence Gerretsen rea-soned that impaired physiological change in pre-eclampticpregnancies may result from the failure of trophoblast to reachthese vessels because of precocious terminal differentiationinto giant cells. It is clear that such scenario would be inline with the intravasation model of spiral artery invasion. Al-though perivascular clustering of giant cells has certainly beenobserved by others including ourselves [92], it is by no meansa frequent occurrence [118]. Also Reisters observations of in-creased apoptosis in perivascular trophoblast in associationwith increased macrophage numbers in the placental bed ofpre-eclamptic patients [133,134], could be fitted into the intra-vasation model, thus explaining the failure of trophoblast tocolonize the spiral arteries.

On the other hand, it is not yet fully clear in how farinterstitial trophoblast invasion is indeed impaired in pre-eclampsia. The first attempts to quantify interstitial tropho-blast volume densities failed to show a difference betweennormal and pre-eclamptic pregnancies [97]. Using a moresophisticated counting system, Naicker et al. [135] did findsignificantly lower interstitial trophoblast invasion in pre-eclampsia. Kadyrov et al. [136] also found lower numbersof interstitial trophoblasts in pre-eclampsia and showed thatthis was not associated with an increased apoptosis rate. Infact the apoptosis rate was lowered in the overall interstitial

952 R. Pijnenborg et al. / Placetrophoblast population of these patients, while Reister hadshown that the apoptosis rate was increased in theperivascular subpopulation of interstitial trophoblast [133].An intriguing finding was that in anaemic women the tropho-blast invasion occurred in higher numbers and reached deeperlayers [136].

Whether impaired perivascular invasion of interstitial tro-phoblast can be the whole story behind restricted physiologicalremodelling must be questioned. The previously discussed en-dovascular migration from the decidua, followed by extravasa-tion, may well be the principal pathway of invasion ofmyometrial spiral arteries. In addition, however, one mightconsider that the initial weakening of vascular smooth muscleby perivascular trophoblast could be impaired in women sub-sequently developing pre-eclampsia, possibly by macrophageattack on interstitial trophoblast or by premature giant cell for-mation. At the moment we have not enough information on thenormal development to decide on the relative impact of possi-ble defects in either of the two invasive pathways in causingdefective remodelling of myometrial spiral arteries.

4.4. Defective endothelial mimicry and exaggeratedmaternal repair responses

After presenting evidence for the acquisition of endothelialmarkers by endovascular trophoblast, Zhou et al. [96] an-nounced that this endothelial mimicry was absent or occurredat a much lower level in pre-eclampsia. The authors postulatedthat the adhesion molecules concerned are important for form-ing vascular connections with uterine vessels, and that down-regulation of these molecules would be a mechanism forimpaired invasion in pre-eclampsia. In view of the controver-sies concerning the trophoblastic expression of endothelialmolecules, this idea e attractive as it might be e should betreated with caution. Furthermore, quantification of immuno-histochemical e in their case immunofluorescent e stainingis notoriously difficult, certainly when it comes to evaluatingstaining intensities. In order to fully substantiate this ideaone should not rely on immunofluorescence only.

Since a maternal repair process obviously takes place in theinvaded spiral arteries, not only involving endothelial restora-tion (Fig. 8) but occasionally also intimal thickening (Fig. 9),one might wonder whether in pre-eclampsia an exaggeratedmacrophage-mediated tissue repair response could lead toelimination of the trophoblast which had earlier become intra-mural. Some investigators would almost certainly be temptedto consider the apparent narrowing of the vessel lumen, whichis sometimes associated with intimal thickening, as patholog-ical. However, since such situation frequently occurs in normalpregnancies as well, there is no indication that such an exag-gerated maternal repair in invaded (decidual!) vessels occursmore frequently in pre-eclamptic cases.

4.5. The relevance of acute atherosis

In the search for vasculopathies in spiral arteries responsi-ble for maternal underperfusion of the placenta, pathologists

nta 27 (2006) 939e958were particularly intrigued by the typical lesion of acute athe-rosis, characterized by vascular necrosis, foam cell infiltration,

cand perivascular leukocytic infiltrations. Although this lesionhad already been described in decidual fragments attached todelivered toxemic placentae [137], Robertson et al. [110]recognized it as a real feature of the placental bed in pre-eclamptic pregnancies which may occur in myometrial aswell as decidual spiral arteries. Especially fascinating wasthe morphological similarity to vascular lesions in rejectedkidney transplants, thereby adding fuel to the then increasinginterest in the immunological problem of pregnancy, which in-cluded the idea of pre-eclampsia as a possible autoimmunedisease. Labarrere [138] subsequently reviewed the histopath-ological features of acute atherosis which might support animmunological involvement, including the deposition of IgMand components of the complement system and of coursethe perivascular mononuclear infiltrations. However, althoughBrosens and Renaer [111] demonstrated a relationship be-tween the lesion and underlying placental infarcts, Khonget al. [139] did not find a worse outcome in pre-eclamptic pa-tients who showed this lesion. This may relate to the fact thatatherosis only occurs in a minor proportion of spiral arteries[140] and therefore may not have a major impact on the totaluteroplacental blood flow.

Since the foam cells had been identified as CD68 immuno-positive macrophages [119], which also contain TNF-a [141],these findings may well be linked with the recent concept thatpregnancy e and a fortiori pre-eclampsia e represents a hy-perinflammatory state [142]. More than 30% of the leuko-cytic cells in the decidua of normal pregnancy aremacrophages. Since the uprising of cytokine research, it be-came clear that macrophages at the placental site are impor-tant cytokine producers and may be pivotal regulatorycells for controlling trophoblast invasion [143]. Histologicalstudies have shown that macrophages form cell extensionswhich may come in close contact with extravillous tropho-blast [144]. Macrophages may have cytolytic effects, presum-ably by the production of cytotoxic cytokines, andoveractivation may possibly be related to placentation defectsand pregnancy complications [145]. In pre-eclampsia, highnumbers of macrophages have been seen to encircle spiral ar-teries with few invaded trophoblasts, while the opposite oc-curs in normal pregnancies [134]. Macrophage-derived foamcells, normally characteristic for atherosis, have been occa-sionally described in physiologically changed spiral arteriescontaining mural trophoblast [146,118], but the real lesionis typically seen in spiral arteries without physiologicalchange. Nevertheless, cytokeratin-immunopositive remnantshave been observed occasionally in atherotic vessel walls[118,119]. Other indications of possible trophoblast-associ-ated vascular remodelling are no longer visible in the necroticarteries. It is, however, a sobering fact that this lesion is alsoreadily found in the decidua vera, a region where trophoblastis absent [76]. It can therefore not be excluded that, as aconsequence to the hyperinflammatory condition of pre-eclampsia, activated macrophages may directly damagedecidual spiral arteries in both decidua basalis and decidua

R. Pijnenborg et al. / Plavera e in serious cases also in the inner myometrium ewhich may or may not interfere with trophoblast invasion.4.6. Uterine NK cells in complicated pregnancies

While uterine NK cells have been postulated to play a directrole in vascular remodelling [103,104], as well as a regulatoryrole in trophoblast invasion [100,102], little information isavailable about possible aberrations in complicated pregnan-cies. Croy et al. [147] reasoned that, because of an associationbetween uterine NK cells and vascular smooth muscle destabi-lization in mice, uterine NK cell deficiency might equally leadto impaired early remodelling of spiral arteries in women.However, the few available reports on pregnancy complica-tions rather suggest an increase in uterine NK cell numbers.Clifford et al. [148] reported an increase of uterine NK cellsin the decidua of women with recurrent miscarriage. In pre-eclampsia, Stallmach et al. [149] found a significant increasein pre-eclamptic women with IUGR, but not in women withpre-eclampsia only. Reister et al. [150], however, did find a sig-nificant increase of CD56 cell numbers in pre-eclampticwomen, particularly near spiral arteries. So far, the latter re-port is the only one that relates to altered perivascular distribu-tion of uNK cells in complicated pregnancies, but thesefindings need to be substantiated. It is clear that the possiblerelationship between increased uNK cell infiltration and defec-tive trophoblast invasion or disturbed vascular remodellingneeds further investigation. The results of such studies mayalso help to clarify the role of uNK cells in normal pregnancy.

5. Epilogue e reconstitution of the arterial wall afterparturition

Because of the drastic changes in spiral artery structure dur-ing pregnancy, one might wonder how the original vessel wallis restored after parturition. It is generally thought that tropho-blast persists in the uterine wall for a week to 10 days postpar-tum [151], but in order to have a complete picture of thevascular reconstitution, a day-to-day follow-up study of theformer placental bed site would be needed. Unfortunately,the only materials available are curettage and hysterectomyspecimens collected in cases of postpartum haemorrhage. Insuch material one may observe a mixture of involutedand subinvoluted vessels, the latter probably being thecause of the haemorrhage. While in normal involution the for-mer uteroplacental arteries show ghost-like hyalinized wallswith collapsed lumina which are lined by endothelium, subin-volution shows widely dilated vessels with thrombus forma-tion, remnants of trophoblast and absence of an endotheliallining [151,152]. There is a hint that in subinvolution sites ar-teries may still contain endovascular trophoblast in their lu-mens indicating a failure of the re-endothelialization whichshould have taken place during the third trimester of preg-nancy [109].

Since during pregnancy the vascular smooth muscle is re-placed by mural trophoblast and fibrinoid, reconstitution ofthe tunica media must take place. Probable precursor cellsare the a-actin-immunopositive myointimal cells which are

953enta 27 (2006) 939e958present in thickened intimae (Fig. 9c). It should be notedthat there is never a complete repair of the original arterial

structure. Indeed, spiral arteries of multiparous uteri show par-tial replacement of their smooth muscle layer by connectivetissue, together with fragmentation and duplication of the elas-tica. Such permanent changes are thought to be related to thehigher birth weights with increasing parity [153].

6. Conclusion

From this overview it will be clear that there are still a lotof controversies and uncertainties about the exact pathwaysand mechanisms of spiral artery remodelling, as well as thespecific effects of endovascular and/or interstitial trophoblast.As a result, plenty of placental bed stories have been toldby different investigators, including ourselves. Althougha lot of experimental research has already been performedon mechanisms of trophoblast invasion, most findings relateto invasive behaviour in general, and thus primarily to intersti-tial invasion, while observations on the specific invasiondirected to spiral arteries are often restricted to a shortparagraph. A major problem is the difficulty of designingexperimental work on isolated placental bed blood vessels.The recently developed in vitro models using explanted spiralartery segments [35] may be helpful in the near future, butone must realize that we are still far from in vitro modellingof vascular function, which necessarily should includehaemodynamics.

One firm conclusion to which everybody will agree is thatvascular remodelling is a vital characteristic of pregnancy, andthat trophoblast plays an important role in the complete phys-iological change of the spiral arteries. However, vascular re-modelling comprises a complex sequence of events, ofwhich in the first place the earliest, decidua-associated alter-ations, should be more precisely defined. This is vital for mak-ing a clear distinction between decidua-induced andtrophoblast-induced remodelling. The relationship of theseearly changes with increasing blood flow and endovascularversus interstitial trophoblast invasion should be clarified,and in particular the role of oxygen in inducing invasive be-haviour specifically with regards to the blood vessels has tobe better understood. In this respect an agreement should beobtained on the expression of specific adhesion molecules byendovascular trophoblast. These basic features need to betaken as a background for understanding defects in invasivebehaviour and failed vascular remodelling in complicatedpregnancies such as pre-eclampsia. In the first place we wouldlike to know in how far defective invasion in pre-eclampsia isto be regarded as an intrinsic defect of trophoblast, or whetherextrinsic factors are responsible for such failures. The lattermay be related to the inflammation theory of pre-eclampsia[142], and in this light the interference of macrophages anduterine NK cells with trophoblast invasion and vascular re-modelling should be further analyzed.

The older generation researchers frequently referred to an-imal studies, in order to strengthen their confidence in the re-ality of the amazing migratory behaviour of trophoblast within

954 R. Pijnenborg et al. / Placblood vessels. When arterial invasion was increasingly ac-cepted as a fact in the human, animal observations tended tobe less frequently quoted by clinically oriented researchers.Nevertheless, we believe that further use of experimental ani-mals will be an enormous help in delineating possible mecha-nisms of invasion and vascular remodelling that may equallyapply to the human. Of course we should be aware of seriousspecies diversities which doubtlessly exist in primate [154]but also in rodent placentation. Spiral artery remodelling inmice and rats has recently been studied more closely[56,57,104,107,155], and there is also renewed interest indeveloping animal models for pregnancy hypertension [156e159]. So far no therapy exists which may cure defective tro-phoblast invasion and vascular remodelling in pre-eclampticwomen, but if some relevant molecules might be discoveredit would be most suitable to test them in such animal models.At present it is impossible to predict whether in a near futuresuch strategy may become reality. Time e and a lot more re-search e will tell.

Acknowledgement

The authors wish to thank Tom Pijnenborg for his adviceand practical help with the illustrations.

References

[1] Dunn PM. Dr William Hunter (1718e83) and the gravid uterus. Arch

Dis Child Fetal Neonatal Ed 1999;80:F76e7.[2] Moore W. The knife man: the extraordinary life and times of John

Hunter, father of modern surgery. London: Bantam Press; 2005. 482 pp.

[3] Hunter W. Anatomia uteri humani gravidi tabulis illustrata e the anat-omy of the human gravid uterus exhibited in figures. Facsimile edition

of the original edition (Birmingham: John Baskerville 1774). Birming-

ham, Alabama: Gryphon Editions; 1980.

[4] Reynolds SRM. Physiology of the uterus. 2nd ed. New York: Paul B.

Hoeber; 1949. 611 pp.

[5] Lundgren N. Studies on the vasculature of the corpus of the human

uterus. Acta Obstet Gynecol Scand 1957;36(Suppl. 4):1e115.

[6] Ramsey EM. The story of the spiral arteries. J Reprod Med

1981;26:393e9.

[7] Brosens I, Robertson WB, Dixon HG. The physiological response of the

vessels of the placental bed to normal pregnancy. J Pathol Bacteriol

1967;93:569e79.[8] Ramsey EM, Donner MW. Placental vasculature and circulation. Stutt-

gart: Georg Thieme Publishers; 1980. 101 pp.

[9] Boyd JD, Hamilton WJ. The human placenta. Cambridge: W. Heffer &

Sons; 1970. 365 pp.

[10] Ramsey EM, Harris JWS. Comparison of uteroplacental vasculature and

circulation in the rhesus monkey and man. Contrib Embryol Carnegie

Inst Wash 1966;38:59e70.[11] Pijnenborg R, Robertson WB, Brosens I, Dixon G. Trophoblast invasion

and the establishment of haemochorial placentation in man and labora-

tory animals. Placenta 1981;2:71e92.

[12] Grosser O. Fruhentwicklung, Eihautbildung und Placentation des Men-

schen und der Saugetiere. Munchen: J.F. Bergmann; 1927. 454 pp.

[13] Boyd JD, Hamilton WJ. Cells in the spiral arteries of the pregnant

uterus. J Anat 1956;90:595.

[14] Boyd JD, Hamilton WJ. Development and structure of the human pla-

centa from the end of the 3rd month of gestation. J Obstet Gynaecol Br

Commonw 1967;74:161e226.

enta 27 (2006) 939e958[15] Hamilton WJ, Boyd JD. Development of the human placenta in the first

three months of gestation. J Anat 1960;94:297e328.

[16] Hamilton WJ, Boyd JD. Trophoblast in human utero-placental arteries.

Nature 1966;212:906e8.

[17] Harris JWS, Ramsey EM. The morphology of human uteroplacental

vasculature. Contrib Embryol Carnegie Inst Wash 1966;38:43e58.

[18] Pijnenborg R, Bland JM, Robertson WB, Brosens I. Uteroplacental ar-

terial changes related to interstitial trophoblast migration in early hu-

man pregnancy. Placenta 1983;4:397e414.

[19] De Wolf F, De Wolf-Peeters C, Brosens I. Ultrastructure of the spiral

arteries in the human placental bed at the end of normal pregnancy.

Am J Obstet Gynecol 1973;117:833e48.

[20] Billington WD. Vascular migration of transplanted trophoblast in the

golden hamster. Nature 1966;211:988e9.[21] Orsini MW. The trophoblastic giant cells and endovascular cells associ-

ated with pregnancy in the hamster, Cricetus auratus. Am J Anat

1954;94:273e331.

[22] Orsini MW. Observations on deciduomata in sterile horns of pregnant

hamsters. J Endocrinol 1968;41:VII.

[23] Browne JCM, Veall N. The maternal placental blood flow in normoten-

sive and hypertensive women. J Obstet Gynaecol Br Empire

1953;60:141e8.

[24] Robertson WB, Brosens I, Pijnenborg R, De Wolf F. The making of the

placental bed. Eur J Obstet Gynecol Reprod Biol 1984;18:255e66.

[25] Brosens IA, Robertson WB, Dixon HG. The role of spiral arteries in the

pathogenesis of preeclampsia. In:Wynn RM, editor. Obstetrics and gyne-

cology annual. New York: Appleton-Century-Crofts; 1972. p. 177e91.

[26] De Wolf F, De Wolf-Peeters C, Brosens I, Robertson WB. The human

placental bed: electron microscopic study of trophoblastic invasion of

spiral arteries. Am J Obstet Gynecol 1980;137:58e70.

[27] Pijnenborg R, Dixon G, Robertson WB, Brosens I. Trophoblastic inva-

sion of human decidua from 8 to 18 weeks of pregnancy. Placenta

1980;1:3e19.

[28] Pijnenborg R, Bland JM, Robertson WB, Dixon G, Brosens I. The pat-

tern of interstitial trophoblastic invasion of the myometrium in early hu-

man pregnancy. Placenta 1981;2:303e16.[29] Pijnenborg R, Robertson WB, Brosens I. The arterial migration of tro-

phoblast in the uterus of the golden hamster. J Reprod Fertil

1974;40:269e80.

[30] Pijnenborg R, Robertson WB, Brosens I. The role of ovarian steroids in

placental development and endovascular trophoblast migration in the

golden hamster. J Reprod Fertil 1975;44:43e51.

[31] Brettner A. Zum Verhalten der Sekundaren Wand der Uteroplacentarge-

fasse bei der Decidualen Reaktion. Acta Anat 1964;57:367e76.[32] Khong TY, Lane EB, Robertson WB. An immunocytochemical study of

fetal cells at the maternaleplacental interface using monoclonal anti-

bodies to keratins, vimentin and desmin. Cell Tissue Res 1986;246:

189e95.

[33] Bulmer JN, Wells M, Bhabra K, Johnson PM. Immunohistological char-

acterization of endometrial gland epithelium and extravillous fetal tro-

phoblast in third trimester human placental bed tissues. Br J Obstet

Gynaecol 1986;93:823e32.

[34] Meekins JW, Pijnenborg R, Hanssens M, McFadyen IR, Van Assche FA.

Immunohistochemical identification of placental bed biopsies and the

implications when studying the spiral artery response to pregnancy.

Hypertens Pregnancy 1994;13:61e9.

[35] Cartwright JE, Kenny LC, Dash PR, Crocker IP, Aplin JD, Baker PN,

et al. Trophoblast invasion of spiral arteries: a novel in vitro model.

Placenta 2002;23:232e5.

[36] Craven CM, Morgan T, Ward K. Decidual spiral artery remodelling

begins before cellular interaction with cytotrophoblasts. Placenta

1998;19:241e52.[37] Craven C, Morgan T, Ward K. Trophoblasts are unnecessary for decid-

ual spiral artery remodeling: lessons from the ectopic pregnancy. J Soc

Gynecol Invest 1997;4(Suppl):137A.

[38] King A, Loke YW. Placental vascular remodelling. Lancet

1997;350:220e1.

[39] Brosens JJ, Pijnenborg R, Brosens I. The myometrial junctional zone

R. Pijnenborg et al. / Placspiral arteries in normal and abnormal pregnancies. Am J Obstet Gyne-

col 2002;187:1416e23.[40] Pijnenborg R. The human decidua as a passage-way for trophoblast

invasion. Trophoblast Res 1998;11:229e41.

[41] Hustin J, Schaaps JP.