Trophoblast invasion: Lessons from abnormally invasive placenta (placenta accreta)

Abstract

The invasion of the uterine wall by extravillous trophoblast is acknowledged as a crucial component of the establishment of pregnancy however, the only part of this process that has been clearly identified is the differentiation of cytotrophoblast (CTB) into the invasive extravillous trophoblast (EVT). The control of invasion, both initiation and termination, have yet to be elucidated and even the mechanism of differentiation is unclear. This review describes our studies which are designed to characterize the intracellular mechanisms that drive differentiation. We have used the over-invasion observed in abnormally invasive placenta (AIP; placenta accreta) to further interrogate this mechanism. Our results show that first trimester CTB to EVT differentiation is accomplished via an epithelial-mesenchymal transition (EMT), with EVT displaying a metastable, mesenchymal phenotype. In the third trimester, while the invasiveness of the EVT is lost, these cells still demonstrate signs of the EMT, albeit diminished. EVT isolated from AIP pregnancies do not however, show the same degree of reduction in EMT shown by normal third trimester cells. They exhibit a more mesenchymal phenotype, consistent with a legacy of greater invasiveness. The master regulatory transcription factor controlling the EMT appears, from the observational data, to be ZEB2 (zinc finger E-box binding protein 2). We verified this by overexpressing ZEB2 in the BeWo and JEG3 trophoblast cell lines and showing that they became more stellate in shape, up-regulated the expression of EMT-associated genes and demonstrated a substantially increased degree of invasiveness. The identification of the differentiation mechanism will enable us to identify the factors controlling invasion and those aberrant processes which generate the abnormal invasion seen in pathologies such as AIP and preeclampsia.

1. Introduction

Trophoblast invasion of the uterus is required for fetoplacental development. Invading trophoblast perform multiple essential functions including the anchoring of the placenta to the uterus, regulating maternofetal immune tolerance and conversion of the maternal spiral arterioles, ensuring adequate blood supply to the intervillous space. The mechanisms and regulation of many of these functions remain to be elucidated. Limited access to tissues has restricted research in the human, and the differences between human and rodent invasion processes constrain investigation using the most common animal models. Nonetheless, exploration of invasion in the human can take advantage of natural experiments, pathologies which demonstrate abnormalities of invasion, including the under-invasion seen in preeclampsia and the over-invasion characteristic of Abnormally Invasive Placenta (AIP, aka placenta creta, accreta, increta, percreta and more recently, Placenta Accreta Spectrum or PAS).

Much of what we know regarding invasion has been observed in preeclampsia, changes such as the absence of integrin switching [1–5] or the loss of matrix metalloproteinases [6,7]. Much less is known about AIP due in part to its relatively low incidence. Incidence has increased from 1 in 30,000 deliveries, as determined in 1937 [8] to approximately 1 in 1000 currently [9–11]. It has become more amenable to investigation, in part through the specialist/referral centers [12]. This review focuses on underlying mechanisms of trophoblast invasion, using AIP as a means of comparative assessment.

2. Placenta previa, uterine damage, access and the route to AIP

Much of what we know about AIP has been inferred from the clinical presentation as placental over-invasion. There are a number of important conclusions we can draw from these epidemiological and observational studies. The two primary risk factors for AIP are placenta previa (placenta implanted over the cervix) and some form of utero-myometrial damage, almost always scarring from one or more lower transverse Caesarean sections (LTCS) [9,10,13]. With respect to the former risk factor, it has been suggested that the thinner endo-myometrial layer around the cervix in placenta previa is more conducive to trophoblast over-invasion, since it presents a reduced barrier to invading trophoblast [14].

Uterine damage, such as that caused by Caesarean section, has been posited to promote AIP by providing an acellular, more sparsely populated or less resistant route for trophoblast permeation through the scar (Fig. 1B). The two risk factors in combination increase the background population-based risk ~100-fold [15].

Fig. 1.

Models of extravillous trophoblast trans-uterine access. The figure shows the potential pathways for extravillous trophoblast (EVT) movement across the decidua from the anchoring villous tip into the myometrium. The figure shows normal pregnancy (A), a pregnancy where uterine damage, such as a Caesarean section, has left a scar across the uterus (B) and a pregnancy where the decidual cell population is altered in terms of either numbers or function (C).

The study by Garmi et al. [16] provides some support for uterine injury as a causative factor. Measuring trophoblast invasion in the presence of decidua, they showed that “injured” decidua increased the extent of trophoblast invasion relative to intact decidua. It is important to note however that it is only the myometrium which is likely to be scarred by Caesarean section, since the endometrium will be renewed following pregnancy and subsequently following menses. The question therefore arises of whether there is such a thing as an “injured” decidua. It seems more probable that a globally abnormal decidua may arise due to other factors, genetic, endocrinological etc. (Fig. 1C).

The concept of altered access is also the explanation presented by Tantbirojn and Parast [17], in which they propose that placenta increta and percreta are not due to intrinsically-generated trophoblast over-invasion, but rather arise secondary to Caesarean scar dehiscence. This enables the entry of chorionic villous tissue into the myometrium, potentially allowing extravillous trophoblast access to the deep myometrium. The increased propensity of AIP placentae towards abruption, and also uterine rupture at the site of an old scar supports this hypothesis.

Other studies [18] have concluded that AIP is not associated with increased capacity for proliferation or invasiveness in trophoblast populations. The conclusion from the combination of these types of studies is that AIP arises primarily from factors such as reduced, absent or abnormal decidua, and/or defective decidualization, leading to increased depth of penetration by (normal) extravillous trophoblast.

3. Abnormal invading trophoblast

Evidence that changes in the invading trophoblast are associated with absent decidualization is obvious in ectopic (most notably tubal) pregnancy. Trophoblast invasion is unimpeded, demonstrating that extravillous trophoblast possess an intrinsic invasive nature [14,19,20], and can differentiate from the parent cytotrophoblast cells in the absence of decidua. There are data showing that trophoblast-specific elements such as growth-, angiogenesis- and invasion-related factors regulate trophoblast invasiveness in AIP [21–24]. Other AIP-associated molecular changes, proposed as factors responsible for trophoblast over-invasion, include abnormalities in secreted extracellular matrix, in secretion of matrix metalloproteinases and in the development of a mesenchymal phenotype [24–27]. There are reasons therefore to believe that a more invasive, or uninhibited (by lack of decidua) invasive trophoblast phenotype is a contributing factor to AIP. While these studies demonstrate up-regulation of trophoblast invasiveness in AIP, in many of these cases it is possible that environmental cues may be responsible for the alterations observed. We believe it is quite probable that both arguments are correct, that both altered access and increased trophoblast invasiveness may be involved in the pathogenesis of AIP.

Is there a reason therefore to think that defects exist at the molecular level beyond the well-recognized risk factors? Aside from the importance of the risk factors, is the obvious fact that the existence of these risk factors in a pregnancy does not automatically lead to AIP. There are many pregnancies in which the presence of placenta previa in subjects who have had multiple Caesarean sections does not result in AIP. Clearly while the risk factors are important pre-disposing components, they do not appear to tell the entire story. It seems likely that some other element or elements, interacting with the effects of placenta previa and uterine damage, complete the equation necessary for AIP causation. This is the root of our molecular investigations into AIP pathogenesis.

4. Exploring the underlying mechanism

Many of the in vitro studies of trophoblast invasion have been based not on an underlying mechanism, but rather on association with isolated invasion-related factors. The wealth of studies examining the under-invasion in preeclampsia has identified many candidate genes involved in the invasion process and therefore possible targets for investigation. Many of these studies are focused on the development of the invasive cells, the extravillous trophoblast and their properties. Alterations in processes such as the conversion of trophoblast from CTB through pre-EVT cell types such as cell column trophoblast, the control of EVT movement through the decidua and the transformation of active EVT into the non-motile multinuclear trophoblast giant cells may all contribute in the development of invasion abnormalities. We have chosen to investigate the differentiation of cytotrophoblast into extravillous trophoblast, the process by which non-motile polarized cells anchored to a basal lamina are converted to non-polarized, anchorage-independent invasive cells.

AIP is an invasion pathology and many investigative attempts have focused on alterations in the extravillous trophoblast (EVT) infiltration of the uterus or factors regulating EVT invasion [28–34]. It important to realize that currently we have incomplete understanding of the processes involved in the normal invasion process. We have minimal understanding of the regulatory elements controlling the promotion and inhibition of invasion, no matter the aberrant control that is a feature of AIP. Research into AIP therefore is, in part, research into the invasion process itself, as well as an attempt to isolate those components which demonstrate abnormal function. There is a substantial literature addressing trophoblast invasion, but it is drawn in large part from research into the under-invasion characteristic of preeclampsia. Much of this invasion literature is focused on the role played by specific genes/proteins. However, in many cases, these genes or proteins have not been examined functionally or have not been shown to vary as a result of invasion pathologies in vivo, raising the question of physiological relevance. This is important because as a complex, multi-component process, it is possible to compromise trophoblast invasion in vitro by modifying a single component, whether or not such a modification occurs in vivo. Nevertheless, it is possible to glean from these reports, information related to the invasion processes, allowing us to build a picture of the components of the normal invasive mechanisms.

We have been investigating the molecular changes in normal and abnormal invasion, starting with studies comparing CTB and EVT in the first and third trimesters [35,36]. We focused on normal processes which might be subverted in AIP (and possibly in preeclampsia). One clear option is disruption of the fundamental mechanism of normal CTB-EVT differentiation, the epithelial-mesenchymal transition (EMT). The idea that the differentiation of cytotrophoblast into extravillous trophoblast might be an EMT has been extant for many years and there has been sporadic evidence presented for this concept [1,37–43]. However, it is only more recently that we have explored this idea to the degree necessary for definitive identification of the process and classification within the range of EMT types already identified [44].

Our initial studies interrogated first trimester CTB and EVT to determine if the differences between them indicated the existence of an EMT as the mechanism of differentiation. We thus examined the expression, in CTB and EVT, of 84 genes associated with EMT, to determine if changes in these genes could provide evidence of an EMT in this differentiation process. The results showed quite clearly that many of the gene expression changes were characteristic of an EMT (Table 1) [35]. On the EMT-associated PCR array we used, there were also genes which showed no change or a change opposite to that which might be expected from other well-characterized EMT types. We attribute this, at least in part, to the breadth and variation within the EMT process. Many of the genes in the PCR array were drawn from the best-investigated examples of EMT, primarily cancer metastasis. Many of these genes might not be expected to form part of a normal trophoblast EMT process. Our data supports that the CTB/EVT differentiation process is a unique type of EMT, separate from those previously identified (gastrulation, wound healing, metastasis). Certainly, it has some unique characteristics, such as the down-regulation of TWIST1. This transcription factor, generally acknowledged as a master EMT regulator, is associated in the trophoblast system with the development of syncytiotrophoblast cells rather than the lineage pathway leading to EVT [45,46]. In addition, EVT not only retain but up-regulate the expression of cytokeratins 7, 14 and 19, elements identified as epithelial markers, which are frequently lost in other EMT types during the transition to a mesenchymal phenotype. This suggests the progression of trophoblast to a metastable cell type within the EMT spectrum (Fig. 2), capable of further movement, forward to a more mesenchymal phenotype or backward, to the epithelial phenotype [47]. We speculate that CTB differentiation may be the molecular archetypal EMT, a “type 0” EMT, as it occurs so early in human development compared to EMT types 1–3 [44,48].

Table 1.

First trimester EMT gene changes (CTB to EVT).

Down-regulated genes		Up-regulated genes
Gene ID	Fold change	Gene ID	Fold change
BMP7	−44.1	FN1	106.9
CDH1	−7.1	ITGA5	139.7
CTNNB1	−5.2	ITGB1	4.3
EGFR	−23.8	KRT14	41.4
FOXC2	−57.8	KRT19	6.2
FZD7	−11.5	KRT7	2.7
JAG1	−30.1	MMP2	356.8
OCLN	−35.2	MMP3	128.9
SNAI1	−2.3	MMP9	160.4
SNAI2	−10.6	NOTCH1	6.4
TWIST1	−8.8	SNAI3	4.3
WNT5A	−14.6	SPARC	4.9
		SPP1	186.3
		TCF4	18.4
		TFPI2	24.4
		TGFB1	47.7
		TGFB2	115.1
		TIMP1	24.5
		VIM	235.2
		WNT5B	3.0
		ZEB2	198.5

The genes in this table all demonstrated significant differential expression in first trimester EVT compared to first trimester CTB controls. Analysis by t-test or Mann-Whitney U test. p < 0.05; n = 6, 6. Data from reference 35.

Fig. 2.

The trophoblast epithelial-mesenchymal transition (EMT) spectrum. The spectrum of the EMT reaches from epithelial to mesenchymal. Cytotrophoblast are at the epithelial end, while first trimester EVT are situated well towards, but not at, the mesenchymal pole. Third trimester EVT show a regression, back along the spectrum towards the epithelial (cytotrophoblastic) end. EVT obtained from AIP pregnancies do not show the same degree of regression and are placed towards the mesenchymal end of the spectrum compared to third trimester normal or placenta previa controls.

5. EMT status in the third trimester

Like preeclampsia, AIP is generally only identified and accessible in the late second and third trimester. While the invasion process has been long-completed at this point, we hypothesized that third trimester EVT cells from AIP pregnancies might still reflect the abnormal process that occurred in the first and second trimesters. We analyzed CTB/EVT differences in normal third trimester pregnancies, as a precursor to the analysis of AIP. When we compared CTB and EVT from normal term pregnancies, we found that while signs of an EMT were still apparent, many of the gene expression changes observed in the first trimester were much reduced compared to the third trimester profile (Table 2) [36]. We concluded that the EVT had regressed from the first trimester position on the EMT spectrum to one closer to that of the CTB (Fig. 2).

Table 2.

Third trimester EMT gene changes (CTB to EVT).

Down-regulated genes		Up-regulated genes
Gene ID	Fold change	Gene ID	Fold change
BMP7	−34.0	FN1	59.8
CDH1	−1.5	IGFBP4	32.4
COL1A2	21.9	ITGA5	38.6
CTNNB1	−6.3	ITGB1	4.0
EGFR	−4.1	KRT19	6.6
F11R	2.6	KRT7	2.7
FOXC2	−3.6	MMP2	203.9
FZD7	−4.1	MMP3	29.6
JAG1	−9.2	PDGFRB	60.8
MMP9	−7.2	SNAI1	5.0
NOTCH1	−2.8	SNAI2	24.7
OCLN	−32.9	SPARC	46.8
SPP1	−5.9	STAT3	2.5
TWIST1	−14.3	TCF4	2.1
ZEB2	−3.3	TFPI2	3.0
		TGFB1	9.1
		TGFB2	34.7
		TIMP1	43.1

The genes in this table all demonstrated significant differential expression in third trimester EVT compared to third trimester normal CTB controls. Analysis by t-test or Mann-Whitney U test. p < 0.05; n = 8, 8. Data from reference 36.

6. Trophoblast differentiation in AIP

In the next stage we undertook analysis of AIP. There are important issues in the investigation of AIP that merit consideration. Although the incidence of AIP is rising, it is still a relatively rare pathology compared to other placental pathologies of pregnancy. Acquiring sufficient samples usually requires access to a referral center. In addition, use of the optimum validation of histopathological analysis is crucial. Clinical, post-surgical reports are inadequate validation of an AIP; histopathological evaluation is the recommended method of assessment and classification, necessary to decide whether a case is truly AIP and to determine the severity of AIP (i.e. accreta, increta, percreta, Grades 1, 2 or 3 on the recent FIGO Clinical Classification System; [49]). Published reports lacking histopathological analysis and/or stratification by AIP grade are always subject to the question of whether results have been obtained from a mixture of normal and pathological samples or a mixture of pathologies. For this reason, we conduct our research using samples from cases where histopathological analysis has confirmed the presence of either placenta increta or placenta percreta (i.e. placental villi within the muscular fibers and sometimes in the lumen of the deep uterine vasculature or villous tissue within or breaching the uterine serosa).

Another important issue is that of controls. National clinical professional organizations recommend delivery of AIP cases at 34–36 weeks, before bleeding or other instability becomes an issue. Use of term pregnancies as controls raises questions of gestational age effects. Use of age-matched cases of preterm birth raises the issue of the causal elements that lead to preterm birth as potentially confounding factors. For our controls we used cells prepared from cases of placenta previa, where the placental is implanted over the cervix and is also often delivered before term, usually, as in AIP, due to potential risk of vaginal bleeding. Unlike preterm birth controls, the previa cases are electively delivered because of the mechanical obstruction presented to delivery by the placenta. There are no metabolic, endocrine or unknown causal issues associated with the previa cases. They are delivered before term to mitigate the risk of events such as abruption/hemorrhage. All our AIP pregnancies also had placenta previa, and were delivered around the same gestational age, making the previa cases ideal controls for AIP pregnancies.

Supporting the use of previas as controls, when we compared EVT from placenta previa with normal term EVT, we found only 3 changes out of 84 genes, 2 of which were changes of less than 1.5-fold. This supports that normal and previa EVT are very similar, at least with respect to this set of EMT-related genes. Using CTB and EVT from the placenta previa cases as controls, we then compared them with the same cells from cases of AIP (all percreta, Grade 3a, 3b of the FIGO classification). While there were no differences between CTB from placenta previa and AIP cases, multiple differences were found between AIP and control EVT (Table 3) [36]. These included loss of CDH1 and increased MMP2, TGFß2 and ZEB2, changes associated with an EMT. EVT obtained from AIP are thus shifted towards the mesenchymal end of the EMT spectrum compared to placenta previa or normal term EVT (Fig. 2), although they clearly show less of a mesenchymal phenotype than first trimester EVT. These data do not permit determination as to whether this resulted from an increased mesenchymal shift in EVT from AIP pregnancies during the first/second trimester, or whether the EVT-AIP were simply less affected by the factors causing regression of EMT status towards the epithelial end of the spectrum during the second/third trimester.

Table 3.

Third trimester EMT gene changes (EVT-previa to EVT-AIP).

Gene ID	Fold change
CDH1	−1.45
CDH2	2.34
COL3A1	3.94
KRT14	1.76
MMP2	3.74
PLEK2	4.87
SNAI2	8.88
SPARC	1.59
SPP1	2.09
STAT3	1.30
TFPI2	2.20
TGFB2	1.85
WNT11	3.32
ZEB2	4.12

The genes in this table all demonstrated significant differential expression in EVT from AIP pregnancies compared to EVT from placenta previa controls. Analysis by t-test or Mann-Whitney U test. p < 0.05; n = 8, 8. Data from reference 36.

7. Regulation of the EMT

As part of the analysis of EMT genes, we analyzed the expression of multiple EMT “master regulator” transcription factors (FOXC2, GSC, TWIST1, SNA1, SNA2, ZEB1, ZEB2) [50,51]. Only ZEB2 (zinc finger E-box binding protein 2) displayed characteristics which matched with both the invasiveness of cells in vivo and with EMT status. The other master regulators either showed no change or changes which were inconsistent with invasiveness. This was surprising, because ZEB2 has generally been regarded as a suppressive transcriptional regulator, associated primarily with the down-regulation of epithelial genes such as CDH1 (E-cadherin) [52–54]. In first trimester trophoblast however, ZEB2 gene expression is increased by almost 200-fold in EVT compared to CTB but, in the third trimester, drops back to a level below that of third trimester CTB [35,36]. In AIP however, EVT levels of ZEB2 remain elevated compared to the corresponding control (placenta previa) EVT, consistent with the more mesenchymal phenotype of cells from the AIP placenta [36].

Drawing from these data, we hypothesized that the ZEB2 transcription factor was responsible for regulating progress of trophoblast cells through the EMT. This is consistent with the levels of ZEB2 in trophoblast cell lines; the non-invasive BeWo and the minimally invasive JEG3 choriocarcinoma lines have low gene expression levels of ZEB2 compared to the much higher level of ZEB2 encountered in the invasive HTR8/SVneo model EVT cell line [36]. To test this hypothesis, we overexpressed ZEB2 in BeWo and JEG3 cells and isolated clonal lines. The two clones showing highest expression of ZEB2 (>40-fold increase in expression) demonstrated the changes in morphology, gene expression and invasiveness characteristic of a mesenchymal shift in EMT status, while lower ZEB2-expressing clones were not affected [55]. This supports our hypothesis with respect to the role of ZEB2 and leads us to conclude that we have identified a major mechanism controlling the CTB/EVT differentiation process, leading to the development of invasive cells. We note also that other systems including the Wnt signaling pathway and transcriptional factors such as STOX1 may play similar roles in regulating EMT-driven trophoblast invasiveness [56–58]; further research is necessary to integrate these elements into a coherent mechanism.

8. Effects of the uterine environment

Several groups have suggested that cell-cell interaction, either through direct contact or through secreted factors, is responsible for regulating invasion. There are reports showing that conditioned medium from decidual natural killer cells (dNK) promotes invasion [59–61]. dNK have been shown to be present at their highest level early in gestation and to decline over gestation [62,63]. However, very reduced dNK cell numbers were observed in near-term AIP compared to previa controls [64], suggesting that any invasion-promoting properties are minimized or no longer relevant later in gestation. By contrast, another study showed no change in dNK in AIP, but an increase in T-regulatory lymphocytes and significantly fewer immature, non-activated dendritic cells [32]. There is no clear evidence for the limitation of invasion via cell-cell interaction although, as noted above, in the absence of a decidual/myometrial layer, trophoblast continues to invade, suggesting some form of trophoblast/decidual interaction is at the root of the restrictions on invasion [59,65–68].

9. Conclusions

Despite the evidence for extra-trophoblastic etiologic factors, most of the limited research being performed into causes of AIP has generally been directed towards discovery of factors that increase trophoblast invasiveness. Less attention has been paid to potential interactions of the invading trophoblast with a reduced or defective decidual layer, but recent research shows that changes in the decidual/myometrial environment surrounding the trophoblast may have major effects on trophoblast function. Investigators have already stepped back from pregnancy to explore the possibility that the uterine environment pre-pregnancy may reflect differences which lead to the pathological effects in preeclamptic pregnancy [69,70]. Similar circumstances may also result in the identification of pre-pregnancy elements which, when added to uterine damage and placenta previa, lead to AIP. As described above, there has been a continuing debate over the question of whether aberrant extravillous trophoblast is the causal element in over-invasion, as opposed to defective decidual/uterine components which enable over-invasion. By extension, it is the possible that those uterine characteristics which influence the invasion process toward AIP may also be apparent in the non-pregnant state. Under these circumstances, intervention to modify these characteristics, pre-pregnancy, might also be possible. To explore these avenues, it is necessary that we understand the mechanisms of over-invasion and the forces driving it, hence the need to determine the molecular etiology.

Our data provide a clearer definition of the EMT as a mechanism for CTB/EVT differentiation. Final confirmation and definition awaits a more in-depth assessment of the gene expression profile in the differentiation process. Nevertheless, we believe we now have a tool by which to assess the extracellular signals that regulate the differentiation status of these cells and, by extension, their invasiveness. The changes in gene expression characteristic of the EMT in trophoblast will enable us to assess the effect of regulatory elements, not simply using one or two parameters but on the basis of the process by which these cells control their phenotype, including morphology and invasiveness. We now have a model of the EMT spectrum, the alterations observed in AIP and the potential for movement within that spectrum leading to changes in invasiveness that are pathophysiologically relevant to both AIP and preeclampsia. It will now be possible to focus on those agents in pathologies which shift cells across the EMT spectrum. It will be instructive to determine whether similar shifts, but toward the epithelial phenotype, are found in EVT from cases of preeclampsia, especially the severe early-onset, placenta-associated form of the disease. There are suggestions that the under-invasiveness observed in preeclampsia may also involve the EMT spectrum, as markers of the reverse process, mesenchymal-epithelial transition, have been observed [71]. The EMT provides a solid mechanistic basis for the abnormally invasive pathologies and for investigations of the defects, trophoblastic or uterine, which are causal.