The promise of placental extracellular vesicles: models and challenges for diagnosing placental dysfunction in utero

Abstract

Monitoring the health of a pregnancy is of utmost importance to both the fetus and the mother. The diagnosis of pregnancy complications typically occurs after the manifestation of symptoms, and limited preventative measures or effective treatments are available. Traditionally, pregnancy health is evaluated by analyzing maternal serum hormone levels, genetic testing, ultrasonographic imaging, and monitoring maternal symptoms. However, researchers have reported a difference in extracellular vesicle (EV) quantity and cargo between healthy and at-risk pregnancies. Thus, placental EVs (PEVs) may help to understand normal and aberrant placental development, monitor pregnancy health in terms of developing placental pathologies, and assess the impact of environmental influences, such as infection, on pregnancy. The diagnostic potential of PEVs could allow for earlier detection of pregnancy complications via noninvasive sampling and frequent monitoring. Understanding how PEVs serve as a means of communication with maternal cells and recognizing their potential utility as a readout of placental health have sparked a growing interest in basic and translational research. However, to date, PEV research with animal models lags behind human studies. The strength of animal pregnancy models is that they can be used to assess placental pathologies in conjunction with isolation of PEVs from fluid samples at different time points throughout gestation. Assessing PEV cargo in animals within normal and complicated pregnancies will accelerate the translation of PEV analysis into the clinic for potential use in prognostics. We propose that appropriate animal models of human pregnancy complications must be established in the PEV field.

Summary sentence Experimental animal models will be essential for defining the opportunity that placental extracellular vesicles may provide for monitoring placental health and function and understanding the pathophysiology of adverse pregnancy outcomes.

Introduction

Placental complications arise in approximately 15% of pregnancies, and due to pregnancy and childbirth, maternal deaths occur in approximately 275,000 cases worldwide annually [1]. In 2010, preterm birth (PTB) was the most common cause of infant mortality and morbidity affecting approximately 15 million babies; Southeastern Asia, South Asia, and sub-Saharan Africa had the highest rates [2]. PTB causes approximately 1 million neonatal deaths each year across the world [3], with surviving infants displaying elevated risks of cardiovascular and respiratory diseases, neurological deficits, and developmental disabilities [4]. Fetal growth restriction (FGR) is the second most common pregnancy complication (impacts ~8% of pregnancies) [5]. FGR increases the risks of intrauterine demise, neonatal morbidity, cognitive delay, and adult onset disease later in life [6, 7]. Women with pre-gestational diabetes (pre-GD) (having type 1 or type 2 diabetes before becoming pregnant) also have a greater risk of exacerbated symptoms during pregnancy, such as diabetic ketoacidosis, myocardial infarctions, retinopathy, and nephropathy, as well as obstetric complications including preeclampsia (PE), uteroplacental insufficiency, preterm labor, shoulder dystocia, and stillbirth [8–11]. Pregnancy complications not only cause emotional stress and trauma on the parents but also create an extreme financial burden for the family and the healthcare system. The annual financial costs are estimated to be $26.2 billion for PTB, $2.18 billion for PE, and $1.8 billion for pregnancy-acquired diabetes (gestational diabetes mellitus; GDM) in the United States [12–14]. Earlier detection of complications, preventative strategies or therapeutics, and placenta-targeted treatments are imperative to improving the in utero environment of the fetus to benefit the long-term health of both the mother and the child.

Diagnosis of a pregnancy complication relies on close monitoring of maternal and fetal health. Obstetricians monitor maternal health by measuring blood pressure, checking vital organs (e.g., renal function), and monitoring fetal health by measuring uterine growth via fundal height and fetal growth directly with ultrasound. Clinicians can also complete a biophysical profile and use Doppler velocimetry to monitor fetal blood flow within the umbilical cord and fetal middle cerebral artery [15]. Although the placenta is essential in pregnancy, no minimally invasive techniques to directly monitor placental health are available. If biomarkers of abnormal placentation could be identified before the pathophysiological complication manifests itself, they may reveal the underlying mechanism(s) that contributes to the insult and provide a targeted approach to develop therapies/treatments.

Pregnancy complications that arise due to a malfunctioning or maldeveloped placenta include PE, early/recurrent pregnancy loss (EPL/RPL), PTB, FGR/intrauterine growth restriction (IUGR), and pre-GD/GDM (Table 1, Supplemental Table 1) [16]. An overview of complications is provided in Table 1 and includes the frequency of the complication within the population, the diagnostic measure to identify the pregnancy complication, and the known pathophysiology for the condition. While GDM is not thought to be a placental disease per se, it may impact fetal well-being as structural and functional alterations can occur in the placenta [17–19]. In addition, maternal infection with vertically transmitted pathogens also gives rise to adverse pregnancy outcomes and may impact placental function or fetal development [20, 21]. Among the implicated pathogens, the TORCHZ group is of particular neonatal concern and consists of the following: Toxoplasma gondii, Other (Listeria monocytogenes, Treponema pallidum, varicella zoster virus, human immunodeficiency virus, enteroviruses and parvovirus B19), Rubella virus, Cytomegalovirus, Herpes simplex virus, and Zika virus [20, 22]. Thus, it is necessary to monitor placental health during and after maternal infection to determine if vertical transmission has occurred and to assess the risk of developing an adverse pregnancy outcome following infection.

Table 1.

Human pregnancy complications.

Pregnancy complication	Percent impact	Current means of identification	Underlying pathology
PE	• 3–5% of pregnancies [177] • Accounts for 14% of pregnancy-associated maternal deaths [177, 178]	• High blood pressure [354] • Proteinuria [354] • Possible renal, liver, pulmonary, and neurological sequelae [177] • PE is sometimes split into two categories with separate pathologies: early onset PE (EPE), <34 weeks gestation and late onset PE (LPE), ≥34 weeks gestation	• Abnormal trophoblast invasion of the maternal decidua and incomplete remodeling of maternal spiral arteries leads to placental ischemia and a pro-inflammatory environment [179] • Often associated with IUGR [177] • EPE: considered a fetal disease with associated placental dysfunction • LPE: a maternal disorder; placenta usually functions properly and is associated with better maternal and fetal outcomes, at least in part based on later delivery (obviating prematurity as a contributor to pathology) [180]
EPL	• 15–25% of all clinical pregnancies [181–183]		• Fetal chromosomal abnormalities are the major cause [184, 185]; however, in cases of normal fetal karyotype, the cause is typically unknown [186–189]
RPL	• ~1% of all women [190]	• RPL is identified by two or more clinical EPLs (pregnancy diagnosis based on ultrasound or tissue/pathology, not chorionic gonadotropin detection alone) [181, 191]	• When uterine anatomical anomalies, genetic factors, antiphospholipid syndrome, or hormonal pathologies are ruled out, more than 50% of patients suffer from unexplained RPL [181, 182, 192]
FGR/IUGR	• 10% of pregnancies [193]	• Ultrasound-estimated fetal weight less than the 10th percentile for that gestational age or a fetus at <10th percentile at birth [193]	• Origins stem from errors in early placental development [194] • Caused by placental insufficiency, genetic syndromes, maternal malnutrition, multiple gestation, teratogens, and oxygen deprivation [134, 195]
			• Maternal vascular malperfusion is considered the leading cause of pathology in FGR placentas and can be accompanied with placental hypoplasia, infarction, and hemorrhage [196]
			• Often associated with PE [177]
Pre-GD	• ~7% of all pregnancies [11] • High risk of maternal morbidity [197]	• Having type 1 or type 2 diabetes prior to start of pregnancy	• Hyperglycemia can impair and disrupt fetal organ development [8] • Placentas from women with diabetes generally have greater surface area, Hofbauer cells, vasculature, and diffusion distance [199]
GDM	• Leading cause of fetal macrosomia [198]	• Glucose intolerance and insulin resistance development during pregnancy	• Placental pathologies include chorionic villus immaturity, high villus density, stromal edema, thicker than average collagen fibers, and diffuse villous stromal calcifications [17], potentially impacting function
PTB	• ~11% of all pregnancies [170] • 75% of perinatal mortality is due to PTB [200]	• Birth before 37 weeks of gestation [170]	• Oxidative stress and inflammation leading to early rupture of membranes are known to be leading causes of PTB, in addition to retroplacental abruption, chronic villitis, and twin gestations [201] • Other causes hypothesized to be involved include vertically transmitted infections, maternal environmental stress, and intra-amniotic inflammation [170, 200]

PEVs may provide an excellent tool with which to monitor placental health and function in human patients. Isolation from fluids is minimally invasive, repeated sampling is feasible, and the ability to monitor the same patient over time provides valuable information as to how the placenta matures, develops, and responds to insult. Most importantly, PEVs contain placenta-specific proteins, which may be used to selectively isolate them from more complex samples, such as blood [23, 24]. However, the roles of PEVs in cellular communication and maternal physiology are not well understood. PEVs are present in maternal blood as early as 6 weeks of gestation [25], and their presence early on makes them attractive molecular packages that may contain a readout of placental health from the early stages of placental development through delivery of the newborn.

The purposes of this review are to (1) introduce EVs, or more specifically PEVs, as a molecular readout of placental health, (2) provide an overview of current information known about human PEVs, (3) discuss animal models used to study pregnancy complications, and (4) discuss future expansion of animal model PEV research to address critical challenges in human PEV research. Although extensive information from human EVs has been obtained from cell cultures and maternal blood sampling, there is a limited understanding of in vivo PEV function in humans and animal pregnancy models. Implementing animal pregnancy models will extend our understanding of trophoblast physiology during pregnancy complications, define PEV cargo and function, and explore the diagnostic and therapeutic potential of PEVs. In vivo studies in experimental pregnancy models are essential to make PEV research translational to a human clinical setting. Due to the breadth of topics this review covers, we apologize for any publications not discussed and refer to more specific reviews wherever possible.

What are extracellular vesicles?

The three main subtypes of extracellular vesicles (EVs) are exosomes (small, 60–80 nm in diameter [26] and large 90–120 nm [26]), microvesicles approximately 100–1000 nm [27], and apoptotic bodies approximately <5 μm [27]. These are distinguished not only by size but also by their route of cellular release (Figure 1) [27–31]. Figure 1 also shows EV release and highlights some unique cargo among the different subtypes. These naming conventions have been greatly debated since it is now well understood that there is overlap in the size and cargo across EV classes [32–34]. The size criteria do not consider the mechanism(s) by which the EV is derived or secreted from the cell. For instance, an EV population isolated based on EV size may encompass both large exosomes and microvesicles; however, these vesicles may differ in their cargo, composition, and biological role. The lipid content may be a more definitive criterion to classify vesicles, as Ouyang et al. [35] demonstrated that phospholipid composition varies between EV classes. Standardized nomenclature across studies regarding the vesicle of interest would greatly strengthen this field of research, as inconsistent nomenclature makes it difficult to compare data across studies. When referencing the work of others in this review, we have used the terminology stated in that publication.

Figure 1.

Schematic diagram of human placental villous structure and PEV biogenesis. Placental structure with increasing magnification (organ level, tissue level, cellular level) is depicted at the left, along with apoptotic body, microvesicle, and exosome biogenesis in a trophoblast cell. A key for molecular elements is at the upper left. Top right: Apoptotic bodies form when a cell is undergoing apoptosis, as depicted by the breakdown of the nucleus and packaging of DNA and organelles. Middle right: Microvesicles are formed via pinching off of the plasma membrane, entrapping molecular cargo. Bottom right: Exosomes are assembled within the secretory pathway. Secretory vesicles (SV) are released by the golgi body and can fuse with endosomes (END). Endosomes can also fuse together or fuse with the lysosome (LYS) for cargo degradation. Late stage endosomes are also referred to as MVBs if they contain intraluminal vesicles (depicted as small light yellow vesicles inside the MVB). All of these vesicles are then released into the maternal bloodstream. Some cargo and membrane proteins specific to the different EV classes are depicted. Lipid composition is depicted in a simplified, grouped manner on the border of the vesicles (green, phosphatidylserine; blue, phosphatidylcholine; purple, phosphatidylethanolamine), with the percent of each lipid content indicated adjacent to the membrane region. The gray membrane of the vesicles represents other lipids that have been identified including sphingomyelin, phosphatidylglycerol, phosphatidylinositol, phosphatidic acid, bis-monoacylglycerophosphate, cardiolipin, lysophosphatidylcholine, and lysophosphatidylethanolamine [35].

A range of isolation techniques have been used to isolate exosomes, including differential ultracentrifugation, size exclusion chromatography, and polyethylene glycol precipitation [33, 36–42]. A major goal of these techniques is to obtain a homogenous population of EVs (i.e., only exosomes), but this remains a challenge given the overlap in size across EV subtypes. It is therefore difficult to compare data across studies, as sample types and isolation techniques vary greatly, especially if the EV populations being analyzed are not homogenous. While there is great scientific interest in exosomes due to their potentially bioactive cargo and ability to be taken up by target cells, recipient cells can also take up microvesicles [43]. Moreover, the quantity of both exosomes and microvesicles present in maternal blood increases throughout gestation [44]. Given the current limitations in isolating a homogeneous population of an EV subtype, it may be more appropriate to globally assess all PEVs to develop diagnostic tools for pregnancy complications. Once techniques are developed to isolate vesicles of a specific EV class, the approach to evaluating PEVs can be modified to assess specific EV subtypes.

EV formation, cargo packaging, and function

Exosomes form within an endosome that is also referred to as a multivesicular body (MVB). As depicted in Figure 1, MVBs can either fuse with the lysosome or the plasma membrane for cellular release. Exosomal surface proteins, such as the tetraspanins, are widely conserved across mammals, as shown in Table 2. Microvesicles form on the cell’s surface where they bleb off from the plasma membrane, and therefore, phosphatidylserine is commonly a component within their membranes [27]. Apoptotic bodies form when a cell undergoes apoptosis, where the cell’s organelles organize into these noninflammatory packages [27]. Microvesicles and apoptotic bodies commonly contain heat shock protein 96 (GP96), actinin-4, and mitofilin [34]. Although there is evidence for conservation of mammalian EV surface markers across EV subtypes, packaging of cargo still remains poorly understood. For more details on packaging of EV cargo and EV secretion pathways, the reader is directed to the following reviews [27, 29, 42, 45–47].

Table 2.

Summary of known EV markers and placenta-specific markers across species.

Species	General EV markers	Placental markers	References
Human	CD9	PLAP	[27, 34, 202–208]
	CD63	PP13
	CD81	Syncytin-1 and -2
	syntenin-1	PAPP-A
	EHD4	HLA-G
	ADAM10	PSG-1
	ESCRT proteins	C19MC
NHP	CD63	PP13	[168, 209–214]
	CD81	Syncytin-1 and -2
	Flotillin-2	PAPP-A
		Mamu-AG
		C19MC
Guinea pig		PLAP	[215, 216]
		Env-cav1
Rabbit	CD9	PLAP	[215, 217, 218]
	CD63	Syncytin-ory1
	CD81
	HSP101
Mouse	CD9	Syncytin-A and -B	[89, 215, 219–222]
	Alix	PLAP
	CD63
	CD81
Rat	CD63	PLAP*	[215, 223–226]
	TSG101
Sheep	CD63	Syn-Rum1	[227–229]
	HSP70
Cattle	CD9	PLAP	[103, 229–231]
	CD63	Syn-Rum1
Pig	CD63		[232]

CD, cluster of differentiation; EHD4, EH domain containing protein 4; ADAM10, a disintegrin and metalloproteinase domain-containing protein 10; ESCRT, endosomal-sorting complexes required for transport.

^*Also expressed in other tissues.

Biologically active nucleic acids, proteins, carbohydrates, and lipids can be packaged into EVs and secreted as a means of intercellular communication [28–30, 48–56]. Extensive research supports the concept that EVs have roles in diverse biological processes including the immune response [57, 58], inflammation [28–30, 59], and transmission of viral infection [28, 48–51, 60]. For example, researchers have shown that EV uptake by recipient cells induces cytokine release [61], inhibits protein translation [53, 62], influences cell proliferation and migration [63], and protects cells from oxidative stress [64]. Additional information on the impact of EVs on recipient cells and EV tropism and therapeutic potential can be found in the following papers [65–70].

PEVs: sample sources, interactions with immune cells, and clinical potential

To highlight the potential information available from EV analysis, scientists have used the terms “circulating biopsy” [71], “fingerprint” [72], and “liquid biopsies” [72, 73]. EVs have received great attention as it has been shown that EV cargo may be altered under diseased and infection states [62, 64, 74], and they can be isolated from minimally invasive fluid samples. PEVs have been detected as early as 6 weeks of gestation [25] (see Table 3 for an overview of the outcomes from human PEV research); however, trophoblasts secrete chorionic gonadotropin into maternal blood shortly after embryo implantation, suggesting that PEVs may encounter maternal cells just as soon.

Table 3.

PEV studies in humans, rodents, cows, pigs, and sheep.

Experimental model	Sources of EVs (experimental and control)		Pregnancy complication	Findings	References
Human	• Plasma from healthy first, second, and third trimester women	• Plasma from healthy, nonpregnant women	—	• Increased quantity of exosomes and PLAP+ exosomes with gestation • Exosomes had a positive impact on endothelial cell wound healing, with early gestation having the greatest impact	[102]
	• Plasma from women with PE	• Control pregnant plasma	PE	• Higher quantities of EVs, ST-derived EVs, and phosphatidylserine/annexin-V-positive EVs in PE patients	[87, 91, 233, 234]
	• HUVEC-conditioned media	• Exosomes isolated from plasma	PE	• Decreased expression of eNOS in HUVECs exposed to exosomes from PE patients • More miR-155 within EVs from PE patients • miR-155 downregulates eNOS expression	[62]
	• Plasma from women with PE	• Plasma from healthy women, gestation-matched	PE	• Elevated levels of exosomes and PLAP+ exosomes, throughout gestation in PE patients	[92]
				• Exosomal miRNA profiles differed significantly
	• Plasma from women with EPE & LPE	• Plasma from gestation age-matched, normotensive women	PE	• Greater exosome quantity in EPE or LPE compared to control • Placental vesicle quantity was elevated in EPE and significantly decreased in LPE, which indicates that the etiology between them is different	[90]
	• PE placentas that underwent either mechanical disruption or perfusion	• Healthy placentas that underwent either mechanical disruption or perfusion	PE	• Increased Flt-1 in PE ST-derived EVs • Decreased endoglin in PE ST-derived EVs	[74]
	• PE placental explant-conditioned media	• Healthy, gestation age-matched placental explant-conditioned media	PE	• Decreased presence of integrins in ST microvesicles from PE placentas may be associated with reduced trophoblast invasion and defective placental vascularization	[235]
	• PE placental explant-conditioned media	• Healthy, gestation age-matched placental explant-conditioned media	PE	• Isolated macro, micro, and nanovesicles with differential ultracentrifugation • Microvesicles from PE placentas were larger in size • PE placentas extruded more micro- and nanovesicles, which contained more Flt-1 than control vesicles • More VEGF was detected in EVs secreted by PE placentas	[93]
	• Sera from women with PE	• Primary term trophoblast cell-conditioned media	PE	• Syncytin-1 and -2 and PLAP are on the membrane of exosomes released by primary trophoblast cells	[145]
		• Sera from gestation age-matched, healthy women		• Less syncytin-2 in exosomes from women with PE compared to control • Decreased internalization of exosomes released by trophoblast cells deficient in syncytin-1 and -2 (via siRNA transfection) compared to controls
	• Plasma from women throughout gestation that later developed LPE	• Plasma from healthy, gestation age-matched women	PE	• Nonsignificant increase in total number of MP over gestation • Over gestation in healthy samples, the quantity of HLA-G+ MPs that contained dsDNA decreased and minimal change in the quantity of PLAP+ MPs with dsDNA was observed • LPE samples contained significantly more total MP than healthy; however, the quantity of HLA-G+ or PLAP+ MPs was not different. This suggests that the ontology of LPE is different than EPE	[236]
	• Placental explant-conditioned media from women with RPL or PE	• Placental explant-conditioned media from healthy, gestation age-matched, women	RPL and PE	• Altered lipid composition in ST microvesicles from RPL and PE compared to controls	[237]
	• Plasma from women with either SGA or FGR fetuses	• Plasma from healthy, gestation age-matched women	FGR	• No difference in the total number of exosomes or PLAP+ exosomes in fetal plasma • Lower percentage of placental exosomes in maternal and fetal plasma from FGR and SGA pregnancies compared to controls • More placental exosomes in fetal than maternal plasma	[238]
	• Plasma from women with either EPE, LPE, or normotensive IUGR	• Plasma from healthy, gestation age-matched women and nonpregnant women	PE and IUGR	• Significantly higher quantities of ST microparticles in EPE plasma, but not LPE or normotensive IUGR, compared to controls	[91]
	• Plasma from women with GDM	• Plasma from healthy, gestation age-matched controls	GDM	• Increase in total exosomes and PLAP+ exosomes from women with GDM compared to controls • Total number of exosomes detected in the plasma of GDM patients at later stages of gestation was negatively correlated with placental weight, while the amount of exosomal PLAP was positively correlated	[56]
				• Increased cytokine release from HUVECs exposed to GDM exosomes than healthy exosomes or baseline
	• Term perfused placenta from women with GDM	• Term perfused placenta from healthy term women	GDM	• 56% of medium/large EVs and 36% of small EVs were PLAP+ • DPPIV, an enzyme that is inhibited in type 2 diabetes mellitus therapy, was co-expressed on PLAP+ EVs. DPPIV retained enzymatic activity on EVs	[239]
				• EVs from GDM samples had greater quantities of DPPIV+ EVs and greater DPPIV activity than controls
	• Plasma from second trimester women with GDM	• Plasma from healthy, gestation age-matched women	GDM	• 78 proteins were differentially expressed in exosomes from women with GDM compared to controls, including proteins involved in metabolic processes and biological regulation	[240]
				• CAMK2beta was more abundant and PAPP-A was less abundant in exosomes from women with GDM compared to controls • Selective enrichment of proteins from the placenta in GDM exosomes compared to controls
	• Serum from first trimester women with GDM	• Serum from healthy, gestation age-matched women	GDM	• PLAP+ EVs were not detected until the 8th week of pregnancy • 17 miRNAs were analyzed by qPCR in total serum EVs	[241]
				• Differential miRNA content between GDM and control EVs. Pathway analysis showed that increased quantities of miRNAs in EVs from GDM were involved in placental development, fetal growth, and insulin and glucose regulation
	• Placental explant-conditioned media from women with GDM	• Placental explant-conditioned media from healthy women	GDM	• Explants from GDM placentas secreted more EVs • Exosomes from controls increased skeletal muscle cell proliferation and migration, in vitro, compared to GDM. Potentially due to differential miRNA expression in exosomes from GDM versus control	[63]
				• miRNA content in exosomes differed from miRNAs in the cells of origin
	• Plasma from women obese or overweight	• Plasma from lean women	Obese/overweight	• There is an increase in exosome and PLAP+ exosome quantity over gestation, which is independent of BMI	[142]
				• There is a correlation between BMI and exosome quantity
				• No significant increase (12–20%) in PLAP+ exosomes throughout gestation
				• Exosomes were taken up by HUVECs and altered HUVEC cytokine release. Cytokine release changed based on the trimester and health status of the woman. Exosomes from obese plasma resulted in significantly more IL-6, TNF-alpha, and IL-8 release. No change in IL-10 was observed.
	• First trimester trophoblast cell-conditioned media incubated in high and low levels of glucose to mimic maternal hyperglycemia	• First trimester trophoblast cell-conditioned media from cells grown under varying oxygen concentrations	Hyperglycemia and hypoxia	• Exposure to hypoxia and high quantities of glucose resulted in increased exosome release • HUVECs released higher quantities of cytokines when exposed to exosomes isolated from cells grown under hypoxia and incubated in high quantities of glucose	[242]
	• HTR8/SVneo-conditioned media grown under normoxia and hypoxia	• Plasma from patients with normal pregnancies, PTB, or PE	PTB and PE	• Exosomes from cells grown under hypoxia negatively impacted endothelial cell wound healing • Unique miRNAs were identified in each condition and three miRNAs were only detected in exosomes from pathological conditions (hypoxic cells, PTB and PE)	[115]
				• Similar exosome miRNA cargo between cells grown under hypoxia and exosome miRNA cargo in early gestation maternal blood from PTB
	• Plasma collected throughout gestation from women who had a PTB	• Plasma of gestation age-matched, healthy women who had a term birth	PTB	• Altered exosomal miRNA cargo between term and PTB (throughout gestation) • Shows that alterations in miRNA expression are present early in gestation, which supports possible use of EVs as a screening tool for risk of PTB	[243]
	• Plasma from women who were undergoing preterm labor	• Plasma from healthy, gestation-age matched women	Preterm labor	• Small RNAseq was performed on whole plasma, EV-depleted plasma, and EVs • Pregnancy-associated miRNA cluster expression, including C14MC and C19MC miRNAs, were expressed at lower levels in preterm labor EV samples compared to healthy pregnancies	[116]
	• Plasma from women who experienced PTB or PPROM	• Plasma from healthy, gestation-age matched women	PTB	• No change in exosome quantity between the three groups • Significantly decreased levels of placental exosomes and different protein cargo in instances of PPROM compared to women with PTB or controls	[244]
Rodent	• Mouse PEVs (syncytin-positive) obtained from freeze–thaw injured placentas	—	PE	• Pregnant mice injected with PEVs had proteinuria and vascular leakage • Nonpregnant mice injected with PEVs resulted in hypertension and proteinuria • Enhancing EV clearance by treatment with the microvesicle-scavenging factor lactadherin prevented the development of this PE-like phenotype	[89]
				• Lactadherin −/− mice had elevated blood pressure, proteinuria, and fewer litters
	• HTR8/SVneo cells grown under hypoxic oxygen	• HTR8/SVneo cells grown under normoxic oxygen	—	• Hypoxia induces increased secretion of small EVs • Protein cargo was different between small EVs from hypoxic and normoxic cells	[226]
				• Injection of the small EVs from cells grown under hypoxic conditions into rats resulted in elevated blood pressure, but had no impact on fetal survival or placental size
	• Human first trimester placenta explant conditioned media	—	—	• Nanovesicles (vesicles smaller than 100 nm) altered the sensitivity of mesenteric artery dilation in pregnant, but not nonpregnant, mice • Vesicles did not alter uterine artery dilation or constriction	[245]
				• Human vesicles preferentially end up in mouse maternal lungs, liver, and kidneys while the feto/placental unit was negative
				• The uterus and cervix were not analyzed
	• Pregnant mouse plasma	—	PTB	• Quantity of exosomes increased with gestation	[221]
				• Mouse plasma exosomes trafficked to the cervix, uterus, fetal membranes, and placenta, and no exosomes were detected in other maternal or fetal tissues
				• Late gestation exosomes had increased quantities of proinflammatory proteins compared to nonpregnant or early gestation exosomes
				• Injection of late gestation exosomes into early gestation mice resulted in PTB
Cow	• Pregnant cow plasma	• Nonpregnant cow plasma	—	• Pregnant plasma exosomes suppressed the expression of the inflammatory cytokines TNF-alpha and IL-6 in BEND cells	[231]
				• Preferential location of miR-499 is in exosomes
				• Inhibition of miR-499 expression in mice resulted in increased uterine inflammation, EPL, and FGR
Pig	• Porcine trophectoderm cell line (PTr2)	• Porcine aortic endothelial cells (PAOEC)	—	• EVs from PTr2 influenced the proliferation of PAOECs • Uptake of heterologous EVs was more likely that autologous EV uptake	[148]
Sheep	• Pregnant sheep sera	• Nonpregnant sheep sera	—	• Exosomal miRNA cargo changed throughout gestation	[228]
				• C14 miRNAs decreased in quantity as gestation continued

MP, microparticle; eNOS, endothelial nitric oxide synthase; HUVEC, human umbilical vein endothelial cells; SGA, small for gestational age; PPROM, preterm premature rupture of membranes; C14MC, chromosome 14 microRNA cluster; C19MC, chromosome 19 microRNA cluster; TNF-alpha, tumor necrosis factor-alpha; IL-6, interleukin-6; HTR8/SVneo, extravillous trophoblast cell line, BEND cells, bovine endometrial epithelial cells; ST, syncytiotrophoblast; DPPIV, dipeptidyl peptidase IV; qPCR, quantitative real time polymerase chain reaction; CAMK2beta, calcium/calmodulin-dependent protein kinase II beta.

The cellular and molecular bases of pregnancy complications are difficult to address within the complexities of an in vivo pregnancy setting, owing to variability in genetics, lifestyle influences, physiology, and differing environmental exposure. Researchers examining human PEVs have focused extensively on characterizing EVs isolated from placental cell cultures (e.g., primary cells and immortalized trophoblast cell lines), placental explants, the perfusate of intact placentas, and maternal peripheral blood. PEVs isolated from placental cell cultures and explants can interact with maternal immune cells [59, 75–77], suggesting that PEVs may modulate maternal immune responses during pregnancy.

Current in vivo and in vitro sources for isolating PEVs

In vitro placental cell cultures provide a means to isolate an enriched PEV population. Stable cell lines can be expanded quickly to generate a homogenous cell population, allowing investigators to define a cell type-specific readout or secretory profile in response to an experimental manipulation (e.g., genetic mutation, infection, hypoxia, drug treatment). However, because some trophoblast cell lines are derived from spontaneously arising tumors or have been immortalized by molecular tools, their gene and protein expression profiles differ from each other and with primary trophoblast cultures [78, 79].

Primary cells provide a more accurate representation of in vivo trophoblasts and can be isolated from gestationally age-matched healthy and maldeveloped placentas for direct comparison. The ability to obtain trophoblasts from first and second trimester placentas, however, may be limited due to constraints surrounding human samples. Until recently, primary term villous cytotrophoblast cells were used only for short-term experiments as they do not proliferate and spontaneously syncytialize [80]. Recent optimization of primate trophoblast cell culture conditions now supports long-term cell proliferation and culture of trophoblast stem (TS) cells derived from primary placental cells or embryos that can be directed toward cell type-specific differentiation [81–84].

Placental explant cultures contain all cell types within the placenta (placental macrophages, cytotrophoblasts, and syncytiotrophoblasts) and therefore provide a more complete system compared to primary cell cultures. Limitations of this culture system include a limited duration of tissue viability [85, 86] and difficulty obtaining placentas from early pregnancies. It also is well accepted that pregnancy complications often stem from errors in early placentation. Therefore, it cannot be known if a complication would have arisen later in gestation when a placenta sample is obtained early in pregnancy. In addition, the large volume of medium used to culture explants can dilute the EV sample [86]. Alternative approaches include obtaining EVs by mechanical scraping the placental villi [61, 87] or perfusing fully intact term placentas [85]. Notably, vaginally delivered term placentas may be confounded by the process of labor.

Overall, in vitro models are useful because they can allow for the rapid generation of PEV-enriched samples, be used to validate antibodies, and provide physiological readout in response to trophoblast insult (i.e., exposure to toxins, hypoxia, or pathogens). These in vitro models also predominantly contain EVs secreted by trophoblasts, whereas maternal blood contains the totality of EVs released by all maternal as well as placental cell types. It is likely that the changes in plasma-derived EVs in pregnancy reflect the maternal adaptation to pregnancy by all maternal organ systems and not specifically placental development. While EVs from nonplacental sources during pregnancy may provide novel biological information, they do not specifically provide direct feedback about placental health per se. Previous studies have demonstrated that the source (bodily fluids, culture media, tissue homogenization) and sample preparation method influence EV function [87, 88]. Therefore, when establishing a PEV model or interpreting the study results, it is important to consider the origin of the PEV population being analyzed.

PEVs isolated from adverse human pregnancies

The analysis of human PEVs isolated from in vitro sources and maternal blood has revealed alterations in PEV cargo between healthy and unhealthy placentas. The isolation methods vary greatly across studies, thus limiting the ability to identify consensus biomarkers associated with an abnormal placental condition. A summary of the findings from studies that have evaluated EV cargo in human and animal pregnancy models in relation to a pregnancy complication is provided in Table 3, and here, we focus on the highlights of quantity and cargo of EVs associated with human pregnancy complications.

Elevated PEV levels offer potential as a way to distinguish complicated from healthy pregnancies. Increased quantities of EVs in maternal blood have been associated with various pregnancy complications as summarized in Table 3 [56, 89–93]. However, there are two concerns with these reports. First, it is unclear whether the research findings are specific to one complication or if they are applicable across pregnancy complications. If it is the latter, elevated EVs would be a general reflection of abnormal placental development and/or function. To determine the diagnostic value of EVs, a comprehensive analysis of EVs isolated from healthy pregnant women and women with a range of pregnancy complications is needed. Second, the techniques used to isolate and quantify (different instruments and/or different setting for the same instrument) EV samples vary widely, which dramatically hinders interstudy comparisons. Until standardized isolation and sample analysis techniques are implemented, variation in EV quantity will continue to be an uncertain indicator of a pregnancy’s health status.

Although human PEV cargo is not well characterized, EVs may contain evidence of placental infection and may serve as means of modulating immune responses. For instance, EVs isolated from trophoblast-conditioned media protected nonplacental cells from viral infection [94–96]. Viral proteins and genomes also have been detected within exosomes and infection was found to alter exosome cargo [60, 97]. As such, clinicians could examine PEVs to learn if a pathogen has breached the maternal–fetal interface as they could directly monitor placental response and health status.

PEVs interact with immune cells during pregnancy

Maternal immune systems inappropriately adapted to pregnancy are associated with pregnancy complications and pregnancy loss [98]. PEVs contain a range of immunoregulatory molecules [99, 100] and interact with maternal immune cells in vitro, which suggests that PEVs may be involved in maternal immune adaptation in pregnancy. EVs are involved in the recruitment of monocytes and macrophages as well as in cytokine and chemokine regulation [76]. Syncytiotrophoblast-derived EVs from healthy placentas suppress and/or promote immunological pathways [77]. Thus, understanding the interaction between the maternal immune system, the fetoplacental unit, and EVs is important. In addition, EVs from pregnant women impact immune cells differently than EVs from nonpregnant women [59, 75]. The dynamic complexities between PEVs and the immune system in a healthy and diseased state support the importance of in vivo models.

Differences in EV tropism for immune cells appear to depend on the source of the EV sample, that is, peripheral blood versus placental tissue. Germain et al. [87] observed strong binding of syncytiotrophoblast microvesicles from term placentas to monocytes in first trimester blood with decreased binding throughout gestation, as determined by enzyme-linked immunosorbent assay (ELISA). However, microvesicles from third trimester human placentas bound preferentially to monocytes and B-cells versus T and NK cells as assessed by Image Stream technology [61]. In contrast, microvesicles isolated from third trimester blood bound to T cells and not B or NK cells via fluorescence-activated cell sorting [59]. These studies co-incubated peripheral blood mononuclear cells (PBMCs) with EV samples in vitro. Since the methodologies used to isolate these EVs impacted tropism and downstream function, in vivo experiments will be important to study the interplay between PEVs and the immune system. EVs obtained from mechanical scraping of term placental villi did not stimulate PBMCs, whereas EV samples obtained from placental perfusate were more stimulatory [61, 87]. Notably, these EVs were obtained from term placentas and represent the end point of pregnancy. Pap et al. [59] found that 50% of microvesicles positive for human leukocyte antigen G (HLA-G) were also positive for Fas ligand (FasL). The authors hypothesized that these two molecules present on the EV surface are involved in maternal immune tolerance [59].

Surveying PEV markers and cargo to identify pregnancy complications

Classic EV markers, such as tetraspanins and ESCRT proteins [27], have been routinely identified in human and animal EVs, as listed in Table 2. As an adjunct to monitoring maternal systemic physiological changes (e.g., blood pressure or proteinuria), clinicians could directly monitor placental health and development by surveying PEVs. The application of PEVs as a prognostic molecular tool, however, is hindered by the current lack of validated placental biomarkers associated with a pregnancy complication in humans or animals. Furthermore, there are few validated placenta-specific EV surface markers to isolate PEVs. The most widely used placenta-specific marker for PEV isolation is placenta alkaline phosphatase (PLAP) [23, 24, 87, 101]. A PLAP ELISA has been used to quantify PEVs from human [25, 90, 102] and bovine samples [103]; however, validation and information regarding cross-reactivity of this antibody are lacking. Despite its use by several labs, there are challenges with specificity as other alkaline phosphatases have been detected in various healthy tissues that can be recognized by PLAP antibodies [104–106]. Validating these PLAP antibodies and making other in-house antibodies, such as NDOG2 and ED822 [74, 87, 107], commercially available will increase reproducibility across studies. HLA-G, another placenta-specific protein, has also been used to isolate and detect PEVs [24, 59], recognizing that this would be an extravillous trophoblast marker. Other placenta-specific protein candidates include syncytin-2, placental protein 13 (PP13), pregnancy-specific glycoprotein 1 (PSG1), and pappalysin-1 (PAPP-A) (Table 2).

Changes in PEV cargo highlight a difference between healthy and pathological placentas, as shown in Table 3. Cuffe et al. [108] discussed the presence of two classes of molecules: “passive” and “bioactive”. Passive molecules are hypothesized to have high predictive potential, whereas bioactive molecules are constitutively secreted by the placenta [108]. This further supports the use of PEVs to monitor placental health. PE is perhaps one of the more well-studied pregnancy complications in which PEV quantities and cargo have been evaluated [109]. Nair and Salomon’s recent review on human GDM [110] discussed systemic and placental changes in EVs. Several studies have focused on RPL and procoagulant microparticles/EVs secreted by platelets and endothelial cells [111–114]; however, we were unable to find any reports regarding the association of PEVs and EPL/RPL. Researchers have observed alterations in PEV cargo between women experiencing PTB compared to controls; however, they have not identified a consistent biomarker [115, 116]. For instance, one study only detected miR-525-5p in EVs from a pathological condition (PTB, PE, or cells grown under hypoxia) [115]. Another study found miR-525-5p to be significantly lower in PTB EVs compared to controls [116]. To understand the changes in PEV cargo associated with the cell’s physiological state, it is necessary to first understand whether EV cargo is selectively packaged. The survey of PEV cargo upon experimental knockout of lysosomal enzymes would provide insight into the critical question of cargo selection that spans all disciplines of the EV field.

The dearth of placenta-specific markers for PEV isolation and the lack of agreement upon biomarkers of a pregnancy complication highlight the need for consistent methodologies and nomenclature in studies that isolate PEVs from these various pregnancy complications. Identification of biomarkers resulting from aberrant placental development may allow for earlier diagnosis and intervention or earlier application of a placental therapy. PEV analysis could also enable better categorization of adverse pregnancy outcomes via specific molecular changes in the placenta rather than by less specific maternal symptoms and fetal measurements.

Animal models of human pregnancy complications

Most PEV research has been performed using human fluid samples and in vitro trophoblast cultures, with relatively few studies in animal models. Due to the complexities of obtaining and working with human samples and the limitations of vitro systems discussed thus far, other systems are needed to advance the study of PEVs. Animal models can provide the rigor and reproducibility that are difficult to achieve with human samples, due to uncontrolled external factors and genetic diversity among clinical patients. Advantages of collecting PEV data from animal models include access to large cohorts raised in controlled environments, rigorous sampling (i.e., the ability to collect samples early in pregnancy and at precise time points), control of the factor(s) causing the pregnancy complication (in some instances), the ability to utilize an animal as its own control during the same or subsequent pregnancy, the potential for longitudinal studies spanning preconception to the delivery of the offspring, and the opportunity for transgenerational studies.

Despite the differences in placentation across animal models, each animal model has strengths and shares similarities to humans that are useful when disentangling the mechanisms underlying pregnancy complications. When selecting an animal model to study a pregnancy complication, the following considerations should be addressed: placentation (i.e., depth of invasion, trophoblast cell organization, immune cell presence), animal husbandry for maintaining a study cohort (without/or without manipulation of an environmental factor), and the specific questions asked during the study (i.e., the importance of the fetus being born precocial; monotocous versus polytocous species). For example, the lack of endometrial trophoblast invasion by the pig placenta and minimal invasion in sheep has limited their utility in modeling PE. Although the pig placenta may not ideally model PE, its similarities in fetal development with humans could be beneficial to study FGR [117]. Furthermore, FGR studies in sheep and rabbit models provide an opportunity for the animal to serve as its own control by evaluating the unmanipulated, opposite uterine horn [5, 118–120]. An overview of placentation is depicted in Figure 2, and the details regarding pregnancy and translatability of each animal model discussed in this review are listed in Table 4 [121–125].

Figure 2.

Experimental models of human pregnancy. Laboratory animal models (NHP, guinea pig, rodent, and rabbit), large animal models (pig, sheep, and cow), and in vitro systems (placental perfusion, tissue/cell culture, and blood) are depicted to represent the experimental models of human pregnancy.

Table 4.

Comparison of placentation across different research models.

Model	Type of placentation	Gestation length (days)	Advantages	Limitations
Human	• Hemochorial	280	• Maternal plasma readily accessible	• Difficult to control environment factors
	• Discoid • Villous organization and extensive spiral artery remodeling • Interstitial extravillous trophoblast invasion		• Placenta samples available from a broad spectrum of adverse pregnancy outcomes • Diverse and extensive literature database • Well established in vitro systems (e.g., cell culture, explants, placental perfusion)	• Highly variable genetics • Restrictions to testing treatments/therapeutics • Precise timing of start of pregnancy can be uncertain • Delay in pregnancy detection limits the ability to obtain samples within first few weeks of pregnancy
			• TS cells available [81–83]
Macaque monkey	• Hemochorial • Bi-discoid placenta • Extensive endovascular extravillous trophoblast remodeling of decidual spiral arteries	165	• Ability to test treatments/therapeutics • Most appropriate model available for human placental physiology, immunology, and endocrine function at the maternal–fetal interface [121, 246, 247]	• Trophoblast interstitial extravillous invasion is superficial in comparison to human • Limited transgenic models • Specialized veterinary expertise and housing required
			• Placental architecture is highly translational [121] • Offspring are born precocial
			• Placental transfer of passive immunity
			• Placental expression of C19MC [203] and nonclassical MHC class I [247]
Guinea Pig	• Hemomonochorial • Discoid • Labyrinthine, invasive [141, 169, 248, 249]	67	• Offspring born precocial [141, 169, 249] • Blastocyst is completely encapsulated within the decidua, similar to human [248]	• Dearth of available antibodies • Lack of transgenic models
			• Passive immunity in late term [250]
			• Substantial trophoblast invasion [5]
			• Similar steroid production and metabolism in the decidua and fetal membranes as the human [169]
Rabbit	• Hemodichorial • Discoid • Labyrinth organization [125, 169]	32	• Fully sequenced genome [122] • Induced ovulator allowing for timed matings [122] • Offspring born precocial [172]	• Placental endocrinology is different than human [169] • Dearth of available antibodies
			• Passive immunity in late term [250] • Housing facilities are readily available
Mouse	• Hemotrichorial	20	• Facile manipulation of genetics [122]	• Offspring not born precocial
	• Discoid • Labyrinthine organization and some spiral artery remodeling [125]		• Ability to test treatments/therapeutics • Timed mating • NK cells present at MFI as with human [251] • Ability to perfuse mid and late gestation placentas [85] • TS cells available [85, 252]	• Placental organization, cell types, and endocrine profile differ compared to humans [253–255] • PLAP is not expressed • Blood flow to the placenta is more limited than in human [121] • Lack of nonclassical MHC expression [251, 256]
				• Shallow implantation compared to rat or human [251]
				• Placental expression of C19MC miRNA is not conserved [203]
				• TS cell isolation and propagation differ from primate [81–84, 257]
				• Murine cytotrophoblast cells are in direct contact with maternal blood, whereas syncytiotrophoblasts are in direct contact in the human [98]
Rat	• Hemotrichorial	22	• Nonclassical MHC expression [251, 256]	• Offspring not born precocial
	• Discoid • Labyrinthine organization • Deep placental implantation [251, 258]		• NK cells at MFI as human [251] • larger size compared to mouse allows practical advantages (tissue availability, surgical procedures) • TS cells available [85, 259]	• Different placental organization and cell types compared to humans [258] • Placental expression of C19MC miRNA is not conserved [203]
				• Less extensive genetic technology and antibody development compared to the mouse [122]
				• Different endocrine profile than humans [255]
Sheep	• Epitheliochorial (synepitheliochorial)	150	• Relatively few offspring per liter [5, 260]	• Minimal trophoblast invasion [260]
	• Cotyledonary [250]		• Offspring born precocial [122] • Large blood samples and surgical manipulations including chronic instrumentation of the fetus are feasible	• Practical limitations on housing of research animals and length of gestation • Different placental cell types and endocrine profile than humans [255]
Cattle	• Epitheliochorial (synepitheliochorial)	280	• Sites of nutrient and waste exchange are villous [124]	• Minimal trophoblast invasion [260]
	• Cotyledonary • Partially nondeciduate [103, 261, 262]		• Macrophages located at the maternal–fetal interface at low levels during the first two-thirds of pregnancy and increase substantially by term [261] • Nonclassical MHC expression towards the end of pregnancy [263]	• Practical limitations on housing of research animals and length of gestation • Different placental cell types and endocrine profile than humans [255]
			• TS cells available [264]
			• Large blood samples and surgical manipulations are feasible
Pig	• Epitheliochorial • Diffuse [262]	114	• Fully sequenced genome [260] • Placental attachment is superficial and interdigitates with the highly folded maternal endometrium • TS cells available [265] • Large blood samples and surgical manipulations are feasible	• Fetal nutrition is predominantly acquired through uterine gland secretions • Passive immunity does not occur until after birth [250] • Unlike humans, there is no syncytiotrophoblast cell type [124] • Practical limitations on housing of research animals and length of gestation

Animal models allow for longitudinal study to identify early biomarkers prior to or at the onset of a complication, which is not possible in humans due to the delays in confirming pregnancy and the need to avoid perturbing early gestation. Identifying changes or biomarkers at the onset of a complication, even before placentation is complete, may be important for diagnosing or preventing a further adverse outcome. Once biomarkers have been identified in an animal model, they can then be validated for tissue specificity and reproducibility within and across species for subsequent translation to humans. For example, miR-210 is more highly expressed in the placentas of mice with PE compared to healthy placentas. It is also expressed in human and ovine placentas [126, 127] with aberrant regulation in human PE and upregulation of the miRNA in hypoxic human placentas. Thus, animal models have led to the identification of a putative biomarker of PE, where upon further refinement of the timing of aberrant expression and identification of target mRNAs may reveal the biological processes contributing to the manifestation of the placental complication.

Animal studies will have an essential role in showing proof of principle for the potential of a therapeutic intervention or diagnostic assay prior to translation into human clinical studies. For example, pregnant guinea pigs and sheep were used to test an experimental placental treatment, in which injection of an adenoviral vector overexpressing VEGF reduced FGR [128, 129]. Due to the positive outcomes in these animal models, this therapy subsequently was transitioned into a clinical trial in Europe (EVERREST project) [130]. Further use of animal models will not only enable the development of improved diagnostics, but they also can provide a platform for developing and evaluating the efficacy of placental therapies [120]. In these animal models, a thorough understanding of the complication as well as the safety of a therapeutic can be appropriately evaluated prior to transition into a clinical trial; however, there must be a thorough understanding of the etiology of that pregnancy complication. The guinea pig and macaque, for example, are both ideal for placenta-targeted therapies as they similarly share a hemochorial type of placentation, bypassing additional maternal layers present in livestock species. Thus, a common workflow in placental research, and other fields, is to identify in rodents, verify in nonhuman primates (NHPs), and then translate to humans. This approach can be implemented in the PEV field by broadly utilizing various experimental animal pregnancy models.

Spontaneous versus induced animal models of adverse pregnancies

While pregnancy complications can develop spontaneously in animal models, they are frequently induced. Similar to the selection of the animal model, there are distinct advantages and limitations for choosing between a spontaneous or induced model to study adverse pregnancy outcomes in animals. For the sake of brevity, we have collated the induction methods of a complication for the various species that are listed in Table 5.

Table 5.

Induced animal models of human pregnancy complications.

	Induction mechanism	NHP	Rabbit	Mouse/rat	Guinea pig	Pig
	Artery ligation	• Abdominal aortic constriction and uterine artery ligation led to hypertension and proteinuria with renal damage and/or impaired function [153, 154, 266]	• Constriction of the maternal aorta below the renal arteries led to proteinuria, kidney damage, and fetal demise [267]	• RUPP led to elevated TNF-alpha levels, hypertension, proteinuria, and FGR [268]
PE	Other	• TNF-alpha injection resulted in proteinuria, hypertension, and elevated sFLT-1 levels [269]	• Fetal hemoglobin injection resulted in proteinuria, fetal demise, and increased apoptosis in the kidneys and placentas [173]	• sFlt-1 injection resulted in hypertension and proteinuria [270] • TNF-alpha infusions resulted in hypertension [271] • Injection of sEng and sFlt-1 [272] • Modified nitric oxide production resulted in hypertension and FGR [273, 274]	• Diet restriction led to ketosis, and minor PE symptoms and pathology [131]	• Fasting resulted in HELLP syndrome, hypertension, and proteinuria [275–277] • Glucose restriction induced hemolysis and increased levels of free heme [278]
	Induction mechanism	NHP	Sheep	Pig	Cow
EPL/RPL	Other	• Low protein diet resulted in poor placental perfusion, miscarriage, and FGR [155]	• Lower insulin and progesterone levels were reported in pregnancy loss during days 18–40 of gestation [279]	• 20–45% implantation failure [136, 138, 139] • Defects in vascular remodeling, inflammation, and altered miRNA expression have been associated with porcine embryonic death [138, 139]	• ~40% implantation failure [135–137] • Genotyping of fetuses that died in utero, and genetic selection of sperm donors provided insight as to genes associated with EPL [135, 137] • EPL is also associated with infection and uterine inflammation [136] • Progesterone levels prior to implantation may contribute to the establishment and maintenance of pregnancy [135]
	Induction mechanism	NHP	Guinea Pig	Rabbit	Mouse/Rat	Sheep	Pig
	Uteroplacental vascular modification	• Ligation of the vessels connecting the primary and secondary discs led to placental insufficiency, fetal death, and FGR [156]	• Uterine artery ligation did not result in fetal demise, but did result in smaller fetuses and impaired organ function [280] • Ablation of uterine artery branches leading to individual placentas [281] and gradual ligation techniques [282] can remain in place for long periods of time and do not result in high fetal mortality	• Uterine artery ligation negatively impacted kidney development in the fetuses [283], which is also seen in human cases of IUGR [284] • Artery ligation resulted in a high prevalence of fetal mortality [118, 119] and negatively impacted renal development [119] • A limitation of uteroplacental vessel ligation is the necessity of a short study duration due to high fetal mortality [285]	• Bilateral uterine artery ligation resulted in asymmetrical IUGR and negatively impacted kidney development in the fetuses [283] • Has been shown to result in altered insulin levels, IGF and IGF binding protein levels, decreased placental weight, hypertension, and delayed growth [120] • A limitation of uteroplacental vessel ligation is the necessity of a short study duration due to high fetal mortality [285]	• Uteroplacental embolization [134, 286] • Single umbilical artery ligation [134, 286]
FGR/IUGR	Other	• Experience placental insufficiency and IUGR similar to humans [121, 155]	• Hypoxia impaired placental development and invasion and resulted in hypertension [287]	• Natural model due to the large litter size (embryo crowding within the uterus can impact growth) [5] • See review [285]	• Injection of sEng and sFlt-1 resulted in FGR [272] • See reviews [120, 288]	• Heat stress decreased placental blood flow and led to IUGR [134, 286] • Sheep living at high elevations develop IUGR [134] • See reviews [117, 289]	• Duration of nutrient restriction is correlated to the severity of IUGR [117] • A limitation of nutrient restriction is lack of hyperleptinemia, as is observed in humans and rats [290] • See review [117]
	Induction mechanism	NHP	Guinea Pig	Rabbit	Mouse/Rat	Sheep	Pig
Pre-GD/GDM	Chemical induced	• Alloxan and STZ [157, 158]	• STZ [291]	• Alloxan and STZ [171, 292]	• Alloxan and STZ [293–296]	• STZ [297, 298]	• STZ [299] did not significantly alter fetal size at birth
	Other		• Useful to model GDM: as the humans, the placenta and fetus are reliant on maternal supply of glucose until late in gestation, which is also the case in humans [249] • Low preconception vitamin D status increases the risk of developing GDM [300]		• Stress, nutritional, and drug manipulations accurately recapitulate insulin resistance, hyperlipidemia, increased inflammation, fetal death, and macrosomia [301] • Limitation: dams had decreased body weight during pregnancy, which is not seen in human GDM [301] • A transgenerational mouse model resulted in metabolic dysfunction, high glucose levels, and lower beta-cell mass in late gestation [302] • Rats fed a high fat, high sugar diet instigated increased glucose and insulin levels but decreased insulin sensitivity. Offspring were not born macrosomic, potentially due to the immature state of rats at birth [303]	• Useful to model GDM, as the placenta and fetus are reliant on maternal supply of glucose until late in gestation, which is also the case in humans [249] • A high fat diet resulted in increased levels of insulin, glucose, and cortisol in the ewe and fetus compared to controls [304]
	Induction mechanism	NHP	Guinea pig	Rabbit	Mouse/rat	Sheep	Cow
PTB	Infection/Inflammation	• Intra-amniotic injection of bacterial proteins, IL-1beta, and TNF-alpha resulted in PTB [169]		• Intrauterine administration of bacterial agents and inflammatory cytokines resulted in PTB [305, 306]	• Inoculation of rodents with inflammatory cytokines and alarmins present during human PTB resulted in PTB [307–310] • Intrauterine administration of bacterial agents and inflammatory cytokines led to PTB in mice [311–313], which had comparable inflammatory cytokine profiles to those observed in human infection associated PTB [311, 314]	• The ability for intra-amniotic modeling of infection as opposed to intrauterine is of great utility, [316, 317] • Intrauterine administration of bacterial agents and inflammatory cytokines [318–320]	• Candida infection resulted in severe placental pathology and is associated with abortion in cows [321] and PTB in women [322–324]
					• Intrauterine inflammation can promote myometrial increases of contraction-associated proteins and trigger cervical remodeling in rats [315]
	Parturition	• Progesterone withdrawal is not a prerequisite for labor, similar to humans [122, 169, 325]	• Increasing progesterone levels in late pregnancy similar to humans [169, 326, 327]	• Labor is dependent on progesterone withdrawal [169, 327]	• Labor is dependent on progesterone withdrawal [169, 326, 327]	• Labor is dependent on progesterone withdrawal [169, 326, 327]
	Other		• Naturally experience preterm birth (~7% compared to 5–11% in human) [169, 170]		• Genetic contributions to PTB in humans have been elucidated through mutant mouse models, such as Calmus et al.’s work on Ehlers–Danlos syndrome [328] • Additional noninflammatory molecules induce PTB in mice [329, 330] • Increased hyaluronan expression has been observed in human and mouse PTB regardless of infection [329, 331]	• Glucocorticoids have been shown to induce preterm labor [318–320]. However, in humans, glucocorticoids are actually used as a treatment in PTB pregnancies to accelerate fetal lung development [332]

Acronyms: nonhuman primate (NHP), soluble Endoglin (sEng), soluble fms-like tyrosine kinase-1 (sFlt-1), hemolysis, elevated liver enzymes, and low platelets (HELLP), insulin growth factor (IGF), streptozotocin (STZ), Reduced uterine perfusion pressure (RUPP).

Spontaneous development of a pregnancy complication in animals is valuable as it suggests there may be mechanistic overlap with humans. Although PE is observed primarily in humans, spontaneous cases have been documented in mice, rats, rabbits, guinea pigs, and monkeys [122, 131, 132]. Here, we briefly describe a few examples that support the use of an animal model for various complications. In general, litter-bearing species can serve as natural models of FGR and IUGR [5, 117, 133]. For example, spontaneous IUGR occurs in 15–20% of swine [134]. Likewise, spontaneous pregnancy loss is common in cattle (~40%) [135–137], pigs (20–45%) [136, 138, 139], and marmoset monkeys (26%) [140]. In a cohort of guinea pigs, 20% of pregnancies spontaneously developed toxemia, and the observations from those animals validated their induced toxemia model [131]. Decreased reproductive efficiency has been observed in both humans and sheep at high altitudes; however, this environmental factor would limit the ability to study FGR in sheep to specific locations and might limit the relevance of such studies to the general human population. While spontaneous instances of these pregnancy complications can be used for experimental modeling, by definition, spontaneous complications are difficult to predict. Events preceding the adverse outcome cannot be efficiently studied and may require larger animal cohorts or specific conditions than is practical for many investigators.

In contrast to the uncertainty in occurrence and timing of a spontaneous complication, researchers can administer precise insults or experimental treatments to control the induction of a pregnancy complication (Table 5). Pregnancy complications may be induced by drug treatment, diet, surgery, or genetic manipulation. Genetic manipulation is commonly used in rodent models to induce a pregnancy complication, where a gene knockin or knockout can aid in further investigating causative genes underlying the development of a complication. Information derived from these genetic mutations then can be translated to other animal models, such as NHPs, that more closely model human pregnancy. The biological relevance of the induced complication must be determined on a species and approach basis. As with any laboratory study, there are limitations to the comparisons that can be made to natural cases of disease. Data gleaned from induced models should be compared to data obtained from spontaneous cases of pregnancy complications whenever possible.

Etiology is important when selecting an approach to induce a pregnancy complication, as the mechanisms impacted may not be translatable to humans. For example, diet-restricted guinea pigs displayed similar symptoms and pathology as those with spontaneous PE [141]; however, the etiologies appear different. Spontaneous PE resulted from uteroplacental ischemia induced by aortic compression caudal to the renal arteries. In contrast, fasting-induced PE led to ketosis and resulted in less severe symptoms and pathology [131]. Notably, similar symptoms of varying severity were observed in this study, and it is important to consider the mechanistic differences underlying the PE symptoms observed as they may relate or differ from the human pathogenesis of the complication.

Animal models of experimental infections during pregnancy

Pathogen infection of a host is species restricted, so ensuring that a pathogen can induce similar pathophysiology in an animal model of human pregnancy is essential to better understanding the downstream implications. Other important considerations for congenital infection models include the route of infection, maternal symptoms, fetal/congenital symptoms, and the role the maternal immune system plays in fighting the infection. Researchers have used in vitro animal and human placental cell culture systems to identify the cell types most susceptible to vertically transmitted pathogens and to unravel the mechanisms behind infection [20]. While in vitro systems have aided in understanding the cellular mechanisms of vertical transmission (e.g., receptors that mediate pathogen trophoblast entry), these mechanisms largely remain elusive. The various animal models that have been used to model TORCHZ infections during pregnancy are broadly summarized in Table 6.

Table 6.

Pathogens in animal models that cause APOs.

Pathogen	Animal model	Observed APO
Brucella	Cattle [333]	Pregnancy loss
	Sheep [333]
Chlamydia	Sheep [334]	Pregnancy loss
	Mice [334, 335]
	Guinea pigs [336]
	Pigs [337]
Cytomegalovirus	Mice [338]	Neurological sequelae, pregnancy loss
	Guinea pigs [338, 339]
	NHP [159, 160]
Hepatitis E virus	Rabbit [340, 341]	Pregnancy loss
Group B Streptococcus	Mice [342] NHP [343]	Pregnancy loss, meningitis, pneumonia, neurological developmental disabilities
Listeria monocytogenes	NHP [161–163]	Pregnancy loss
	Guinea pigs [254, 344]
Rubella	Rats [345, 346]	Pregnancy loss, neonatal demise, ocular abnormalities
Toxoplasma gondii	Sheep [347, 348]	Pregnancy loss
	Mice [349]
	NHP [350]
Zika virus	NHP [164–166]	Pregnancy loss, fetal malformations
	Mice [351]
	Guinea pigs [352]
	Pigs [353]

APOs, adverse pregnancy outcomes.

Current knowledge of PEVs in animal pregnancy models

This section provides a brief overview of PEV studies that have been performed using mouse and livestock pregnancy models, with more details presented in Table 3. Similar to humans [102, 142], the total number of exosomes and PLAP-positive vesicles isolated from blood increased throughout bovine gestation (Table 3) [103]. Sequencing of miRNAs from bovine exosomes revealed unique expression profiles across trimesters [103]. The placental miRNA profile also changes throughout gestation in humans [143] and rhesus macaques [144]. These data suggest that despite minimal placental invasion in the cow, placental exosomes similarly circulate in the maternal bloodstream as observed in humans. Thus, there may be conservation in marker expression and function of cargo given the similarities in EV miRNA profiles during gestation.

Data collected from mouse and human studies support that EV clearance may impact pregnancy health status. Excess vesicles in lactadherin^−/− pregnant mice resulted in elevated blood pressure, proteinuria, and fewer litters, suggesting that EV clearance opposes the development of PE-like symptoms (Table 3) [89]. Moreover, elevated levels of PEVs or PEVs from injured murine placentas can induce PE symptoms when injected into pregnant mice [89]. This impact supports and expands upon human data. An in vitro human trophoblast culture study found that less syncytin-1 and -2 cellular expression resulted in decreased exosome uptake and, thus, an excess of released EVs [145]. Interestingly, placentas from women with PE expressed less syncytin-1 and -2 than controls [145]. Germain et al. [87] also reported elevated levels of circulating free and fewer bound syncytiotrophoblast microvesicles in patients with PE than healthy subjects. The overlap in findings support the use rodent models can offer in terms of unraveling the impact of EVs.

Observations from rodent pregnancy models have revealed that PEVs may serve as a means of communication at the maternal–fetal interface. In mice, fetal and maternal exosomes trafficked across the maternal–fetal interface and fetal exosomes impacted maternal cell function [146]. Similarly, PEV trafficking to maternal cells also has been shown in large animal pregnancy models. A recent study showed that the binucleate trophoblast cells of the ruminant placenta also secrete exosome-like vesicles to the maternal uterine epithelium and connective tissue [147]. Similarly, in vitro study showed that porcine trophectoderm cell lines secreted EVs that influenced the proliferation of porcine aortic endothelial cells (Table 3). This supports fetomaternal cross-talk in the pig [148], a phenomenon similarly observed in sheep [149]. In addition, EVs isolated from ovine uterine fluid are taken up by embryos/trophectodermal cells and vice versa [149, 150]. EV uptake by these cells suggests that maternal–fetal communication occurs very early in pregnancy and shows the potential to assess embryo-derived PEVs as early as the pre/peri-implantation stages. Investigation of the earliest stages of development could reveal how these EVs may be altered in EPL.

While the results of in vitro PEV studies are intriguing, the prevalence of PEVs in ovine and porcine maternal blood remains unclear. PEVs are expected to be present in ovine maternal circulation as they have been detected in bovine maternal blood, and these species share similar placental architecture. Further isolation of PEVs from all stages of pregnancy, including embryo-derived EVs, across livestock species will help to understand complications that arise from errors in the establishment of pregnancy as well as maldeveloped placentation. Although there are few published animal PEV studies, the similarities in findings between animal models and human in vitro data further support the need for additional in vivo animal studies.

Future perspectives: expansion of PEV research in animal pregnancy models

The studies discussed in the previous section represent all current, but limited, publications on PEVs in animal pregnancy models. Studies with animal models provide an opportunity to improve our understanding of the consequences of placental complications through comprehensive study of PEVs and their cargo. Additional studies in the animal models discussed above, especially in those with a hemochorial placenta, are needed to identify biomarkers and expand our knowledge of PEV cargo and function. The development of PEV animal models is especially important to elucidate the impact EVs have on the maternal immune system and maternal physiology in healthy and complicated pregnancies, as this cannot be studied in vitro.

The use of NHPs in PEV research would be particularly valuable as there are extensive similarities between humans and NHPs as listed in Table 4. Macaques share a similar hemochorial type of placentation with extensive remodeling of decidual spiral arteries by endovascular trophoblasts [121, 123]. Unlike rodent models but similar to humans, NHPs express miRNAs from the primate-specific chromosome 19 microRNA cluster (C19MC) (Table 4). miRNAs of the C19MC are almost exclusively expressed in the placenta and have been detected within human EVs [35]. The C19MC miRNAs have roles in placental function and are aberrantly expressed in pregnancy complications [151, 152]. Several pregnancy complication paradigms are already in place with NHPs [121, 153–167]. Applying the study of PEVs to these established models will enable advances not feasible in human pregnancy research—for example, monitoring vertical pathogen transmission by PEV analysis with timed infection studies. Investigators have recently characterized and validated rhesus and cynomolgus macaque TS cell lines that can be differentiated into syncytiotrophoblasts and extravillous trophoblasts [84, 168]. These cell lines may have tremendous value in terms of identifying PEV biomarkers of infection and disease.

Guinea pigs may also offer utility in PEV research as they have a discoid, hemomonochorial, labyrinthine placenta and are relatively low cost compared to NHPs. Their longer gestation (~68 days) compared to the mouse and rat (~20 and ~22 days, respectively) allows for enhanced longitudinal sampling. They also naturally experience PTB at approximately 7% rate (the human rate is 5–11%) [169, 170]. Hence, guinea pigs may be a useful model for biomarker identification as well as drug development as an intermediate model between rodents and NHPs.

The use of rabbits could be beneficial as they are induced ovulators [122, 171], which allows for early and precisely timed pregnancy sample collection. Rabbits, like humans, have a syncytial trophoblast layer [122], and their genome has been fully sequenced. Their relatively short gestation (~32 days) allows for short studies that can assess the impact of pregnancy on the fetus as the offspring are born precocial [172], a feature more similar to humans than rodents. A representative example of the rabbit model being used to understand a pregnancy complication is a study in which injection of fetal hemoglobin resulted in proteinuria, fetal demise, and increased apoptosis in the kidneys and placentas [173]. This study helped elucidate the impact fetal hemoglobin may have on PE in humans and showed the efficacy of alpha1-microglobulin (A1M) as a therapy to alleviate PE-like symptoms [173].

There are additional advantages and limitations of a model that are particularly relevant to PEV biomarker identification. Animal models with smaller litter sizes, such as the macaque or the sheep, allow for more focused biomarker detection for singleton pregnancies as in humans. In animals with large litters, some fetuses may normally develop, while others are resorbed. If healthy fetuses are present, it may be difficult to parse out and identify a biomarker of the pregnancy complication. Animals that mature more quickly typically have shorter gestations and allow for transgenerational study design. Moreover, animal pregnancy models are advantageous as they provide the ability to survey PEVs in relation to fetal growth and development over time, as well as in association with offspring physiology throughout their lifespan in a manageable time frame.

Next steps in establishing animal pregnancy models for PEV research

A major limitation in the use of PEVs to diagnose pregnancy complications is the lack of information regarding early, predictive markers [108], which makes the identification of longitudinal markers difficult. As illustrated in Table 2, there is overlap between general human EV markers and those of the animal models; however, only a few studies have specifically looked at PEVs in animal models despite similarity in some placental markers. For PEV research to be translational, we propose the following goals:

1. Rigorous assessment of placenta-specific markers in longitudinal in vivo studies, in additional cohorts for repeatability, and across species to ensure translatability.

2. Validation of antibodies that are subsequently made commercially available for use across labs and species (when applicable).

3. Thorough assessment of the prognostic potential of a biomarker associated with a pregnancy complication in various animal models.

4. Development of a database for placenta-specific markers and biomarkers of pregnancy complications using high-throughput techniques (i.e., next-generation RNA sequencing, mass spectrometry for proteomics, and lipidomics).

5. Preclinical evaluation of a biomarker associated with a human pregnancy condition in human samples (retrospective and prospective studies).

Animal models will enable the development of datasets with predictive markers because researchers will have control over sample collection and the timing of the insult. Researchers can also use the established animal models to determine the predictive power of potential biomarkers that were identified in humans [108]. The harrows of identifying a single biomarker in humans also support the need for induced pregnancy complication studies because a marker consistently identified across species with varied disease severity has translational potential. Overall, animal models can greatly strengthen the PEV field in terms of studying and prospectively identifying pregnancy complications.

Current questions and future opportunities in the PEV field

Having discussed the opportunities in pregnancy complication research with a range of animal models, there remain questions in PEV research that are cross-cutting, regardless of the species used, including research with human clinical samples.

• ▪ Is PEV cargo selectively packaged?—If so, how?

• ▪ How are the presence of PEV membrane proteins and cargos altered when EVs are derived from a diseased placenta?

• ▪ Do the embryo and the placenta use PEVs to communicate to maternal cells before and during implantation?

• ▪ What role(s) do PEVs have in regulating maternal immune adaptation to pregnancy? Are PEVs essential for the successful establishment of pregnancy?

• ▪ Do PEVs from a maldeveloped placenta cause or contribute to a pregnancy complication by triggering a maternal physiological or immune response, or is their presence a manifestation of the impact of the pregnancy complication on placental function?

PEVs hold the promise of future prognostic and diagnostic development as they can provide high “clinical predictive power” [101] for pregnancy complications. Unraveling the mechanisms of cargo packaging is crucial to understand how EV cargo of a malfunctioning placenta may be altered in comparison with those derived from a healthy placenta. There is currently a debate in the EV literature as to what “exosomes” are, and whether these truly can be isolated from a complex EV population [32]. Consistent nomenclature and standard techniques to isolate EVs would allow comparison among studies. There are currently three EV databases, EVpedia [174], ExoCarta [175], and Vesiclepedia [176]; however, a database specifically for PEV research would enable meta-analysis of results, allow for marker identification reported across pregnancy complication research, and enable the field to quickly advance. Since placental development is continuous and gene expression changes throughout pregnancy, it is important that investigators develop a database of EV cargo from all stages of pregnancy. The NIH Human Placenta Project would be an excellent platform to support such a database.

In conclusion, representative in vitro and in vivo animal models are necessary to identify biomarkers of pregnancy complications. A better understanding of PEV biology will allow deeper insights into placental function and development throughout gestation, help to identify maldeveloped and/or infected placentas, and potentially underpin development of placental therapeutics. We propose that in order to achieve these advances, appropriate animal models of human pregnancy complications must be established.

Supplementary data

Supplementary data are available at BIOLRE online.

Conflict of interest

The authors have declared that no conflict of interest exists.