Maternal circulating miRNAs that predict infant FASD outcomes influence placental maturation

Maternal gestational circulating microRNAs, predictive of adverse infant outcomes, including growth deficits, following prenatal alcohol exposure, contribute to placental pathology by impairing the EMT pathway in trophoblasts.


Introduction
Prenatal alcohol exposure (PAE) is common (1,2,3). Between 1.1% and 5% of school children in the United States are conservatively estimated to have a fetal alcohol spectrum disorder (FASD) (4). Consequently, FASD, due to PAE, is the single largest cause of developmental disabilities in the United States and worldwide (5) and a comorbid factor in a number of other prevalent developmental neurobehavioral disabilities, including attention deficit/ hyperactivity and autism spectrum disorders (6).
PAE can result in decreased body weight, height, and/or head circumference in infants. Consequently, infant growth deficits are a cardinal diagnostic feature for fetal alcohol syndrome (7), which represents the severe end of the FASD continuum. However, although well recognized as a diagnostic feature, the mechanistic linkage between PAE and growth restriction remains unclear. In 2016, as part of our effort to identify maternal diagnostic biomarkers of the effect of PAE, we reported that elevated levels of 11 distinct miRNAs in maternal circulation during the second and third trimesters distinguished infants who were affected by in utero alcohol exposure (heavily exposed affected [HEa]) from those who were apparently unaffected at birth by PAE (heavily exposed unaffected [HEua]) or those who were unexposed (UE) (8). In that study, we predicted, based on bioinformatics analyses, that these HEa  Placental development involves maturation of cytotrophoblasts at the tips of anchoring villi into invasive extravillous trophoblasts, as well as fusion of cytotrophoblasts into multinucleate, hormoneproducing syncytiotrophoblasts (9). Maturation into extravillous trophoblasts, which invade the maternal decidua and remodel the uterine spiral arteries into low-resistance high-flow vessels that enable optimal perfusion for nutrient and waste exchange, requires cytotrophoblasts to undergo EMT (10). Impaired placental EMT, as well as orchestration of the opposing mesenchymal-epithelial transition pathway, has been found in conditions resulting from placental malfunction, primarily preeclampsia (11,12,13,14,15,16). Although there have been no previous studies directly investigating the effects of PAE on placental EMT, a rodent study demonstrated that PAE, during a broad developmental window, reduced the number of invasive trophoblasts within the mesometrial triangle, a region of the uterine horn directly underlying the decidua (17). Furthermore, both human and rodent studies have found PAE disrupts placental morphology and interferes with cytotrophoblast maturation, as with preeclampsia (18,19,20,21). Disrupted trophoblast maturation, seen in these conditions, is associated with aberrant expression of placental hormones, primarily human chorionic gonadotropin (hCG) (22,23,24,25).
Our study is the first to report that PAE interferes with expression of core placental EMT pathway members. Using rodent and primate models of gestation, as well as complementary miRNA overexpression and knockdown studies in vitro, we also provide evidence that HEa miRNAs, which predict infant growth deficits due to PAE, collectively but not individually, mediate PAE's effects on placental EMT through their effects on cytotrophoblast maturation and cellular stress. In a mouse model of pregnancy, a single combined exposure to the murine-expressed HEa miRNAs resulted in placental EMT inhibition and diminished placental and fetal growth. Collectively, these data suggest that elevated HEa miRNAs may represent an emergent maternal stress response that triggers fetal growth restriction, although subgroups of HEa miRNAs may compete to protect against the loss of EMT. Moreover, most members of the group of HEa miRNAs have also been implicated in other placental insufficiency and growth restriction syndromes, giving rise to the possibility that growth restriction syndromes may share common etiological mediators.

HEa miRNAs explain variance in infant growth outcomes due to PAE
Given the association of individual HEa miRNAs with gestational pathologies, we sought to determine if circulating HEa miRNA levels could explain the variance in sex and gestational age-adjusted neonatal height, weight, and head circumference in our Ukrainian birth cohort, which are growth measures sensitive to in utero Inset colored circles represent the corresponding sex and gestational age-adjusted growth parameters these miRNAs were correlated with. Of the 22 studies queried, 11 (50%) used unbiased screenings for miRNA expression. (B, C) Heat map of mature HEa miRNA expression (B) and pri-HEa miRNA expression (C) across different tissues resulting from secondary analysis of publicly available RNA-sequencing data. Legend depicts row-centered Z-score. environment (57). We found that eight of the HEa miRNAs each significantly explained between 7% and 19% of infant variation in these growth measures (Table 1). Furthermore, seven of these miRNAs were also associated with fetal growth restriction and preeclampsia as identified by our literature review (Fig 1A). Interestingly, a multivariate statistical regression model that accounted for levels of all 11 HEa miRNAs together, explained a far greater proportion of infant variance, between 24% and 31%, in all three growth measures than accounting for them individually (Table S1), suggesting HEa miRNAs collectively account for the variance in infant growth outcomes.
HEa miRNAs are transcribed preferentially in the placenta Data extracted from publicly available gene expression profiling datasets (58) show that HEa miRNAs and their unprocessed precursor transcripts, HEa pri-miRNAs, are enriched in placenta compared with other tissues, suggesting that the placenta itself transcribes these miRNAs and may be a significant contributory tissue to maternal circulating HEa miRNAs (Fig 1B and C). Moreover, because HEa miRNAs are also associated with gestational pathologies caused by poor placental invasion, these HEa miRNAs may also contribute to the placental response to PAE. We, therefore, assessed in rodent and primate models, whether PAE could result in impaired EMT, and if HEa miRNAs could explain the effects of PAE on placental EMT-associated gene expression.

HEa miRNAs moderate placental EMT impairment in PAE models
EMT, in trophoblasts, is characterized by the disappearance of epithelial markers such as E-Cadherin and the appearance of mesenchymal markers such as the intermediate filament, vimentin, a process that is controlled by the expression of key mesenchymal determination transcription factors, Snail1 and 2 and TWIST, as extensively described (10,14,15,59,60,61,62). These five markers have been used to assess EMT in a variety of model systems, so our studies used these markers to assess the effects of alcohol and HEa miRNAs on trophoblast EMT.
In the first analysis, using a murine model of PAE that mimicked moderate to binge-type alcohol consumption throughout early and mid-pregnancy, we fractionated GD14 placenta into three zones: the cytotrophoblast-and syncytiotrophoblast-rich labyrinth zone, the glycogen-and spongiotrophoblast-rich junctional zone, and the decidual zone comprising the endometrial contribution to the placenta (Fig 2A). Multivariate analysis of variance (MANOVA) for expression of these five core genes in the EMT pathway within placental trophoblasts revealed a significant effect of ethanol exposure on EMT pathway member expression selectively within the labyrinth zone (Pillai's trace statistic, F (5,21) = 6.85, P < 0.001, Fig 2B) but not within the junctional or decidual zones. Post hoc univariate ANOVA indicated ethanol exposure specifically elevated CDH1 (F (1,25) = 7.452, P = 0.011), which encodes epithelial E-Cadherin, whereas expression of the pro-mesenchymal transcription factor SNAI1, which encodes Snail1, was significantly reduced (F (1,25) = 21.022, P = 0.0001). We also observed a significant interaction between fetal sex and PAE on expression of SNAI2, which encodes Snail2 (F (1,25) = 2.18, P = 0.047) and a trend towards decreased expression of the terminal mesenchymal marker VIM (vimentin, F (1,25) = 2.749, P = 0.11), whereas there was no effect on TWIST expression (Fig 3A-E). Consistent with our gene expression data, E-Cadherin protein levels were significantly elevated in the labyrinth zone of PAE placenta (F (1,24) = 31.63, P = 0.0005), whereas not in the junctional or decidual zones (Figs 3F and S1A and B). However, when we controlled for expression of the eight mouse homologs of HEa miRNAs as a covariate, using multivariate analysis of covariance (MANCOVA), ethanol's effect on EMT became marginally nonsignificant (Pillai's trace, F (5,21) = 2.713, P = 0.068) (Fig 2C), suggesting that these miRNAs partially mediate effects of PAE on EMT pathway members in mice. Interestingly, PAE limited to the periconceptional period in rats also influenced expression of EMT core transcripts (Figs S2B and S3A-E).
To determine if PAE's effects on EMT pathway members in placenta are broadly conserved throughout mammalian evolution, we adopted a nonhuman primate (macaque) model of moderate to binge-type alcohol consumption. Placental tissues were isolated from GD85, GD110, and GD135 placenta (Fig 2D), which spans the human equivalent of the mid-second to mid-third trimester ( Fig  S2C). There was a significant effect of ethanol exposure on expression of core EMT mRNA transcripts by MANOVA (Pillai's trace statistic, F (4,9) = 4.229 P = 0.045, Fig 3B). Consistent with our findings in mouse, post hoc univariate ANOVA indicated that in primate placenta, ethanol exposure significantly increased CDH1 expression (F (1,12) = 4.866, P = 0.048), whereas VIM expression was significantly reduced (F (1,12) = 12.782, P = 0.0004), suggesting that, as in the mouse, PAE also impairs EMT in the primate placenta. Interestingly, there was no effect on SNAI2 or TWIST expression (Fig 3G-J). As in mice, accounting for expression of HEa miRNAs together as a covariate abolished the significant effect of PAE on EMT, although to a greater degree than mice (Pillai's trace, F (1,1) = 1.605, P = 0.425, Fig 2E). Interestingly, accounting for expression of individual HEa miRNAs did not explain the effects of PAE on placental EMT, suggesting that HEa miRNAs act in concert to mediate the effect of PAE on EMT in the primate placenta ( Fig 2F).
Collectively, our data suggest PAE-induced impairment of EMT in the trophoblastic compartment of placentae is conserved between rodents and nonhuman primates and that HEa miRNAs, particularly in primates, may moderate the effect of PAE on placental EMT. Consequently, subsequent studies focused on the collective role of HEa miRNAs, either on basal or on alcohol-influenced placental trophoblast growth, invasion, and the maturation of physiological function.

HEa miRNAs impair EMT in a model of human cytotrophoblasts
To investigate whether HEa miRNAs collectively interfere with the EMT pathway, as suggested by our in vivo data, we examined the effects of transfecting HEa miRNA mimics and antagomirs into BeWO cytotrophoblasts ( Fig 4A). We initially overexpressed each of the 11 HEa miRNAs individually, to determine whether any of them could influence the EMT pathway. We did not observe any significant effects (Fig S4), consistent with our findings in the primate PAE model that individual miRNAs did not explain the effects of ethanol on EMT. In contrast, transfection of pooled HEa miRNAs into cytotrophoblasts significantly increased CDH1 expression (F (1,36) = 30.08, P < 0.0001). Interestingly, expression of the pro-mesenchymal transcription factors TWIST and SNAI1 were also significantly reduced, but only in the context of concomitant 320 mg/dl ethanol treatment, pointing to an interaction effect between HEa miRNAs and ethanol (F (1,36) = 5.650 and 5.146, respectively, P = 0.023 and P = 0.029, Fig 4B-E). Consistent with our qPCR data, transfection of HEa miRNAs also significantly increased E-Cadherin protein expression (F (1,20) = 33.86, P < 0.0001, Fig 4F). We were unable to detect SNAI2 transcript expression or vimentin protein expression in these cells, consistent with previous reports (63).
We next sought to determine if more restricted subsets of HEa miRNAs could recapitulate the effects of HEa miRNAs collectively on EMT. Thus, we overexpressed hsa-miR-222-5p and hsa-miR-519a-3p, which are implicated in preeclampsia and fetal growth restriction, as well as hsa-miR-885-3p, hsa-miR-518f-3p, and hsa-miR-204-5p, which are implicated in preeclampsia, fetal growth restriction, and spontaneous abortion or preterm labor ( Fig S5A). In contrast to the collective action for all HEa miRNAs, exposure to each of these pools resulted in significant decreases in CDH1 expression (F (2,12) = 20.12, P = 0.0001). The pool including hsa-miR-885-3p, hsa-miR-518f-3p, and hsa-miR-204-5p also significantly increased Snai1 SNAI1 (F (2,12) = 4.604, P = 0.0328; Dunnett's post hoc P = 0.0497, Fig S5B-E). These data suggest that HEa miRNAs include subgroups of miRNAs that have the potential to partly mitigate the effects of elevating the entire pool. However, the potential protective effects of these subgroups are masked by the collective function of the entire group of HEa miRNAs. Whereas transfection of HEa miRNA mimics increased CDH1 expression, transfection of pooled antagomirs to HEa miRNAs significantly reduced CDH1 expression, only in the context of 320 mg/dl ethanol coexposure ( HEa miRNA × 320 mg/dl EtOH interaction, F (1,36) = 13.51, P = 0.0008; post hoc Tukey's honest significance difference (HSD), P = 0.005, Fig 4G). However, expression of TWIST was also decreased with ethanol co-exposure, and there was no significant difference in E-Cadherin protein expression relative to the control ( Fig 4H-K). Thus, our data suggest that increasing HEa miRNA levels impairs EMT pathway members in cytotrophoblasts, whereas inhibiting their action has a more restricted effect on EMT pathway members.

Antagomirs prevent HEa miRNAs' inhibition of EMT
We next investigated if pretreating cytrophoblasts with pooled HEa-miRNA antagomirs could prevent inhibition of the EMT pathway caused by transfecting HEa miRNA mimics. Pretreatment of cytotrophoblasts with HEa miRNA antagomirs prevented the elevation in CDH1 caused by transfection with HEa miRNA mimics (post hoc Tukey's HSD, n = 10 samples per group, P = 0.004). Likewise, pre-transfection with HEa miRNA antagomirs also prevented HEa miRNA mimic-induced reduction of SNAI1 and VIM expression (post hoc Tukey's HSD, n = 10 samples per group, P = 0.007 and P < 0.0001, respectively) ( Fig 6A-D).
As with cytotrophoblasts, pre-transfection with HEa miRNA antagomirs prevented HEa miRNA mimic-induced reduction of VIM, SNAI1, and SNAI2 expression in extravillous trophoblasts (post hoc Tukey's HSD, n = 10 samples per group, P < 0.0001, Fig 6E-H). Thus, our data suggest that pretreating cells with HEa miRNA antagomirs prevents inhibition of EMT pathway members resulting from transfection with HEa miRNA mimics in cytotrophoblasts and extravillous trophoblasts.

HEa miRNAs impair extravillous trophoblast invasion
Functionally, inhibition of the EMT pathway should reduce trophoblast invasiveness. Thus, we performed a transwell invasion assay using HTR8 extravillous trophoblasts transfected with HEa miRNA mimics and antagomirs. Although ethanol exposure by itself did not impair trophoblast invasion (Fig S6), there was a marginally significant interaction effect between ethanol exposure and HEa miRNA mimic transfection (F (1,28) = 3.418, P = 0.075). Thus, a planned comparison indicated that transfection with HEa miRNA mimics significantly reduced trophoblast invasion in the context of 320 mg/ dl ethanol co-exposure, relative to the control mimics (t(14) = 2.762, P = 0.015), consistent with our data demonstrating HEa miRNAs interfere with the EMT pathway ( Fig 7A). Contrastingly, transfecting HEa miRNA antagomirs increased invasion in the context of 320 mg/ dl ethanol co-exposure, although this effect was only marginally significant (t(14) = 1.805, P = 0.093, Fig 7B).

HEa miRNAs retard trophoblast cell cycle progression
Given the proliferative nature of cytotrophoblasts and the intimate relationship between EMT and cell cycle (64,65), we assessed the effects of ethanol and HEa miRNAs on BeWO cytotrophoblast cell cycle. After pulse-labeling the cells with the nucleic acid analog, EdU, for 1-h, we found that individually transfecting six of the HEa miRNA mimics increased EdU incorporation (unpaired t test, P < 0.05, false discovery rate [FDR] correction), suggesting an overall increased rate of DNA synthesis (Fig S7A). Contrastingly, simultaneous transfection of HEa miRNAs significantly reduced EdU incorporation (F (1,26) = 59.69, P < 0.0001), mirroring the effects of increasing concentrations of ethanol (R 2 = 0.304, P = 0.012) (Fig S7B and A).
To further characterize the coordinated effect of HEa miRNAs on cytotrophoblast cell cycle, we pulse-labeled the cells with EdU for 1-h and, post-fixation, labeled them with 7AAD to segregate cells into three groups: G 0 /G 1 (7AADlow, EDU−), S (EDU+), and G 2 /M (7AADhigh, EDU−). Both 120 mg/dl and 320 mg/dl ethanol exposures significantly decreased the proportion of cells in the S-phase, whereas 320 mg/dl exposure increased the proportion of cells in the G 2 /M-phase, consistent with the observed reduction in the rate of DNA synthesis (Fig S7C). Similar to the effects of ethanol exposure, pooled HEa miRNA mimic administration also significantly decreased the proportion of cells in the S-phase (F (1,28) = 52.78, P < 0.0001), whereas increasing the proportion of cells the G 2 /M-phase (F (1,28) = 8.395, P = 0.007) and exacerbated alcohol's effects on the cell cycle ( Fig 8C). Interestingly, pooled HEa miRNA antagomir administration also reduced the proportion of cells in the S-phase (F (1,26) = 14.98, P = 0.0007) and increased the proportion of those in the G 2 /M-phase (F (1,26) = 12.38, P = 0.002) ( Fig 8D).

HEa miRNAs have minimal effect on cell survival
We next investigated whether ethanol-and HEa miRNA-induced changes in cell cycle were related to an increase in cell death. Only the 320 mg/dl dose of ethanol exposure demonstrated a slight, but marginally significant effect, of increasing lytic cell death (t(18) = 2.022, P = 0.054), although there was no effect on apoptosis (Fig S8A  and B). However, the changes in cell cycle following transfection of individual or pooled HEa miRNA mimics were not mirrored by changes in lytic cell death. Nevertheless, two HEa miRNAs, hsa-miR-671-5p and hsa-miR-449a, did significantly increase apoptosis (unpaired t test, P < 0.05, FDR correction) (Fig S8C and D).
Contrastingly, transfection of four HEa miRNA antagomirs individually, significantly increased lytic cell death (unpaired t test, all P < 0.05, FDR correction), with the antagomir to hsa-miR-491-3p also increasing apoptotic cell death (t(14) = 3.383, P = 0.004, FDR correction, Fig S8C and D). Likewise, transfection of pooled HEa miRNA antagomirs increased lytic cell death (F (1,36) = 11.40, P = 0.002) but did not cause increased apoptosis (Fig S8E-H). Taken together, our data suggest that whereas ethanol exposure may increase cytotrophoblast death, increased levels of HEa miRNAs have minimal effects on cell death, suggesting that their effect on cell cycle and the EMT pathway is independent of any effect on cell survival.
HEa miRNAs modulate cytotrophoblast differentiation-associated Ca 2+ dynamics HEa miRNAs' effects on EMT pathway member expression, coupled with cell cycle retardation, indicate that HEa miRNAs influence trophoblast maturation. To model HEa miRNAs' effect on hormoneproducing and calcium-transporting syncytiotrophoblasts (66), we used a well-established protocol of forskolin-induced syncytialization of BeWO cytotrophoblasts (67,68). As expected, forskolin treatment induced fusion/syncytialization of cytotrophoblasts resulting in a greater average cell size in the forskolin + HEa miRNA   The effect of ethanol on baseline calcium was abrogated by HEa miRNAs, whereas HEa miRNAs + forskolin was not significantly different to forskolin alone, indicating that forskolin and HEa miRNAs may be affecting similar calcium pathways. The conversion of cytrophoblasts to syncytiotrophoblasts is accompanied by an increase in endoplasmic reticulum, which could increase calcium-buffering capabilities in response to ethanol stress on the cells; thus, HEa miRNA-induced syncytialization pathways may be protective against ethanol stress.
Extracellular ATP has been shown to inhibit trophoblast migration (69) and can directly stimulate increased intracellular calcium elevations through purinergic receptors ubiquitously present on trophoblasts (70). Both HEa miRNA and ethanol administration significantly increased intracellular calcium in response to acute ATP administration (F (1,426) = 10.34 and F (1,386) = 16.30, P = 0.001 and P < 0.0001, respectively) ( Fig 10B). This may be indicative of a lack of down-regulation of purinergic receptors required in trophoblast migration as part of the interrupted EMT pathway. Forskolin-induced maturation decreased calcium response to ATP (F (1,386) = 50.72, P < 0.0001) (Fig 10C) and prevented the HEa miRNA-induced increase in ATP response. These data agree with previous studies showing increased nuclear trafficking of ionotropic receptor P2X7 and more localized P2X4 expression over placental development, which may decrease the overall calcium influx in response to ATP (71).

HEa miRNAs promote syncytialization-dependent hormone production
Transfection of HEa miRNA mimics did not change CGA (encodes Chorionic gonatropin alpha), CGB (encodes Chorionic gonadoptropin beta), or IGF2 (encodes Insulin-like growth factor 2) transcript expression relative to the control in non-syncytialized trophoblasts. However, following forskolin-induced syncytialization of BeWO cytotrophoblasts (Fig 11A), HEa miRNA mimics significantly increased expression of CGA and CGB (post hoc Tukey's HSD, n = 10 samples per group, P = 0.001 and 0.005, respectively). Consistent with our previous results, HEa miRNA mimics also increased CDH1 expression in both cytotrophoblasts and syncytiotrophoblasts (F (1,20) = 5.286, P = 0.032); there was also a main effect of syncytialization on CDH1 expression, as has been previously reported (F (1,36) (Fig 11F). On the other hand, there was no effect of HEa miRNA antagomirs on CGA and CGB expression, although we did observe a decrease in IGF2 transcript expression, following syncytialization, relative to controls (post hoc Tukey's HSD, n = 10 samples per group, P = 0.001) (Fig 11G-J).
Given that HEa miRNAs promote syncytialization-dependent hormone production, we next investigated maternal plasma levels of intact hCG in our Ukraine birth cohort. Plasma hCG levels were nonsignificantly increased in the second trimester of HEa group mothers relative to their UE counterparts, consistent with previous studies (72). During the third trimester, however, hCG levels remained significantly elevated in HEa group mothers compared with the UE group (median test, n = 23 samples in HEa group and n = 22 for HEua and UE groups, P = 0.03) (Fig 12). Furthermore, there was no significant difference of gestational age at blood draw between the different groups indicating the increased level of hCG in the HEa group was not confounded by gestational age at which blood was sampled (Fig S10) (73). Interestingly, both alcohol and hCG levels were negatively associated with gestational age at delivery (GAD), with a significant interaction between periconceptional alcohol exposure and hCG levels on GAD (Table S2). Taken together, our data suggest HEa miRNAs may contribute to PAE-dependent increases in hCG levels during pregnancy.
Following tail vein administration of two human-specific sentinel miRNAs, miR-518f-3p and miR-519a-3p, we found a high biodistribution of both miRNAs in the placenta, comparable with levels seen in the liver and spleen (Fig S11A and B). Thus, to determine whether HEa miRNA's effects on fetal growth could result from their actions on the placenta, we quantified the placental expression of core EMT members in the GD18 placentas of control and HEa miRNA fetuses. HEa miRNA administration significantly reduced expression of mesenchymal-associated transcript VIM (F (1,14) = 14.23, P = 0.002) and SNAI2 (F (1,14) = 5.99, P = 0.028) with a significant sex by HEa miRNA interaction effect on SNAI1 (F (1,66) = 5.55, P = 0.034) and CDH1 (F (1,14) = 6.01, P = 0.028) (Fig 14A-E). Interestingly, and in line with our in vitro findings whereby HEa miRNAs promoted syncytialization-dependent cell fusion and hCG production, HEa miRNA administration significantly increased expression of the mRNA transcript for SynB, a gene that is important for syncytiotrophoblast maturation (F (1,66) = 4.11, P = 0.047) (Fig 14F).

Discussion
We previously reported that gestational elevation of 11 maternal plasma miRNAs predicted which PAE infants would exhibit adverse outcomes at birth (8). These HEa miRNAs were elevated throughout mid and late-pregnancy, encompassing critical periods for fetal development, and were predicted to target the EMT pathway (8). In this study, we tested this prediction by adopting rodent and macaque gestational moderate alcohol self-administration paradigms. Despite differences in their placental anatomy (74,75,76,77), we are the first to report that PAE impairs placental EMT across species, indicating a conserved effect of PAE on placental development. In addition, we found that HEa miRNAs collectively, but not individually, mediated the effects of PAE on core EMT pathway members and that, together, they inhibited EMT in human trophoblast culture models. Although we assessed the effects of HEa miRNAs on core EMT components (10,14,15,59,60,61,62), analysis of their 39 UTRs indicates that these are unlikely to be the direct targets of HEa miRNA action. Additional studies will be needed to dissect out the signaling networks that connect HEa miRNAs to the assessed EMT components.
Interestingly, HEa miRNAs also promoted syncytialization (forskolin)dependent hCG expression, mirroring the elevation of third trimester maternal hCG levels in the PAE group within our clinical cohort. This late-gestation elevation of hCG levels may serve as a compensatory mechanism to prevent the preterm birth associated with PAE, as hCG during late gestation is hypothesized to promote uterine myometrial quiescence (78,79). In support of this hypothesis, we found significant negative associations between both hCG levels and alcohol consumption with GAD. Furthermore, there was a significant interaction between periconceptional alcohol exposure and hCG levels, with higher hCG levels corresponding to a smaller effect of alcohol exposure at conception on GAD, indicating that hCG moderates the effect of alcohol on age at delivery (Table S2).
Since HEa miRNAs collectively prevented trophoblast EMT, we hypothesized that, as a functional consequence, these maternal  miRNAs would also inhibit fetal growth. When we delivered 8 of the 11 HEa miRNAs known to be present in mouse, to pregnant dams during the period of placental branching morphogenesis and endometrial invasion, and when EMT is particularly active, we found that HEa miRNAs reduced fetal growth. Importantly, ethanol exposure during this period has also been shown to result in fetal growth deficits and dysmorphia in rodent PAE models (80,81), suggesting that maternal miRNA-mediated deficits in trophoblast invasion may mediate some of the effects of PAE on fetal growth. In support of this, we found placentas from the HEa miRNA-treated group had impaired expression of core EMT pathway members. This disruption of placental EMT may also have implications for placental vascular dynamics, as we have also previously observed in mouse models (82). The nonhuman primate tissue analyzed here was also derived from animals that were characterized in vivo using MRI and ultrasound imaging, which demonstrated that maternal blood supply to the placenta was lower in ethanol-exposed animals compared with controls and that oxygen availability to the fetal vasculature was reduced (83).
HEa miRNAs may mediate other pregnancy-associated pathologies, aside from PAE. We identified numerous studies that reported increased circulating and placental levels of at least 8 of 11 HEa miRNAs in gestational pathologies arising from placental dysfunction. For example, elevated levels of one HEa miRNA, miR-519a-3p, a member of the placentally expressed C19MC family cluster, was reported in the placentae of patients with preeclampsia, recurrent spontaneous abortion, and intrauterine growth restriction (29,30,45,46). Interestingly, collective overexpression of the 59 C19MC miRNAs inhibits trophoblast migration, explaining their enrichment in the non-migratory villous trophoblasts and suggests their downregulation is necessary for maturation into invasive extravillous trophoblasts (84). Thus, a greater understanding of the placental roles of HEa miRNAs may also help disentangle the etiology of other pregnancy complications. We also observed that overexpression of more restricted subsets of HEa miRNAs associated with preeclampsia, fetal growth restriction, and spontaneous abortion or preterm labor also partly promoted EMT transcript signatures, contrasting with the collective inhibitory action of HEa miRNAs as a whole. Thus, elevation of some subsets of HEa miRNAs may constitute a compensatory mechanism aimed at minimizing placental pathologies, although their potential protective effects are masked by the collective elevation of HEa miRNAs.
Although we did not investigate the effects of PAE on EMT in nonplacental organs, it is likely that PAE broadly disrupts EMT in multiple fetal compartments. Developmental ethanol exposure has been shown to inhibit the EMT-dependent migration of neural crest progenitors involved in craniofacial development, explaining the facial dysmorphology seen in fetal alcohol syndrome and FASDs (85,86). Outside of its effects on the neural crest, PAE is significantly associated with various congenital heart defects, including both septal defects and valvular malformations (87,88,89,90). Given that development of heart depends on EMT within the endocardial cushions (91,92), disruption of endocardial EMT could explain both the valvular and septal malformation associated with PAE.
Collectively, our data on HEa miRNAs suggest miRNA-based interventions could minimize or reverse developmental effects of PAE and other placental-related pathologies. miRNA-based therapeutic approaches have been advanced for other disease conditions (93,94). However, our data also suggest the effects of combinations of miRNAs are not a sum of their individual effects. Functional synergy between clusters of co-regulated miRNAs may be a common feature in development and disease. For instance, in 2007, we presented early evidence that ethanol exposure reduced miR-335, -21, and -153 in neural progenitors and that coordinate reduction in these miRNAs yielded net resistance to apoptosis following ethanol exposure (95). In that study, we also showed that coordinate knockdown of these three miRNAs was required to induce mRNA for Jagged-1, a ligand for the Notch cell signaling pathway, an outcome that was not recapitulated by knocking down each miRNA individually (95). More recently, combined administration of miR-21 and miR-146a has been shown to be more effective in preserving cardiac function following myocardial infarction than administration of either of these miRNAs alone (96). Although miRNA synergy has not been explored in detail, these data show that new biology may emerge with admixtures of miRNAs and that therapeutic interventions may require the use of such miRNA admixtures rather than single miRNA molecules, as have been used in clinical studies to date.
In conclusion, we have observed that a set of 11 miRNAs, predictive of adverse infant outcomes following PAE, collectively mediate the effects of alcohol on the placenta. Specifically, elevated levels of these miRNAs together, but not individually, promote an aberrant maturational phenotype in trophoblasts by inhibiting core members of the EMT pathway and promoting cell stress and syncytialization-dependent hormone production. Although extensive research has established circulating miRNAs as biomarkers of disease, our study is one of the first to show how these miRNAs explain and control the disease process themselves. Functionally, we find that these miRNAs are clinically correlated with measures of fetal development and directly cause intrauterine growth restriction when administered in vivo. Our work suggests that a greater understanding for the role of HEa miRNAs during development, and their role in coordinating the EMT pathway in the placenta and other developing tissues, will benefit the understanding of FASDs and other gestational pathologies and potentially lead to effective avenues for intervention.

Mouse model of PAE
C57/BL6J mice (The Jackson Laboratory) were housed under reverse 12-h dark/12-h light cycle (lights off at 08:00 h). PAE was performed using a previously described limited access paradigm of maternal drinking (97,98). Briefly, 60-d-old female mice were subjected to a ramp-up period with 0.066% saccharin containing 0% ethanol (2 d), 5% ethanol (2 d), and finally 10% ethanol for 4-h daily from 10:00 to 14:00 beginning 2 wk before pregnancy, continuing through gestation (Fig S2A). Female mice offered 0.066% saccharin without ethanol during the same time period throughout pregnancy served as controls. Tissue from the labyrinth, junctional, and decidual zone of male and female gestational day 14 (GD14) placentae were microdissected, snap-frozen in liquid nitrogen, and stored at −80°C preceding RNA and protein isolation.

Rat model of PAE
Outbred nulliparous Sprague Dawley rats were housed under a 12-h light/12-hour dark cycle. PAE in Sprague Dawley was conducted according to our previously published exposure paradigm (20,100). Briefly, dams were given a liquid diet containing either 0% or 12.5% ethanol (vol/vol) from 4 d before mating until GD4 (Fig S2B). Dams had ad libitum access to the liquid diet 21 h daily and consumed equivalent calories. Water was offered during the remaining 3 h of the day. On GD5, liquid diets were removed and replaced with standard laboratory chow. On GD20, the placentas were immediately separated into the labyrinth and junctional zone, snap-frozen in liquid nitrogen and stored at −80°C preceding RNA isolation.

Nonhuman primate model of PAE
As previously described in detail (83), adult female rhesus macaques were trained to orally self-administer either 1.5 g/kg/d of 4% ethanol solution (equivalent to six drinks/d) or an isocaloric control fluid before time-mated breeding. Each pregnant animal continued ethanol exposure until gestational day 60 (GD60, term gestation is 168 d in the rhesus macaque) (101). Pregnancies were terminated by cesarean section delivery at three different time points; GD85, GD110, or GD135 (Fig S2C). The macaque placenta is typically bilobed with the umbilical cord insertion in the primary lobe and bridging vessels supplying the fetal side vasculature to the secondary lobe ( Fig 2D showing gross placenta anatomy) (102). Full thickness tissue biopsies (maternal decidua to fetal membranes) were taken from both the primary and secondary lobes of the placenta (Fig 2E showing H&E section of placenta). Samples were immediately snap-frozen in liquid nitrogen and stored at −80°C preceding RNA isolation.

Cell death analysis
BeWO cell culture was harvested 48 h post transfection. Media was subjected to lactate dehydrogenase (LDH) detection using the Pierce LDH Cytotoxicity Assay kit (Cat No. 88953; Thermo Fisher Scientific), according to the manufacturer's instructions, for lytic cell death quantification. The Promega Caspase-Glo 3/7 Assay system (Cat No. G8091; Promega) was used to quantify apoptotic cell death.

Invasion assay
At 24 h post-transfection and/or ethanol exposure, HTR8 cells were serum-starved for an additional 18 h. Subsequently, HTR8 cells were seeded onto transwell permeable supports precoated with 300 μg/ml Matrigel (Cat No. 354248; Corning). After 24 h, cells remaining in the apical chamber were removed with a cotton swab. Cells that invaded into the basal chamber were incubated with 1.2 mM 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) for 3 h, and the precipitate solubilized with 10% SDS in 0.01 N HCl. Absorbance intensities were read at 570 nm in a Tecan Infinite 200 plate reader.

Metabolic flux analysis and calcium imaging
BeWO cells (10,000/well) were plated into Seahorse XF96 Cell Culture Microplates (Cat No. 103275-100; Agilent Biotechnology). The OCR, a measure of mitochondrial respiration, and ECAR, a measure of glycolysis, were measured using the Seahorse XFe96 flux analyzer (Seahorse Bioscience). At the time of assay, the cell culture medium was replaced with the appropriate prewarmed Seahorse XF Base Medium (Cat No. 102353-100; Agilent Biotechnology). OCR and ECAR parameters were measured using the Seahorse XFp Cell Energy Phenotype Test kit (Cat No. 103275-100; Agilent Biotechnology). Metabolic stress was induced by simultaneous treatment with 1 μm oligomycin and 0.125 μM carbonyl cyanide p-[trifluoromethoxy]phenyl-hydrazone (FCCP).
BeWO cells were also plated onto glass coverslips in 24-well plates at a density of 30,000 cells/well. After exposure to ethanol and/or forskolin in culture, the cells were prepared for calcium imaging. After replacement of the culture media with external imaging media (154 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 0.5 mM MgCl 2 , 5 mM glucose, and 10 mM Hepes, pH 7.4), the cells were loaded for 35 min at 37°C with the calcium indicator dye fluo-4 AM (Cat No. F14201; Thermo Fisher Scientific), at a final concentration of 5 μM fluo-4 AM in 0.1% DMSO. After incubation, the cells were washed to remove remaining extracellular fluo-4 and imaged at 40× using confocal microscopy (FV1200-equipped BX61WI microscope; Olympus Corporation). Time-lapse images were acquired at a frequency of 0.5 Hz. Individual cells were manually outlined, and area and mean fluorescence intensity were obtained for each cell (FIJI image processing package) (105). To determine the functional calcium range of each cell, at the end of imaging, the cells were exposed to 5 μM ionomycin and 10 mM EGTA (0 mM external Ca 2+ , F range = F ionomycin − F EGTA ). Baseline fluorescence was determined by averaging the lowest five consecutive fluorescence values during the initial 5 min (F baseline ), which was then expressed as a percentage of F range (ΔF baseline = (F baseline − F EGTA )/F range × 100). Maximal intracellular calcium response to 100 μM ATP was determined by averaging the highest three consecutive fluorescence values during ATP application (F ATP ) and determining the amount of fluorescence as a percentage of F range (ΔF ATP = (F ATP − F EGTA )/F range × 100).

Quantitative reverse transcriptase-polymerase chain reaction analysis
Total RNA was extracted from tissue, as well as BeWO Scientific). The data presented correspond to the mean 2 −ΔΔCt after being normalized to the geometric mean of β-actin, hypoxanthine-guanine phosphoribosyltransferase 1 (HPRT1), and 18s rRNA. Expression data for miRNA was normalized to the geometric mean of miR-25-3p, miR-574-3p, miR-30b-5p, miR-652-3p, and miR-15b-5p. For each primer pair, thermal stability curves were assessed for evidence of a single amplicon, and the length of each amplicon was verified using agarose gel electrophoresis. A list of primers and their sequences is presented in Table S3.

Western immunoblotting analysis
Protein was extracted using 1× RIPA lysis buffer (MilliporeSigma) supplemented with Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific). Tissue was homogenized using the Branson Sonifier 150. Protein concentration was determined using Pierce BCA protein assay kit (Thermo Fisher Scientific), and 30 μg of protein was loaded onto a 4%-12% Bis-Tris (Cat No. NPO323BOX; Invitrogen/Thermo Fisher Scientific), size-fractionated at 200 V for 35 min, and transferred to a PVDF membrane using the iBlot transfer system (Invitrogen/Thermo Fisher Scientific). Blots with protein from cultured cells were blocked with 5% nonfat dry milk in tris-buffered saline containing Tween-20 (TTBS) for 1 h and incubated overnight with primary antibody. The blot was then washed and incubated with an HRP-conjugated goat anti-rabbit or anti-mouse IgG (Invitrogen) at dilution 1:1,000 for 1-h, then developed using PerkinElmer Western Lightning Plus Chemi ECL (PerkinElmer) and visualized using a CCD camera (Fluorchem Q, Alpha Innotech). Blots with protein from homogenized tissue were dried overnight, rehydrated in methanol, stained with REVERT Total Protein Stain, and developed with the Odyssey CLx Imaging System (LI-COR). Blots were then blocked with Odyssey Blocking Buffer (TBS) for 1 h and incubated overnight with primary antibody. The blot was then washed and incubated with IRDye 800CW secondary antibody (Cat No. 925-32210; LI-COR). The following antibodies were used: β-Actin HRP (Cat No. sc-47778; Santa Cruz Biotechnology); Goat anti-Mouse IgG (H+L) Secondary Antibody, HRP (Cat No. 62-6520; Thermo Fisher Scientific); Goat anti-Rabbit IgG (H+L) Secondary Antibody, HRP (Cat No. 65-6120; Thermo Fisher Scientific); purified Mouse Anti-E-Cadherin (Cat No. 610181; BD Biosciences); and Rabbit antivimentin antibody (EPR3776) (Cat No. ab 924647; Abcam). Protein levels were quantified using the densitometric analysis package in FIJI image processing software (105).

ELISA
The second and third trimester maternal plasma samples were collected as part of a longitudinal cohort study conducted in two regions of Western Ukraine as part of the Collaborative Initiative on FASDs (CIFASD.org) between the years 2006 and 2011, as previously reported (8). Plasma, at a 1:1,000 dilution, was subjected to hCG detection using Abcam's intact human hCG ELISA kit (Cat no. ab100533) following the manufacturer's protocol.

Literature review
We conducted a literature review for HEa miRNAs and their associated gestational pathology using the National Institute of Health's PubMed search interface. For each miRNA, the following search parameters were used: ½miRX OR miR X OR miRNA X OR miRNAX or miRNX AND MeSH Term where X represents the miRNA of interest and automatic term expansion was enabled. The following MeSH terms, and related search terms (in brackets), were used: Fetal Growth Retardation

Statistical analyses
Linear regression models were used to estimate associations between infant growth measures and miRNA expression levels, gestational age at blood draw, the interaction between subjectcentered miRNA expression level and gestational age at blood draw, and child sex. Spearman correlations between infant growth measures and subject-centered miRNA expression levels were also calculated. Linear regression models were also used to estimate the associations between gestational at birth and logtransformed hCG levels, ethanol intake, the interaction between log-transformed hCG levels and ethanol intake, gestational at blood draw, and child sex. Statistical analysis and graphs were generated with GraphPad Prism 6 software (GraphPad Software, Inc), SPSS v24, or R version 3.3.1. Results are expressed as the mean ± SEM or alternatively as box and whisker plots with the bounds of the box demarcating limits of first and third quartile, a median line in the center of the box, and whiskers representing the total range of data. The overall group effect was analyzed for significance using one-way MANOVA, one-way or two-way ANOVA with Tukey's HSD or Dunnett's multiple comparisons post hoc testing when appropriate (i.e., following a significant group effect in one-way ANOVA or given a significant interaction effect between experimental conditions in two-way ANOVA), to correct for a family-wise error rate. A two-tailed t test was used for planned comparisons. For experiments characterizing the individual effects of HEa miRNAs against the control miRNA or antagomirs, individual two-tailed t test with 5% FDR correction was applied to account for multiple comparisons. All statistical tests, sample sizes, and post hoc analysis are appropriately reported in the results section. A value of P < 0.05 was considered statistically significant and a value of 0.1 < P < 0.05 was considered marginally significant.

Study approval
Human study protocols were approved by the institutional review boards at the Lviv National Medical University, Ukraine, and the University of California San Diego as well as Texas A&M University in the United States. Research was conducted according to the principles expressed in the Declaration of Helsinki with written informed consent received from participants before inclusion in the study. All rodent experiments were performed in accordance with protocols approved by the University of New Mexico Institutional Animal Care and Use Committee (IACUC), the Texas A&M University IACUC, and the University of Queensland Animal Ethics Committees. All procedures involving nonhuman primate research subjects were approved by the IACUC of the Oregon National Primate Research Center (ONPRC), and guidelines for humane animal care were followed. The ONPRC abides by the Animal Welfare Act and Regulations enforced by the US Department of Agriculture.

Author Contributions
AM Tseng: investigation, methodology, and writing-original draft, review, and editing. AH Mahnke: investigation, methodology, and writing-review and editing. AB Wells: formal analysis and writing-review and editing. NA Salem: methodology and writing-review and editing. AM Allan: investigation, methodology, and writing-review and editing. VHJ Roberts: investigation, methodology, and writing-review and editing. N Newman: investigation, methodology, and writing-review and editing. NAR Walter: investigation, methodology, and writing-review and editing. CD Kroenke: investigation, methodology, and writing-review and editing. KA Grant: investigation, methodology, and writing-review and editing. LK Akison: investigation, methodology, and writing-review and editing. KM Moritz: investigation, methodology, and writing-review and editing. CD Chambers: supervision, investigation, methodology, and writing-review and editing. RC Miranda: supervision, investigation, methodology, and writing-review and editing.