Comprehensive Analysis of the Placenta–Cortex Transcriptomic Database Reveals a Neuroactive Ligand–Receptor Dysregulation After Prenatal Alcohol Exposure

Abstract

Neuroplacentology is an emerging field of research supporting that the placenta actively contributes to the fetal brain development through the release of bioactive molecules. Recent angiogenesis-focused data showed that prenatal alcohol exposure (PAE) disrupts inter-organ gene expression between the placenta and fetal cortex. The present study aimed to perform the first comprehensive and untargeted analysis of a murine placenta–cortex transcriptomic database of PAE. Gene lists from a recently NCBI-deposited PAE Placenta–Cortex transcriptomic database were analyzed using g:Profiler for unbiased functional profiling querying Gene Ontology, KEGG, and Reactome databases. Genes intersecting with cell–cell communication terms were submitted to STRING and ShinyGO analyses to identify enriched protein–protein interactions and pathways. Several ligand or receptor candidates were then validated by Western blot. g:Profiler revealed 21 enriched GO functional maps, seven KEGG pathways, and six Reactome pathways, of which 11 were related to cell-to-cell communication. STRING analysis exhibited substantial protein–protein interaction enrichments supporting that proteins belonging to the functional maps and pathways are biologically connected. Notably, 38 ligands or receptors from endocrine families including angiotensinogen, leptin, somatostatin, or PACAP were identified. Western blot analysis of protein candidates showed different validation patterns. In particular, the PACAP receptor family confirmed transcriptomic findings and revealed sex-dependent PAE-impacted expression profiles. The present study indicates that PAE is associated with alterations in the transcriptomic placenta–cortex expression profile, including changes in the expression ratios of several ligands and/or receptors implicated in key physiological pathways such as energy balance, vascular development, and neurogenesis. These transcriptomic associations suggest that altered placenta–fetal brain signaling at the gene expression level may be involved in alcohol-induced neurodevelopmental disorders, highlighting the need for future functional validation studies.

1. Introduction

The placenta is an ephemeral organ which plays a pivotal role in the growth of the fetus by providing nutrients, oxygen, removing waste products, and also by constituting a selective and/or protective barrier for the passage of hormones, toxic agents, and infectious microorganisms [1]. In particular, the placenta is a source of endocrine factors involved in the control of major physiological functions of the pregnant woman and the fetus such as food intake, lactation preparation, or immunity [2].

Neuroplacentology has emerged as a new field of research focusing on the role of placenta on fetal brain development [3]. Arising evidence indicates that the placenta directly contributes to the development and functions of the fetal brain through the release of bioactive molecules [4,5,6,7]. As an example, a targeted repression of the placental angiogenic factor CD146 impairs the development of the fetal brain vasculature with neurodevelopmental consequences on differentiating oligodendrocytes, a cell population migrating along microvessels [5]. Similarly, it was recently shown that placental overexpression of Igf1 impairs the development of the striatum in the fetus [8] and that placental neuroactive steroids are engaged in the development of fetal brain structures [9,10,11] whereas high levels of placental interleukin (IL)-1β are associated to a high risk of neurodevelopmental impairments in the progeny [12]. Altogether, these recent advances on neuroplacentology support the existence of a functional placenta–fetal brain axis which, when dysregulated, may contribute to neurodevelopmental troubles. Consequently, the identification of placental factors involved in brain development would have a value as predictive biomarkers of neurodevelopmental disorders [13].

Prenatal alcohol exposure (PAE) is a leading cause of neurodevelopmental troubles, and it has been well established that alcohol exposure during pregnancy can impact major neurodevelopmental processes such as neurogenesis, synaptogenesis, or myelination [14]. In the last decade, it was shown that in human, alcohol consumption during pregnancy is associated with a marked disorganization of the cortical vasculature in the fetus [15]. Such an effect is associated with the dysregulation of the expression of angiogenic placental factors [6,16]. Consistent with these clinical data, mouse models of fetal alcohol spectrum disorder (FASD) showed that PAE modifies the inter-organ expression ratio of more than 25 angiogenic genes [7] and that a targeted repression of PlGF in the placenta mimics PAE-induced cortical disorganization while, reciprocally, placental PlGF overexpression partially rescued the effects of PAE on the cortical vasculature of fetuses [6]. Taken together, these data suggest that PAE impairs placenta–brain communication and that, if so, the deleterious effect of PAE would be wider than the recently evidenced dysregulation of angiogenic factors [7]. In order to test this hypothesis, the aim of the present study was to perform the first global and unbiased analysis of the murine Placenta–Cortex transcriptomic signature of PAE.

2. Results

2.1. g:Profiler Analysis of the Placenta–Cortex PAE Database

An inter-organ transcriptomic Placenta/Cortex database of PAE was recently deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) with the accession number GSE241836 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE241836, accessed on 1 September 2023), and a list of 1582 genes were found dysregulated between the control and PAE signatures [7]. To perform a comprehensive and unbiased enrichment analysis, this gene list (Supplementary Table S1) was submitted to g:Profiler, which enables the exploration of high-throughput genomic data with a simultaneous mining of Gene Ontology Resource, KEGG pathway and Reactome databases [17]. After evaluating the Molecular Function (MF), Biological Process (MP), and Cell Component (CC) GO classifications, g:Profiler analysis identified 68, 345, and 84 GO terms that were significantly enriched, respectively (Figure 1A). Regarding the KEGG pathway and Reactome pathway databases, analysis of the PAE Placenta–Cortex signature revealed seven and six enriched processes, respectively. The results for the 510 enriched terms and pathways can be found in Supplementary Table S2.

Figure 1.

g:GOST functional profiling of the Placenta–Cortex PAE database. (A) Manhattan plot illustrating the enrichment analysis after high-throughput genomic data analysis with Gene Ontology Resource, KEGG pathway database, and Reactome pathway database sources. The x-axis represents enriched terms grouped according to data sources and color-coded (e.g., GO:Molecular Function in red; GO:Biological Process in yellow, GO:Cellular component in green, KEGG pathways in pink, and Reactome pathways in blue). The y-axis displays the adjusted p-values expressed as negative log10 scale. The most relevant terms of Gene Ontology Resource are capped. (B) Table representing the most relevant driver terms from each database. Significance is expressed as adjusted p-value and –log10 (adj p). For each term, the observed intersection between query and term size is compared to expected intersection size to allow calculation of fold enrichment. A color code is applied to represent the importance of fold enrichment of each term or pathway (insert).

In order to determine whether the comprehensive list of Gene Ontology terms had significant interactions, the 497 GO terms were then submitted to a GO context organization analysis. The initial stage of this process involved the aggregation of significant terms into sub-ontologies based on their relationships. The subsequent stage focused on the identification of the leading term that gives rise to other significant functions in the neighborhood. This strategy refined the term list returned by g:Profiler and indicated that the enriched-GO terms from the GO:MF, GO:BP, and GO:CC classifications could be reduced to 8, 11, and 2 context maps, respectively (Figure 1B). Each of these maps was identified by leading terms corresponding to Gene Ontologies that are represented as capped in Figure 1A. These relevant GO terms, as well as the enriched KEGG and Reactome pathways previously identified, are described in Figure 1B, which includes the statistical values and fold enrichment for these entities. For each leading term, the related context map was edited and provided in the Supplementary Dataset S1 folder. As an example, the context map relative to the term ‘Protein binding’ is shown in Figure 2. The leading term ‘Protein binding’ (GO:0005515; marked with an arrow) is linked to 31 enriched GO terms (Figure 2; from green to yellow frames). All associated terms are defined (name, reference number and adjusted p-values), and their relations within the ontology are visualized (Supplementary Dataset S1). The higher the number of connected GO terms for a given map, the stronger the impact on a given process. Here, the term ‘Protein binding’ appears at the second level within ‘Molecular function’ (marked with an arrowhead). It is noteworthy that ‘Protein binding’ is directly connected to ‘signaling receptor binding’ (GO:0005102; p = 2.015 × 10⁻⁸), which is, in turn, a direct parent of ‘growth factor receptor binding’ (GO:0070851; p = 2.651 × 10⁻²) and ‘receptor ligand activity’ (GO:0048018; p = 4.815 × 10⁻⁴): These three GO terms, indicated by red dotted frames on Figure 2, are evidently associated with cell-to-cell communication.

Figure 2.

GO context organization analysis. The most significantly enriched GO terms from each of the three GO classifications (MF, BP and CC) are reorganized with g:Profiler based on their ontologies and a calculation of the enrichment significance when excluding up- and downstream terms of the ontology. Resulting GO context maps are presented based on enriched GO-terms defined relation (connecting edges between term nodes). On this figure, the “Protein binding” GO context map is visualized, which constitutes one out of the eight maps generated by the algorithm. The leading GO-term is indicated by an arrow. All significantly enriched GO-terms are displayed with a colored border that corresponds to their enrichment p-values (from the highest (dark green) to the lowest (yellow)). Grey-bordered GO-terms are not significantly enriched after this second step of analysis but allow connecting the enriched terms to the root of the domain (i.e., ‘molecular function’; arrowhead). For each term, g:Profiler provides the term ID, description, and enrichment p-value. Each row is representative of the hierarchical structure of the Gene Ontology tree. Each context association map is provided at high resolution in the Supplementary Dataset S1 folder.

Subsequently, an in-depth analysis of each context map revealed that the enrichment analysis of the Placenta–Cortex PAE inter-organ signature uncovered six enriched terms/pathways related to cell-to-cell communication, such as ‘Protein binding’ (GO: MF:0005515; p = 7.333 × 10⁻⁵⁷; fold enrichment = 1.5), ‘Neuroactive ligand–receptor interaction’ (KEGG:04080; p = 0.00084; fold enrichment = 1.92), ‘Signaling by GPCR’ (REAC:R-MMU-372790; p = 0.030; fold enrichment = 1.58), or ‘Neuropeptide signaling pathway’ (GO:0007218; p = 0.019; fold enrichment = 3.27; Figure 1B; red dotted frames). In addition, the present non-targeted analysis showed significant enrichment for the term ‘Circulatory system process’ (GO:0003013; p = 0.00043; fold enrichment = 2.04).

In order to refine the aforementioned examination of the GO context organization analysis, the utilization of pie charts was employed to illustrate the number of associated GO terms for each leading term (Figure 3), with adjusted p values (Supplementary Table S4). For instance, among the eight context maps resulting from the GO:MF resource analysis, five leading terms are connected to one or several terms (Figure 3A). As illustrated, the leading term ‘Protein binding’ is associated with 31 related terms, while the leading term ‘ATP-dependent activity, acting on DNA’, is connected to a single term (Figure 3A; green dotted square). Conversely, three leading terms were found to be devoid of associated GO terms and were consequently annotated in grey text (Figure 3A). A parallel examination was conducted on the context maps resulting from the GO:BP resource and the GO:CC resource analysis (Figure 3B,C). Additionally, a thorough examination of the pathways that were found to be significantly enriched in the KEGG and Reactome databases during the initial functional profiling phase was conducted (Figure 4), with adjusted p values (Supplementary Table S5). The analysis reveals the number of genes present in the Placenta–Cortex PAE inter-organ signature that are found in each enriched pathway. For example, the ‘Neuroactive ligand/receptor interaction’ pathway (KEGG:04080; p = 0.00084) includes 49 genes that belong to the Placenta–Cortex PAE inter-organ signature (Figure 4A; green dotted square). In a similar manner, the ‘Signaling by GPCR’ pathway (REAC:R-MMU-372790; p = 0.03092) encompasses 69 genes from our query (Figure 4B; green dotted square). For all the enriched GO terms, KEGG, and Reactome pathways mentioned in Figure 4A,B, complete intersecting gene lists were provided in Supplementary Table S3. In aggregate, the exhaustive g:Profiler analysis revealed that, in addition to vascular dysregulation, other processes such as those involved in cell-to-cell communication and ligand/receptor interactions display altered inter-organ expression ratios in the PAE Placenta–Cortex signature.

Figure 3.

Pie chart representation of GO terms associated with the leading terms from the organization maps. (A) GO context analysis of the enriched GO-terms from the GO:Molecular function (MF) database generated height-organization maps. Each map is identified by its leading GO-term. For example, the Protein binding context map is represented in blue and regroups 31 significantly enriched GO-terms connected to GO-term Protein binding. (B) Number of significantly enriched GO-terms connected to each of the 11 leading GO-terms resulting from the GO context analysis of the GO:Biological Process (BP) database. (C) Number of significantly enriched GO-terms connected to the 2 leading GO-terms resulting from the GO context analysis of the GO:Cell Component (CC) database. Green dotted line frames represent GO context maps in relation to intercellular communication or circulatory system. Leading terms with no resulting connection to any enriched GO term are designated in grey.

Figure 4.

Pie chart representation of genes numbers within the enriched pathways resulting from the g:GOST analysis of KEGG and Reactome pathways. (A) Number of intersecting genes per significantly enriched KEGG pathway database. For example, 49 genes from our query are found in the pathway Neuroactive ligand/receptor interaction. (B) Number of intersecting genes per significantly enriched Reactome pathway database. For example, 69 genes from our query are found in the pathway Signaling by GPCR.

2.2. Identification of Protein–Protein Interactions Within Enriched Entities

A STRING analysis was performed to determine the degree to which proteins encoded by query genes belonging to each leading GO term, or KEGG/Reactome pathway functionally interact (Table 1). The results yielded predicted protein–protein interaction (PPI) scores within the specified groups (Table 1). For instance, STRING analysis of the leading term ‘Neuroactive ligand–receptor interaction’ (KEGG:04080) revealed an association network that exhibited a substantial PPI enrichment, involving 38 proteins that were linked by 73 interactions. This finding suggests that these proteins are biologically connected as a functional group (p < 1.0 × 10⁻¹⁶ versus a random set of proteins of the same size and degree distribution drawn from the genome). A careful inspection of the PPI-plot revealed the presence of several proteins belonging to families of neuropeptides and/or their receptors, including VIP/PACAP and somatostatin. The list also includes some neurotransmitter receptors, such as GRIN1 ‘Glutamate Ionotropic Receptor NMDA Type Subunit 1’. As a second example, STRING analysis of the Reactome pathway ‘Signaling by GPCR’ (REAC:R-MMU-372790) revealed a significant PPI enrichment (p < 1.0 × 10⁻¹⁶) with 58 proteins involved in 165 edges. Of particular interest was the observation that the PPI-plot also demonstrated the presence of VIP/PACAP and somatostatin receptors (Supplementary Dataset S2). For some terms such as “Dilated cardiomyopathy” (KEGG:05414; p = 0.325) or “Arrhythmogenic right ventricular cardiomyopathy” (KEGG:05412; p = 0.0591), STRING analysis showed no significant PPI enrichment (Table 1). Altogether, the STRING analysis showed significant protein–protein enrichments for all leading GO resource terms and KEGG/Reactome pathways associated with cell-to-cell communication. Furthermore, these data call attention to the pivotal role of proteins functioning as ligands or receptors in the context of cell communication.

Table 1.

STRING analysis of protein–protein interactions (PPI) of intersecting genes between our query and identified GO terms, and KEGG/Reactome pathways after g:Profiler context organization analysis. Data are presented as follows: number of nodes (i.e., intersecting genes/proteins dysregulated between Control and PAE Placenta–Cortex signatures), number of observed connecting edges, number of expected edges and PPI enrichment p-value. Lines of the table highlighted in green represent protein clusters related to cell–cell communication and selected for KEGG pathway enrichment analysis. The four highlighted protein–protein interaction lists are provided as tabular forms in the Supplementary Dataset S2.

Source	Term_ID	Term_Name	Number of Nodes:	Number of Edges:	Average Node Degree:	Avg. Local Clustering Coefficient:	Expected Number of Edges:	PPI Enrichment p-Value:
GO:MF	GO:0005515	Protein binding	871	2903	6.67	0.373	2424	<1.0 × 10−16
GO:MF	GO:0140110	Transcription regulator ativity	149	158	2.12	0.347	40	<1.0 × 10−16
GO:MF	GO:0008094	ATP-dependent activity, acting on DNA	9	20	4.44	0.802	1	<1.0 × 10−16
GO:MF	GO:0003677	DNA binding	193	401	4.16	0.393	118	<1.0 × 10−16
GO:MF	GO:0046873	Metal ion transmembrane transporter activity	44	89	4.05	0.402	6	<1.0 × 10−16
GO:BP	GO:0065007	Biological regulation	931	3066	6.59	0.372	2483	<1.0 × 10−16
GO:BP	GO:0042592	Homeostatic process	136	269	3.96	0.411	166	1.1 × 10−13
GO:BP	GO:0006796	Phosphate-containing compound metabolic process	161	312	3.88	0.458	213	1.57 × 10−10
GO:BP	GO:1901696	Response to nitrogen compounds	88	189	4.3	0.447	86	<1.0 × 10−16
GO:BP	GO:0003013	Circulatory system process	48	45	1.88	0.5	16	2.17 × 10−9
GO:CC	GO:0110165	Cellular anatomical structure	1327	4537	6.84	0.375	4198	0.000000102
KEGG	KEGG:04080	Neuroactive ligand-receptor interaction	38	73	3.84	0.533	4	<1.0 × 10−16
KEGG	KEGG:04713	Circadian entrainment	7	4	1.14	0.714	1	0.0362
KEGG	KEGG:05414	Dilated cardiomyopathy	6	2	0.667	0.667	1	0.325
KEGG	KEGG:04015	Rap 1 signaling pathway	18	33	3.67	0.59	6	3.22 × 10−15
KEGG	KEGG:05412	Arrhythmogenic right ventricular cardiomyopathy	4	1	0.5	0.5	0	0.0591
KEGG	KEGG:04725	Cholinergic synapse	7	9	2.57	0.357	2	0.0000456
KEGG	KEGG:05032	Morphine addiction	5	3	1.2	0.6	0	0.0148
REAC	REAC:R-MMU-112316	Neuronal system	40	110	5.5	0.644	20	<1.0 × 10−16
REAC	REAC:R-MMU-112315	Transmission across chemical synapses	20	49	4.9	0.658	9	<1.0 × 10−16
REAC	REAC:R-MMU-380320	Recruitment of NuMA to mitotic centrosomes	6	4	1.33	0.556	0	0.000000794
REAC	REAC:R-MMU-380270	Recruitment of mitotic centrosome proteins	5	4	1.6	0.667	0	0.0000000797
REAC	REAC:R-MMU-380287	Centrosome maturation	5	4	1.6	0.667	0	0.0000000797
REAC	REAC:R-MMU-372790	Signaling by GPCR	58	165	5.69	0.517	25	<1.0 × 10−16
INPUT QUERY		1583 terms	1473	5049	6.86	0.376	4708	0.000000464

2.3. KEGG Enrichment Analysis of Leading GO, KEGG, and REAC Terms

A ShinyGO analysis was then conducted on the four relevant sub-ontologies that appeared to be associated with cell communication or vascular processes: ‘Metal ion transmembrane transporter activity’ GO:MF:0046873, ‘Circulatory system process’ GO:BP:0003013, ‘Neuroactive ligand–receptor interaction’ KEGG:04080, and ‘Signaling by GPCR’ REAC:R-NMU-372790. The goal was to determine if, by interrogating the KEGG pathway database, the four gene lists revealed enriched pathways of interest and to evaluate the overlap of these pathways across the four queries (Figure 5).

Figure 5.

KEGG pathway enrichment analysis for the four identified GO terms and KEGG/Reactome pathways. (A–D) Charts representing significantly enriched pathways obtained by querying KEGG pathway database with the list of genes constituting each of the four networks identified. For each KEGG pathway: The size of the spot indicates the number of query genes found in the pathway; The color indicates the −log10(FDR) value. (A) Analysis of the GO:0046873 “Metal ion transmembrane transporter activity”. (B) Analysis of the GO:0003013 “Circulatory system process”. (C) Analysis of the pathway KEGG:04080 “Neuroactive ligand–receptor interaction”. (D) Analysis of the pathway REAC:R-MMU-372790 “Signaling by GPCR”. (E) Data table indicating rank, −log10(FDR), number of genes, Fold enrichment and gene list for each query. Each gene name is indicated in color, with genes expressed in the cortex vs. the placenta shown in green, and genes expressed in the placenta vs. the cortex shown in red.

For instance, Figure 5A depicts the results obtained from analyzing the leading term ‘Metal ion transmembrane transporter activity’ GO:MF:0046873 which, according to g:profiler, exhibited a 2.27-fold enrichment (p = 6.249 × 10⁻⁶) and comprises 55 genes that intersect with the Placenta–Cortex PAE inter-organ signature (Figure 1). Here, ShinyGO revealed that these 55 genes can predict 49 enriched KEGG pathways (the 20 pathways with the highest −log10(FDR) are displayed on the figure; Figure 5A). Interestingly, “Neuroactive ligand–receptor interaction” was significantly enriched for each of the four queries and was among the three most enriched for “Circulatory system process” (12 intersecting query genes; 11.6-fold enrichment; rank 3; FDR = 2; Figure 5B,E), “Neuroactive ligand–receptor interaction” (48 intersecting query genes; 55.9-fold enrichment; rank 1; FDR = 7.317; Figure 5C,E), and “Signaling by GPCR” (33 intersecting query genes; 27.29-fold enrichment; rank 1; FDR = 1.4412; Figure 5D,E). Detailed data for all enriched pathways are available in Supplementary Dataset S3.

The edition of the “Neuroactive ligand–receptor interaction” KEGG graphs, which are based on the ShinyGO analysis of the three inputs, revealed several families of dysregulated ligands and/or receptors in the PAE Placenta–Cortex transcriptomic signature, some of which are shared by the three graphs (Figure 6). Among these ligands and receptors, angiotensinogen and its receptors were found dysregulated as previously shown and validated in a targeted angiogenesis study [7]. Moreover, other ligands and/or receptors were found dysregulated such as serotonin receptors, gastrin-releasing peptide and its receptor, glutamate receptors, or the VIP/PACAP family receptors (Figure 6). Altogether, this data pointed out new candidate signaling pathways dysregulated by PAE between the placenta and the fetal brain.

Figure 6.

Ligand/receptor families belonging to the “Neuroactive ligand–receptor interaction” pathway found simultaneously enriched by the ShinyGO analysis of the leading GO, KEGG, and REAC terms. Boxes represent gene-encoded receptors or ligands. Circles represent non-protein ligands (metabolites). The 38 genes encoding ligands and/or receptors exhibit an altered inter-organ expression balance between the Control and PAE Placenta–Cortex signatures and present in the KEGG:04080 pathway “Neuroactive ligand–receptor interaction” are displayed in color: Red boxes are query genes intersecting ONLY with KEGG:04080 “Neuroactive ligand–receptor interaction” pathway; Blue boxes visualize query genes intersecting with both KEGG:04080 “Neuroactive ligand–receptor interaction” AND GO:0003013 “Circulatory system process”; Green boxes visualize query genes intersecting with both KEGG:04080 “Neuroactive ligand–receptor interaction” AND REAC:R-MMU-372790 “Signaling by GPCR”; Yellow boxes represent query genes intersecting with the three pathways.

2.4. Protein Validation of Different Ligand and/or Receptors Found Dysregulated by the Transcriptomic Analysis Between the Placenta and the Fetal Cortex After PAE

Global analysis of the PAE Placenta–Cortex signature indicated a marked dysregulation of cell–cell communication with several ligand and/or receptors with altered inter-organ expression ratios. In order to validate these transcriptomic data at the protein level, we performed Western blot studies in control and alcohol-exposed placentas and their matched fetal cortices at E20 on different ligands or receptors, i.e., PACAP, PAC1R, VIP1R (Figure 7 and Figure 8), LEPTINR (Figure 9), and SSTR2 (Supplementary Figure S1). In the control group, PACAP protein expression is not significantly different between the placenta and the cortex of both male and female embryos (Figure 7A,C; Supplementary Table S8). In contrast, PAE tends to induce different PACAP protein expression between the two organs in males (p = 0.0938; effect size: 0.56; Figure 7D). No effect of PAE was found regarding PACAP expression between the placenta and the fetal cortex of females (Figure 7D). When considering a given organ, PAE tends to reduce PACAP expression in placentas of both male and female fetuses (p = 0.0513; effect size: 0.43, Figure 7E; Supplementary Table S8). In fetal cortices, PAE markedly decreased PACAP expression in males (p < 0.05; effect size: 0.60) while no effect was found in females (Figure 7F). Altogether, these data suggest that PAE alters the inter-organ expression balance of PACAP protein between the placenta and the fetal cortex with a more pronounced effect in males.

Figure 7.

Protein validation of PACAP expression in paired placenta/cortex extracts from control and alcohol-exposed mice. (A,B) Western blots visualizing the relative expression of PACAP in placentas and paired cortices of males (A) and females (B) fetuses from control and prenatal alcohol-exposed mice. (C) Quantification in the control group of the relative placenta/cortex expression of PACAP in males (left), females (middle) and pooled sexes (right). (D) Quantification in the alcohol group of the relative placenta/cortex expression of PACAP in males (left), females (middle) and pooled sexes (right). (E) Effect of PAE on the relative PACAP expression in the placenta from male (left), female (middle) and pooled (right) fetuses. (F) Effect of PAE on the relative PACAP expression in the cortex from male (left), female (middle) and pooled (right) fetuses. p = 0.0513; p = 0.0938; * p < 0.05 using the Wilcoxon test (C,D) and the Mann–Whitney test (E,F).

Figure 8.

Protein validation of PAC1 and VIPR1 expression in paired placenta/cortex extracts from control and alcohol-exposed mice. (A,B) Western blots visualizing the relative expression of PAC1 in placentas and paired cortices of male (A) and female (B) fetuses from control and alcohol-exposed mice. (C) Quantification in the control group of the relative placenta/cortex expression of PAC1 in males (left), females (middle), and pooled sexes (right). (D) Quantification in the alcohol group of the relative placenta/cortex expression of PAC1 in males (left), females (middle), and pooled sexes (right). (E) Effect of PAE on the relative PAC1 expression in the placenta from male (left), female (middle), and pooled (right) fetuses. (F,G) Western blots visualizing the relative expression of VIPR1 in placentas and paired cortices of male (F) and female (G) fetuses from control and prenatal alcohol-exposed mice. (H) Quantification of the relative placenta/cortex expression of VIPR1 in males and females of the control group. (I) Quantification of the relative placenta/cortex expression of VIPR1 in males and females of the PAE group. (J) Comparison of VIPR1 expression in male and female placentas of control and PAE mice. (K) Comparison of VIPR1 expression in female and male cortices of control and PAE fetuses. * p < 0.05 using the Wilcoxon test (C,D,H,I) and the Mann–Whitney test (E,J,K).

Figure 9.

Protein validation of LEPTINR expression in paired placenta/cortex extracts from control and alcohol-exposed mice. (A,B) Western blots visualizing the relative expression of LEPTINR in placentas and paired cortices of male (A) and female (B) fetuses from control and alcohol-exposed mice. (C) Quantification in the control group of the relative placenta/cortex expression of LEPTINR in males (left), females (middle), and pooled sexes (right). (D) Quantification in the alcohol group of the relative placenta/cortex expression of LEPTINR in males (left), females (middle), and pooled sexes (right). (E) Effect of PAE on the relative LEPTINR expression in the placenta from male (left), female (middle) and pooled (right) fetuses. (F) Effect of PAE on the relative LEPTINR expression in the cortex from male (left), female (middle), and pooled (right) fetuses. * p < 0.05, ** p < 0.01 using the Wilcoxon test (C,D) and the Mann–Whitney test (E,F).

PAC1R is the high affinity receptor of PACAP [18]. In both prenatal treatment groups, Western blot studies performed at GD20 showed that PAC1R is expressed more in the placenta than in the fetal cortex in both female and male fetuses (Figure 8A–D). PAE increased PAC1R expression in placentas to a similar degree in both male and female fetuses and this effect was significant when both sexes were pooled (p < 0.05; effect size: 0.507; Figure 8E). In addition to PAC1R, PACAP also binds VIPR1 and VIPR2 receptors [18]. In contrast to PAC1R, Western blots showed that in the control group, VIPR1 is more expressed in the cerebral cortex of both females (p = 0.0625; EF: 0.904) and males (p < 0.05; effect size: 0.89; Figure 8H) and a similar expression profile is found in the alcohol group. In contrast to PAC1R, VIPR1 expression was not significantly different in the placenta or fetal cortex of PAE mice compared to controls (Figure 8J,K).

LEPTINR was another receptor found dysregulated in the interorgan transcriptomic analysis. Western blot analysis showed that in both control and alcohol groups LEPTINR is significantly more expressed in the fetal cortex than in placentas (Figure 9A,D; Supplementary Table S8). This expression pattern was observed in males, females, and when both sexes were pooled (p < 0.01; effect size: 0.886 for the control group; p < 0.01; effect size: 0.89) for the alcohol group although dispersion was more pronounced in females (Figure 9C,D). When considering a given organ, no significant effect of PAE was found even if, when compared to the control group, the median values of LEPTINR were lower in the alcohol placenta and higher in the alcohol cortex of both sexes (Figure 9E,F). Finally, in similar experiments, SSTR2 was also tested and slightly detected by Western blot in either placenta or fetal cortex. No significant effects were found regarding this receptor (Supplementary Figure S1).

Altogether, protein validation experiments performed on several ligand or receptor candidates (PACAP, PAC1R, VIPR1, LEPTINR, and SSTR2) showed different protein expression profiles associated to different PAE-induced effects, thus contributing to refine the transcriptomic analysis.

3. Discussion

3.1. PAE and the Placenta–Brain Axis

The deleterious effects of alcohol consumption during pregnancy on the developing fetal brain has been demonstrated for a long time and numerous studies have highlighted the effects of alcohol on major neurodevelopmental functions such as neurogenesis [19], apoptosis [20], or cell differentiation [21]. On the other hand, the brain is not the only organ impacted by PAE. Several research groups recently reported that PAE markedly impairs the transcriptomic profile of the placenta, in both clinical [9] and animal models of PAE [22,23]. For example in rat, using a moderate drinking model, Savage and co-workers showed that alcohol modified the expression of placental genes encoding calcium binding proteins, matrix metalloproteinases, and the angiogenic factor PlGF [24], while, in mouse, Pinson and co-workers showed that PAE upregulated the expression of members of the Notch signaling pathway involved in the control of angiogenesis [23]. Finally, in a prospective cohort in Cape Town, a RNA-Seq analysis revealed abnormal cell-type markers in placenta from the FASD group [25]. The emergence of neuroplacentology supports the observation that the effects of PAE on the development of the fetal brain vasculature could be linked to placental dysfunction [5,7]. Indeed, first, the expression of several angiogenic factors expressed in the placenta, such as PlGF, CD146, or VEGFR1, are dysregulated by PAE [5,6]. Second, these factors are released and exist in circulating forms [26,27] and third, it has been recently shown that a targeted repression of placental CD146 and PlGF mimics some PAE-induced vascular defects in the fetal cerebral cortex [5,6]. Conversely, placental overexpression of PlGF rescued the effects of PAE on the fetal cerebrocortical vasculature [6]. Consistent with these data, the recent publication of the first inter-organ transcriptomic signature focused on vascular development showed that the expression of several angiogenic factors was dysregulated between these two organs [7]. Considering that alcohol exerts multiple effects in a wide variety of cell types, all these studies support the interpretation that the dysregulation of the placenta–brain axis by PAE adversely impacts cell–cell communication beyond angiogenesis.

3.2. Global Analysis of the PAE Placenta–Cortex Transcriptomic Database Confirmed the Vascular-Targeted Interorgan Signature and Unveiled New Dysregulated Processes

The comprehensive and unbiased enrichment analysis of the inter-organ transcriptomic Placenta–Cortex database for PAE (NCBI; accession number GSE241836; (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE241836, accessed on 1 september 2023) constitutes the first high-throughput analysis of transcriptomic data with a simultaneous mining of Gene Ontology Resource, KEGG and Reactome databases. Several highlights raised from this analysis were noted. First, GO terms found significantly enriched are structured within a limited number of context maps. Among them, the context map “Circulatory system process” is consistent with the previous targeted studies performed in FASD models [5] and in human [6,9], which supported that PAE disrupts the angiogenic communication between the placenta and the fetal brain. Second, 11 of the 34 enriched context maps identified as exhibiting significant GO relations in the global transcriptomic interorgan analysis are related to cell–cell communication and ligand–receptor binding, such as Protein binding, Metal ion transmembrane transporter activity, Signaling receptor complex adaptor activity, or Signaling by GPCR. Moreover, the STRING analysis revealed substantial PPI enrichments indicating that proteins resulting from the expression of the PAE-dysregulated genes were sharing common functional/mechanistic processes. Among them, several ligands and/or receptors belonging to the somatostatin, VIP/PACAP, LEPTIN, glutamate, or progesterone families were found dysregulated in the global analysis of the Placenta–Cortex transcriptomic signature. Interestingly, except for the Agt/Agtr family, which contributes to the control of angiogenesis, most of these ligand/receptor families were not identified in the targeted angiogenesis transcriptomic analysis [5]. Overall, these data suggest that, in addition to vascular development, PAE impairs other physiological functions involved in a placental–fetal brain communication.

3.3. The Comprehensive Analysis of the PAE Placenta–Cortex Signature Unveils Candidates for Dysregulated Neuroactive Pathways

A survey of the literature provides different interpretations regarding the developmental relevance of the ligand–receptor families dysregulated by PAE in the Placenta–Cortex transcriptomic signature. For example, human placenta is known to be a source of non-neuronal acetylcholine and the main functions of the cholinergic system in placenta include regulation of intracellular amino acid uptake, blood flow regulation, and release of prostaglandin hormones [28]. Concerning the somatostatin family, several somatostatin receptors are expressed by trophoblasts in the placenta [29,30]. Moreover, the presence of high levels of somatostatin in the umbilical artery suggest that the neuropeptide detected in the feto–placental circulation originates from the fetus [31]. Whereas the effects of somatostatin in the developing brain are well documented, studies investigating the effects of somatostatin on placenta remain few. For example, in vitro experiments showed that somatostatin stimulates the migration of trophoblasts [29]. In the present study, the inter-organ transcriptomic analysis suggested that the SSTR2 receptors were dysregulated. However, SSTR2 was slightly detected, and no such effect was retrieved at the protein level. Altogether, these data suggest that somatostatin receptors would not be prioritized for functional validation studies. LEPTIN is a protein hormone involved in the control of energy balance [32]. In human placenta, LEPTIN is expressed by trophoblasts, and recent studies showed that its expression is controlled by PlGF and VEGF [33], two angiogenic factors dysregulated by PAE [6]. In terms of function in the placenta, it was reported that LEPTIN promotes invasion of cytotrophoblasts [34]. Protein validation showed that the LEPTIN receptor is significantly more expressed in the fetal cortex than in the placenta. Considering that circulating LEPTIN is detected in the umbilical vein [35], investigating the effects of gain/loss of function experiments targeting members of the LEPTIN family on a placenta–brain axis would make sense [36]. Based on protein validation experiments, the PACAP family appeared particularly interesting. Indeed, several studies on the PACAP family demonstrated trophic effects of these neuropeptides during neurodevelopment [37]. In the placenta, PACAP is expressed by stroma cells surrounding the blood vessels within villi [38], and it is also detected in the umbilical vein blood [39]. In the developing central nervous system, PACAP receptors are expressed from embryonic day 9.5 in the neural tube, and the neuropeptide has been shown to exert neurotrophic effects [40]. As an example, PACAP has been shown to regulate forebrain neural stem cells and neurogenesis [41]. Protein validation showed that PACAP, PAC1R, and VIPR1 receptors are differently expressed between the placenta and the fetal cortex, and that, in addition, PAE differently impacted their expression. Altogether, protein validation experiments performed on different candidates unveiled by the transcriptomic analysis of the PAE Placenta–Cortex signature revealed different validation profiles, paving the way for targeted functional validation studies.

3.4. Towards Functional Validations of the Candidate Dysregulated Pathways

Omics technologies including transcriptomics analysis represent an exceptional source of large-scale data with promising value to elucidate new dysregulated processes involved in pathologies. However, a key challenge remains the biological validation of bioinformatics transcriptomics-based analysis [42]. The present study provides to our knowledge the first list of neuroactive candidates linking PAE and neuroplacentology. These factors would be indicators of a PAE-induced endocrine placenta–cortex dysfunction. In order to strengthen the validity of this list of factors, we performed a Western blot validation of several dysregulated ligands or receptors from different neuroactive families. Protein validation gave several indications: First, Western blot experiments confirmed that, as found by microarrays, PACAP, PAC1R, VIP1R, and LEPTINR were differentially expressed between the two organs whereas SSTR2 was not. For example, PAC1R is mostly expressed in the placenta while VIP1R is mainly detected in the fetal cortex. Second, regarding a given organ, PAE differentially impacted the expression of some proteins. For example, prenatal ethanol exposure decreased PACAP expression in the placenta of both females and males whereas it increased PAC1R expression. In the fetal cortex, PAE decreased PACAP expression only in males and no effect was found regarding VIPR1. Taken together, these data indicate that even if there are several, not all candidates listed after the transcriptomic analysis of the PAE interorgan signature are associated with a protein dysregulation. Such a protein validation appears a pre-requisite before performing gain/loss of function studies, as we recently carried out for the angiogenic function [5].

3.5. Limitations and Perspectives of the Study

Design of the database: The main objective of the present study was to re-analyze with a comprehensive approach the gene list resulting from the recently published and deposited PAE-Placenta/Cortex signature (NCBI’s Gene Expression Omnibus and accessible through the Series accession number GSE241836) [7]. Regarding the implementation of the study, the bulk approach was prioritized for two reasons: (i) to favor emergence of robust modifications in the PAE signature, and (ii) when the project was engaged, no clear evidence was in favor of targeting a given cell type (for example trophoblasts) rather than another one (for example endothelial cells). However, the bulk strategy also presents limitations such as missing fine but biologically important cell-specific dysregulations. Furthermore, protein validation experiments revealed sex-specific differences for one neuroactive candidate. The transcriptomic database was built with mixed sexes (equal male/female ratio). This choice was justified by several reasons: (i) previous studies from our research group showed no sex differences [21,43], and (ii) epidemiological studies showed no statistically significant difference between sexes [44]. However, in the light of publications showing a sex dysmorphism of the placenta [45], the present study on PAE and neuroplacentology paves the way of “cell-targeted and sex” interaction studies such as a trophoblasts/oligodendrocyte precursors transcriptomic signature of PAE.

Identification of cardiovascular enriched pathways: Whereas the aim of the present study was to analyze a Placenta–Cortex transcriptomic database of PAE, we were surprised to see that heart-related pathways were also significantly enriched. These results suggest that pathways involved in a placenta–brain axis may be also shared with cardiovascular development. This hypothesis could make sense when considering that (i) PAE is known to affect cardiovascular development in human [46], (ii) in animal models, PAE predisposes to cardiomyopathy [47,48], and (iii) previous data evidenced several dysregulated placental angiogenic factors [5,6]. Among research perspectives, it would be interesting to use a gain/loss of function approach to explore the contribution of angiogenic placental genes on PAE-induced cardiomyopathies.

3.6. Neuroplacentology and Clinical Considerations of FASD

Emerging evidence on neuroplacentology supports that placental factors may have a biomarker value of neurodevelopmental outcomes [13]. Regarding FASD, despite active primary prevention, alcohol consumption during pregnancy remains a leading cause of disability and neurodevelopmental disorders in the world [14]. FAS, the most severe expression of FASD, can be early diagnosed based on characteristic cranio-facial dysmorphisms. However, the overwhelming majority of FASD children do not display these dysmorphisms yet will progressively develop neurobehavioral troubles frequently misdiagnosed until school-age [49]. Individuals with FASD will experience life-long educational and socio-professional consequences. Because the placenta is an ephemeral organ destroyed at birth, the identification of placental factors possibly involved in the development of the fetal brain may have a diagnosis value. The present study provides a list of neuroactive candidates dysregulated by PAE-between the placenta and the fetal brain and paves the way for functional validation studies. Behind FASD, a recent study on infants showed that some of the placental factors that we found dysregulated are also dysregulated in children with autism spectrum disorder [50], highlighting the placenta’s value in diagnosis and precision medicine [51].

In conclusion, the present study provides the first global and unbiased analysis of the murine placenta–cortex transcriptomic profiles associated with prenatal alcohol exposure. Beyond coordinated alterations in gene expression related to vascular development in both organs, the present data suggest that alcohol exposure could be associated with broader changes in placenta–cortex transcriptomic relationships. In particular, several neuroactive candidate genes involved in the control of energy balance and neurodevelopment have been identified. Although based on transcriptomic associations, these findings strongly support that neuroplacentology constitutes a promising research avenue to propose hypotheses on processes involved in perinatal and long-term PAE-induced neurodevelopmental pathologies, warranting future functional validation.

4. Materials and Methods

4.1. Mouse Model of Prenatal Alcohol Exposure

All mice were housed in a temperature-controlled room (21 ± 1 °C) with a 12 h/12 h light/dark cycle (lights on from 7 a.m. to 7 p.m.) with free access to food and tap water. Animal procedures followed the guidelines of the French Ethical Committee and the European Directive (Council Directive 2010/63/EU, license no. 21CAE035), authorization no. APAFIS#22136-2019092013438607 v4 from the Ministère de l’Enseignement Supérieur, de la Recherche et de l’Innovation delivered the 03/06/2020 (Ethics committee CENOMEXA). Experiments were carried out under the supervision of authorized investigators (B.J.G., authorization no. 7687 from the Ministère de l’Agriculture et de la Pêche). The procedures for in vivo treatment of pregnant mice, as well as tissue collection and preparation, were described in our previous publication [5]. In brief, twelve National Marine Research Institute (NMRI) mice were provided by Janvier (Le Genest-Saint-Isle, France). Pregnant mice received daily subcutaneous injections of either sodium chloride (NaCl 9‰) or alcohol (3 g/kg, Fisher Scientific) diluted in NaCl (50%, v/v) from gestational day (GD) 15 to GD20. Afterwards, placentas and corresponding fetal cortices were collected at GD20, and total RNA was isolated using the NucleoSpin^® RNA Plus kit (Macherey-Nagel, Hoerdt, France). No specific criteria of exclusion were set. One-color comparative microarray analyses were performed with the Whole Mouse Genome Oligo 4_44K Microarray (G2519F-014868, Agilent Technologies, Les Ulis, France).

4.2. Constitution of the Placenta–Cortex Transcriptomic Database of PAE

The database of the murine Placenta–Cortex transcriptomic signature after PAE was previously established and deposited in the NCBI’s Gene Expression Omnibus (GEO; [7]) and accessible through the Series accession number GSE241836 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE241836, accessed on 1 september 2023). Briefly, a whole mouse genome microarray of paired placentas and cortices arising from different litters and from four independent experiments was performed to establish two transcriptomic inter-organ “placenta–cortex” signatures for control and PAE groups on gestational day 20 (Figure 10A–C). Practically, one-color comparative hybridization was performed using Whole Mouse Genome Oligo 4_44K Microarray (G2519F-014868, Agilent Technologies) to compare gene expression profiling. cRNAs were synthesized from 100 ng total RNA, labeled using Quick Amp Labeling Kit (Agilent Technologies), and hybridized on microarrays at 65 °C for 17 h. Raw hybridization data, evaluated on every probe 5 µm-sized array, using DNA microarray scanner G2565CA (Agilent Technologies), were extracted with Feature Extraction Software 10.5.1.1 (Agilent Technologies), then transferred to Genespring^® (GX 12.6 software, Agilent Technologies) for data processing (normalization) and data mining. Data were normalized by 75th percentile and in each array, outlier spots and those exhibiting a heterogeneous signal or not above background were discarded. Data scale-up was performed at “baseline to median of all samples”. Volcano-plots were generated supported by statistical significance of differential gene expression (fold change cut-off = 2) performed with Student’s t-test with Benjamini Hochberg correction for multiple testing to check the False Discovery Rate (FDR; p value < 0.05) [7]. Venn analysis of the two interorgan signatures was used to generate six lists of genes (Figure 10B): (1) transcripts over-expressed (at least two fold) in the cortex versus placenta in control mice (300 genes), (2) transcripts over-expressed (at least two fold) in the cortex versus placenta in PAE mice (593), (3) transcripts under-expressed (at least two fold) in the cortex versus placenta in control mice (111), (4) transcripts under-expressed (at least two fold) in the cortex versus placenta in PAE mice (209), (5) genes over-expressed in placenta versus cortex in both signatures but whose levels were modified by at least 40% under ethanol treatment (177), and (6) genes over-expressed in cortex versus placenta in both signatures but whose levels were modified by at least 40% under ethanol treatment (192) [7]. The 40% threshold was applied to filter genes common between the Control and PAE inter-organ signatures and whose interorgan ratio was modified by PAE. A +/−40% cut-off was chosen to ensure high robustness and stringency of the PAE effects. This cut-off is compatible with the range of variations previously reported regarding the effect of PAE on genes from the VEGF family [6,15]. The processing of these data was performed as follows: (i) removal of genes from RIKEN project list (cDNAs without gene names) or unknown, (ii) selection of genes found dysregulated between the two signatures in at least 3 or 4 independent experiments (statistical threshold of 0.75) and resulted in a list of 1582 genes representative of the Placenta–Cortex transcriptomic signature of PAE. These 1582 genes were submitted to g:Profiler for a global analysis (Figure 10D–G).

Figure 10.

Experimental design of the study. (A) PAE protocol applied on pregnant mice prior to microarray gene expression analysis (Agilent). (B) Venn analysis used to sort the genic probes exhibiting a ≥2-fold expression variation (upper or lower) between cortex versus placenta in control and/or alcohol groups (GeneSpring). (C) Establishment of PAE Placenta–Cortex signatures, based on data filtration to retain genes with variations in at least 3 or 4 independent experiments [7]. (D) Multi-database analysis of the PAE transcriptomic placenta–cortex gene list interrogating Gene Ontology, KEGG and Reactome databases simultaneously. The output indicates the number of enriched terms and/or pathways (g:GOSt). (E) Analysis of relations between the most significant GO terms identified within each GO classification (Molecular Function, Biological Process, and Cell Component). The output consists of context maps organized as ontologies and linking GOs to a leading term (g:profiler/GO Context). (F) Leading terms with no significant relations are excluded from further analysis. (G) Among the 11 GO terms and 7 KEGG and 6 Reactome pathways identified, four are related to intercellular communication. (H) The four pathways exhibit significant protein–protein interactions (STRING). (I) The gene lists of each of the four GO/pathways related to intercellular communication were submitted to enrichment analysis (ShinyGO). The “Neuroactive ligand–receptor interaction” pathway is commonly enriched for three gene lists queries. (J) Comparison of KEGG networks allows identification of candidate ligands and receptors. (K) Protein validation of a PAE-dysregulated ligand/receptor pathway by Western blot and placenta/cortex correlation analysis.

4.3. g:Profiler Analysis

The gene list of the Placenta–Cortex transcriptomic signature after PAE was submitted to g:Profiler ([17]; https://biit.cs.ut.ee/gprofiler/gost, accessed on 1 september 2023) for functional profiling. gProfiler’s g:GOST tool performs a statistical enrichment analysis to identify overrepresentation of Gene Ontology terms (http://geneontology.org), KEGG pathways (https://www.genome.jp/kegg/pathway.html, accessed on 1 september 2023), and Reactome Pathways (https://reactome.org/; Figure 10D). Concerning the Gene Ontology Resource analysis, significantly enriched GO terms were extracted according to each classification: Molecular Function, Biological Process, and Cell Component. These terms were subsequently submitted to a GO context analysis to reduce the list of terms to the most relevant and to reorganize the significant terms according to their relationships (Figure 10E). The output consists of ‘GO context maps’ representing sub-ontologies of GO and derived from a leading term (Figure 10F). Concerning KEGG pathway and Reactome pathway databases, the g:GOSt tool directly provides significantly-enriched pathways. The settings used in g:Profiler are reported in Supplementary Table S6.

4.4. STRING Analysis

Query genes intersecting with the g:Profiler output (GO terms and enriched KEGG and Reactome pathways; Figure 10G) were submitted to a STRING protein–protein interaction networks analysis (PPI) to identify significantly enriched functional protein–protein interactions (https://string-db.org/; Figure 10H). The analysis consisted of identifying predicted associations for a group of proteins based on bioinformatic sources including text mining, databases, experiments, co-expression, neighborhood, gene fusion, and co-occurrence [52,53]. The outputs are (i) a graphical network of predicted associations, displaying proteins as nodes and interactions as edges. Confidence of PPI is symbolized by edge thickness, (ii) a statistical analysis which provides the average node degree (reflecting the number of interactions of proteins within the network) and the PPI enrichment p-value (indicating that connections are not random). Such an enrichment indicated that the proteins were at least partially biologically connected as a group. The settings used in STRING are reported in Supplementary Table S6.

4.5. ShinyGO Analysis

Query genes intersecting with the g:Profiler output (GO terms and enriched KEGG and Reactome pathways; Figure 10G) were submitted to a ShinyGO analysis (https://bioinformatics.sdstate.edu/go/, accessed on 1 september 2023) in order to search for functional pathways (Figure 10I). The outputs are (i) charts representing significantly enriched pathways. For each KEGG pathway: the size of the spot represents the number of query genes found in the pathway; the color indicates the −log10(FDR) value, (ii) a hierarchical clustering tree based on shared genes between pathways. Bigger dots indicate more significant p-values. The settings used in ShinyGO are reported in Supplementary Table S6.

4.6. Western Blot Validation

At GD20, placentas and matched E20 cortices from three different litters were homogenized in ice-cold lysis buffer (Cell Signaling Technology, Danvers, MA, USA). The homogenates were centrifuged (18,000× g; 20 min), and the resulting supernatants were used for Western blot analysis (Figure 10K). For each cortical and placental sample, fifty micrograms of protein extract, determined via the Bradford assay, were denatured at 100 °C for 5 min in Laemmli buffer (0.5 M Tris-HCl, pH 6.8; 8% SDS; 0.5% bromophenol blue; 10% glycerol; 10% β-mercaptoethanol) and separated on a 10% SDS–polyacrylamide gel. After electrophoresis, proteins were transferred onto nitrocellulose or PVDF membranes. Depending on the primary antibody, membranes were incubated in different blocking solutions (1× TBS; 0.05% Tween 20; 5% nonfat dry milk, or 1× TBS; 0.05% Tween 20; 5% BSA) at room temperature for 1.5 h and then, incubated overnight with the primary antibodies listed in Supplementary Table S7. Following incubation with the appropriate HRP-conjugated secondary antibodies (Supplementary Table S7), protein signals were detected using an enhanced chemiluminescence ECL Plus system (Amersham Biosciences Europe GmbH, Freiburg, Germany). Band intensities were quantified using a blot analysis system (Bio-Rad Laboratories, Marne-la-Coquette, France) and normalized with GAPDH serving as a loading reference. Commercial molecular weight markers (SeeBlue prestained standard, Invitrogen, Carlsbad, CA, USA) were used as references. All original images for western blot are available in Supplementary Dataset S4.

4.7. Statistical Analyses

Statistical analysis of the PAE Placenta–Cortex transcriptomic database was done using GeneSpring GX (Agilent, Santa Clara, CA, USA) and detailed with its previous deposit in the NCBI’s Gene Expression Omnibus (GEO; accession number GSE241836, (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE241836) and publication [7]. To perform functional profiling, g:GOSt uses multiple testing correction and applies a tailor-made algorithm g:SCS for reducing significance scores when interrogating Gene Ontology Resource, KEGG, and Reactome databases with a hypergeometric test to calculate the significance of the specified functional terms in the provided input gene list. GO context relies on a two-stage, hybrid term list-filtering algorithm for GO enrichment results. This algorithm takes into account the underlying topology of annotations without introducing additional hyperparameters, while reducing the term lists retrieved by g:Profiler. Statistical analysis of the predicted protein–protein interactions (PPIs) was conducted using the STRING Network Analysis database (https://string-db.org). Network significance was assessed using the PPI enrichment p-value provided by STRING, which compares the number of observed network edges to the expected number for a random protein set. FDR-adjusted p-values reported by the STRING platform were used to assess statistical significance. Statistical analysis of enriched biological pathways was performed using ShinyGO analysis. Statistical significance was calculated by using hypergeometric p-value with multiple-testing correction yielding FDR; a significant FDR cutoff was fixed at 0.05. Statistical analysis of relative protein expression by Western blot was performed using Prism (GraphPad Software, La Jolla, CA, USA). Data were represented as box and whisker plots with bars indicating upper and lower values. A pair was defined as placenta and fetal cortex samples from the same fetus, and the unit of analysis was the individual fetus for both paired and independent comparisons. Paired comparisons between placenta and fetal cortex were analyzed with a two-tailed Wilcoxon matched-pairs signed-rank test, separately for males, females, and pooled groups. Comparisons between independent groups were analyzed using a two-tailed Mann–Whitney test. Statistical significance was set at p < 0.05, with a trend defined as 0.05 < p < 0.1 [54]. For both the Mann–Whitney and Wilcoxon tests, effect size was expressed as the correlation coefficient r, following Prism-compatible conventions:

$$r={\displaystyle \frac{Z}{\sqrt{N}}}$$

For the Mann–Whitney test, N corresponds to the total number of observations (N = N₁ + N₂). For the Wilcoxon matched-pairs signed-rank test, N corresponds to the number of matched pairs with non-zero differences. The magnitude of r was interpreted using Cohen’s thresholds: absolute values ≥0.10 indicate a small effect, ≥0.20 a medium effect, ≥0.30 a large effect, ≥0.40 a very large effect, and ≥0.71 a huge effect [55]. Data for each experimental group are detailed in Supplementary Table S8.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijms27041819/s1.

Author Contributions

Conceptualization, B.J.G., C.S., C.D. and M.L.; Data curation, A.F.-M., B.J.G., C.D., C.S. and M.L.; Formal analysis, A.F.-M., B.J.G., C.B.-L., C.D., C.S., D.D.S., F.M., G.N., M.L., S.G. and S.M.; Funding acquisition, B.J.G. and S.M.; Investigation, A.F.-M., C.B.-L., C.D., C.S. and M.L.; Methodology, A.F.-M., C.B.-L., C.D., C.S. and M.L.; Supervision, A.F.-M. and B.J.G.; Validation, F.M., G.N., S.G., D.D.S. and S.M.; Writing, original draft, A.F.-M., B.J.G., C.D., C.S. and F.M.; Writing—review and editing: A.F.-M., B.J.G., C.B.-L., C.D., C.S., D.D.S., F.M., G.N., M.L., S.G. and S.M. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

The animal study protocol was approved on 6 march 2020 by the Ethics Committee of the French Ministry of Education and Research (authorization no. APAFIS#22136-2019092013438607 v4 from the Ministère de l’Agriculture et de la Pêche) and was supervised by authorized investigators (BJG, authorization no. 7687 from the French Ministry of Agriculture and Fisheries).

Informed Consent Statement

Not applicable.

Data Availability Statement

The data discussed in this publication were deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE241836 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE241836, accessed on 27 August 2023).

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This work was supported by Normandy University (UMR1245), Rouen University, Institut National de la Santé et de la Recherche Médicale (INSERM, UMR1245; UMR-S U1237), Rouen University Hospital, Fondation de France, ANR AlcoBrain. C.S. was a recipient of a PhD grant from Rouen University.