Placental effects on the maternal brain revealed by disrupted placental gene expression in mouse hybrids

Abstract

The mammalian placenta is both the physical interface between mother and fetus, and the source of endocrine signals that target the maternal hypothalamus, priming females for parturition, lactation and motherhood. Despite the importance of this connection, the effects of altered placental signalling on the maternal brain are insufficiently studied. Here, we show that placental dysfunction alters gene expression in the maternal brain, with the potential to affect maternal behaviour. Using a cross between the house mouse and the Algerian mouse, in which hybrid placental development is abnormal, we sequenced late-gestation placental and maternal medial preoptic area transcriptomes and quantified differential expression and placenta-maternal brain co-expression between normal and hybrid pregnancies. The expression of Fmn1 and Drd3 was significantly altered in the brains of females exposed to hybrid placentas. Most strikingly, expression patterns of placenta-specific gene families and Drd3 in the brains of house mouse females carrying hybrid litters matched those of female Algerian mice, the paternal species in the cross. Our results indicate that the paternally derived placental genome can influence the expression of maternal–fetal communication genes, including placental hormones, suggesting an effect of the offspring's father on the mother's brain.

1. Background

The placenta is a unique, transient organ shared by two organisms. Placental morphology is surprisingly diverse across vertebrates and is subject to rapid evolutionary change and convergent evolution [1,2]. In most eutherian mammals, including mice and humans, successful blastocyst implantation relies on endometrial invasion by the embryonic trophoblast cells that give rise to the mature placenta [3]. As such, the placenta provides the closest physical and molecular link between mother and offspring seen in any animal [4]. This intimate connection promotes an array of maternal–fetal interactions, including bidirectional hormonal regulation and even the exchange of entire cells. These interactions are not spatially limited but extend to both the fetal and the maternal brain [5,6].

Throughout pregnancy, the placenta mediates the regulation of resource allocation, immune tolerance, fetal development and, importantly, hormonal priming of the maternal brain. A key subset of placenta-secreted molecules is involved in priming maternal physiology for parturition and lactation, and promoting the onset of maternal behaviours in late gestation [7,8]. In rodents, the medial preoptic area (MPoA) in the hypothalamus is thought to be the primary neural target of these placental molecules [5,9,10]. The MPoA has been characterized as the central hub of parenting behaviour [11]: receptors for key pregnancy hormones and neurotransmitters, including oestrogen, prolactin and dopamine, are highly expressed in this nucleus and interact with ligands of both maternal and placental origin [12].

Two classes of placental genes are of particular importance to the interaction between the placenta and maternal brain: imprinted genes (IGs) and placenta-specific gene families (PSFs). Several IGs and PSF genes are expressed in the same placental compartment, especially in the placental endocrine compartment [13,14]. IGs are exclusively or predominantly expressed from one allele, and are highly expressed in placenta and brain. An allelic expression is determined by heritable epigenetic marks (imprints) in maternal and paternal germ cells, such that some IGs are maternally silenced and paternally expressed, whereas others are paternally silenced and maternally expressed [15]. During pregnancy, IGs are critical to placental development and function, maintaining the balance between maternal supply and embryonic demand, and regulating maternal–fetal exchange [14].

PSFs arose through lineage-specific gene duplication events during placental evolution [16]. In rodents, these are the prolactin gene family (placental lactogens, Prls), placental cathepsin proteases and their inhibitors, and pregnancy-specific glycoproteins [17–19]. PSF gene products are mainly expressed from the placental endocrine compartment and many are secreted into the maternal bloodstream; key functions include placental development, immunoregulation, and physiological and neurological priming of the mother [5,9,16]. Most notably, a subset of Prls binds prolactin receptors in the maternal MPoA, leading to the proposal that placental hormones directly affect maternal endocrine state and behaviour [5,10]. IGs are implicated in regulating PSF secretion via their effects on the structure and function of the placental endocrine compartment [20]. However, our current understanding of the role of IGs in PSF signalling is rudimentary, and the relationship between gene expression in placenta and the maternal MPoA is uncharted.

The majority of the placenta, including the endocrine compartment, is fetally derived. Placental representation of both parental genomes sets the stage for conflict (maternal–paternal and parent–offspring), and for coadaptation (mother–offspring), with IGs uniquely positioned to mediate both types of interactions [21–23]. However, while evolutionary models for IG expression abound [24], empirical studies of the interaction between paternally derived placental signals and signal reception in the maternal brain are rare [8,25,26].

Here, we use a natural hybrid system to explore the effects of placental dysregulation on gene expression in the maternal brain. Over- or under-growth that depends on the direction of the cross is a signature of disrupted imprinting in mammalian hybrids [27]. This pattern is documented in several orders, with the best-studied examples in the rodent genera Peromyscus, Phodopus and Mus [28–30]. Parent-of-origin growth effects in the cross between the house mouse, Mus m. domesticus (Dom) and the Algerian mouse, M. spretus (Spret), were first described over 20 years ago: placentas are undersized when the mother is Dom and the father is Spret, and severely oversized in the reciprocal cross, with more extreme size effects in both directions of the backcross [28]. Subsequent studies confirmed altered expression and methylation of candidate IGs, and disrupted placental organization [31,32]. Specifically, the placental endocrine compartment (or junctional zone) was shown to be reduced and disorganized [28,31]. However, the extent of placental misexpression has not been measured on a genomic scale, and this system's potential to uncover the maternal consequences of altered placental signalling has not been considered.

We recently showed that maternal responsiveness to pups is significantly altered in Dom females with newborn hybrid relative to conspecific litters [33]. These effects of litter genotype were only detected during the first 24 h postpartum, suggesting that altered behaviour in mothers of hybrid litters is the residual consequence of abnormal placental signalling rather than a response to phenotypic differences between hybrid and conspecific pups [33]. Here, by comparing MPoA expression between females of the same species that differ only in the type of pregnancy/placenta they experience (hybrid versus conspecific), we specifically isolate the effect of placental gene expression differences on the maternal brain (electronic supplementary material, figure S1). Characterization of altered gene expression at the maternal–fetal interface provides insight into the mechanisms of maternal–fetal communication, the contribution of the paternal genome to this interaction, and identifies promising candidate genes for future evolutionary and biomedical work.

2. Material and methods

(a). Animals and tissue collection

Mice used in this study were maintained on a 12 : 12 light : dark cycle with lights on at 9.00, and were provided with 5001 Rodent Diet and water ad libitum. All animal procedures were approved by the Oklahoma State University IACUC under protocol 141-AS. Mus m. domesticus (Dom) was represented by the wild-derived inbred strain WSB/EiJ (Jackson Laboratory) and Mus spretus (Spret) was represented by the wild-derived inbred strain SFM/Pas (Montpellier Wild Mice Genetic Repository). We conducted three crosses (female shown first): Dom × Dom (Dom pregnancy), Dom × Spret (hybrid pregnancy) and Spret × Spret (Spret pregnancy). The reciprocal cross (Spret × Dom) was attempted 25 times but was never successful. Mice were paired between 17.00 and 18.00, left undisturbed for two nights and split on the morning of the second day. The second night was counted as embryonic day 0 (e0). Females were weighed after two weeks to confirm pregnancy but were otherwise left undisturbed. Pregnant females (n = 5/type of pregnancy) were euthanized by cervical dislocation between 10.00 and 11.00 on an embryonic day 17–18 (e17.5) and the maternal brain was extracted. We chose this late gestation time point based on prior work demonstrating that infusion of placental lactogen into the MPoA of near-term females promotes the onset of maternal care at parturition [5,9]. Embryos were separated from placentas, the maternally derived placental decidual layer was removed [34], and embryos and dissected placentas were weighed. All tissues were transferred to RNAlater, and stored at −20°C until microdissection and RNA extraction.

(b). Brain microdissection and RNA extraction

The maternal MPoA was localized using the mouse brain atlas (figures 26–33 in [35]), and microdissected by sectioning the RNAlater-perfused brain at 100 µm on a Leica CM 1950 cryostat, followed by dissection under a dissecting microscope in chilled PBS droplets for improved visibility of brain microstructure. DNA was extracted from embryonic tissue using the DNeasy Blood & Tissue Kit (Qiagen, USA) followed by PCR for the Y-linked gene, Zfy1, to determine sex. Placentas from one male and one female per litter were used for RNA extraction (n = 5 males per hybrid cross, n = 5 males per conspecific cross, n = 4 females per hybrid cross and n = 5 females per conspecific cross). RNA was extracted from all tissues immediately after microdissection using the RNeasy Plus Universal Mini Kit (Qiagen) for MPoA, and the AllPrep RNA/DNA Mini Kit (Qiagen) for placenta. RNA was stored at −80°C until sequencing.

(c). Transcriptome data processing and analysis

Maternal MPoA and placental transcriptomes were sequenced on the Illumina HiSeq 4000 platform, producing greater than 30 million, 150 bp paired-end reads per sample. To improve comparability, all placenta samples (Dom, Spret and hybrid) were mapped to a pseudo-hybrid genome, generated using the genome preparation tool in the programme SNPsplit (Babraham Bioinformatics [36]). MPoA samples were mapped to their corresponding genomes (WSB/EiJ_v1 for Dom MPoA, SPRET/EiJ_v1 for Spret MPoA [37]) using HISAT2 2.1 [38]. Post-processing of alignments (filtering and downsampling) was done using SAMtools 0.1.19 [39]. Transcript quantification and annotation were done using StringTie 1.3.3 [40] with gene annotation information retrieved from Ensembl (ftp://ftp.ensembl.org). To confirm equivalent expression from Dom and Spret alleles in the hybrid placenta we split the hybrid alignment files, separating reads originating from the Spret and Dom alleles. Splitting was performed using SNPsplit. Because the maternal brain is exposed to placental signals of both sexes simultaneously, male and female placental expression was analysed jointly. Differential expression was tested with DESeq2 1.16.1 [41]. Normalized read count tables produced by DESeq2 were used in subsequent co-expression analyses. GO-term and pathway overrepresentation analyses were performed using the PANTHER gene list analysis tool [42]. Additional information on RNAseq data processing and analysis is provided in electronic supplementary material, text S1.

(d). Evolutionary rates for selected genes

For evolutionary analysis of selected gene sequences, pairwise dN/dS (the per site ratio of non-synonymous to synonymous substitutions) between Dom and Spret was calculated using YN00 implemented in PAML 4.8 [43]. Further analysis of genes with dN/dS > 1 was performed with CodeML implemented in PAML 4.8 [43], including sequences from related Mus subspecies and species (Mus m. musculus, Mus m. castaneus, Mus caroli and Mus pahari). Detailed methods for evolutionary rates analyses are provided in electronic supplementary material, text S1.

3. Results

(a). Hybrid placental expression is globally closer to the maternal than the paternal species

As expected, large percentages of placental genes, including IGs, were differentially expressed (DE) in all three pairwise comparisons: hybrid versus Dom, 17.5% (3207/18 298 genes; 14/135 IGs), hybrid versus Spret 26.1% (4837/18 529 genes; 23/135 IGs), Dom versus Spret, 38.2% (7292/19 079; 32/135 IGs) (figure 1; electronic supplementary material, figures S2–S4, dataset S1). Consistent with less DE in the hybrid versus Dom than in the hybrid versus Spret comparison, hybrid samples clustered slightly closer to Dom in the diagnostic PCA plot (electronic supplementary material, figure S5). Notably, approximately twice as many genes in hybrid placentas were uniquely DE relative to Spret (3271) as opposed to Dom (1641) (figure 1a). Thus, the general expression pattern in hybrid placentas was more similar to the maternal species. Dom-like expressed genes were enriched for multiple immune related pathways, together with angiogenesis, vascular development and haemostasis related terms, among others (electronic supplementary material, dataset S1).

Figure 1.

DE gene analysis between Mus m. domesticus (Dom), M. spretus (Spret) and hybrid placentas. (a) Overlap of DE genes between the comparisons hybrid versus Dom and hybrid versus Spret; up = genes expressed higher in hybrids compared to parental species, down = genes expressed lower in hybrids compared to parental species. Results of pathway overrepresentation in text boxes. Volcano plot of DE results of (b) hybrid versus Dom and (c) hybrid versus Spret placentas. (Online version in colour.)

(b). Placenta-specific gene families and imprinted genes are enriched among differentially expressed genes in hybrid placentas

From the perspective of M. m. domesticus females carrying hybrid litters, abnormal placental expression constitutes all genes that are DE relative to Dom placentas. We divided these genes into three types. Transgressively expressed genes are significantly over- or underexpressed relative to both parental species; genes with intermediate expression are significantly overexpressed relative one parent but underexpressed relative to the other; genes with Spret-like expression are uniquely DE relative to Dom (figure 1a).

The 275 transgressively overexpressed genes were significantly enriched for B-cell receptor activation and integrin cell surface interaction pathways (figure 1a; electronic supplementary material, dataset S1). The 167 transgressively underexpressed genes were significantly enriched for pathways that involve Prls, including prolactin and growth hormone receptor signalling, epidermal growth factor (ERBB) signalling, and cytokine signalling (figure 1a; electronic supplementary material, dataset S1). Prls and other PSF genes were highly overrepresented among DE genes in hybrids compared to both Dom (Fisher's exact test: p < 0.001, odds ratio = 5.65) and Spret (p < 0.001, odds ratio = 3.41). Thirty-three PSF genes, including 14/22 Prls, were DE hybrid versus Dom placenta (electronic supplementary material, table S1). Of these, the majority (29/33) were underexpressed relative to Dom placenta.

IGs were also significantly overrepresented among hybrid DE genes compared with both Spret (Fisher's exact test: p = 0.02, odds ratio = 0.59) and Dom (p = 0.03, odds ratio = 0.55). Three IGs (Tnfrsf23, Phlda2 and Klf14) were transgressively overexpressed, two (Ascl2 and Sfmbt2) were transgressively underexpressed, and two (Tspan32 and Th) were significantly DE compared with both parental species but intermediate between the two (electronic supplementary material, table S2). Four of these DE IGs belong to the same imprinting cluster (IC2) on the distal part of mouse chromosome 7 (dist7) (MouseBook, https://www.mousebook.org, 22 March 2018), and are normally maternally expressed (electronic supplementary material, table S2 and dataset S1). We note that Tnfrsf23 is expressed in decidual cells at the junction between the maternally derived decidua and the extraembryonic placenta [44]. The IGs with Spret-like expression were all overexpressed relative to Dom and 4/6 are normally paternally expressed. The latter pattern contrasts with other DE IGs, of which only 1/7 is normally paternally expressed (electronic supplementary material, table S2).

(c). Altered medial preoptic area expression in mothers exposed to hybrid placentas

To test for effects of abnormal placental expression on the maternal brain, we compared e17.5 MPoA expression of Dom females carrying hybrid litters (MPoA-hy) and Dom females carrying conspecific litters (MPoA-dom) (electronic supplementary material, figures S6–S8 and dataset S2). Several genes were DE between MPoA-hy and MPoA-dom: Dopamine receptor 3 (Drd3) (LFC = −0.47, padj < 0.001), Formin 1 (Fmn1) (LFC = −0.74, padj < 0.001) and Calneuron 1 (Caln1) (LFC = −0.36, padj = 0.02) had lower expression in MPoA-hy. Notably, Drd3 had Spret-like expression in MPoA-hy (figure 2a). Interestingly, Cathepsin-R (Ctsr) (LFC = 0.21, padj < 0.001), a normally placenta-specific PSF gene was expressed at low levels in MPoA-hy but was not expressed in MPoA-dom (figure 2; electronic supplementary material, dataset S2 and table S3).

Figure 2.

DE gene analysis between Mus m. domesticus MPoA during normal pregnancy (MPoA-dom), Mus m. domesticus MPoA during hybrid pregnancy (MPoA-hy) and M. spretus MPoA during normal pregnancy (MPoA-spret). (a) Overlap of DE genes between the three comparisons. (b) Volcano plot of DE results of MPoA-hy versus MPoA-dom. Significantly DE genes with FDR ≤ 0.05 and log2 fold change ≥0.2 are depicted in red and labelled with the gene symbol, PSF genes in yellow with a black border. (Online version in colour.)

(d). Evidence for an effect of the placenta's paternal genome on maternal medial preoptic area

The MPoA is thought to be a primary target of placental lactogens [7] and the maternal hypothalamus is sensitive to altered placental expression of IG Phlda2 [8]. Therefore, differences in placenta-MPoA co-expression between hybrid and Dom pregnancies inform MPoA response to abnormal placental expression. One hundred and seventy-six genes were uniquely co-expressed between placenta and MPoA in hybrid pregnancies (figure 3a; electronic supplementary material, dataset S3). These include 45 genes that were DE between hybrid and Dom placentas, 16 that were DE compared with both parental species' placentas, and Ctsr, which was uniquely co-expressed in hybrid pregnancies (figure 3; electronic supplementary material, dataset S3). Interestingly, whereas 50 co-expressed genes were common to Dom and Spret pregnancies, 228 (greater than 4.5× more) were common to hybrid and Spret pregnancies (figure 3a; electronic supplementary material, dataset S3). There was pathway overrepresentation overlap between this Spret-like gene set and genes with transgressive misexpression in hybrid placenta (figure 1a), including prolactin and growth hormone receptor signalling and ERBB signalling (figure 3; electronic supplementary material, dataset S3). Moreover, PSFs were significantly overrepresented among Spret-like co-expressed genes (Fisher's exact test: p < 0.001, odds ratio = 12.57).

Figure 3.

Co-expression between MPoA and placenta for Mus m. domesticus during normal pregnancy (Dom pregnancy), Mus m. domesticus during hybrid pregnancy (hybrid pregnancy) and M. spretus during normal pregnancy (Spret pregnancy). (a) Overlap of co-expressed genes between the three pregnancy types. A secondary Venn diagram shows overlap with DE genes in hybrid versus Dom placentas. (b) Correlation between PSF gene expression in placenta and MPoA: red = Dom pregnancy ($R_{\mathrm{adj}}^2=0.38$, p = 0.01), blue = Spret pregnancy ($R_{\mathrm{adj}}^2=0.9$, p < 0.001), yellow = hybrid pregnancy ($R_{\mathrm{adj}}^2=0.95$, p < 0.001). Inset boxplot shows total PSF gene expression in MPoA (sum of normalized PSF counts per litter size).

Although expression levels were far lower in the MPoA (range = 10–589, mean = 100 normalized counts, LFC to placenta: range = −9.54 to −14.13, mean = −11.12) than in the placenta (range = 20–1 127 603, mean = 97 988 normalized counts), the apparent shift in Dom hybrid pregnancies towards Spret-like placenta–MPoA co-expression of PSFs was striking. To explore this relationship further, we tested for correlated expression of PSF genes between placenta and MPoA. We found very strong, positive correlations for hybrid ($R_{\mathrm{adj}}^2=0.95$, p < 0.001) and Spret pregnancies ($R_{\mathrm{adj}}^2=0.9$, p < 0.001) and a significant but, surprisingly, a weaker positive correlation for Dom pregnancies ($R_{\mathrm{adj}}^2=0.38$, p = 0.01; figure 3b). For all three pregnancy types, the $R_{\mathrm{adj}}^2$ estimates for the PSF correlations were significantly larger than expected by chance (300 random correlations: $R_{\mathrm{adj}(\mathrm{meam})}^2=0.244$; one-sample t-test, hybrid $R_{\mathrm{adj}}^2:t_{299}=-60.94,$ FDR < 0.001; Dom $R_{\mathrm{adj}}^2:t_{299}=-11.94,$ FDR < 0.001; Spret $R_{\mathrm{adj}}^2:t_{299}=-56.8,$ FDR < 0.001). Additionally, total PSF expression (sum of all PSF read counts divided by litter size) was significantly higher in MPoA-hy relative to MPoA-dom but statistically indistinguishable from MPoA-spret (one-way ANOVA: F_2,12 = 5.44, p = 0.02, Tukey HSD: MPoA-hy versus MPoA-dom: p = 0.016, MPoA-hy versus MPoA-spret: p = 0.18, MPoA-dom versus MPoA-spret: p = 0.36; figure 3b). These patterns suggest that the placenta's paternally inherited genome (in this case from M. spretus) influences gene expression in the maternal MPoA.

(e). Evidence for positive selection on three placenta-specific gene family genes

PSF genes exhibit accelerated evolutionary rates, potentially driven by maternal–fetal conflict [45]. We therefore screened the top 10 co-expressed PSF genes with the highest expression in MPoA, together with Ctsr, for pairwise dN/dS > 1 between Dom and Spret (electronic supplementary material, table S4). Three genes, Prl8a6, Tpbpb and Ctsr, met this criterion (electronic supplementary material, figure S9 and table S5) and were tested for positive selection including one-to-one orthologues from four related Mus subspecies and species.

Codon site-based models detected evidence for positive selection on all three genes. We found significant positive selection on codon sites across the whole tree for Ctsr (LRT_(M1a–M2a) = 9.4, FDR = 0.03) and Tpbpb (LRT_(M1a–M2a) = 10.6, FDR = 0.02), but no selection specifically restricted to the Dom or Spret lineages (electronic supplementary material, table S5). Prl8a6 did not show positive selection on codon sites across the whole tree (LRT_(M1a−M2a) = 3.14, FDR = 0.33) but we detected positive selection on one codon site within the Dom lineage (LRT_(bsA1−bsA) = 5.55, FDR = 0.05; electronic supplementary material, table S5).

4. Discussion

Molecular communication between the placenta and the maternal brain is crucial for the expression of maternal behaviour in rodents [5,8,10]. In humans, altered regulation of placental IGs is associated with prenatal depression [46], a predictor of lower growth rate and higher disease risk in infants [47]. In this study, we used a hybrid mouse model to characterize the extent to which placental disruption influences gene expression in the maternal brain. Several maternally expressed IGs were transgressively misexpressed in the hybrid placenta. In Mus m. domesticus females carrying hybrid litters, we found altered PSF expression in the placenta, and in the maternal MPoA. Surprisingly, the expression of these genes was highly correlated between the two tissues, and was M. spretus-like in the MPoA. This suggests that paternally inherited alleles in the placenta exert substantial influence on expression in the maternal brain. Collectively, our results reveal the reciprocal effects of mothers on offspring and offspring on mothers, mediated in both cases by the placenta.

(a). Maternal effects on placental expression

Global patterns of expression in the placenta were strongly associated with maternal genotype. In hybrids, half as many genes were DE relative to normal Dom placentas as opposed to Spret placentas. Notably, the more than 1600 genes with Dom-like expression in the hybrid placenta were highly enriched for terms associated with immunity and regulation of blood flow, both of which are essential to placental mediation between mother and embryo [3]. Because maternal vasculature is incorporated into the placenta, whole placenta transcriptomes necessarily include some transcripts of maternal origin and detection of Tnfsrs23 transcripts in all samples indicates that some cells from the maternal decidua remained following dissection. However, maternal blood flow within the placenta is under the direct control of placental cell lineages; trophoblast giant cells invade and replace maternal vascular endothelium, limiting maternally derived tissue to blood [48]. Likewise, Tnfsrs23-expressing cells are adjacent to fetal trophoblast giant cells at the junction between fetal and maternal tissues [44]. Therefore, while contamination from maternal transcripts may contribute to Dom-like expression in hybrid placenta, it is unlikely to bias the expression of such a large number of genes. The regulatory effects of maternal hormones, and of maternally inherited genes in the placenta, are non-mutually exclusive alternative explanations. For example, because paternal X chromosome inactivation is maintained in mouse placenta [49], maternally inherited X-linked genes are strong candidates for modulating autosomal expression in both sexes. While disentangling maternal effects (sensu [50]) from the effects of maternally inherited genes is a challenge for future studies, we note that the match between maternal genotype and placental expression of genes that modulate maternal immune tolerance and angiogenesis is consistent with the expectation of molecular coadaptation between mother and offspring [22,51], and the well-established effect of maternal environment on placental function [52].

(b). Altered placenta-specific gene family and imprinted gene expression in the placenta

Maternal adaptation to pregnancy relies to a great extent on placental signalling. Thus, altered expression of genes encoding or influencing placental signalling molecules can ultimately affect maternal physiological and behavioural response to pregnancy [5,8]. Misexpression in the hybrid placenta was substantial. However, the most striking pattern we found was the reduced expression of a large number of PSFs. In mice, most of these genes are expressed from the placental endocrine compartment and many are found in maternal plasma during pregnancy [15]. In this hybrid mouse model, the endocrine compartment is markedly reduced when the mother is Dom [28]. Thus, reduced abundance of PSF producing cell types probably contributes to an overall reduction in PSF expression.

IGs are thought to modulate PSF expression, primarily through effects on placental endocrine cell abundance, with maternally expressed genes (MEGs) repressing and paternally expressed genes (PEGs) promoting cell proliferation [19]. Two such MEGs, Phlda2 and Ascl2, were transgressively misexpressed in hybrid placentas. Altered expression of either of these genes in laboratory mouse models results in an undersized endocrine compartment, altered glycogen energy stores and reduced PSF gene expression [53,54]. Indeed, Phlda2 and Ascl2 seem to be critical co-regulators of placental endocrine compartment development [19]. Collectively, our results are consistent with the proposed role of IGs in placental signalling [13,19,55], and identify MEG misexpression as a candidate mechanism for the undersized endocrine compartment and a consequent global reduction in PSF expression in the hybrid placenta.

(c). The effects of hybrid placental dysfunction on the maternal brain

Altered signalling in hybrid placentas has the potential to affect the maternal brain. We found subtle but significant differences in the expression of four genes in the MPoA of Dom females exposed to hybrid relative to conspecific placentas. Both Fmn1 (Formin1) and Caln1 (Calneuron1) were underexpressed. In the brain, Fmn1 is involved in the formation of adherens junctions and in linear actin cable polymerization [56]. The formation of adherens junctions is important in the maintenance of the blood brain barrier (BBB) [57]. During pregnancy, the permeability of the BBB is increased by placenta-derived factors to which the maternal brain must respond in order to maintain this barrier [58]. Reduced expression of Fmn1 therefore suggests alterations in BBB adaptation during hybrid pregnancies. Caln1 encodes a neuron-specific protein with sequence similarities to calcium-binding calmodulins. While altered expression of a calcium-binding protein could indicate alterations in neuronal activity in the MPoA exposed to hybrid placentas, the functional effects such a small reduction in the expression are uncertain.

Drd3 (Dopamine receptor D3) was also underexpressed compared to Dom mothers, but not to Spret mothers, in the hybrid pregnancy MPoA. DRD3, a D2-like receptor with a generally inhibitory role, is implicated in treatment-resistant major depression [59] and Drd3 knock-out mice exhibit a suite of anxiety- and depressive-like behaviours with similar but milder effects in heterozygous knock-outs [59,60]. Given that the action of dopamine in the MPoA is critical for the expression of maternal behaviour in rats [61], and hypothalamic dopamine is altered in a mouse model for postpartum depression [62], reduced Drd3 expression in the MPoA might cause deficits in maternal behaviour. Notably, we found that Dom mothers of hybrid relative to conspecific litters are slower to retrieve their newborn pups to the nest and spend less time in the nest on the first night after parturition [33]. Whether reduced dopamine–Drd3 binding in the prepartum MPoA contributes to this reduction in maternal responsiveness is an intriguing question for future study.

Ctsr (Cathepsin R), a placenta-specific cathepsin, was the only gene that was upregulated in the MPoA exposed to hybrid placentas. The difference in expression, although significant, was so small that the biological relevance is questionable. However, unlike other PSF genes, expression of Ctsr in the maternal brain was unique to the hybrid pregnancy. Interestingly, loss of the IG Peg3 leads to de-repression of several PSF members, including Ctsr, in the fetal and adult brain [63]. Although Peg3 was not misexpressed in the hybrid placenta, the transgressively overexpressed MEG, Phlda2, was recently shown to perturb maternal behaviour and neural gene expression when its dosage was altered in mouse placenta [8]. In particular, overexpression of placental Phlda2 reduced postpartum nurturing behaviour [8], an effect of a single gene manipulation that is strikingly similar to early postpartum deficits in maternal behaviour in Dom mothers of hybrid litters [33]. Since Phlda2 and Ascl2 jointly regulate the development of the endocrine compartment [19], it is likely that transgressive misexpression of both genes in hybrid placenta impacts the maternal brain via effects on placental hormone expression.

(d). Paternal effects on the maternal brain

The hybrid placenta expresses both maternally derived (Dom) and paternally derived (Spret) alleles. Thus, females pregnant with hybrids are exposed to gene products from a foreign paternal genome. In the MPoA of Dom females exposed to hybrid placentas, we found a substantial subset of genes, including PSF genes and Drd3, with expression patterns that differed from Dom mothers with conspecific litters, but closely matched those of Spret mothers. A surprisingly large number of genes were co-expressed between placenta and MPoA in hybrid and Spret pregnancies but not in Dom pregnancies. In particular, placental and MPoA PSF gene expression were highly correlated in hybrid pregnancies and in Spret pregnancies, while Dom pregnancies showed a weaker correlation. Likewise, total MPoA PSF gene expression was Spret-like in hybrid pregnancies.

Together, these results suggest that PSF expression levels in the maternal MPoA are driven by placental expression levels of the same genes and that PSF, and potentially Drd3, expression in the MPoA is influenced by paternally inherited alleles in the placenta. Because PEGs are, by definition, expressed from the paternal allele, PEGs with Spret-like expression in the hybrid placenta are candidates for these paternal effects on maternal MPoA expression. For example, while Igf2 is best known for promoting placental cell proliferation and invasion, it also influences placental endocrine function with downstream effects on maternal physiology [64].

Pregnancy requires substantial investment from the mother, which is offset by costs to her capacity to invest in future offspring [65]. However, when offspring are sired by multiple males, selection favours fathers who extract maximal maternal resources for their own offspring [65]. Haig and colleagues proposed that these asymmetries in the reproductive interests of males and females, and the coefficients of relatedness between mothers and offspring (always 0.5) versus fathers and offspring (0.5 or 0), should promote parental antagonism, played out at the molecular level between maternally and paternally expressed IGs in the placenta [20,66]. Because placental endocrine signals promote maternal investment in current offspring, placental hormones are also proposed players in both parental and mother–offspring conflicts [55,67]. Consistent with a history of antagonistic coevolution, PSFs, in general, are the fastest evolving genes in the rodent placenta [45]. We report a similar signature of selection on three PSF genes that are co-expressed in the hybrid placenta and the maternal MPoA.

Trivers [67] described placental hormones as the molecular equivalent of begging calls. Here, we show for the first time that the expression of Prls and other PSFs is highly correlated between the placenta and maternal brain. While the function of PSFs in the brain is undefined, placental genotype-dependent differences between Dom females in the strength of the correlation and the number of co-expressed genes indicate that the relationship is driven by the placenta not the mother. Moreover, the Spret-like co-expression patterns of PSF genes in mothers of litters sired by Spret males implicate the paternally inherited genome as the driver of these placental begging calls, which are echoed in the maternal brain.

5. Conclusion

Evolutionary theoreticians have modelled mammalian pregnancy as both intimate cooperation and antagonistic struggle between two genetically distinct organisms [21,65,68]. Whether driven by conflict or coadaptation, it is clear that the placenta is the mediator of these complex interactions between mother and offspring. Here, we concentrated on placental effects on the maternal brain during the final stages of pregnancy, when it is believed to be a critical source of signal molecules that prime female physiology and behaviour for motherhood [7,10]. We found both hybrid placental misexpression with the potential to disrupt maternal–fetal communication and altered expression in the brains of mothers exposed to hybrid placentas. Global expression in the hybrid placenta seems to be dominated by the maternally derived genome and/or driven by maternal effects. However, maternal–placental communication genes co-expressed in maternal brain and placenta show elevated evolutionary rates, consistent with antagonistic coevolutionary processes. The expression of a proportion of transcripts of these genes from a foreign paternal genome in the placenta has the potential to affect the maternal brain and explain postpartum effects on maternal behaviour [33]. In addition to the effects of placental disruption on the maternal brain, expression differences between the parental species in this hybrid system reveal an unanticipated influence of the placenta's paternal genome on the maternal brain. These paternal effects on the maternal brain could play a major role in the expression of maternal behaviour and the quality of maternal care, and open novel avenues of research in both evolutionary and biomedical fields.

Ethics

All animal procedures were approved by the Oklahoma State University IACUC under protocol 141-AS.

Data accessibility

RNA-seq data from this study have been submitted to the NCBI Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under accession no. GSE126469.

Authors' contributions

P.C. conceived of the study. P.C. and L.A. designed the study. L.A. generated and analysed the data, and wrote the manuscript with input from P.C.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by the National Science Foundation (grant no. NSF-IOS714 1558109 to P.C).