Biological model of maternal auts2a contribution
To examine the role and importance of maternal auts2a, we used the medaka (Oryzias latipes), a model species in which a single ortholog of human AUTS2 (named auts2a in fish) is present10. Because the long Auts2 isoform (i.e., encoding the N-terminal region of the protein) is maternally-inherited16, involved in neuronal differentiation6, required for participation in PRC1.5-AUTS2-regulated gene expression4,5 and known to be expressed in nuclei and cytoplasm 6 , we decided to target the N-terminal part of the medaka Auts2a protein. Using CRISPR/Cas9-based genome editing we generated two different mutant lines (auts2a 2D and auts2a D338k) (Fig. 1a-b). Because a full mortality was observed in homozygous -/- auts2a 2D mutants during gastrulation, the auts2a D338k line was used in the present study. The auts2a D338k line harbors a deletion of 338,136 bp spanning exons 3-7 and is missing large intronic regions (introns 2-7) (Fig. 1a).
To specifically determine maternal auts2a contribution and analytically separate the impact of maternal and paternal auts2a parental expression, we crossed auts2a -/- females with auts2a +/+ males to generate heterozygous fish lacking maternal auts2a contribution (MCH- group). We also performed the opposite crossing to generate heterozygous fish with maternal auts2a contribution (MCH+ group) (Fig. 1c). Comparing these heterozygous individuals lacking or not maternal auts2a contribution allowed us to specifically assess the impact of auts2a maternal contribution in fish exhibiting the same genotype. This heterozygous state is highly relevant as it also corresponds to documented AUTS2 mutations in human patients3,7,15,17 . In addition, we also used wild-type (WT group) and -/- (mutant group) individuals in which both maternal and zygotic gene contribution were either absent (mutant group) or present (WT group).
Intergeneration control of macroscopic traits
The negative effect of Auts2 mutation/deletion on survival19, body size and head3,5,20 development has been reported in various vertebrate models as well as in humans but the specific contribution of maternal gene expression to these phenotypes has never been quantified. To characterize the contribution of maternal auts2a to developmental success and macroscopic traits, we monitored embryo survival throughout development, as well as head and body size in adult progeny. No difference in embryo survival could be observed until the 34-somite stage when WT individual exhibited a significantly higher survival (chi square, P=5.7 e-20) than the three other groups (MCH+, MCH-, mutant) (Fig. 1d). The same difference was also observed at the end of somitogenesis (chi square, P=1.2 e-24). At hatching, while the difference between MCH+ and WT groups remained significant (chi square, P=1.6 e-14), we observed further differences with a significantly lower survival in MCH- in comparison to MCH+ (chi square, P=9.4 e-9) and WT groups (chi square, P=5.7 e-42). Similarly, we observed a significantly lower survival in mutants in comparison to MCH+ (chi square, P=1.0 e-14) and WT groups (chi square, P=5.7 e-42). In contrast, no difference was observed between mutant and MCH- groups (chi square, P=0.17). In adult individuals, we observed significant differences in head (Tukeys’s post hoc test, P< 1.0E-7) and body (Tukeys’s post hoc test, P=1.4 e-6) size between WT and mutant groups with a reduced head and body size in the mutant group compared to WT (Fig. 1e). A similar profile was observed between MCH+ and MCH- that exhibited the same genotype with a significantly reduced head (Tukeys’s post hoc test, P=2.8 e-5) and body (Tukeys’s post hoc test, P=0.047) size in the MCH- group compared to MCH+. In contrast, no differences in head or body size were observed between WT and MCH+ groups or between MCH- and mutant groups, indicating that the maternal auts2a contribution, alone, is sufficient to drive these macroscopic phenotypes.
Intergeneration control of progeny behavior
To determine the intergenerational impact of maternal auts2a contribution to long-term (i.e., adult) progeny behavior we used a novel tank test and a recognition test to estimate anxiety-like behavior and environment recognition abilities, respectively (Fig. 2a). To minimize possible tank effects, the crossing scheme (Fig. 2b) was slightly modified (see methods for details) in order to raise half of the WT individuals with fish of the MCH+ group and the other half with fish of the MCH- and mutant groups. Fish were genotyped only after behavioral tests.
In the novel tank test, the time spent in the lower zone of the tank was used as a parameter to estimate anxiety-like behavior21. We observed a significantly reduced anxiety-like behavior in the mutant group in comparison to both WT (Tukeys’s test, P=1.4E-3) and MCH+ (Tukeys’s test, P=0.014) groups (Fig. 2c). Similarly, we observed a significantly reduced anxiety-like behavior in the MCH- group in comparison to both WT (Tukeys’s test, P=3.5E-3) and MCH+ (Tukeys’s test, P=0.035) groups (Fig. 2c). In contrast, no difference in anxiety-like behavior was observed between WT and MCH+ groups or between MCH- and mutant groups, indicating that the lack of maternal auts2a contribution, alone, is sufficient to drive reduced anxiety-like behavior in the long-term.
To determine their ability to recognize their environment, fish were placed back in the test tank on day 2 (Fig. 2a) and their velocity was used as an estimator of their exploratory behavior as velocity typically decreases in a known environment21. After 10 minutes in the test tank, fish of the WT and MCH+ groups, but not of the MCH- and mutant groups, exhibited a reduced exploratory behavior in comparison to day 1 (Fig. 2d). After 60 minutes in the tank test, all experimental group exhibited a significantly reduced exploratory behavior in comparison to day 1 (Fig. 2d). Together, these results indicate that groups with maternal auts2a contribution exhibit a higher ability to recognize their environment, regardless of their genotype. This shows that maternal auts2a contribution, alone, is sufficient to drive enhanced environmental recognition in the progeny.
Maternal auts2a contribution-dependent developmental gene expression sequence
To determine the impact of maternal auts2a contribution on the gene expression sequence during neurodevelopment, embryo transcriptome was studied a late neurula stage (stage 1818) and during late embryonic brain formation (stage 2918) (Supplementary tables 1-4). Differences in gene expression were assessed between fish with or without maternal auts2a contribution. In order to separate the impact of maternal contribution and fish genotype (i.e., the number of auts2a alleles in the genome) comparisons were made among heterozygous individuals (MCH analysis, comparison of MCH+ and MCH- groups), or regardless of the genotype (MCA analysis, comparison of MCA+ and MCA- groups). For the later, the maternal contribution group (MCA+) included both WT and MCH+ individuals, while the no maternal contribution group (MCA-) included both MCH- and mutant individuals (Fig. 3a). We first focused on heterozygous individuals (MCH) in which only the maternal auts2a contribution, but not the genotype, varied. This model corresponds to pathological cases in which only one copy of the human AUTS2 gene is usually mutated3,7. We observed major differences in gene expression between MCH+ and MCH- groups at both stages with 487 and 277 differentially expressed gene (DEG) at stages 18 and 29, respectively (Fig. 3b). The higher number of DEG at stage 18, in comparison to stage 29, indicates that the lack of maternal auts2a contribution drives major differences in gene expression early during development (i.e., at late neurula stage). At this stage, we observed a significantly higher proportion (66 %) of up-regulated genes in the MCH- group in comparison to MCH+ (Fig. 3c). In contrast, we did not observe any difference in the proportion of up- and down-regulated genes at stage 29. This suggests that, during neurulation, maternal auts2a contribution regulates gene expression mostly through gene repression rather than gene activation.
When analyzing maternal auts2a contribution regardless of the genotype (MCA analysis) we identified a similar number of DEG at both stages (n=549 and 568 at stage 18 and 29, respectively) (Fig. 3b). We also observed a similar proportion of up-regulated and down-regulated genes at both stages (Fig. 3c). These results obtained in MCA individuals in which the impact of both maternal contribution and genotype are assessed are in sharp contrast with the results obtained above in MCH individuals in which only maternal auts2a contribution is assessed. These observations are fully consistent with the high number of DEG (Fig. 3b, yellow sectors) that are found when comparing MCA+ and MCA- groups, but not when comparing MCH+ and MCH- groups (Fig. 3b, blue sectors). These genotype-dependent genes, that are not regulated in MCH fish are thus likely to be regulated by zygotic, rather than maternal, auts2a.
Together, these results show that maternal auts2a contribution, alone, drives major differences in gene expression especially during early development (i.e., at late neurula stage) and mostly through gene repression rather than activation.
Dysregulated genes associated with neuropathologies
To reveal any existing link between auts2a maternal contribution and molecular mechanisms or genes linked to human neuropathologies, we used the DisGeNET database that covers more than 24 000 diseases and traits, 17 000 genes and 117 000 genomic variants22. At stage 18, a highly significant proportion (47%) of the 319 human orthologs of DEG originating from the MCH analysis were associated with neuropathologies, in comparison to 32% for a random selection of 319 genes in the genome (Fig. 4a, Supplementary table 5). This strong association with neuropathologies was observed for both up-regulated and down-regulated genes at stage 18 (Fig. 4a). In contrast, no significant association between DEG and neuropathologies was observed at stage 29 using the 186 human orthologs of DEG at this stage. Among the genes identified at both stages, we identified genes associated with various pathologies including autism spectrum disorder (ASD), behavioral abnormalities, epilepsy, hypotonia, intellectual disability, microcephaly and schizophrenia (Fig. 4a). At stage 18, maternal auts2a contribution, alone, was able to trigger differences in the expression of a significant number of genes associated with the 7 above-listed classes of neuropathologies (Fig. 4a).
For DEG identified in the MCA analysis, we observed a significantly higher proportion of genes associated with neuropathologies, in comparison to a random selection of genes in the genome, for both stages 18 and 29 (Fig. 4b). Despite the higher number of DEG identified in the MCA analysis, the overrepresentation of neuropathology-associated genes at stage 18 was however less significant than the overrepresentation found in the MCH analysis, and was not significant for down-regulated gene in the MCA analysis. This suggests that the dysregulation of genes associated with human neuropathologies during neurulation is for the most part driven by maternal auts2a contribution even though we cannot rule out a possible contribution of zygotic auts2a. During late embryonic brain formation (i.e., stage 29), the dysregulation of genes associated with neuropathologies appears to be mostly zygotically regulated even though we cannot rule out a milder contribution of maternal auts2a that would be statistically difficult to reveal due to the lower number of DEG in the MCH analysis at this stage.
Dysregulated molecular signaling pathways during neurodevelopment
To further characterize dysregulated molecular processes in the absence of maternal auts2a contribution, we specifically investigated the seven conserved metazoan signaling pathways23 and the components that comprise the basement membrane and adhesome complex (i.e., extracellular matrix). These pathways are known to orchestrate complex cell and tissue interactions regulating development, including neurodevlopment24–27. When comparing heterozygous individuals lacking (MCH-) or not (MCH+) maternal auts2a contribution, we observed a dysregulation of genes belonging to all pathways, with the exception of the Hedgehog pathway (Supplementary table 6). At stage 18, the number of genes dysregulated was especially high for the RTK/growth factors (iqgap3, hbegfa, rab8b, rab25a, rab25b, rab42, rrad, rraga, rras, rras2), TGF-β (amh, bmp2, bmper, brinp3b, nomo, smad9) and canonical Wnt (mfrp, nkd2a, ved, vwde) pathways, as well as for the extracellular matrix (itga6b, hmcn2, lamc1, lamb2, lamb2l, nid1a, nid2a) (Fig. 5a, blue bars). Similar results, even though less pronounced, were observed when analyzing maternal auts2a contribution regardless of the genotype (MCA analysis) (Fig5. a, yellow bars). At stage 29, the comparison of MCH+ and MCH- groups led to the identification of dysregulated genes in all pathways, with the exception of the Hedgehog pathway (Fig. 5a, blue bars). The number of dysregulated genes was however much lower (1-3 genes) with the exception of the Notch-Delta pathway in which 5 genes (her8.2, her12, hey1, nle, tuba8l2) were dysregulated. In contrast to stage 18, we observed a more pronounced dysregulation of most pathways when the comparison was made regardless of the genotype (MCA analysis) (Fig5. a, yellow bars) indicating that these pathways are more predominantly regulated by zygotic auts2a. This was however much less pronounced for the Notch-Delta pathway indicating that this pathway remains mostly regulated by maternal auts2a contribution at stage 29.
Together, our results show that maternal auts2a contribution acts through the regulation of most evolutionarily conserved metazoan pathways, especially RTK/growth factors, TGF-β, canonical Wnt, as well as extracellular matrix components at late neurula stage. We also observed that maternal auts2a contribution modulates Notch-Delta pathway gene expression during late embryonic brain formation.
Dysregulated transcription factors during neurodevelopment
Because maternal auts2a contribution was able to influence the expression of several hundreds of genes during neurodevelopment, we specifically investigated transcription factors and monitored the orthologs of all inventoried human transcription factors (TFs)28. When comparing heterozygous individuals lacking (MCH-) or not (MCH+) maternal auts2a contribution, we observed a dysregulation of a large number of TFs at both stage 18 (n=24) and stage 29 (n=12) (Supplementary table 7). At stage 18, we observed the up-regulation of 7 transcriptional activators (msantd1, spi1b, ebf2, mef2aa, sox10, bnc1, meox2a) and the down regulation of 2 transcriptional repressors (hbp1, znf703) in MCH- individual, consistent with the hypothesis that maternal auts2a contribution acts mostly through gene repression at late neurula stage. It should be noted that among the 24 dysregulated TFs, 12 are classified as regulators and cannot be specifically associated with either repression or activation. Among the different categories of TFs dysregulated at stage 18 in absence of maternal auts2a contribution, some were of specific interest due to their known function in human brain development. This was especially the case for the homeodomain and HMG/Sox categories (displayed in yellow brown and pink, respectively in Fig. 5b). DLX2 and SIX3, that belong to the Homeodomain TF category, are known activators of transcription. DLX2 is involved in interneuron differentiation and forebrain development, and plays a role in craniofacial patterning and morphogenesis29. DLX2 is also a negative regulator of the Notch-Delta pathway that is dysregulated in absence of maternal auts2a contribution30. Six3 participates in anterior brain formation through the modulation of Wnt1 and Shh31. In the HMG/Sox category, SOX10 plays a major role in neural crest and peripheral nervous system development32,33 while mutations in the CIC genes have been associated with brain tumor, epilepsy and mental retardation34,35. Finally, our analysis reveals srcap, a AT hook TF acting upstream of the Notch pathway. Mutations in human SRCAP cause Floating-Harbor, a rare disorder characterized by short stature, language deficits, and dysmorphic facial features, which are also observed in AUTS2 syndrome3,36.
When incorporating WT and mutant groups in the analysis (MCA analysis), the number of dysregulated TFs was less important suggesting that their regulation by auts2a at stage 18 is mostly under maternal auts2a rather than zygotic control. In addition, the absence of dysregulated TFs in the HMG/Sox and AT hook categories in the MCA analysis suggests that TFs belonging to these categories are mostly, if not exclusively, under maternal auts2a control. For the Homeodomain category, the only dysregulated TF (six4) in the MCA analysis is not dysregulated in the MCH analysis, indicating that dysregulated Homeodomain TFs (dlx2, six3) are also exclusively regulated by maternal auts2a.
These observations at stage 18 are in sharp contrast with our observations at stage 29 for which a much higher number of dysregulated TFs was observed in the MCA analysis (n=43) in comparison to the MCH analysis (n=12). Because wild-type and mutant groups are included in the MCA analysis, this observation suggests that, in contrast to stage 18, zygotic auts2a acts through a large number of specific TFs. Interestingly, three TFs belonging to the bHLH category (sim2, atoh7, hey1) are found in both MCA and MCH analyses suggesting that they are mostly, if not exclusively, under maternal auts2a control. Sim2 is a regulator of transcription known to be a master gene of central nervous system (CNS) development37. Interestingly, SIM2 is ubiquitinated by RING ubiquitin ligase, a component of the PRC1.5-AUTS2 complex38.
Our analysis reveals that the lack of maternal auts2a contribution is sufficient to trigger the dysregulation of several key TF families, including Homeodomain, HMG/Sox, AT Hook, and bHLH during development in a stage dependent-manner. More specifically, dysregulated genes belonging to these TFs categories are major player of nervous system development and, for some of them, associated with neuropathologies.
Intergenerational transmission mechanisms
To gain insight into the mechanisms of intergenerational determination, we investigated the molecular cargo (i.e., maternally inherited mRNAs, or maternal mRNAs) in eggs originating from females lacking (mutant) or not (wild-type) maternal auts2a contribution. Because of the role of AUTS2 in regulating gene expression, we reasoned that auts2a could regulate gene expression in the oocyte, leading to the accumulation of specific maternal factors that could, after being inherited by the zygote, act to directly or indirectly regulate gene expression during neurodevelopment. We therefore performed RNA-seq analysis in 1-cell stage embryos originating from mothers exhibiting (WT group) or not (mutant group) auts2a contribution (Supplementary table 8). We decided to focus on maternal mRNAs exhibiting a depletion in mutant eggs under the hypothesis that these factors would be more likely to intergenerationally regulate neurodevelopment under maternal auts2a control. We identified 30 transcripts exhibiting a significant depletion in the absence of maternal auts2a contribution (Fig. 6a, Supplementary table 8). Collectively, these transcripts have known or predicted functions related to cellular communication, cell migration, cell proliferation, DNA and RNA binding, and differentiation, intracellular trafficking that are consistent with a regulatory role during embryonic development. Among these depleted maternal mRNAs, msi2a and hdac1 were of particular interest. The musashi RNA binding protein 2 (Msi2) is a translational regulator that targets genes involved in development and cell cycle regulation. Msi2 is believed to play a role in the proliferation and maintenance of stem cells, including neuronal progenitor, in the central nervous system39,40. Similarly, we observed a significant depletion in hdac1 mRNA in 1-cell stage embryos originating from mothers lacking auts2a contribution. Hdac1 is an epigenetic regulator known as a transcriptional repressor and acting as a component of the histone deacetylase complex. In all investigated animal species, Hdac1 is maternally inherited. In mice, Hdac1 regulates cell cycle and zygotic genome activation41,42 and is critical for preimplantation development43. In zebrafish, hdac1 plays major role in central and peripheral nervous system development44. Our observations suggest that maternal auts2a acts through maternally inherited factors accumulated in the oocyte that regulate, in turn, early embryonic development and neurodevelopment. We hypothesize that this intergenerational effect is mediated, at least in part by maternally inherited mRNAs encoding transcriptional regulators such as hdac1 and msia2.
Evolutionary conserved genes linked to behavior and nervous system development
In the present study we provide a unique vertebrate example demonstrating that maternal, but not paternal, invalidation of a gene linked to numerous human neuropathologies (i.e., 149 according to the DisGeNET database) can trigger major neurodevelopmental and long-term behavioral differences in the progeny. We therefore aimed at identifying other genes tightly associated with human neuropathologies that could exhibit similar features. We reasoned that relevant genes had to be expressed in the oocyte and used maternal inheritance as a criterion to search publicly available gene expression data. We used the mouse and the zebrafish, two popular model vertebrate species that are heavily used for biomedical purposes. We identified 11,521 and 22,738 maternally inherited transcripts in mice and zebrafish, respectively. To be able to subsequently link candidate genes to human pathologies, zebrafish and mouse lists were filtered for common orthologs (Supplementary table 9), which led to the identification of a list of evolutionarily conserved maternal mRNAs. These conserved maternally inherited genes were subsequently filtered for gene that could be associated with either “Behavior” (GO:0007610) or “Nervous system development (NSD)” (GO:0007399) in either humans, mice (Mus musculus), rat (Rattus norvegicus), or zebrafish (Danio rerio) using the gene ontology database (http://geneontology.org/ , see methods for details). We were able to identify 107 maternally-inherited genes associated with “Behavior” and “NSD” GO terms in both mice and zebrafish (Fig. 7a). We then used the DisGeNET database to identify genes linked to human pathologies. We also used the Disease Pleiotropry Index (DPI) that ranges between 0 and 1 and reflects the diversity of associated diseases, a DPI of 1 corresponding to the highest diversity. Among the 107 identified genes, 92 had a human ortholog present in DisGeNET database (Supplementary table 10). The DPI of these genes (green box plot, Fig. 7b) was significantly higher than the DPI of genes not associated with “Behavior” or “NSD” GO terms (grey box plot, Fig. 7b). Similarly, the DPI of the genes with “Behavior” and “NSD” GO terms (green box plot, Fig. 7b) was significantly higher than the DPI of the genes associated only with “NSD” (yellow box plot, Fig. 7b). We then focused on the genes associated with a higher diversity of diseases (i.e., DPI > 0.75). Among the evolutionarily conserved maternal genes associated with “Behavior” and “NSD” GO terms, 49% (45/92) had a DPI above 0.75, including AUT2 and EP300, another member of the PRC1.5-AUTS2 complex. The proportion of genes with a DPI above 0.75 was significantly lower for genes associated only with either one of these two GO terms, and for genes not associated with either one of these categories. These results indicate that AUTS2 and EP300, its partner in the PRC1.5-AUTS2 complex that transcriptionally regulates CNS genes, belong to a group of 45 evolutionarily conserved maternally-inherited genes that are associated with “Behavior” and “NSD” and linked to a wide diversity of human pathologies.