Post-transcriptional regulation by the exosome complex is required for cell survival and forebrain development via repression of P53 signaling

The neocortex of the mammalian brain is radially structured into six neuronal layers and multiple functional domains that form the structural basis for sensorimotor processing and intellectual ability. In early cortical development, apical progenitors (APs) in the ventricular zone (VZ) function as neural stem cells (NSCs) and produce neurons via direct and indirect neurogenesis in a specific temporal order, which will make up the different cortical layers (Götz and Huttner, 2005; Kriegstein et al., 2006). In direct neurogenesis, APs divide asymmetrically to generate new APs and neurons. Neurons produced during early neurogenesis are distributed mainly in the lower cortical layers (LLs) L6 and L5. In indirect neurogenesis, APs divide to self-renew and produce basal progenitors (BPs) that undergo limited cycles of symmetric divisions to generate more neurons with upper layers (ULs) L4-L2 (Pontious et al., 2008).

Optimal regulation of gene expression is crucial for establishing the intricate balance between the rate of proliferation and differentiation of neural progenitor cells as well as cell viability and apoptosis. Transcriptional regulation plays a central role in controlling gene expression. However, regulation of gene expression is not limited to the transcriptional level. Post-transcriptional mechanisms, such as the regulation of RNA stability contribute in sharpening the expression of genes during development.

The evolutionarily conserved RNA exosome is an essential factor that modulates gene expression during development (Januszyk and Lima, 2014; Kilchert et al., 2016). The ring-like structured exosome complex contains 11 different exosome component (Exosc) subunits, including nine structural subunits (Exosc1-9) and two catalytic subunits (Exosc10, Dis3) (Januszyk and Lima, 2014; Kilchert et al., 2016). Together with structural counterparts, Exosc10 and Dis3 are able to degrade numerous RNAs using their ribonuclease activity. This makes the entire exosome complex indispensable for controlling the richness of RNAs, degrading malfunctional or mis-configured RNAs. The integrated Exosc subunits, which interact through composite surfaces with their co-factors, are essential for targeting the exosome to specific RNAs for degradation, therefore conferring functional specificity (Lubas et al., 2011; Januszyk and Lima, 2014; Kilchert et al., 2016; Lim et al., 2017; Puno and Lima, 2018; Schmid and Jensen, 2019).

Of note, mutations of the Exosc genes have been found in various human brain disorders, including corpus callosum hypoplasia, cerebellar atrophy, abnormal myelination, pontocerebellar hypoplasia with cerebellar and spinal motor neuron degeneration, and intellectual disability (Wan et al., 2012; Boczonadi et al., 2014; Di Donato et al., 2016; Burns et al., 2018; Morton et al., 2018; Fasken et al., 2020), suggesting important roles for the exosome complex in neural development.

To investigate the possible involvement of the exosome complex in brain development, we used a conditional knockout (cKO) of the RNA exonuclease subunit Exosc10 from early brain development in transgenic mice. Transcriptional profiling of the Exosc10cKO cortical tissue revealed that Exosc10 suppresses the expression of large sets of genes involved in various processes of brain development, including cell death-related pathways. RIP-seq and RNA degradation analyses uncovered that Exosc10 directly binds to and induces degradation of P53 signaling-related transcripts. Phenotypically, we found that elimination of Exosc10 leads to a massively enhanced apoptosis, reduced neurogenesis and dysgenesis of cortical layers, with the first effect being rescued by inhibition of P53 signaling. Overall, this study provides new insights into the post-transcriptional regulatory mechanism mediated by the RNA exosome complex, which acts upstream of P53 signaling in apoptosis suppression in brain development.

Mutations in human Exosc2, Exosc3, Exosc8 and Exosc9 genes, and their associated brain disorders imply important roles for the exosome complex in neurodevelopment (Wan et al., 2012; Boczonadi et al., 2014; Di Donato et al., 2016). In further support of the involvement of the exosome complex in mammalian brain development, exosome genes are prominently expressed in the developing mouse cortex, especially in the ventricular zone (VZ) (Fig. 1A; Fig. S1A-J). Extracting the published scRNA-seq dataset of the mouse developing cortex (Telley et al., 2016) also confirmed the highest expression of exosome genes in apical progenitors (APs), which are found in the VZ (Fig. 1B; Fig. S1K-S). As Exosc10 is the exonuclease subunit of the RNA exosome complex (Fig. S1A), we aimed to study the role of Exosc10 in neural development by generating and characterizing the cortical phenotype of Exosc10 conditional knockout (cKO) mice.

To investigate the consequences of loss of Exosc10 in brain development, we bred mice bearing conditional inversion (COIN) alleles of Exosc10 (Economides et al., 2013; Pefanis et al., 2015) with different Cre lines, including telencephalon-specific FoxG1-Cre (Hébert and McConnell, 2000) and cortex-specific Emx1-Cre (Gorski et al., 2002), to generate the corresponding cKO mutants: cKO_FoxG1-Cre and cKO_Emx1-Cre. Telencephalon-specific cKO_FoxG1-Cre embryos exhibited an absolute absence of the telencephalon at E17.5 (Fig. S2A, arrow). To examine formation of the telencephalon at early stages, immunohistochemistry analysis was performed with antibodies against Sox2, Pax6, HuCD and NeuN in forebrain tissue of control and cKO_FoxG1-Cre embryos between E10.5 and E12.5 (Fig. S2B,C). The expression of these markers was found in telencephalon (Tel), diencephalon (Di) and mesencephalon (Mes) of control embryos, whereas their expression was seen only in Di and Mes structures in cKO_FoxG1-Cre mutants. This finding suggests that the deletion of Exosc10 at the onset of telencephalon formation in cKO_FoxG1-Cre embryos results in the failure of telencephalon formation, which makes this mouse line inappropriate for further investigations. At P6, the cortex in cKO_Emx1-Cre mice was significantly smaller than that of the control mice; however, we were still able to examine cortical development under Exosc10 deficiency (Fig. 1C,D).

The cKO mutants at P6 had visually smaller cortical size and thinner cortical layers than that of controls, as revealed by Satb2 immunostaining (Fig. 2A,G). To explore cortical layer formation in detail, we performed immunohistochemistry and evaluated the expression of cortical layer (L)-specific markers, such as reelin (L1; Fig. 2B), Cux1 (L2/3; Fig. 2C), Ctip2 (L5; Fig. 2D), Sox5 (L5/6; Fig. 2E) and Tbr1 (L6; Fig. 2F).

The population of reelin⁺ L1 neurons generated in the cortical hem and ventral telencephalon, where the Emx1 promoter is not active (Gorski et al., 2002), seems to be preserved in the cortex of cKO_Emx1-Cre mutants (Fig. 2B,G). In contrast, the numbers of Cux1⁺ L2/3, Ctip2⁺ L5, Sox5⁺ L5/6 and Tbr1⁺ L6 neurons, which are generated from cortical progenitors, were significantly reduced in the cKO_Emx1-Cre cortex (Fig. 2C-G). Together, these results show that the expression of Exosc10 is required for normal forebrain development and cortical layer formation.

The diminished population of neurons in the cKO cortex at postnatal stage promoted us to investigate the consequences of the loss of Exosc10 expression on neurogenesis at early embryonic stages. Consistent with the decreased number of neurons in cortical layers at P6 (Fig. 2), the Exosc10-ablated cortex at E13.5 displayed a diminished number of HuCD⁺ and NeuN⁺ neurons in cortical plate (CP) (Fig. 3A,B,D,E). Remarkably, the thickness of the germinal zones, i.e. the ventricular zone (VZ) and sub-ventricular zone (SVZ) (Fig. 3A-C), in the control cortex was comparable with that of the mutant cortex (Fig. 3A,B). Accordingly, the number of apical progenitors (Pax6⁺ or Sox2⁺) in VZ and basal progenitors (Tbr2⁺) in SVZ was comparable in cKO and control cortices (Fig. 3A-E). These findings suggest that the deletion of Exosc10 might lead to an enhanced apoptosis or defect in neuronal differentiation in early cortical development, causing the observed brain microcephaly at the postnatal stage.

The immunohistochemical analysis with the apoptosis marker activated caspase 3 (Casp3) revealed that Exosc10 ablation engenders intense apoptosis in Exosc10cKO_Emx1-Cre cortex at E13.5, especially in the rostromedial area (Fig. 4A-C), where the Cre recombinase activity was found to be highest (Gorski et al., 2002; Narayanan et al., 2015; Nguyen et al., 2018). A dramatic increase in apoptosis was already evident in the mutant cortex at E11.5 (Fig. 4D). Apoptotic cells were observed in the entire mutant cortex, albeit more dominant in the basal side of cortical wall, suggesting that neurons were the most affected population of cells therein (Fig. 4A-C). Notably, there was no difference in the number of Casp3⁺ apoptotic cells between control and Exosc10cKO cortices at postnatal stages, indicating that Exosc10 expression is required for cell viability only at embryonic stages of cortical development (Fig. S3).

Double immunohistochemical analyses of Casp3 and markers for apical progenitors (Sox2; Fig. 4A), basal progenitors (Tbr2; Fig. 4B) and neurons (NeuN; Fig. 4C) confirmed that apoptosis was found in all three cell populations in the cKO_Emx1-Cre cortex (indicated by filled arrows in Fig. 4A-C,E). Notably, the highest cell death rate was identified in NeuN⁺ neurons (Fig. 4E). Altogether, our findings indicate that Exosc10 is crucial for cell viability in early cortical development.

To understand the molecular mechanisms underlying control of cortical development by Exosc10 and to identify the Exosc10 direct target transcripts, RNA sequencing (RNA-seq) as well as RNA immunoprecipitation sequencing (RIP-seq), were performed. In the RNA-seq, RNAs isolated from the E12.5 Exosc10cKO_Emx1-Cre and control cortices were sequenced (Fig. S4A,C). We found that loss of Exosc10 caused upregulation of 1031 genes and downregulation of 844 genes (adjusted P<0.05; Fig. 5A, Table S1). Gene ontology (GO) analysis of the upregulated genes reflected involvement of various brain development processes (Fig. 5B, Table S2). Consistent with the increased apoptosis, cell death-related pathways, such as regulation of apoptotic signaling pathway, and signal transduction by P53 mediators were upregulated in the cKO cortex (Fig. 5C, Table S2).

In our RIP-seq experiment, RNAs were purified from the E12.5 wild-type mouse cortex (Fig. S4B,D). By sequencing the Exosc10-bound RNAs, binding enrichment of Exosc10 on 3159 transcripts was identified (adjusted P<0.05; Fig. 5B, Table S3). GO analysis revealed that the Exosc10-bound transcripts participate in various processes of brain development (Fig. 5D, Table S4). To identify candidates for functional analysis, we compared the RNAs upregulated in the Exosc10cKO cortex in RNA-seq experiment with transcripts bound by Exosc10 in RIP-seq analysis. We identified 144 transcripts common to the results of both experimental analyses (Fig. 5E, Table S5). Interestingly, those intersectional 144 upregulated genes showed an enrichment in the GO term ‘neuron death’ (Fig. 5F, Table S6).

In accordance with our immunohistochemical data showing increased numbers of Casp3⁺ apoptotic cells, many genes involved in P53 apoptosis signaling (e.g. Ccng1, Sesn2, Pmaip1, Bbc3 and Aen) were upregulated in the cortex of Exosc10cKO mutants (Fig. 5A; Fig. 6B). Furthermore, the results from our RIP-seq analysis revealed that transcripts of many apoptosis-related genes (e.g. Ccne1, Ccng2, Tsc2, Bbc3, Apaf1 and Aen) were bound by Exosc10 (Fig. 5B; Fig. 6A). Thus, our findings raise the possibility that Exosc10 inhibits apoptosis by directly suppressing expression of P53 signaling effector genes.

Among the transcripts belonging to the P53 pathway, Aen (apoptosis enhancing nuclease) and Bbc3 (BCL2 binding component 3; also known as PUMA, P53-upregulated modulator of apoptosis) were bound by Exosc10 (Fig. 6A). Their upregulated expression in the Exosc10cKO cortex was first revealed by RNA-seq (Fig. 6B), then confirmed by qPCR (for Bbc3, Aen and Atr; Fig. 6C) and immunohistochemical analyses (for Aen; Fig. 6D,E), making them strong candidates for mediating regulation of apoptosis by Exosc10.

To ascertain the functional effect of Exosc10 binding, we examined whether Exosc10 deletion influences decay of these identified transcripts. For this purpose, an RNA degradation assay was performed using cultured cortical NSCs derived from E12.5 Exosc10 COIN/COIN embryos. NSCs were treated with either soluble Tat-Cre recombinase to knockout Exosc10 (Exosc10KO) or vehicle as a control group (Fig. 6F; Fig. S5). An actinomycin D-based method was used to halt de novo transcription (Yoon et al., 2017). The cells were harvested and qPCR was performed to quantify the transcript level of Aen and Bbc3 in cultured NSCs before (0 h) and after 5 h and 17 h treatment of actinomycin D (Fig. 6F). Compared with control, a higher stability of Aen transcripts in Exosc10KO NSCs was observed after 5 h and 17 h treatment of actinomycin D (Fig. 6G). The higher RNA stability of Bbc3 in mutant NSCs than that in controls was seen after 17 h of actinomycin D treatment (Fig. 6H). These results indicated that Exosc10 directly binds and degrades transcripts of Aen and Bbc3.

To consolidate our observation that Exosc10 may regulate apoptosis via suppression of the P53 apoptosis pathway, in vivo rescue experiments were performed. We used a P53 inhibitor, pifithrin-α (PFTα), that is known to inhibit P53-dependent activation of P53-targeted genes (Komarov et al., 1999). PFTα was injected daily starting from 9.5 days post coitum (d.p.c.), and the PFT-treated Exosc10cKO animals were examined at E13.5 (Fig. 7A). Whereas 52.0±3.5% of cells in the PFT-untreated (non-injected) Exosc10cKO cortex were apoptotic, the percentage of Casp3⁺ cells decreased to 31.0±9.2% upon PFTα treatment (Fig. 7B,C). Thus, the observed apoptotic phenotype in Exosc10cKO mutants was largely rescued by inhibition of the P53 pathway in the developing cortex. Among the P53 pathway-related genes, we compared the expression of Bbc3 and Aen from control and cKO cortices, which were treated with either vehicle or P53 inhibitor. Remarkably, PFTα treatment does not significantly rescue the aberrant upregulation of Aen upon the loss of Exosc10 in the developing cortex (Fig. 7D). Treatment with PFTα, however, decreases the expression of Bbc3, which is upregulated in cKO cortex (Fig. 7E). This is in line with the evidence that Bbc3 (but not Aen) is a direct target of P53, as the promoter region of Bbc3 contains P53-binding sites and can be directly activated by P53 (Han et al., 2001).

To address whether Exosc10 regulates cortical development partly via suppression of P53 signaling, pregnant mice between 9.5 and 15.5 days post coitum (d.p.c.) were intraperitoneally injected daily with PFTα solution. Owing to the perinatal lethality of PFTα-treated animals, the brain samples were collected at E18.5 for phenotype analysis (Fig. 8A). The expression of Satb2, which marks the majority of projection neurons in all cortical layers and areas (Alcamo et al., 2008; Britanova et al., 2008), was then examined in control, and in cKO with and without PFTα treatment. The treatment with P53 inhibitor did not influence the size of wild-type (control) cortex, as indicated by Satb2 expression. Remarkably, when compared with vehicle-treated cKO embryos, PFTα-treated embryos had a significantly larger cortex. Concurrently, PFTα administration in cKO mutants resulted in an increase in the number of Satb2⁺ neurons (Fig. 8B-F). The findings indicate that treatment with the P53 inhibitor partly rescues the aberrant cortical morphology in mutants.

Altogether, these results reveal that loss of Exosc10 during early cortical development causes aberrantly enhanced expression of P53 pathway-related transcripts such as Aen and Bbc3, and causes increased apoptosis similar to that observed after P53 overexpression (Fig. 8G). These findings demonstrate that the balance between the RNA exosome complex and P53 signaling activity is essential for cell survival and for normal cortical development (Fig. 8G).

Here, we have investigated the possible function of the RNA exonuclease Exosc10 in brain development. We demonstrate that expression of Exosc10 is crucial for controlling cell survival and cortical development. Our findings indicate that Exosc10 performs an essential function in controlling P53-mediated apoptosis signaling by directly degrading the P53 signaling-related transcripts such as Aen and Bbc3 (Fig. 8G).

Post-transcriptional regulations such as RNA modification and RNA stability are emerging as mechanisms that are essential for regulation of gene expression. The best known mechanism of RNA modification, so far, is the methylation of nitrogen 6 in adenosine (N6-methyladenosine, m6A). The m6A writer complex, consisting of Rbm15, Wtap, Mettl3 and Mettl14, is responsible for addition of methyl groups to RNA (Liu et al., 2014; Wang et al., 2016). On the other hand, the m6A eraser Alkbh5 can remove methyl groups installed on RNA (Zheng et al., 2013). YTH proteins and eIF3 can serve as m6A readers to recognize m6A (Meyer et al., 2015; Patil et al., 2016). Recent studies suggest important roles for m6A methylation in the modulation of RNA stability in brain development (Yoon et al., 2017; Li et al., 2018; Wang et al., 2018; Widagdo and Anggono, 2018; Flamand and Meyer, 2019).

m6A deficiency by Mettl14 cKO in the developing mouse brain increased the stability of NSC transcripts, causing lengthening of the cell cycle of NPCs and prolongation of cortical neurogenesis further at postnatal stage (Yoon et al., 2017). The m6A depletion by Mettl14 cKO also changed the levels of modified histones and of transcripts encoding histone-modifying enzymes, indicating that m6A-dependent control of epigenetic program alterations is involved in neurodevelopment (Wang et al., 2018).

Deletion of m6A reader Ythdf2 in Ythdf2 KO mice leads to late embryonic lethality. Neurogenesis was declined significantly with concurrent reduction in the number of Tbr2⁺ BPs leading to a thinner cortical plate in Ythdf2 KO embryos (Li et al., 2018). It should be noted, however, that m6A was shown to impact not only RNA stability but also other features of RNA metabolic processes, such as translation, splicing and transport of RNA (Dominissini et al., 2012; Meyer et al., 2012; Wang et al., 2014, 2015; Xiao et al., 2016). Thus, the precise roles of RNA stability in brain development are still largely unknown.

Functional investigations of genes encoding exosome subunits in model systems indicated that most exosome subunits are required for viability from yeast to man (Januszyk and Lima, 2014; Kilchert et al., 2016; Morton et al., 2018; Fasken et al., 2020). Although the precise role of the exosome complex in neural development is not known, mutations in four out of the 11 exosome subunit genes in humans have been found to be associated with neurodevelopmental and psychiatric disorders. Notably, mutations in Exosc2 are associated with intellectual disability (Di Donato et al., 2016), whereas mutations in Exosc3 (Wan et al., 2012; Rudnik-Schoneborn et al., 2013; Zanni et al., 2013; Eggens et al., 2014; Halevy et al., 2014), Exosc8 (Boczonadi et al., 2014) and Exosc9 (Burns et al., 2018) cause different types of cerebellar hypoplasia that lead to severe neurodegeneration and lethality. Mutations of these exosome factors were also associated with other brain defects, such as corpus callosum hypoplasia, cerebellar atrophy and abnormal myelination, as well as pontocerebellar hypoplasia with cerebellar and spinal motor neuron degeneration (Wan et al., 2012; Rudnik-Schoneborn et al., 2013; Zanni et al., 2013; Boczonadi et al., 2014; Eggens et al., 2014; Halevy et al., 2014; Di Donato et al., 2016; Burns et al., 2018). The findings in these human genetics studies indicate that the RNA exosome is crucial for normal neural development and cognition (Morton et al., 2018; Fasken et al., 2020).

Among Exosc subunits, the role of Exosc10 in biological processes is the most investigated. Exosc10 has been shown to stimulate mRNA turnover (Van Dijk et al., 2007), 3′ pre-rRNA processing (Knight et al., 2016), and decay of long non-coding and enhancer RNAs (eRNAs and lncRNAs) (Pefanis et al., 2015) with its absence causing RNA processing defects in yeast (Carneiro et al., 2007) and increased vulnerability to DNA damage (Rolfsmeier et al., 2011; Marin-Vicente et al., 2015; Domingo-Prim et al., 2019). Exosc10 functions with a co-factor, such as the NEXT complex, that recognizes and degrades RNA in DNA/RNA hybrid or RNA/RNA hybrid configuration, or eRNAs/lncRNAs (Lubas et al., 2011; Lim et al., 2017; Puno and Lima, 2018; Schmid and Jensen, 2019). Studies in cultured cell lines and in transgenic mice show that human Exosc10 is crucial for cell cycle (Blomen et al., 2015). More recent in vivo work reported that Exosc10 controls the onset of spermatogenesis in male germ cells (Jamin et al., 2017). Accordingly, Exosc10cKO mutant mice show small testes and impaired differentiation of germ cells, and exhibit reduced fertility (Jamin et al., 2017). However, whether or not Exosc10 is essential for brain development has remained unclear. Our findings reveal that the function of Exosc10 is required for the development of the forebrain. During early corticogenesis, Exosc10 is indispensable for cell viability and cortical layer formation. The requirement of the exosome complex in cell survival identified using the Exosc10cKO_Emx1-Cre cortex could possibly explain the aforementioned neurodegeneration caused by mutations of human Exosc genes.

The ring-like structured exosome complex contains eleven evolutionarily conserved subunits, including nine structural subunits (Exosc1-9) and two catalytic subunits (Exosc10 and Dis3) (Januszyk and Lima, 2014; Kilchert et al., 2016). The expression pattern analysis (Fig. S1B-J) revealed that many exosome subunits (e.g. Exosc1, Exosc2, Exosc3, Exosc5, Exosc9 and Exosc10) are widely expressed in developing mouse cortex. Remarkably, expression of some subunits is found to be limited to the VZ (Exosc8) or SVZ (Exosc7). This raises the question of whether all the components are required for the RNA exonuclease activity of the exosome complex. Even though our understanding of the functions of the exosome complex and its subunits in development has improved, several key questions remain unanswered. For example, is the composition of the exosome complex restricted to eleven subunits? Also, what is the contribution of individual subunits in formation and action of the exosome complex? Whether lineage-restricted subunits exist that lead to dynamic combinatorial assembly of exosome complexes, producing their biological specificity, remains to be determined. Efforts to resolve these and other questions would stimulate continuous interest in this area of research.

P53 is a well-known master regulator of numerous developmental events. It triggers expression of various downstream genes, some of which promote growth arrest and DNA repair, whereas others are involved in apoptosis (Mendrysa et al., 2011; Jain and Barton, 2018). The massive apoptosis observed in the early developing cortex of Exosc10cKO_Emx1-Cre embryos culminated in a severe reduction of cortical size.

To gain a transcriptome-wide insight into the role of exosome complex in corticogenesis, we carried out RNA-seq analysis of cKO cortices and Exosc10 RIP-seq. Our findings revealed upregulation of several genes associated with the P53 apoptosis pathway. Interestingly, cardinal components of apoptosis-associated P53 signaling, Aen (Kawase et al., 2008) and Bbc3 (Han et al., 2001; Jeffers et al., 2003), were identified in our RNA-seq and RIP-seq analyses. The RNA degradation assay carried out further highlighted that Exosc10 directly degrades transcripts of Aen and Bbc3.

Aen possesses exonuclease activity to degrade both DNA and RNA (Lee et al., 2005). In P53 signaling-dependent apoptosis, DNA damage signals lead to translocation of Aen into the nucleolus, causing nucleolar disruption (Kawase et al., 2008). Subsequently, Aen degrades DNA and RNA, amplifying the damage signal and inducing apoptosis (Kawase et al., 2008). The promoter region of Bbc3 contains P53-binding sites and can be directly activated by P53 (Han et al., 2001). Bbc3 is part of the BH3-only BCL-2 family proteins, which have been found to localize to mitochondria in response to apoptotic stimuli, where they induce mitochondrial apoptosis (Huang and Strasser, 2000; Lomonosova and Chinnadurai, 2008). Bbc3 has been shown to be required for γ-irradiation-induced cell death in the developing brain, and P53 is not able to induce apoptosis in the absence of Bbc3 (Jeffers et al., 2003). Our data also highlight that Exosc10 regulates cell viability in developing cortex by repressing the distinct P53-dependent apoptosis signaling pathways, including those caused by DNA/RNA damage and γ-irradiation signals. Notably, PFTα treatment significantly rescues the aberrant upregulation of Bbc3 (but not Aen) upon the loss of Exosc10 in the developing cortex (Fig. 7D,E). This proves that Bbc3 (but not Aen) is a direct target of P53.

Previous studies indicated that accumulation of RNA/DNA hybrids or noncoding RNAs as eRNAs/lncRNAs could induce cellular genomic instability leading to P53 activation and cell death (Pefanis et al., 2015; Wolin and Maquat, 2019). In addition, RNA exosome is important for DNA DSB repair as the lack of Exosc10 leads to accumulation of DNA breaks and P53 activation (Pefanis et al., 2015; Domingo-Prim et al., 2019). The findings suggest that other mechanisms in addition to the increased expression of Aen and Bbc3 cause the hyperactivation of P53 signaling in response to the defect of RNA exosome activity.

In addition to the apoptotic P53 signaling pathway, we examined the oxidative stress signaling – one of the well-known cell death-triggering pathways in the developing brain (Green, 1998; Ikonomidou, 2009). Expression of genes encoding the main components of this pathway (e.g. BAX, BH3 and Cytochrome C) was unchanged in our RNA-seq analysis, and the pathway itself was not found in a corresponding GO study (Tables S1 and S2). Thus, our findings suggest that Exosc10 inhibits apoptosis mainly by suppressing the activity of the apoptotic P53 signaling pathway.

Pharmacological inhibition of P53 signaling rescued the described defects in cell viability in the Exosc10cKO mutants (Fig. 7A-C), suggesting that the Exosc complex negatively regulates P53 signaling during early cortical development. It is worth noting, however, that the inhibition of P53 signaling was not able to restore the normal thickness of cortical layers in Exosc10cKO mutants (Fig. 7A-C). Therefore, the observed drastic reduction in the size of the cortical plate cannot be singularly ascribed to the increased apoptosis but other unreported perturbations may contribute to the observed defective neurogenesis. In agreement with this assertion, our RNA-seq and RIP-seq data suggest that Exosc10 might directly suppress expression of many neuronal differentiation-associated genes. Possible defects in neurogenesis and neuronal differentiation in the cortex-specific Exosc10cKO mutants will be in focus in a separate study.

Overall, our findings indicate a crucial role for Exosc10 in P53 pathway-mediated apoptosis, in which the binding of Exosc10 to the mRNAs of the P53 signaling mediators Aen and Bbc3 confers their rapid turnover. Our study indicates that suppression of P53 signaling by the exosome complex is essential for normal cell survival and brain development (Fig. 8G).

Conditional inversion (COIN) alleles for Exosc10 (COIN/COIN) (Economides et al., 2013; Pefanis et al., 2015), FoxG1-Cre (Hébert and McConnell, 2000) and Emx1-Cre (Gorski et al., 2002) mice (Mus musculus) were maintained in a C57BL6/J background. Animals were handled in accordance with the German Animal Protection Law.

A list of antibodies is provided in the supplementary Materials and Methods.

Detailed descriptions have been provided previously (Narayanan et al., 2015) and more detail can be found in the supplementary Materials and Methods.

Detailed descriptions have been provided previously for RNA-seq (Narayanan et al., 2015; Nguyen et al., 2018), RIP-seq (Yoon et al., 2017; Xie et al., 2019) and bioinformatics analyses (Narayanan et al., 2015; Nguyen et al., 2018). In RNA-seq experiments, RNA was obtained from cortex from five control and five Exosc10cKO embryos at E12.5. cDNA libraries were prepared using the TruSeq RNA Sample Preparation v2 Kit. DNA was quantified using a Nanodrop spectrophotometer, and its quality was assessed using an Agilent 2100 Bioanalyzer.

Exosc10 RNA-seq was performed using Magna RIP Kit (Merck Millipore) according to the manufacturer's instructions. In brief, dissociated cells from E12.5 cortex on 10 cm dish were lysed in 400 µl of complete RIP lysis buffer-containing protease inhibitors and RNase inhibitor. Exosc10 protein was pulled down using a Dynabeads-associated Exosc10 antibody. A mock pull-down was carried out with normal rabbit IgG (Cell Signaling Technologies). The immunoprecipitated complex was washed intensively and the pulled down RNA was extracted using Trizol reagent. Purified RNA was sequenced at the Transcriptome and Genome Analysis Laboratory (TAL) (University of Goettingen, Germany).

Data obtained from RNA- and RIP-seq were processed with the help of the Galaxy web platform (Afgan et al., 2018) and further analyzed using Webgestalt (http://www.webgestalt.org) (Wang et al., 2013, 2017; Zhang et al., 2005) and DAVID Bioinformatics Resources 6.8 (Huang et al., 2009). Gen sets from RNA- and RIP-seq were compared using Venny 2.1 (Oliveros). Base calling, fastq conversion, quality control and read alignments were all achieved as outlined for RIP-Seq. Reads were aligned to mouse genome mm10 and counted using FeaturesCount (http://bioinf.wehi.edu.au/featureCounts/). Differential expression was assessed using DESeq2 from Bioconductor (Love et al., 2014). Functional GO enrichment analyses were performed using ToppGene (Chen et al., 2009).

Mouse NSCs were isolated from E12.5 Exosc10 COIN/COIN cortices and cultured in NSC culture medium containing KO DMEM/F12 (Invitrogen), StemPro Neural Supplement (Invitrogen), Glutamax (Invitrogen), penicillin/streptomycin (Invitrogen), 20 ng/ml FGF2 (Invitrogen) and 20 ng/ml EGF (Invitrogen) on culture dishes precoated with 0.1% gelatin as described previously (Tuoc and Stoykova, 2008; Tuoc et al., 2013). TAT-Cre recombinase (Excellgen) (1 µM) was added to fresh NSC culture medium for 26 h to achieve Exosc10 KO.

The assay was performed as described previously (Yoon et al., 2017; Xie et al., 2019). Briefly, Exosc10 COIN/COIN mouse NSCs were cultured to about 70-90% confluence. Some wells were treated with 1 µM TAT-Cre (Excellgen) 26 h beforehand to achieve Exosc10 KO; cells with vehicle solution served as controls. Actinomycin D (5 mM, Sigma) was supplemented to fresh NSC culture medium. Subsequently, cells were harvested at different time points (0 h, 5 h and 17 h) by washing once with PBS and detaching using Tripsin/EDTA (Sigma). For transcript quantification, RNA was extracted from the NSC samples. qPCR was carried out to quantify the transcript level of target genes. The experiment was performed in triplicate and normalized to internal 18S. Fold changes of transcript targets between Exosc10 KO and control NSCs were compared at different time points (0 h, 5 h and 17 h) after actinomycin D treatment. Fold changes at 5 h and 17 h were normalized to those at 0 h (before actinomycin D treatment).

Exosc10cKO_Emx1-Cre mouse embryos were subjected to the p53-inhibitor Pifithrin-α (PFT-α) (Komarov et al., 1999) by intraperitoneal injection of 2.2 mg/kg PFT-α (Selleckchem) into the pregnant mother at between E9.5 and E12.5 or between E9.5 and E15.5. Embryonic brains were isolated at E13.5 or E18.5 and immunohistochemistry was performed.

Dorsal views of forebrains of mutant and control mice were photographed under a dissection microscope. Cortical anterior-posterior axis (AP), cortical surface and midline lengths from the digitized images were measured with Fiji software to make comparison between mutants and controls. For further details, see the supplementary Materials and Methods.

IHC quantification was performed using anatomically matched coronal sections. In most cases, cell counts of six matched sections were averaged (control/cKO). For quantitative analyses of IHC signal intensity of cytoplasm-staining markers, fluorescent images of selected areas of the cortex were used. Color images were converted to gray scale and the fluorescent signal intensity values were measured using the Analyze/Measure function of Fiji software. The signal intensity from the background next to the tissue was subtracted from the measured intensity for normalization. For further details, see the supplementary Materials and Methods.

Imaging was performed with an Axio Imager M2 (Zeiss) with a Neurolucida system (Version 11; MBF Bioscience) and a confocal fluorescence microscope (TCS SP5; Leica). Images were further analyzed with Adobe Photoshop and Fiji. Statistical analyses were carried out using Student's t-test. Graphs are plotted as mean±s.e.m. An unpaired t-test was carried out on the average from at least three biological replicates. All details of statistical analyses and description for histological experiments are presented in Table S7 and in the supplementary Materials and Methods.

We acknowledge T. Huttanus and M. Blessmann for their expert animal care and support. We also thank F. Guillemot, A. Nave, A. P. McMahon, A. Jones and O. Machon for providing reagents.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.188276.reviewer-comments.pdf