Reciprocal control of translation and transcription in autism spectrum disorder

Abstract

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social communication and the presence of restricted patterns of interest and repetitive behaviors. ASD is genetically heterogeneous and is believed to be caused by both inheritable and de novo gene variations. Studies have revealed an extremely complex genetic landscape of ASD, favoring the idea that mutations in different clusters of genes interfere with interconnected downstream signaling pathways and circuitry, resulting in aberrant behavior. In this review, we describe a select group of candidate genes that represent both syndromic and non‐syndromic forms of ASD and encode proteins that are important in transcriptional and translational regulation. We focus on the interplay between dysregulated translation and transcription in ASD with the hypothesis that dysregulation of each synthetic process triggers a feedback loop to act on the other, which ultimately exacerbates ASD pathophysiology. Finally, we summarize findings from interdisciplinary studies that pave the way for the investigation of the cooperative impact of different genes and pathways underlying the development of ASD.

Autism‐spectrum disorder (ASD) is a genetically complex and heterogenous disorder. Here, the authors review a select group of ASD‐associated genes and discuss the interplay between dysregulated translation and transcription in the pathophysiology of ASD.

Glossary

4E‐BP

eIF4E‐binding protein

ADHD

attention deficit and hyperactivity disorder

ADNP

activity‐dependent neuroprotective protein

aNCSs

adult neural stem cells

APOE

apolipoprotein E

ASD

autism spectrum disorder

BDNF

brain‐derived neurotrophic factor

BNIP3

BCL2 and adenovirus E1B 19‐kDa‐interacting protein 3

Brd4

bromodomain‐containing protein 4

CACNA1E

calcium voltage‐gated channel subunit alpha1 E

CACNB1

calcium voltage‐gated channel auxiliary subunit beta 1

CaMKII

Ca²⁺/calmodulin‐dependent protein kinase II

CNS

central nervous system

CNVs

copy number variants

CREB

cAMP response element‐binding protein

CSF

cerebrospinal fluid

CTGF

connective tissue growth factor

CYFIP1

cytoplasmic FMRP‐interacting protein 1

DA

dopamine

DAT

dopamine transporter

eIF4A

eukaryotic translation initiation factor 4 A

eIF4E

eukaryotic initiation factor 4E

eIF4G

eukaryotic translation initiation factor 4 G

EN1

engrailed 1

EN2

engrailed 2

ERK

extracellular signal‐regulated kinase

Fmr1

fragile X mental retardation 1

FXS

fragile X syndrome

GAP

GTPase‐activating protein

GAPDH

glyceraldehyde 3‐phosphate dehydrogenase

Grm5

glutamate metabotropic receptor 5

HDVAS

Helsmoortel‐Van der Aa syndrome

HIF1a

hypoxia‐inducible factor 1‐alpha

HITS‐CLIP

high‐throughput sequencing of RNAs isolated by cross‐linking immunoprecipitation

IGF‐1

insulin‐like growth factor 1

IL‐6

interleukin 6

KO

knockout

LIF

leukemia inhibitory factor

LTD

long‐term depression

LTP

long‐term potentiation

MeCP2

methyl‐CpG‐binding protein 2

MEK

MAPK/ERK kinase

mGluR

metabotropic glutamate receptor

MNK

MAP kinase‐interacting kinase

mTORC1

mammalian/mechanistic target of rapamycin complex 1

Myl2

myosin regulatory light chain 2

NMDAR

N‐methyl‐D‐aspartate receptor

Nrxn3

neurexin 3

PABP

poly(A)‐binding protein

P‐CREB

phosphorylated CREB

PDE4D

cAMP‐specific 3',5'‐cyclic phosphodiesterase 4D

PDK1

3‐phosphoinositide‐dependent protein kinase 1

PI3K

phosphoinositide 3‐kinase

PIP3

phosphatidylinositol (3,4,5)‐trisphosphate

PSD95

postsynaptic density protein 95

PTEN

phosphatase and tensin homolog

PV

parvalbumin

Rack1

receptor for activated C kinase 1

REDD1

regulated in development and DNA damage responses 1

RHEB

ras homolog enriched in brain

RYK

receptor tyrosine kinase

S6K1

ribosomal protein S6 kinase beta‐1

SCN1A

sodium voltage‐gated channel alpha subunit 1

SIRT1

sirtuin 1

SLC6A4

solute carrier family 6 member 4

SWI/SNF

SWItch/Sucrose non‐fermentable

TH

tyrosine hydroxylase

TrKB

tropomyosin receptor kinase B

TSC1

tuberous sclerosis complex 1

TSC2

tuberous sclerosis complex 2

ZFP161

zinc finger protein 161

Introduction

Autism spectrum disorder (ASD) comprises a class of heterogeneous neurodevelopmental diseases characterized by three salient features: early‐onset difficulties in communication, reduced social interaction, and repetitive, restricted behavior/interests (McPartland & Volkmar, 2012; Kulage et al, 2014). Individuals diagnosed with ASD often present with co‐morbid psychiatric and medical disorders, such as attention deficit and hyperactivity disorder (ADHD), intellectual disability, epilepsy, anxiety, eating problems, and alterations in mood and sleep and mood (Zoghbi & Bear, 2012; Mannion & Leader, 2013; Romero et al, 2016). In many cases, the co‐morbid conditions lead to more severe impairments as a result of the cumulative effects of these different disorders, as well as a synergistic action of the various molecular pathways involved.

ASD is highly heritable and yet is a genetically heterogeneous disorder (Hallmayer et al, 2011). Over the last few decades, great progress has been made in understanding the genetic etiology of ASD, spurred by a wide array of genome‐wide analyses and approaches (Iossifov et al, 2012; Neale et al, 2012; O’Roak et al, 2012; Sanders et al, 2012). These genetic studies have helped to reveal neurobiological mechanisms, as well as the contribution of environmental and epigenetic factors (Kim & Leventhal, 2015) involved in the pathophysiology of ASD. Advances in genetic and genomic methodologies have identified a large number of ASD‐associated de novo mutations, including single‐base pair mutations (single‐nucleotide variants; SNVs) and a number of copy number variants (CNVs) that disrupt protein‐coding genes (Sanders et al, 2012; Iossifov et al, 2014), as well as mutations located within intronic and intergenic regions (Zhou et al, 2019b). In this heterogeneous genetic architecture, syndromic forms of ASD (Box 1) caused by highly penetrant single‐gene mutations represent only a minority of the ASD cases (Sztainberg & Zoghbi, 2016). In contrast, most of the cases are idiopathic with common variants contributing to a significant fraction (40–60%) of genetic liability of ASD (Klei et al, 2012; Gaugler et al, 2014; Zhou et al, 2019b).

Box 1. Syndromic and non‐syndromic ASD.

ASD is a heterogeneous heritable neuropsychiatric disorder that can be classically categorized into two types: syndromic and non‐syndromic autism. Syndromic autisms are caused by mutations in a particular gene or a set of genes and are manifested within the context of neurological syndromes, such as Fragile X Syndrome, Tuberous Sclerosis, or Rett Syndrome. Non‐syndromic autism, which comprises a vast majority of autism cases, is not linked to other neurological diseases (or syndromes) but is also heritable with a number of genes that are linked to this class of autism.

To date, these large‐scale sequencing efforts certainly have contributed to unveiling the genetic etiology of ASD and have advanced our understanding of potential neurobiological mechanisms that give rise to this disorder. From this perspective, several studies have emphasized the categorization of ASD risk gene groups by their putative functions: ASD‐related mutations tend to be significantly enriched in genes involved in synaptic function and structure, and in genes encoding for transcripts that are important in transcriptional and translational regulation (De Rubeis et al, 2014; Krumm et al, 2014; de la Torre‐Ubieta et al, 2016).

Given the extremely complex genetic landscape of ASD, it has been proposed that mutations in these different clusters of genes interfere with interconnected downstream signaling pathways and circuitry in a cell type‐specific manner, resulting in ASD pathophysiology (Santini & Klann, 2014; de la Torre‐Ubieta et al, 2016; Mullins et al, 2016; Wang et al, 2018). The identification of a cluster of genes encoding for transcripts involved in different phases of mRNA translation (Darnell et al, 2011; Santoro et al, 2012) supports one of the unifying theories on ASD according to which dysregulation of translational control represents a common endpoint of familial and sporadic ASD‐associated signaling pathways (Kelleher & Bear, 2008; Darnell & Klann, 2013; Santini & Klann, 2014). Consistent with this notion, several signaling pathways upstream of translation, including PI3K (phosphoinositide 3‐kinase)/Akt/mTORC1 (mammalian/mechanistic target of rapamycin complex 1) and ERK (extracellular signal‐regulated kinase) are dysregulated in several types of ASD (Sonenberg & Hinnebusch, 2009; Darnell & Klann, 2013), resulting in increased net de novo protein synthesis, which in turn leads to altered synaptic plasticity and ASD‐like phenotypes (Auerbach et al, 2011; Gkogkas et al, 2013a; Santini et al, 2013). However, the physiological outcome of increased translation of specific mRNAs differs depending on the different regulatory processes and ASD‐linked genes tightly coalesce in modules that implicate distinct biological functions during development, including early transcriptional regulation and synaptic development (Parikshak et al, 2013). Interestingly, recent studies on monogenic forms of ASD suggest that not all excessively translating mRNAs contribute to pathological changes, raising the hypothesis that some of these mRNAs may be protective adaptations (Thomson et al, 2017). On the other hand, excessive translation of certain transcripts may result in widespread epigenetic misregulation (Korb et al, 2017) creating a positive feedback loop between translational and transcriptional processes that exacerbate neuronal dysfunction in ASD (Box 2).

Box 2. In need of answers.

• ASD‐associated genes are often regulators of the expression of a large group of other genes with multiple molecular functions. No gene operates alone. The extremely complex genetic landscape of ASD favors the idea that mutations in different clusters of genes interfere with interconnected downstream signaling pathways and circuitry. Which synaptic and non‐synaptic regulatory processes are particularly susceptible to altered translation and transcription and are some downstream biological processes more relevant than others to drug treatments?

• Does the epigenetic misregulation resulting from an excessive translation of certain transcripts act at a critical time in development, and when would treatment be most effective?

• We have hypothesized a bidirectional regulation of translation and transcription in ASD and that mutations in genes that result in dysregulation of one of those processes triggers a positive feedback loop to act on the other, which exacerbates ASD pathophysiology. Is there a specific brain region where to identify a core set of mistranslated mRNAs that mediate the complex level of dynamic regulation between translation and transcription underlying ASD pathophysiology?

• The direct effect of specific proteins encoded by genes implicated in ASD, regulating the translation of transcriptional regulators is a likely way to explain many of the changes in mRNA levels reported in ASD. The possibility that the observed regulatory interplay between translation and transcription may not be direct for some of those genes arise the question on whether changes observed in one process from disruptions in the other could likely represent a protective and compensatory mechanisms rather than a positive feedback loop exacerbating ASD pathophysiology, aiming further investigation.

Here, we examine the evidence for interplay between dysregulated translation and transcription in ASD. We review a select group of ASD‐associated genes representative of both syndromic and non‐syndromic forms of ASD that are involved in either translational or transcriptional regulation and describe the molecular function of their protein products, and the ASD‐associated clinical phenotypes caused by disruption of these genes. Then, we summarize findings from interdisciplinary studies that have queried the impact of dysregulated translational control on transcription and dysregulated transcriptional control on translation with respect to ASD pathophysiology.

Regulation of translation and ASD

Examination of the different genetic studies and transcriptomic analyses (Voineagu et al, 2011; De Rubeis et al, 2014; Gupta et al, 2014) that have been performed to date indicates that ASD‐associated genes are often regulators of the expression of a large group of other genes with multiple molecular functions. The phosphatidylinositol 3‐kinase (PI3K)‐mechanistic target of rapamycin (mTOR) signaling cascade is one of the most commonly studied pathways where mutations occur and give rise to ASD‐related phenotypes (Huber et al, 2015). Overactivation of the PI3K‐mTOR pathway has been described in fragile X syndrome (FXS) (Sharma et al, 2010; Hoeffer et al, 2012), and mutations in genes encoding for negative regulators of the mTOR pathway, such as tuberous sclerosis complex 1 (TSC1), tuberous sclerosis complex 2 (TSC2), and phosphatase and tensin homolog (PTEN) (Kwon et al, 2006) are consistent with increased mTOR activity in cells isolated from individuals with ASD (Onore et al, 2017). Because mTOR complex 1 (mTORC1) is a well‐known regulator of translation (Hay & Sonenberg, 2004; Ma & Blenis, 2009; Zoncu et al, 2011), deletion or loss‐of‐function mutations in genes encoding for proteins involved in regulating mTORC1 lead to alteration of the translational control machinery, ultimately resulting in increased de novo translation, altered protein synthesis‐dependent synaptic plasticity, and the appearance of behavioral defects consistent with ASD (Kelleher & Bear, 2008; Ebert & Greenberg, 2013; Santini & Klann, 2014). In the next section, we will briefly summarize the impact of dysregulated translational control on transcription focusing on three ASD‐associated genes involved in translation regulation: fragile X mental retardation 1 (FMR1), phosphatase and tensin homolog (PTEN), and tuberous sclerosis complex (TSC).

Fragile X mental retardation protein (FMR1)

Transcriptional silencing of the fragile X mental retardation 1 (FMR1) gene results in the loss of fragile X mental retardation protein (FMRP) and gives rise to fragile X syndrome (FXS) (Verkerk et al, 1991). This form of intellectual disability represents one of the most prominent single‐gene causes of ASD, accounting for an estimated 1% to 6% of all cases (Schaefer & Mendelsohn, 2013). Reduced expression or absence of FMRP leads to atypical brain development and dendritic spine anomalies (Irwin et al, 2000), resulting in cognitive and motor dysfunction in FXS individuals such as hyperactivity, anxiety, attention deficit disorder, speech perseveration, impulsive behavior, and stereotypical and repetitive movements (Martin & Bell, 1943; Dykens et al, 1989; Hodapp et al, 1990; Cordeiro et al, 2011).

FMRP is an mRNA‐binding protein that operates as regulator of translation (Darnell & Klann, 2013), among other functions (Contractor et al, 2015). Studies of Fmr1 knockout (KO) mice indicate that FMRP has a crucial role in the activity‐dependent regulation of mRNA transport (Dictenberg et al, 2008), translation, and stability in neurons (Bassell & Warren, 2008; Darnell et al, 2011). Evidence from cells derived from FXS patients and FXS model mice suggests that FMRP represses the initiation and elongation steps of translation, which results in a net increase in protein synthesis when expression of FMRP is absent in FXS (Darnell & Klann, 2013). With respect to translation initiation, FMRP interacts with cytoplasmic FMRP‐interacting protein 1 (CYFIP1), which associates and sequesters the cap‐binding protein eukaryotic initiation factor 4E (eIF4E), thereby inhibiting the translation initiation of specific mRNAs (Napoli et al, 2008). FMRP also regulates translation elongation, where it has been shown that FMRP stalls ribosomal translocation on specific mRNAs during elongation (Darnell et al, 2011), although the molecular mechanisms responsible for this regulation remain to be determined.

Although the mechanisms of translational control by FMRP have been extensively described (Bassell & Warren, 2008; Costa‐Mattioli et al, 2009; Darnell et al, 2011) and several FMRP‐regulated transcripts and pathways have been identified (Darnell et al, 2001, 2011), the mRNAs with differential translation that contribute to neuronal function or that impact gene expression have not been evaluated comprehensively in FXS, although several recent studies provided significant progress in this direction (Ceolin et al, 2017; Korb et al, 2017; Thomson et al, 2017; Liu et al, 2018). New insights into genome‐wide gene expression changes resulting from FMRP deficiency come from a study on adult neural stem cells (aNCSs) derived from wild‐type and Fmr1‐deficient mice (Liu et al, 2018). Using simultaneous high‐resolution ribosome profiling and transcriptomic analysis, Liu and colleagues described diverse expression changes at both mRNA and translation levels, revealing a translational buffering mechanism by FMRP. A form of buffering whereby the loss of FMRP leads to increased translation of mRNAs, which then stimulates a feedback decrease in the levels of mRNA, in order to conserve the amount of protein produced. Accordingly, FMRP deficiency results in dysregulation of numerous mitosis and neurogenesis genes primarily at the RNA expression level, whereas the expression of many synaptic genes is mostly dysregulated at the translation level (Liu et al, 2018). Moreover, exaggerated translation of epigenetic or transcriptional transcripts, which are direct targets of FMRP, may indirectly result in the upregulated expression of transcription regulators such as Ndn, an intronless gene encoding for a growth suppressor protein in post‐mitotic neurons (Liu et al, 2018).

Several direct FMRP target mRNAs encoding transcription factors and chromatin modifiers (Darnell et al, 2011; Korb et al, 2017) have been identified by using high‐throughput sequencing of RNAs isolated by cross‐linking immunoprecipitation (HITS‐CLIP). Loss of FMRP in mice results in elevated expression of several chromatin‐associated proteins, including bromodomain‐containing protein 4 (Brd4). The upregulation of Brd4, as well as of other chromatin‐associated FMRP targets such as MLL1 and p300, results primarily in an increase in histone modifications (i.e. acetylation and methylation) and active chromatin (Korb et al, 2017). In addition, Brd4 acts as activator hub for several FMRP targets (Rahman et al, 2011; Shen et al, 2015) and its loss is correlated with the decreased expression of synaptic genes (Korb et al, 2015, 2017; Liu et al, 2018), which are increased in FXS (Darnell et al, 2011; Korb et al, 2017). Finally, inhibition of Brd4 using the small molecule JQ1 reversed several of the gene expression changes observed in FMRP‐lacking neurons and alleviated many phenotypes displayed by Fmr1 KO mice (Korb et al, 2017). Interestingly, many of the chromatin‐associated FMRP targets are also ASD‐linked genes; mutations in these genes result in either increased or decreased expression (Korb et al, 2017). Taken together, these observations provide insight into FMRP‐dependent translatome and transcriptome changes in FXS, suggesting that FMRP functions by regulating translation to modify transcription (Fig 1). Misregulation of gene expression at both the transcription and translation levels likely contribute to FXS, creating a positive feedback loop which exacerbates molecular and neuronal dysfunction. Given these observations, targeting the synergistic dysregulation of translational and transcriptional mechanisms, perhaps using transcription inhibitors (Korb et al, 2017), to rebalance global changes in gene expression may represent a successful therapeutic approach for treatment of FXS.

Figure 1. Dysregulation of translation and transcription in FXS.

(Left) FMRP regulates the translation of certain mRNAs, including synaptic and chromatin‐associated proteins, such as Brd4 (red) and p300 (yellow), which facilitate the initiation and elongation phases of transcription by binding to activated chromatin at acetylated lysine residues. (Right) Dysregulation of translation in FXS impacts gene expression at the transcription level creating a feedback loop which exacerbates both molecular and neuronal dysfunction.

Phosphatase and tensin homolog (PTEN)

Originally identified as a cancer predisposition gene (Liaw et al, 1997), the phosphatase and tensin homolog (PTEN) gene located on chromosome 10q23 is an ASD risk gene (Butler et al, 2005; Buxbaum et al, 2007; Varga et al, 2009). Human genetic studies showed that ASD and macrocephaly are associated with germline heterozygous PTEN mutations, with PTEN‐associated ASD accounting for approximately 10% of macrocephalic ASD (Butler et al, 2005; Varga et al, 2009; Frazier et al, 2015). Nearly all individuals with PTEN‐associated ASD exhibit severe cognitive deficits, seizures, motor dysfunction, and verbal ability delay (Varga et al, 2009; Child & Cascino, 2013; Frazier et al, 2015), consistent with the macrocephaly white matter abnormalities that have been reported (Frazier et al, 2015).

PTEN encodes for a lipid phosphatase (PTEN) with specificity toward phosphatidylinositol (3,4,5)‐triphosphate (PIP3) (Parsons, 2004), which acts as a negative regulator of the PI3K/Akt/mTORC1 signaling pathway (Maehama & Dixon, 1998). PTEN inhibition of PIP3 results in decreased phosphorylation of Akt and germline PTEN mutations lead to reductions in PTEN protein levels, increased Akt phosphorylation, and subsequently a constitutively active PI3K/Akt/mTORC1 signaling pathway (Tan et al, 2011). Activated Akt also phosphorylates and inhibits TSC2 (also known as tuberin) (Inoki et al, 2002; Potter et al, 2002), thus removing the inhibition of mTORC1 by the TSC1/2 complex (Fig 2). Due to the important role of PI3K/Akt/mTORC1 pathway in the regulation of cell growth, survival, and proliferation, it is not surprising that PTEN inactivation has been observed in a broad spectrum of human cancers. However, ASD‐associated PTEN mutations affect Akt inhibition differently than those appearing in adults with predominantly cancer phenotypes and seem not to substantially abrogate PTEN activity in vivo (Rodríguez‐Escudero et al, 2011).

Figure 2. Mutations in PTEN and TSC1/2 are associated with dysregulation of translation and transcription.

(Left) PTEN regulates translational/transcriptional systems, playing an important role in neuronal proliferation, migration, survival, morphology, and plasticity. As a lipid phosphatase, PTEN catalyzes the removal of the D3 phosphate from PIP3, which acts as a negative regulator of the PI3K/Akt/mTORC1 signaling pathway, whereas PTEN’s protein phosphatase activity targets a number of substrates, including MAPK signaling, which PTEN inhibits. The TSC1–TSC2 complex, through its GTPase‐activating protein activity towards the small G‐protein Rheb, is downstream effectors of the PI3K/Akt pathway and a critical negative regulator of mTORC1, regulating translation, transcription, and other cellular processes. (Right) Mutations in both PTEN and TSC1/2 trigger the activation of mTORC1, thereby regulating translation, transcription, and other cellular processes. Exaggerated mTORC1 signaling results in disinhibited protein synthesis. The loss of either PTEN or TSC1/2 activity directly impact gene transcription and the translation of specific mRNAs, such as TH and CTGF, and a number of synaptic scaffolding proteins such as PSD‐95 and SAPAP1.

Several studies have focused on the role of PTEN mutations in ASD by generating gene disruption models in mice to investigate PTEN neurobiology and unveil the cellular processes regulated by this phosphatase (reviewed in Tilot et al, 2015). PTEN heterozygous mice represent the most investigated model for PTEN loss in the CNS (Tilot et al, 2015). However, endophenotypes displayed by conditional PTEN knockout mice more closely resemble PTEN‐related ASDs and provide insight into the different brain regions and cell types involved in ASD. Nse‐Cre mice, where PTEN has been deleted selectively in a subset of neurons in cortex and hippocampus, exhibit social deficits and increased repetitive and perseverative behaviors (Kwon et al, 2006; Napoli et al, 2012) suggesting that PTEN simultaneously regulates proliferation and connectivity in the CNS. Pan‐neural PTEN loss results in hyperproliferation and increased cell size, which are due to the overactivation of Akt/mTORC1 signaling (Kwon et al, 2001; Amiri et al, 2012). Elevated mTORC1 signaling also is responsible for increased branching of the dendritic arbor, thickening of dendritic processes, and ectopic axonal projections exhibited by PTEN null neurons (Kwon et al, 2006; Fraser et al, 2008). Thus, morphological aberrations of neurons lacking PTEN may be responsible for the significant changes in firing properties and synaptic plasticity observed in different PTEN mouse models (Luikart et al, 2011). For example, PTEN heterozygous KO mice exhibit decreased long‐term potentiation (LTP) and impaired NMDA receptor‐dependent long‐term depression (LTD) at hippocampal CA1 synapses (Wang et al, 2006). GFAP‐Cre PTEN conditional KO mice also display decreased LTP at CA1 synapses (Fraser et al, 2008) as well as impaired metabotropic glutamate receptor‐dependent (mGluR)‐dependent LTD at dentate granule cell synapses(Takeuchi et al, 2013). However, it has been suggested that the loss of PTEN may not induce de novo spinogenesis but instead may increase synaptic activity by inducing morphological and functional maturation of spines (Haws et al, 2014). While all these studies imply that increased spine density or maturation underlie the changes in synaptic plasticity, mice with a CamKIIα‐Cre‐dependent postnatal deletion of PTEN exhibit LTP and LTD deficits at CA1 synapses in the hippocampal neurons, but no alterations in either neuronal or dendritic morphology, suggesting that these phenotypes are not necessarily dependent on each other (Sperow et al, 2012). Moreover, rapamycin, the pharmacological inhibitor mTORC1 signaling, rescues the neuronal hypertrophic phenotypes and behavioral abnormalities associated with loss of PTEN, but surprisingly the loss of S6K1, a downstream effector of mTORC1 and ERK does not result in the same outcome (Kwon et al, 2003; Chalhoub et al, 2006; Zhou et al, 2009).

Given the increasing evidence that common/shared biological pathways or brain circuits may account for the cellular and molecular mechanisms involved in ASD development, it has been proposed that although loss of PTEN in neurons is sufficient to cause social deficits in mice, one cannot exclude the possibility that other ASD‐related genes are necessary for the development of the complex autistic phenotype seen in patients carrying PTEN mutations (Zhou & Parada, 2012). Indeed, the deletion of PTEN alters the expression of the ASD‐related genes FMRP and MeCP2. The activation of the PI3K/Akt/mTORC1 pathway in the Nse‐Cre, PTEN mouse model, for example, results in a decreased number of synaptic scaffolding proteins such as PSD‐95 and SAPAP1, decreased mGluR, and increased FMRP expression (Backman et al, 2001; Kwon et al, 2001). Moreover, both PTEN and transcription factor MeCP2 reciprocally regulate the expression of each other via microRNAs (Lyu et al, 2016), 20–25‐nucleotide‐long noncoding RNAs that modulate gene expression and development by post‐transcriptionally targeting RNA‐induced silencing complexes (Bartel, 2004). Lyu and colleagues reported that deletion of PTEN leads to increased phosphorylation of Serine 133 of CREB (P‐CREB) and increased expression of its target microRNA‐132, which in turn inhibits the expression of MeCP2 by targeting the 3′UTR of MeCP2 mRNA. The loss of PTEN thus results in lower MeCP2 expression. Interestingly, it has been shown that microRNA‐132 inhibitors largely blocked the effects of CREB on dendrite maturation (Magill et al, 2010). On the other hand, knockdown of MeCP2 leads to upregulation of microRNA‐137, which in turn represses expression of PTENP (Lyu et al, 2016). Thus, PTEN is downregulated when MeCP2 is knocked down.

Additional evidence for association between ASD‐linked PTEN mutations and dysregulated translational/transcriptional systems underlying ASD comes from a study on PTEN missense (PTEN^m3m4) knock‐in transgenic mice (He et al, 2015b). This study is based on the evidence that conditional Pten deletion in dopaminergic neurons of mice results in elevated expression of tyrosine hydroxylase (TH), the key enzyme in the dopamine (DA) biosynthesis pathway, associated with increased striatal DA content (Domanskyi et al, 2011). PTEN normally acts to suppress tyrosine hydroxylase (TH) transcription through the inhibitory effect on PI3K and P‐CREB but also inhibits TH phosphorylation by suppressing the ERK pathway. There is thus an increase in TH expression and phosphorylation in the cerebrum and in the striatum of PTEN^m3m4 mice (He et al, 2015b). Dopamine (DA) exerts a pivotal role in the modulation of locomotive function, learning and memory, and social interactions (Wise, 2004; Shohamy & Adcock, 2010; Rice et al, 2011; Matthews et al, 2016; Torquet et al, 2018). Dysregulation of the dopaminergic system has been linked to ASD pathophysiology, including exaggerated repetitive and perseverative behaviors (McDougle et al, 2005; DiCarlo et al, 2019). An augment in TH synthesis and function leads to the release of too much dopamine in the prefrontal cortex and midbrain, which can result in repetitive and perseverative behaviors (He et al, 2015a). Conditional deletion of PTEN in dopaminergic neurons results in neuronal hypertrophy, an increased number of dopaminergic neurons and social behavioral deficits (Clipperton‐Allen & Page, 2014). Consistent with studies of the PTEN^m3m4 mice, the ASD‐linked PTEN mutations H93R, F241S, and D252G were not able to suppress TH when overexpressed in PC12 cells. In addition, ectopic expression of two other PTEN missense mutations that either lack phosphatase function in general (C124S) or only lipid phosphatase function (G129E) failed to suppress TH in PC12 cells (He et al, 2015b). Together, these findings suggest that ASD‐associated PTEN mutations can enhance TH expression in two ways: the loss of the lipid phosphatase activity of PTEN will induce TH transcription and translation through the dysregulated PI3K/CREB pathway and the loss of the protein phosphatase activity of PTEN will enhance the phosphorylation of TH through the ERK pathway (He et al, 2015b). Therefore, the loss of both lipid and protein phosphatase activity of PTEN results in a synergic effect on TH regulation leading to ASD.

Finally, microarray studies have revealed changes in gene transcription profiles following PTEN depletion or overexpression (Matsushima‐Nishiu et al, 2001; Carver et al, 2011; Mulholland et al, 2012; Chen et al, 2014), and PTEN was proposed to repress transcription by maintaining a condensed chromatin structure through a physical interaction with histone H1. Loss of PTEN results in the displacement of histone H1, chromatin decondensation and the elevation of H4K16 acetylation (Chen et al, 2014), which opens chromatin structure for transcription (Chen et al, 2014).

Tuberous sclerosis complex (TSC)

Tuberous sclerosis complex (TSC) is an autosomal‐dominant inherited disorder caused by loss‐of‐function mutations in either TSC1 or TSC2 genes (Tsai & Sahin, 2011). TSC affects 1/6,000 newborns worldwide with ASD occurring in more than 50% of the individuals (Ehninger & Silva, 2011; Han & Sahin, 2011; Gipson et al, 2013) and accounting for 1–4% of all cases of ASD (Fombonne, 2003). Clinical manifestations associated with TSC are variable with respect to symptoms and disease severity, which is in part dependent on which TSC gene is affected. TSC patient populations are characterized by high prevalence of epilepsy (∼ 90%), intellectual disability, ASD, ADHD, sleep disruption, and additional psychiatric features in adults, such as anxiety and mood disorders (Ehninger & Silva, 2011; Han & Sahin, 2011). Neuropathologies associated with TSC include formation of cortical tubers, which are characterized by dysplastic neurons with immature electrophysiological properties (Cepeda et al, 2012), subependymal nodules, and subependymal giant cell astrocytomas (DiMario, 2004). Extensive studies of the TSC1 and TSC2 genes have revealed a wide spectrum of mutations in TSC individuals. These mutations comprise a mix of missense, nonsense, insertions, and deletions involving nearly all exons present in both the TSC1 and TSC2 genes (Rosset et al, 2017), and their impact on the clinical phenotypes is extremely variable (Ehninger & Silva, 2011; Han & Sahin, 2011). Individuals with heterozygous mutations in either of the TSC1 or TSC2 genes have a ~ 100‐fold increase in the probability of being diagnosed with ASD compared with the general population, which is phenotypically similar to ASD than other etiologies at the behavioral level (Hunt & Shepherd, 1993; Smalley, 1998; Ehninger & Silva, 2011).

TSC1 and TSC2 encode for hamartin (also referred to as TSC1) and tuberin (also referred to as TSC2), respectively, which are canonical components of the mTOR pathway (Tsai & Sahin, 2011; Lipton & Sahin, 2014). As mentioned above (see PTEN section), phosphorylation of TSC2 by Akt inhibits its function. This relieves the inhibition of TCS1/2 on mTORC1 and promotes its activity (Inoki et al, 2002; Takei et al, 2004). In neurons, TSC1 and TSC2 proteins form a heterodimer that regulates protein synthesis and cell size (Kwiatkowski & Manning, 2005). TSC1 functions as a regulator of TSC2 stability, preventing TSC2 degradation (Benvenuto et al, 2000), whereas TSC2, a GTPase‐activating protein (GAP), inactivates Rheb, a Ras family GTPase, and other small G proteins (Kwiatkowski & Manning, 2005). The phosphorylation of TSC2 by Akt results in Rheb activation, which in turn triggers the activation of mTORC1, thereby regulating translation, transcription, and other cellular processes (Kwiatkowski & Manning, 2005). Without a normally functioning TSC1/2 complex, mTORC1 is hyperactive, resulting in disinhibited protein synthesis and subsequent cell growth (Ruvinsky & Meyuhas, 2006; Wullschleger et al, 2006). Thus, mutations in TSC1 and TSC2 are consistent with the notion that dysregulated neuronal translation underlies several forms of ASD (Fig 2) (Ehninger & Silva, 2011).

Human neurons derived from TSC2‐deficient pluripotent stem cells exhibited developmental abnormalities that recapitulate pathological hallmarks of cortical malformations in patients such as altered synaptogenesis and synaptic transmission paralleled by molecular changes in pathways associated with ASD. TSC2 deletion leads to exaggerated mTORC1 signaling during differentiation of human neurons in a gene‐dosage‐dependent manner. Pharmacological inhibition of mTORC1 corrects developmental abnormalities and synaptic dysfunction, supporting the notion that the lack of TSC2 causes alterations that are dependent on hyperactive mTORC1 signaling (Costa et al, 2016). Synaptic abnormalities were also observed in astrocyte‐specific Tsc1 deletion mice and in the Eker rat model of TSC (Von Der Brelie et al, 2006). Several different heterozygous TSC1 or TSC2 rodent models exhibit behavioral abnormalities, including social interaction and learning deficits (Goorden et al, 2007, Waltereit et al, 2011, Von Der Brelie et al, 2006), which are reversed by acute rapamycin treatment. The behavioral deficits displayed by the various TSC model mice are often correlated with impaired synaptic plasticity. For example, mice with a heterozygous deletion of Tsc2 (TSC2 ^+/−) have deficient hippocampal mGluR‐LTD and decreased translation of Arc mRNA (Auerbach et al, 2011). Similarly, postnatal deletion of Tsc1 in hippocampal CA1 neurons in mice results in impaired mGluR‐LTD and increased excitatory synaptic function (Bateup et al, 2011). Notably, the hyperactivation of the mTORC1 pathway results in reduced mGluR‐LTD and reduced de novo protein synthesis in TSC model mice. To note, this is in contrast to Fmr1 KO mice, which exhibit increased net de novo translation, enhanced mTOR signaling and exaggerated mGluR‐LTD at CA1 synapses (Sharma et al, 2010). Moreover, inhibition of mTORC1 and augmentation of mGluR5 signaling in Tsc2 ^+/− mice restored mGluR‐LTD and normalized de novo protein synthesis. These findings suggest that TSC1/TSC2 and FMR1 mutations have opposite effects on the tight regulation of translation that underlies proper synaptic function. It has been suggested that TSC1/TSC2 and FMRP may differentially regulate the translation of different pools of mRNA and that their dysregulated transcripts may have a different outcome in terms of synaptic function. An alternative possibility is that enhanced mTORC1 activity in TSC may result in increased FMRP activation, resulting in suppression of specific mRNAs (Ebert & Greenberg, 2013). In line with this hypothesis, deleting Fmr1 in Tsc2 ^+/− mice prevented synaptic and behavioral deficits in the double mutant mice (Auerbach et al, 2011).

A delicate balance of mTORC1 activity also has been shown to be required for proper myelination in oligodendrocytes, where both inactivation and hyperactivation of mTORC1 results in a hypomyelination phenotype (Lebrun‐Julien et al, 2014; Carson et al, 2015). Interestingly, disruption of myelination during development has been implicated in a range of neurodevelopmental disorders including TSC. Moreover, TSC patients with ASD display impaired white matter integrity compared with TSC patients without ASD (Lewis et al, 2013; Peters et al, 2013). Similarly, mice lacking neuronal TSC1/2 have a hypomyelination phenotype (Meikle et al, 2007; Lebrun‐Julien et al, 2014; Carson et al, 2015). It has been shown that the loss of functional TSC1/2 in neurons results in blockade of oligodendrocyte development and myelination, which is mediated by neuronal connective tissue growth factor (CTGF) (Ercan et al, 2017). CTGF has been shown to bind to, antagonize insulin‐like growth factor, and stimulate oligodendrocyte development and, thus, can indirectly block the activation of mTORC1 signaling in oligodendrocytes (Stritt et al, 2009). mTORC1 in oligodendrocytes promotes initiation of myelination, myelin thickness by controlling lipogenesis, and translation of myelin proteins (Lebrun‐Julien et al, 2014). In addition, CTGF is highly expressed and secreted from neurons lacking TSC1/2 and blocks the development of oligodendrocytes. Thus, genetic deletion of CTGF in TSC1‐KO neurons improves myelination. Finally, TSC2‐deficient neurons exhibit a decrease of serum response factor (SRF) protein, which is the transcriptional repressor of the CTGF gene (Stritt et al, 2009) (Fig 2). Taken together these findings indicate that in addition to its well‐known role in the regulation of protein synthesis, TSC1/2 affects the transcription machinery in the context of myelination, adding an additional layer of regulatory control during development.

Regulation of transcription and ASD

Gene expression is regulated by orchestrated mechanisms that include the action of thousands of transcription factors, epigenetic factors (Day & Sweatt, 2011; Cholewa‐Waclaw et al, 2016; Marshall & Bredy, 2016) and three‐dimensional genomic factors such as chromatin structure or nuclear organization (Medrano‐Fernández & Barco, 2016; Rajarajan et al, 2016; Watson & Tsai, 2017) that act at the transcriptional and post‐transcriptional level. Perturbations of these transcriptional control programs likely causes a broad range of neurodevelopment diseases processes, including ASD. Several mutations associated with ASD have been shown to disrupt components and functions of signaling networks involved in the control of gene transcription, such as regulation of chromatin and expression of effector genes that act on synapses or that control calcium influx into neurons (Ebert & Greenberg, 2013). Herein, we summarize findings related to three ASD‐associated genes that act as key sensors of transcription regulation and highlight their impact on translational control underlying ASD pathophysiology: methyl‐CpG‐binding protein 2 (MECP2), activity‐dependent neuroprotective protein (ADNP), and engrailed 2 (EN2).

Methyl‐CpG‐binding protein 2 (MECP2)

Rett syndrome (RTT) is a neurodevelopmental disorder often associated with ASD and characterized by compromised brain function, severe intellectual disability, language and learning disabilities, repetitive stereotyped hand movements, and developmental regression (Liyanage & Rastegar, 2014). De novo mutations in the methyl‐CpG‐binding protein 2 (MECP2) gene, a transcriptional regulator, account for more than 95% of RTT cases (Amir et al, 1999; Chahrour & Zoghbi, 2007; Neul et al, 2010). Due to the X‐linked nature of the MECP2 gene, RTT is the leading cause of intellectual disability in females, affecting ~ 1:10,000 girls worldwide, with rare cases reported in males (Shahbazian & Zoghbi, 2002; Moog et al, 2003; Liyanage & Rastegar, 2014). Either loss or duplication of MECP2 results in neurodevelopmental disorders characterized by multiple autistic behaviors (Amir et al, 1999). Loss‐of‐function mutations in MECP2 can cause RTT (Chahrour & Zoghbi, 2007), whereas gain‐of‐function mutations result in MECP2 duplication syndrome (Van Esch et al, 2005), which is observed more frequently in males and is characterized by developmental delay, motor dysfunction, epilepsy, anxiety, frequent respiratory infections, and early death (Ramocki et al, 2010). Several studies suggest an association between mutations in MECP2 and ASD (Beyer et al, 2002; Shibayama et al, 2004; Swanberg et al, 2009), although whether MECP2 mutations identified in patients with ASD interfere with normal functions of MeCP2 remained unknown.

MeCP2 belongs to the DNA methyl‐binding protein (MBP) family (Lewis et al, 1992), and it is an epigenetic regulator that acts on chromatin structure (Nan et al, 1998) with crucial functions in controlling neuronal morphology, synaptic transmission, and synaptic plasticity (Chao et al, 2007; Monteggia & Kavalali, 2009). Although, initially proposed as a transcriptional repressor (Nan et al, 1998), evidence suggests that MeCP2 acts by facilitating global transcription (Chahrour et al, 2008; Ben‐Shachar et al, 2009). Indeed, MeCP2 has recently emerged as a multifunctional protein that bridges epigenetic marks in the DNA (cytosine methylation/hydroxymethylation) with chromatin structure and the regulation of gene expression (Fasolino & Zhou, 2017). The transcriptional regulatory role of MeCP2 appears to depend on its interacting protein partners, such as CREB, cohesin, and HP1, which contribute to MeCP2 multiple functions (de Paz & Ausió, 2017): transcription activation, chromatin loop formation, and chromatin architecture modulation, respectively (Fuks et al, 2003; Chahrour et al, 2008; Lyst & Bird, 2015). Moreover, de novo phosphorylation of MeCP2 at serine 421 (S421) by a CaMKII‐dependent mechanism controls the ability of MeCP2 to regulate dendritic patterning, spine morphogenesis, and the activity‐dependent induction of Bdnf transcription (Zhou et al, 2006) (Fig 3).

Figure 3. Regulation of transcriptional and translational machinery by MeCP2.

(Left) MeCP2 facilitates global transcription by interacting with its protein partners such as transcription activators, and proteins acting on chromatin loop formation and chromatin architecture modulation. De novo phosphorylation of MeCP2 at serine 421 (S421) by a CaMKII‐dependent mechanism controls the ability of MeCP2 to regulate dendritic patterning, spine morphogenesis, and the activity‐dependent induction of Bdnf transcription. MECP2 regulates translation initiation by interacting with components of the pre‐initiation complex, which is consistent with its ability to bind mRNA, regulates ribosome biogenesis by interacting with nucleolin, which in turn regulates rRNA synthesis/ribosome biogenesis by controlling RNA polymerase I activity, and promotes the posttranscriptional processing of particular miRNAs, such as miR‐199a which positively controls mTOR signaling by targeting inhibitors for mTOR signaling (e.g. PDE4D, HIF1a, SIRT1). (Right) Hypothetical connection between MeCP2‐mediated transcriptional regulation and mTORC1 activity, highlighting the cooperative network between translational control and transcriptional co‐regulation underlying ASD pathophysiology. Both loss of function mutations or duplication of MECP2 result in dysregulation of both processes, ultimately altering ribosome biogenesis (e.g. nucleolin), neuronal plasticity and synapses (e.g. reduced expression of CAMKIIα and PSD95), and spine formation. Impaired Akt/mTORC1 signaling is associated to reduced de novo translation and posttranscriptional processing of particular miRNAs in MeCP2‐mediated ASD.

Apart from mutations in the MECP2 coding region, mutations in the 3’UTR and increased DNA methylation of the MECP2 promoter have been found in ASD patients (Shibayama et al, 2004; Loat et al, 2008; Nagarajan et al, 2008; Xu et al, 2012). Misexpression of the MeCP2 protein is thought to contribute to neuropathology by causing dysregulation of plasticity. Findings from mice carrying mutations that abolish expression of MeCP2 strongly suggest that it regulates plasticity in development and adulthood (Krishnan et al, 2017; Karaca et al, 2018). Nevertheless, how MECP2 mutations affect the plasticity that underlies learning and memory is poorly understood. It has been suggested that MeCP2 is required for long‐term memory formation and that it controls the learning‐induced transcriptional response of hippocampal neurons required for memory consolidation (Karaca et al, 2018). Furthermore, Mecp2 mutations impair auditory cortical plasticity and maternal vocal perception in adult female mice by altering parvalbumin‐expressing cortical inhibitory interneurons (PV) (Krishnan et al, 2017), which appear to be especially vulnerable to loss of MeCP2 (Ito‐Ishida et al, 2015; Morello et al, 2018). It has been suggested that MeCP2 acts in cortical PV interneurons to coordinate stimulus‐specific, experience‐dependent cortical plasticity, thereby reshaping the output of deep layer pyramidal neurons and that mutations in MECP2 disrupt cortical activity by arresting modulation of cortical PV networks (Lau et al, 2020). Consistent with these findings, MeCP2 overexpression results in atypical response properties of the auditory cortex in mice (Zhou et al, 2019a).

As mentioned above, proper regulation of protein synthesis in neurons is crucial for the organization of synaptic structure and synaptic plasticity (Richter & Klann, 2009), and aberrant regulation of the mTOR pathways have been associated with molecular defects in different neurodevelopmental disorders. Evidence of the relationship between MECP2 mutations, translational control, and ASD comes from studies on Mecp2 mutant mice (Ricciardi et al, 2011) and human embryonic stem cell model of RTT (Li et al, 2013). Notably, pre‐symptomatic Mecp2 mutant mice exhibit downregulated Akt/mTORC1 signaling and reduced protein synthesis, consistent with deficits in translation (Fig 3). In addition, MECP2 may promote translation initiation by interacting with components of the pre‐initiation complex, which is consistent with its ability to bind mRNA (Young et al, 2005). The impaired translational control in Mecp2 mutant mice is not restricted to a specific subset of transcripts but affects molecules with pivotal roles in neuronal activity and plasticity (i.e. CAMKIIα and PSD95), as well as proteins with no clear role in neuronal function (i.e. Rack1 and GAPDH) (Ricciardi et al, 2011). Consistent with these findings, MECP2 deficiency in neurons is associated with compromised protein synthesis (Li et al, 2013), suggesting that reduced translation in MECP2‐deficient neurons contributes to the ASD phenotypes in RTT. As mentioned above, reduced de novo protein synthesis in MECP2 mutant neurons is accompanied with impaired Akt/mTORC1 signaling; conversely, its activation by exogenous growth factors such as BDNF and IGF‐1 (insulin‐like growth factor 1) or by depletion of PTEN (a negative regulator of PI3K, an upstream component of the AKT/mTOR pathway) promoted protein synthesis and ameliorated ASD phenotypes in RTT mutant neurons (Li et al, 2013). Furthermore, studies on RTT human cerebellum showed that common MECP2 mutations (T158M, R255X) affect pathways that impinge on ribosome biogenesis. Ribosomal RNA (rRNA) synthesis is a rate‐limiting step for ribosome biogenesis and it has been suggested that nucleolin, a regulator of rRNA transcription and processing might be a potential MeCP2 target (Olson et al, 2018). Nucleolin regulates rRNA synthesis/ribosome biogenesis by controlling RNA polymerase I activity, a process that is also controlled by mTORC1–S6K1 signaling. Although studies on murine models reported that the 28S and 18S rRNA transcripts are reduced in murine Mecp2‐deficient neurons (Gabel et al, 2015) and that the mTOR signaling is impaired in RTT models (Ricciardi et al, 2011), human RTT cerebellum neurons exhibited elevated rRNA transcripts and mTORC1–S61K signaling, pointing toward a potential over‐activation of these processes (Olson et al, 2018). Further evidence that links MeCP2 to mTORC1 signaling comes from a study investigating MeCP2 and its role in posttranscriptional processing of particular microRNAs (miRNAs) in RTT (Tsujimura et al, 2015). miR‐199a, a MeCP2‐regulated miRNA, positively controls mTOR signaling by targeting inhibitors of mTOR activity ameliorating RTT neuronal phenotypes, whereas miR‐199a inhibition blocks MeCP2 function and its genetic deletion led to a reduction of mTOR activity resulting in RTT phenotype in mice (Tsujimura et al, 2015).

All together, these studies suggest that both RTT and MECP2 duplication syndrome are among the growing number of neurodevelopmental disorders that are linked to dysregulated mTORC1 signaling, highlighting the cooperative network between translational control and transcriptional co‐regulation underlying ASD pathophysiology (Fig 3).

Activity‐dependent neuroprotective protein (ADNP)

Activity‐dependent neuroprotective protein (ADNP) represents a leading de novo mutated gene causing a constellation of features associated with a particular form of ASD termed Helsmoortel‐Van der Aa Syndrome (HVDAS). Heterozygous mutations in ADNP, which encodes a transcription factor, on the long arm of chromosome 20 (20q13.13), account for the etiology of 0.17% of patients with ASD (Helsmoortel et al, 2014) and ADNP levels in the plasma are significantly correlated with IQ (Malishkevich et al, 2016). Individuals with ADNP mutations share features such as global developmental delay, intellectual disability, facial anomalies, behavioral, and motor disturbances, as well as congenital cardiac defects, feeding difficulties, and visual problems (Helsmoortel et al, 2014; Pascolini et al, 2018). Moreover, the evidence that ADNP is required for brain formation (Pinhasov et al, 2003) coupled with the finding that a major phenotypic outcome of Adnp‐haploinsufficency in mice leads to cognitive impairments, placed ADNP as a key transcription factor required for normal brain function (Vulih‐Shultzman et al, 2007) (Fig 4).

Figure 4. Disruption of ADNP and EN2 intracellular pathways causes dysregulation of transcriptional and translational control in ASD.

(Left) ADNP and EN2, which act as key transcription regulators, also impact translation by interacting with factors such as eIF4E. ADNP binds directly to eIF4E, suggesting an active role of ADNP in cap‐dependent translational control differentially regulating the expression of several mRNAs in age‐dependent and sex‐dependent manner, such as SlLC6A4 (serotonin transporter), and CACNA1E (calcium channels) and BECN1 (autophagy regulator). ADNP binds and regulates ZFP161 and his mouse homolog zf5, which act as a transcriptional activator of DAT, IL‐6, LIF, and a transcriptional repressor of FMR1 by binding to promoters of these genes. EN2 contains a binding site for eIF4E and triggers the rapid phosphorylation of eIF4E and eIF4E‐binding protein (4E‐BP). EN2 modulates the pleiotropic effects of IGF‐1 by altering S6K1 activation and attenuates S6K1 phosphorylation and activation via mTORC1‐dependent and mTORC1‐independent pathways downstream of PI3K. (Right) Perturbations of transcriptional and translational control results in disruption of multiple signaling networks involved in the control of gene transcription and protein synthesis. Subsequent changes in the expression of synaptic proteins and proteins that control calcium influx into neurons results in altered E/I balance in ASD. Excessive EN2 signaling alters excitation/inhibition (E/I) balance by acting on glutamatergic and GABAergic dendritic branching, which is mediated by an increase in mTORC1‐dependent protein synthesis.

Originally associated with neuroprotection and neuroglia interactions (Gozes, 2007), ADNP is a member of the SWI/SNF (SWItch/Sucrose Non‐Fermentable) chromatin remodeling complex, with predominantly nuclear localization (Mandel & Gozes, 2007; Helsmoortel et al, 2014) and controls the expression of more than 400 genes during embryonic development (Mandel & Gozes, 2007). Besides the direct interaction with the SWI/SNF complex, ADNP is involved in multiple molecular mechanisms and its transcriptional control function is also associated with myosin regulatory light chain 2 (Myl2) (Mandel & Gozes, 2007), globin (Dresner et al, 2012), and the major Alzheimer’s disease risk gene apolipoprotein E (APOE) (Mandel & Gozes, 2007) in a sex‐dependent manner (Malishkevich et al, 2016). Therefore, ADNP shows a tight association with neurodevelopmental processes, exerting a key role in regulating cognitive functions which are not limited to ASD, but extend to schizophrenia (Merenlender‐Wagner et al, 2015) and Alzheimer's disease (Malishkevich et al, 2016).

ADNP can be localized to the cytoplasm and the axon and, thus, is not confined to the cell nucleus (Gennet et al, 2008), suggesting additional, non‐transcriptional activities of ADNP. Consistent with this idea, there is evidence that ADNP plays a role in the initiation step of protein synthesis (Malishkevich et al, 2015). ADNP has been reported to bind directly to eIF4E, suggesting an active role of ADNP in cap‐dependent translational control (Fig 4). Haploinsufficient ADNP (Adnp ^+/‐) mice mimic human HVDAS in terms of altered gene expression patterns and synapse density, as well as compromised developmental, motor, and cognitive ability (Vulih‐Shultzman et al, 2007; Amram et al, 2016). In contrast, complete deficiency in ADNP results in neural tube closure defects and death during gestation in Adnp ^−/− homozygous embryos (Mandel & Gozes, 2007). Altered eIF4E expression in the hippocampus of Adnp ^+/− male mice has been reported. The levels of eIF4E show a transient increase at a young age and are subsequently decreased, suggesting changes in translational control early in development (Malishkevich et al, 2015). Reduced Adnp levels thus correlate with increased eIF4E expression in mice. Interestingly, ADNP expression is reduced in female postmortem human hippocampal tissue (Malishkevich et al, 2015), but whether this changes eIF4E expression has not been investigated. In addition to the age‐dependent changes in eIF4E expression, ADNP seems to differentially regulate the expression of several mRNAs in a sex‐ dependent manner (Amram et al, 2016). It has been suggested that differences in ADNP expression may account for the sexual divergence associated with ASD (Malishkevich et al, 2015). Adnp‐deficient genotype shows sexual dichotomy and ADNP expression is modulated during the estrous cycle in the hypothalamus (Furman et al, 2004) in mice. Adnp ^+/− mice exhibit impaired hippocampal expression of key ASD‐linked genes including the serotonin transporter (SlLC6A4), the calcium channel (voltage‐dependent calcium channel, CACNB1), the calcium channel (CACNA1E), and the autophagy regulator, BECN1 (Beclin1), in a sex‐dependent manner. Finally, the ADNP‐interaction partner, zinc finger protein 161 homolog (ZFP161) contributes to the complex interplay between transcription and translation. ADNP binds and regulates ZFP161 and its mouse homolog zf5, which act as a transcriptional activator of dopamine transporter (DAT; SLC6A3), interleukin 6 (IL‐6), leukemia inhibitory factor (LIF), and a transcriptional repressor of FMR1 by binding to promoters of these genes (Malishkevich et al, 2015; Gozes, 2018) (Fig 4), arising questions on the impact of ADNP–ZFP161–FMR1 interaction on translational control. These observations suggest that ADNP is a multitasking regulatory protein at the interface between transcriptional and translational processes that underlie sexual divergence and pathophysiology associated with ASD.

Engrailed 2 (EN2)

Several studies suggest an association between intronic polymorphisms of human Engrailed 2 (EN2) gene and non‐syndromic forms of ASD (Benayed et al, 2005, 2009; Hnoonual et al, 2016). Human EN2 maps to distal chromosome 7 (7q36.3), a chromosomal region that has been linked to ASD (Allen et al, 1997; Liu et al, 2001; Alarcón et al, 2002). Notably, in ASD subjects, several SNPs in EN2 function as a transcriptional activator, resulting in increased levels of gene expression (Courchesne et al, 2001; Bauman & Kemper, 2005; Benayed et al, 2009). In addition, variations in the EN2 protein expression and abnormal cerebellar structure have been identified in ASD patients (Courchesne et al, 2001; Bauman & Kemper, 2005). The cerebellum is one of the first structures of the human brain to differentiate, and histopathological changes in its neuronal structure, such as the loss of Purkinje cells, have been observed in a number of post‐mortem brains of ASD individuals (Bailey et al, 1998; Palmen et al, 2004; Allen, 2005). Evidence suggests that very early cerebellar developmental defects may account for a number of neurological abnormalities reported in ASD individuals (see review by Bolduc and Limperopoulos (2009)). Indeed, cerebellar abnormalities have also been described in human and mouse studies of ASD‐associated phenotypes in tuberous sclerosis (Asano et al, 2001).

EN2 belongs to the homeobox gene family and encodes a homeodomain‐containing transcription factor, which has a key role in the early development of the cerebellum. During neurodevelopment, EN2 regulates cerebellar growth and is important in the organization of the cerebellum, as well as the midbrain/hindbrain region (Sgaier et al, 2007). EN2 and its homolog EN1 have been shown to have different regulatory roles at different stages of development involving both transcriptional and translational activities (see review by Spatazza et al (2013)). Indeed, EN2 mRNA expression is downregulated during the first postnatal week in the hippocampus and in the cerebellum, a period associated with a sudden increase in synaptogenesis in rodents (Bury & Sabo, 2011). Several mouse models with alterations in EN2 expression have been generated to investigate its role in ASD pathophysiology. Interestingly, both homozygous En2 knockout (En2 KO) and En2 transgenic mice exhibit abnormal cerebellar development, such as cerebellar hypoplasia and decreased number of cerebellar neuronal cells (i.e. Purkinje cells), as well as cognitive impairments (Kuemerle et al, 2007; Tripathi et al, 2009), which are believed to be due to abnormal postnatal cerebellar development (Gharani et al, 2004). Finally, En2 KO mice exhibit ASD‐like behavioral traits such as decreased sociability, spatial learning deficits, and increased seizure susceptibility (Cheh et al, 2006; Tripathi et al, 2009; Brielmaier et al, 2012; Provenzano et al, 2014) associated with hippocampal dysregulation of neurofibromin‐dependent pathways (Provenzano et al, 2014).

Protein synthesis is crucial for spine morphology, as well as long‐term dendritic and synaptic plasticity (Klann & Dever, 2004; Sutton & Schuman, 2005; Lo & Lai, 2020), whose dysregulation has been associated with several forms of ASD (Hoeffer et al, 2012; Gkogkas et al, 2013b; Santini et al, 2013; Xu et al, 2020). Several studies reported that EN2 directly regulates translation. EN2 interacts with the translational machinery and activates cap‐dependent mRNA translation in axonal growth cones, a function that is crucial for axon turning (Brunet et al, 2005). Brunet et al, (2005) reported that secreted En2 generates an external gradient promoting growth cone turning in Xenopus. This depends on an interaction of internalized En2 with eIF4E and is blocked by protein synthesis but not transcription inhibitors. The interaction between En2 and eIF4E triggers the rapid phosphorylation of eIF4E and eIF4E‐binding protein (4E‐BP) (Brunet et al, 2005; Osborne & Borden, 2015) resulting in enhanced cap‐dependent translation (Fig 4). A mutant form of En2 lacking the putative eIF4E‐binding domain had no effect on growth cones. In addition, another study suggests that EN1 (the EN2 orthologue) increases mRNA translation in hippocampal cells, whereas EN2 is subjected to a tight regulation during development (Soltani et al, 2017).

Excessive EN2 signaling alters excitation/inhibition (E/I) balance by acting on glutamatergic and GABAergic dendritic branching, which is mediated by an increase in mTORC1‐dependent protein synthesis (Provenzano et al, 2014; Soltani et al, 2017) (Fig 4). Transcriptome analysis identified over 800 genes differentially expressed in the cerebellum and hippocampus of En2 KO mice and showed a significant convergence of neurobiological pathways previously linked to ASD pathology. Among the differentially expressed genes, Grm5, Nrxn3, and Scn1a, which encode for mGluR5, a neuronal adhesion protein of the Neurexin (NRXN) family, and the voltage‐gated sodium channel alpha subunit (SCN1A), respectively, are of particular interest for ASD (Sgadò et al, 2013). Finally, it was demonstrated that there is a functional interaction between EN2 expression and IGF1 signaling (Rossman et al, 2014), whose reduction in cerebrospinal fluid (CSF) of ASD patients has been reported (Riikonen et al, 2006), suggesting that EN2 modulates the pleiotropic effects of IGF‐1 by altering S6K1 activation in vitro (Rossman et al, 2014). It has been speculated that reduced EN2 expression increases S6K1 phosphorylation and activation via mTORC1‐dependent and mTORC1‐independent pathways downstream of PI3K, whereas in physiological conditions, EN2 attenuates mTORC1‐mediated proliferation (Rossman et al, 2014). Translational signaling pathways, including phosphorylation of eIF4E and the PI3K/Akt/mTORC1 and ERK (extracellular signal‐regulated kinase) signaling cascades, represent a common feature in various ASD‐associated gene alterations (Ebert & Greenberg, 2013), and it is intriguing that deletion of EN2, a transcription factor, shows dysregulated Akt‐mTORC1‐S6K1 signaling, which may account for increased mRNA translation and spine formation (Fig 4).

Conclusion and future outlook

The evidence highlighted in this review suggests that there is a complex level of dynamic regulation between translation and transcription that likely contribute to ASD pathophysiology. It should be noted that the interplay between translation and transcription is presented here as positive feedback loop that exacerbates ASD pathophysiology, but it is possible that this interplay could be compensatory and protective, which should be investigated in the future. We propose that there is bidirectional regulation of translation and transcription in ASD and that mutations in genes that result in dysregulation of one of those processes triggers a positive feedback loop to act on the other, which exacerbates ASD pathophysiology. This raises the possibility that rebalancing the system by targeting key players in this mutual regulatory loop could restore protein homeostasis, and ultimately, normal neuronal function. From this perspective, isolating and interpreting the changes in mRNA translation that affect neuronal physiology in terms of pathways, synaptic properties, and epigenetic outcomes represent a challenging but plausible route toward developing effective treatment in ASD. Moreover, it provides a mechanistic framework for the investigation of the synergistic contributions of different ASD‐associated genes and pathways underlying the development of ASD phenotypes.

Conflict of interest

The authors declare that they have no conflict of interest.

EMBO reports (2021) 22: e52110.