The role of altered translation in intellectual disability and epilepsy

Abstract

A very high proportion of cases of intellectual disability are genetic in origin and are associated with the occurrence of epileptic seizures during childhood. These two disorders together effect more than 5% of the world’s population. One feature linking the two diseases is that learning and memory require the synthesis of new synaptic components and ion channels, while maintenance of overall excitability also requires synthesis of similar proteins in response to altered neuronal stimulation. Many of these disorders result from mutations in proteins that regulate mRNA processing, translation initiation, translation elongation, mRNA stability or upstream translation modulators. One theme that emerges on reviewing this field is that mutations in proteins that regulate changes in translation following neuronal stimulation are more likely to result in epilepsy with intellectual disability than general translation regulators with no known role in activity-dependent changes. This is consistent with the notion that activity-dependent translation in neurons differs from that in other cells types in that the changes in local cellular composition, morphology and connectivity that occur generally in response to stimuli are directly coupled to local synaptic activity and persist for months or years after the original stimulus.

1. Introduction:

Intellectual disability (ID) is a collection of diverse neurodevelopmental disorders united by impaired intellectual abilities (van Bokhoven, 2011). There are three clinical criteria for ID: an intelligence quotient below 70; impairment in two or more adaptive behaviors, including communication, self-care, social skills, and self-direction; and evidence that the neurological symptoms began prior to the age of 18 (American Psychiatric Association, 2013). ID affects an estimated 2–3% of the population, about 85% of whom have a mild form. While the cause of ID can be environmental, an enormous number of diverse genetic causes have been identified. These include chromosomal aneusomies and other chromosomal abnormalities, as well as mutation or deletion of a diverse array of individual genes. More than 450 individual genes have been implicated in ID (van Bokhoven, 2011). In total, an estimated 20% of mild cases and 65% of more severe cases of ID can be explained by genetic factors (Rauch et al., 2006).

Epilepsy is a common neurological disorder in which enhanced, synchronized neuronal firing in the brain leads to recurring spontaneous seizures (Myers and Mefford, 2015). Epilepsy is estimated to affect about 4% of the population. While about 20–30% of epilepsy cases are the result of a specific insult such as stroke, tumor, or brain trauma, the remainder are thought to be of genetic origin (Hildebrand et al., 2013). As with ID, there is a diverse array of clinical phenotypes and syndromes among epilepsy patients, which fall into the categories of genetic generalized epilepsy (GGE), focal epilepsy, and epileptic encephalopathy (EE). The GGE syndromes, which tend have the least severe comorbidities, are characterized by generalized seizures that involve both sides of the brain. At the other extreme, the EEs are often early onset, characterized by severe refractory seizures, and tend to occur with significant comorbid neurological disorders including ID (Myers and Mefford, 2015).

Epilepsy and ID are mutually comorbid in a high number of patients. Epilepsy occurs in about in about 20% of patients with ID. Conversely, ID occurs in about 25% of patients with epilepsy. Despite this high degree of mutual comorbidity, research regarding the intersection of these two diseases is underrepresented in the publications of both fields (Shankar et al., 2018). Both ID and epilepsy also frequently occur with Autism Spectrum Disorder (ASD). ASD is the most common comorbid phenotype in patients with ID, occurring in about 40% of patients (La Malfa et al., 2004). Estimates for ASD in epilepsy patients range from 6% to 27% (Jeste and Tuchman, 2015).

Given the high rate of combined occurrence of ID with epilepsy, and also with ASD, there are likely to be common pathways that, when perturbed by genetic mutations, result in these conditions. One such pathway is the translation of mRNAs. ID- and epilepsy-causing mutations occur at every step in the metabolism and translation of mRNA, including mRNA processing, initiation, elongation, tRNA modification, mRNA stability/degradation, and upstream regulation. These are listed in Table 1 with the mutations sorted according to the processing step (In Table 1 and Figures 3–6: Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause ID or epilepsy, but sometimes cause the other are colored purple or orange respectively). The prevalence of these genetic disorders varies enormously from relatively common diseases like Fragile X Syndrome (FXS), to diseases where only a handful of patients with genetic variants have been identified. In addition, mutations in other regulators of translation have been identified in other neurological and neurodegenerative diseases (Kapur et al., 2017, Kapur and Ackerman, 2018). Perhaps not surprisingly, these genetic diseases reveal the essential nature of translation in neurological function and show that any dysregulation of translation can have severe consequences and lead to a diverse set of clinical manifestations. Nevertheless, why there is such a high prevalence of translation-related proteins in neurological disorders is not understood. This review will explore one possible explanation: the unique role that activity-dependent translation plays in neurons.

Table 1:

Translation Regulators Implicated in Epilepsy and Intellectual Disability

	Protein	Epilepsy	Intellectual Disability	Translation Step	Secondary Translation Step	Notable Comorbidities	References
mRNA Processing
	RBFOX1/3	Yes	No	Alternative Splicing	mRNA Stability		Lal et al., 2013a, Lal et al., 2013b, Lal et al., 2015
	DHX30	Few	Yes	RNA Helicase		ASD	Lessel et al., 2017
	DDX3X	Some	Yes	RNA Helicase		Microcephaly	Lennox et al., 2020
Initiation
	rpL10	Some	Yes	Initiation		ASD, Microcephaly	Thevenon et al., 2015, Zanni et al., 2015, Brooks et al., 2014, Bourque et al., 2018
Elongation: Elongation Factors
	eEF1A2	Yes	Yes	Elongation		ASD, Microcephaly	Kaur et al., 2019, Cao et al., 2017, Lopes et al., 2016, McLachlan et al., 2019, Lam et al., 2016, Nakajima et al., 2015, Veeramah et al., 2013, Inui et al., 2016, De Rinaldis et al., 2020
	eEF1B2	Some	Yes	Elongation			Lopes et al., 2016, McLachlan et al., 2019, Larcher et al., 2020, Najmabadi et al., 2011
	eEF1D	No	Yes	Elongation			McLachlan et al., 2019
	DPH1	Some	Yes	Elongation			Loucks et al., 2015, Sekiguchi et al., 2018, Urreizti et al., 2020, Alazami et al., 2015
Elongation: tRNA
	AARS	Yes	Yes	tRNA Charging		Microcephaly, Myelination Defects	Nakayama et al., 2017, Simons et al., 2015
	FARS2	Yes	Yes	tRNA Charging		Mitochondrial Defects	Almalki et al., 2014
	PARS2	Yes	Yes	tRNA Charging		Microcephaly, Myelination Defects	Yin et al., 2018
	ADAT3	Few	Yes	tRNA Processing		Microcephaly	Alazami et al., 2013, Ramos et al., 2019, El-Hattab et al., 2016, Salehi Chaleshtori et al., 2018
	ALKBH8	Yes	Yes	tRNA Processing			Monies et al., 2019
	PUS3	Yes	Yes	tRNA Processing		Microcephaly	Shaheen et al., 2016, de Paiva et al., 2019, Abdelrahman et al., 2018
	PUS7	No	Yes	tRNA Processing		Microcephaly	Shaheen et al., 2019, Darvish et al., 2019, de Brouwer et al., 2018
	NSUN2	Few	Yes	tRNA Processing	mRNA Stability	Microcephaly	Martinez et al., 2012, Komara et al., 2015, Abbasi-Moheb et al., 2012, Khan et al., 2012
	FTSJ1	No	Yes	tRNA Processing			Guy et al., 2015, Takano et al., 2008, Freude et al., 2004
	TRMT1	Yes	Yes	tRNA Processing			Najmabadi et al., 2011, Davarniya et al., 2015, Zhang et al., 2020
mRNA Stability
	Pumilio1/2	Yes	Yes	mRNA Stability	Initiation	Microcephaly, Ataxia	Bonnemason-Carrere et al., 2019, Gennarino et al., 2018, Wu et al., 2015
	DDX6	No	Yes	mRNA Stability	Initiation		Balak et al., 2019
mRNA Stability: Nonsense Mediated Decay
	UPF3B	Some	Yes	Nonsense Mediated Decay	Termination		Xu et al., 2013, Tejada et al., 2019, Tzschach et al., 2015, Laumonnier et al., 2010
Multi-Level Regulators: FMRP
	FMRP	Some	Yes	Multiple Levels		ASD	Richter et al., 2015, De Rubeis and Bagni, 2010
	PQBP1	No	Yes	Elongation	FMRP-binding	Microcephaly	Kalscheuer et al., 2003, Rejeb et al., 2011
	CYFIP1/2	Yes	Yes	Initiation	FMRP-binding		Zweier et al., 2019, Zhong et al., 2019, Nakashima et al., 2018, Peng et al., 2018, Butler, 2017, Huang, 2016
Upstream Regulators
	TSC1/2	Yes	Some	Upstream Regulation	mTOR Signaling	Brain Lesions, ASD, Renal/Cardiac Symptoms	Curatolo and Bombardieri, 2008
	PTEN	Yes	No	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018
	PIK3CA	Yes	No	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018
	AKT3	Yes	Some	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018, Conti et al., 2015
	DEPDC5	Yes	Some	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018, Scheffer et al., 2014, Dibbens et al., 2013, Iffland et al., 2019, Picard et al., 2014, Ishida et al., 2013
	NPRL2	Yes	No	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018, Iffland et al., 2019
	NPRL3	Yes	Some	Upstream Regulation	mTOR Signaling		Marsan and Baulac, 2018, Iffland et al., 2019
	mTOR	Yes	Some	Upstream Regulation	mTOR Signaling		Mirzaa et al., 2016, Marsan and Baulac, 2018, Lim et al., 2015, Gordo et al., 2018, Epi et al., 2013
	SynGAP	Yes	Yes	Upstream Regulation		ASD, Schizophrenia	Hamdan et al., 2009, Hamdan et al., 2011, Rauch et al., 2012, Vlaskamp et al., 2019

Primarily Epilepsy

Primarily Epilepsy, Some ID

Epilepsy and ID

Primarily ID, Some Epilepsy

Primarily ID

Figure 3: ID- and epilepsy-linked genes with roles in translation elongation.

eEF1A is responsible for delivering aminoacylated tRNA to the ribosome. eEF1A2 is a neuron and muscle specific isoform of this protein. Release of the tRNA results in GTP hydrolysis release of GDP-bound eEF1A2 from the ribosome (Dever and Green, 2012, Kaur et al., 2019). Members of the eEF1 complex actively exchange this GDP for a GTP, allowing eEF1A2 to perform another cycle of tRNA delivery (McLachlan et al., 2019). Translocation of tRNA from the A-site to the P-site, requires the action of eEF2 containing a unique diphthamide modification. Among other proteins, DPH1 is required for diphthamide synthesis (Liu et al., 2012). Proteins linked to ID and epilepsy are colored as shown in the legend. Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause one condition, but sometimes cause the other are colored purple or orange.

Figure 6: FMRP and binding partners regulate mRNA translation through various mechanisms.

A. Phosphorylated FMRP inhibits mRNA translation by stalling elongation of mRNA bound ribosomes (Narayanan et al., 2007, Niere et al., 2012, Nalavadi et al., 2012). B. Phosphorylated FMRP recruits miRNA and to mRNA. The CCR4-NOT deadenylase complex is then recruited leading to microRNA-mediated degradation of the mRNA (Muddashetty et al., 2011). C. FMRP and CYFIP1 form a translation initiation inhibitory complex with CYFIP1 acting as a non-canonical 4E-BP (Napoli et al., 2008). D. PQBP1 recruits FMRP and its target mRNAs to ribosomes, leading to enhanced translation initiation of particular mRNAs (Wan et al., 2015). E. FMRP binds to the C-terminus of the Slack Na⁺ -activated K⁺ channel, leading to channel activation (Brown et al., 2010, Zhang et al., 2012). Proteins linked to ID and epilepsy are colored as shown in the legend. Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause one condition, but sometimes cause the other are colored purple or orange.

Activity-dependent translation in neurons differs from regulation of translation in other cells types in that it does not simply represent a reversible homeostatic response of signaling cascades to changes in the environment. Instead, it is coupled to changes in local cellular composition, morphology and connectivity that are long-term or permanent even months or years after the original stimulus. Two well-studied examples are the long-term potentiation (LTP) and long-term depression (LTD) of synaptic connections, which represent increased or decreased neuronal excitability in response to specific patterns of synaptic activity, respectively (Younts et al., 2016, Baltaci et al., 2019, Ebert and Greenberg, 2013, Waung and Huber, 2009). In both, a specific pattern of neuronal stimulation leads to a signaling cascade that, along with other more short-term changes, increases local translation. For example during LTD, synaptic proteins such as the AMPA subunit GluR2 are synthesized and incorporated directly into the synapse (Waung and Huber, 2009). This in turn alters neuronal excitability by creating a feedback loop that further regulates both local translation and synaptic excitability (Younts et al., 2016, Baltaci et al., 2019, Ebert and Greenberg, 2013, Waung and Huber, 2009). This feedback loop of localized activity-dependent translation, which is critical both for synaptogenesis and for learning and memory, helps to explain why mutations in components of translation affect neurological function so profoundly.

Mutations in genes that encode regulators of translation can affect activity-dependent translation in several distinct ways. Depending on the normal function of the encoded protein, a loss-of-function mutation can alter the translation of mRNA targets by causing; i) a constitutive increase or decrease in translation, ii) diminished activity-dependent translation, or iii) enhanced activity-dependent translation. These possibilities are schematized in Figure 1, which also provides examples of genes that when mutated result in each type of outcome. Generally, each of these leads to either diminished or enhanced LTP or LTD but can also alter other aspect of neuronal plasticity including constitutive changes in neuronal excitability and sensitivity to sensory inputs. For the purposes of this review, a protein involved in activity-dependent translation will be defined as any regulator of translation that has a demonstrated effect on an activity-dependent processes such as LTP or LTD, or a translation regulator whose function is known to be altered by neuronal stimulation.

Figure 1: Disorders of translation.

Schematic of different ways in which gene mutations can affect translation of mRNAs, including diminished or enhanced activity-dependent translation and constitutive increases in translation. Examples of gene that when mutated result in each type of dysfunction are provided at right.

This review will explore the series of steps in the translation of mRNA and how genetic perturbations of these pathways can lead to both ID and epilepsy. One theme that emerges is that the mutant genes encoding proteins with key roles in activity-dependent translation are much more likely to result in epilepsy coupled to ID than those that play no role in activity-dependent aspects of translation. Consistent with this, most translation regulators with mutations primarily causing epilepsy, play a role in activity-dependent translation (see Table 2). The exception to this rule is the diverse set of proteins that regulate tRNA processing and charging, although it is possible that this set of proteins is connected to activity-dependent translation by a currently unknown mechanism. In contrast, many of the translation regulators with mutations causing ID are not known to play any role in activity-dependent translation.

Table 2:

Activity-Dependent Translation is Correlated with Epilepsy

Protein	Documented role in activity-dependent translation?	Epilepsy	ID	Effect of loss of function on activity-dependent translation	Functional effect of loss of function	References
Documented role in activity-dependent translation
AKT3	Yes	Yes	Yes	Diminished	Diminished LTP	Zhang et al., 2019a
CYFIP1/2	Yes	Yes	Yes	Constitutive	Enhanced LTD	Bozdagi et al., 2012
eEF1A2	Yes	Yes	Yes	Enhanced or Diminished	Enhanced LTD, Diminished LTP	Mendoza et al., 2020, Tsokas et al., 2005
Pumilio1/2	Yes	Yes	Yes	Constitutive	Constitutive LTP	Dong et al., 2018, Driscoll et al., 2013
SynGAP	Yes	Yes	Yes	Constitutive	Enhanced LTD, Diminished LTP	Paul et al., 2019, Barnes et al., 2015
TSC1/2	Yes	Yes	Some	Constitutive	Diminished LTD, Diminished LTP	Auerbach et al., 2011, Switon et al., 2017, von der Brelie et al., 2006
mTOR	Yes	Yes	Some	Diminished	Diminished LTD, Diminished LTP	Jaworski and Sheng, 2006
PIK3CA	Yes	Yes	No	Diminished	Diminished LTP	Saw et al., 2020
PTEN	Yes	Yes	No	Constitutive	Diminished LTD, Diminished/Enhanced LTP	Takeuchi et al., 2013, Fraser et al., 2008
RBFOX1/3	Yes	Yes	No	Enhanced	Enhanced LTP	Tomassoni-Ardori et al., 2019
FMRP	Yes	Some	Yes	Constitutive	Enhanced LTD	Richter et al., 2015, De Rubeis and Bagni, 2010, Richter and Zhao, 2021
UPF3B	Yes	Some	Yes	Constitutive	Diminished activity-dependent differentiation	Jolly et al., 2013
Predicted role in activity-dependent translation
DEPDC5	Predicted	Yes	Some	Constitutive	Unknown	Panchaud et al., 2013
NPRL3	Predicted	Yes	Some	Constitutive	Unknown	Panchaud et al., 2013
NPRL2	Predicted	Yes	No	Constitutive	Unknown	Panchaud et al., 2013
PQBP1	Predicted	No	Yes	Constitutive	Unknown	Wan et al., 2015
No known role in activity-dependent translation
ALKBH8	No	Yes	Yes	None	None
AARS	No	Yes	Yes	None	None
FARS2	No	Yes	Yes	None	None
PARS2	No	Yes	Yes	None	None
PUS3	No	Yes	Yes	None	None
TRMT1	No	Yes	Yes	None	None
AARS	No	Yes	No	None	None
FARS2	No	Yes	No	None	None

2. Overview of Translation:

Figure 2 illustrates each of the stages of translation that will be covered in this review. After transcription, RNAs that encode proteins undergo several processing steps before they become mature mRNA. These include the addition of a 5’ 7-methylguanosine cap and a 3’ poly-adenosine (poly-A) tail. Both modifications regulate mRNA degradation and participate, directly or indirectly, in the initiation of translation (Kapur et al., 2017). Another processing step for the pre-mRNA for many genes is splicing. Spliceosomes remove introns from the mRNA while retaining exons. Alternative splicing creates different combinations of exons in the mature mRNA, leading to multiple isoforms of the resultant protein (Reviewed in Lee and Rio, 2015). Translation itself is a cyclical process with three distinct phases: initiation, elongation, and termination. Initiation is the most complex of these phases, during which a series of eukaryotic initiation factors (eIFs) aid in the assembly of the functional 80S ribosome on mRNA (Reviewed in Jackson et al., 2010, Sonenberg and Hinnebusch, 2009). During this step, the 5’ and 3’ ends of the mRNA are looped into a circular structure, a process mediated by Poly(A) Binding Protein (PABP) which binds eIFs attached to the 5’ end.

Figure 2: Overview of eukaryotic translation.

Outline of various steps in translation that have a corresponding section in the text. Newly transcribed pre-mRNA are processed by addition of a 5’ 7-methylguanosine cap and a 3’ poly-A tail (Lee and Rio, 2015). In addition, spliceosomes remove introns from the mRNA while retaining exons (Lee and Rio, 2015). Translation itself is a cyclical process with three distinct phases: initiation, elongation, and termination. During initiation a series of eIFs aid in the assembly of the functional 80S ribosome on mRNA (Jackson et al., 2010, Sonenberg and Hinnebusch, 2009). During elongation, eEFs mediate the recruitment of charged tRNA to the ribosome, the addition of amino acids to the nascent peptide, and the resetting of the ribosome for a new elongation cycle (Dever and Green, 2012). Termination of translation is mediated by eukaryotic release factors (eRFs) and occurs on reaching a stop codon in the mRNA (Dever and Green, 2012). The stability of mRNAs is controlled by maintenance of the poly-A tail, recruitment of miRNAs, and nonsense mediated decay (NMD) (Iwakawa and Tomari, 2015, Kishor et al., 2019). Various steps of translation are regulated by both multi-level and upstream regulators.

The central step of translation is the cycle of elongation, which is mediated through eukaryotic elongation factors (eEFs). Charged tRNA enters the aminoacyl (A) site of the ribosome. If the anticodon of the tRNA matches the mRNA codon at the A site, a peptide bond is formed between the nascent peptide and the amino acid bound to the tRNA in the A site, and the mRNA shifts by one codon. Finally, the deacylated tRNA is released from the E site and the cycle repeats (Reviewed in Dever and Green, 2012).

Termination of translation is mediated by eukaryotic release factors (eRFs) and occurs on reaching a stop codon in the mRNA (Reviewed in Dever and Green, 2012). Even after the release of the ribosome, PABP is believed to keep the mRNA circularized. This produces more efficient translation by increasing the rate of re-initiation and restarting the translation cycle (Jackson et al., 2010).

A final process that controls translation is the regulation of mRNA stability. This is determined by several distinct pathways, including maintenance of the poly-A tail, recruitment of miRNAs, and nonsense mediated decay (NMD) (Reviewed in Iwakawa and Tomari, 2015, Kishor et al., 2019). Each of these processes has been reviewed extensively and detailed information can be found as cited above. In addition, the steps of translation described above are regulated by both multi-level and upstream regulators.

2.1. Activity-Dependent Translation

Activity-dependent translational control in neurons is analogous to signaling cascades that occur in all cells, but specifically links the activation of ion channels produced by neuronal activity to changes in translation. This translational regulation often occurs through post-translational modification of translation machinery by kinases or phosphatases, thereby altering rates of translation (Choe and Cho, 2020). For example, dephosphorylation of eIF2α during LTP removes translational repression, a process known as disinhibition, leading to increased general translation (Costa-Mattioli et al., 2009). Alternatively, phosphorylation of eEF2 during LTD leads to inhibition of general translation but increased translation of a subset of mRNAs (Park et al., 2008). Such regulation of translation by post-translation modifications allows rapid changes in protein expression that can alter neuronal excitability. This, in turn, may create a feedback loop that results in long-term changes to both translation and synaptic excitability (Younts et al., 2016, Baltaci et al., 2019, Ebert and Greenberg, 2013, Waung and Huber, 2009).

3. The Role of Translation in Intellectual Disability and Epilepsy:

3.1. mRNA Processing

3.1.1. Alternative Splicing

Mutations that alter the splicing of pre-mRNAs can result in epilepsy. The disease-causing genes identified thus far do not encode components of the spliceosome itself but encode regulatory proteins that that direct which exons will be included or excluded from mature mRNA. These include RNA Binding Fox-1 Homolog (RBFOX) 1 and its homolog RBFOX3, both of which regulate exon inclusion by binding to specific motifs in the introns of pre-mRNA. As is the case for many translation regulators, RBFOX1 also has a secondary function to regulate the stability of mRNAs. Specifically, the presence of RBFOX1 binding motifs in the 3’ untranslated region (UTR) of mRNAs increase the stability of these transcripts (Ray et al., 2013). This regulation of mRNA stability is performed by cytoplasmic RBFOX1, whereas alternative splicing is regulated by nuclear RBFOX1. A particularly high proportions of mRNAs for genes linked to cortical development and autism have been found to be targets of RBFOX1 (Lee et al., 2016). RBFOX1 and RBFOX3 mutations have been implicated as causative or contributing factors in many types of epilepsy, including GGE, childhood benign epilepsy, and Rolandic epilepsy (Lal et al., 2013a, Lal et al., 2013b, Lal et al., 2015). Very few patients with RBFOX1 and RBFOX3 mutations, however, develop ID, although ASD has been reported in some patients (Lal et al., 2013b). In addition, the targets of mRNA stability regulation

Although neuronal stimulation can affect the choice of which splice variants come to be expressed (Liu and Kaczmarek, 1998), such changes are not usually considered a general feature of normal activity-dependent translation and typically occur only at specific stages of development. Clearly, however, a change in splice variants can alter the way a neuron responds to stimulation. For example, overexpression of RBFOX1 alters the pattern of splicing of the mRNA for Brain-Derived Neurotrophic Factor (BDNF) receptor TrkB, leading to reduced BDNF-dependent LTP (Tomassoni-Ardori et al., 2019).

3.1.2. RNA Helicases

Another mRNA processing step is the unwinding of RNA secondary structures by RNA helicases. The family of DExH-box ATP-dependent RNA helicases (DHX) unwinds RNA secondary structures at a variety of steps during mRNA metabolism including transcription, transport, and translation. One member of this family, DHX30, participates in mRNA unwinding during translation itself. Several de novo mutations in DHX30 impair mRNA binding or ATP-hydrolysis, and lead to reduced global translation associated with the formation of stress granules. Patients with these mutations have ID (Lessel et al., 2017). Although DHX30 is also thought to play a role in mitochondrial ribosome assembly, normal mitochondrial function was observed in these patients, suggesting that it is the inhibition of global translation that leads to ID. Like the RBFOX proteins, DHX30 is not predicted to serve a role in activity-dependent translation. Consistent with the trend described throughout this review, only a small number of patients with DHX30 mutations develop epilepsy (Lessel et al., 2017). Comorbidities such as ASD, however, occurred in some patients.

Another RNA helicase implicated in ID is DEAD-box (DDX) 3X. DDX3X plays a role in mRNA metabolism and translation by regulating mRNAs with long or complex 5’ UTRs and acting as an eIF4E inhibitory protein (Shih et al., 2008, Lai et al., 2008). Mutations in DDX3X have been shown to disrupt its RNA helicase activity, leading to the formation of RNA granules and altered translation (Lennox et al., 2020). These pathogenic mutations cause ID with some patients also developing epilepsy, ASD, and microcephaly (Lennox et al., 2020). Although, there is evidence DDX3X may play a role in cell-stress responses, there is no documented role in activity-dependent translation (Lai et al., 2008).

3.2. Initiation

While initiation is one of the most intricate steps in translation, few of the primary initiation factors have been implicated in neurological diseases, possibly because of the essential role that these proteins play in every cell in the body. One exception is eIF2, which, when mutated, results in neurodegeneration rather than neurodevelopmental disorders (Kapur et al., 2017). Nevertheless, one ribosomal protein (rp), rpL10, has been implicated in both ID and epilepsy. rpL10 participates in the biogenesis of the 60S ribosomal subunit as well as the joining of the two subunits on the mRNA. It binds directly to rpS6 and to the nuclear export factor Nmd3P to form the interface between the ribosomal subunits (Pachler et al., 2004). Mutations in rpL10 have been identified in patients with a diverse array of clinical manifestations. The shared phenotype among most of these patients is, however, X-linked syndromic ID, while the most common comorbidities are epilepsy and microcephaly (Thevenon et al., 2015, Zanni et al., 2015, Brooks et al., 2014, Bourque et al., 2018). Other patients develop ASD rather than ID, although rpL10 mutations may only be a contributing factor to this phenotype (Chiocchetti et al., 2011, Klauck et al., 2006). The disease-causing rpL10 mutations can be either gain- or loss-of-function, leading either to increased or decreased numbers of actively translating polysomes respectively, likely because the mutations enhance or suppress the joining of the ribosomal subunits (Zanni et al., 2015). It is not yet known, however, whether rpL10 plays any role in activity-dependent translation.

Mutations in several other proteins that influence the initiation of translation result in ID and epilepsy (e.g. CYFIP1, DDX6, Pumilio). These proteins function as inhibitory eIF4E binding proteins (4E-BPs) or otherwise participate in protein complexes with eIF4E to regulate translation. An interaction between eIF4E and eIF4G is required for the normal recruitment of mRNA to the ribosome during cap-dependent translation (Richter and Sonenberg, 2005, Sonenberg and Hinnebusch, 2009). 4E-BPs modulate this interaction to inhibit translation in various cell types, including neurons (Richter and Sonenberg, 2005). 4E-BPs interact with eIF4 to prevent its binding to either the 5’ cap of mRNA or eIF4G. This interaction can be untethered from mRNA to inhibit general translation or can be tethered to specific mRNAs (as is the case for CYFIP1) to selectively inhibit translation (Richter and Sonenberg, 2005, Napoli et al., 2008, De Rubeis et al., 2013, Panja et al., 2014). Phosphorylation of the 4E-BPs causes the release of eIF4E and restored translation (Richter and Sonenberg, 2005). Other proteins that form complexes with eIF4E (like DDX6 and Pumilio) can recruit the CCR4-NOT deadenylase complex, leading to microRNA-mediated degradation of an mRNA (Nishimura et al., 2015, Wang et al., 2015, Kamenska et al., 2016, Chauderlier et al., 2018). Finally, some proteins (like DDX6) that interact with eIF4E can enhance translation by promoting the recruitment of eIF4E to the mRNA (Wang et al., 2015). The role of these proteins in neurological disease will be discussed later in the context of their other regulatory functions.

3.3. Elongation

3.3.1. Elongation Factors

Two classes of proteins involved in translation elongation have been implicated in ID and epilepsy. The first class are elongation factors and their modifiers. These proteins include various components of the eEF1 complex, and Diphthamide Biosynthesis 1 (DPH1), which is responsible for a unique post-transcriptional modification on eEF2. The second class is the aminoacyl-tRNA synthetases and modifiers that will be discussed in the next section.

The eEF1 complex is responsible for delivering aminoacylated tRNAs to the ribosome, and the central functional component of this complex is eEF1A (Dever and Green, 2012, Kaur et al., 2019) (Figure 3). The eEF1A2 isoform illustrated in Figure 3 is the less prevalent of two independently-coded isoforms but is selectively expressed in neurons and muscles after development. Mutations in eEF1A2 have been shown to cause ID and epilepsy in many patients, with some patients showing additional comorbidities including ASD and microcephaly (Kaur et al., 2019, Cao et al., 2017, Lopes et al., 2016, McLachlan et al., 2019, Lam et al., 2016, Nakajima et al., 2015, Veeramah et al., 2013, Inui et al., 2016, De Rinaldis et al., 2020). These include both homozygous loss-of-function mutations and de novo heterozygous mutations, which mouse models suggest cause gain-of-function (Nakajima et al., 2015, Veeramah et al., 2013, Inui et al., 2016, Davies et al., 2017, Davies et al., 2020).

Two other components of the eEF1 complex are eEF1B2 and eEF1D, which are both guanine exchange factors (McLachlan et al., 2019) (Figure 3). Mutations in eEF1B2 primarily cause ID, although seizures have been observed in some patients (Lopes et al., 2016, McLachlan et al., 2019, Larcher et al., 2020, Najmabadi et al., 2011). There is also some evidence from large-scale screens that mutations of eEF1D also contribute to ID (McLachlan et al., 2019).

The eEF1 complex has been shown to play a role in activity-dependent translation, and components of this complex are subject to regulation by protein kinases (Sasikumar et al., 2012). Activation of metabotropic glutamate receptors during LTD leads to a phosphorylation-dependent dissociation of eEF1A2 from the remaining members of the complex, preventing the recycling of eEF1A2 bound GDP back to GTP (Mendoza et al., 2020). In addition, the translation of eEF1A has been shown to be rapidly upregulated by mTOR signaling during LTP (Tsokas et al., 2005). Interestingly, these two effects regulate translation in opposing directions in LTD and LTP, suggesting a complex role for eEF1A in regulating activity-dependent translation. In addition to being the member of the eEF1 complex most intimately related to activity-dependent translation, eEF1A2 is the only member whose mutation almost always results in epilepsy, supporting a link between dysregulated activity-dependent translation and epilepsy.

The eEF2 protein is responsible for the translocation step of translation. It acts as a sensor of neuronal activity and plays a downstream role in the regulation of translation during synaptic plasticity. Depending its characteristics, neuronal activity can lead to either an increase or decrease in eEF2 phosphorylation, resulting in a decrease or increase in eEF2 activity and translation, respectively (McCamphill et al., 2015). eEF2 also plays a role in coupling very local synaptic activity in dendrites with local changes in translation (Sutton et al., 2007). eEF2 is modified post-translationally by a unique highly conserved modification of a histidine residue to diphthamide. This modification is carried out the DPH1 enzyme (Figure 3). In mice, genetic knockout of DPH1 results in reduced elongation efficiency, as well as in errors of translation caused by inappropriate frame shifting during translocation (Liu et al., 2012). In humans, homozygous DPH1 mutations cause syndromic ID, and about half of these patients also experience epilepsy. No other comorbidities are consistently reported (Loucks et al., 2015, Sekiguchi et al., 2018, Urreizti et al., 2020, Alazami et al., 2015). Factors that influence the activity of DPH1 or whether it plays a role in activity-dependent translation in neurons is not known.

3.3.2. tRNA charging

Another key step in elongation is the shuttling to the ribosome of the individual amino acids linked to the tRNAs. The charging of the tRNAs with their cognate amino acids is carried out by aminoacyl-tRNA synthetases (aaRS). Unique aaRS are responsible for the charging of each type of amino acid to its corresponding tRNA. Before this charging takes place, however, tRNAs undergo post-transcriptional modifications catalyzed by a variety of proteins. Given the essential role of tRNA function in translation, it is not surprising that mutations in both aaRS proteins and the tRNA modifying proteins result in a large number of disease states with diverse neurological manifestations (Kapur et al., 2017, Kapur and Ackerman, 2018) (Figure 4).

Figure 4: Altered tRNA modifications and charging cause ID and epilepsy.

The shuttling of amino acids to the ribosome by charged tRNA is a key step in elongation. Before charging, tRNA undergo numerous post-transcriptional modifications by a large array of proteins. After modification, they are charged with amino acids by aaRS. Unique aaRS are responsible for the charging of each type of amino acid to its corresponding tRNA (Kapur et al., 2017, Kapur and Ackerman, 2018). Proteins linked to ID and epilepsy and the residues these proteins modify are colored as shown in the legend. Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause one condition, but sometimes cause the other are colored purple or orange.

The activity of several aaRS have been linked to epilepsy and ID. These include Alanyl-tRNA Synthetase (AARS), Prolyl-tRNA Synthetase 2 (PARS2), and Phenylalanine-tRNA Synthesis 2 (FARS2), and mutations in each of these three genes produce epilepsy with ID (Nakayama et al., 2017, Simons et al., 2015, Almalki et al., 2014, Yin et al., 2018, Almannai et al., 1993). Mutations in AARS and PARS2 are also associated with microcephaly and defects in myelination, while those in FARS2 result mitochondrial defects (Nakayama et al., 2017, Simons et al., 2015, Almalki et al., 2014, Yin et al., 2018). In addition to binding their cognate amino acid, some aaRS have an additional editing function, hydrolyzing the aminoacyl bond if an incorrect amino acid becomes bound. Thus, in theory, mutations could impair translation either by simply reducing rates of incorporation of the amino acid into nascent proteins or by decreased editing of improper aminoacylation. Evidence suggests, however that decreased aminoacylation alone is primary cause of abnormal translation, because the editing function persists in some of the disease-causing mutants (Nakayama et al., 2017, Simons et al., 2015).

As is shown in Figure 4, mutations in multiple enzymes that modify tRNA post-transcriptionally cause ID, but only some of these also cause epilepsy. Adenosine Deaminase tRNA Specific 3 (ADAT3) is responsible for deamination of adenosine to inosine at position 34 in tRNA (Alazami et al., 2013). Mutations have been shown to cause ID and microcephaly, with some patients developing epilepsy (Alazami et al., 2013, Ramos et al., 2019, El-Hattab et al., 2016, Salehi Chaleshtori et al., 2018). ADAT3-related ID is particularly prevalent because about 1% of people in Saudi Arabia are estimated to be carriers of one autosomal recessive mutation, V144M (Ramos et al., 2019). Another post-transcriptional modifier, Alkylation Repair Homolog 8 (ALKBH8) is a dioxygenase responsible for modifying wobble uridines of tRNA anticodons with several specific modifications, and mutations in this enzyme lead to ID with a high incidence of epilepsy (Monies et al., 2019). Two other proteins, Pseudouridine Synthase (PUS) 3 and PUS7 are responsible for the formation of pseudouridine from uridine at specific positions on tRNA (Shaheen et al., 2016, Shaheen et al., 2019, Darvish et al., 2019, Lecointe et al., 2002, Lecointe et al., 1998). Mutations in both proteins cause ID and microcephaly, although only PUS3 mutations have been reported to cause epilepsy (Shaheen et al., 2016, Shaheen et al., 2019, Darvish et al., 2019, de Paiva et al., 2019, Abdelrahman et al., 2018, de Brouwer et al., 2018).

NOP2/Sun RNA Methyltransferase 2 (NSUN2) is an RNA methylase responsible for methylation of cytosine at positions 47–50 in tRNA (Martinez et al., 2012, Hussain et al., 2013) (Figure 4). It also methylates mRNA and non-coding RNAs important for miRNA-dependent protein degradation. Mutations in NSUN2 cause syndromic ID and microcephaly, although few patients develop epilepsy (Martinez et al., 2012, Komara et al., 2015, Abbasi-Moheb et al., 2012, Khan et al., 2012). Such loss-of-function mutations can lead to both improper tRNA methylation and changed rates of protein synthesis by altering mRNA levels (Hussain et al., 2013). Which of these functions is most critical for the clinical phenotypes is, however, not known. Another RNA methylase, RNA 2’-O-Methyltransferase (FTSJ1), is required for 2’-O-methylatation of C32 and G34 of the tRNA for phenylalanine, and possibly other tRNAs (Guy et al., 2015). Loss-of-function mutations in this enzyme cause nonsyndromic X-linked ID in the absence of epilepsy (Guy et al., 2015, Takano et al., 2008, Freude et al., 2004). Finally, tRNA Methyltransferase 1 (TRMT1) dimethylates guanine 26 of most tRNAs. TRMT1 mutations that prevent this modification cause both ID and epilepsy (Najmabadi et al., 2011, Davarniya et al., 2015, Zhang et al., 2020)

One unifying aspect of these tRNA regulating proteins is that none have so far been shown to be modified by neuronal stimulation or play a role in activity-dependent translation. This category of proteins is the major exception to the hypothesis that dysregulated activity-dependent translation and epilepsy are critically linked, a theme of this review. Nevertheless, there may exist yet undiscovered links between neuronal stimulation and the activity of tRNA regulating proteins (Arif et al., 2017). Moreover, given the central role that tRNA play in translation, it is highly probable that there are multiple independent links between the dysregulation of tRNA and epilepsy. For example, the aaRS responsible for charging tRNA are strongly linked to epilepsy with ID, while the proteins responsible for modifying tRNA have a more variable association, suggesting that different mechanisms are responsible for the epilepsy associated with each category of protein.

3.4. mRNA Stability

Another set of regulators of translation are the RNA-binding proteins Pumilio1, Pumilio2. These have a dual function in that they determine the stability/degradation of some mRNAs to which they bind, but also independently control the initiation of the translation of other mRNAs. The pathway by which Pumilio1/2 proteins promote the degradation of the mRNAs to which they bind is shown in Figure 5A. This requires their interaction with eIF4E-transporter protein (4E-T). Pumilio and 4E-T form a translation repression complex (Zahr et al., 2018). In turn, 4E-T binds to eIF4E, resulting in a bridge between the 3’ and 5’ termini of mRNA. Subsequent degradation of the mRNA occurs by recruitment of the CCR4-NOT deadenylase complex and microRNA-mediated gene repression (Nishimura et al., 2015, Wang et al., 2015, Kamenska et al., 2016, Chauderlier et al., 2018) (Figure 5A).

Figure 5: Pumilio and DDX6 both regulate mRNA degradation and translation initiation.

A and C. Pumilio and DDX6 both repress translation by causing degradation of mRNA (Zahr et al., 2018, Nishimura et al., 2015). Through separate mechanisms, both bridge the 5’ and 3’ ends of mRNA by binding 4E-T. The CCR4-NOT deadenylase complex is then recruited leading to microRNA-mediated degradation of the mRNA (Nishimura et al., 2015, Wang et al., 2015, Kamenska et al., 2016, Chauderlier et al., 2018). B. Pumilio leads to repression of translation initiation. After binding to a conserved RNA binding domain, Pumilio interacts with PABP, thereby preventing its normal interaction with eIF4G and inhibiting translation (Weidmann et al., 2014, Chagnovich and Lehmann, 2001). D. DDX6 promotes translation initiation. YBX1 recruits DDX6 to the 3’ UTR. DDX6 then bridges the 5’ and 3’ ends of mRNA by recruiting eIF4E, leading to increased translation (Wang et al., 2015). Proteins linked to ID and epilepsy are colored as shown in the legend. Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause one condition, but sometimes cause the other are colored purple or orange.

Pumilio1/2 also repress initiation by interacting with PABP, which binds to poly-A tails to promote translation (Figure 5B). Although Pumilio does not interfere with the binding of PABP to mRNA, it suppresses its ability to enhance translation, most likely by preventing PABP binding to eIF4G, which results in diminished translation initiation (Weidmann et al., 2014, Chagnovich and Lehmann, 2001).

Loss-of-function Pumilio1 mutations lead to Pumilio1-Associated Developmental Disability, Ataxia, and Seizure (PADDAS) syndrome, which, as the name suggests, is associated with epilepsy, ataxia, and microcephaly in addition to early-onset ID. Milder mutations may lead only to adult-onset ataxia (Bonnemason-Carrere et al., 2019, Gennarino et al., 2018). A link between Pumilio2 and epilepsy has been found in animal models. Mice lacking Pumilio2 have spontaneous seizures and display a number of other behavioral abnormalities (Siemen et al., 2011). These mice also have altered expression of genes involved in synaptic structure and transmission, many of which are known epileptogenic factors (Follwaczny et al., 2017). Consistent with this, Pumilio2 levels are reduced in the hippocampus and temporal neocortex in patients with drug-refractory temporal lobe epilepsy (Wu et al., 2015).

Pumilio1 and Pumilio2 are clearly associated with activity-dependent translation in the brain, where they regulate the expression of proteins involved in LTP, such as the AMPA glutamate receptor 2 (GluR2), as well as proteins more generally associated with neuronal function and neurological disease (Dong et al., 2018, Bohn et al., 2018). There is also evidence that Pumilio expression is regulated by neuronal activity, and that Pumilio2 acts as a regulator of intrinsic excitability by repressing the expression of voltage-gated sodium channels (Dong et al., 2018, Kaye et al., 2009, Driscoll et al., 2013). Thus, Pumilio proteins provide a ready example of proteins associated with activity-dependent translation that cause epilepsy and ID when mutated.

Like the Pumilio proteins, DEAD-Box Helicase 6 (DDX6) has a dual function in that it regulates both the stability/degradation of mRNAs and the initiation of their translation (Figure 5). Unlike Pumilio1/2, however, DDX6 is not an RNA-binding protein but an RNA helicase and a component of ribonucleoprotein (RNP) granules called processing bodies (P-Bodies). The pathway by which DDX6 regulates mRNA degradation is shown in Figure 5C. It is similar to that of Pumilio1/2, in that, with several other proteins, it forms a repression complex with 4E-T (Nishimura et al., 2015). 4E-T binds to eIF4E, bridging the 3’ and 5’ termini of mRNA and leading to mRNA degradation, through recruitment of the CCR4-NOT deadenylase complex and microRNA-mediated gene repression (Nishimura et al., 2015, Wang et al., 2015, Kamenska et al., 2016, Chauderlier et al., 2018). The interaction of DDX6 with 4E-T can also lead to repression of translation of mRNA to which they are linked by their coordinated storage in P-bodies (Balak et al., 2019, Kamenska et al., 2016).

The mechanism by which DDX6 regulates initiation is quite distinct from that of Pumilio1/2. The association of DDX6 with the translation machinery enhances rather than suppresses the initiation of translation (Figure 5D). This is mediated by the protein YBX1, which can recruit DDX6 to stem-loops in the 3’ UTR of mRNAs. It then recruits eIF4E, thereby promoting increased translation initiation (Wang et al., 2015). De novo mutations in DDX6 have been associated with ID but not with epilepsy (Balak et al., 2019), and unlike Pumilio1/2, there is as yet no known link between DDX6 and activity-dependent translation.

3.5. Nonsense Mediated Decay

A pathway for degrading mRNA that is distinct from those shown in Figure 5 is Nonsense Mediated Decay (NMD). This is responsible for degrading mRNAs that have premature termination sites because of mutations or aberrant transcription. NMD, however, also plays a role in the regulation of several normal transcripts, which carry typical NMD motifs. The translation of these normal transcripts is regulated by the cellular level of NMD, which is determined by the downregulation or alternative splicing of proteins that carry out NMD (Wittmann et al., 2006, Lejeune, 2022). Transcripts regulated by NMD include ARHGAP24, which encodes Rho GTPase-activating protein 24, a member of the Rho family of proteins that have known linkages to ID through their roles in actin remodeling (Nguyen et al., 2012), and GADD45B (Growth arrest and DNA-damage-inducible βα) which regulates activity-dependent DNA demethylation (Xu et al., 2013).

Mutations in Up-Frameshift 3B (UPF3B), one the proteins required for NMD, are well established as a cause of X-linked ID with some patients also developing epilepsy (Xu et al., 2013, Tejada et al., 2019, Tzschach et al., 2015, Laumonnier et al., 2010). UPF3B is responsible for recognizing the exon junction complex (EJC) to trigger NMD (Raimondeau et al., 2018). Analysis of gene expression in patient cells reveals that about 5% of the genome is dysregulated by UPF3B loss-of-function mutations (Nguyen et al., 2012). Some of these dysregulated genes are canonical targets of NMD, such as ARHGAP24 (Xu et al., 2013, Nguyen et al., 2012).

Animal models have suggested that UPF3B plays a key role in the response of neurons to altered activity and that it regulates mRNA levels in an activity-dependent manner. Levels of UPF3B increase during development of the mouse brain, but to different extents and with different subcellular localization in different neurons (Jolly et al., 2013). Activation of hippocampal neurons with high potassium leads to decreased UPF3B mRNA levels within two hours, and loss of UPF3B increases the expression of canonical NMD targets, causing increased proliferation and dedifferentiation (Jolly et al., 2013). Mouse models of UPF3B loss-of-function have also identified changes in many other NMD targets that have functions related to neuronal differentiation and maturation (Huang et al., 2018).

While there exist other Up-Frameshift (UFS) proteins that control NMD, there is evidence that, uniquely among these proteins, UPF3B also plays a role in canonical translation termination (Neu-Yilik et al., 2017). UPF3B reduces termination efficiency by impairing the ability of eRFs to recognize the stop codon and hydrolyze peptidyl-tRNA. UPF3B can also bind to mRNA and ribosomes, causing the post-termination complex to dissociate. Another Up-Frameshift protein, UPF2, may competitively inhibit these functions by binding to UPF3B (Neu-Yilik et al., 2017). In addition, there exists another closely related Up-Frameshift protein, UPF3A, which, in theory, could compensate for some effects of the loss of UPF3B. Nevertheless, it appears that UPF3A is either much less effective at triggering NMD or that it actually inhibits NMD (Nguyen et al., 2012, Kunz et al., 2006, Shum et al., 2016).

3.6. FMRP-Related Regulators

3.6.1. FMRP

The most common genetic cause of inherited ID is Fragile X Syndrome (FXS), which is comorbid with epilepsy and ASD (Richter et al., 2015, De Rubeis and Bagni, 2010). FXS is caused by loss of the RNA binding protein, Fragile X Mental Retardation Protein (FMRP). This typically occurs by expansion of a CGG repeat region in the fmr1 gene that codes for FMRP, leading to its transcriptional silencing (Verkerk et al., 1991). It is well-established that FMRP influences the synthesis of new proteins by binding a subset of neuronal mRNAs, usually by suppression of translation, although for some mRNA targets, translation is stimulated by FMRP binding (Figure 6). Suppression of translation of FMRP-bound mRNAs can occur through several distinct mechanisms, including stalling ribosomes (Figure 6A), preventing initiation (Figure 6C), or altering the stability of transcripts (Figure 6B) (Richter et al., 2015, De Rubeis and Bagni, 2010, Darnell et al., 2011, Richter and Zhao, 2021). Neuronal stimulation can alleviate this suppression. Accordingly, in animal models, loss of FMRP leads to a small overall increase in protein synthesis, as well as loss of certain forms of activity-dependent protein synthesis. For example, loss of FMRP alters LTD produced by activation of metabotropic glutamate receptors (mGluR-dependent LTD) and its link to an increased protein synthesis (Ebert and Greenberg, 2013, Waung and Huber, 2009, Auerbach et al., 2011).

The way FMRP influences activity-dependent translation is determined by its post-translational modifications, specifically phosphorylation and ubiquitination. Phosphorylated FMRP generally represses translation, and its dephosphorylation by PP2A phosphatase after mGluR stimulation increases translation (Figure 6A) (Narayanan et al., 2007, Niere et al., 2012, Nalavadi et al., 2012). Ubiquitination followed by degradation of FMRP also contributes to the removal of translation repression after neuronal stimulation (Hou et al., 2006, Nalavadi et al., 2012). Both dephosphorylation and ubiquitination are required for mGluR-dependent changes in protein translation and LTD (Niere et al., 2012, Hou et al., 2006). Additionally, FMRP regulation of mRNA stability via interaction with microRNAs (miRNA) plays an important role in activity-dependent translation. Recruitment of miRNAs to mRNA is facilitated by phosphorylation of FMRP (Figure 6B). Dephosphorylation of FMRP following mGluR stimulation leads to release of miRNA from the mRNA (Muddashetty et al., 2011). The targets of this method of regulation, such as the NMDA receptor subunit NR2A, play important roles in synaptic plasticity (Edbauer et al., 2010). These extensive links between FMRP, activity-dependent translation, and synaptic plasticity may provide an explanation for the occurrence of ID in Fragile X Syndrome.

FMRP also binds the mRNAs for several ion channels that determine the intrinsic excitability of neurons. Both large scale screens and targeted validations have determined that FMRP binds the mRNAs of several ion channels including voltage-dependent K⁺ channels (Kv3.1, Kv4.2), Na⁺-activated K⁺- channels (K_Na1.1 (Slack)), and Ca²⁺-activated K⁺- channels (K_Ca1.1 (BK )) (Darnell et al., 2011, Gross et al., 2011, Lee et al., 2011, Strumbos et al., 2010, Darnell et al., 2001). Loss of FMRP disrupts the normal changes in channel expression that occur in response to neurotransmitters (Gross et al., 2011, Lee et al., 2011) or to changes in sensory inputs (Strumbos et al., 2010).

3.6.2. FMRP-Binding Proteins

FMRP forms complexes with several other proteins that also regulate translation and that are also implicated in ID and epilepsy. One of these is Polyglutamine-Binding Protein 1 (PQBP1). Experiments in flies have shown that PQBP1 binds FMRP and recruits it, with its target mRNAs, to ribosomes leading to enhanced translation initiation of particular mRNAs (Figure 6D) (Wan et al., 2015). Patients with PQBP1 loss-of-function mutations develop ID and microcephaly (Kalscheuer et al., 2003, Rejeb et al., 2011). Although PQBP1 has not directly been shown to play a role in activity-dependent translation, its interaction with FMRP suggests that this is very likely to be the case.

A second set of FMRP-binding partners linked to both ID and epilepsy are the Cytoplasmic FMR1 Interacting Proteins CYFIP1 and CYFIP2 (Schenck et al., 2001, Zhang et al., 2019b), which are both highly expressed throughout the brain. Of the two, CYFIP1 has been investigated more intensely and found to play mutually exclusive roles in the regulation of translation and actin nucleation. CYFIP1 forms a translation initiation inhibitory complex with FMRP, eIF4E, and PABP, acting as a non-canonical 4E-BP in this complex by binding directly to eIF4E (Figure 6C) (Napoli et al., 2008). Exposure of neurons to BDNF has been shown to release CYFIP1 from this complex, thereby releasing FMRP from the inhibitory complex and increasing translation of FMRP targets at synapses (Napoli et al., 2008, De Rubeis et al., 2013, Panja et al., 2014).

The molecular functions of CYFIP2 are less well established. Its potential role in the regulation of translation has not been fully explored, but it contains the same eIF4E binding motif as CYFIP1 and could therefore play some of the same roles in activity-dependent regulation of translation. The finding of decreased CYIFIP2 levels in Alzheimer’s disease patients and evidence from a mouse model of CYFIP2 haploinsufficiency, suggest that it regulates the translation of Amyloid Precursor Protein and Calmodulin-Dependent Protein Kinase II (CaMKII) (Napoli et al., 2008, Tiwari et al., 2016).

De novo CYFIP2 mutations have been implicated directly in both ID and epilepsy (Zweier et al., 2019, Zhong et al., 2019, Nakashima et al., 2018). Both CYFIP1 and CYFIP2 have been identified as candidate genes in West syndrome, which is characterized by infantile spasms, ID, and often intractable epilepsy (Peng et al., 2018). Microdeletions of the locus containing CYFIP1 and three other genes causes ID, developmental delay, and ASD in patients (Butler, 2017). There is indirect evidence that CYFIP1 may be the disease-causing gene at this locus. In mice, loss of CYFIP1 function leads to neuronal abnormalities, including high numbers of immature spines, thinning of the myelin sheath (as seen in FXS), increased inhibitory synapse number, increased inhibitory synaptic transmission, and changes in the expression of known FMRP targets and genes involved in epilepsy (De Rubeis et al., 2013, Nebel et al., 2016, Davenport et al., 2019, Pathania et al., 2014, Bozdagi et al., 2012). Finally, CYFIP1 mRNA and protein levels are increased in human patients with intractable Temporal Lobe Epilepsy and in a lithium-pilocarpine rat model of this disease (Huang, 2016).

In addition, to binding the mRNAs for several ion channels, FMRP also directly binds several ion channels themselves. One of these is the Slack Na⁺-activated K⁺ channel, which is activated by the binding of FMRP to the cytoplasmic C-terminus of the channel (Figure 6E) (Brown et al., 2010, Zhang et al., 2012). Additionally, there is evidence that Slack channels may play an independent role in activity-dependent translation (Zhang et al., 2012). Slack mutations are also associated with epilepsy and severe ID, providing another connection between FMRP, activity-dependent translation, and epilepsy-associated ID (Barcia et al., 2012, Heron et al., 2012, Kim and Kaczmarek, 2014). Other channels to which FMRP has been found to bind include BK channels (Deng et al., 2013, Kshatri et al., 2020), Kv1.2 (Yang et al., 2020), Kv4.3 (Zhan et al., 2020), K_Ca2.2 (Deng et al., 2019),Cav2.2 (Ferron et al., 2014), and Cav3.1 (Zhan et al., 2020). The direct binding of translation modulators such as FMRP to ion channels, which shape firing patterns by responding immediately to changes in membrane potential, may provide positive or negative feedback loops during activity-dependent translation.

3.6.3. Activity-Regulated Local Translation

One aspect of protein synthesis particular to neurons is localized mRNA translation that occurs far from the soma. Most cells are compact and, while domain-specific translation does occur, such cells can easily transport protein synthesized in the cytoplasm to all parts of the cell. In contrast, neurons have axons and dendrites that extend well beyond the soma. Extensive literature provides evidence of postsynaptic translation in dendrites and spines, and there is some evidence of translation in presynaptic terminals (Younts et al., 2016, Rangaraju et al., 2017). While the long-term effects of LTP and LTD may affect the entire cell, the changes, including activity-dependent translation, occur locally at synapses (Younts et al., 2016, Baltaci et al., 2019, Ebert and Greenberg, 2013, Waung and Huber, 2009). Some of the proteins discussed in this section are relevant to this local activity-regulated translation. In particular, FMRP has been shown to be critical for the normal regulation of local translation in dendrites and spines following LTD (Ifrim et al., 2015). The dysregulation of local translation in FXS is one possible contributing factor to the disease phenotype and provides clues for how mutations in other activity-dependent regulators of translation discussed here could cause related phenotypic outcomes.

3.7. Upstream Regulators

The final category of translation regulatory proteins are upstream regulators that guide the overall process of translation. The mammalian Target of Rapamycin (mTOR) is a serine/threonine protein kinase that, when activated, can phosphorylate multiple proteins in the initiation and elongation steps of translation, leading to rapid stimulation of translation in response to external signals (Wang and Proud, 2006). The pathways that regulate mTOR activity are illustrated in Figure 7 and constitute one of the central hubs of translation regulation in the cell. The canonical mTOR pathway signals through Protein Kinase B (AKT), Tuberous Sclerosis Proteins (TSC) 1/2, and Ras Homolog Enriched in Brain (RHEB). In this pathway, activation of AKT and RHEB stimulate translation, while TSC1/2 act as inhibitors. Other signaling pathways such as the MEK/ERK pathway have significant crosstalk with mTOR signaling (Mendoza et al., 2011).

Figure 7: Mutations in members of the mTOR and related pathways cause ID and epilepsy.

The mTOR signaling pathway is a central hub in the regulation of translation and cell growth (Wang and Proud, 2006). In the canonical mTOR pathway activation of AKT and RHEB stimulate translation, while TSC1/2 act as inhibitors. Other signaling pathways such as the MEK/ERK pathway have significant crosstalk with mTOR signaling (Mendoza et al., 2011). While growth factors control much of mTOR regulation, amino acids levels also regulate mTOR through the members of the GATOR complex (DEPDC5, NPRL2, and NPRL3) (Panchaud et al., 2013). Proteins linked to ID and epilepsy are colored in as shown in the legend. Mutations leading exclusively to ID or epilepsy are colored blue or yellow respectively. Mutations that cause both conditions are colored red. Mutations that primarily cause one condition, but sometimes cause the other are colored purple or orange.

The mTOR pathway is a hub for neurological disorders. Mutations in mTOR itself, as well as members of the mTOR pathway and proteins that activate mTOR through ERK signaling result in ID with epilepsy (Switon et al., 2017, Mirzaa et al., 2016). The most common of these disorders is tuberous sclerosis complex (TSC). TSC results from mutations in the genes TSC1 and TSC2 and affects about 1 in 5800 births (Curatolo and Bombardieri, 2008). TSC1/2 inhibit the activation of mTOR by RHEB, so loss-of-function in TSC leads to increased mTOR signaling and altered translation (Li et al., 2006). Many disease-causing mutations have been identified in these two genes, mostly in TSC2. While the disorder is characterized by the formation of benign, noninvasive lesions in many organ systems, severe neurological symptoms occur in many patients. The primary physical manifestation in the brain is the presence of cortical tubers, but other types of brain lesions also occur. The most common neurological symptoms are epilepsy, ID, and ASD. About 90% of patients develop epilepsy, which is often early-onset and tends to increase in severity over time. About half of patients have ID, although the severity varies greatly. Non-neurological comorbidities include renal and cardiac symptoms that occur in varying frequency and severity (Curatolo and Bombardieri, 2008).

Mutations in several other mTOR regulators result in epilepsy. The first class of these regulators form a pathway upstream of TSC. Phosphatase and Tensin Homolog (PTEN) and Phosphatidylinositol-4,5-Bisphosphate 3-Kinase Catalytic Subunit Alpha (PIK3CA) are upstream regulators of AKT3, which phosphorylates TSC to inhibit its activity. Brain somatic mutations in all three of these proteins lead to focal epilepsies (Marsan and Baulac, 2018, Conti et al., 2015). Some patients with AKT3 mutations also develop ID (Conti et al., 2015). The other class of these regulators are members of the GTPase-Activating Protein Activity Toward Rags 1 (GATOR1) protein complex, which senses amino acid levels and inhibits mTOR signaling upon amino acid deprivation (Panchaud et al., 2013). Mutations in members of this complex, DEP Domain Containing 5 (DEPDC5), Nitrogen Permease Regulator-Like (NPRL) 2, and NPRL3, can lead to focal epilepsies (Marsan and Baulac, 2018, Scheffer et al., 2014, Dibbens et al., 2013, Iffland et al., 2019). Mutation of DEPDC5 can also lead to dominant nocturnal frontal lobe epilepsy (Picard et al., 2014, Ishida et al., 2013). Some patients with mutations in DEPDC5 and NPRL3 also develop ID as a comorbidity (Scheffer et al., 2014, Dibbens et al., 2013, Iffland et al., 2019, Picard et al., 2014). Whether these mutations cause dysregulation of mTOR-dependent translation has not been investigated directly.

Mutations in mTOR protein itself can also lead to both epilepsy and ID. Deep sequencing in patients with focal cortical dysplasia type II (FCDII), which is characterized by intractable epilepsy, revealed somatic gain-of-function mTOR mutations in 16% of patients (Marsan and Baulac, 2018, Lim et al., 2015). Using in-utero electroporation to mimic these somatic mTOR mutations in mice caused spontaneous seizures that could be rescued using the mTOR inhibitor rapamycin, consistent with the hypothesis that increased mTOR activity leads to FCDII (Lim et al., 2015). Mutations in mTOR signaling have also been found to cause epilepsy and ID in other conditions including Smith-Kingsman Syndrome and Lennox-Gastaut Syndrome (Mirzaa et al., 2016, Gordo et al., 2018, Epi et al., 2013). Abnormal mTOR signaling has also been observed in Fragile X syndrome as well as in RETT syndrome, a condition caused by defects in DNA methylation that results in ID and seizures (Ricciardi et al., 2011, Sharma et al., 2010). There is evidence from FMRP knockout mice that mTOR signaling is responsible for the increased global translation seen in FXS (Troca-Marin et al., 2012). This alone, however, does not appear to account for changes in mGluR-dependent LTD, which, as stated above is a characteristic of the FXS phenotype in mice (Troca-Marin et al., 2012).

Proper functioning of the mTOR pathway is critical for both LTP and LTD, and mutations in many components of the pathway result in dysregulated activity-dependent translation (Switon et al., 2017). Depending on which member of the mTOR pathway is mutated, translation levels downstream of mTOR may be increased or decreased. Those mutations that reduce mTOR-dependent translation almost universally lead to diminished LTP or LTD. For example, mutations in mTOR itself, which cause decreased translation, lead to loss of both LTP and LTD (Jaworski and Sheng, 2006). Similarly, PIK3CA and AKT3 mutations, which also reduce downstream translation, both lead to diminished LTP (Zhang et al., 2019a, Saw et al., 2020). Perhaps surprisingly, mutations that cause increased mTOR signaling and increased downstream translation, such as those in TSC1/2 and PTEN, also generally cause deficits of both LTP and LTD (Auerbach et al., 2011, Switon et al., 2017, Takeuchi et al., 2013, Fraser et al., 2008, von der Brelie et al., 2006).

Synaptic GTPase-Activating Protein (SynGAP) is another upstream regulator of translation and is expressed primarily in neurons (Gamache et al., 2020, Paul et al., 2019). It is a major component of postsynaptic densities and exists in multiple isoforms, a result of alternative-splicing of its mRNA. In response to neuronal stimulation it undergoes changes in its phosphorylation state and subcellular localization. Its biological function is to regulate the activity of small signaling GTPases including Ras and Rap, and such changes in SynGAP can therefore influence multiple cellular processes, including suppression of the ERK/RHEB/mTOR pathway (Wang et al., 2013) (Fig. 6). Thus, knockdown of SynGAP levels in cultured cortical neurons lead an overall increase in protein synthesis as well as increased levels of GluR2-lacking AMPA receptors. These receptors are incorporated into postsynaptic membranes leading to enhanced synaptic strength. SynGAP has therefore been proposed to be a key regulator of normal synaptic homeostasis (Wang et al., 2013).

Patients with SynGAP deletions or mutations develop severe ID (Gamache et al., 2020, Hamdan et al., 2009, Hamdan et al., 2011, Rauch et al., 2012, Vlaskamp et al., 2019) Although a variety of estimates have been made, a majority of these patients also develop epilepsy and ASD (Gamache et al., 2020, Hamdan et al., 2009, Hamdan et al., 2011, Rauch et al., 2012). These mutations tend to be de novo autosomal dominant mutations, resulting in loss-of-function. More recently, SynGAP mutations have also been found as a potential risk factor in schizophrenia, although this role has not been fully confirmed (Gamache et al., 2020).

Recent findings have suggested that there is crosstalk between the regulation of translation by SynGAP and that of the FMRP pathway (Paul et al., 2019, Barnes et al., 2015). The phenotype of reduced SynGAP expression, as occurs in SynGAP^+/− in mice, resembles that of Fmr1^−/y mice in that both have increased basal protein synthesis, and mGluR-dependent LTD in both strains is independent of new protein synthesis (Barnes et al., 2015). At least in part, this is because the heterozygous knockout of SynGAP reduces levels of FMRP early in development (Paul et al., 2019). Moreover, FMRP binds SynGAP mRNA and inhibits its translation. These interactions lead to changes in levels of both regulatory factors, and in the characteristics of LTP, LTD and activity-dependent protein synthesis at different stages in development.

4. Conclusions: Translational Regulation and Synaptic Proteins

All the translation regulators we have discussed play a critical role in neurological function. While some of the genetic mutations described result in clinical phenotypes outside the brain, they all lead to specific neurological dysfunction in the form of ID or epilepsy (Table 1). In addition, mutations in many other regulators of translation not discussed here have been identified in other neurological and neurodegenerative diseases (Kapur et al., 2017, Kapur and Ackerman, 2018). This review has explored the clinical manifestations of mutations in each of the regulators of translation.

Protein components of translation can be divided loosely into two groups, those with well documented roles in translation that is specifically altered by neuronal activity and those for which no such role has been documented. One emergent theme is that mutations in proteins with well-described roles in activity-dependent translation almost universally cause epilepsy. This is illustrated in Table 2, which links many of these mutations to their effects on diminished, enhanced or constitutive activity-dependent translation (see Figure 1), on LTP or LTD, as well as their effects on epilepsy and ID. One potential exception is the tRNA-regulating proteins, in which mutations cause epilepsy but are not known to play a role in activity-dependent translation (Table 2). A comparison of Tables 1 and 2 reveals that while mutations in the proteins that control activity-dependent translation are also associated with ID, intellectual function can be readily impaired by many additional translation regulators with no known role in activity-dependent translation, and that these do not cause epilepsy.

One central question that arises from these analyses is why are so many translation regulators associated with neurological disease while having apparently much more minor effects in other tissues? One partial answer is that some of these proteins have neuronal-specific isoforms, suggesting their function in neurons differs slightly from that in non-neuronal tissue. In addition, some of the proteins discussed here, such as NOVA1/2, CYFIP2, and eEF1A2, are selectively expressed in neurons (Kaur et al., 2019, Zhang et al., 2019b, Ule et al., 2003). Nevertheless, many others, including rpL10, DPH1, Pumilio, FMRP and members of the mTOR pathway, are ubiquitously expressed in cells throughout the body (Pachler et al., 2004, Liu et al., 2012, Bohn et al., 2018, Troca-Marin et al., 2012). Another likely explanation is that, in contrast to most non-neuronal tissues, the effect of abnormal translation events may not be reversed by simply restoring normal translation. Specifically, the improper synthesis of synaptic components or ion channels may result in irreversible changes in the function of the neurons in that circuit, leading to both the hyperexcitability of the circuit and loss of its normal role in processing memory and learning. This concept is illustrated in Figure 8.

Figure 8: Model linking activity-dependent translation to epilepsy and ID.

A. Neuronal activity at synapses leads to changes in the translation of specific proteins, including ion channels, receptors, metabolic proteins, and proteins involved in axonal, dendritic, or synaptic growth. These changes in protein translation lead to long term changes in the excitability of neurons, as in LTP and LTD, and in the stability or loss of synaptic connections. This feedback loop between neuronal activity and activity-dependent translation plays a vital role in neuronal function. B. When proteins regulating activity-dependent translation are mutated this feedback loop is broken. Often, activity can no longer regulate translation because translation is constitutively increased to its post-activity level (as shown here) or translation is stuck at its pre-activity level. Both of these scenarios can lead to altered LTD or LTP. As described in this review, this frequently leads to epilepsy and ID.

Human mutations have revealed the critical role that translation plays in neuronal activity and the vulnerability of neurons to abnormal translation. Further experimental exploration of these links could provide important clues about how dysregulation of translation leads to specific neurological disorders and provide potential therapeutic avenues for the treatment of patients with mutations in the factors that control distinct stages of mRNA translation.