Article Text
Abstract
Background: Autism is a common childhood onset neurodevelopmental disorder, characterised by severe and sustained impairment of social interaction and social communication, as well as a notably restricted repertoire of activities and interests. Its aetiology is multifactorial with a strong genetic basis. EIF4E is the rate limiting component of eukaryotic translation initiation, and plays a key role in learning and memory through its control of translation within the synapse. EIF4E mediated translation is the final common process modulated by the mammalian target of rapamycin (mTOR), PTEN and fragile X mental retardation protein (FMRP) pathways, which are implicated in autism. Linkage of autism to the EIF4E region on chromosome 4q has been found in genome wide linkage studies.
Methods and results: The authors present evidence that directly implicates EIF4E in autism. In a boy with classic autism, the authors observed a de novo chromosome translocation between 4q and 5q and mapped the breakpoint site to within a proposed alternative transcript of EIF4E. They then screened 120 autism families for mutations and found two unrelated families where in each case both autistic siblings and one of the parents harboured the same single nucleotide insertion at position −25 in the basal element of the EIF4E promoter. Electrophoretic mobility shift assays and reporter gene studies show that this mutation enhances binding of a nuclear factor and EIF4E promoter activity.
Conclusions: These observations implicate EIF4E, and more specifically control of EIF4E activity, directly in autism. The findings raise the exciting possibility that pharmacological manipulation of EIF4E may provide therapeutic benefit for those with autism caused by disturbance of the converging pathways controlling EIF4E activity.
Statistics from Altmetric.com
Key points
Implication of EIF4E in autism
Molecular evidence: (physical mapping of a breakpoint in a boy with severe autism)-the breakpoint disrupts two proximal downstream putative exons of EIF4E
Further evidence coulc be provided by validating a EIF4E promoter mutation in two autistic children
The mutation increases EIF4E promoter activity
There is widespread expression of a putative exon downstream the breakpoint region in amygdala, hippocampus and cerebral cortex, regions of the brain implicated in autism (Northern Blot Analysis)
Expression of EIF4E in unchanged in autism lymphoblastoid cell lines
Autism (OMIM 209850) is a common childhood onset neurodevelopmental disorder, characterised by severe and sustained impairment of social interaction and social communicative abilities, as well as a notably restricted repertoire of activities and interests. Although multifactorial in origin, autism has a strong genetic basis. For autistic disorder, monozygotic twin concordance is 60%, and when a broader spectrum of autistic features is considered, concordance rates approach 90%, in contrast to 0% and 10% for dizygotic twins.1 Autism is clinically heterogeneous with up to 10% of cases associated with well defined neurological disorders such as tuberous sclerosis and fragile X syndrome.1 Genome wide linkage studies yielded several linkage peaks including 2q, 5p, 7q, 11p 17q, and, of interest here, 4q.2 3 4 5 Rare mutations have been found at several loci including the neuroligin, neurexin and SHANK3 genes.6 7 8 Microscopic chromosomal rearrangements are seen in 3–6% of autism,9 submicroscopic copy number variations (CNVs) are found in at least 10% of sporadic cases, but <2% in familial cases, and many individual loci remain to be identified.10 11
EIF4E activity is the rate limiting component of eukaryotic translation initiation, which directs ribosomes to the mRNA 5′ cap structure for initiation of protein synthesis. In brain, EIF4E activity is fundamental to the regulation of lasting alterations in synaptic strength or plasticity, and of long term potentiation (LTP): these are important in learning and memory.12 13 Increased activity in these systems can lead to repetitive, perseverative behaviour patterns.12 EIF4E is highly conserved across species and no germline mutations have been reported to date.
Synaptic translation is the final step in pathways already implicated in autism. Deregulation of PTEN/PI3K and tuberous sclerosis pathways, which converge on mammalian target of rapamycin (mTOR), an upstream regulator of EIF4E, have been implicated in conditions co-morbid with autism.14 Likewise, in fragile X syndrome, absence of FMRP upregulates synaptic translation through failure of recruitment of CYFIP1, a newly recognised EIF4E binding protein.15 16 Local protein synthesis is required for synaptic plasticity, a mechanism that underpins learning and memory.17 18 19 20 21
Regulation of EIF4E activity is known to play a key role in learning and memory through its control of translation within the synapse.12 13 Genome wide linkage studies in autism patients have shown linkage to the region containing the EIF4E locus on chromosome 4q.2 4 Here we present evidence from three independent families that strongly implicate EIF4E in autism.
Methods
Subjects
The child with the translocation was assessed using an informal clinical interview and the Autism Diagnostic Observation Schedule (ADOS; module 1 of generic version) at 4 years 9 months, and the Autism Diagnostic Interview Revised (ADI-R) at 6 years 7 months.22 23 Blood samples were obtained from the translocation carrier and his parents gave informed consent. The study protocol was approved by the Grampian local research ethics committee. A lymphocyte cell line was derived from the translocation carrier by transformation with Epstein–Barr virus (EBV) (European Collection of Cell Cultures, Health Protection Agency Culture Collections, Salisbury, UK).
Cases of non-syndromic autism were obtained from the Autism Genetic Research Exchange collection (http://www.agre.org/). Cases were selected on the basis of availability of DNA from both parents and two siblings with a definite diagnosis of autism (120 multiplex families).
One thousand and twenty control DNA samples from healthy individuals with no evidence of neuropsychiatric disorder were screened for the mutations found in autism cases. Putative mutations identified in autism cases were screened for in family members.
Cytogenetic analysis
Conventional cytogenetic analysis was performed on GTG banded metaphase nuclei at the 550 band level. Nuclei were obtained from a lithium heparinised peripheral blood sample following standard cytogenetic culture and harvest protocols.24 Fluorescent in situ hybridisation (FISH) was performed with commercial unique sequence telomere specific probes (Vysis), bacterial artificial chromosomes (BACs), and fosmid clones from the regions flanking the cytogenetic breakpoints. The BACs and fosmids were selected using the University of California at Santa Cruz (UCSC) Genome Bioinformatics Browser (http://genome.ucsc.edu/) and obtained from BACPAC Resources (Children’s Hospital Oakland Research Institute). Genomic DNA was labelled by direct incorporation of fluorochromes by nick translation (Vysis nick translation kit). BACs were hybridised for 24 h, followed by 2 min washes in 0.4× SSC/0.1% IPEGAL CA-630 (Sigma) at 72°C visualised at 100× magnification (Zeiss Neofluar objective) using an epifluorescence microscope (Zeiss Axiscop) and an Applied Imaging analysis system using the MacProbe version 4.3 software.
Chromosome flow sorting
Derivative chromosomes were separated from their homologues by dual laser flow sorting at the Molecular Cytogenetics Group, University of Cambridge. Generation of chromosome specific paint probes followed previously described methods.25 26
Detailed analysis of the breakpoint regions
DNA from der (4) and der (5) was amplified using the GenomePlex Single Cell Whole Genome Amplification Kit (WGA4) from Sigma-Aldrich. PCR primer pairs from across the region of interest were used to amplify DNA of derivative chromosomes. Primer pairs were selected to walk across both breakpoints to characterise the translocation by direct sequencing. CNV analysis on the Affymetrix human Gene-Chip 10 K array was used to exclude other cryptic rearrangements.
Mutation analysis
Direct sequencing was used to examine the coding regions and the promoter of the EIF4E gene. PCR products were purified by a Y-100 column (Fisher Scientific UK, Loughborough, UK) and direct sequencing was performed using the Big Dye Terminator v. 3.1 Cycle Sequencing Kit (Applied Biosystems, Warrington, UK). Sequencing reactions were analysed using an ABI 3100 Genetic Analyzer and results were analysed using the programs SEQUENCHER 3.1.1 (Gene Codes Corp, Michigan, USA). Putative mutations were validated by sequencing DNA from the affected sibling (and both parents) showing a variation, and 200 controls. Variants found in affected sibs but not controls were additionally screened for in 1020 anonymous control samples using denaturing high performance liquid chromatography (dHPLC) on a transgenomic WAVE apparatus, with a positive control on each run.
Binding reactions and electrophoretic mobility shift assays
Binding reactions with 80 fmol 32P 5′ end-labelled double-stranded oligonucleotides were performed in 16 mM Hepes-KOH (pH 8), 16% glycerol, 80 mM KCl, 0.16 mM EDTA, 0.8 mM DTT and 10 mg/ml HeLa cell nuclear extract (Abcam). After 30 min incubation on ice, reactions were analysed by 5% polyacrylamide gel electrophoresis, and visualised using autoradiography or a Fuji Phosphoimager with AIDA software for quantitation. Double stranded DNA molecules used were the wild type genomic sequence 5′-TTTCCTCTTACCCCCCCTTCTGGAGCGGTT (C7-4EBE) and the derivative C8-4EBE with an additional C added to the C7 stretch element. Where indicated, 200-fold or 500-fold excess of cold double stranded competitor DNA was added.
Luciferase assays
A 410 base pair EIF4E promoter fragment spanning the region from a PstI site up to the major transcription initiation site (position +1 in fig 3a) was amplified using the Roche Expand High Fidelity polymerase chain reaction (PCR) kit, with oligonucleotides to create a C7-4EBE and a C8-4EBE version. PCR fragments were inserted into pGEM-T easy (Promega, Southampton, UK) and verified by sequencing. The primers contained KpnI and HindIII restriction sites that were used to insert the promoter fragments into the firefly luciferase reporter vector pGL3 basic (Promega). Near confluent HeLa cells grown in 24 well plates were transfected with 0.7 μg pGL3 based reporter, and with 100 ng pRL-Tk (Promega) expressing Renilla luciferase. Cells were lysed 24–48 h after transfection and luciferase activities were determined using a dual luciferase assay (Promega). Firefly luciferase activity was standardised with respect to Renilla luciferase. Transfections were done in triplicates and each transfection was measured three times. Shown are the average and standard deviation calculated from the three parallel transfections.
Results
Routine cytogenetic screening identified a de novo balanced 46,XY,t(4;5)(q23;q31.3) translocation in a boy with classic autism (fig 1a). There is no family history suggestive of autistic traits, he had no dysmorphic features other than a double hair whorl on the crown, and no malformations. On the Autism Diagnostic Observation Schedule-Generic (Module) the child scored 21 points on algorithm items, and on the Autism Diagnostic Interview Revised (2000)22 23 he scored at ceiling level for most algorithm items with a total of 28 points in the domain of reciprocal social interaction, 14 points for non-verbal communication, and 8 points for repetitive behaviours. Thus, he demonstrated a typical and severe autistic phenotype and on all measures, DSM IV criteria for the diagnosis of autistic disorder were fulfilled. Development was apparently normal up until the age of 2 years. This was followed by a period of severe regression characterised by loss of speech, social interaction and communication skills, and the development of stereotyped and repetitive patterns of behaviour. At the age of 6.5 years his speech consisted of little other than stereotyped vocalisations, but did not display comorbid global developmental delay.
FISH analysis with BACs and fosmid clones localised the sites of the breakpoints in immortalised lymphocytes. BAC clone RP11-911N10 spans the breakpoint on chromosome 4 (fig 1b,c), and fosmids mapped the breakpoint on chromosome 5 to a 47.6 kb interval (142,854,992–142,902,586) (data not shown). Fine mapping was performed by PCR using DNA amplified from flow sorted derivative chromosomes.25 26 The breakpoint was further characterised by Sanger sequencing27 (fig 1d). Chromosome translocation occurred by breakage and re-ligation of chromosomes 4 and 5 and involved the addition of few (1–4) residues at the breakpoint. CNV analysis using the GeneChip Human Mapping 10 K 2.0 Array (Affymetrix) excluded other cryptic rearrangements (data not shown).
The breakpoint on chromosome four is located in a region linked to autism2 and maps 56 kb downstream of the EIF4E reference sequence. The breakpoint on chromosome 5 is not in a linked region, the nearest gene being NR3C1. Families with heterozygous mutation of NR3C1 are reported with hypertension, hypokalaemia and female masculinisation, but not autistic features.28 By contrast, our subject is normotensive with a normal urinary screen for catecholamines. Thus our finding supports previous data implicating the chromosome 4q region in autism.
To explore further a role of EIF4E in autism, we screened for mutations in 120 multiplex families with two autistic siblings obtained from the Autism Genetic Research Exchange collection (AGRE). In two independent families, direct sequencing revealed a heterozygous single base C insertion in the EIF4E promoter region in the proband. In both of these families, the variant was also present in the second autistic sibling and the father. The variant was not found in 1020 anonymous control samples. In fig 2, the chromatograms from the analysis of one family are shown, clearly revealing the presence of the insertion in the father and the two affected offspring. Inheritance of the mutation is identical for the second family (data not shown). AGRE SNP data from Affymetrix 10 k 2.0 arrays confirmed that the two families are unrelated. This sequence variant is not present in 2040 control chromosomes. It is located in a region previously identified as the EIF4E basal promoter element (4EBE) that binds hnRNPK and contains a stretch of 7 C nucleotides (C7-4EBE)29 30 (fig 2b). The C insertion extends it to an 8 C nucleotide run (C8-4EBE). A number of rare base changes were also present in the autism families (see supplemental table S1).
To determine whether this mutation changes the properties of the EIF4E promoter we performed binding studies with short double stranded DNA fragments derived from 4EBE, using HeLa cell nuclear extract as source of binding factors. Electrophoretic mobility shift assays (EMSA) demonstrate that the C8-4EBE DNA sequence variant has an increased affinity for an abundant nuclear protein, probably hnRNPK.30 This is revealed in binding experiments where excess unlabelled C8-4EBE DNA competes more efficiently with binding of the nuclear factor to 32P-labelled C7-4EBE DNA then an identical excess of unlabelled C7-4EBE DNA (fig 3a).
Promoter activity assays were performed using a 0.4 kb EIF4E promoter fragment with either a C7- or a C8-4EBE upstream of the firefly luciferase reporter gene, resulting in a construct where luciferase expression is exclusively controlled by one or the other of these elements. Both variants were transfected into HeLa cells and showed a significant increase in expression compared to the promoterless control construct. Importantly, the single C insertion (C8-4EBE) resulted in a significant two-fold increase in EIF4E promoter activity compared to the wild type (C7-4EBE) promoter (fig 3b). Thus this mutation causes deregulation of EIF4E expression.
All four of the affected children with this promoter variant scored highly within all domains of the ADI-R interview22 and had pronounced language delay. One showed regression. The ADI-R is not designed to provide an index of severity, but scores indicate that all subjects had a high level of symptoms.
Discussion
Our findings implicate the EIF4E gene in the pathogenesis of autism, and provide a strong case for further study of EIF4E and related pathways in autism. Our evidence is twofold. First, the chromosomal translocation in the boy with autism implicates the region containing EIF4E in autism. Second, we observed a heterozygous C insertion in the EIF4E promoter in two further unrelated families. In both families, this mutation changes a basal promoter element of EIF4E, shown to be the binding site of hnRNPK.30 Importantly, the mutation leads to a higher affinity for its binding protein, and causes a twofold increase in promoter activity.
Genomic data indicate that the EIF4E gene has at least two different transcription start sites. The transcript ENST00000280892, a member sequences of the human EIF4E consensus CDS set CCDS34031, is transcribed from the proximal promoter that contains 4EBE (fig 2B),29 31 whereas transcription of NM_001968.3 is initiated about 1.5 kb further upstream. Our data implicate the C insertion in the development of autism and suggest increased activity of the proximal promoter EIF4E promoter as a mechanism.
In both families, the C insertion is shared by a reportedly unaffected parent and two children diagnosed with autism. Variable penetrance and expression is frequently seen in families with autism. Likewise, in tuberous sclerosis, a disorder caused by TSC1 and TSC2 mutations associated with aberrant mTOR signalling, variation in severity of phenotype is common even within families. Furthermore, these parents may have mild behavioural features that are not catalogued with the AGRE collection. Of the five cases of autism we describe with deregulation of EIF4E, all had a high level of symptoms, and two of five exhibited regression. In contrast to findings at other autism loci, none of the affected cases with mutations in this study had either associated mental retardation or epilepsy.
The translocation may have its effect in two ways. The effect of downstream brain specific regulatory elements may be disrupted by the translocation. The translocation also disrupts an alternative EIF4E transcript. The GENSCAN algorithm32 predicts an EIF4E transcript NT_016354.401 that encodes a larger protein encompassing the EIF4E reference sequence including the cap binding domain required for the role of EIF4E in translation initiation. NT_016354.401 has three additional downstream exons, and two of these exons map downstream of the breakpoint (fig 4). Hybridisation to probes 2778989 and 2778996 on the Affymetrix Human Exon 1.0 ST array indicates that the two proximal downstream exons of NT_016354.401, which flank the breakpoint, are expressed at elevated levels in the cerebellum. Our own data revealed widespread expression of the terminal exon in the amygdala, hippocampus and cerebral cortex, regions of the brain implicated in autism (not shown). Disruption of NT_016354.401 by the translocation would most likely result in the degradation of the mRNA by a mechanism related to nonsense mediated mRNA decay, and cause downregulation of EIF4E activity.
Clearly, further investigation of function is limited by the unavailability of brain material from the affected individuals. EIF4E levels in transformed lymphocytes derived from the translocation case and the children carrying the C insertion were similar to levels in control transformed lymphocyte cell lines (not shown). Expression of the four genes flanking the breakpoint (fig 1) was similar in the lymphocyte cell line from the translocation case and in control cells (not shown). We reviewed the available genomic data across the EIF4E locus for evidence of alternative isoforms. Our analysis of EST data revealed that there is little alternative splicing, with some transcripts containing a cryptic second exon. Differences between EIF4E transcripts are concentrated at the 5′ end and 3′ ends. Differences at the 5′ end reflect at least in part the different promoter usage. In addition, a significant proportion of mRNAs lack the first exon. Differences at the 3′ end are mostly due to mRNAs ending prematurely and lacking the terminal exons. Many of these differences may be artificially caused by the methods used for analysis. However, it is possible that there might be several EIF4E transcripts, some of which may be brain specific. Future investigations will focus on identifying such transcripts and on the effects of deregulating EIF4E expression on synapse function in appropriate models.
A number of independent lines of evidence support a role for EIF4E in autism. First, it is located in a linkage hot spot implicated through linkage studies.2 Second, EIF4E activity is regulated by the highly conserved PTEN/PI3K and tuberous sclerosis pathways. These pathways converge on mammalian target of rapamycin (mTOR), an upstream regulator of EIF4E. In tuberous sclerosis complex (TSC), where 25–50% have autistic features, mutations in TSC1 and TSC2 remove inhibition of mTOR and increase EIF4E activity.13 33 Similarly, individuals with germline mutations in PTEN often have associated autistic features.14 Mice with knockouts of the intracellular receptor mediator of rapamycin activity, Fkbp12, display repetitive and other behavioural features like those found in autism.34 Deregulation of these signalling pathways can result in abnormalities of brain growth and synaptic plasticity in a manner analogous to fragile X syndrome, a learning disability disorder with prominent autistic features, where inactivation of FMRP causes upregulation of synaptic translation mediated through CYFIP1, a recently recognised EIF4E binding protein.16 Deregulation of glutamate signalling is also seen in both fragile X and TSC. Cap dependent translation is active during mGluR-LTD, and both MEK-ERK and PI3K-mTOR signalling pathways regulate EIF4E activity.35 Indeed, on the basis of the association of mutations in FMRP, TSC1/2 and PTEN with autism, Kelleher et al36 hypothesised that “defects in translational repression may represent a possible mechanism leading to autistic phenotypes”. Finally, mutation in another subunit of the translation initiation complex, EIF2B, causes the paediatric neurological disorder leucoencephalopathy with vanishing white matter.37 We previously reported decreased white matter and increased grey matter in autism.38 Thus, variation in EIF4E or interacting proteins is consistent with the white matter variation observed in autism.
Regulation of synaptic plasticity by EIF4E is highly complex and probably governed by as yet unknown additional EIF4E binding proteins.12 In oncogenesis, increased EIF4E activity results in specific upregulation of translation of particular mRNAs that are normally inefficiently translated.13 Subtle deregulation, either up or down, of synaptic EIF4E through modification of translation of specific brain transcripts may significantly impact on delicate processes such as synaptic consolidation. Penetrance and expressivity of such genetic variants will depend upon genetic background and environmental factors, possibly at specific stages in development. This could account for our finding of reportedly asymptomatic carriers of the insertion mutation among the parents of the affected cases. Larger studies are now required to determine the prevalence and penetrance of EIF4E mutations.
EIF4E then is the end point of a number of pathways already implicated in autism. Our work showing direct involvement of EIF4E therefore provides key additional support for Kelleher’s protein synthesis hypothesis for autism.36 The drug rapamycin suppresses EIF4E expression through its effects on mTOR. Exciting preliminary observational data from trials of rapamycin in TSC suggest that rapamycin may not only control tumour growth, but may also improve behaviour.39 Our findings raise the exciting possibility that in cases of autism caused by deregulation of translation, pharmacological manipulation EIF4E expression—either directly or, for example, through manipulation of mTOR signalling—could provide therapeutic benefit.
Acknowledgments
We thank Helen Murdoch for putting us in contact with the translocation case; Elaine Durward, Ben Milner, Ronggai Li, Guoqing Liu and Lynne Gray for technical assistance; Caroline Clark and Christine Bell for technical advice and Dee Rasalam for advice on diagnostic criteria for autism; the Medical Genetics and Tau Rx groups for sharing physical resources; and Irwin McLean and Ian Booth for invaluable comments. We gratefully acknowledge the family of our translocation case, participating AGRE families and the Autism Genetic Resource Exchange (AGRE) Consortium. The Autism Genetic Resource Exchange is a program of Cure Autism Now and is supported, in part, by grant MH64547 from the National Institute of Mental Health to Daniel H Geschwind (PI). Flow sorting was made possible by a Wellcome Trust grant to Malcolm Ferguson-Smith for the Cambridge Resource Centre for Comparative Genomics.
REFERENCES
Supplementary materials
web only appendix
Files in this Data Supplement:
Footnotes
▸ Additional table is published online only at http://jmg.bmj.com/content/vol46/issue11
Funding This project was funded by the Scottish Government Chief Scientist Office, NHS Grampian Endowments and an anonymous grant to the University of Aberdeen Development Trust.
Competing interests None declared.
Patient consent Obtained.
Provenance and Peer review Not commissioned; externally peer reviewed.
Linked Articles
- Miscellaneous