De Novo variants in EEF2 cause a neurodevelopmental disorder with benign external hydrocephalus

Abstract

Eukaryotic translation elongation factor 2 (eEF2) is a key regulatory factor in gene expression that catalyzes the elongation stage of translation. A functionally impaired eEF2, due to a heterozygous missense variant in the EEF2 gene, was previously reported in one family with spinocerebellar ataxia-26 (SCA26), an autosomal dominant adult-onset pure cerebellar ataxia. Clinical exome sequencing identified de novo EEF2 variants in three unrelated children presenting with a neurodevelopmental disorder (NDD). Individuals shared a mild phenotype comprising motor delay and relative macrocephaly associated with ventriculomegaly. Populational data and bioinformatic analysis underscored the pathogenicity of all de novo missense variants. The eEF2 yeast model strains demonstrated that patient-derived variants affect cellular growth, sensitivity to translation inhibitors and translational fidelity. Consequently, we propose that pathogenic variants in the EEF2 gene, so far exclusively associated with late-onset SCA26, can cause a broader spectrum of neurologic disorders, including childhood-onset NDDs and benign external hydrocephalus.

Introduction

The EEF2 gene (MIM: 130610) encodes the eukaryotic translation elongation factor 2 (eEF2), a catalyst required for the elongation of polypeptide chains during mRNA translation (1). It coordinates structural changes in the ribosome—from its pre- to post-translocational state—that result in the displacement of messenger RNA (mRNA) and transfer RNA (tRNA) molecules. Specifically, after peptide bond formation, eEF2 catalyzes the GTP-dependent translocation of a peptidyl-tRNA from the A to the P site and a deacylated-tRNA from the P to the E(xit) site while moving the ribosome one codon in the 3′ direction along the mRNA. In line with its fundamental role in protein synthesis, eEF2 is ubiquitously expressed. The functional conservation of eEF2 mirrors the structural conservation of this gene sequence among eukaryotes, from yeast to human (2,3). The eEF2 is composed of five domains involved in polypeptide elongation. Domains I and II cluster together to form the GTP-binding pocket. The interactions between domains I, II, III and V with Helices 98 (the Sarcin/Ricin loop or SRL) and 43 (the GTPase associated center or GAC) of the large ribosomal subunit rRNA coordinate and stimulate the GTPase activity of eEF2 and translocation. The interaction of domain IV Helix 69 is thought to help coordinate the movement of eEF2 with the peptidyl and deacylated tRNAs in the ribosome (4).

To date, the heterozygous p.(Pro596His) substitution was the only known variant of the EEF2 gene associated with human disease, specifically the autosomal dominant late-onset spinocerebellar ataxia 26 (SCA26 [MIM: 609306]) (5). This variant was carried by 24 affected individuals and two asymptomatic individuals (age range: 26–60 years) of a six-generation Norwegian family with pure cerebellar ataxia (5). Atrophy of the cerebellum was demonstrated in 11 affected individuals (5). Functional studies in yeast demonstrated that the SCA26 p.(Pro596His) variant (P580H in the EFT2 gene in yeast) resulted in specific translational fidelity defects and a greater susceptibility to proteostatic stress (5). As the EFT2 P580H substitution did not cause gross destabilization or mislocalization of the protein, it was concluded that it retained a degree of biological function that is compatible with life (5). We identified de novo EEF2 missense variants in three children with neurodevelopmental delays and various structural brain abnormalities, including benign external hydrocephalus. Additionally, we provide in vitro and in vivo evidence that these variants alter cell growth and translational fidelity, resulting in a broader spectrum of EEF2-related neurodevelopmental disorders (NDD).

Results

De novo EEF2 variants are likely deleterious

De novo missense variants in the EEF2 gene (NM_001961.3) were identified in peripheral blood samples of three unrelated individuals presenting with NDD (Table 1). None of the variants were found in the individual’s parents, none of whom were consanguineous, nor did any have first degree relatives with a NDD. No other likely pathogenic variants were identified in additional candidate genes. EEF2 variants were localized throughout the gene (Fig. 1A) and occurred in highly conserved amino acids, considering 12 species (Supplementary Material, Fig. S1). None of the EEF2 variants were present in the Genome Aggregation Database (gnomAD) (6). Population data available in gnomAD indicated that the gene is highly constrained for predicted loss-of-function (pLoF) and missense variation (LOEUF = 0.172, misZ score = 4.878) (6). Furthermore, all missense variants were predicted to be likely pathogenic by a CADD score above 20 (7). Structurally, all de novo missense EEF2 variants mapped to functionally important locations within the protein (Supplementary Material, Fig. S2). Specifically, p.(Cys388Tyr) is positioned at a critical interface between domains I and II, p.(Val28Met) is at the base of domain I where it directly contacts the SRL, and p.(His769Tyr) is located in domain V, which is known to interact with the GAC at the tip of Helix 43.

Table 1.

De novo missense variants in EEF2

Individual	cDNA changea	Amino acid changeb,c	Protein domain	CADD score	Corresponding yeast variantd
1	c.82G>A	p.(Val28Met)	Protein synthesis factor, GTP-binding; small GTP-binding protein domain	28.5	V28M
2	c.1163G>A	p.(Cys388Tyr)	Not in a domain	33	C372Y
3	c.2305C>T	p.(His769Tyr)	Translation elongation factor EFG/EF2, C-terminal	24.2	Q753Y

^aGenBank: NM_001961.3.

^bGenBank: NP_001952.1.

^cNone of these variants is present in the gnomAD.

^dFor complete description of the eEF2 yeast variants, please see Supplementary Material, Table S1.

Figure 1.

De novo pathogenic EEF2 variants (A) cause a NDD with minor craniofacial dysmorphisms (B) and structural brain abnormalities (C). (A) EEF2 transcript (NM_001961.3) and human de novo EEF2 variants within or adjacent to specific domains of the eEF2 protein (NP_001952.1). (B) Individuals 1, 2 (B1) and 3 (B2) had a large head with a prominent forehead. (C) Axial, T2-weighted MR images of individuals 1 (C1, upper panel), 2 (C2) and 3 (C3), showing mild enlargement of lateral ventricles and external CSF spaces. Sagittal, T1-weigthed MR image of individual 1 (C1, lower panel) illustrates the diffuse thinning of the corpus callosum.

Individuals with de novo missense EEF2 variants share neurodevelopmental delay and other structural and functional nervous system abnormalities

All three male individuals of 3, 6 and 9 years of age had neurodevelopmental delays (Table 2). Ages of walking were 22, 24 and 26 months and ages at first words were 12, 15 and 24 months. Neurologic abnormalities (2/3) included hypotonia, gait instability, poor motor coordination and seizures. Morphological central nervous system (CNS) abnormalities were also frequent (3/3) (Fig. 1C). All individuals had some widening of cerebrospinal fluid (CSF) spaces, especially dilated lateral and third ventricles. Individual 1 had additional abnormalities of the cerebral morphology, i.e. diffuse thinning of the corpus callosum and left temporo-occipital focal dysplasia. Behavioral problems (1/3) included autistic behavior. At physical examination, all individuals had a head relatively larger than expected for height. Short stature could be observed in individual 3. Furthermore, they shared facial dysmorphisms, namely a prominent forehead with a high hairline, small and low-set ears with a prominent helix and antihelix, deep set eyes with narrow palpebral fissures, a short nose, a thin upper lip and a small (prominent) chin (Fig. 1B). Interestingly, both individuals 1 and 2 had fine and sparse scalp hair, sparse eyebrows and nail dysplasia (hypoplastic and dystrophic toenails). Individual 3 had, fast growing hair and nails, except for the toenails of the 5th toes that were hypoplastic. Minor hand abnormalities (3/3) included bilateral palmar creases, short and tapering fingers, and 5th finger clinodactyly. Mild 2–3 toe syndactyly was observed in individual 1. Ophthalmologic abnormalities (2/3) included strabismus requiring surgery and myopia.

Table 2.

Genotype and phenotype of individuals with de novo pathogenic variants in the EEF2 gene

Individuals	1	2	3
Gender/Age at examination	M/3 y 10 m	M/9 y	M/6 y 4 m
Genotype
cDNA changea	c.82G>A	c.1163G>A	c.2305C>T
Protein changeb	p.(Val28Met)	p.(Cys388Tyr)	p.(His769Tyr)
Inheritance	de novo	de novo	de novo
Phenotype
Growthc
Gestational age at birth	39 w	37 w	38 w
Length at birth (centile range)	NA	52 cm	NA
Weight at birth (centile range)	3345 g (25–50th)	3440 g	2470 g (3rd–15th)
HC at birth (centile range)	NA	35 cm	NA
Age at measurements	3 y 10 m	6 y/9 y	6 y 4 m
Height (centile range)c	101 cm (~50th)	111 cm (~20th)/130 cm (~30th)	102 cm (−2.5 SD)
Weight (centile range)c	18 kg (~85th)	19 kg (~25th)/26.4 kg (30th)	15.5 kg (−2.2 SD)
HC (centile range)c	52 cm (85-97th)	52.7 cm (80th)/53.4 cm (75th)	52.3 cm (~75th)
Prenatal and neonatal history
Congenital abnormalities	Periscrotal hypospadias	NA	−
Other neonatal problems	Jaundice requiring phototherapy	NA	−
Psychomotor development
Motor delay	+	+	+
Age at walking	22 m	24 m	26 m
Speech delay	+	+	+
Age at first words	15 m	24 m	12 m
Intellectual disability	−	−	−
Degree of intellectual disability	n.a.	n.a.	Mild
Neurologic and psychiatric features
Neurological abnormalities	+ (Hypotonia, unsteady gait, high stepping)	+ (Poor motor coordination)	−
Brain abnormalities (brain MRI)	+ (Mild–moderate enlargement of lateral and third ventricles, diffuse thinning of CC, left temporo-occipital focal dysplasia)	+ (Mild–moderate enlargement of the lateral and third ventricles and external CSF spaces)	+ (Mild enlargement of the lateral ventricles and external CSF spaces)
Seizures (age of onset/type)	+ (2 y 6 m/2 febrile seizures)	−	−
Abnormal EEG (age/result)	− (2 y 7 m/normal)	n.a.	n.a.
Behavioral problems	−	+ (Autistic behavior)	−
Dysmorphic features
Craniofacial dysmorphisms	+ (Minor)	+ (Minor)	+ (Minor)
Hands	+ (Bilateral single transverse palmar creases; short fingers; clinodactyly of the 5th finger)	+ (Clinodactyly of the 5th finger)	+ (Tapering fingers)
Feet	+ (Mild 2–3 toes syndactyly)	−	−
Other abnormalities
Abnormal vision	+ (Strabismus requiring surgery)	+ (Myopia)	−
Skin/hair/nails abnormalities	+ (Fine sparse scalp hair, sparse eyebrows, hypoplastic and dystrophic toenails, capillary malformations)	+ (Fine sparse scalp hair, fast growing and brittle toe nails)	+ (Fast growing hair and nails, hypoplastic nails of 5th toes)
Musculoskeletal system abnormalities	−	+ (Mild joint laxity)	−

Abbreviations are as follows: +, present; −, absent; F, female; M, male; y, years; m, months; w, weeks; SD, standard deviation; HC, head circumference; CC, corpus callosum; CSF, cerebrospinal fluid; NA, not available and n.a., not applicable.

^aGenBank: NM_001961.3.

^bGenBank: NP_001952.1.

^cPercentile range; if the percentile is <3rd or >97th, SD is indicated.

De novo EEF2 variants alter cellular growth and sensitivity to translational inhibitors

Often used as a genetic model for human diseases, the yeast Saccharomyces cerevisiae protein translation apparatus is highly similar to that of higher eukaryotes. The eEF2 amino acid sequence is 66% identical with 85% homology between yeast and humans at the DNA and protein levels, respectively (2,3), making this organism a viable substitute for functional analysis. Thus, a yeast model system was used to investigate the functional consequences of eEF2 variants. Yeast eEF2 mutants were constructed using a strain (generously donated by Dr Terri Kinzy) in which both endogenous copies of the paralogous eEF2 genes, EFT1 and EFT2, were disrupted and replaced with a low-copy plasmid bearing the EFT2 gene (8). A multiple sequence alignment was used to determine equivalent codons within the Homo sapiens and S. cerevisiae eEF2 sequence. If the position of the variant was not conserved, two yeast strains were constructed: one containing the equivalent H. sapiens codon and the other, the allele of the affected individual (Supplementary Material, Table S1). Additionally, the P580H and H699N alleles were employed as positive controls because their functional defects have been previously characterized in yeast systems (5,9,10).

Growth curves were collected in order to determine whether the sole expression of variant eEF2 would have an effect on cell growth rates. Most eEF2 variants demonstrated overall slower growth, i.e. increased doubling times, indicating disruption to normal growth but retention of viability (Fig. 2A). Surprisingly, the V28M allele enhanced cellular growth rates relative to the wild-type (WT) eEF2.

Figure 2.

Growth phenotypes of eEF2 model yeast strains showed altered cellular growth and sensitivity to translational inhibitors. (A) Growth curves of WT and variant yeast strains in rich media. Of the de novo variants, V28M exhibited an increased growth rate, while C372Y and Q753Y showed varying decreased growth rates. The growth rate of the control Q753H was equivalent to the WT, validating the yeast model approach. Error bars denote standard deviation. (B) Doubling times of all model strains in media with varying translational inhibitors. Doubling time was calculated by dividing ln(2) by the growth rate, which was found by non-linear regression of the exponential phase of each growth curve. Missing values indicate strains that were inviable. Errors denote 95% confidence intervals. (C) Heat map illustrating the effects of each translational inhibitor on growth of the variant yeast strains relative to WT. Cycloheximide: V28M sensitive; C372Y, Q753Y resistant. Hygromycin B: C372Y lethal; Q753Y sensitive; V28M resistant. Anisomycin: V28M sensitive; C372Y, Q753Y no change. Paromomycin: C372Y lethal; Q753Y sensitive; V28M no change. Heat map values were generated by calculating the log₂ of the fold-change relative to WT for each variant’s doubling time in the presence of each translational inhibitor. This was then subtracted from the log₂(FC) for growth of the corresponding variant in rich media.

Translational inhibitors can be used as probes to indicate specific aspects of the protein synthetic process that may be altered by variants. Paromomycin is an aminoglycoside whose binding to the small subunit rRNA increases misreading by decreasing the ability of the ribosome to proofread incoming aminoacyl-tRNAs, and also by inhibiting translocation (11). Hygromycin B is also an aminoglycoside that binds close to the decoding center near the top of helix 44 of small subunit rRNA. Although it does cause some miscoding, its major effect is to inhibit translocation by sequestering peptidyl-tRNA in the ribosomal P site through increased affinity of the A site for aminoacyl-tRNA (11). Cycloheximide inhibits eEF2-mediated translocation by blocking the ability of deacylated tRNA to enter the E-site of the large ribosomal subunit (12). Lastly, anisomycin binds to the A-site of the large subunit, inhibiting aa-tRNA accommodation into the A site (11).

Consistent with the central role played by eEF2 in the process of translocation, hygromycin B, paromomycin and cycloheximide all had effects on at least one of the strains, whereas anisomycin tended to have either small inhibitory (V28M and H699N) or no effects on cell growth (C372Y, Q753Y and P580H) (Fig. 2C). In validation of the humanized strain employed as a control for Q735Y, it did not demonstrate altered growth compared with the WT (Fig. 2A), indicating that the growth phenotypes observed were specific to the patient-derived allele. Notably, however, variant-specific phenotypes were observed. For example, the V28M variant conferred resistance to the hygromycin B; in contrast, the Q753Y and H699N variants conferred sensitivity to this drug, and the C372Y and P580H were completely inviable at the concentration employed in these experiments (Fig. 2C). A similar profile was observed for paromomycin, consistent with its chemical similarity with hygromycin B. The sensitivity/resistance profiles to cycloheximide tended to be the reverse of the aminoglycosides, with the exception of the H699N variant (Fig. 2C).

De novo EEF2 variants confer allele-specific effects on translational fidelity

To further probe the function effects of the eEF2 variants, dual-luciferase assays were performed in the yeast models to determine their abilities to accurately translate the genetic code (Fig. 3A). Analysis of programmed −1 ribosomal frameshifting (−1 PRF) recoding demonstrated that the de novo variants exhibited no significant change in −1 PRF frequency while recapitulating previous findings that P580H and H699N cause an increase. Furthermore, translational recoding of the programmed +1 ribosomal frameshifting (+1 PRF) signal did not appear to differ between any of the strains (Fig. 3B) Termination codon readthrough (TCR) is also an indicator of changes in translational fidelity. As seen in Fig. 3B, several eEF2 variants result in a decrease in the background level of TCR, including C372Y and Q753Y, with V28M exhibiting a similar rate to that of the WT strain. The previously reported P580H and H699N also decreased the amount of TCR.

Figure 3.

Translational recoding in eEF2 (yeast) model strains showed patient-derived variants alter translational fidelity. (A) Control and experimental vector designs for dual-luciferase assay. Dual-luciferase reporters are bicistronic constructs that contain two luciferase genes separated by a spacer region (21). These reporters can be easily engineered to insert different translational recoding signals within this spacer, thereby affecting the translation of the second luciferase gene. Recoding signals that can be utilized include RNA sequences that program ribosomes (PRF) to slip by one base in either the 5′ (−1 PRF) or 3′ (+1 PRF) direction. Canonical termination codons can also be used to determine the amount of TCR. The reporters are designed where the second gene is only translated if recoding occurs. The amount of both translated proteins can be assayed using chemiluminescence and their ratio can be compared with that of a control reporter containing no recoding signal. This yields the frequency of recoding events that occur during translation of that experimental reporter. The recoding signals used to determine the translational fidelity were the −1 PRF signal from the human immunodeficiency virus (HIV-1) (22) and the +1 PRF signal derived from the S. cerevisiae Ty1 retrotransposable element (23). These sequences were employed because they have been well characterized and are frequently used to measure translational recoding. The ability of ribosomes to misread the UGA termination codon was used to measure TCR, since it is the most prone to endogenous readthrough of the three stop codons (24), thus enabling maximization of signal-to-noise ratios. Using these different recoding signals enables the analysis of the overall translational fidelity profile of each variant eEF2 model. (B) Translational recoding in eEF2 (yeast) model strains: −1 frameshifting, +1 frameshifting and termination codon read-through in yeast strains. Of the de novo variants, C372Y (ns) and Q753Y (ns) exhibited increased −1 PRF, no strains exhibited altered +1 PRF, and C372Y and Q753Y showed decreased TCR. Data represented as fold-change values relative to WT. Dotted line marks a fold-change of 1, which would indicate no difference between WT and variant. Box plots show the median, interquartile range and range of data. P-values were calculated by Tukey’s multiple comparisons tests between the WT and each variant. ns: non-significant, ^*P<0.05, ^**P<0.01, ^***P<0.001.

Discussion

We identified de novo pathogenic EEF2 variants in three unrelated children with a NDD and structural brain anomalies. Here, we propose that the EEF2 gene, so far exclusively associated with the autosomal dominant adult-onset SCA26, is involved in a broader spectrum of NDDs.

Firstly, in vitro and in vivo findings support a deleterious effect of the de novo EEF2 variants. Bioinformatic analyses indicated that the substitution of the three conserved amino acids are likely damaging and all de novo variants were not previously reported in healthy population-based cohorts (6), supporting the low frequency of these missense variants in EEF2. Furthermore, this gene is highly intolerant to pLoF variation (6). Additionally, investigation of yeast models of eEF2 patient-derived mutants showed these EEF2 variants confer growth and translational fidelity defects. The observation of variant-specific differences in both translational fidelity and sensitivity/resistance to translational inhibitors points to differences in the effects of the variants on different aspects of eEF2 functions. For example, the p.(Cys388Tyr) and p.(His769Tyr) variants promoted better recognition of stop codons (TCR) than even WT cells. This may be explained by decreased ability of eEF2 to compete with the elongation termination factors eRF1/eRF3, either by decreased affinity for the ribosome or by decreased GTPase activity. In contrast, the p.(Val28Met) mutant did not affect the aspects of translational fidelity assayed in this study. Of particular interest, the p.(Val28Met) variant is located at a critical site of interaction with the SRL which plays a central role in the translocation process. Phenotypically, it is unique in that it grows faster than WT and shows altered sensitivities to translational inhibitors that are very different than the other mutants. The p.(Cys388Tyr) variant is positioned at an important interface between domains I and II. It may affect folding of the GTP binding pocket and/or the structural flexibility of the molecule (13). The p.(Cys372Tyr) variant, which corresponds to p.(Cys388Tyr) of individual 2, conferred the most pronounced phenotypic defects in yeast. The p.(His769Tyr) substitution is located at a critical interaction between eEF2 and the distal tip of the large subunit ribosomal RNA Helix 43, the GAC. This variant conferred altered sensitivities to translational inhibitors and decreased termination codon misreading. It may affect the affinity of eEF2 for the ribosome, and/or ribosomal stimulation of the eEF2 GTPase activity by the ribosome. Lastly, the SCA26 p.(Pro596His) substitution maps to the critical interaction between domain IV and H69. This variant stimulated −1 PRF and conferred increased fidelity at termination codons, suggesting that it inhibits eEF2 function, perhaps by decreasing the affinity of the protein for the ribosome.

Secondly, individuals with de novo EEF2 variants present with a childhood-onset, variable phenotype consisting of developmental delay/intellectual disability, non-specific craniofacial dysmorphisms and abnormalities of the brain morphology, including benign external hydrocephalus. EEF2 is associated with diverse neurodevelopmental presentations, ranging from psychomotor developmental delay with normal intelligence (individuals 1 and 2) to mild ID (individual 3). In these individuals, motor delay was more pronounced than speech delay. The three individuals also shared structural brain abnormalities, in particular a mild to moderate enlargement of the lateral and third ventricles and external CSF spaces. Interestingly, individual 1 manifested additional cerebral morphology abnormalities, such as cortical dysplasia and an abnormal corpus callosum. Craniofacial dysmorphisms included relative macrocephaly due to benign external hydrocephalus with a prominent forehead, low-set ears with a prominent helix, deep set eyes, short nose and thin upper lip. Ectodermal abnormalities included fine sparse scalp hair and nail dysplasia in individuals 1 and 2 and fast growing hair and nails in individual 3. Although the number of patients identified so far is limited, overlapping mild clinical features in three patients seems to exemplify the variable expressivity of EEF2-related disorders, because the phenotype of these children differs substantially from that of previously reported adults presenting with SCA26 (5), in particular due to the childhood age of onset, occurrence of the intellectual disability and distinct morphological CNS abnormalities. Remarkably, developmental delay/intellectual disability, short stature, dysmorphic facial features, abnormal head circumference (microcephaly or macrocephaly), CNS malformations and ectodermal abnormalities were previously described in individuals carrying biallelic variants in the DPH1 (14–16) or DPH2 genes (17), which result in a deficiency in the biosynthesis of diphthamide, a critical post-translational modification in eEF2 that enables the regulation of its function.

Altogether, the fundamental role of eEF2 in gene expression, the abnormal cell growth pattern, translation inhibitors sensitivity and translation fidelity in yeast, and the observed delayed neurodevelopment and brain structure anomalies, support the pathogenicity of the three de novo EEF2 variants. As described above, we hypothesize that these missense variants result in distinct functional effects, based on allele-specific phenotypes in yeast. Since de novo variants in EEF2 result a childhood-onset, more severe and more variable phenotype than the autosomal dominant late-onset SCA26, we propose that the phenotype of the EEF2-related disorders is a consequence of the specific pathogenic variants that distort the different eEF2 intrinsic functions and its interactions with the ribosome, which differently influence the ability to accurately translate the transcriptome and subsequently disrupts growth on various organ levels. A multicenter effort to further clinically and molecularly characterize a growing number of individuals with an EEF2-related disorder will shed light on the predictable broad spectrum of these disorders. In order to facilitate this, a website will be available for collection of in-depth phenotypic data of individuals carrying likely pathogenic EEF2 variants, which aims to corroborate the observed phenotypic heterogeneity and establish firm genotype–phenotype correlations.

Materials and Methods

Patients

Two unrelated Dutch individuals were enrolled at the outpatient clinic of the Department of Human Genetics, Radboudumc (Nijmegen, The Netherlands). One additional individual was identified with the collaboration of GeneDx through GeneMatcher (18). Likely pathogenic variants were detected by performing clinical exome sequencing of DNA extracted from peripheral blood samples of the affected individuals and parents (trio analysis). This study was approved by the institutional review board Commissie Mensgebonden Onderzoek Regio Arnhem-Nijmegen (CMO-nr 2018-4736) and informed consent for enrollment was obtained from all legal representatives.

Exome sequencing

Individual 1

Using genomic DNA from the proband and parents, the exonic regions and flanking splice junctions of the genome were captured using the IDT xGen Exome Research Panel v1.0. Massively parallel (NextGen) sequencing was done on an Illumina system with 100 bp or greater paired-end reads. Reads were aligned to human genome build GRCh37/UCSC hg19, and analyzed for sequence variants using a custom-developed analysis tool. Additional sequencing technology and variant interpretation protocol has been previously described (19). The general assertion criteria for variant classification are publicly available on the GeneDx ClinVar submission page (http://www.ncbi.nlm.nih.gov/clinvar/submitters/26957/).

Individuals 2 and 3

Exome sequencing was performed as previously described (20).

The variants identified in this study were submitted to the Leiden Open Variation Database (https://databases.lovd.nl/shared/genes/EEF2).

Yeast strains

EFT2-6xHis variants were generated by site-directed mutagenesis of pJD2490 using the Aligent QuikChange Lightning Kit (Aligent Technologies, catalog #210519) per the manufacturer’s instructions. Mutagenic oligos were synthesized by Genewiz (Supplementary Material, Table S2). EFT2 strains were generated by alkali cation transformation of EFT2-6xHis encoding plasmids into yJD995 (MATa ade2 ura3 his3 leu2 trp1 eft::HIS3 eft2::TRP1 +YCpEFT1-URA3) using negative 5-FOA selection to shuffle out the endogenous EFT1 plasmid (Supplementary Material, Tables S3 and S4). Once viable colonies were obtained, cells were grown in YPAD and the plasmids were isolated using the GeneJET Plasmid Miniprep Kit (Thermo Scientific, Catalog # K0503) and transformed into Stellar competent cells (Clontech, catalog # 636766) to generate plasmid for sequencing. Bacterial colonies were miniprepped using the GeneJET kit and the plasmids were sequence-verified by Genewiz using Sanger sequencing.

Translational fidelity assays

EFT2 variant yeast strains were transformed with dual-luciferase reporter constructs as above. Transformants were selected auxotrophically for the URA3 gene encoded on the reporter plasmid. For each assay, reporter-transformed yeast cultures were inoculated in synthetic-complete medium lacking uracil and allowed to grow to late logarithmic phase. Strains were then diluted to an optical density at 600 nm (OD₆₀₀) of 0.4 in fresh medium and allowed to grow for 2 h. Yeast strains were lysed in Passive Lysis Buffer (Promega) and the lysate was split into three wells of a white 96-well plate. Dual-luciferase assays were conducted with a Promega GloMax Multi+ Detection System using the Dual-Luciferase Reporter Assay System (Promega, Catalog # E1960) as described by the manufacturer. For each strain, the Firefly and Renilla luminescences were found by averaging the luminescence for the three triplicate wells. Ratios of programmed ribosomal frameshifting (PRF) for each strain were calculated by dividing the ratios of Firefly to Renilla luminescence of the test reporter by the read-through luminescence ratio for that strain. Assay replicates were normalized by dividing recoding percentages by that of the WT strain to give fold-change (FC) values. Experiments were repeated with 10 biological replicates for each reporter construct and data were analyzed by ordinary one-way ANOVA using GraphPad Prism. P-values were calculated from Tukey’s multiple comparisons tests between the WT and each variant. ^*P<0.05, ^**P<0.01, ^***P<0.001. Graphs show the median, interquartile range and range of data for each genotype.

Assays of cellular growth

Yeast strains were grown to logarithmic phase in YPAD medium at 30°C then diluted to an OD₆₀₀ of 0.05 in 100 μL of YPAD medium or YPAD medium with hygromycin B, cycloheximide, anisomycin or paromomycin (20 μg/mL, 50 ng/mL, 15 μg/mL and 5 mg/mL, respectively) in two wells of a sterile 96-well plate. Yeast strains were grown in a Synergy HTX plate reader (BioTek, catalog #S1A) at 30°C, where the OD₆₀₀ of each well was taken every 15 min for 48 h, with shaking for 10 min preceding each measurement. The average absorbance of wells containing only medium or medium with translational inhibitor was subtracted from each measurement, which were then averaged across the two triplicate wells for each strain at every time point. Non-linear regression was performed with GraphPad Prism to fit the exponential phase to an exponential growth function. Doubling time was calculated by diving ln(2) by the rate parameter found through non-linear regression. Heat map values were generated by calculating the log₂ of the FC for each variant relative to WT in the presence of each translational inhibitor, which was then subtracted from the log₂(FC) for growth of corresponding variant in rich media. Experiments were repeated with three biological replicates.

Funding

Dutch Organization for Health Research and Development (ZON-MW grants 917-86-319 and 912-12-109 to B.B.A.d.V.) and the National Institutes of Health (R01 GM 117177 to J.D.D.).