De novo and inherited monoallelic variants in TUBA4A cause ataxia and spasticity.

Alpha-tubulin 4A encoding gene (TUBA4A) has been associated with familial amyotrophic lateral sclerosis (fALS) and fronto-temporal dementia (FTD), based on identification of likely pathogenic variants in patients from distinct ALS and FTD cohorts. By screening a multicentric French cohort of 448 unrelated probands presenting with cerebellar ataxia, we identified ultra-rare TUBA4A missense variants, all being absent from public databases and predicted pathogenic by multiple in-silico tools. In addition, gene burden analyses in the 100,000 genomes project (100KGP) showed enrichment of TUBA4A rare variants in the inherited ataxia group compared to controls (OR: 57.0847 [10.2- 576.7]; p = 4.02 x10-07). Altogether, we report 12 patients presenting with spasticity and/or cerebellar ataxia and harboring a predicted pathogenic TUBA4A missense mutation, including 5 confirmed de novo cases and a mutation previously reported in a large family presenting with spastic ataxia. Cultured fibroblasts from 3 patients harboring distinct TUBA4A missense showed significant alterations in microtubule organisation and dynamics, providing insight of TUBA4A variants pathogenicity. Our data confirm the identification of a hereditary spastic ataxia disease gene with variable age of onset, expanding the clinical spectrum of TUBA4A associated phenotypes.

The exon junction complex is required for DMD gene splicing fidelity and myogenic differentiation.

Deposition of the exon junction complex (EJC) upstream of exon-exon junctions helps maintain transcriptome integrity by preventing spurious re-splicing events in already spliced mRNAs. Here we investigate the importance of EJC for the correct splicing of the 2.2-megabase-long human DMD pre-mRNA, which encodes dystrophin, an essential protein involved in cytoskeletal organization and cell signaling. Using targeted RNA-seq, we show that knock-down of the eIF4A3 and Y14 core components of EJC in a human muscle cell line causes an accumulation of mis-splicing events clustered towards the 3' end of the DMD transcript (Dp427m). This deregulation is conserved in the short Dp71 isoform expressed ubiquitously except in adult skeletal muscle and is rescued with wild-type eIF4A3 and Y14 proteins but not with an EJC assembly-defective mutant eIF4A3. MLN51 protein and EJC-associated ASAP/PSAP complexes independently modulate the inclusion of the regulated exons 71 and 78. Our data confirm the protective role of EJC in maintaining splicing fidelity, which in the DMD gene is necessary to preserve the function of the critical C-terminal protein-protein interaction domain of dystrophin present in all tissue-specific isoforms. Given the role of the EJC in maintaining the integrity of dystrophin, we asked whether the EJC could also be involved in the regulation of a mechanism as complex as skeletal muscle differentiation. We found that eIF4A3 knockdown impairs myogenic differentiation by blocking myotube formation. Collectively, our data provide new insights into the functional roles of EJC in human skeletal muscle.

Titin copy number variations associated with dominant inherited phenotypes.

BACKGROUND: Titinopathies are caused by mutations in the titin gene (TTN). Titin is the largest known human protein; its gene has the longest coding phase with 364 exons. Titinopathies are very complex neuromuscular pathologies due to the variable age of onset of symptoms, the great diversity of pathological and muscular impairment patterns (cardiac, skeletal muscle or mixed) and both autosomal dominant and recessive modes of transmission. Until now, only few CNVs in TTN have been reported without clear genotype-phenotype associations. METHODS: Our study includes eight families with dominant titinopathies. We performed next-generation sequencing or comparative genomic hybridisation array analyses and found CNVs in the TTN gene. We characterised these CNVs by RNA sequencing (RNAseq) analyses in six patients' muscles and performed genotype-phenotype inheritance association study by combining the clinical and biological data of these eight families. RESULTS: Seven deletion-type CNVs in the TTN gene were identified among these families. Genotype and RNAseq results showed that five deletions do not alter the reading frame and one is out-of-reading frame. The main phenotype identified was distal myopathy associated with contractures. The analysis of morphological, clinical and genetic data and imaging let us draw new genotype-phenotype associations of titinopathies. CONCLUSION: Identifying TTN CNVs will further increase diagnostic sensitivity in these complex neuromuscular pathologies. Our cohort of patients enabled us to identify new deletion-type CNVs in the TTN gene, with unexpected autosomal dominant transmission. This is valuable in establishing new genotype-phenotype associations of titinopathies, mainly distal myopathy in most of the patients.

SpliceAI-visual: a free online tool to improve SpliceAI splicing variant interpretation.

SpliceAI is an open-source deep learning splicing prediction algorithm that has demonstrated in the past few years its high ability to predict splicing defects caused by DNA variations. However, its outputs present several drawbacks: (1) although the numerical values are very convenient for batch filtering, their precise interpretation can be difficult, (2) the outputs are delta scores which can sometimes mask a severe consequence, and (3) complex delins are most often not handled. We present here SpliceAI-visual, a free online tool based on the SpliceAI algorithm, and show how it complements the traditional SpliceAI analysis. First, SpliceAI-visual manipulates raw scores and not delta scores, as the latter can be misleading in certain circumstances. Second, the outcome of SpliceAI-visual is user-friendly thanks to the graphical presentation. Third, SpliceAI-visual is currently one of the only SpliceAI-derived implementations able to annotate complex variants (e.g., complex delins). We report here the benefits of using SpliceAI-visual and demonstrate its relevance in the assessment/modulation of the PVS1 classification criteria. We also show how SpliceAI-visual can elucidate several complex splicing defects taken from the literature but also from unpublished cases. SpliceAI-visual is available as a Google Colab notebook and has also been fully integrated in a free online variant interpretation tool, MobiDetails ( https://mobidetails.iurc.montp.inserm.fr/MD ).

Expanding SPTAN1 monoallelic variant associated disorders: From epileptic encephalopathy to pure spastic paraplegia and ataxia.

PURPOSE: Nonerythrocytic αII-spectrin (SPTAN1) variants have been previously associated with intellectual disability and epilepsy. We conducted this study to delineate the phenotypic spectrum of SPTAN1 variants. METHODS: We carried out SPTAN1 gene enrichment analysis in the rare disease component of the 100,000 Genomes Project and screened 100,000 Genomes Project, DECIPHER database, and GeneMatcher to identify individuals with SPTAN1 variants. Functional studies were performed on fibroblasts from 2 patients. RESULTS: Statistically significant enrichment of rare (minor allele frequency < 1 × 10-5) probably damaging SPTAN1 variants was identified in families with hereditary ataxia (HA) or hereditary spastic paraplegia (HSP) (12/1142 cases vs 52/23,847 controls, p = 2.8 × 10-5). We identified 31 individuals carrying SPTAN1 heterozygous variants or deletions. A total of 10 patients presented with pure or complex HSP/HA. The remaining 21 patients had developmental delay and seizures. Irregular αII-spectrin aggregation was noted in fibroblasts derived from 2 patients with p.(Arg19Trp) and p.(Glu2207del) variants. CONCLUSION: We found that SPTAN1 is a genetic cause of neurodevelopmental disorder, which we classified into 3 distinct subgroups. The first comprises developmental epileptic encephalopathy. The second group exhibits milder phenotypes of developmental delay with or without seizures. The final group accounts for patients with pure or complex HSP/HA.

RFC1 nonsense and frameshift variants cause CANVAS: clues for an unsolved pathophysiology.

Cerebellar ataxia, neuropathy and vestibular areflexia syndrome (CANVAS) is an inherited late-onset neurological disease caused by bi-allelic AAGGG pentanucleotide expansions within intron 2 of RFC1. Despite extensive studies, the pathophysiological mechanism of these intronic expansions remains elusive. We screened by clinical exome sequencing two unrelated patients presenting with late-onset ataxia. A repeat-primer polymerase chain reaction was used for RFC1 AAGGG intronic expansion identification. RFC1 mRNA expression was assessed by quantitative reverse transcription-polymerase chain reaction. We identified the first two CANVAS affected patients who are compound heterozygous for RFC1 truncating variants (p.Arg388* and c.575delA, respectively) and a pathological AAGGG expansion. RFC1 expression studies in whole blood showed a significant reduction of RFC1 mRNA for both patients compared to three patients with bi-allelic RFC1 expansions. In conclusion, this observation provides clues that suggest bi-allelic RFC1 conditional loss-of-function as the cause of the disease.

Evaluating the Transition from Targeted to Exome Sequencing: A Guide for Clinical Laboratories.

The transition from targeted to exome or genome sequencing in clinical contexts requires quality standards, such as targeted sequencing, in order to be fully adopted. However, no clear recommendations or methodology have emerged for evaluating this technological evolution. We developed a structured method based on four run-specific sequencing metrics and seven sample-specific sequencing metrics for evaluating the performance of exome sequencing strategies to replace targeted strategies. The indicators include quality metrics and coverage performance on gene panels and OMIM morbid genes. We applied this general strategy to three different exome kits and compared them with a myopathy-targeted sequencing method. After having achieved 80 million reads, all-tested exome kits generated data suitable for clinical diagnosis. However, significant differences in the coverage and PCR duplicates were observed between the kits. These are two main criteria to consider for the initial implementation with high-quality assurance. This study aims to assist molecular diagnostic laboratories in adopting and evaluating exome sequencing kits in a diagnostic context compared to the strategy used previously. A similar strategy could be used to implement whole-genome sequencing for diagnostic purposes.

Clinical phenotype and loss of the slow skeletal muscle troponin T in three new patients with recessive TNNT1 nemaline myopathy.

BACKGROUND: Congenital nemaline myopathies are rare pathologies characterised by muscle weakness and rod-shaped inclusions in the muscle fibres. METHODS: Using next-generation sequencing, we identified three patients with pathogenic variants in the Troponin T type 1 (TNNT1) gene, coding for the troponin T (TNT) skeletal muscle isoform. RESULTS: The clinical phenotype was similar in all patients, associating hypotonia, orthopaedic deformities and progressive chronic respiratory failure, leading to early death. The anatomopathological phenotype was characterised by a disproportion in the muscle fibre size, endomysial fibrosis and nemaline rods. Molecular analyses of TNNT1 revealed a homozygous deletion of exons 8 and 9 in patient 1; a heterozygous nonsense mutation in exon 9 and retention of part of intron 4 in muscle transcripts in patient 2; and a homozygous, very early nonsense mutation in patient 3.Western blot analyses confirmed the absence of the TNT protein resulting from these mutations. DISCUSSION: The clinical and anatomopathological presentations of our patients reinforce the homogeneous character of the phenotype associated with recessive TNNT1 mutations. Previous studies revealed an impact of recessive variants on the tropomyosin-binding affinity of TNT. We report in our patients a complete loss of TNT protein due to open reading frame disruption or to post-translational degradation of TNT.

Association between prophylactic angiotensin-converting enzyme inhibitors and overall survival in Duchenne muscular dystrophy-analysis of registry data.

AIMS: To estimate the effect of prophylactic angiotensin-converting enzyme inhibitors (ACEi) on survival in Duchenne muscular dystrophy (DMD). METHODS AND RESULTS: We analysed the data from the French multicentre DMD Heart Registry (ClinicalTrials.gov: NCT03443115). We estimated the association between the prophylactic prescription of ACEi and event-free survival in 668 patients aged 8 to 13 years, with normal left ventricular function, using (i) a Cox model with intervention as a time-dependent covariate, (ii) a propensity-based analysis comparing ACEi treatment vs. no treatment, and (iii) a set of sensitivity analyses. The study outcomes were overall survival and hospitalizations for heart failure (HF) or acute respiratory failure. Among the 668 patients included in the DMD Heart Registry, 576 (mean age 6.1 ± 2.8 years) were eligible for this study, of whom 390 were treated with ACEi prophylactically. Death occurred in 53 patients (13.5%) who were and 60 patients (32.3%) who were not treated prophylactically with ACEi, respectively. In a Cox model with intervention as a time-dependent variable, the hazard ratio (HR) associated with ACEi treatment was 0.49 [95% confidence interval (CI) 0.34-0.72] and 0.47 (95% CI 0.31-0.17) for overall mortality after adjustment for baseline variables. In the propensity-based analysis, 278 patients were included in the treatment group and 834 in the control group, with 18.5% and 30.4% 12-year estimated probability of death, respectively. ACEi were associated with a lower risk of death (HR 0.39; 95% CI 0.17-0.92) and hospitalization for HF (HR 0.16; 95% CI 0.04-0.62). All other sensitivity analyses yielded similar results. CONCLUSION: Prophylactic ACEi treatment in DMD was associated with a significantly higher overall survival and lower rates of hospitalization for HF.

Objective Evaluation of Clinical Actionability for Genes Involved in Myopathies: 63 Genes with a Medical Value for Patient Care.

The implementation of high-throughput diagnostic sequencing has led to the generation of large amounts of mutational data, making their interpretation more complex and responsible for long delays. It has been important to prioritize certain analyses, particularly those of "actionable" genes in diagnostic situations, involving specific treatment and/or management. In our project, we carried out an objective assessment of the clinical actionability of genes involved in myopathies, for which only few data obtained methodologically exist to date. Using the ClinGen Actionability criteria, we scored the clinical actionability of all 199 genes implicated in myopathies published by FILNEMUS for the "National French consensus on gene Lists for the diagnosis of myopathies using next generation sequencing". We objectified that 63 myopathy genes were actionable with the currently available data. Among the 36 myopathy genes with the highest actionability scores, only 8 had been scored to date by ClinGen. The data obtained through these methodological tools are an important resource for strategic choices in diagnostic approaches and the management of genetic myopathies. The clinical actionability of genes has to be considered as an evolving concept, in relation to progresses in disease knowledge and therapeutic approaches.

ASC-1 Is a Cell Cycle Regulator Associated with Severe and Mild Forms of Myopathy.

OBJECTIVE: Recently, the ASC-1 complex has been identified as a mechanistic link between amyotrophic lateral sclerosis and spinal muscular atrophy (SMA), and 3 mutations of the ASC-1 gene TRIP4 have been associated with SMA or congenital myopathy. Our goal was to define ASC-1 neuromuscular function and the phenotypical spectrum associated with TRIP4 mutations. METHODS: Clinical, molecular, histological, and magnetic resonance imaging studies were made in 5 families with 7 novel TRIP4 mutations. Fluorescence activated cell sorting and Western blot were performed in patient-derived fibroblasts and muscles and in Trip4 knocked-down C2C12 cells. RESULTS: All mutations caused ASC-1 protein depletion. The clinical phenotype was purely myopathic, ranging from lethal neonatal to mild ambulatory adult patients. It included early onset axial and proximal weakness, scoliosis, rigid spine, dysmorphic facies, cutaneous involvement, respiratory failure, and in the older cases, dilated cardiomyopathy. Muscle biopsies showed multiminicores, nemaline rods, cytoplasmic bodies, caps, central nuclei, rimmed fibers, and/or mild endomysial fibrosis. ASC-1 depletion in C2C12 and in patient-derived fibroblasts and muscles caused accelerated proliferation, altered expression of cell cycle proteins, and/or shortening of the G0/G1 cell cycle phase leading to cell size reduction. INTERPRETATION: Our results expand the phenotypical and molecular spectrum of TRIP4-associated disease to include mild adult forms with or without cardiomyopathy, associate ASC-1 depletion with isolated primary muscle involvement, and establish TRIP4 as a causative gene for several congenital muscle diseases, including nemaline, core, centronuclear, and cytoplasmic-body myopathies. They also identify ASC-1 as a novel cell cycle regulator with a key role in cell proliferation, and underline transcriptional coregulation defects as a novel pathophysiological mechanism. ANN NEUROL 2020;87:217-232.

Correction: The genomic and clinical landscape of fetal akinesia.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

The genomic and clinical landscape of fetal akinesia.

PURPOSE: Fetal akinesia has multiple clinical subtypes with over 160 gene associations, but the genetic etiology is not yet completely understood. METHODS: In this study, 51 patients from 47 unrelated families were analyzed using next-generation sequencing (NGS) techniques aiming to decipher the genomic landscape of fetal akinesia (FA). RESULTS: We have identified likely pathogenic gene variants in 37 cases and report 41 novel variants. Additionally, we report putative pathogenic variants in eight cases including nine novel variants. Our work identified 14 novel disease-gene associations for fetal akinesia: ADSSL1, ASAH1, ASPM, ATP2B3, EARS2, FBLN1, PRG4, PRICKLE1, ROR2, SETBP1, SCN5A, SCN8A, and ZEB2. Furthermore, a sibling pair harbored a homozygous copy-number variant in TNNT1, an ultrarare congenital myopathy gene that has been linked to arthrogryposis via Gene Ontology analysis. CONCLUSION: Our analysis indicates that genetic defects leading to primary skeletal muscle diseases might have been underdiagnosed, especially pathogenic variants in RYR1. We discuss three novel putative fetal akinesia genes: GCN1, IQSEC3 and RYR3. Of those, IQSEC3, and RYR3 had been proposed as neuromuscular disease-associated genes recently, and our findings endorse them as FA candidate genes. By combining NGS with deep clinical phenotyping, we achieved a 73% success rate of solved cases.

The FLNC Ala1186Val Variant Linked to Cytoplasmic Body Myopathy and Cardiomyopathy Causes Protein Instability.

Filamin C-related disorders include myopathies and cardiomyopathies linked to variants in the FLNC gene. Filamin C belongs to a family of actin-binding proteins involved in sarcomere stability. This study investigates the pathogenic impact of the FLNC c.3557C > T (p.Ala1186Val) pathogenic variant associated with an early-onset cytoplasmic body myopathy and cardiomyopathy in three unrelated patients. We performed clinical imaging and myopathologic and genetic characterization of three patients with an early-onset myopathy and cardiomyopathy. Bioinformatics analysis, variant interpretation, and protein structure analysis were performed to validate and assess the effects of the filamin C variant. All patients presented with a homogeneous clinical phenotype marked by a severe contractural myopathy, leading to loss of gait. There was prominent respiratory involvement and restrictive or hypertrophic cardiomyopathies. The Ala1186Val variant is located in the interstrand loop involved in intradomain stabilization and/or interdomain interactions with neighbor Ig-like domains. 3D modeling highlights local structural changes involving nearby residues and probably impacts the protein stability, causing protein aggregation in the form of cytoplasmic bodies. Myopathologic studies have disclosed the prominent aggregation and upregulation of the aggrephagy-associated proteins LC3B and p62. As a whole, the Ala1186Val variant in the FLNC gene provokes a severe myopathy with contractures, respiratory involvement, and cardiomyopathy due to protein aggregation in patients' muscles.

Primary mitochondrial disorders and mimics: Insights from a large French cohort.

OBJECTIVE: The objective of this study was to evaluate the implementation of NGS within the French mitochondrial network, MitoDiag, from targeted gene panels to whole exome sequencing (WES) or whole genome sequencing (WGS) focusing on mitochondrial nuclear-encoded genes. METHODS: Over 2000 patients suspected of Primary Mitochondrial Diseases (PMD) were sequenced by either targeted gene panels, WES or WGS within MitoDiag. We described the clinical, biochemical, and molecular data of 397 genetically confirmed patients, comprising 294 children and 103 adults, carrying pathogenic or likely pathogenic variants in nuclear-encoded genes. RESULTS: The cohort exhibited a large genetic heterogeneity, with the identification of 172 distinct genes and 253 novel variants. Among children, a notable prevalence of pathogenic variants in genes associated with oxidative phosphorylation (OXPHOS) functions and mitochondrial translation was observed. In adults, pathogenic variants were primarily identified in genes linked to mtDNA maintenance. Additionally, a substantial proportion of patients (54% (42/78) and 48% (13/27) in children and adults, respectively), undergoing WES or WGS testing displayed PMD mimics, representing pathologies that clinically resemble mitochondrial diseases. INTERPRETATION: We reported the largest French cohort of patients suspected of PMD with pathogenic variants in nuclear genes. We have emphasized the clinical complexity of PMD and the challenges associated with recognizing and distinguishing them from other pathologies, particularly neuromuscular disorders. We confirmed that WES/WGS, instead of panel approach, was more valuable to identify the genetic basis in patients with "possible" PMD and we provided a genetic testing flowchart to guide physicians in their diagnostic strategy.

Identification and in vivo functional investigation of a HOMER2 nonstop variant causing hearing loss.

DFNA68 is a rare subtype of autosomal dominant nonsyndromic hearing impairment caused by heterozygous alterations in the HOMER2 gene. To date, only 5 pathogenic or likely pathogenic coding variants, including two missense substitutions (c.188 C > T and c.587 G > C), a single base pair duplication (c.840dupC) and two short deletions (c.592_597delACCACA and c.832_836delCCTCA) have been described in 5 families. In this study, we report a novel HOMER2 variation, identified by massively parallel sequencing, in a Sicilian family suffering from progressive dominant hearing loss over 3 generations. This novel alteration is a nonstop substitution (c.1064 A > G) that converts the translational termination codon (TAG) of the gene into a tryptophan codon (TGG) and is predicted to extend the HOMER2 protein by 10 amino acids. RNA analyses from the proband suggested that HOMER2 transcripts carrying the nonstop variant escaped the non-stop decay pathway. Finally, in vivo studies using a zebrafish animal model and behavioral tests clearly established the deleterious impact of this novel HOMER2 alteration on hearing function. This study identifies the fourth causal variation responsible for DFNA68 and describes a simple in vivo approach to assess the pathogenicity of candidate HOMER2 variants.

Long-Reads Sequencing Strategy to Localize Variants in TTN Repeated Domains.

Titin protein is responsible for muscle elasticity. The TTN gene, composed of 364 exons, is subjected to extensive alternative splicing and leads to different isoforms expressed in skeletal and cardiac muscle. Variants in TTN are responsible for myopathies with a wide phenotypic spectrum and autosomal dominant or recessive transmission. The I-band coding domain, highly subject to alternative splicing, contains a three-zone block of repeated sequences with 99% homology. Sequencing and localization of variants in these areas are complex when using short-reads sequencing, a second-generation sequencing technique. We have implemented a protocol based on the third-generation sequencing technology (long-reads sequencing). This new method allows us to localize variants in these repeated areas to improve the diagnosis of TTN-related myopathies and offer the analysis of relatives in postnatal or in prenatal screening.

A National French Consensus on Gene List for the Diagnosis of Charcot-Marie-Tooth Disease and Related Disorders Using Next-Generation Sequencing.

Next generation sequencing (NGS) is strategically used for genetic diagnosis in patients with Charcot-Marie-Tooth disease (CMT) and related disorders called non-syndromic inherited peripheral neuropathies (NSIPN) in this paper. With over 100 different CMT-associated genes involved and ongoing discoveries, an important interlaboratory diversity of gene panels exists at national and international levels. Here, we present the work of the French National Network for Rare Neuromuscular Diseases (FILNEMUS) genetic diagnosis section which coordinates the seven French diagnosis laboratories using NGS for peripheral neuropathies. This work aimed to establish a unique, simple and accurate gene classification based on literature evidence. In NSIPN, three subgroups were usually distinguished: (1) HMSN, Hereditary Motor Sensory Neuropathy, (2) dHMN, distal Hereditary Motor Neuropathy, and (3) HSAN, Hereditary Sensory Autonomic Neuropathy. First, we reported ClinGen evaluation, and second, for the genes not evaluated yet by ClinGen, we classified them as "definitive" if reported in at least two clinical publications and associated with one report of functional evidence, or "limited" otherwise. In total, we report a unique consensus gene list for NSIPN including the three subgroups with 93 genes definitive and 34 limited, which is a good rate for our gene's panel for molecular diagnostic use.

Analysis of Dp71 contribution in the severity of mental retardation through comparison of Duchenne and Becker patients differing by mutation consequences on Dp71 expression.

The presence of variable degrees of cognitive impairment, extending from severe mental retardation to specific deficits, in patients with dystrophinopathies is a well-recognized problem. However, molecular basis underlying mental retardation and its severity remain poorly understood and still a matter of debate. Here, we report one of the largest study based on the comparison of clinical, cognitive, molecular and expression data in a large cohort of 81 patients affected with Duchenne muscular dystrophy (DMD) and Becker muscular dystrophy (BMD) bearing mutations predicted to affect either all dystrophin products, including Dp71 or all dystrophin products, except Dp71. In addition to the consistent data defining molecular basis underlying mental retardation in DMD, we show that BMD patients with MR have mutations that significantly affect Dp71 expression or with mutations located in exons 75 and 76. We also show that mutations upstream to exon 62, with DMD phenotype, predicted to lead to a loss-of-function of all dystrophin products, except Dp71 isoform, are associated, predominantly, with normal or borderline cognitive performances. Altogether, these reliable phenotype-genotype correlations in combination with Dp71 mRNA and protein expression studies, strongly indicate that loss-of-function of all dystrophin products is systematically associated with severe form of MR, and Dp71 deficit is a factor that contributes in the severity of MR and may account for a shift of 2 SD downward of the intelligence quotient.