|
nif:isString
|
-
This study was conducted with the approval of the University of Washington’s institutional review board. Adults gave written consent, parents gave written permission for their minor children to participate in the study, and additionally, school-age children gave written assent and preschool-age children gave oral assent. Extensive family history interviews were conducted with the participating adults in each family to obtain background information regarding presence of an SSD diagnosis and history of speech therapy services for the interviewed persons themselves as well as other family members. In addition, each adult filled out a questionnaire regarding her/his educational, developmental, and health history. Parents provided details regarding the developmental history of each of their children. Copies of any available written assessment reports were obtained. Affectation status was assigned based on this information and, for young children who had not yet been professionally assessed for the presence of SSD, additionally on performance on standardized and nonstandardized speech measures. In a few cases where sufficient evidence was not available, unknown affectation status was assigned. Family A consists of 24 members in three generations with a familial SSD consistent with CAS (Fig 1; note that the text refers to individual ID numbers with the family identifier as a prefix for clarity). All participants are of European descent, with a small admixture of Japanese descent in six of the participants. Phenotypes and DNA were available for two founders, four adult offspring and their spouses, and 13 grandchildren, 11 of whom could be classified with respect to CAS affectation. The oldest grandchild, A-301, was unable to contribute DNA or participate in the testing; only his developmental history was available. The proband, ID A-304, age 10 years at the time of testing, had a history of severe CAS requiring intense and prolonged speech therapy. The grandfather, ID A-101, reported receiving speech services as an elementary school student whereas the grandmother, A-102, did not report receiving such services. Both grandparents reported individuals biologically related to them with difficulties in the area of speech and language acquisition. No written records were available regarding the grandparents’ speech development. Two of their four participating adult offspring (A-206, A207) had received speech services for five or more years during their early elementary and middle school years. Of the 14 grandchildren, four (A-304, A-305, A-310, A-311) had previously been given a diagnosis of CAS and were currently receiving speech therapy or had completed their course of speech therapy, two (A-312, A-314) were diagnosed based on the speech testing conducted as part of this study, two (A-301, A302) had been diagnosed with a mild speech delay not consistent with CAS as preschoolers, one was too young (15 months) to be diagnosed unambiguously, and five had never received an SSD diagnosis of any type. Details regarding the behavioral findings have been reported previously [42].
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0153864.g001 Family diagram for Family A.Square shape = male, circle shape = female, black fill = affected, white fill = unaffected,? = affectation status unknown, arrow = proband, HCS = Illumina HumanCytoSNP-12v2, HCE/1-0 = Illumina HumanCoreExome-12v1-0_B. Numbers underneath each symbol are individual IDs. Boxes around an ID identify individuals with SNP array data. Filled boxes indicate IDs that also have whole exome sequence data.
Family B also has a history of familial CAS. The family consists of 39 members in five generations, all of European descent except for six individuals with an admixture of African American descent (Fig 2). DNA was available for 14 participants (B-202, B-204, B-205, B-206, B-301, B-302, B-303, B-308, B-311, B-404, B-405, B409, B-410, B-505). Questionnaire and interview information was available for these participants and also for B-506 and B- 507. All of these participants except B-206, B-405, and B-410 participated in behavioral testing.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0153864.g002 Family diagram for Family B.Square shape = male, circle shape = female, black fill = affected, white fill = unaffected,? = affectation status unknown, arrow = proband, HCE/1-1 = Illumina HumanCoreExome-12v1-1_B. Numbers underneath each symbol are individual IDs. Boxes around an ID identify individuals with SNP array data. Filled boxes indicate IDs that also have whole exome sequence data.
The proband, B-403, was 14 years old at the time of testing. He was born at term after an uncomplicated pregnancy and delivery and passed regularly scheduled health, vision, and hearing checks throughout the preschool years. He began receiving speech and expressive language services at age 2;5 due to severe delays in these areas. His diagnosis of CAS at this time was based on severely impaired articulation skills in the presence of severe oral apraxia, not further described in the assessment report. At age 3;8, speech testing with the Structured Photographic Articulation Test II (SPAT-II) [53] resulted in a standard score of 66 (population mean = 100, SD = 30; 1st percentile, far below normal limits), consistent with a severe SSD. His consonant inventory was extremely restricted, consisting of only /d, b, m, n/. Oral motor testing showed deficits in imitating tongue movements. Language testing using Clinical Evaluation of Language Fundamentals-Preschool (CELF-P) [54] showed an Auditory Comprehension standard score of 95 (37th percentile, within normal limits) and an Expressive Comprehension standard score of 50 (1st percentile, far below normal limits). At age 5;8, the proband underwent an occupational therapy evaluation that revealed severe fine motor deficits, especially in grasping and eye-hand coordination skills, qualifying him for services in this area, whereas his gross motor development was found to be within normal limits. Upon entering school, the proband showed difficulty with reading and spelling. For instance, at age 8;10, when tested with the Woodcok-Johnson Tests of Achievement III [55], he obtained a standard score of 63 in Broad Written Language and 64 in Broad Reading (both 1st percentile). The ASHA technical report on CAS lists a small consonant inventory, poorer expressive than receptive language skills, oral apraxia, fine motor deficits, and difficulty with written language as frequently co-occurring conditions (http://www.asha.org/docs/html/PS2007-00277.html). Testing at age 14 showed severe difficulty with nonword imitation [45–47], especially in the form of rearranged phoneme sequences. During diadochokinetic testing, his syllable durations for the monosyllables /pa/, /ta/, and /ka/ were longer than expected for his age, indicative of slow syllable production speed (z = -1.11, -1.84, and -1.13, respectively), but excessively long for the trisyllable /pataka/ (z = -4.32), indicative of severe difficulties with motor planning of complex sequences. Increased difficulty with multisyllabic diadochokinetic tasks, compared to monosyllabic tasks, was reported in our previous studies of children and adults with CAS histories [39, 42, 52, 56]. A similar history of SSD and delays in written language was also reported by his mother, B-302, and another relative, B-311, whereas B-409 reported a history of SSD only in the absence of difficulties with written language, and B-405, a history of difficulties with written language in the absence of SSD. Two family members, B-101 and B-201, both deceased, were reported to have had severely disordered speech during childhood but written records were not available for them or any other members in generations I and II. Because of her concerns that the severe speech disorder could be of genetic origin, the proband’s mother had sought genetic testing for the proband and herself five years prior to participating in this study. According to the clinical report, a microarray analysis of 622 loci using 1,887 BAC clones was performed on DNA derived from peripheral blood. Two interstitial duplications, separated by a normal intervening sequence, were detected on 15q26.3 ([CTD-3210F22, RP11-947PI-631H11]x3, [RP11-262p8, RP11-654A16]x2, [RP11-20G13, CTD-3221M10]x3), summing to 2 Mb in size. The centromeric duplication contains the entire FAM169B gene and the telomeric one, part of the MEF2A gene. Fluorescence in situ hybridization (FISH) analysis using two BAC clones from the two regions (CTD-3210F22, RP11-20G13) showed a pattern consistent with duplication. The same duplication was also found in the mother’s DNA using microarray analysis. The clinical significance of this abnormality could not be determined at the time of the clinical report. Family A provided more direct phenotypic observations and fewer missing samples, compared to Family B. Therefore, the main focus of this study was placed on Family A and data from Family B were used for purposes of comparison.
Complementary genomic approaches were selected because the genetic etiologies of CAS cases in the literature to date include not only a point mutation [13, 14, 37] but also deletions and duplications [26, 35–37]. To investigate the presence of single, relatively rare alleles in the families, we conducted linkage analysis. To detect duplications and/or deletions, we performed copy-number variation (CNV) analysis. Identity-by-descent (IBD) analysis was used to investigate more common segregating variants in Family A, where the grandfather had received speech therapy during childhood but the possibility of childhood speech difficulties in the grandmother could not be ruled out completely. Whole exome sequencing (WES) followed by variant filtering was performed in both families to identify candidate variants. Because of greater statistical power to detect linkage in Family A, compared to Family B, selected candidate variants were genotyped and checked for segregation in Family A only.
DNA was extracted from peripheral blood using standard laboratory procedures. The samples passed quality control checks for sample swaps and incorrectly specified parentage. Because of the phenotypic overlaps with the previously described KE family where a point mutation in the FOXP2 gene caused a severe speech and language disorder [13, 14], this gene was ruled out by exclusion mapping [57] prior to genome-wide analysis procedures. The University of Washington (UW) Center for Mendelian Genomics (CMG) provided single nucleotide polymorphism (SNP) genotypes based on three arrays, as well as WES. In Family A, genotypes for eight participants (Fig 1) were obtained using the Illumina HumanCytoSNP-12v2 array (henceforth HCS) with 298,563 markers. Genotypes for these and eight additional participants (Fig 1) were obtained using the Illumina HumanCoreExome-12v1-0_B array (henceforth HCE/1-0) with 538,448 markers. In Family B, all 14 available DNA samples (Fig 2) were genotyped using the Illumina HumanCoreExome-12v1-1_B (henceforth HCE/1-1) with 542,585 markers. In Family A, DNA samples from two cousins, ID A-304 and A-312, both with a diagnosis of CAS and highly informative based on position in the family pedigree, and samples from the two grandparents were selected for WES (Fig 1). Similarly, B-202, B-311, and B-404 were selected for WES in Family B (Fig 2). Following methods previously described in detail [58], the NimbleGen in-solution SeqCap EZ Exome Library v2.0 (Roche, Basel, Switzerland) was used to capture the exome and adjoining regions, following the manufacturer’s instructions. Short-read sequencing was done on an Illumina HiSeq 2000 platform. For Family B, to evaluate whether the previously reported duplications on 15q26.3 segregated with the disorder, one probe within each of the duplicated regions (Hs02820990_cn, located within FAM169B at bp 98,981,473, and Hs01667266_cn, located within MEF2A at bp 100,250,891), and two control probes (Hs03312008_cn at bp 97,806,447 and Hs05387770_cn at bp 101,256,220) were typed in seven strategically selected samples.
Prior to the SNP-based linkage analyses in the two families, power analysis with 1,000 simulations was conducted using the SLINK package [59, 60]. Under the assumed model of autosomal dominant inheritance, there was one case of nonpenetrance in Family A (A-202) and one in Family B (B-308). As in other genome-wide family-based studies with similar mode of inheritance and evidence for reduced penetrance [61], we assumed parameters of penetrance = 0.50 in the two high-risk genotypes and 0.01 in the low-risk genotype. A simple reduced penetrance model similar to this that allows for sporadic cases works well in situations where the penetrance is unknown but incomplete, outside information to inform the parameters further is not available, and the genotype-phenotype relationship is likely to have at least some complexity [62]. In Family A, the resulting maximum log odds (LOD) score in the power analysis at theta = 0 was 2.75 with the grandfather coded as affected and the grandmother, as unaffected, and 2.45 with both grandparents coded with unknown affectation status. The maximum LOD score in Family B was 2.21. Although both these maxima are below the traditional LOD score requirement of 3 for declaring strong evidence of autosomal linkage [63], this threshold was designed to be conservative, and is actually overly conservative [64, 65]. In addition, with current easy access to sequence data, the original concern about cost of follow-up no longer carries the same concern as it did when the original threshold was proposed. The SNP markers were checked for genotyping errors using the PLINK [66] and PEDSTATS [67] packages and SNPs with genotyping errors were removed from the analysis. Files were formatted for MCMC linkage analysis and an ideal set of SNPs was chosen for a marker panel with the Pedigree-Based Analysis Pipeline (PBAP) [68], targeting marker spacing of 0.5 centimorgan (cM), minor allele frequency (MAF) > 0.2, and LD between markers < 0.04. Minor allele frequencies (MAFs) for the SNP arrays were based on the 1000 Genomes Project Europeans (http://www.1000genomes.org). Genetic locations (cM) were obtained from the Rutgers Maps, Build 134 [69] to establish marker order. These positions were then converted to positions based on the Haldane map function to comply with the requirements of the analysis methods. Affectation status for the grandparents in Family A was conservatively set to unknown; two additional models, each with one grandparent coded as affected, were run. MCMC-based linkage analysis was conducted with the gl_auto and gl_lods programs of the MORGAN 3.2 package [70–72]. The gl_lods program calculates LOD scores based on the phenotype information, penetrance model, and the inheritance vectors that are estimated by gl_auto for each marker given the available pedigree constellation, the marker data, and the genetic map. For gl_auto, the run conditions were 100,000 total run iterations, 15% burn-in iterations, and 2,000 saved iterations. Chromosomal regions retained for further analysis were required to have LODmax scores > 1. The approximate 95% confidence interval (CI) about the peak was defined as the region between the boundaries about the peak where LOD = LODmax− 1 [73]. For CNV analyses in the two families, two sources of input were used. First, genotypes from the exomes were entered into the Copy Number Inference from Exome Reads (CoNIFER) package [74]. For CoNIFER-based CNV discovery, reads from each exome were split into up to two consecutive 36mers and mapped using the single-end mode of mrsFAST [75], then aligned to the hg19 reference genome. Reads per kilobase per million (RPKM) values were calculated and targets with a median RPKM of 1 were excluded. Standardized RPKM values were calculated and a single value decomposition (SVD) algorithm was applied. The output from this analysis, SVD-ZRPKM, was used as the normalized relative copy number of a given exon in a sample. To exclude naturally occurring regions that are duplicated or repeated in the genome, CNVs were filtered using a 50% reciprocal overlap mask. The second source of input for CNV analysis was the set of 16 Illumina HCE/1-0 (Family A) and 14 Illumina HCE/1-1 (Family B) SNP genotypes. Here, we calculated CNVs with two software packages, PennCNV [76] and cnvHap [77]. PennCNV uses a hidden Markov model (HMM) approach, incorporating several types of information including total signal intensity, allelic intensity ratio at each marker, distance between SNPs, and allele frequencies. To avoid biased results, we did not use pedigree information [78]. Like PennCNV, cnvHap uses an HMM approach but additionally incorporates chromosome-wide haplotypic information and cluster-based models of allele frequencies at each marker position. Specifically regarding the previously reported deletion regions on chrs 2 and 16 [26, 35–37], Illumina HCE/1-0 and HCE/1-1 genotypes from two affected members per family were examined for presence of heterozygous genotypes. In Family A only, IBD analysis was performed using the HCS genotypes and the BEAGLE software package, Version 3.3.2 [79]. The SNP base calls were normalized to the forward genomic reference strand and converted to PLINK [66] format with the participants coded as unrelated. The unphased genotypes of 165 unrelated HapMap3 Caucasians (CEU) were merged by PLINK with the genotypes of the eight participants. Duplicated SNPs and SNPs with inconsistent locations were deleted. The genotypes of the participants and the HapMap3 Caucasians were phased as unrelated subjects in BEAGLE. The fastIBD routine of BEAGLE was then used to estimate the shared haplotype frequencies among all pairs, inputting default parameters. Ten haplotype pairs were sampled for each participant during each iteration of phasing. Very rare shared haplotypes between pairs (a threshold of a fastIBD score of 1.0e-10) are likely to be identical by descent. The results of ten independent FastIBD analyses were combined. Exclusive regions of haplotype sharing unique to affected participants were compared to the results from linkage analyses. Specifically, a region shared exclusively by the six affected grandchildren selected for SNP typing was required to be shared in all 15 pairwise comparisons. To determine IBD sharing with one of the grandparents, the region in question was required to be shared by the grandparent and all six selected grandchildren. Selected variants in Family A were tested for segregation using single-marker parametric linkage analysis based on the same parameters as the genome-wide multipoint linkage analysis, here using MERLIN [80] with customized bit size to accommodate the pedigree size. This step was repeated for two additional models, one with the grandfather but not the grandmother coded as affected, and one with the reverse affectation assignment.
Exome Variant Annotation, Filtering, and Single-Variant Genotyping: Exome variants were annotated using ClinVar (http://www.ncbi.nlm.nih.gov/clinvar) and Seattle Seq 137 (http://snp.gs.washington.edu/SeattleSeqAnnotation137/HelpHowToUse.jsp), Variant Effect Predictor, Release 76 [81, 82], and searched with GEMINI [83]. All DNA physical map locations reported in this study refer to the hg19 reference genome. In the exome sequences, an important filtering criterion was position within regions implicated in linkage analysis. Because of the assumption of autosomal dominant inheritance, heterozygous variant genotypes were prioritized. In Family A, the possibility that the children inherited causal variants from either of the two grandparents was considered. Based on the assumption that the causative change is relatively rare in the population, allele frequencies in control exomes obtained to date by the National Heart, Lung and Blood Institute’s (NHLBI) Exome Sequencing Project (ESP) (http://evs.gs.washington.edu/EVS/) and the 1000 Genomes project for European as well as all populations were consulted to prioritize MAFs of 15% or lower. To maximize reliability, variants with read depths < 10 and variants that failed quality control by GATK [84] were excluded. The average read depth of the retained variants was 73.5. Variants were further evaluated with respect to their functions (e.g., missense, coding-synonymous), using the in-house Genome Variation Server, and predicted functional effects (e.g., benign, possibly damaging), using PolyPhen [85] and the Combined Annotation Dependent Depletion (CADD) scores [86]. Genotyping of selected candidate variants was done using polymerase chain reactions in a thermal cycler (DNA Engine Tetrad 2; MJ Research) followed by Sanger sequencing using an ABI 3130xl DNA Analyzer for capillary electrophoresis and ABI BigDye fluorescent dye terminator cycle sequencing kits (Applied Biosystems, Grand Island, NY). In one case, (NIPBL variant), ExoSAP-IT purified PCR products were submitted to Genewiz (Seattle, WA) for Sanger sequencing on ABI 3730xl DNA Analyzers. To obtain genotypes of two variants from C4orf21, ExoSAP-IT purified PCR products from all available Family B members were submitted to GenScript (Piscataway, NJ) for sequencing on 3730xl DNA Analyzers. For exome variant filtering and single-marker linkage analysis, we considered not only variants in the regions with positive evidence for linkage but also variants in previously reported candidate regions for CAS [13–15, 18–24, 26, 35–38] including regions implied in reading/spelling disorders due to reported comorbidities with CAS [9–11]. For exome variant filtering, variants shared by all affected individuals in one or both families were considered most plausible.
BEAGLE https://faculty.washington.edu/browning/beagle/b3.html#beaglev4 cnvHap http://www.imperial.ac.uk/people/l.coin CoNIFER http://conifer.sourceforge.net GEMINI http://faculty.washington.edu/wijsman/software.shtml MORGAN https://www.stat.washington.edu/thompson/Genepi/MORGAN/Morgan.shtml PBAP (Pedigree Based Analysis Pipeline) http://faculty.washington.edu/wijsman/software.shtml PEDSTATS http://csg.sph.umich.edu/abecasis/Pedstats/download/ PennCNV http://penncnv.openbioinformatics.org/en/latest/ PLINK http://pngu.mgh.harvard.edu/~purcell/plink/download.shtml
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}