Dissecting the Clinical Heterogeneity of Autism Spectrum Disorders through Defined Genotypes Authors: Hilgo Bruining, Leo de Sonneville, Hanna Swaab, Maretha de Jonge, Martien Kas, Herman van Engeland, Jacob Vorstman Published: 2010, PLOS ONE DOI: 10.1371/journal.pone.0010887 ABSTRACT The etiology of autism spectrum disorders (ASD) is largely determined by different genetic factors of variable impact. This genetic heterogeneity could be a factor to explain the clinical heterogeneity of autism spectrum disorders. Here, a first attempt is made to assess whether genetically more homogeneous ASD groups are associated with decreased phenotypic heterogeneity with respect to their autistic symptom profile. The autistic phenotypes of ASD subjects with 22q11 deletion syndrome (22q11DS) and ASD subjects with Klinefelter Syndrome (KS) were statistically compared to the symptom profile of a large (genetically) heterogeneous ASD sample. Autism diagnostic interview-revised (ADI-R) variables were entered in different statistical analyses to assess differences in symptom homogeneity and the feasibility of discrimination of group-specific ASD-symptom profiles. The results showed substantially higher symptom homogeneity in both the genetic disorder ASD groups in comparison to the heterogeneous ASD sample. In addition, a robust discrimination between 22q11-ASD and KS-ASD and idiopathic ASD phenotypes was feasible on the basis of a reduced number of autistic scales and symptoms. The lack of overlap in discriminating subscales and symptoms between KS-ASD and 22q11DS-ASD suggests that their autistic symptom profiles cluster around different points in the total diagnostic space of profiles present in the general ASD population. The findings of the current study indicate that the clinical heterogeneity of ASDs may be reduced when subgroups based on a specific genotype are extracted from the idiopathic ASD population. The current strategy involving the widely used ADI-R offers a relatively straightforward possibility for assessing genotype-phenotype ASD relationships. Reverse phenotype strategies are becoming more feasible, given the accumulating evidence for the existence of genetic variants of large effect in a substantial proportion of the ASD population. INTRODUCTION Autism spectrum disorders (ASDs) delineate a group of behaviorally-defined disorders including autism, PDD-NOS, and Asperger syndrome. Many efforts are being made to address the clinical heterogeneity of ASDs. At the same time, the diversity of genetic findings in the past decade indicate that ASDs should also be considered genetically heterogeneous [1] , [2] . This raises the question to what extent the clinical heterogeneity can be explained by the underlying genetic heterogeneity of ASDs. In this study we will address this issue through the assessment of the homogeneity of the ASD phenotype in genetically more homogenous samples. In addition to the growing number of genetic ASD susceptibility loci with small effect sizes, recent studies have described new “causative” genetic variants in ASDs that are assumed to have a large impact on ASDs [3] , [4] . They are thought account for about 10–20% of ASD cases [1] , [4] , [5] . These risk variants are likely to show incomplete penetrance and imperfect segregation with disease as most variants have also been observed in non-autistic controls [5] , [6] . Furthermore, several ASD variants have been shown to cause brain disorders other than ASD, including schizophrenia, mental retardation and epilepsy [1] , [6] . This combination of incomplete penetrance and pleiotropic phenotypes could indicate that these loci cause a global disruption in brain development, making it more vulnerable to develop a range of different brain disorders. Efforts are required to distinguish distinct aspects of those brain disorders that are caused by these genetic variants with large effect, versus aspects that result from various other (environmental and/or genetic) hits. A logical starting point would be to assess whether at all, ASD cases ascertained for a particular genetic variant display distinct autistic characteristics. This model can be considered probable when cases carrying the same genetic variant are found to share particular (combinations of) symptoms in higher frequencies than most cases in the idiopathic ASD population. The probability of genetic ASD subphenotypes can be illustrated by Rett syndrome (RTT). A more homogeneous profile of autistic symptoms together with non-autistic symptoms has led to the description of the RTT genetic subphenotype that is formally classified in the DSM-IV-TR as an ASD subtype. RTT is a progressive neurodevelopmental disorder that manifests in girls during early childhood [7] , [8] . Mutations in MECP2 gene are found in more than 95% of classic RTT cases [9] , [10] . Patients with RTT appear to develop normally up to 6–18 months of age. Deceleration of head growth is often the presenting symptom. This is often accompanied by general growth retardation, weight loss, and a weak posture and ataxia [7] . Social withdrawal and loss of language become apparent at early age. Most patients progressively develop stereotypic hand wringing or washing movements. Other frequent autistic features include expressionless face, hypersensitivity to sound, lack of eye-to-eye contact and unresponsiveness to social cues [8] . This illustrates that both specific autistic and non-autistic features characterize RTT. Importantly, the RTT autistic features are also present among the general population of autistic individuals though probably in a much lower frequency. The modest recurrence of most identified large risk variants so far precludes the inclusion of adequate carrier numbers to evaluate the specificity of the autistic subphenotype per variant. Each of these variants on its own represents only a small proportion (at most 1–2%) of the ASD population [4] , [11] . Genetic disorders such as RTT that are frequently associated with ASD have associated features such as congenital malformations or somatic disorders that enhance the chance of clinical detection. Therefore, a focus on ASD subjects ascertained for particular a well defined genetic disorder enables the inclusion of larger numbers. Importantly, similar to the newly discovered genetic variants of large effects, most genetic disorders are associated with ASD only in a fraction of affected subjects, thus the defining variants in these disorders also display incomplete penetrance. This warrants a focus on subsets of individuals with a particular genetic disorder that are diagnosed with ASD which could possibly precipitate the impact of a particular variant on autistic symptomatology [1] . As a proof of concept we studied the ASD phenotype of ASD subjects with 22q11 deletion syndrome (22q11DS) and Klinefelter syndrome (KS, 47 XXY). 22q11DS and KS subjects without an ASD classification were excluded. 22q11DS and KS are relatively frequent disorders affecting 1–2,000–4,000 and 1–700 respectively [12] – [14] . Both are clinically defined genetic disorders like RTT and increased rates of ASD have been described in both 22q11DS and KS subjects [15] – [17] . The presence of both disorders has also previously been described among populations of subjects with ASD [18] – [20] . The structure of the ASD phenotype associated with 22q11DS and KS was compared to a large a large genetically heterogeneous sample of ASD subjects in different statistical analyses involving standard autistic measurements. The analyses aimed to assess differences in symptom homogeneity and the feasibility of differentiation of group-specific ASD-symptom profiles. RESULTS We compared the autism symptom profile of 39 subjects with 22q11DS and ASD and 14 subjects with KS and ASD to a large genetically heterogeneous sample of 372 ASD subjects (further referred to as the heterogeneous ASD sample). Autism diagnostic interview-revised (ADI-R) algorithm variable scores were entered in the analyses to evaluate differences in symptom homogeneity and the feasibility of ASD genetic subphenotype discrimination. Table 1 shows the basic characteristics of the 22q11DS-ASD, KS-ASD and heterogeneous ASD samples. Correlations in the heterogeneous ASD sample between IQ scores and the different ADI-R outcome scores were virtually negligible (−0.1040. The final sample consisted of 372 verbal subjects. Study participants ranged in age from 4 to 20 years. Similar to the ASD cases obtained from the 22q11 and XXYDS cohorts, all subjects out of the heterogeneous ASD sample had been evaluated in consensus meetings to confirm ASD diagnosis through the interviews and all subjects met ADI-R thresholds. IQ had been assessed by means of the Dutch versions of the Wechsler scales (WPSSI WISC III and WAIS) [25] – [27] in the KS and 22q11DS sample and in a significant part of the ASD heterogeneous group (65%). IQ scores of the heterogeneous ASD sample that could not be assessed with the Wechsler scales have been assessed with the RAVEN Progressive Matrices [28] the Mullen Scales of Early Learning [29] or the Snijders-Oomen non-verbal intelligence test-Revised [30] . No difference in intelligence level between those with and those without ASDs had been found in both the 22q11DS and KS samples in the original psychiatric surveys (see [16] , [17] ). MEASUREMENTS ADI-R subscale and symptom variables were entered in the statistical analyses for phenotype comparisons. The ADI–R is an established ‘gold standard’ in diagnostic/phenotypic evaluations of autism. It is an extensive clinical interview administered to the parents. The interview focuses on the three core or so called “content” domains of autism (i.e. qualitative abnormalities in social interaction (S), qualitative abnormalities in communication (C) and stereotyped and repetitive behaviors (R) [24] . ADI-R items are coded for these domains and also for an “age of onset” domain. A classification of an autism spectrum disorder is applied when scores in all domains are met or when scores are met in two core domains and meet criteria on the “age of onset” domain, but are one point away from meeting autism criteria in the remaining core domain. Reliability of the ADI-R in a population with mild to moderate mental retardation has been established [31] . The ADI-R may also be used to assess profiles of autistic symptomatology [32] , [33] . The ADI–R algorithm is composed of 37 symptom “items”. These items were originally selected as a minimum for optimal ASD classification [24] . ADI-R labels consist of 2 to 5 items and are directly related to the DSM-IV-TR criteria of an Autistic Disorder. As a result, 12 ADI-R labels are used. Each ADI-R domain consists of 4 labels, eg S1-4, C1-4 and R1-4. Items are coded as 0 (ASD behavioural symptom specified not present), 1 (specified behaviour not sufficient to code “2”) or 2 (specified ASD symptom present). Maximum label scores thus range from 4–10. An overview of the description of the ADI-R items, labels and the ADI-R domains of the algorithm is provided in Tables S1 and S2 . STATISTICAL ANALYSES Symptom homogeneity was operationalized as the (inverse of the) mean number of ADI-R algorithm items within each ASD group on which the subjects scored clearly in the autistic range, i.e. ADI-R item score  = 2. Thus, per subject, the number of items with score  = 2 were counted. A lower number of ADI-R items that reached the autism criterion can be considered indicative of a relative reduction in symptom heterogeneity. Differences between groups in number of ADI-R items on which the autism criterion was reached were analyzed by means of a univariate analysis of variance, with post hoc Bonferroni multiple comparisons. Discriminant analyses (DA) were performed to determine whether the genetic disorder ASD subsamples could be differentiated from the heterogeneous ASD sample on the basis of ADI-R variables. The analyses addressed the question to what extent ADI-R label and/or ADI-R symptom item profiles could successfully discriminate 22q11DS or KS ASD profiles from the heterogeneous ASD sample profile (i.e., 22q11DS+ASD vs. heterogeneous ASD, and KS+ASD vs. heterogeneous ASD). In addition, three group DAs were performed to explore the separation of the three groups by means of 2 discriminant functions. For all discriminant analyses, stepwise Wilks' lambda was used, with probability of F for entry and removal of variables set at 0.5 and 0.10 respectively. For classification the within-group covariance matrices were used and prior probability was set to equal for all groups. SUPPORTING INFORMATION ADI-R algorithm items sorted by labels and domains. (0.07 MB DOC) ADI-R algorithm items sorted by number. (0.05 MB DOC) Classification matrix of the 3-group discriminant analysis of heterogeneous ASD versus KS-ASD versus 22q11DS-ASD. (0.04 MB DOCX) Description and discriminant function coefficients of ADI-R items extracted in the 3-group discriminant analysis of heterogeneous ASD versus KS-ASD versus 22q11DS-ASD. (0.04 MB DOC) Overview of extracted ADI-R items with subsequent labels in the different discriminant analyses. DA1  = 22q11DS-ASD versus heterogeneous ASD. DA2  =  KS-ASD versus heterogeneous ASD. D3  = 3-group group comparison of 22q11DS-ASD versus KS-ASD versus heterogeneous ASD. (0.05 MB DOC) Results for stability calculations of DAs. ADI-R items extracted in the 4 additional DAs, sorted by correspondence/item number. To verify whether the stepwise analysis of the three groups DA did provide stable solution, the stepwise DA to separate the three groups has been repeated another four times, with half of the sample of subjects with heterogeneous ASD. Thereto the sample was divided in four quarters, called Q1 to Q4. The DA has been performed with inclusion of the heterogeneous subsamples Q1+Q2, Q3+Q4, Q1+Q3, and Q2+Q4. The ‘solutions’ are compared with each other and with the result of the DA presented in the paper that included the total sample of subjects with heterogeneous ASD. The comparison is focused on the number and type of items that are extracted. Table S6 shows that the solutions are highly similar. The number of items extracted varies between 10 and 15. There are 8 items that appear in each DA, there is 1 items that appears in 3 DAs, there are 7 items that appear in 2 DAs, there is 1 item appearing in only one DA. Comparing the four solutions with the results of the original DA presented in the paper shows that all of the 12 items extracted in this DA, all show up in one or more one of the other DAs. Eight items of the original DA show up in all other analyses, the other four items of the original DA appear in at least two of the other DAs. (0.06 MB DOC) REFERENCES [1] Abrahams BS Geschwind DH 2008 Advances in autism genetics: on the threshold of a new neurobiology. Nat Rev Genet 9 341 355 [2] Muhle R Trentacoste SV Rapin I 2004 The genetics of autism. Pediatrics 113 e472 e486 [3] Beaudet AL 2007 Autism: highly heritable but not inherited. Nat Med 13 534 536 [4] Sebat J Lakshmi B Malhotra D Troge J Lese-Martin C 2007 Strong association of de novo copy number mutations with autism. Science 316 445 449 [5] Bucan M Abrahams BS Wang K Glessner JT Herman EI 2009 Genome-wide analyses of exonic copy number variants in a family-based study point to novel autism susceptibility genes. PLoS Genet 5 e1000536 [6] Mefford HC Sharp AJ Baker C Itsara A Jiang Z 2008 Recurrent rearrangements of chromosome 1q21.1 and variable pediatric phenotypes. N Engl J Med 359 1685 1699 [7] Ben Zeev GB 2007 Rett syndrome. Child Adolesc Psychiatr Clin N Am 16 723 743 [8] Nomura Y Segawa M 2005 Natural history of Rett syndrome. J Child Neurol 20 764 768 [9] Trappe R Laccone F Cobilanschi J Meins M Huppke P 2001 MECP2 mutations in sporadic cases of Rett syndrome are almost exclusively of paternal origin. Am J Hum Genet 68 1093 1101 [10] Amir RE Van den Veyver IB Wan M Tran CQ Francke U 1999 Rett syndrome is caused by mutations in X-linked MECP2, encoding methyl-CpG-binding protein 2. Nat Genet 23 185 188 [11] Freitag CM 2007 The genetics of autistic disorders and its clinical relevance: a review of the literature. Mol Psychiatry 12 2 22 [12] Abramsky L Chapple J 1997 47,XXY (Klinefelter syndrome) and 47,XYY: estimated rates of and indication for postnatal diagnosis with implications for prenatal counselling. Prenat Diagn 17 363 368 [13] Bojesen A Juul S Gravholt CH 2003 Prenatal and postnatal prevalence of Klinefelter syndrome: a national registry study. J Clin Endocrinol Metab 88 622 626 [14] Tezenas Du MS Mendizabai H Ayme S Levy A Philip N 1996 Prevalence of 22q11 microdeletion. J Med Genet 33 719 [15] Antshel KM Aneja A Strunge L Peebles J Fremont WP 2007 Autistic spectrum disorders in velo-cardio facial syndrome (22q11.2 deletion). J Autism Dev Disord 37 1776 1786 [16] Bruining H Swaab H Kas M van Engeland H 2009 Psychiatric characteristics in a self-selected sample of boys with Klinefelter syndrome. Pediatrics 123 e865 e870 [17] Vorstman JA Morcus ME Duijff SN Klaassen PW Heineman-de Boer JA 2006 The 22q11.2 deletion in children: high rate of autistic disorders and early onset of psychotic symptoms. J Am Acad Child Adolesc Psychiatry 45 1104 1113 [18] Konstantareas MM Homatidis S 1999 Chromosomal abnormalities in a series of children with autistic disorder. J Autism Dev Disord 29 275 285 [19] Marshall CR Noor A Vincent JB Lionel AC Feuk L 2008 Structural variation of chromosomes in autism spectrum disorder. Am J Hum Genet 82 477 488 [20] Wassink TH Losh M Piven J Sheffield VC Ashley E 2007 Systematic screening for subtelomeric anomalies in a clinical sample of autism. J Autism Dev Disord 37 703 708 [21] Constantino JN Davis SA Todd RD Schindler MK Gross MM 2003 Validation of a brief quantitative measure of autistic traits: comparison of the social responsiveness scale with the autism diagnostic interview-revised. J Autism Dev Disord 33 427 433 [22] Lord C Rutter M Goode S Heemsbergen J Jordan H Mawhood L 1989 Autism diagnostic observation schedule: a standardized observation of communicative and social behavior. J Autism Dev Disord 19 185 212 [23] Frazer KA Murray SS Schork NJ Topol EJ 2009 Human genetic variation and its contribution to complex traits. Nat Rev Genet 10 241 251 [24] Lord C Rutter M LeCouteur A 1994 Autism Diagnostic Interview-Revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. J Autism Dev Disord 24 659 685 [25] Wechsler D 1989 Wechsler preschool and primary scale of intelligence-revised. New York Psychological Corporation [26] Wechsler D 1997 Wechsler Adult Intelligence Scale-Third Edition. San Antonio, TX Psychological Corporation [27] Wechsler D Kort W Compaan EL Bliechrodt N Resig WC 2002 WISC-III NL, Handleiding. London Psychological coorporation unlimited [28] Raven JC 1995 Colored progressive matrices sets I and II. Oxford Oxford Psychologists Press Ltd [29] Mullen EM 1995 Mullen Scales of Early Learning. Circle Pines, MN American Guidance Service Inc [30] Snijders JT Tellegen PJ Winkel M Laros JA 2009 SON-R 2, 5-7 Niet-verbale Intelligentietest - Revisie [SON-R 2, 5-7 Snijders-Oomen Non-verbal Intelligencetest-Revised]. Lisse, the Netherlands Swets & Zeitlinger [31] de Bildt A Sytema S Ketelaars C Kraijer D Mulder E 2004 Interrelationship between Autism Diagnostic Observation Schedule-Generic (ADOS-G), Autism Diagnostic Interview-Revised (ADI-R), and the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) classification in children and adolescents with mental retardation. J Autism Dev Disord 34 129 137 [32] Brune CW Kim SJ Salt J Leventhal BL Lord C 2006 5-HTTLPR Genotype-Specific Phenotype in Children and Adolescents With Autism. Am J Psychiatry 163 2148 2156 [33] Kates WR Antshel KM Fremont WP Shprintzen RJ Strunge LA 2007 Comparing phenotypes in patients with idiopathic autism to patients with velocardiofacial syndrome (22q11 DS) with and without autism. Am J Med Genet A 143A 2642 2650