Background
Autism spectrum disorder (ASD) is characterized by tremendous clinical variability and causal heterogeneity. Historical efforts to behaviorally subtype ASD have been largely unsuccessful due to lack of meaningful treatment implications by subtype and inadequate consensus regarding clinical phenotype [
1,
2]. Recent efforts have targeted the genetic causes of ASD to explore biologically defined subtypes [
3,
4]. Advances in genetic sequencing technology have improved our ability to identify disease-causing mutations [
5]. Chromosomal abnormalities, copy number variants (CNVs), and disruptive single nucleotide variants (SNVs), including nonsense, frameshift, and splice site mutations, have been associated with increased risk of ASD [
6‐
9]. Most recently, work relating ASD risk to de novo disruptive SNVs suggests these single point mutations account for approximately 10% of ASD cases [
6,
8]. These discoveries have prompted a shift in ASD research; instead of using extensive phenotyping prior to sequencing ASD populations, researchers have begun by identifying genes of interest in affected individuals and then exploring phenotype in specific gene cohorts [
10].
The application of this genetics-first approach to subtyping ASD has successfully identified similar medical, behavioral, and dysmorphic features shared by individuals with disruptive variants in high-confidence ASD risk genes, such as
CHD8,
ADNP,
SCN2A, and
DYRK1A (e.g., [
11‐
13]). Dual-specificity tyrosine phosphorylation-regulated kinase 1A, or
DYRK1A, is a highly conserved gene in the Down syndrome critical region of chromosome 21 [
6,
14], and appears to play a major role in brain development, specifically neurogenesis, neural plasticity, and cellular death [
15].
DYRK1A haploinsufficiency was initially identified for its role in intellectual disability, which is clinically defined as childhood onset of significant cognitive and adaptive impairment [
16]. Recurrent disruptions to
DYRK1A have been found in as many as 0.5% of cases with ASD [
14,
17]. In
Drosophila models, truncating mutations to
DYRK1A (
Drosophila ortholog termed the Minibrain (
Mnb) gene) result in microcephaly, including intact but smaller brain structures [
18].
Dyrk1A-null mouse models (−/−) displayed growth deficiencies resulting in mid-gestational death [
19]. Mice heterozygous for
Dyrk1A (+/−) survived to adulthood but presented with reduced growth, developmental delays, motor and learning difficulties, and atypical behaviors, including anxiety [
19,
20]. A consistent clinical phenotype appears in humans. Reported cases to date have exhibited microcephaly and intellectual disability; other features, including seizures, speech and motor delays, feeding difficulties, and distinct facial dysmorphology, have been noted as well [
13,
15,
21‐
24].
Studies relying on medical record review have reported ASD diagnoses in up to 40% of cases with
DYRK1A mutations, but many remaining cases have features consistent with ASD, such as stereotypies, reduced eye contact, and social anxiety [
15,
21]. Few studies have conducted diagnostic evaluations of ASD and ASD-related symptoms as part of a clinical phenotyping battery. Diagnostic rates may be as high as 88% when ASD is directly evaluated as part of the study assessment process [
13].
Literature on
DYRK1A haploinsufficiency supports its association with ASD risk and suggests a complex phenotype that includes distinct dysmorphology as well as cognitive, neurological, and medical impairments. However, studies to date have only reported categorical descriptions of phenotype, which note the presence or absence of a common phenotype, such as ASD or no ASD. Quantitative assessments of ASD-related features in large cohorts and in relation to other ASD cohorts have not been examined. While previously published reports have noted an emerging phenotypic profile, variability in clinical presentation remains. Measurable data on medical, developmental, and behavioral characteristics are needed to better understand small variations in phenotype between individuals. Furthermore, varied clinical presentations of individuals with
DYRK1A mutations have yet to be examined in the context of their familial phenotypic profile as a measure of remaining genetic background. This approach has been applied to other developmental disorders and to CNVs associated with ASD, but has yet to be applied to disruptive SNVs associated with ASD [
25‐
28].
The aim of the proposed study was to examine a large cohort of cases with
DYRK1A mutations, provide a summary of phenotype, and compare recurrent medical and behavioral features to (1) large idiopathic ASD samples and (2) a cohort with disruptive mutations to a different ASD-associated gene,
CHD8. Alongside
DYRK1A,
CHD8 is one of the most recurrent genes with disruptive SNVs implicated in ASD and provides a comparison group ascertained in the same way as the cases with
DYRK1A mutations in this sample [
6,
8]. Detection of phenotypic differences between these two groups could inform understanding of different biological profiles of ASD and illuminate key features unique to each disrupted gene. This study also explored the contribution of genetic background to phenotypic variability among individuals with disruptive
DYRK1A mutations.
Discussion
This study of the DYRK1A haploinsufficiency phenotype, compiling previously published and newly identified cases, confirms a phenotype characterized by microcephaly, intellectual disability, speech delay, motor difficulties, feeding difficulties, and vision abnormalities. A common facial gestalt included deep-set eyes with a hooded appearance, slightly upslanted palpebral fissures, tubular-shaped nose with pronounced broad tip, high nasal bridge, prominent brow with high anterior hairline, retrognathic jaw, and small chin. Dysmorphic feet, including proximal placement of the first toe, syndactyly of the second and third toe, and unusually long and/or crooked toes, and protruding, post-rotated ears with overfolded, thick helices were also commonly observed. Those with de novo disruptive SNVs and chromosomal rearrangements did not differ in clinical features.
Of those case studies where ASD was mentioned and/or evaluated, 43% of probands received an ASD diagnosis. Among 15 cases who received gold-standard ASD assessment, rates increased to 73%. Additionally, features common to ASD, such as stereotypic and anxious behaviors, were noted in many cases where reference to an ASD diagnosis was absent. This suggests rates of ASD in DYRK1A cohorts may be higher in reality than reported in the total sample of DYRK1A cases published to date.
There are several reasons for the potential underestimated prevalence rate of ASD among published DYRK1A cases. First, most previously published cases relied on medical records, which varied greatly in the detail provided and discussion of comprehensiveness of prior evaluations; as such, it is unknown whether ASD was either evaluated and ruled out, or not evaluated at all. Second, it can be difficult to tease apart symptoms of ASD from those of intellectual disability and speech impairments without specialized training and experience with differential diagnosis within developmental disabilities, particularly in children with complex medical histories. Additionally, establishing an ASD diagnosis may not be the most pressing concern for families (and perhaps providers) given the array of impairments and medical conditions that often accompany children with a DYRK1A mutation. As DYRK1A haploinsufficiency continues to be explored within ASD risk, these factors need to be considered when determining rates in this population.
In an effort to situate the
DYRK1A phenotype in the context of ASD, we found the
DYRK1A group (Pub-SNV, UW-SNV, and Pub-CHR groups combined) exhibited significantly higher incidence of key features compared to those with idiopathic ASD: intellectual disability, speech delay, motor difficulties, vision abnormalities, feeding difficulties, and microcephaly. Frequency of these features also significantly differed between the
DYRK1A group and the comparison group with idiopathic ASD and IQ below 70. This is consistent with prior evidence that disruptive SNVs and CNVs often result in significantly more impairing comorbidities than in idiopathic ASD [
6,
8]. Notably, when those with
DYRK1A mutations who were originally ascertained for an ASD diagnosis were compared to the idiopathic group (also ascertained for ASD), the profile remains the same. This provides further support that the phenotype commonly exhibited in individuals with
DYRK1A disruptions and ASD is indeed distinct from idiopathic ASD. The co-occurrence of five or more of these phenotypic features in
DYRK1A cases (79% of total sample, 89% of those ascertained for ASD) provides support for further exploration of
DYRK1A haploinsufficiency in an individual presenting with concerns of ASD and this combination of phenotypic features.
Prior publications of
DYRK1A mutation cases have relied on categorical data to describe clinical phenotype. Our exploration of a quantitative phenotype suggested that
DYRK1A haploinsufficiency is differentiable from idiopathic ASD by measures of cognition, adaptive skills, and head size and distinguishable from a different ASD-associated gene mutation,
CHD8 by head size. It is possible that further phenotypic differences exist which have not been detected by current diagnostic tools given limits to the level of resolution inherent in clinical assessment. Markers relying on quantitative, brain-based measures may reveal gene-specific profiles. For instance, recent work highlights divergent information processing systems for children with 16p11.2 CNVs [
50] and children with an early-emerging disruptive SNV [
51]. Considering the intellectual disability associated with
DYRK1A haploinsufficiency, a passive, noninvasive neuroimaging approach may help illuminate neuroendophenotypes that link the behavioral phenotype to the underlying neural mechanisms.
Exploring quantitative phenotype in UW-SNV participants illuminated phenotypic heterogeneity among individuals. While
DYRK1A mutations significantly impact functioning in a number of domains, the severity of impairment varied among individuals. Family background may, in part, contribute to this variability. While still exploratory, variability in parental phenotype corresponded with variability observed in probands with
DYRK1A haploinsufficiency. Most striking were familial patterns on measures of head circumference. Even with the range of microcephaly, probands with the smallest head sizes were related to parents with smaller head sizes compared to other parents within the UW-SNV group. Physiological characteristics are among the most highly correlated between parents and children in typically developing populations, ranging from 0.5 to 0.7 [
52,
53]. Our findings suggest that, even in the presence of a de novo, disruptive
DYRK1A mutation, parental phenotype may still impact their affected child’s presentation. Of course, secondary genetic events, embryonic or early developmental influences, and treatment must also be considered as potential factors contributing to the variability.
Our findings must be considered in the context of limitations of this study. First, information available for previously published cases varied widely. While some case reports provided detailed record of psychiatric history, others only included medical history, which also varied in its extensiveness. Full assessment history was unknown for previously published cases, raising questions whether phenotypic features left out of a case report were previously ruled out and confirmed absent or were not assessed. These variations highlight the importance of consistency in phenotypic assessment across future DYRK1A phenotype studies to ensure comprehensive and accurate phenotyping efforts. Second, those with DYRK1A mutations who participated in the same quantitative assessment battery remain small in number. Larger sample sizes are indeed needed to better understand the quantitative phenotype of DYRK1A haploinsufficiency and potential variability between affected individuals. Also, while comparison of DYRK1A mutation cases to idiopathic ASD provides important confirmation of distinct comorbidities within ASD, it is important to acknowledge that individuals in the idiopathic group may, with future advances in our understanding of the genetics of ASD, no longer be identified as idiopathic. The idiopathic group analyzed in this study likely represents a population with fewer syndromic features than populations with ASD and other genetic events. Thus, further studies and larger samples of other ASD-associated gene mutations are needed to further distinguish how the DYRK1A haploinsufficiency phenotype differs from that of other disruptive gene events. Future studies of this population should also aim for greater specificity in phenotypic characterization in efforts to better understand DYRK1A haploinsufficiency as a unique clinical profile. Continued study of ASD-associated genes, including DYRK1A, will allow for improved understanding of ASD subtypes and inform future approaches to personalized treatment.