CDK13 -related disorder: a deep characterization of speech and language abilities and addition of 33 novel cases

Speech and language impairments are central features of CDK13 -related disorder. While pathogenic CDK13 variants have been associated with childhood apraxia of speech (CAS), a systematic characterisation of communication has not been conducted. Here we examined speech, language, non-verbal communication skills, social behaviour and health and development in 41 individuals with CDK13 -related disorder from 10 countries (male = 22, median-age 7 years 1 month, range 1 – 25 years; 33 novel). Most participants used augmentative and alternative communication (AAC) in early childhood (24/41). CAS was common (14/22). Performance varied widely across intellectual ability, social behaviour and expressive language skills, with participants ranging from within average through to the severely impaired range. Receptive language was signi ﬁ cantly stronger than expressive language ability. Social motivation was a relative strength. In terms of a broader health phenotype, a quarter had one or more of: renal, urogenital, musculoskeletal, and cardiac malformations, vision impairment, ear infections and/or sleep disturbance. All had gross and ﬁ ne motor impairments (41/41). Other conditions included mild-moderate intellectual disability (16/22) and autism (7/41). No genotype-phenotype correlations were found. Recognition of CAS, a rare speech disorder, is required to ensure appropriately targeted therapy. The high prevalence of speech and language impairment underscores the importance of tailored speech therapy, particularly early access to AAC supports.


INTRODUCTION
CDK13 is part of the family of over 30 cyclin-dependent kinases (CDKs) expressed in humans [1].CDK13 is involved in transcription and posttranscriptional processing [2] and plays a critical role in embryonic development [3].CDK13 expression is detectable in the heart, brain and craniofacial area [3].
The neurodevelopmental profile includes average intellectual ability through to moderate intellectual disability (ID), autism spectrum disorder (hereafter autism), attention-deficit/hyperactivity disorder (ADHD), epilepsy and sleep disturbances [14].
Speech and language are reported as among the most commonly impacted areas of neurodevelopment in CDK13related disorder [5,6,8,10,13,14].Yet whilst communication difficulties are ubiquitously reported, there has been no specificity to the clinical diagnoses, with very general terms such as 'speech and language delay' being used.Nor have studies used standardised measures/assessment protocols.A pathogenic variant in CDK13 was recently associated with a rare and severe speech disorder, childhood apraxia of speech (CAS), in a gene discovery cohort of children ascertained for CAS [13].However, there has been no systematic reverse phenotyping evaluation of speech or language deficits in a cohort of individuals with pathogenic CDK13 variants to date, to confirm this association with CAS.Further, the absence of a comprehensive speech and language evaluation in this population limits prognostic counselling and the provision of targeted intervention.Here, we systematically characterise speech and language abilities, and examine possible genotype-phenotype correlations, in children with CDK13-related disorder using standardised outcome measures.

METHODS Participants
Inclusion criteria were a molecularly confirmed pathogenic diagnosis of CDK13-related disorder.Exclusion criteria were the existence of other pathogenic variants in other genes associated with neurodevelopmental disorders.Participants were recruited internationally via an online CDK13 support group or via their treating clinical geneticist from French, Dutch, German, English and Spanish speaking backgrounds.The Royal Children's Hospital Human Research Ethics Committee provided ethical approval (HREC 37353 A).Caregivers provided written informed consent for their children to participate, even in the case of the young adults in the study.

Health and development
Families completed caregiver questionnaires concerning individuals' health and developmental history (Supplementary Table 1).Caregiver questionnaires were completed in the participants' language: English, French, Dutch, German, and Spanish.Results were confirmed with a case history via telehealth and provision of additional reports (e.g., cognitive assessments, electroencephalogram results, and autism diagnostic reports), a successful method employed previously [15][16][17][18].All Englishspeaking, verbal participants (i.e., those who used primarily spoken words to communicate) also completed a telehealth assessment with a university-trained speech pathologist.

Adaptive behaviour and motor skills
The Vineland Adaptive Behaviour Scales (VABS II/III) caregiver version, was completed online for English-, Spanish-(third edition) and French-speaking (second edition) participants [19,20].The VABS II/III provides standardised scores for communication, socialisation, self-care, activities of daily living, motor skills and an overall adaptive behaviour score (that does not include the motor skills subtest).

Language and social communication
The VABS II/III communication domain, with receptive, expressive and written skill subdomains, assessed language in English-, French-and Spanish-speaking individuals.English-and Dutch-speaking caregivers of participants younger than 2 years-old completed the Communication and Symbolic Behaviour Scales Developmental Profile (CSBS-DP) standardised questionnaire [21].Children 4-16 years old completed the Children's Communication Checklist -Second Edition (CCC-2) [22].The CSBS-DP and CCC-2 assess speech, receptive and expressive language, non-verbal communication and social communication abilities.These tools were not available in other languages.
The Social Responsiveness Scale-2 (SRS-2) was completed by Englishand Dutch-speaking families [23].The SRS-2 caregiver questionnaire is standardised across three versions from pre-school-(>2 years) and schoolaged children, to adulthood.The SRS-2 measures social behaviour based on the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-V) autism diagnostic criteria, and examines domains of [23]: social awareness, social cognition, social communication, social motivation, restricted interests, and repetitive behaviour [24].

Alternative communication methods and therapy
Minimally verbal children were defined as having less than 50 spoken words [17,25] and were assessed using the Inventory of Potential Communicative Acts (IPCA) caregiver questionnaire [26].This assessment investigates communication behaviours used by individuals of all ages, such as; facial expression, body movement, gesture, and augmentative and alternative communication (AAC) (e.g., sign language, communication devices) across a range of functions including greeting, protesting, and commenting.Caregiver surveys provided information on current therapy goals and AAC systems.

Speech
Verbal children were assessed using standardised tools which examined performance across speech domains of: articulation, phonology, stuttering, dysarthria and CAS.
Articulation disorder (distorted production of a speech sound, e.g., a lisp), phonological delay (where a child is persisting in the use of speech sound error patterns made by >10% of younger children, e.g., fronting of fricatives or velars; gliding, etc) and phonological disorder (atypical speech sound error patterns, defined as errors made by <10% of children in the general population, e.g., initial consonant deletion, backing of sounds) were diagnosed using the Phonology and Inconsistency subtests of the Diagnostic Evaluation of Articulation and Phonology (DEAP) [27] and confirmed during a five minute conversational speech sample.Stuttering was measured with a Likert scale rating from 0 (no stuttering) to 9 (severe stuttering) based on the conversational sample [28].Ratings for CAS were made using the American Speech and Hearing Association CAS Technical Report Protocol consensus features [29].CAS diagnoses are made based on three criteria: (i) inconsistent speech production; (ii) disrupted and prolonged co-articulatory transitions; and (iii) prosodic errors.Features of these three criteria were operationally defined and rated using a checklist, previously applied in other populations with genetic disorder [15-18, 30, 31].Dysarthria was rated in the presence of neuromuscular tone disruption to one or more of the sub-systems for speech (e.g., phonatory, articulatory) as well as the presence of specific dysarthric features (e.g., hypernasality), as rated using the Mayo Clinic Dysarthria Classification System [32].Dysarthria and CAS ratings were made based on these operationalised criteria using single word responses to the DEAP Phonology and Inconsistency subtests, the conversational speech sample, and diadochokinetic speech tasks (e.g., 'pataka') [27,30].Clinician and caregiver reports documented speech diagnoses for non-English speaking individuals, who were not able to be directly assessed.
The Intelligibility in Context Scale (ICS) [33] was completed by caregivers to assess how well a participant is understood (intelligibility) based on their speech in the past month, with different communication partners (e.g., friends, family members) on a scale of 1 (never understood) to 5 (always understood) (Supplementary Fig. 1).

Feeding and nonspeech oral motor skills
English-speaking children aged 6 months to 7 years completed the Child Oral and Motor Proficiency Scale (ChOMPS) [34].This assessment separates eating and drinking skills into: complex movement patterns (e.g., licking food from lips), basic movements (e.g., sitting), oral motor coordination (e.g., moving jaw to chew), and fundamental oral skills (e.g., closing lips).Caregivers of children who drooled completed the Drooling Impact Scale [35], whereby the frequency and impact of drooling was rated 1 (none) through 10 (all the time).The structure and function of the oral articulators was assessed [36] to support interpretation of the speech and feeding results.

Statistical analyses
Non-parametric statistical analyses were conducted due to the data not being normatively distributed.A Wilcoxon Signed Rank tests compared individual differences between VABS II/III receptive and expressive language scores, CCC-2 domains, and SRS-2 domains.To explore genotype-phenotype associations, a Mann Whitney test compared VABS II/III adaptive behaviour and communication scores between groups with different variants.A Kruskal-Wallis test compared VABS II/III domains.Ages, VABS II/III, CCC-2, CSBS-DP, SRS-2, and ICS data were reported using descriptive statistics detailing central tendency (mean, median) and variability (SD).
Most participants had missense variants (n = 37) (Fig. 1).Seventeen had the same missense variant (n = 17, c.2525A > G, p.Asn842Ser).Six other participants shared a further missense variant (c.2149 G > A, p.Gly717Arg) and a further fifteen participants had other missense variants.Of the four participants who did not have missense variants, three had truncating variants (IDs 39, 40, 41) and one had a splice site variant (ID 38).Thirty-nine were confirmed de novo and two were of unknown inheritance   Musculoskeletal problems were apparent (15/31), although findings were heterogeneous.Hypotonia was common in infancy (11/41).Small stature and difficulties gaining weight were frequent (11/40).
Education.Twenty-three participants were school aged or older at the time of assessment.Two children were home-schooled, 13 attended special schools and eight attended mainstream settings.VABS II/III scores from participants with the same variant (c.2525 A > G, p.Asn842Ser, n = 15), were compared to the rest of the cohort (n = 22) (Fig. 2).There was no statistically significant difference between participants with this variant and the rest of the cohort on the adaptive behaviour composite score (p = 0.39, p > 0.05) or their communication score (p = 0.36, p > 0.05).However, when descriptively assessed via a boxplot (Fig. 2), the participants with the same variant tended to be more similar to one another on their adaptive behaviour composite score (SD = 15.97)than the rest of the group (SD = 23.39).

Language and social communication
At a group level, average receptive language skills (mean = 10.20) were significantly stronger than average expressive language ability (mean = 8.71) (p = 0.03, p < 0.05) on the VABS II/III.Overall communication scores (test standard score mean = 100, SD = 15) indicated generally low communication skills (mean = 63.76),however scores ranged from within average limits to severely impaired (Fig. 2, Table 3).
CSBS-DP emotion and eye gaze (mean = 9.75), words (mean = 8) and understanding (mean = 7.75) were in the average range (Table 4).High variability in the group was also observed across CSBS-DP subdomains.
Language skills on the CCC-2 (n = 22) and CSBS-DP (n = 4) were low across most subdomains (mean = 10, SD = 3) (Table 4).CCC-2 subdomain scores ranged between −1 to −2 SD of the mean, except for speech, syntax and semantic scores which were greater than −2 SD (Supplementary Fig. 2).Speech was the lowest subdomain and was significantly different to all subdomains except syntax (p < 0.05, Table 4).High variability amongst the group was observed in all subdomains.
SRS-2 T-scores demonstrated a range of social communication abilities, from within normal limits (9/27) to severely impaired (10/ 27) (normative mean = 60, SD = 10) (Fig. 3, Table 3).Moderate to severe scores indicate a high likelihood of autism (15/27), however only five had confirmed diagnoses of autism and four other individuals had sensory processing disorder.Restricted and repetitive behaviours were moderately impaired (mean = 70.53)and were significantly different from social motivation (p = 0.0003, p < 0.01).All other social communication domains were mildly impaired.
More than half of the cohort used AAC (24/41) (Table 3).KWS was commonly the sole AAC system (11/24) and was used by younger children (<3 years old, 4/11) or those with verbal communication who on occasion used single signs while speaking (6/11), with only one older participant (>3 years old) using KWS as their primary communication system.All other AAC users had graphic AAC systems (Table 3).Four minimally verbal participants older than 3 years old, when children typically learn to combine words, did not have an AAC system.
Eleven minimally verbal participants completed the IPCA (Supplementary Table 3).In terms of symbolic communication, almost half the group used speech to greet and farewell others and seek attention (5/11) and used sign to request 'more' (6/11).Graphic AAC was mostly used for requesting objects (4/11).Participants also used symbolic gesture (e.g., hugging, pointing) to request objects and seek comfort and answer yes or no.Challenging behaviours were exhibited when participants felt angry (6/11).
The fourteen participants with CAS had features across all three diagnostic criteria (Fig. 4) [29,30].The most prevalent CAS features distinct from dysarthric features were inconsistent production of the same phoneme (consonant or vowel) across different words (92.86%), difficulty sequencing sounds and syllables (85.71%), and increased errors with increased word length and complexity (71.43%).Participant 3 had a history of CAS that had largely resolved.One participant (ID 36) had a mild stutter [28].Four of the eight participants with dysarthria also had CAS.Dysarthric features were seen across all speech dimensions (Fig. 4).Only five participants had a speech disorder diagnosis in isolation (CAS 1/5, dysarthria 2/5, articulation  For non-English speaking participants who were not able to be assessed over telehealth (n = 19), 10 caregivers reported clinically diagnosed speech disorders and nine individuals were minimally verbal.Of the 10 with clinically diagnosed speech disorders, articulation disorder was most common (9/10), followed by CAS (6/10), phonological delay (5/10) and disorder (5/10), and dysarthria (4/10).
Intelligibility ranged from 1 (never understood) to 5 (always understood) across a variety of communication partners (n = 41) (Supplementary Fig. 1).Participants were most intelligible to their caregivers (mean = 3.98, sometimes to usually understood), and least intelligible to strangers (mean = 2.65, rarely to sometimes understood).

Feeding and nonspeech oral motor skills
The ChOMPS (n = 17) indicated that feeding difficulties were almost universal (16/17) in children younger than 7 years old (Supplementary Fig. 3).Most participants (15/17) had highly concerning feeding skills (<5th percentile).Only one participant had feeding skills within normal limits.Complex movement patterns (e.g., licking food off the top lip) were descriptively most challenging, while basic movement patterns were a strength (e.g., bringing a bottle to mouth).Drooling prevalence ranged from never drooling to frequent drooling, with drooling generally delayed and resolving only by the late primary school years.
Oral motor skills were impaired in all participants able to complete testing (21/21, Table 3).Greatest difficulty was seen in moving the tongue vertically and horizontally (14/21 and 13/21, respectively), as well as in rounding the lips (10/21), and coordinating two or more non-speech movements (13/21, e.g., bite then lick lips).

DISCUSSION
Here we provide the most comprehensive characterisation of speech and language in CDK13-related disorder.With the addition of 33 novel cases to the existing 60 cases in the literature, we also provide a description of over a third of all published cases of CDK13-related disorder to date.
Speech production was substantially more impaired when compared to other communication domains, such as social communication.Speech disorder was the most prevalent phenotypic feature, where CAS was dominant (63.6%) and considerably more prevalent than general population diagnostic frequencies (0.1%) [37,38].CAS frequently co-occurred with other speech disorders.Despite the frequency of CAS, only one participant was receiving a CAS-specific intervention.This lack of recognition of CAS may be hindering opportunities for more targeted therapy with negative implications for longer-term outcomes.Expressive syntax (the arrangement of words to form sentences), can also be impacted by severe speech disorder [39].Development of speech was generally protracted, e.g., not combining words until after 15 months of age, with most participants using AAC to support their communication needs while verbal speech developed.Some participants remained minimally verbal or had severely impaired speech intelligibility, requiring AAC aides into adolescence.Comprehensive AAC supports are required so that individuals can meet all their communication needs where verbal speech is inadequate.
Historically, the non-specific terms speech/language and delay/ disorder have been used interchangeably to describe the features observed in CDK13-related disorder.Our systematic characterisation of specific speech and language diagnoses is critical for the provision of tailored interventions.Our findings suggest access to AAC in the early years, with ongoing support for AAC into adolescence if needed, is of paramount importance to optimise communication outcomes.Typically developing children are immersed in their language system from birth and say their first words around 12 months of age.A child with CDK13-related disorder should be exposed to both verbal language and AAC before their first birthday to allow for optimal learning opportunities with a trained speech pathologist [40].AAC is not used as a replacement for verbal development, but rather it is known to support verbal development [41] and particularly support growth of expressive vocabulary and grammar [42].AAC should continue to be implemented if the child cannot be understood by different individuals across communication settings (e.g., school, home, with friends), so as to meet all of their communication needs.Consequently, AAC systems that can execute a range of communication functions should be considered.Further, a combination of AAC systems can be used, known as a multimodal approach, such as KWS and a high-tech graphic AAC [43].
The most common speech disorder in this group, CAS, disrupts motor planning and programming for speech.In line with this speech motor involvement, fine and gross motor impairment were also widespread.The frequency of co-occurring fine, gross and speech motor disorders implicates an underlying mechanism of disordered movement planning abilities in CDK13-related disorder.Evidence for the motor involvement in CDK13-related disorder is consistent with neurobiological evidence showing high CDK13 expression in the cerebellum; which is responsible for precision of speech sounds and physical movement amongst other skills [44][45][46].Further, almost half of those with MRI findings in our cohort had hypoplasia of the corpus callosum.Callosal aberrations have also been implicated in speech disorder [47].Further evidence is required to better understand the neurobiological bases of CDK13-related disorder and their association to speech disorder.
Six female participants had average expressive and/or receptive language skills.These participants demonstrate that speech and language disorders may dissociate, given that all of this group had speech disorders in the presence of intact language abilities.Participants with moderate to severe social behavioural impairment all had similarly impaired language skills.Yet there were also participants with average social behaviour and impaired language ability.Hence, impaired social behaviour was always associated with impaired language skills, but not vice versa.It is important to acknowledge that speech and language impairment can be present in the absence of ID, with linguistic behaviours having their own biologically driven neurological pathways [48].
Receptive language was a relative strength when compared to expressive language ability.In genetic conditions with a high prevalence of CAS systematically characterised to date, stronger receptive language skills compared to expressive language skills     have not been seen across cohorts [15][16][17][18].This suggests that receptive language may be a strength for individuals with CDK13related disorder, at least relative to other genetic conditions involving CAS that are understood at this time [15][16][17][18].A limitation of our study was that assessment tools and access to trained clinicians for examining speech and language in individuals from non-English speaking backgrounds were more limited than those available for English-speaking individuals.
Moderate ID generally corresponded with moderate to severely impaired language skills.However, three participants with very low to average FSIQ had moderate to severely impaired language skills.Consequently, intellectual and language ability are typically congruent, but can be distinct from one another in some individuals, indicating distinct neurobiological pathways underpinning language impairment and intellectual disability.
The incidence of ID was less in this cohort than previously reported [14], with around one quarter of assessed participants having borderline to average FSIQ.However, this cohort may be biased, as caregivers may only self-refer to a speech and language study for children with stronger language and intellectual ability.Additionally, previous literature largely characterises individuals drawn from cohorts of children ascertained for ID, so here we broaden the phenotype of CDK13-related disorder with the inclusion of individuals without ID.
The occurrence of other neurodevelopmental conditions such as autism and ADHD was consistent with previously published cases [14].However, most participants with moderate to severe social behaviour impairment in this study did not have a clinical diagnosis of autism.For the first time, sensory processing disorder was identified as a commonly occurring feature of CDK13-related disorder, with over one-quarter of the cohort affected.However, sensory processing disorder and autism are difficult to differentially diagnose, especially on a background of intellectual and language impairment, and sensory processing disorder is not considered a DSM-V diagnosis [24,49].The range of co-occurring neurodevelopmental conditions highlights the importance of systematic neuropsychological assessment, to provide optimal, individualised support.
With regards to genotype-phenotype correlations, there was little evidence to indicate that genetic variants were closely associated with specific phenotypes.Of the 17 participants who shared the same variant, considerable heterogeneity emerged in intellectual, language, speech, and medical presentations.
The health and medical profile in CDK13-related disorder was expanded here with our addition of 33 novel cases to the literature.Feeding problems had a significant impact in infancy and early childhood.Similarly, renal, urogenital, and musculoskeletal malformations, and vision impairment were more common than cardiac malformations in our cohort who, as noted earlier, may have been a more biased group.Cryptorchidism was present in our cohort (38% of males), having been recently described in individuals with pathogenic CDK13 variants [14].
We are the first to characterise sleep disturbances in CDK13related disorder, highlighting prevalent insomnia features.Sleep quality and duration can also negatively impact receptive and expressive language skills [50].
In terms of clinical impacts of our work, neuropsychological assessments are recommended to assess cognitive abilities, given the incidence of ID and other neurodevelopmental conditions.Likewise, occupational and physiotherapy are warranted as fine and gross motor impairment was ubiquitous.Critically, our work suggests speech pathology services should be sought to implement AAC in early childhood, and then provide targeted speech and language therapy (e.g., evidence-based CAS therapy) when verbal speech develops.
In conclusion, we characterise speech and language in CDK13related disorder and identify CAS as a common feature.Until this study, CAS had only been described in one individual (ID 27) in the literature, included in two previous studies [13,14].The profile of speech, language and ID, on the background of significant health disorders, emphasises the importance of comprehensive, multidisciplinary assessment and intervention for individuals with CDK13-related disorder.

Development.
Most participants learned to sit and walk after the expected milestones of 7 (35/41) and 15 months (33/41), respectively (Tables 2, 3 for milestones).Twenty-six participants had delayed acquisition of first words (>15 months) and four had not yet said their first words (aged between 2-to 12-years).Similarly, only eight participants made short sentences at the expected age (2-3 years), and 19 participants (aged up to 15 years old) were not yet combining words.The remaining 15 participants began combining words between 5 to 7 years of age.Neurodevelopmental conditions.Co-occurring neurodevelopmental conditions were common ( de novo, ^= unknown inheritance, + = feature present, -= feature absent, # = surgery, ASD atrial septal defect, CT computerised tomography, MRI magnetic resonance imaging, NA not assessed, PDA patent ductus arteriosus, VSD ventricular septal defect.NM_003718.Most participants had accessed speech therapy (39/41).Many accessed physiotherapy (36/41) and occupational therapy (36/41) for gross and fine motor impairments.Adaptive behaviour and motor skills A range of profiles was noted in the range of adaptive behaviour composite scores (mean = 61.86).No single domain of daily living, socialisation or communication, was significantly different to any other (p = 0.26).

Table 2 .
Development, co-occurring diagnoses, and therapy supports in this cohort.

Table 3 .
Speech and language phenotype in children with CDK-13 related disorder (n