Heterozygous loss-of function variants in DOCK4 cause neurodevelopmental delay and microcephaly

Neurons form the basic anatomical and functional structure of the nervous system, and defects in neuronal differentiation or formation of neurites are associated with various psychiatric and neurodevelopmental disorders. Dynamic changes in the cytoskeleton are essential for this process, which is, inter alia, controlled by the dedicator of cytokinesis 4 (DOCK4) through the activation of RAC1. Here, we clinically describe 8 individuals (7 males and one female) with variants in DOCK4 and overlapping phenotype of mild to severe global developmental delay. Additional symptoms include coordination or gait abnormalities, microcephaly, unspeci�c brain malformations, hypotonia and seizures. Five individuals carry missense variants (four of them detected de novo) and three individuals carry null variants (two of them maternally inherited). Molecular modelling of the heterozygous missense variants suggests that the majority of them affect the globular structure of DOCK4. In vitro functional expression studies in transfected Neuro-2A cells showed that all missense variants impaired neurite outgrowth. Furthermore, Dock4-knock-out Neuro-2A cells also exhibited defects in promoting neurite outgrowth. Furthermore, a sex-specic expressivity is possible, as the only female of the cohort was comparatively mildly affected. Our results, including clinical, molecular and functional data, suggest that loss-of-function variants in DOCK4 cause a variable spectrum of a novel neurodevelopmental disorder with microcephaly.


Introduction
Global developmental delay (DD), often leading to intellectual disability (ID), has a prevalence of about 2% and is among the most frequent indications for genetic testing (Sheridan et al. 2013).Despite continuous improvements in sequencing technologies, a genetic cause for DD/ID can only be found in 27 to 42% of routine diagnostic cases (Gilissen et al. 2014; Aspromonte et al. 2019).The high genetic heterogeneity is one reason for the incomplete diagnostic rate.According to the SysNDD (Kochinke et al. 2016) database, pathogenic variants in 1,780 (gene statistics from October 26, 2023) different genes are currently established as causes of diverse neurodevelopmental disorders (NDD).However, as pathogenic variants in novel NDD genes are still being identi ed at a high pace (Kaplanis et al. 2020), it is becoming obvious that not all NDD genes have been discovered yet.
Neurons form the basic anatomical and functional structures of the nervous system and consist of two structural components: a cell soma, and neurites that include a single axon and dendrites.The length of the neurons varies depending on their function and can be 10,000 to 15,000 times the diameter of the neuronal cell body (Larsen et al. 2016).The formation and function of dendrites and neurites is highly dependent on dynamic changes in the cytoskeleton that are regulated by small GTPases (Nimchinsky et al. 2002).Disease causing variants in genes that are involved in this process, such as RHOA (MIM: #618727), CDC42 (MIM: #616737) and RAC1 (MIM: #617751), are associated with a NDD.In addition, dysregulation of guanine nucleotide exchange factors that regulate RAC1 activity are also associated with DD/ID.For example, bi-allelic null variants in DOCK6 (MIM: #614219) and DOCK7 (MIM: #615859), members of the dedicator of cytokinesis (DOCK) family, are a cause for the Adams-Oliver syndrome (Sukalo et al. 2015) and epileptic encephalopathy (Perrault et al. 2014), respectively.Furthermore, Wiltrout et al. reported three individuals with DD, hypotonia, coordination or gait abnormalities and bi-allelic variants in DOCK3 (MIM: #618292) (Wiltrout et al. 2019).Together with DOCK3, DOCK4 belongs to the DOCKB subfamily and has four conserved domains: the Nterminal located src Homology-3 (SH3) domain, a proline-rich C-terminus and the dock homology regions (DHR) 1 and 2; the latter is responsible for binding and activation of RAC1 (Shi 2013).The DOCK4-dependent RAC1 activation has been reported as important for neurite differentiation (Xiao et al. 2013) and dendritic spine formation (Ueda et al. 2013).
In this study, we report eight individuals with an overlapping phenotype of mild to severe DD/ID, harboring different variants in DOCK4.We used structural in silico modeling and a neurite outgrowth assays with Neuro-2A cells that were transiently transfected with an overexpression plasmid for DOCK4 or after CRISPR-Cas9-mediated Dock4-knock-out to assess possible impact of the variants.

Materials and Methods
Research cohort and identi cation of variants By using match making platforms (Sobreira et al. 2015), personal communication and a literature review, nine different families and 17 individuals from the literature (for further details see Supplemental Information) with potential causative DOCK4 variants were assessed.Genotypic and detailed phenotypic information was obtained from referring collaborators for eight individuals, which were included into the clinical description, using a standardized questionnaire assessing family history, clinical history, genetic testing, variant details, EEG, brain imaging and medication (see Table S1).All individuals underwent single or trio exome or genome sequencing using local protocols.Maternity and paternity were con rmed for all individuals and Sanger sequencing was used for segregation analysis where appropriate.As no causative variants in a known disease gene were identi ed that explains DD/ID of the individuals, the data were examined in a scienti c approach, including parental sequence data if available.
All variants were prioritized considering allele frequency in gnomAD (Karczewski et al. 2020) (Wiel et al. 2019)) and involvement of candidate genes in neuronal processes.Apart from the DOCK4 variants observed, none of the individuals described had other remarkable ndings that likely explain the phenotype.All variants in DOCK4 are described with regard to the human reference genome version GRCh37 and to the transcript NM_014705.4 and have been classi ed according to the criteria of the American College of Medical Genetics (ACMG) (Richards et al. 2015) and the recommendations of the Sequence Variant Interpretation Working Group (ClinGen).

DOCK4 expression plasmids
Expression Plasmids were generated as described previously (Rahimi et al. 2022), by using full-length human DOCK4 open reading frame (ORF; GeneBank: BC117689; MHS6278-211690438) that was obtained from Horizon Discovery BioSciences (Cambridge, United Kingdom).Fragments of the DOCK4 were ampli ed by PCR (TableS2) from DOCK4 ORF by using Q5 High Fidelity DNA Polymerase (New England Biolabs, Frankfurt am Main, Germany) and were assembled together with the pcDNA3.1_mRFPVector by using the NEBuilder HiFi DNA Assembly Cloning kit (New England Biolabs) according to the manufactures recommendation to obtain a plasmid that express a DOCK4 (GeneBank sequence: NM_014705) mRFP fusion protein.Mutagenesis was performed by using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs) according to the manufacturer's recommendation.All sequences were veri ed by Sanger sequencing.
Control cells were generated by using the Alt-R CRISPR-Cas9 Mouse Control Kit (Integrated DNA Technologies).24 hours post transfection, cells were collected and seeded at a density of 0.8 cells per 100 µl into 96-well plates.After 10-12 days of culture, wells containing only one colony were split into two wells of a 96-well plate and positive clones con rmed by DNA sequencing were further propagated.Knockout clones were nally con rmed by Western blotting.

Neurite outgrowth assay
After transfection, cells were collected and subjected to FACS sorting by using a BD FACSAria II SORP (BD Biosciences, Heidelberg, Germany).15000 of mRFP-selected cells were seeded per well into a 96-well plate and culture media was exchanged after 4 hours to Neurobasal A medium containing GlutaMAX, antibiotics and 2% B27 supplement (Thermo Fisher Scienti c).For DOCK4 KO cells, which were not transfected, 7500 cells per well in a 96-well plate were seeded.Afterwards cells were cultured (37°C and 5% CO2 in humidi ed air) and monitored in a Celldiscoverer 7 (Zeiss, Oberkochen, Germany).Neurite length was determined after 36 hours in differentiation media by using NeuronJ (Meijering et al. 2004) implemented in ImageJ.

Molecular Modeling
The

Statistical analysis
Statistical analysis was carried out using SPSS (IBM, Armonk, USA; Version: 24.0.0.2 64-bit).We used a one-way ANOVA with the Games-Howel post hoc test for multiple comparisons.If not stated otherwise, data is presented as mean ± SEM.A p-value < 0.05 was presumed to be statistically signi cant.

Clinical description
In this study we report on eight individuals with heterozygous rare variant in DOCK4.A summary of the clinical symptoms of each individual is presented in Table 1 and a full clinical description can be found in Table S1 and in the Supplemental Information.Additional ndings include microcephaly in individuals 2, 4 and 6.Although minor facial dysmorphisms have been reported for four individuals (individuals 2, 5, 6 and 8), there was no apparent shared facial gestalt (Table S1).
It should be noted that we also identi ed two compound heterozygous missense variants in a fetus with microcephaly, rhombencephalosynapsis and microlissencephaly (Table S1).As the pregnancy was aborted, no further clinical information is available.As the assumed inheritance is not consistent with the rest of the cohort, we have not included this case in the clinical description.Nonetheless, this fetus had the most severe structural anomalies of the cerebellum and brainstem (for details see Supplemental Information).

Genetic ndings
In total, we identi ed eleven variants in DOCK4, eight missense and three nonsense variants (Fig. 1A).In four individuals, we detected a de novo missense variant.Segregation analysis of individual 5 was not possible because he was adopted.Two null variants were maternally inherited and the inheritance of the null variant of individual 7 is unknown (Fig. 1B).All variants were absent from gnomAD, except the de novo variant from individual 4 (c.4013G> A, p.Arg1338Gln), which was detected in gnomAD once, and the two inherited missense variants p.Pro253Leu and p.Val420Met (detected compound heterozygous with the variant p.Val1042Ala) that were observed with an allele frequency of 0.034% and 0.004% in gnomAD, respectively.It should be noted that the paternally inherited variant p.Pro253Leu was detected together with the de novo variant p.Met1044Thr in individual 2.
Unfortunately, phasing of the de novo variant was not possible, but it is more likely that this variant is also located on the paternal allele (i.e. in cis) (Jónsson et al. 2017).For both of the identi ed null variants, it can be assumed, that the mRNA is degraded by nonsense-mediated mRNA decay (Lindeboom et S1).The missense variants are located in the proline rich C-terminal end, in the DHR-1, DHR-2 and proximal of the DHR-2 domain (Fig. 1).A clustering of disease causing missense variants in a speci c region of DOCK4 was not observed.Only several small parts of DOCK4 seem to be highly depleted of missense variants in the general population (Fig. 1A).

DOCK4 missense variants affect the globular structure of DOCK4
To better understand the impact of the missense variants in DOCK4, we performed a structural analysis.As there is no experimental structure of DOCK4 available to date, we used a model generated by AlphaFold-2 for the structural interpretation of the variants.Inspection of the AlphaFold scores indicates a reliable model for residues 1-1587, whereas the low con dence scores for residues 1588-1966 suggests that the C-terminus is disordered and lacks a de ned three-dimensional structure.Structural modelling of the variants in the globular domain reveals that all of them cause a destabilization of the DOCK4 fold.
Depending on the molecular origin of destabilization, the variants can roughly be divided into three categories: introduction of steric clashes, loss of hydrophobic interactions and loss of polar interactions.For each category, the corresponding variants are described below, and a representative variant is shown in Fig. 2. (i) Introduction of steric clashes: The p.Thr982Ile exchange induces steric clashes with Leu923 due to the longer isoleucine sidechain present in the variant (Fig. 2B,C).Similar steric problems emerge from the longer methionine sidechain in the p.Val420Met and the β-branched threonine sidechain in the p.Met1044Thr variant.(ii) Loss of hydrophobic interactions: In the p.Ile1067Thr variant, the isoleucine is replaced by a shorter threonine.This results in a loss of hydrophobic interactions with Val1106 and Lys1109 (Fig. 2D,E).A similar loss of hydrophobic interactions is observed for the p.Pro253Leu and p.Val1042Ala variants.(iii) Loss of polar interactions: In the wild-type protein, the sidechains of Arg1338 and Asn1411 form a hydrogen bond (Fig. 2F).This polar interaction cannot be formed by the shorter glutamine sidechain caused by the p.Arg1338Gln variant (Fig. 2G).Similarly, the Arg853-Glu817 salt-bridge is lost in the p.Arg853His variant.
The variant p.Lys1962Asn is located in the nonglobular C-terminus of DOCK4.Such regions may harbor linear motifs that mediate posttranslational modi cation and/or protein-protein interactions (Kumar et  As the in silico predictions suggests that all variants affect DOCK4 function, we tested this assumption experimentally.Huang et al. (Huang et al. 2019) demonstrated that the variants p.Arg853Leu and p.Asp946_Lys1966delinsValSer* (described as 945VS) lead to a signi cantly reduced DOCK4 dependent activation of Rac1 and Rap1 that resulted in a compromised function in promoting neurite outgrowth in Neuro-2A cells.Therefore, we generated plasmids expressing DOCK4 and harboring the variants of interest, transfected them into Neuro-2A cells and monitored neurite outgrowth in the presence of differentiation media.In addition, we tested the gnomAD variant p.Pro1733Ala, and as an additional positive control the variant p.Pro1718Leu that exhibits an impaired function in activating RAP1 (Yajnik et al. 2003).As expected, overexpression of wild-type DOCK4 promoted neurite formation compared to cells transfected with the vector alone (Fig. 3).Moreover, the gnomAD variant p.Pro1733Ala was indistinguishable from the wild-type, supporting its non-pathogenic nature, and the positive controls p.Arg853Leu and p.Pro1718Leu resulted in a signi cant reduced neurite outgrowth.More important, all of the variants studied from the individuals also signi cantly impaired neurite formation in Neuro-2A cells, including the two variants identi ed in the fetal case.This was demonstrated by both, the total and longest neurite length (Fig. 3B, C).Next we experimentally tested whether loss of DOCK4 also affects neurite outgrowth in Neuro-2A cells.Therefore, we generated Neuro-2A Dock4 knock-out cells by using the Alt-R CRISPR-Cas9 system utilizing two different guide RNAs (ko1 and ko2) and one unspeci c control guide RNA (C: control).Dock4 knock-out was con rmed by DNA sequencing and western blot (Figure S1).The neurite outgrowth assay demonstrated that ko1 and ko2 cells had signi cant shorter neurites (total and longest neurite length) compared to control and wild-type cells (Fig. 4).Taken together, our experiments demonstrated that null variants and deleterious missense variants lead to impaired function in promoting neurite growth, supporting the association of DD / ID and the variants of the individuals of the present cohort.

Discussion
In the present study, we describe eight individuals with de novo and inherited heterozygous variants (in total eleven different variants, Fig. 1) in DOCK4.The affected individuals exhibited a NDD with mild to severe DD, microcephaly and coordination or gait abnormalities.According to ACMG criteria (Richards et al. 2015), ve variants can be classi ed as (likely) pathogenic and six as variants of uncertain signi cance (Table S1).We could demonstrate that null variants (Fig. 4) and the seven missense variants p.Pro253Leu, p.Val420Met, p.Thr982Ile, p.Val1042Ala, p.Met1044Thr, p.Ile1067Thr and p.Lys1962Asn impair neurite formation in Neuro-2A cells.Structural modelling of the variants suggested that all of them cause a destabilization of the DOCK4 fold.Of note, we also tested the variant c.593G > C, p.Ser198Thr (absent from gnomAD) that we identi ed in a female individual with isolated generalized epilepsy (maternally inherited).As the variant was indistinguishable from the wild type (total neurite length: D4-wild-type: 110 ± 4.2µm, p.Ser198Thr: 99.8 ± 3.2 µm; p = 0.77; longest neurite length: D4-wild-type: 64.6 ± 2.5µm, p.Ser198Thr: 56.2 ± 1.5 µm; p = 0.183), we excluded this individual from the cohort.Unfortunately, we could not test the effect of the variant p.Arg1338Gln (individual 4), who was the only patient alive with a severe phenotype (Table 1).However, molecular modelling indicates that the p.Arg1338Gln exchange destabilizes the DOCK4 fold (Fig. 2F, G).Notably, p.Arg1338Gln is the only variant of the analyzed set, which is located close to the DOCK4-Rac1 interface (Fig. 2A), may therefore more critically affect RAC1 binding, and possibly results in a more severe phenotype.Noteworthy, the exchange of the amino acids 1359-1361 and 1373-1374 that are close to the RAC1 binding site, to an alanine in DOCK3 (corresponding to codons 1321-1323 and 1335-1336 of DOCK4: NM_014705.4)disrupted Rac activation (Namekata et al. 2010).
The genes of the DOCKB subfamily DOCK3 and DOCK4 are important for brain development by activating RAC1 (Shi 2013).Interestingly, we could also nd a notable overlap of symptoms of the individuals in in the present cohort with those harboring biallelic disease-causing variants in DOCK3 (Wiltrout et al. 2019) (Figure S2), supporting DOCK4 as a NDD gene.In particular, coordination or gait abnormalities, e.g.ataxia were present in all DOCK3 individuals and in the majority of individuals (individuals 1, 2, 3, 5 and 8) in the present study.
Based on the detection of rare de novo and inherited missense variants in DOCK4, we hypothesized an autosomal dominant mode of inheritance, although we could not rule out an autosomal recessive mode based on the impairment of DOCK4 function by the variants, i.e. de novo variants would lead to a more severe loss of function compared to inherited variants.With our assay, we could not distinguish between the effects of the de novo and inherited missense variants, strengthening the hypothesis of an autosomal dominant mode of inheritance.This is also supported by the detection of a heterozygous de novo variant in DOCK4 in seven individuals with NDD in other studies (see Table S1).Interestingly, Dock4 knockout in mice leads to early embryonic lethality (Abraham et al. 2015), which is frequently observed in NDD genes, such ATP2B1 (Okunade et al. 2004), with an autosomal dominant mode of inheritance.In contrast, knock out of the paralogue Dock3 is not embryonic lethal in mice, but adult animals show a cerebral accumulation of autophagic vacuoles and a disorganization of the axonal cytoskeleton (Chen et al. 2009).Another possible mode of inheritance could result from a hypomorphic variant on the second allele, in addition to a null variant (individuals 6-8), as described for the TAR Syndrome (Albers et al. 2012).In fact, we could detect additional coding SNPs p.Pro1733Ala (rs150569245) and p.Val1914Ile (rs12705795) in individual 6 and 8, respectively.However, although we have not tested the effect of the SNP p.Val1914Ile, it is unlikely that both SNPs are hypomorphic, as we could demonstrate that the variant p.Pro1733Ala does not affect DOCK4 function (Fig. 3).Therefore, this mode of inheritance is unlikely for DOCK4.
Nevertheless, based on preliminary information from the fetus (F1 in Table S1), an autosomal recessive DOCK4-associated disorder may also be possible and will need to be investigated in the future.
In family 7, we detected a null variant in DOCK4 that did not co-segregate with the symptomatic mother and sister (Fig. 1B, Table S1; a genetic cause for the symptoms of the mother and the sister has not yet been fully investigated).Furthermore, the phenotype of individual 7 is milder (mild DD, no coordination or gait abnormalities, learning di culties and no ID) and thus different from the rest of the cohort.Hence, it is unclear, whether this variant is causative for the symptoms of individual 7.
However, it is not known whether the variant is inherited or whether the father is symptomatic, as there is no contact with the father.More important, all other individuals of the cohort are male, which could indicate a sex-speci c expressivity.Pagnamenta et al. (Pagnamenta et al. 2010) described a family with eight individuals (two females and six males) harboring a DOCK4 truncating deletion (p.Asp946_Lys1966delinsValSer*).The two clinically characterized females had an unremarkable development and were diagnosed with dyslexia.Two of the affected males had DD and were diagnosed with autism.The IQ was in the normal range.Three other males had signi cant problems in reading and spelling and one male was diagnosed with Asperger disorder (no developmental milestones were available).This report indicates an intrafamilial variability of DOCK4 variants, with males more severely affected than females.Guo et al. (Guo et al. 2021) investigated autism spectrum-disorder like behavior in conditional Dock4 knockout mice and also observed sex-speci c effects.For example, knockout males showed higher anxiety levels and poorer working memory compared to knockout female mice.Noteworthy, overexpression of Rac1 restored excitatory synaptic transmission and corrected the impaired social behavior of Dock4 knockout mice.Taken together, these ndings provide preliminary evidence for sex-speci c variable expressivity within autosomal dominant DOCK4-related NDD.However, this assumption can only be con rmed in a larger cohort.
Regarding the underlying pathomechanism, our data, including in silico structural modeling, suggest a loss-of-function mechanism for both missense and null variants.In addition, both Dock4 knockout and the missense variants investigated resulted in impaired function in promoting neurite outgrowth in Neuro-2A cells.A gain-of-function mechanism of missense variants is unlikely because overexpression of wild-type DOCK4 results in increased neurite outgrowth capabilities and not the opposite.Furthermore, loss of DOCK4 function is compensated by overexpression of the DOCK4 interacting partner RAC1, as demonstrated in vitro (Huang et al. 2019) and in vivo (Guo et al. 2021), further suggesting a loss-of-function mechanism.
In summary, the overlapping phenotype of eight Individuals, the structural modelling, the role of DOCK4 in the central nervous system, and the proven impact of the variants on neuronal outgrowth prompt us to add heterozygous null variants and deleterious missense variants in DOCK4 as a monogenetic cause of an NDD with microcephaly.

Internet Resources
al. 2016) causing haploinsuffuciency.Based on gnomAD (Karczewski et al. 2020), DOCK4 has a reduced amount of missense and null variants compared to its size of 1,966 amino acids (pLI = 0.83, Z score = 2.52) in a population without severe, early-onset phenotypes, indicating a mild constraint.The missense variants from individuals 2, 3 and 4 are predicted to be deleterious by CADD and MutPred2 (Table

Table 1
Of note, the father of individual 2, who transmitted one variant in DOCK4 (Fig.1B), has a bipolar disorder, attention de cit hyperactivity disorder, anxiety and dyslexia.Different coordination or gait abnormalities were reported in ve out of seven individuals (note: individual 4 is not assessable due to severe motor delay).In detail, individual 1 had ataxia and tremor, individual 2 had ataxia and spasticity, individual 3 had dystonia and spastic hemiplegia, individual 5 had dystonia and individual 8 had cerebral palsy.Other neurological symptoms were muscular hypotonia (individuals 2, 5 and 8) and different types of seizures.Individual 1 had generalized tonic-clonic seizures that regress under LPA treatment.Individual 8 had an abnormal EEG that is consistent with a focal onset seizure.Seven individuals underwent MRI and four had non-speci c brain abnormalities.Individual 2 had mild cerebellar ectopia and individual 3 had a complex volume loss in the left periventricular region, extending into basal ganglia and part of brainstem, with thinning of the corpus callosum and suggestion of incomplete myelination in surrounding areas.Individual 5 had mild diffuse cerebral atrophy and minimal white matter gliosis and individual 8 bilateral cerebral hemispheric white matter volume loss and gliosis.
developmental milestones until the rst epileptic seizure at 4 years of age, followed by moderate to severe developmental delay.Cognitive impairment was highly variable, ranging from learning di culties with no ID (individuals 3 and 7) to mild (individuals 2 and 6) and moderate ID (individuals 1 and 5).Individual 4 was at the most severe end of the spectrum with severe ID and incomplete head control, incomplete rolling over and no ability to sit at the age of 3 years and 11 months.Behavioral abnormalities were less common, with a diagnosis of autism or autistic behavior in individuals 5 and 6, and attention disorder and anxiety in individual 7.