1. IntroductionIntellectual disability (ID) is a major medical and socio-economic problem owing to its high incidence in the general population. Mutations in about 1500 different genes have been associated with ID [1,2], while pathogenic mechanisms and the molecular basis of gene dysfunction in ID remains to be elucidated. So far, functional studies have mainly focused on single gene defects, such as the Fragile X syndrome. Here, we propose a systematic approach to gain pathway-based insights into mechanisms leading to cognitive dysfunction in humans.Based on human genetic studies, we selected 45 genes (Table 1) to generate and characterize mutant mice models, of which 39 were selected based on the frequency of pathogenic variants in ID patients. The proteins encoded by these genes act in various biological processes, such as transcription regulation, epigenetic modification, synaptic transmission, or influencing the excitatory/inhibitory balance of central nervous system (CNS) activity. In patients with ID, 31 genes of this selection carry loss-of-function (LoF) variants or deletions, with two genes displaying additional gain-of-function (Gof) mutations, two genes carrying splicing variants, and 12 genes carrying missense variants. Most of these pathogenic variants cause a syndromic ID disorder. Six other genes potentially involved in ID were included, namely ASCC3 found mutated in a single family by performing a large-scale genomic study [3], EHMT2 homologous to EHMT1 causing Kleefstra syndrome, and CDK8. CDK8 encodes a key subunit of the Mediator complex involved in transcriptional processes and harbouring pathogenic variants in several other subunits, causing syndromic or non-syndromic ID. Two additionally included genes were found mutated in single patients with ID (CDC27, mIR124a-2). We also added STARD8, which is deleted or disrupted in patients with a contiguous gene deletion and craniofrontonasal syndrome [4,5], and encodes a Rho GTPase-activating protein, like OPHN1 whose LoF causes X-linked ID [6]. Thirty of the selected genes are located on an autosome and 15 on the X chromosome. Thus, we generated, characterized and compared male mouse models carrying either a mutation identical to a patient with ID or a LoF allele to better understand the function of a candidate ID gene. Different types of strategies were followed for generating the appropriate models (see Section 2 and Figure S1). Subsequently, we assessed the viability of the mutant mouse lines. Twenty-seven lines were further investigated using a standardized behavioural screen, focusing on males since a wide range of disorders associated with ID are X-linked ID and 33% of genes in the present study are located on the X-chromosome. Several tests were used to assess a wide range of functions or their pathologies, including circadian activity, neurological reflexes and specific motor abilities, anxiety-related behaviour, sensorimotor gating, and learning and memory processes. For some mouse lines, additional tests were performed to further characterize abnormalities observed or to extend phenotypic traits related to individual genes. 4. DiscussionIn the present study, we generated 50 mutant lines for 45 genes clinically or potentially relevant for further understanding ID in humans. About 66% of the mutant lines generated were homozygous/hemizygous lethal, providing evidence that these genes are essential for normal development and survival. Embryonic phenotyping of homozygous/hemizygous/heterozygous lethal lines revealed several abnormalities, including pronounced craniofacial and skeletal defects, severe ganglia hypoplasia, and abnormal nervous or sensory system development in line with congenital malformations observed in the corresponding human syndromes. For example, we found that Setbp1−/− mutants displayed palatal and vertebral skeletal defects, reduced DRG and abnormal nasopharyngeal opening, reproducing some aspects of the SETBP1 Disorders (also known as Mental Retardation, Autosomal Dominant 29), characterized by ID and distinctive facial features [61,87]. We also found in adult Setbp1+/− mice several behavioural alterations including muscle weakness and altered PPI reminiscent of symptoms observed in patients with a similar mutation type [88]. This model is of interest for better understanding the physiopathology of this new syndrome. Our results from Med25−/− embryos revealed several abnormalities including exencephaly, anophthalmia or cyclopia and telencephalon hypoplasia, in line with those observed in humans with MED25 mutations such as Basel-Vanagaite-Smirin-Yosef syndrome characterized by severely delayed psychomotor development resulting in ID, as well as variable eye, brain, cardiac, and palatal abnormalities [47]. Our results obtained for Tubb3M388V/+ embryos are in line with those previously reported for the Tubb3R262C/R262Cmouse model and with a spectrum of abnormalities including hypoplasia of oculomotor nerves and dysgenesis of the corpus callosum and anterior commissure observed in human syndromes with TUBB3 mutations [66,67,89], supporting the role of TUBB3 in axonal guidance and maintenance.Large scale standardized behavioural phenotyping of 27 mutant lines carrying mutations in genes involved in ID in humans revealed unique gene/phenotype behavioural profiles based on activity patterns. Interestingly, genes mutated in the hypoactive group are altered either in Kleefstra syndrome (Ehmt1, Mbd5, Nr1i3) or Rett syndrome (Mecp2) or CDKL5 Deficiency Disorder, CDD (Cdkl5). Kleefstra syndrome is characterized by ID, childhood hypotonia, severe expressive speech delay and a distinctive facial appearance with a spectrum of additional clinical features including autistic-like behavioural problems and cardiac defects [25,26,50,90]. Autosomal Dominant Mental Retardation 1/2q23.1 deletion syndrome, caused by pathogenic MBD5 variants, shares several phenotypic traits with Kleefstra syndrome [36,37]. Similarly, patients with a pathogenic variant in CDKL5 (CDKL5 Deficiency Disorder, CDD) or MECP2 (Rett syndrome) present with overlapping clinical features.We report here that Ehmt1+/−, Ehmt1+/−/Ehmt2+/−and Mbd5+/− mice show hypoactivity and learning and memory deficits in several situations, extending previously reported data in Ehmt1+/−mice [72,91], and recapitulating several cognitive, autistic and hypoactivity features observed in Kleefstra syndrome and 2q23.1 deletion patients, supporting the use of Ehmt1+/− and Mbd5+/− as good mammalian models for Kleefstra and Autosomal Dominant Mental Retardation 1/2q23.1 deletion syndromes.We found only limited and weak behavioural changes in Ehmt2+/− and Nr1i3−/− mutants. EHMT2 and its paralog EHMT1 encodes a histone methyltransferase and act together in protein complexes responsible for deposition of mono- and di-methylated forms of Histone 3 Lysine 9 (H3K9me/me2). These methylation marks are associated with gene silencing in euchromatin [92]. Since LoF variants in EHMT1 give rise to Kleefstra syndrome, it is tempting to speculate that EHMT2 is a candidate for syndromic ID as well. However, our data from Ehmt1+/−, Ehm2+/− and Ehmt1+/−/Ehmt2+/− mutants show that the main phenotypic traits are linked to the Ehmt1+/− mutation, in line with a previous study where we showed that unlike Ehmt1+/−, Ehmt2+/− did not present the marked increase of H3K9me2/3 [76] reduces the strength of the hypothesis linking EHMT2 with Kleefstra spectrum disorders and is potentially associated ID. Indeed, several EHMT2 LoF alleles have been reported without convincing evidence for involvement in human genetic disorders. For NRI3, a single de novo missense mutation (c.740T>C [p.Phe247Ser]) was identified in a patient with core symptoms of Kleefstra syndrome [50]. In our study, Nr1i3−/− mice displayed only a slight decrease in contextual fear conditioning. In line with behavioural data, the assessment of hippocampal neuronal morphology in Nr1i3−/− mice did not reveal any gross abnormality concerning neurite length, branching or excitatory synapse density (not shown). Elsewhere, the quantification of ectopic wing vein formation in Drosophila [50] revealed that the EHMT overexpression phenotype was almost completely rescued by heterozygous LoF mutations in EcR/Nr1i3, and overexpression of EcR/Nr1i3 enhanced EHMT-induced ectopic vein formation, providing strong evidence of a synergistic relationship between EHMT and EcR/NR1I3, and that NR1I3 per se has reduced incidence. Combined, the results could suggest that in the patient affected, the (c.740T>C [p.Phe247Ser]) single amino acid substitution is not a loss of function mutation and has a different effect on the protein. Additionally, the patient carried a de novo MTMR9 missense variant (c.310T>G [p.(Ser104Ala)]; NM_015458.3) of uncertain significance [50].Several mutant lines showed a characteristic hyperactivity phenotype. Il1rapl1−/y, Ptchd1−/y, and ArxDup24/y mice displayed substantial hyperactivity and stereotypic behaviour, and either increased exploration or reduced anxiety. These mutant lines also had altered learning and memory abilities. In humans, IL1RAPL1 and PTCHD1 mutations are found in X-linked ID or X-linked autism spectrum disorders [32,58,59]. Among behavioural features of these syndromes, hyperactivity, stereotypies, and altered learning abilities are commonly present. Interestingly, motor problems including psychomotor delay and hypotonia present in patients with PTCHD1 or ARX mutations were also found in our mutants, which displayed either decreased muscle strength (Ptchd1−/y) or altered grasping and reaching, reflecting fine-tuned motor abilities (ArxDup24/y) [9,10,93]. In human syndromes, mutations are hemizygous substitutions or deletions (IL1RAPL1), hemizygous deletion or insertion (PTCHD1), or hemizygous c.428_451dup24 duplication (ARX). Our mutant lines reproduced some of these mutation types.Mutations in ANKRD11, ATP6AP2 and PRPS1 have also been associated with several neurodevelopmental disorders with ID in humans [7,11,94,95]. ATP6AP2 mutations are found in X-linked ID with Parkinsonism and spasticity, and PRPS1 mutations in Arts syndrome, X-linked recessive Charcot-Marie-Tooth disease 5 and X-linked non-syndromic hearing loss [11,55,56,57,95,96,97,98]. Mutations are either hemizygous splice site mutations that leads to a LoF (ATP6AP2), or substitutions leading mainly to GoF, but also to LoF (PRPS1), causing different behavioural symptoms including hyperactivity, stereotypies and altered cognitive abilities. On the other hand, heterozygous deletions or splice site mutations in the ANKRD11 gene have been found in patients with KBG syndrome, characterized by macrodontia, distinctive craniofacial and skeletal anomalies, short stature, and neurological problems including ID [7,94]. Hyperactivity, anxiety, and hearing loss have also been described [99]. In the present study, we generated Atp6ap2−/y, Prps1−/y and Ankrd11−/− mutant lines and found them embryonic lethal. We then generated and characterized the neuronal specific lines Ankrd11Camk2a, Atp6ap2Camk2a/y, and Prps1Camk2a/y. Ankrd11Camk2a and Atp6ap2Camk2a/y displayed substantial hyperactivity and stereotypic behaviour, increased exploration and reduced anxiety, and altered learning and memory. Interestingly, Atp6ap2Camk2a/y also showed decreased startle response in line with motor problems including hypotonia found in patients. Our data reproduced some of the behavioural phenotypes observed in patients and support, at least in part, the specific effect of the deletion on excitatory neuronal cells. Ankrd11+/− mutants also showed hearing loss displaying increased ABR thresholds [100], in line with hearing problems reported in some KBG patients [99]. Among the hyperactive lines, Prps1Camk2a/y showed only subtle phenotypes displaying increased stereotypic behaviour and increased working memory. We assume that neuronal loss-of-function per se is unlikely to model syndrome-related behavioural traits. In line with this observation, we found that Prps1Csp4/y mice, a glial specific conditional KO, displayed increased stereotypic behaviour during the initial exploration in the actimetric cages, and decreased motor performance in the rotarod (Table S1).In the present study we also analysed the effect of LoF variants in several candidate ID genes. One of these genes, CDK8, encodes an important regulator of the multi-subunit Mediator complex, involved in transcriptional processes. Mutations in other subunits of the Mediator complex were previously identified in syndromic and non-syndromic ID. Cdk8−/− mice are lethal; therefore, we generated and analysed Cdk8Camk2a mice. These mutants displayed hyperactivity in several situations, a trend to increased stereotypic behaviour, hypotonia reflected by decreased muscle strength, and altered recognition memory. While our studies were ongoing, various heterozygous CDK8 missense mutations were reported to cause a syndromic developmental disorder characterized by hypotonia, ID, and behavioural abnormalities [13]. Affected individuals tended to have learning disability, autistic features and attention deficit-hyperactivity disorder (ADHD). In vitro functional studies showed that the mutations strongly attenuated CDK8 kinase activity, supporting a dominant-negative mechanism of pathogenesis for CDK8 substitutions [13]. Interestingly, our data clearly recapitulate most behavioural alterations described in humans [13], and suggest that these alterations are neuronal specific.ASCC3 was identified as a candidate gene for autosomal recessive ID, with a potentially pathogenic missense variant (p.(S1564P)) identified in a single family [3]. In the present study, we show that Ascc3−/− mice are lethal, and Ascc3Camk2amutants showed hyperactive behaviour and increased rearing in the actimetric cages, and reduced anxiety-related behaviour in the elevated plus maze. We suggest that neuronal homozygote deletion of ASCC3 gene is not sufficient to induce profound behavioural alterations. The family affected by the ASCC3 variant was also reported to have mild ID [3].

Across this study, applying an extensive behavioural pipeline allowed us to identify different classes of genes for which mutations caused several behavioural alterations. The heterogeneity of phenotypes and penetrance are reminiscent of the effects of mutations observed in human syndromes with ID. The class of genes with increased activity includes several Camk2a-conditional KO lines specific to excitatory neurons. It can be suggested that the hyperactive phenotype might be due, at least in part, to the Camk2a promoter driving Cre recombinase. Hyperactivity and stereotypies are almost common phenotypes in human syndromes with ID. In addition, our constitutive Il1rapl1−/y and Ptchd1−/y, and ArxDup24/y mice also showed substantial hyperactivity in different situations. Finally, behavioural phenotyping of the Camk2a-Cre reporter line did not show any obvious sign of hyperactivity or another relevant behavioural alteration (not shown). These arguments reduce the strength of the hypothesis that hyperactivity observed in the Camk2a-mutant lines is likely related to the Camk2a promoter driving Cre recombinase. Nevertheless, it is noteworthy that Camk2-Cre specific mutations for constitutively lethal lines are partially expressed (in the glutamatergic neurons). Additional data from mutants bred on other reporter lines would increase our knowledge about the effect of mutations described in this study depending on their expression in other cellular compartments. Conditional mutants with combined expression in different cell types might potentially display extensive or stronger phenotypic traits affecting more functions.

Our functional findings are based on a thorough behavioural exploration of gene mutations related to ID in mice, focussing on males. It should be emphasized that around 33% of genes involved in ID (and those generated here) are X-linked genes. That is the main reason why only males were characterized in this study, although this might also be considered a limitation. The extension of phenotyping to females might have increased the strength of our findings. In this regard, in the context of another worldwide effort designed to generate and characterize null mouse models for all the genes of the mouse genome, we also characterized mutant females with some of the genes from the present collection. For example, Ptchd1−/− females also displayed several behavioural abnormalities including hyperactivity and cognitive deficits, extending the data observed in Ptchd1−/y males [86].