Large-Scale Functional Assessment of Genes Involved in Rare Diseases with Intellectual Disabilities Unravels Unique Developmental and Behaviour Profiles in Mouse Models

Major progress has been made over the last decade in identifying novel genes involved in neurodevelopmental disorders, although the task of elucidating their corresponding molecular and pathophysiological mechanisms, which are an essential prerequisite for developing therapies, has fallen far behind. We selected 45 genes for intellectual disabilities to generate and characterize mouse models. Thirty-nine of them were based on the frequency of pathogenic variants in patients and literature reports, with several corresponding to de novo variants, and six other candidate genes. We used an extensive screen covering the development and adult stages, focusing specifically on behaviour and cognition to assess a wide range of functions and their pathologies, ranging from basic neurological reflexes to cognitive abilities. A heatmap of behaviour phenotypes was established, together with the results of selected mutants. Overall, three main classes of mutant lines were identified based on activity phenotypes, with which other motor or cognitive deficits were associated. These data showed the heterogeneity of phenotypes between mutation types, recapitulating several human features, and emphasizing the importance of such systematic approaches for both deciphering genetic etiological causes of ID and autism spectrum disorders, and for building appropriate therapeutic strategies.


Introduction
Intellectual 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 Table 1. List of mutated genes, human variants, their functional implication in human syndromes, and mouse models generated in this study for these genes. LoF and GoF indicate loss-of-function and gain-of-function, respectively. SNPs indicate Single Nucleotide Polymorphism.  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.

Materials and Methods
The procedures carried out in this project were performed in agreement with the EC directive 2010/63/UE86/609/CEE, submitted to the French Ethics Committee 017 (Com'Eth) and received accreditation under number 2012-139.

Embryonic Pipeline
The viability/sub-viability of mutants was assessed by crossing heterozygous mice and scoring the offspring's genotypic distribution. Selected lines with homozygous/hemizygous (or heterozygotes) scores below the Mendelian ratio were analysed using the well-defined embryo pipeline that we developed in the IMPC [81]. Briefly, embryos were collected at specific developmental stages to determine their window of lethality ( Figure 1B) and further characterize their developmental defects.

Embryonic Pipeline
The viability/sub-viability of mutants was assessed by crossing heterozygous mice and scoring the offspring's genotypic distribution. Selected lines with homozygous/hemizygous (or heterozygotes) scores below the Mendelian ratio were analysed using the well-defined embryo pipeline that we developed in the IMPC [81]. Briefly, embryos were collected at specific developmental stages to determine their window of lethality ( Figure 1B) and further characterize their developmental defects.

Behavioural Phenotyping Strategy
For each mutant line, cohorts of wildtype (WT) and mutant male mice were generated for phenotyping (8-12 mice per genotype). Mice were group-housed (2-4 per cage) and allowed 1-2 weeks acclimation in the phenotyping area with controlled temperature (21-22 °C) under a 12-12 light-dark cycle (light on at 07 a.m.), with food and water available ad libitum. Behavioural testing was performed in 10 to 13-week-old adults and carried out in agreement with EC directive 2010/63/UE86/609/CEE, and under the ethics committee accreditation number 2012-139.
Animals underwent a standardized behavioural screen composed of tests to assess sensory and motor abilities, biological rhythm, pain sensitivity, anxiety-related behaviour, sensorimotor gating, learning and memory, social behaviour, and susceptibility to seizures according to the ARRIVES guidelines [82]. Behavioural protocols are described in the supplementary materials; most of them are thoroughly detailed in a recent volume of current protocols [83]. The order of tests was carefully defined from the least to the most stressful test to reduce the potential influence of repeated testing ( Figure 1C). Animals underwent a standardized behavioural screen composed of tests to assess sensory and motor abilities, biological rhythm, pain sensitivity, anxiety-related behaviour, sensorimotor gating, learning and memory, social behaviour, and susceptibility to seizures according to the ARRIVES guidelines [82]. Behavioural protocols are described in the Supplementary Materials; most of them are thoroughly detailed in a recent volume of current protocols [83]. The order of tests was carefully defined from the least to the most stressful test to reduce the potential influence of repeated testing ( Figure 1C).

Statistical Analysis
For each mutant line, phenotyping data were analysed using unpaired Student t-tests or repeated measures analyses of variance (ANOVA) with one between factor (genotype) and one within factor (time, etc.). Qualitative parameters (e.g., clinical observations) were analysed using the χ2 test. The level of significance was set at p < 0.05. A gene/phenotype heat-map was drawn based on p-values for each parameter (Table S1: Gene/phenotype heatmap).
In addition, other analyses were performed on all mutant lines related to categorized biological functions identified using appropriate parameters selected from the different behavioural tests. Fifty-seven (57) parameters were distributed in 10 categories representing biological functions (Table S2: categories of biological functions and related parameters). For each mutant line, a phenotype score (corresponding to the ratio of number of parameters having a phenotype) was calculated per biological function (Table S3: the phenotype scores). We considered that a parameter presents a phenotype if its adjusted p-value (using Benjamini-Hochberg method to control the false discovery rate within each biological function) was less than 0.05 (Supplementary Materials). Based on these phenotypic scores, Principal Component Analysis (PCA) was performed and a heatmap with cluster representation drawn (see the Section 3).

Gene/Phenotype Relationship
Fifty mutant lines were generated for 45 genes (Table 1 for genes and abbreviations). The viability of homozygous/hemizygous mutant mice was assessed by crossing the heterozygous mice. From the 50 mutant lines analysed, 66% displayed a homozygous/hemizygous lethality, 8% were sub-viable with less than the expected Mendelian progeny ratio [81], and 26% were viable ( Figure 1A). Concerning the lines carrying a point mutation, seven out of nine presented homozygous lethality, and haploinsufficiency or autosomal dominant lethality was also found for Med12 and Tubb3. Finally, of 16 lines with mutated genes on the X-chromosome (all alleles), seven were hemizygous lethal, eight viable and one showed sub-viability ( Figure 1A).
This high rate of lethality/sub-viability could be expected, as these genes were selected based on their specific involvement in neurodevelopmental disorders. Indeed, the overall rate of homozygous lethality assessed in IMPC is approximately one out of three mutant genes [81] but increases dramatically when disease-related or essential genes are mutated [84].
Among the sub-viable mutant lines, Med17 −/− showed decreased body weight, breathing difficulties, and died between P0 and eight weeks after birth. Necropsy examination revealed abdominal and pulmonary haemorrhage, hypoplasia of the thymus and heart failure.
These data strongly support the fact that around 65% of genes involved in ID are essential for survival at normal Mendelian ratios. To better understand the specific involvement of these genes in CNS functions, we generated brain specific conditional mutants.
Assessing the viability of mutant lines with a tissue specific Camk2a reporter revealed that five out of six lines were viable and one out of six lines was sub-viable.
We pursued the analysis with the standardized behavioural phenotyping of 27 mutant lines to detect a wide range of phenotypic traits affecting different CNS functions. As eight mutant lines were generated for X-linked genes with features found only in males, we focused the adult study on male mutant mice. The first observation of the gene/phenotype heatmap revealed three main classes of mutants based on activity phenotypes observed in the different behavioural tests (Table S1). The first group of mutated genes inducing a substantial increase in spontaneous activity, the second group of mutant lines with decreased spontaneous activity, and the third group of mutant lines with no change.

Hyperactivity Group
The hyperactivity group includes Cdk8 Camk2a/Camk2a (hereafter named Cdk8 Camk2a ), Ankrd11 Camk2a/Camk2a (named hereafter Ankrd11 Camk2a ), Atp6ap2 Camk2a/y , Il1rapl1 −/y , Prps1 Camk2a/y , Ptchd1 −/y , Arx Dup24/y , and Ascc3 Camk2a/Camk2a (hereafter named Ascc3 Camk2a ). These eight mutant lines showed a substantial increase in locomotor activity and stereotypic behaviour in different situations including actimetric cages (increased number of beam-breaks), the open field (higher distance travelled over the 30 min test) (Figure 2), the Y-maze, and social tests (increased number of visits) [85,86]. They also showed other behavioural alterations. For example, all Atp6ap2 Camk2a/y [85], Il1rapl1 −/y and Ptchd1 −/y [86] mutants had altered contextual and cued fear conditioning, displaying reduced percentage of freezing both during the context and the cued testing sessions (Figure 2). Ankrd11 Camk2a also had decreased contextual fear conditioning. In addition, Il1rapl1 −/y mice had altered spatial learning in the water maze with reduced number of platform crosses, while Ptchd1 −/y had decreased working memory both in the Y-maze and the object recognition tasks [86] (Figure 2). They also showed altered motor abilities, evidenced by decreased muscle strength in Ptchd1 −/y and Cdk8 Camk2a mutants ( Figure 2) and decreased startle response in the Atp6ap2 Camk2a/y mice [85].
This group of hypoactive mutants also had other behavioural alterations. Mbd5 +/− had decreased contextual and cued fear conditioning with decreased freezing performance, improved sociability exploring more the congener than an object, but had altered social memory displaying no preference of novel congener, and decreased startle response (Figure 4). On the other hand, Ehmt1 +/− , Ehmt1 +/− /Ehmt2 +/− had altered recognition memory with recognition index around the chance level, altered social memory (for Ehmt1 +/− ) and increased startle response (Figure 4). Cdkl5 −/y displayed decreased prepulse inhibition (PPI) and increased thermal pain threshold (Figure 4), while Mecp2 −/y mutants had decreased startle reactivity and PPI, showed substantial tremors (100% of accuracy), and had a decreased thermal pain threshold. Finally, all these hypoactive lines, except Mecp2 −/y , which showed the opposite phenotype, displayed a trend of increased anxiety with either  This group of hypoactive mutants also had other behavioural alterations. Mbd5 +/− had decreased contextual and cued fear conditioning with decreased freezing performance, improved sociability exploring more the congener than an object, but had altered social memory displaying no preference of novel congener, and decreased startle response (Figure 4). On the other hand, Ehmt1 +/− , Ehmt1 +/− /Ehmt2 +/− had altered recognition memory with recognition index around the chance level, altered social memory (for Ehmt1 +/− ) and increased startle response (Figure 4). Cdkl5 −/y displayed decreased prepulse inhibition (PPI) and increased thermal pain threshold (Figure 4), while Mecp2 −/y mutants had decreased startle reactivity and PPI, showed substantial tremors (100% of accuracy), and had a decreased thermal pain threshold. Finally, all these hypoactive lines, except Mecp2 −/y , which showed the opposite phenotype, displayed a trend of increased anxiety with either Figure 2. Selected behavioural alterations in the hyperactivity group. Cdk8 Camk2a , Ankrd11 Camk2a , Atp6ap2 Camk2a/y , Ilrapl1 −/y and Ptchd1 −/y [86] showed increased activity in the circadian activity and open field tests reflected by a higher number of beam breaks and distance travelled compared to wildtypes. Data are expressed as the mean ±SEM of front and back successive beam breaks (circadian activity) and distance (open field) across time, and analysed using repeated measures ANOVA, followed by t-tests for each time point. Cdk8 Camk2a , Ankrd11 Camk2a , and Ilrapl1 −/y showed altered learning performance in the water maze (decreased number of platform crosses for Ilrapl1 −/y ), fear conditioning (decreased percentage of freezing for Ankrd11 Camk2a and Ilrapl1 −/y ) or object recognition (Cdk8 Camk2a ), or motor deficits in the grip test (Cdk8 Camk2a and Ptchd1 −/y ). Data are expressed as mean ± SEM% freezing (fear conditioning), or scattergrams with the median for number of platform crosses (water maze), recognition index (object recognition), or muscle strength (grip test), and analysed using either repeated measures ANOVA followed by t-tests, or Student t-tests for single time points. * p < 0.05, ** p < 0.01, *** p < 0.001 vs. WT; § p < 0.05 vs. the chance level (only WT displayed good performance while mutants performed at the hazard level).

PCA and Cluster Analysis
Based on association studies, additional statistical analysis was performed on the 21 mutant lines whose genes are closely associated with ID (six genes were excluded). A graphical representation of phenotype scores was done using a heatmap combined with a dendrogram showing the arrangement of mutant line clusters produced by hierarchical clustering (Figure 5 and Supplementary materials). Correlated variables were grouped together on a circle of correlations. The smaller an individual coordinate on an axis, the smaller its contribution to the component. The three first components explained 70.27% of the variance ( Figure 5A). Data are mean ± SEM (fear conditioning, startle reactivity) or scattergrams with the median (object recognition, social test, Y-maze, marble burying, hot plate). Data are analysed using either repeated measures ANOVA followed by t-tests or Student's t-test for single time points. * p < 0.05, ** p < 0.01, *** p < 0.001 vs. WT; § p < 0.05, § § p < 0.01, § § § p < 0.001 vs. the chance level; £ vs. Ehmt1 +/− /Ehmt2 +/− .

PCA and Cluster Analysis
Based on association studies, additional statistical analysis was performed on the 21 mutant lines whose genes are closely associated with ID (six genes were excluded). A graphical representation of phenotype scores was done using a heatmap combined with a dendrogram showing the arrangement of mutant line clusters produced by hierarchical clustering (Figure 5 and Supplementary Materials). Correlated variables were grouped together on a circle of correlations. The smaller an individual coordinate on an axis, the smaller its contribution to the component. The three first components explained 70.27% of the variance ( Figure 5A). Dyrk1a Dlx5−6/+ , Wdr62 −/− , Prps1 Camk2a/y , Cdk8 Camk2a , Ankrd11 Camk2a , Arx Dup24/y , and Cntnap2 −/− . The last group of mutants displayed a few changes or no phenotype. PCA was performed to visualise potential links between biological functions and mutant line similarities ( Figure  5C,D, and Supplementary materials). On the one hand, a group of lines including Cdkl5 −/y , Il1rapl1 −/y , Ptchd1 −/y , Atp6ap2 Camk2a/y , Mbd5 +/− , and Ehmt1 +/− displayed alterations mainly in activity, repetitive behaviour, novelty exploration and anxiety (Axis 1). On the other hand, mutant lines including Cdk8 Camk2a , Mecp2 −/y and Setbp1 +/− showed deficits in motor abilities and pain sensitivity, while Atp6ap2 Camk2a/y and Ilrapl1 −/y showed learning and memory deficits. The first observation of Figure 5B shows a group of mutant lines, including Atp6ap2 Camk2a/y , Il1rapl1 −/y , Ehmt1 +/− , Mbd5 +/− , Cdkl5 −/y , and Ptchd1 −/y , with a high number of functional alterations. The second cluster includes gene mutations with a moderate number of altered functions, and includes Mecp2 −/y , Setbp1 +/− , Entpd1 −/− , miR137 +/− , Dyrk1a Dlx5−6/+ , Wdr62 −/− , Prps1 Camk2a/y , Cdk8 Camk2a , Ankrd11 Camk2a , Arx Dup24/y , and Cntnap2 −/− . The last group of mutants displayed a few changes or no phenotype. PCA was performed to visualise potential links between biological functions and mutant line similarities ( Figure 5C,D and Supplementary Materials). On the one hand, a group of lines including Cdkl5 −/y , Il1rapl1 −/y , Ptchd1 −/y , Atp6ap2 Camk2a/y , Mbd5 +/− , and Ehmt1 +/− displayed alterations mainly in activity, repetitive behaviour, novelty exploration and anxiety (Axis 1). On the other hand, mutant lines including Cdk8 Camk2a , Mecp2 −/y and Setbp1 +/− showed deficits in motor abilities and pain sensitivity, while Atp6ap2 Camk2a/y and Ilrapl1 −/y showed learning and memory deficits.

Discussion
In 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 Tubb3 M388V/+ embryos are in line with those previously reported for the Tubb3 R262C/R262C mouse 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 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 Arx Dup24/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 (Arx Dup24/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 Ankrd11 Camk2a , Atp6ap2 Camk2a/y , and Prps1 Camk2a/y . Ankrd11 Camk2a and Atp6ap2 Camk2a/y displayed substantial hyperactivity and stereotypic behaviour, in-creased exploration and reduced anxiety, and altered learning and memory. Interestingly, Atp6ap2 Camk2a/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, Prps1 Camk2a/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 Prps1 Csp4/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 Cdk8 Camk2a 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 Ascc3 Camk2a mutants 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 Arx Dup24/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].

Conclusions
The results of the present study allowed us to establish a broad gene-phenotype relationship map for a wide range of genes involved in several neurodevelopmental disorders with ID, or potentially involved in ID. Several of the mutant lines studied reproduced, as far as possible, the human mutation types, and displayed strong phenotypic similarities with patient features, constituting interesting genetic tools to better understand human syndromes with ID and making it possible to establish new potential therapeutic strategies.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/biomedicines10123148/s1. Figure S1: Mouse models generated and phenotyped in the Gencodys consortium; Figure S2: Cranial and cervical vertebra abnormalities of Setbp1 −/− mice at E15.5 and E18.5; Figure S3: Cranial nerve and associated ganglia, as well as dorsal root ganglia abnormalities in Tubb3M388V/M388V mutant embryos; Figure S4: Severe craniofacial abnormalities in Med25 −/− mutant embryos and fetuses; Table S1: Gene/phenotype heatmap drawn from main parameters of different tests from 27 characterized mutant lines; Table S2: Biological function categories and related parameters, established for PCA analysis and cluster representation; Table S3: The phenotype scores calculated for each mutant line and functional category to perform PCA analysis and cluster representation.

Data Availability Statement:
The data presented in this study are available upon request.