Biallelic mutations in the death domain of PIDD1 impair caspase-2 activation and are associated with intellectual disability

PIDD1 encodes p53-Induced Death Domain protein 1, which acts as a sensor surveilling centrosome numbers and p53 activity in mammalian cells. Early results also suggest a role in DNA damage response where PIDD1 may act as a cell-fate switch, through interaction with RIP1 and NEMO/IKKg, activating NF-κB signaling for survival, or as an apoptosis-inducing protein by activating caspase-2. Biallelic truncating mutations in CRADD—the protein bridging PIDD1 and caspase-2—have been reported in intellectual disability (ID), and in a form of lissencephaly. Here, we identified five families with ID from Iran, Pakistan, and India, with four different biallelic mutations in PIDD1, all disrupting the Death Domain (DD), through which PIDD1 interacts with CRADD or RIP1. Nonsense mutations Gln863* and Arg637* directly disrupt the DD, as does a missense mutation, Arg815Trp. A homozygous splice mutation in the fifth family is predicted to disrupt splicing upstream of the DD, as confirmed using an exon trap. In HEK293 cells, we show that both Gln863* and Arg815Trp mutants fail to co-localize with CRADD, leading to its aggregation and mis-localization, and fail to co-precipitate CRADD. Using genome-edited cell lines, we show that these three PIDD1 mutations all cause loss of PIDDosome function. Pidd1 null mice show decreased anxiety, but no motor abnormalities. Together this indicates that PIDD1 mutations in humans may cause ID (and possibly lissencephaly) either through gain of function or secondarily, due to altered scaffolding properties, while complete loss of PIDD1, as modeled in mice, may be well tolerated or is compensated for.


Introduction
Biallelic missense mutations in the gene encoding caspase and RIP adaptor with death domain, CRADD (also known as RAIDD), have been reported to cause intellectual disability (ID) with enlarged head circumference (MRT34; MIM 614499; ref. 1 ). Mild lissencephaly has also been reported in affected individuals [2][3][4][5][6] , and megalencephaly without obvious cortical lamination defects in Cradd null mice 2 . Together with PIDD1 (p53-Induced Death Domain protein 1), CRADD can activate caspase-2-an endopeptidase thought to regulate apoptosis upon DNA damage in the so-called PIDDosome complex [see ref. 7 reviewed in refs. 8,9 ]. However, recent studies suggest a crucial role in p53 activation upon centrosome accumulation 10,11 . Alternatively, PIDD1 can also interact with RIP1 and NEMO upon DNA damage and promote nuclear factor κB (NF-κB) activation for cell survival and sterile inflammation 12,13 . Several reports implicate caspase-2 in neuronal cell death caused by NGF or beta-amyloids 8,14 . Whether these effects of caspase-2 require PIDDosome formation remains less clear 15 . Caspase-2 can also process tau protein leading to its accumulation in dendritic spines, which have been found to be increased in cognitively impaired mice and humans 16 . Moreover, loss of caspase-2 associates with increased anxiety behavior, but not memory loss in mice 17 . Together, these studies indicate that caspase-2 activity controls neuronal stress responses and cognitive brain function. As PIDDosome-independent activation of caspase-2 has been reported in hematopoietic cells, fibroblasts, and neurons 14,18 , it remains unclear if PIDD1 is needed for normal brain function.
Recently, we reported the identification of a biallelic nonsense mutation (Chr11:799453G>A; NM_145886.3: c.2587C>T; p.Gln863*) within the PIDD1 gene in two Pakistani families segregating with non-syndromic ID 19 . PLINK: pair-wise PI_HAT values between members of the two families <0.1, suggested that the two families are not closely related. Linkage analysis for the two families gave a LOD score of 3.7. The variant is not present in the Greater Middle East Variome (GMEV) 20 dataset, and in the gnomAD dataset (v2.1.1; gnomad.broadinstitute.org) of 246,378 chromosomes, a single (heterozygous) occurrence is reported (allele frequency = 4E−6). The mutation is predicted to truncate the protein midway through the Death Domain (DD). As the mutation occurs within the last exon, nonsense-mediated mRNA decay is not predicted, and it is anticipated that the etiopathological mechanism is through the loss of function of the DD, but this has not been formally tested. Three additional ID families with biallelic PIDD1 mutations have since been identified, including the missense mutation chr11:799846G>A; NM_145886.3: c.2443C>T; p.Arg815Trp (2 (heterozygous) out of 266,508 alleles in gnomAD; MAF = 7.5E−6), a splice acceptor site mutation chr11: 800015C>T; NM_145886.3: c.2275-1G>A (not present in gnomAD) 21 , and a nonsense mutation chr11:800770>A; NM_145886.3: c.1909C>T; p.Arg637* (4 (heterozygous) out of 152,564 alleles in gnomAD; MAF = 2.6E−5). The Arg815Trp mutation, also within the DD, is predicted to be pathogenic, as is the splice variant. Pedigrees and mapping information are shown in Fig. 1. The constitutive processing of PIDD1 protein is demonstrated in Fig. 2, indicating the position and effect of the reported mutations relative to the cleaved PIDD1 and its known functional domains.
Using cell biological assays we demonstrate that all these mutations act as hypomorphs, severely affecting the ability of PIDD1 to activate p53 in response to extra centrosomes. Behavioral studies in mice do support a role for the PIDDosome in anxiety behavior but did not reveal cognitive deficits. We conclude that in mice loss of PID-Dosome function is well tolerated or compensated for, while truncating mutations found in humans may lead to additional effects, related to the scaffolding function of PIDD1 in multiprotein complexes, such as those affecting inflammation, chromosome segregation, translesion synthesis, or nucleolar function and dynamics (reviewed in ref. 11 ).

Materials and methods
Cloning, mutagenesis, cell culture, and transfections HEK293T human embryonic kidney cells (ATCC CRL-3216 TM ) were used in different experiments. Full-length PIDD1 (NM_145886) was amplified using reverse transcriptase PCR (RT-PCR) from cDNA generated from total RNA extracted from HEK293T cells (all reagents from Thermo Fisher scientific, Waltham, MA, USA), and cloned into mammalian expression vector pDEST47 and pDEST53, C-terminally and N-terminally GFP-tagged, respectively, using the Gateway TM cloning system (Thermo Fisher Scientific, Waltham, MA, USA). In addition, an N-terminally tagged GFP construct was generated in pDEST53 with just the DD 95 amino acid residues of PIDD1 ( Supplementary Fig. 1). For the Nterminally tagged GFP constructs and C-terminally tagged GFP constructs, p.Q863* and p.R815W mutations were introduced through PCR-based site-directed mutagenesis according to the manufacturer's instructions (Quikchange Lightning site-directed mutagenesis kit; Agilent Technologies, Santa Clara, CA). Full-length wildtype (WT) CRADD (NP_001307028) was amplified using RT-PCR from cDNA generated from total RNA extracted from HEK293T cells (all reagents from Thermo Fisher Scientific, Waltham, MA, USA), and was cloned in pDsRed-Monomer-N1 vector (Clontech, Mountain View, CA). For co-localization experiments, in order to cotransfect the WT-CRADD_DsRed plus WT-PIDD1-CT-GFP, WT-CRADD_DsRed plus PIDD1-CT-GFP_Q863*, and WT-CRADD_DsRed plus PIDD1-CT-GFP R815W fusion proteins in HEK293T cells, either PolyFect (Qiagen) or polyethylenimine based transfection reagents were used (jetPEI ® , PolyPlus, Illkirch, France) following the manufacturers' instructions. Similarly, for coimmunoprecipitation (Co-IP) studies, the GFP-PIDD1-DD constructs and WT-CRADD_DsRed were used (see Supplementary Table 1 for primer details).

Cell harvesting and protein extraction
One T75 flask of HEK293T cells at >90% confluence were used for all Co-IP experiments. Cells were washed with 500 μL of ice-cold phosphate buffer saline (PBS) for each flask, and harvested with a soft sterile cell scraper. Cell pellets were washed three times by resuspending the cells gently in ice-cold PBS and centrifugation at 1500 × g at 4°C for 5 min. Cell pellets were stored at −80°C until use.
For protein isolation, pellets were resuspended in lysis buffer (50 mM HEPES-KOH, 100 mM KCl, 2 mM EDTA, B Location and overlap of HBD regions, and locations of the identified mutations relative to the exonic structure of PIDD1. The known PIDD1 protein cleavage products, N-PIDD1, C-PIDD1, and CC-PIDD1 are shown, as reported by Tinel et al. (2007). MRI images for AS110 IV-3. C Coronal T2 sequence shows subtle smooth gyration and shallow sulcation affecting the frontal and temporal lobes, compatible with lissencephaly. D Axial T2 sequence shows subtle smooth gyration and shallow sulcation affecting the frontal and temporal lobes, compatible with lissencephaly. E Axial T2 sequence shows subtle smooth gyration and shallow sulcation affecting the frontal lobes, compatible with lissencephaly. MRI of the Manipal-1 proband (II-3) showing pachygyria in F T1 axial, G T2 axial, and H sagittal section showing mildly dysmorphic (short and thick) corpus callosum. 0.1% NP40, 10% glycerol, 0.25 mM Na 3 VO 4 , 50 mM β-glycerolphosphate, 1 mM NaF, 1 mM DTT) and incubated on ice for 30 min. Cell debris were spun down for 15 min at 14-160,000 × g at 4°C. In all, 10 U DNAse I per mL of lysate (Fermentas, Waltham, MA) were added. Protein concentrations were quantified using bicinchoninic acid protein assay kit (Pierce TM , Thermo Fisher Scientific, Waltham, MA). The concentration of protein in the lysate was typically~5-10 mg/mL. An aliquot of lysate was also collected for western blot analysis and stored at −80°C until used for further experiments.

Cell-based protein degradation
In order to check the stability of PIDD1 protein carrying the identified missense, nonsense, and splice mutations, in comparison to WT PIDD1, we transfected HEK293T cells as described previously, and after 24 h added 10 µg/mL cycloheximide (CHX; Sigma-Aldrich, St Louis, MO) to prevent translation of new protein, plus or minus proteosomal inhibitor MG-132, and harvested at the indicated times. Samples collected at each time point were then analyzed by western blot, using anti-FLAG-M2 antibody (Sigma-Aldrich, St Louis, MO).

Data analysis
Bio-Rad's Quantity One ® ID analysis software and Image Lab TM software were used to process gel images. For cellular images Olympus FluoView software to calculate Pearson's correlation coefficient (PCC) for the colocalization of chromocenters (DAPI) and recombinant protein (GFP & DsRed monomer N1) was used. The descriptive statistical data analysis was calculated using Microsoft Excel and two-tailed Student's t-test for statistical significance was calculated using GraphPad online tool (GraphPad Software Inc., San Diego, CA).

Exon-trap assay for c.2275-1G>A mutation
Exon 15 of the wild-type PIDD1 gene, as well as portions of its flanking regions (235 bp upstream and 368 bp downstream), were PCR-amplified from the genomic DNA of an unrelated (and unaffected) individual, restriction digested with SalI and ligated into the MCS of the pET01 vector (MoBiTec GmbH). The pET01 MCS is located between two bordering exons. In order to generate the pET01-PIDD1 vector with the non-synonymous mutation in intron 11 (c.2275-1G>A), site-directed mutagenesis was performed. Briefly, PCR using mutagenic primers was used to induce the base substitution, followed by digestion of the parental template using DpnI, phosphorylation of the mutagenic amplicon by T4 PNK (NEB), and circularization using Quick T4 DNA Ligase (NEB, Ipswich, MA). Following verification by Sanger sequencing, the wild type, mutant, and control (empty pET01) vectors were separately transfected (Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA)) into HEK293T cells. Twenty-four hours post-transfection, RNA was extracted and, after first-strand cDNA synthesis with random hexamers, RT-PCR was performed using primers complementary to the pET01 exons that border the MCS. Following gel electrophoresis, bands were excised, purified, and submitted for Sanger sequencing. All primers are listed in Supplementary Table 2.

PIDD1 overexpression and functional analysis
C-terminally FLAG-tagged versions of either WT PIDD1 or PIDD1 point mutants Q863*, R815W, c.2275-1G>A (splice acceptor), or G876S, a variant found in heterozygous form in an unrelated ID patient (John B. Vincent, unpublished data) were generated by sitedirected mutagenesis and cloned into pcDNA5 expression plasmid by PCR cloning. The c.2275-1G>A mutant variant missing exon 15 was generated by producing a PCR fragment using a CMV promoter-specific forward primer and a selective hPIDD1_BamHI_Exon_DD reverse primer (see Supplementary Table 3) with an additional C-Terminal BamHI restriction site to clone the PCR fragment back into the original pcDNA5 plasmid, after BamHI digestions, thereby keeping the Cterminal FLAG-tag intact. To allow reconstitution of PIDD1 KO cells 10 silent mutations were additionally introduced in the PAM sequence to neutralize the originally used sgRNA (ccccatcaccatccgctatcggc) in these cells (FAVA REF G&D).
The functionality was assessed by transiently transfecting the constructs into the PIDD1 KO U2OS or A549 cell lines 10 in duplicates, using Metafectene (Biontex Laboratories GmbH, Munich, Germany) as described in the user manual (six-well plate, 400,000 cells/well). After overnight transfection, one subset of the duplicate was treated with 2 μM ZM447439 for 48 h, while the other was left untreated. Subsequently cells were lysed and immunoblotted for MDM2 (MAY-113; Invitrogen, Thermo Fisher Scientific, Waltham, MA) and GAPDH (CS14C10; Cell Signaling Technology, Danvers, MA). Construct expression and processing was detected in HEK293T cells using Sigma anti-FLAG-M2 antibody. For immunoprecipitation, FLAG-M2 agarose beads (Sigma) were used to enrich FLAG-tagged PIDD1 variants.

Animal experiments
Mice were generated as previously described (Pidd1 tm1.Anvi ) 18 , with exons 3-15 removed. KO mice and WT littermates underwent a physical exam at the age of 3 and 12 months. For the 3-month-old group, 10 male WT and 9 male KO mice were studied. For the 12-monthold group, 8 female WT mice and 8 female KO mice were studied. This exam included observation of ambulation in a novel cage, hearing (response to finger snap) and visual ability (visual cliff), condition of fur and whiskers, body weight, motor skills (beam walk), and muscular strength (wire hang test). Ambulation was observed for 5 min, looking for overall activity, stereotypies, seizures, and other abnormal behavior. The beam walk test was performed on square wooden sticks with 1 and 2 cm diameter and 1 m length. Mice were placed in the center, and the time needed to reach the end was measured (three repetitions, best taken). The wire hang test was performed by placing mice on a wire grid, which was inverted after the mouse grabbed hold. Maximum duration was 60 s. For younger (3 months) mice the open field, light-dark, and elevated plus maze tests were also performed. Mice were transferred to the ante-room of the testing facility 24 h before the commencement of experiments. The free access to food and water, the climate, and the light-dark cycle were kept constant. The tests were performed between 9 a.m. and 1 p.m. All tests were video monitored and evaluated by an experimenter blinded to the genotype of the animals. The procedures for the open field, elevated plus maze, and light-dark choice tests were performed as published earlier 22 in accordance with the recommendations of EMPRESS (European Mouse Phenotyping Resource of Standardised Screens; http://empress.har. mrc.ac.uk).
In short, the open field arenas had a size of 50 × 50 cm and were illuminated to 150 lux. Mice were observed over 10 min, measuring time, distance traveled, and number of entries into three subfields: center (central 16% of overall area), border (8 cm along walls), and intermediate.
The light-dark test was performed in the same arenas with a black box inserted, which covered 1/3 of the area. Light was set to 400 lux and mice tested for 5 min measuring time, distance traveled, and number of entries into the light compartment.
The elevated plus maze test consisted of four arms, two closed (20 cm walls) and two open arms, each 50 × 5 cm in size elevated about 80 cm above ground. Exploratory activity on the open arms was tested over 5 min at 180 lux.

Bioinformatic transcriptomic analyses
In order to better understand the possible connection between PIDD1 and the brain developmental phenotypes observed in individuals with mutations disrupting PIDD1 function, we used publically available transcriptomic datasets (gene expression microarray and RNAseq) for PIDD1, and co-expression with related genes to explore commonalities in temporaspatial expression. Firstly, we used data from the Genotype-Tissue Expression (GTEx) project to explore gross anatomical gene expression for humans through the GTEx Portal (https://www. gtexportal.org/home/). Secondly, we used Allen Brain Atlas data (www.brainspan.org) to look at temporospatial mRNA expression for human brain from 8 weeks postconception to 40 years of age, using RNAseq and gene expression microarray 23 , and from RNAseq data from the PsychEncode dataset 24 . Adult Human transcriptomic comparisons were performed with the Allen Human Brain Atlas data 25 . Co-expression analysis of BrainSpan datasets was performed as described previously 26 (https://hbaset. msl.ubc.ca/). Thirdly, we used single-cell RNAseq data from fetal and postnatal mouse brains, for which cell types have been classified according to spatial and taxonomical cluster analysis (www.mousebrain.org) 27,28 . Data were analyzed from the single-gene perspective (PIDD1/Pidd1 alone), also with PIDDosome interactors (PIDD1/Pidd1+ CRADD/Cradd+CASP2/Casp2), or with a gene set of PIDD1 interactors (P gene set: PIDD1, CRADD, CASP2, MADD, FADD) plus genes relevant to lissencephaly (L gene set: RELN, TUBA1A, NDE1, KATNB1, CDK5, ARX, DCX, LPHN1, LPHN2, LPHN3).

Results
Familial history and clinical observations for PIDD1 mutation families Figure 1A shows pedigree information and segregation pattern of five consanguineous families with mutations in PIDD1. Family M278, with two affected siblings, has a splice acceptor mutation (Chr11:800015C>T; NM_145886.3: c.2275-1G>A), just ahead of the penultimate (14th) exon. The variant is not present in the gnomAD database. Family M8700004, with three affected siblings, has a missense mutation (Chr11:799846G>A; NM_145886.3:c.2443C>T; p.Arg815Trp), converting a residue situated within the DD that is positively charged across eutherian mammalian evolution to a non-polar residue (Supplementary Fig. 2). This variant (rs758859772) is present in the gnomAD database, with an allele frequency of <0.0001, with no homozygous occurrences. Indian family Manipal-1 has not been reported previously. The female proband presents with mild ID, and has the homozygous nonsense mutation (Chr11:800770>A; NM_145886.3: c.1909C>T; p.Arg637*; see Supplementary Fig. 3 for electropherograms)). This variant (rs578222814) is present in the gnomAD database, with an allele frequency of <0.0001, with no homozygous occurrences. None of these variants are present in the GMEV.
A more detailed clinical analysis of the five families is provided in the Supplementary Information and the genomic location on Chr11 as well as intron-exon structure of the PIDD1 gene are depicted in Fig. 1B. Briefly, eight out of eleven affected family members (for whom clinical information was available) were reported to have epilepsy (generalized tonic-clonic seizures). Clinical magnetic resonance imaging (MRI) reports were available for two affected family members with the Gln863* mutation, as well as the proband from Manipal-1 with Arg637*. Although both reports from AS110 indicated normal anatomy, our own reexamination of the images indicated: in AS110 IV-3, a 30-year-old male, subtle lissencephaly of the cortical layers of the cerebral hemispheres, most pronounced in bilateral frontal lobes (Fig. 1C-E); in AS105 III-3, a 5-year-old male, mild decreased volume of the body and splenium of the corpus callosum. MRI from Manipal-1 II-3 shows broad smooth gyri in bilateral frontotemporal lobes, with reduced sulcations, suggestive of the lissencephaly-pachygyria complex. Thus, as with CRADD, mutations in PIDD1 may also be associated with lissencephaly. Although normal gait was reported for all available affected members, bradykinesia was evident in AS110 IV-1 and IV-3, and slurred speech was reported for family M8700004. Affected individuals were classified as moderate to severe intellectually disabled. Vision and hearing were reported as normal, although one individual was reported as having convergent strabismus. Aggressive behavior was a common feature in the Gln863* families, but not the splice site family. Psychosis was also reported for the affected Arg815Trp family members by the mother, and confirmed after evaluation by a psychiatrist.

Mutations in the PIDD1 DD cause mis-localization of CRADD
As with the nuclear pore protein Nup98/96, PIDD1 protein auto-processes itself by an intein-like mechanism, which gives rise to a leucine-rich PIDD1-N and PIDD1-C fragment containing the DD. PIDD1-C is further autocleaved into PIDD1-CC 29 . Figure 2A highlights the autoprocessing sites and residue count of the resulting cleaved PIDD1 proteins. All four mutations (plus the G876S variant) are either within or ahead of the DD (Q863*, R815W, G876S, R637*), or impacting the DD (c.2275-1G>A) by potentially impairing splicing, leading to a premature stop and protein truncation (Fig. 2B).

DD mutations do not prevent auto-proteolysis or impact stability of PIDD1
First, we explored the relevance of the splice acceptor site mutant. Since access to fresh RNA from the Iranian family M278 was not feasible, and thus we were unable to check directly the effects on splicing of the c.2275-1G>A mutation and any resulting changes to the PIDD1 open reading frame, we opted to construct an exon-trap construct as a model system to investigate the likely effects of the mutation and confirm in silico predictions of exclusion of exon 15 from the mRNA. Exon-trap analysis showed skipping of exon 15, with direct splicing from exon 14 to the terminal exon 16 (Supplementary Fig. 4) resulting in the substitution Arg759Gly with premature termination at codon 760 (p.Arg759Glyfs*1) leading to the exclusion of the entire DD (Fig. 2B).
Next, we transiently transfected HEK293T cells with Cterminal FLAG-tagged PIDD1 in its WT or various mutant forms. This analysis revealed normal processing of PIDD1 in case of the R815W or G876S variant, where PIDD1-FL, -C and -CC were present at similar levels. However, PIDD-Q863* appears to be poorly expressed and showed little PIDD1-FL or PIDD1-CC species, both only visible under extended exposure (Fig. 4A). The splice acceptor mutation (c.2275-1G>A), however, appeared to affect mainly the formation of PIDD1-CC. As a note on the side, probing the same lysates with the PIDD1-specific antibody clone ANTO1 suggests that the epitope recognized must lie within the DD downstream of Q863 (Fig. 2B). Together we conclude that single point mutations G876S and R815W do not affect autoprocessing or protein stability, while mutations that truncate the DD can interfere with protein stability (Q863*) or interfere with autoprocessing and efficient generation of PIDD1-CC, as in case of c.2275-1G>A, that leads to exon skipping and premature stop in position 760.
Cycloheximide-chase assays were performed to check whether the PIDD1 mutant proteins undergo accelerated protein degradation; however, it appears that WT and mutant PIDD1 are relatively stable, and, even when tracked over 4 h (Fig. 4B) or 24 h (Fig. S1 (D), do not show substantial degradation, and any differences between WT and mutants are minimal. Thus, the PIDD1 mutations identified do not significantly impact either autoprocessing or protein stability.

Pathologic PIDD1 mutants fail to interact with CRADD and activate caspase-2
It was previously shown that PIDD1 activates caspase-2 in the PIDDosome in a CRADD-dependent manner in response to cytokinesis failure or accumulation of extra centrosomes 10 . In turn, caspase-2 cleaves MDM2 which leads to p53 activation and p21 induction. Hence, we tested if PIDD1 mutants can still interact with CRADD. To this end, FLAG-tagged versions of PIDD1 were overexpressed in 293T cells followed by immunoprecipitation analysis. This analysis revealed that exogenous PIDD1 can interact with endogenous CRADD in its WT or G876S mutant form, while the premature stop or splice site-induced truncation mutants lost interaction with CRADD and the R815W point mutant showed strongly reduced binding (Fig. 5A). Consistent with this WB of N-terminally GFP-tagged PIDD1-Q863* showed a faster migrating PIDD1-FL band compared to WT, but similar size to WT for a stronger band at~75 kDa, likely representing the PIDD1-N cleavage product (Supplementary Fig. 2A). The Co-IP of GFP-tagged PIDD1-DD with CRADD demonstrated ablation of interaction for the Q863* and R815W mutations (Supplementary Fig. 2B, C).
To explore the potential consequences of mutation for PIDDosome function we performed reconstitution assays where PIDD1-deficient U2OS osteosarcoma and A549 lung-adenocarcinoma cells were used to transiently express sgRNA resistant WT, mutant, or truncated PIDD1 proteins. Cleavage of MDM2 by active caspase-2 was used as a functional readout in western analyses. Using this assay, we can conclude that all PIDD1 mutants bar the G876S variant have lost the ability to activate caspase-2, and hence constitute loss-of-function alleles. The G876S mutant appears to act as a mild hypomorph in this assay (Fig. 5B).

Mouse phenotyping
The physical examination of the Pidd1-deficient mice did not reveal any differences between KO and WT genotypes, neither at 3 nor at 12 months of age. Overall, KO mice tended to start ambulation of the novel cage slower than WTs, but did not show abnormal behavior or signs of distress or panic. Fur and whiskers were generally well, body weight was similar (25 ± 1.6 vs. 27 ± 1.3 at 3 months and 31 ± 4.1 vs. 29 ± 3.5 at 12 months for WT and KO, respectively. Data represent mean ± SD). Vision was intact at 3 months of age, but impaired at 12 months of age. However, this is normal for the C57BL/6N strain. At 3 months of age, four out of nine KO mice did not stay the entire time in the wire hang test, while all WT mice did. However, the difference did not reach statistical significance. At 12 months of age, one out of eight KO and two out of eight WT mice dropped before 60 s. The light-dark and elevated plus maze tests did not reveal any difference ( Supplementary  Fig. 5). By contrast, a mild but statistically significant anxiolytic-like effect was observed in the open field test (see Fig. 6).

Bioinformatic transcriptomic analyses
Gross anatomical transcriptional surveying through the GTEx data shows broad expression of human PIDD1 mRNA in multiple tissues, but with higher expression in brain (cerebellum) (Supplementary Fig. 6). BrainSpan exon microarray and RNAseq data analysis and Psy-chEncode RNAseq data analysis for PIDD1 alone, for PIDDosome genes (PIDD1+CRADD+CASP2), or for PIDDosome and lissencephaly related (P+L) genes, shows higher expression and co-expression during early human development (see heat maps in Supplementary Fig. 7). Coexpression of PIDD1+CRADD+CASP2 in six adult brains (through area under the curve (AUC) analysis) is mainly restricted to corpus callosum, cerebellum, and thalamus ( Supplementary Fig. 8).
Analysis of the single-cell RNAseq data from developing mouse brains 27,28 shows overall sparse neuronal expression for Pidd1, but with a small cluster in Slc17a6-coexpressing (Slc17a6 is used as a marker for glutamatergic neurons) forebrain neurons at age E12−16 postconception days (pcd), but with higher density transcription among cells classified as radial glia, neural crest, choroid, and ectodermal (see Supplementary Fig. 9a Mouse brain cells reference map, and 9b). This is in clear contrast to other PIDDosome components Cradd, and Casp2, which are more abundantly and ubiquitously expressed in neuronal and non-neuronal cells (Supplementary Fig. 9c, d). Co-expression analysis of Pidd1 + Cradd + Casp2 is highly specific. Ranking co-expression of the three genes by area under the curve (AUC) analysis for postnatal mouse 27 (but data not available for fetal mouse) shows most of the co-expression of Pidd1, Cradd, and Casp2 being in sympathetic and peripheral neurons and glia (e.g. in cholinergic sympathetic neurons (SYCHO1), and cholinergic enteric neurons (ENT8) and glia (ENTG1,3,5)). Granule neuroblasts from dentate gyrus is the only cell cluster from the central nervous system (CNS) with significant co-expression (p < 0.05) (Supplementary Table 4).
It should be noted that gene transcription profiles do not necessarily correlate directly with expression of the encoded protein; however, such data are frequently used as a proxy for Fig. 5 The effect of PIDD1 mutations on interacton with CRADD and caspase-2 activation. A PIDD1 immunoprecipitation pull down of CRADD. HEK293T cells transfected with vectors encoding GFP or FLAG-tagged WT or mutant PIDD1 were lysed after 24 h, and subjected to immunoprecipitation, followed by western analysis using antibodies recognizing the FLAG-tagged PIDD1, or CRADD, in the IP sample or flow through. B Functional effect of PIDD1 mutation on Caspase-2 activity through MDM2 cleavage assay. PIDD1-deficient U2OS or A549 cells generated by CRISPR/Cas9 technology were transiently transfected with the same constructs used in A and treated with the AuroraB kinase inhibitor ZM447439 to induce cytokinesis failure and PIDDosome activation. Cell lysates were probed with an antibody recognizing the cleaved variant of MDM2. Reprobing with an antibody recognizing GAPDH was done to confirm comparable protein loading. Fig. 6 Open field testing on Pidd1 knockout mice. Open field testing on Pidd1 KO mice (n = 9) and WT littermates (n = 10), as performed on 3-month-old male mice, measuring the time spent in the center of the field, the number of visits to the center of the field, and the total distance traveled by the mice during the testing perion. The Mann-Whitney U test was performed, and the asterisk indicates statistical significance (p < 0.05). Other tests performed for which results did not reach statistical significance are provided in Supplementary Fig. 5.
protein expression data, given the challenges of obtaining such high-resolution information for proteins.

Discussion
Mutations in the CRADD gene have been linked with lissencephaly, a condition in which defective neuronal migration leads to cortical underdevelopment and reduced or absent gyri, resulting in ID and/or other neurodevelopmental disorders 2,3 . Examination of MRI from two affected individuals with the PIDD1 Gln863* mutation did show subtle signs of lissencephaly in one (the elder) of the two individuals; however, the images available to us were poor quality. Brain MRI from the Manipal-1 proband with the Arg637* mutation also indicated lissencephaly, more specifically pachygyria, as well as shortened/thickened corpus callosum. This adds weight to lissencephaly being a common feature for both CRADD and PIDD1-related disorders. Both CRADD and PIDD1 proteins form part of the complex known as the PIDDosome-a regulator of cell death in response to DNA damage 7 . Of note, PIDD1 and CRADD have been reported to activate caspase-2 in various cellular contexts, including DNA damage or centrosome amplification in cancer cells as well as amyloid-induced neuronal toxicity 8,11 . Remarkably, constitutive post-translational selfprocessing of PIDD1 gives rise to PIDD1-N (~48 kDa) and PIDD1-C (~51 kDa), which is further processed into PIDD1-CC (~37 kDa), needed for caspase-2 activation within the PIDDosome 29 . PIDD1-N contains a leucinerich domain while PIDD1-C contains a DD from residues 774-893. Serine residues at S446 (generating PIDD1-C) and S588 (generating PIDD1-CC) are involved in this autoprocessing mechanism 29 . In our study we have shown that none of the genetic alterations reported here prevent self-processing of PIDD1, but differences in stability become apparent in PIDD1-CC variants with a truncated DD. This suggests that the effect of the PIDD1 mutations that have been associated with ID 19,21 are most likely due to the loss of function of PIDD1-CC, as PIDD1-C can be still be generated at rates comparable to WT (Fig. 4). Caspase-2 recruitment and activation in the PIDDosome is critically dependent on PIDD1-CRADD interaction, as a complete loss of caspase-2 activation was observed in the several CRADD and PIDD1 mutants 32 . Here, either directly using Co-IP or indirectly using MDM2 cleavage as the readout, we confirm the loss of PIDD1-CRADD interaction due to PIDD1 DD mutations R815W and Q863*, while the G876S variant (found only in heterozygous form in a single unrelated affected individual) has no significant impact. Despite showing relatively poor amino acid residue conservation across vertebrate families at R815 (see Supplementary Fig. 3), the R815W DD substitution shows this effect to a similar degree compared with the Q863*. Consistently, these mutants were no longer able to process MDM2 effectively. Currently, we have not documented that mutant PIDD1-C is also functionally impaired. However, as NF-κB activation by PIDD1-C also depends on DD-DD interaction within RIP1 and mutant PIDD1 fails to interact with CRADD, it is fair to speculate that this function of PIDD1-C would also be compromised to some degree.
Degradation assays using cycloheximide suggest that there is negligible effect of the PIDD1 mutations on protein stability or longevity. This supports our conclusions that the effect of these mutations is through loss of DD function. However, it is quite possible that other mutations in PIDD1 will impact the protein stability, as has been observed for the CRADD Gly128Arg mutation 2 .
Although the phenotypic consequences of Pidd1 knockout in mice did not recapitulate the cognitive deficit seen in the patients with biallelic disruption of the PIDD1 C-terminal DD, there were mild but significant anxiety phenotypes noted. A recent meta genome-wide association study across several anxiety-related neurodevelopmental disorders (autism spectrum disorder, attention deficit/hyperactivity disorder (ADHD), obsessive compulsive disorder, and Tourette Syndrome) including over 93,000 study participants showed genome-wide significant association at PIDD1 33 . Further, it has been reported that a risk allele for ADHD at PIDD1 (rs7104929, allele G) correlates with significantly increased methylation at promoter and enhancer epigenetic marks and decreased transcription of the gene 34 . Recent whole-genome sequencing efforts for autism spectrum disorder proband/ mother/father trio through the Autism Speaks/MSSNG study also has some evidence for PIDD1 as a rare risk geneout of~10,000 trios sequenced, a single affected individual (AU2798301) is reported to show compound heterozygous variants: maternal allele: chr11:802296G>C; NM_145887: exon6:c.C1075G; p.(Arg359Gly); paternal allele: chr11: 802684G>A, NM_145887:exon4:c.C917T:p.Pro306Leu. The respective allele frequencies in the gnomAD control database are 8E−6 and 4E−6, with no homozygotes observed. Predictions for these missense variants are given as "high" and "medium" damaging, respectively (https://research.mss.ng). It is also worth noting that copy number variants across PIDD1 reported in the DECIPHER database (compiled in ref. 34 ) include 10 duplications and 1 triplication, 5 of which are de novo, in individuals with ID, global developmental delay, autism, delayed speech and language development, and macrocephaly with postnatal growth retardation. Thus, there is accumulating evidence of a role for PIDD1 in neurodevelopmental disorders spanning multiple diagnoses, and with a variety of inheritance patterns.
It is worth noting that, for Cradd−/− mice, Di Donato et al. 2 reported no significant CNS phenotype, and megaloencephaly but no cortical lamination defect, contrasting with the lissencephaly and ID reported in CRADD mutation patients. Brain anatomical evidence for the human PIDD1 mutations comes from both PIDD1 families for which MRI brain images were available, and Pidd1−/− mouse brain anatomical analysis has not been performed yet, but it is also clear that, as with CRADD/cradd, the ablation of mouse Pidd1 does not appear to cause the degree of cognitive deficit seen in humans with disruption of the Cterminal PIDD1 DD and PIDD-CC function. It should be noted that the fetal mouse brain cell-type RNAseq analyses suggest that Pidd1 expression in neurons is very restricted, and the degree of transcriptional overlap between Pidd1, Cradd, and Caspase-2 (Casp2) in CNS cells is very low. This could explain the lack of CNS-related phenotypes for the Cradd and Pidd1 mice, but could also limit any study of PIDD1 mutations in neuronal cells. We posit that there may be higher overlapping expression of these genes in developing human CNS, which may be relevant to the etiopathology of PIDD1 mutations. The anatomical RNAseq and microarray data from the human BrainSpan study 23 and PsychEncode RNAseq study 24 certainly suggest that coexpression of PIDD1, CRADD, and CASP2 is high in many brain regions in the developing human brain (see Supplementary Fig. 7). However, as with mice 28 , expression of PIDD1 in specific cell types in human fetal cerebrum and cerebellum is low and very sparse, but co-expression data and data for other brain regions are not yet available 35 .
Whether the etiological pathway for PIDD1 and CRADD mutations in neurodevelopmental disorders is united through the PIDDosome is not yet fully established; however, as the limited clinical brain anatomical data available for our PIDD1 mutation families appears to echo those reported for CRADD mutations, this is certainly something that needs to be further explored. Future work will focus on whether Pidd1 mice also share such features, regardless of the lack of a clear cognitive deficit. However, the phenotypic commonalities observed in humans thus far suggest that further studies into other PIDDosome pathway components as disease or risk genes for neurodevelopmental disorders are clearly warranted.