MYT1L haploinsufficiency in human neurons and mice causes autism-associated phenotypes that can be reversed by genetic and pharmacologic intervention

MYT1L is an autism spectrum disorder (ASD)-associated transcription factor that is expressed in virtually all neurons throughout life. How MYT1L mutations cause neurological phenotypes and whether they can be targeted remains enigmatic. Here, we examine the effects of MYT1L deficiency in human neurons and mice. Mutant mice exhibit neurodevelopmental delays with thinner cortices, behavioural phenotypes, and gene expression changes that resemble those of ASD patients. MYT1L target genes, including WNT and NOTCH, are activated upon MYT1L depletion and their chemical inhibition can rescue delayed neurogenesis in vitro. MYT1L deficiency also causes upregulation of the main cardiac sodium channel, SCN5A, and neuronal hyperactivity, which could be restored by shRNA-mediated knockdown of SCN5A or MYT1L overexpression in postmitotic neurons. Acute application of the sodium channel blocker, lamotrigine, also rescued electrophysiological defects in vitro and behaviour phenotypes in vivo. Hence, MYT1L mutation causes both developmental and postmitotic neurological defects. However, acute intervention can normalise resulting electrophysiological and behavioural phenotypes in adulthood.


INTRODUCTION
Autism spectrum disorder (ASD) is a common neurodevelopmental disorder (NDD) characterised by behavioural changes, including altered social patterns [1].ASD is often associated with coexisting conditions, including epilepsy, intellectual disability, and hyperactivity.Gene mutations affecting neuronal communication confer increased risk for ASD and offer possible therapeutic targets [2].However, the genetic heterogeneity of ASD is enormous, and multiple transcriptional regulators have recently been associated with this group of disorders.Indeed, mouse models show that mutations of chromatin remodelers, such as Chd8 and BAF complex members like Smarcc2, can induce behavioural phenotypes [3][4][5][6], suggesting a potential causal role in ASD.Yet, their contribution to disease and their clinical relevance often remains elusive [7], which limits the development of targeted treatments for gene regulator-associated mental disorders at the time of diagnosis [8,9].
Of the 91 chromatin or gene regulators most strongly associated with ASD (category 1; [10]), MYT1L is specifically and continuously expressed in virtually all neurons throughout life [10][11][12].MYT1L is a conserved zinc finger transcription factor and mutations have been reported in patients diagnosed with intellectual disability, schizophrenia, epilepsy, and ASD [13][14][15][16][17][18], suggesting that MYT1L-mediated gene regulation may be important in preventing NDDs including ASD.Indeed, 98% (50 out of 51) of currently reported cases with heterozygous MYT1L deletion or loss-of-function mutations were diagnosed with ASD and/or intellectual disability [19].Besides behavioural features, several patients with MYT1L mutations also display developmental delays, obesity, seizures, and brain malformations [19].MYT1L was one of the three original factors able to directly reprogram fibroblasts into functional neurons upon overexpression [20].MYT1L is a transcriptional repressor that can enhance neuronal identity in vitro by actively repressing several developmental pathways, including WNT and NOTCH.This is achieved in part through recruitment of epigenetic silencers such as SIN3/HDAC [21][22][23].Unexpectedly, reprogramming experiments have also revealed that MYT1L binds and represses several non-neuronal gene programs, such as muscle and fibroblast genes, suggesting a role as pan-neuronal safeguard that silences other lineage-specific genes [21,24,25].Indeed, recent studies described that Myt1l haploinsufficiency, caused by frameshift mutation of Myt1l exon 15 or deletion of exon 9, induced altered brain development and behaviour phenotypes in mice [26,27].However, a number of important questions remain unanswered.For example, it is unclear whether distinct MYT1L mutations cause overlapping phenotypes, as suggested based on patient reports.Further, no studies present the effects of MYT1L depletion in human neurons.Finally, the molecular mechanisms causing the neurological phenotypes, and whether they are amenable to intervention, are unknown.
Here, we used genetically-engineered mice and human induced neurons to investigate the role of MYT1L during development and as a new preclinical platform to study MYT1L-associated NDDs.Using these translational models, we show that MYT1L deficiency is sufficient to induce autism-associated phenotypes, ranging from deregulation of gene expression and delayed neurogenesis to cortical thinning and altered behaviour.MYT1L target genes, including WNT and NOTCH regulators, were activated early upon MYT1L depletion in human neurons and chemical pathway inhibition could rescue associated differentiation defects.We found that MYT1L loss also caused upregulation of non-neuronal genes, including the cardiac sodium channel SCN5A, and triggered an unexpected neuronal network hyperactivity that can be rescued by MYT1L overexpression or SCN5A knockdown.Furthermore, acute application of the approved drug lamotrigine rescued the electrophysiological phenotypes in postmitotic mouse and human neurons, as well as behaviour phenotypes in adult mice, providing a potential therapeutic avenue for patients with MYT1L syndrome.

MATERIALS AND METHODS Statistical analysis
No statistical methods were used to predetermine sample size.All values including replicates and statistical analysis are available in the Supplementary Tables S2-8, 13.Cells for in vitro analysis were randomly selected and mice were randomly allocated to treatment groups.The investigators were blinded for patch clamp recordings, other experiments were not randomised.

Experimental design
To model MYT1L loss we generated a germline mutation in mice harbouring a 7 bp frameshift in Myt1l exon 6.In addition, we engineered a conditional heterozygous MYT1L allele in human ESCs, allowing cremediated removal of exon 7 in human neurons.We observed the expected reduction of full-length MYT1L protein in human and mouse neurons.In mutant mice we found a truncated MYT1L protein isoform that lacked nuclear localisation and function, confirming the validity of our loss-offunction models.

MYT1L immunoprecipitation and mass spectrometry
One hemisphere of P0 mice was flash-frozen in liquid nitrogen and used for each MYT1L immunoprecipitation as described previously [21].Bound proteins were enzymatically digested and peptides were analysed by mass spectrometric analysis (Fusion Orbitrap; Thermo Fisher Scientific).

Primary tissue dissection and cell culture
Mouse glial cells were isolated from forebrains of wild-type CD1 (Jackson Laboratories) at P0 as described here [21].For primary neuronal cultures, the hippocampus or the cortex of P0 Myt1l-mutant or control mouse pups were isolated as published previously [28].For overexpression or knockdown, the cells were transduced with lentivirus encoding indicated constructs on day in vitro 3 (DIV3).

EdU labelling and immunostaining of mouse brain sections
Timed-pregnant females were intraperitoneally injected at E14.5 with EdU (30 mg/ kg body weight) (Life Technologies).After 20 h embryonic brains were harvested.Q-fraction analysis followed previously established practices [6].The number of TBR2+ and SOX2+ cells was determined in midcortical 265 µm-wide segments using an in-house developed Fiji macro.All quantifications were carried out at equivalent anteroposterior positions between genotypes.

Morphology analysis
For analysis of cortical thickness, P0 brain sections were aligned across genotypes using subcortical anatomical landmarks for orientation, and thickness was measured at a 45°angle from the dorsal midline.Morphology analysis of cultured neurons was performed with Fiji using SNT plugin [30,31].Nuclear FLAG-MYT1L localisation was performed in primary mouse neurons at DIV 11 upon overexpression of indicated constructs at DIV 3.

Behaviour experiments
The mice were housed in a temperature-controlled vivarium maintained on a 12-h light-dark cycle and tests were conducted during the light cycle.Procedures were performed at the Interdisciplinary Neurobehavioral Core of Heidelberg University and approved by the Regierungspräsidium in Karlsruhe (G-287/20, G-105/16).
Multi-electrode array (MEA) and drug treatment MEA measurements were performed using the Maestro Pro multiwell device with the Axis Navigator software and 48-well plates (all from Axion Biosystems).For acute treatments, 10 µM lamotrigine (Targetmol) was added per well and the plate was measured before and 2 h after the treatment.Data was analysed using meaRtools [45].

MYT1L represses non-neuronal gene expression and is essential for survival in mice
To address whether MYT1L mutations can cause complex neurodevelopmental defects, we generated a mouse model with a 7 bp deletion in exon 6 of Myt1l (Supplementary Fig. S1A-F).This frameshift mutation induces a premature STOP codon at amino acid (aa) 78, similar to a patient with a nonsense mutation at aa 75 in human MYT1L [15].Resulting offspring exhibited expected reduction in full-length MYT1L protein compared to (+/+) controls (Fig. 1A and Supplementary Fig. S1G, H).However, we detected a truncated MYT1L isoform, most likely expressed from an internal methionine at aa 99 (Supplementary Fig. S1G, I) (Supplementary Table S1).This isoform failed to translocate to the nucleus due to lack of the nuclear localisation signal and therefore is expected to be non-functional (Supplementary Fig. S1J, K).Myt1l (+/−) mice are viable and fertile, but homozygous Myt1l (−/−) deletion resulted in postnatal lethality (Fig. 1B).This shows that MYT1L is essential for survival and validated our loss-of-function model.
To investigate changes in cell composition and gene expression, we performed single cell RNA sequencing (scRNA-Seq) of the prefrontal cortex at birth (Fig. 1C-E and Supplementary Fig. S2) (Supplementary Table S2).We did not observe loss or emergence of new cell clusters, but we identified Myt1l mutantspecific subpopulations by calculating the likelihood of each cell to be in a neighbourhood enriched in either control or mutant cells using MELD [40] (Fig. 1C and Supplementary Fig. S2C, D).Only minor changes in cell type ratios occurred between mutant and control mice (Fig. 1D and Supplementary Fig. S2E), but we found fewer Tbr2-and Sema3c-cell populations, which mark newly-formed migrating neurons in the subventricular zone (SVZ).Additionally, Cdca7+ interneurons were reduced in (−/−) mice, whereas cortical layer I neurons (Reln+ ) were increased.Analysis of the differentially expressed genes showed that MYT1L depletion resulted in global gene upregulation that was more pronounced in homozygous mutants (Fig. 1E).We observed increased expression of many non-neuronal gene programs upon MYT1L depletion, even in cortical layer neurons (Supplementary Fig. S2F).To identify MYT1L target genes, we performed CUT&RUN [43] in wildtype mice at embryonic (E) day E18.5, postnatal (P) day 0 (P0) and 3-months (adult) (Fig. 1F, G and Supplementary Fig. S2G, H) (Supplementary Table S3).We found a large overlap of bound sites and enrichment of the MYT1L DNAbinding motif (AAAGTT) at different developmental stages (Fig 1H and Supplementary Fig. S2I, J).Intersecting MYT1L target genes with deregulated genes at P0 revealed significant upregulation within different neuronal populations, such as Satb2 or Fezf2 cortical cell populations (Fig. 1I and Supplementary Fig. S2K).This underscores the role of MYT1L as a transcriptional repressor and suggests that, despite induction of neuronal identity, Myt1lmutant cells failed to silence inappropriate gene expression programs causing a confused transcriptional identity in neurons in vivo.
Impaired neurogenesis in Myt1l-mutant mice MYT1L can enhance neuronal identity in vitro by repressing negative regulators of neurogenesis [21], such as Notch and Wnt, which upon dysregulation could affect neuronal differentiation dynamics and brain structure.Therefore, we performed an EdU pulse-chase alongside co-staining for the proliferation marker, Ki67, to determine the quit (Q) fraction of cortical cells exiting the cell cycle over a 20 h period.Indeed, Ki67−EdU+ cells were significantly decreased, corresponding to a decrease in Q fraction of ~37% in Myt1l (-/-) embryos at E15.5 (Fig. 2A).SOX2+ neural stem cells and TBR2+ neural progenitors were unchanged at E15.5 (Supplementary Fig. 3A).At birth, SOX2+ stem cells were increased in Myt1l (+/−) and (−/−) mice (Fig. 2B), while TBR2+ progenitors did not change.This mimics overexpression of the NOTCH effector Hes1 that increases the stem cell pool after birth and results in a thinner cortex [46].Hence, we studied the brain anatomy of Myt1l-mutant mice and found that overall brain size was unchanged at birth while brain weight was lower, resulting in an increased length to weight ratio (Fig. 2C and Supplementary Fig. S3B, C).Furthermore, the cortices were ~10% (+/−) or ~15% (−/−) thinner than controls, respectively (Fig. 2C and Supplementary Fig. S3C).These results suggest that failure to repress anti-neurogenic programs impaired neurogenesis, which might in turn cause structural brain abnormalities and might explain brain malformations found in some patients [19].

Gene expression changes in mutant mice resemble those of ASD patients
Next, we examined the effects of MYT1L deficiency at the molecular level across brain development using bulk RNA-Seq of mouse cortices at E18.5 (coinciding with peak Myt1l expression in the mouse brain (Supplementary Fig. S1A [11])), P0 (birth), P22 (juvenile), and 3-months (adult).We found several hundred genes differentially expressed, with distinct clusters of deregulated genes early (E18.5 and P0) and late (P22 and adult) in development (Fig. 3A and Supplementary Fig. S4A) (Supplementary Table S4).Many of the deregulated genes did not overlap with previously-annotated MYT1L targets [21] or the MYT1L CUT&RUN targets (Fig. 1F and Supplementary Fig. S2G), implying that many changes in bulk samples are indirect.In heterozygous brain samples, MYT1L target genes were both up and downregulated, nevertheless ~60% of direct MYT1L target genes were upregulated upon homozygous Myt1l mutation (Supplementary Fig. S4B).This is in line with the upregulation of MYT1L target genes in (+/−) and (−/−) cortical neurons based on single cell analysis (Fig. 1E), suggesting that MYT1L predominantly acts as a repressor during neuronal development.In accordance with the decreased Q fraction at E15.5, gene expression at E18.5 revealed downregulation of neurogenesis-associated GO terms and upregulation of cell division (Fig. 3B).Furthermore, we observed upregulation of early-foetal gene expression signatures and concomitant decrease of late-foetal signatures at E18.5 [47] (Fig. 3C and Supplementary Fig. S4D).This suggests that MYT1L deficiency, as observed in other mouse models of ASD [5], causes delays in brain development.We observed continued downregulation of neuronal GO terms and an increase in signalling and non-neuronal terms upon Myt1l mutation in adult mice (Fig. 3B and Supplementary Fig. S4C).Gene set enrichment analysis (GSEA) highlighted several non-neuronal cell fate signatures among the upregulated genes, and indicated an overall enrichment of genes implicated in epilepsy, schizophrenia, and ASD among the deregulated genes in Myt1l-mutant brains [48,49] (Supplementary Fig. S4D, E).Interestingly, mice harbouring our Myt1l exon 6 mutation displayed gene up-and downregulation that overlapped with Chd8 haploinsufficient mice [6], but only partly with recently-published exon 15 [26] and exon 9 [27] Myt1l-mutant mice (Supplementary Fig. S4F).Finally, we discovered that genes upregulated in ASD patients are also upregulated in our MYT1Ldeficient mice, while genes downregulated in Myt1l-mutant mice are also downregulated in ASD patients [5,50] (Fig. 3D and Supplementary Fig. S4G).Overall, our data show that MYT1L deficiency in mice leads to delays in neurodevelopment and deregulation of gene expression as observed in ASD patients.
Behavioural hyperactivity and male-specific social deficits upon Myt1l mutation in mice Next, we performed behaviour analyses using our Myt1l mouse model.We observed a trend towards increased vocalisation in early postnatal MYT1L-deficient mice, but the number and type of analysed calls were not significantly changed (Supplementary Fig. S5A).In line with hyperactivity phenotypes reported in several MYT1L patients [13-16, 51, 52], we observed increased locomotion in adult mice in their home cages, as well as in juvenile mice in open field and elevated plus-maze tests (Fig. 3E and Supplementary Fig. S5B-D).Mutant mice spent ~70% more time in centre regions and open arms in the latter two tests, respectively.In addition, the explorative rearing behaviour was significantly increased in mutant mice (Supplementary Fig. S5D).This indicates that MYT1L-deficient mice are less anxious, a phenotype reported in other ASD mouse models [53,54].We also observed a significant decrease in marble burying and grooming events, which are used to evaluate repetitive behaviours but also represent decreased anxiety-like behaviours (Supplementary Fig. S5D, F).Finally, social chamber experiments were performed to investigate the social behaviour of our Myt1l-mutant mice.All animals displayed the expected preference for investigating chambers with a familiar littermate as opposed to empty chambers.However, while both wildtype and female Myt1lmutant mice spent more time exploring the chamber with a newly-added unfamiliar mouse, MYT1L-deficient males did not exhibit this expected preference for social novelty (Fig. 3F and Supplementary Fig. S5E).In summary, Myt1l depletion causes several striking behavioural changes in tests used to model ASDlinked behaviour including hyperactivity that was found in male and female mice and male-specific social deficits.

Delayed neurogenesis and non-neuronal gene expression in MYT1L-deficient human neurons
To dissect the role of MYT1L in human neurons we engineered human embryonic stem cells (hESCs) with a heterozygous conditional knockout allele of MYT1L exon 7 (+/fl) (Fig. 4A and Supplementary Fig. S6).MYT1L (+/fl) hESCs were directed towards electrically mature human neurons using NGN2 transcription factor-mediated differentiation [29].This allows to study developmental and synaptic processes linked to neurological disorders in a conditional manner [55][56][57].Cre-mediated MYT1L (+/−) deletion resulted in ~50% protein reduction, mimicking MYT1L haploinsufficiency found in patients (Fig. 4B, Supplementary Fig S6G).First, we assessed the transcriptional effects of MYT1L depletion compared to Δcre-transduced isogenic controls at early and late maturation timepoints (Supplementary Table S5).MYT1L-mutant neurons exhibited a decrease in neuronal GO terms and an enrichment of genes implicated in epilepsy, schizophrenia, and ASD among the deregulated genes [48,49] (Fig. 4C and Supplementary Fig. S7A).GO terms associated with differentially upregulated genes strongly overlapped between our mouse and human models (Supplementary Fig. 7B) (Supplementary Table S6), indicating conserved effects.Overall, MYT1L depletion resulted in stronger gene activation early (week 1) and late (week 6) after neuronal induction (Fig. 4D and Supplementary Fig. S7C).The MYT1L DNA-binding motif (AAAGTT) was enriched at genes upregulated upon MYT1L depletion (Fig. 4E and Supplementary Fig. S7D, E).Several transcription factors regulating non-neuronal cell fates, such as mesoderm-specific PRRX1 or glia switch regulator SOX9 [58,59], harbour MYT1L motifs and were upregulated in MYT1L-deficient neurons (Fig. 4F).GSEA revealed upregulation of non-neuronal gene expression programs, such as muscle cell fate, upon MYT1L depletion (Supplementary Fig. S7F).Ingenuity Pathway Analysis (IPA) confirmed that inappropriate developmental programs, such as myogenesis and cardiogenesis, were activated following MYT1L mutation, suggesting that they are continuously repressed by MYT1L (Fig. 4G and Supplementary Fig. S7G).On the other hand, neuronal transcription factors such as VAX2 or EN1 that do not contain MYT1L motifs in their promoter showed decreased expression after one week, but compared to controls exhibited an unexpected increased expression at week six (Fig. 4F).This expression pattern was mirrored in pathways involved in neuronal differentiation and function, including synaptogenesis and calcium signalling, that were initially repressed but later increased in MYT1L-deficient neurons (Fig. 4G and Supplementary Fig. S7G).Importantly, our findings regarding neuronal gene deregulation were recapitulated in neurons derived from an additional independently-engineered MYT1L (+/fl) hESC clone (Supplementary Fig. S6 and S7).Surprisingly, the delay in neuronal gene expression did not affect neuron complexity after ten days or six weeks of induced neurogenesis (Supplementary Fig. S8).During neuronal reprogramming of mouse fibroblasts MYT1L promoted neuronal fate by repressing anti-neurogenic genes such as Id3 as well as WNT and NOTCH pathway members [21].Here, we observed increased expression of these genes in MYT1L-deficient human neurons during neuronal induction (Fig. 4F).Combined application of chemical inhibitors targeting the WNT and NOTCH pathways restored TUJ1 protein to control levels and partially normalised gene expression changes in MYT1L mutants based on RNA-Seq measurements, particularly by decreasing expression of genes that are upregulated upon MYT1L depletion (Fig. 4H and Supplementary Fig. S9A,  B) (Supplementary Table S7).Moreover, we identified several proneuronal transcription factors that normally peak early during induced neurogenesis and that were downregulated in MYT1Lmutant neurons at week one, but upregulated at week six (Supplementary Fig. S9C).Expression of these factors could be normalised in MYT1L-deficient neurons by WNT and NOTCH inhibition at week one (Supplementary Fig. S9D).Overall, this indicates that heterozygous deletion of MYT1L during human induced neurogenesis causes deregulation of autism-associated genes and activation of non-neuronal target genes that result in neurogenesis delays at least in part mediated by increased WNT and NOTCH signalling.

Synaptic and network hyperactivity in MYT1L-mutant human and mouse neurons
To investigate whether MYT1L haploinsufficiency affects neuronal function in human cells, we studied the electrophysiological Fig. 1 MYT1L is crucial for survival and repression of genes in mice.A MYT1L full-length protein levels in brain lysates from mice harbouring a CRISPR-induced 7 bp frameshift germline mutation in exon 6 (first protein coding exon) of Myt1l after birth normalised to control.Representative Western blot images using indicated antibodies are shown; n ≥ 6. B Survival pie chart from (+/−) x (+/−) offspring at birth (8 litters P0; n = 12 (+/+, black), 31 (+/−, teal), 15 (−/−, yellow)) and after weaning (7 litters P21; n = 15 (+/+), 17 (+/−), 0 (−/−)) indicated that homozygous Myt1l mutants die after birth.C Single cell RNA-Seq of the prefrontal cortex at birth and likelihood of observing MYT1L-deficient (+/−) and control (+/+) cell enrichment in populations of specific cell types using MELD.D Ratio of MYT1L-deficient (+/−) and control (+/+) cells in indicated cell types annotated based on reference data from [77] properties of MYT1L (+/−) neurons at the population level following six weeks of maturation using multi-electrode array (MEA) [60,61].Unexpectedly, we observed that spike firing doubled, and simultaneous network firing activity tripled, in human MYT1L-mutant neurons compared to controls (Fig. 5A).The altered electrophysiological activity directly correlated with MYT1L protein depletion, occurred as early as three weeks after neuronal induction and was maintained until the end of a long-term experiment at week 15 (Supplementary Fig. S10A-C).The electrophysiological phenotype could be recapitulated in an independently-engineered MYT1L (+/fl) hESC clone (Supplementary Fig. S10D).Next, we tested whether these functional changes were conserved in primary mouse neurons.To that end, we investigated electrophysiological properties of cultured hippocampal neurons derived from Myt1l-mutant and control mice using MEA.Recapitulating our findings in human MYT1L (+/−) neurons, we observed increased firing and network hyperactivity.These effects scaled with MYT1L depletion, with an average spike firing increase of ~700% (−/−) and ~400% (+/−) compared to control (Fig. 5B and Supplementary Fig. S11A).Of note, overexpression of MYT1L 99-1187 in wildtype neurons did not alter electrophysiological activity on MEA, ruling out a potential dominant negative effect of the truncated MYT1L isoform further confirming that our Myt1l-mutant mice are a loss-of-function model (Supplementary Fig. S11B).Increased MEA firing activity could also be observed in primary neurons derived from cortices of MYT1L-deficient mice at birth, indicating that defects are not restricted to specific brain regions, but affected at least two areas implicated in ASD [62] (Supplementary Fig. S11C).To correlate the observed network hyperactivity with synaptic activity of single cells, we performed patch clamp recordings.In line with the observed network hyperactivity, the amplitude and frequency of spontaneous excitatory postsynaptic currents (EPSCs) and of miniature EPSCs (mEPSCs; recorded in the presence of tetrodotoxin (TTX)) was increased in MYT1L-deficient cultured primary mouse neurons (Supplementary Fig. S11D, E), respectively.Like in the human model, we did not observe significant changes in neuronal morphology of cultured MYT1L-deficient primary  3 Autism-associated changes in gene expression and behavioural phenotypes upon Myt1l mutation in mice.A Heatmap of deregulated genes upon Myt1l mutation across all replicates and stages, scaled by gene (rows).B Selected top gene ontology (GO) terms and p-values of genes that were down-(blue) or upregulated (red) upon Myt1l mutation in the cortex of mice compared to control at indicated timepoints during development.C GSEA plots show depletion of mid-foetal (blue) and enrichment of early-foetal (red) neural development-related gene sets among genes deregulated in Myt1l-mutant brains at E18.5.NES, normalised enrichment score; FDR, false discovery rate.D Overlap of genes up-or downregulated upon Myt1l mutation in mouse cortex at E18.5 and P22 and genes that are up-or downregulated in ASD patient brains determined by GeneOverlap.For RNA-Seq analysis n ≥ 5 E18.5, n ≥ 4 P0 for (+/+, black), (+/−, teal) and (−/−, yellow), respectively; n = 6 P22, n ≥ 3 adult for (+/+) and (+/−), respectively.E Open field test showed that Myt1l (+/−) mutants covered more distance and spent more time in centre regions compared to control animals at P23; n = 25 for (+/+) and n = 29 for (+/−) from three independent cohorts.F Social chamber experiments showed male-specific decrease of time Myt1l (+/−) mutants spent to explore a novel mouse compared to control mice at the age of one month; n = 10 for (+/+) and (+/−) from three independent cohorts.Bar graphs show mean values, data points from individual animals are displayed, error bars = SEM, Mann-Whitney test **p < 0.01, ***p < 0.001, ****p < 0.0001, NS not significant.neurons (Supplementary Fig. S11F).We also performed electrophysiological brain slice recordings from pyramidal neurons located in the hippocampal CA1 region of one-month-old mice.Consistent with our in vitro data, we observed an increased spontaneous EPSC frequency in Myt1l mutants compared to controls (Fig. 5C).Intrinsic properties, such as resting membrane potential and evoked action potential features, appeared unchanged (Supplementary Fig. S12A).The frequency of spontaneous inhibitory postsynaptic currents (sIPSCs, recorded in the presence of CNQX and D-APV) was increased in MYT1L-deficient pyramidal neurons, indicating that the increased excitation did not result from decreased inhibition (Supplementary Fig. S12B).Our experiments show that continuous MYT1L depletion impairs neuronal function and results in an unexpected electrophysiological hyperactivity phenotype in human and mouse neurons.

Genetic and pharmacologic rescue of neuronal network activity in vitro and behaviour phenotypes in vivo
We observed similar electrophysiological network hyperactivity phenotypes based on MEA measurements in primary mouse and stem cell-derived human induced neurons upon MYT1L mutation, indicating a conserved underlying molecular mechanism.Since activation of non-neuronal target genes upon MYT1L mutation was observed in human and mouse models, this might contribute to the overlapping phenotypes.To identify such targets, we analysed genes that: (i) were bound by MYT1L in vivo as determined by CUT&RUN (Fig. 1F and Supplementary Fig. S2G) (Supplementary Table S3); (ii) were deregulated in primary mouse or human induced neurons upon heterozygous MYT1L deletion determined by RNA-Seq (Supplementary Table S5 and 8); (iii) contributed to the upregulation of non-neuronal terms based on IPA (Fig. 4 and Supplementary Fig. S4 and 7); and (iv) were lowly expressed in the brain compared to other tissues, based on public expression datasets [63].Strikingly, 78% (30 out of 38) of these non-neuronal targets were upregulated upon MYT1L depletion in mouse and human neurons (Fig. 6A and Supplementary Fig. S13A).The top upregulated target is TFAP2B, which, upon mutation, can cause Char syndrome, characterized by facial, ductal, and hand anomalies [64].Interestingly, some patients with MYT1L mutations exhibit mild facial dysmorphisms [19,65].In addition, we found an almost two-fold increase of the cardiac voltage-gated sodium channel SCN5A, which upon mutation can cause long QT syndrome [66], among the top 10 upregulated target genes.SCN5A displayed a low basal expression in human and mouse neurons, and increased expression in MYT1L-deficient neurons could alter electrophysiological properties.MYT1L remains expressed in mature neurons, suggesting that it also functions in postmitotic neurons (Supplementary Fig. S1A).We therefore asked whether Myt1l overexpression (OE) in postmitotic MYT1Ldeficient neurons could restore repression of target genes, such as Scn5a.Myt1l OE significantly increased the expression of the neuronal marker Tuj1 compared to GFP OE controls (Fig. 6B and Supplementary Fig. S13B, C).Importantly, Myt1l OE decreased the expression of target genes that were upregulated in MYT1L-mutant neurons, specifically Hes1 and Scn5a (Fig. 6C).Strikingly, we found that Myt1l OE also restored neuronal network activity to control levels (Fig. 6D).These results suggest that the continuous deregulation of gene expression in postmitotic neurons caused by MYT1L mutation contributes to electrophysiological hyperactivity phenotypes, and that these could be restored even after neurodevelopment is complete.To test whether SCN5A upregulation caused the electrophysiological hyperactivity phenotype, we performed shRNA-mediated SCN5A knockdown in both mouse primary and human induced neurons, which reduced SCN5A levels by 80%, and measured neuronal network activity using MEA (Supplementary Fig. S13D).Strikingly, Scn5a knockdown reduced neuronal network hyperactivity in (+/−) and (−/−) mouse primary neurons to wild type levels (Fig. 6E).This could be recapitulated by SCN5A knockdown in MYT1L (+/−) human induced neurons (Fig. 6G), indicating that SCN5A upregulation contributes at least in part to the electrophysiological network hyperactivity in mouse and human MYT1L-mutant neurons.In order to find a potential therapeutic intervention, we decided to test whether lamotrigine, a sodium channel blocker and FDAapproved antiepileptic drug [67,68], could rescue MYT1L deficiency-induced phenotypes.Indeed, acute application of lamotrigine normalised electrophysiological network hyperactivity of both human and mouse MYT1L-deficient neurons towards control levels (Fig. 6F, H).Finally, to study if neuronal network hyperactivity contributes to the behaviour phenotypes observed upon Myt1l mutation and to test whether these are also amenable to pharmacological intervention in vivo, we injected mice with  -/-+/+ +/--/  lamotrigine.We found that acute lamotrigine treatment in Myt1lmutant mice normalised anxiety and hyperactivity behaviour in the open field test, as well as locomotion and rearing in home cage observations, towards untreated control mice (Fig. 6I and Supplementary Fig. S13E).Overall, these results show that, besides neurodevelopmental defects, MYT1L mutation causes upregulation of non-neuronal genes and neuronal network hyperactivity in mouse and human neurons.Acute lamotrigine treatment in postmitotic neurons and adult mice rescued both the electrophysiological network hyperactivity and the tested behaviour phenotypes (Fig. 6J), indicating that electrophysiological effects contribute, at least in part, to altered behaviour in mice.Therefore, MYT1L mutation not only impacts development, but also affects neuronal function and behaviour in adulthood, suggesting that targeted treatments may also benefit patients at later stages of development.

DISCUSSION
Unlike most neuropsychiatric disease-associated chromatin regulators, the transcription factor MYT1L is specifically expressed in virtually all neurons and remains expressed throughout life [11,12].However, whether and how MYT1L mutations can cause neuropsychiatric disease remains poorly understood.In this study, we present evidence that MYT1L haploinsufficiency is sufficient to cause phenotypes associated with NDDs and ASD in human neurons as well as in mouse models due to failure to repress target genes resulting in (i) delays during developmental neurogenesis and (ii) synaptic malfunction in mature neurons.MYT1L deficiency in our human and mouse models caused significant deregulation of genes associated with epilepsy, schizophrenia, and ASD, which have been diagnosed in patients carrying MYT1L mutations [13][14][15][16].We also found that gene expression changes in Myt1l-mutant mouse brains directly resembled those observed in ASD patient brains, demonstrating that MYT1L deficiency can lead to ASD-associated transcriptional profiles.Additional phenotypes linked to ASD and MYT1Lassociated NDDs include neurodevelopmental delays and changes in brain anatomy.Using transcriptomic analysis, we observed altered maturation dynamics in MYT1L-deficient human neurons upon transcription factor-mediated neuronal differentiation in vitro and impaired neurogenesis in the prefrontal cortex of Myt1l-mutant mice in vivo.Even at E18.5, Myt1l-mutant mice displayed activation of early-foetal gene expression programs and downregulation of mid-foetal programs, which has also been described in other ASD models [5].Previous studies showed that acute depletion of Myt1l or the closely related family member Myt1 by shRNA treatment impaired neurogenesis in utero by repressing Hes1 expression and Notch signalling [21,69].Here, we expand these findings by showing that germline Myt1l mutation impaired cell cycle exit of cortical progenitors during development and increased the number of SOX2+ neural stem cells in the SVZ at birth, which resembles Hes1 overexpression phenotypes [46].Indeed, combined chemical inhibition of WNT and NOTCH could restore early induction of proneuronal genes in human induced neurons.This confirms impaired neurogenesis upon MYT1L deficiency and could explain the thinned cortex observed in Myt1l-mutant mice and neurodevelopmental delays in human and mouse models.One limitation of our study is that we modelled only frameshift mutations predicted to result in premature STOP codons of both human and mouse MYT1L, similar to a reported patient with a nonsense mutation at aa 75 [15].Our mouse model presented a frameshift in exon 6 [70] that generated non-functional MYT1L protein isoforms missing essential domains such as the nuclear localisation signal, while conditional deletion of exon 7 in our human model resulted in nonsense-mediated RNA decay.Of note, both models exhibited similar gene deregulation and electrophysiology phenotypes, suggesting that loss-of-function results in overlapping defects independent of the mutation type.Notably, two recent studies describe additional Myt1l mouse models; Chen et al. generated a frameshift mutation in exon 15 [26], and Kim et al. used an exon 9 excision mutant [27].All three models have in common that homozygous Myt1l knockout results in postnatal lethality and altered brain morphology, emphasising the crucial role of MYT1L for development.In addition, all three studies found behavioural phenotypes including hyperactivity.In line with our study, Kim et al. reported decreased anxiety.Chen et al. reported male-specific impaired social behaviour, which we also found in our model although using different assays and developmental stages.Nevertheless, the three Myt1l-mutant mouse models exhibit distinct gene expression and neurodevelopmental defects.In contrast to our study, Chen et al. report an increased Q fraction and a decrease in SOX2+ neural stem cells in Myt1l-deficient mice at E14.5 and suggested MYT1L mainly acts as a transcriptional activator.However, in a recent preprint the same authors found that MYT1L bound and repressed promoters and enhancers of genes involved in early neuronal development programs [71].Here, using single cell transcriptomics, we show that MYT1L-bound genes are indeed activated in cortical neurons upon loss of MYT1L, supporting its role as a repressor.In line with this, both Chen et al. and our study found increased expression of early neurodevelopmental gene signatures later in development.Notably, all studies report a significant overlap of deregulated genes with ASD gene sets.Hence, while the differences between these three Myt1l mouse models might reflect the distinct nature of mutations, genetic mouse background, and experimental conditions, their overlap emphasises the important role of MYTL for neurodevelopment.In addition to loss-of-function variants, several de novo missense variants, indels and genomic duplications and deletions encompassing MYT1L have been reported in individuals with neuropsychiatric disease [19,65].Since MYT1L patients display diverse phenotypes, including both macro-and microcephaly, and are diagnosed with different neuropsychiatric disorders like ASD and schizophrenia, future studies will be needed to clarify how specific mutations or genetic backgrounds affect these phenotypes by engineering additional mouse and human stem cell models or by generating patient-derived neurons from induced pluripotent stem cells.
Alongside destabilization of neuronal cell identity, which manifested as gene expression changes and neurogenesis delays, we observed striking electrophysiological phenotypes in MYT1Ldeficient neurons.Unexpectedly, primary mouse and induced human neurons displayed a three to four-fold increase in neuronal network activity, underpinned by an increase in sEPSC amplitude and mEPSC frequency.Since, the frequency of sIPSCs was increased in MYT1L-deficient pyramidal neurons, increased network activity is not likely caused by decreased inhibition in our model.However, in newborn Myt1l (−/−) mice, we observed a slight increase in the fraction of cortical layer I neurons along with decreased Cdca7+ interneuron populations based on single cell analysis.Therefore, the effect of MYT1L mutations on inhibitory neurons remains elusive and requires future studies.Network phenotypes could be explained by multiple factors, such as changes in neuronal morphology or synapse density as well as general mechanisms regulating neurotransmitter release, including deregulation of voltage-gated sodium and calcium channels [72][73][74].Interestingly, the upregulation of channels that are normally not specifically expressed in neurons, such as SCN5A, could also impair normal synaptic transmission in MYT1L-mutant neurons.We therefore tested whether MYT1L overexpression, shRNA-mediated SCN5A knockdown or lamotrigine treatment, which is reported to block sodium and calcium channels [67,68], could rescue the MYT1L mutation-induced network hyperactivity.Strikingly, MYT1L overexpression decreased expression of target genes such as Scn5a and Hes1 in postmitotic MYT1L-deficient mouse neurons, and MYT1L overexpression and SCN5A knockdown both rescued electrophysiological network hyperactivity phenotypes.In addition, acute application of lamotrigine normalised not only the network hyperactivity phenotypes of postmitotic MYT1L-deficient human and mouse neurons, but also several behavioural hyperactivity phenotypes observed in mice.This suggests that specific MYT1L haploinsufficiency-associated phenotypes can be rescued by genetic intervention or smallmolecule drugs even later in development.
Overall, we present the first evidence that MYT1L mutations destabilise neuronal cell fate and function, and are sufficient to cause ASD-associated phenotypes in human and mouse models.Hence, failure to silence non-neuronal gene expression in neurons represents a novel mechanism that, at least in part, could contribute to ASD aetiology.MYT1L is a unique neurodevelopmental disease-associated transcription factor that is specifically expressed in virtually all neurons throughout life.It is therefore tempting to speculate that active lifelong repression of nonneuronal programs is an evolutionarily-conserved pathway critical for the prevention of brain disorders.Interestingly, MYT1L levels decrease during aging in both mice and humans (Supplementary Fig. S1A), and loss of neuronal cell identity has recently been suggested to play a role in Alzheimer's disease models [75,76], which suggests that neurodegeneration could also be regulated by the novel MYT1L-mediated mechanism presented here.Finally, we show that unexpected electrophysiological network hyperactivity upon MYT1L deficiency in post-mitotic mouse and human neurons, and associated behaviour phenotypes in mice, can be acutely targeted by the approved drug lamotrigine, which could provide an opportunity for therapeutic interventions.
Fig.3Autism-associated changes in gene expression and behavioural phenotypes upon Myt1l mutation in mice.A Heatmap of deregulated genes upon Myt1l mutation across all replicates and stages, scaled by gene (rows).B Selected top gene ontology (GO) terms and p-values of genes that were down-(blue) or upregulated (red) upon Myt1l mutation in the cortex of mice compared to control at indicated timepoints during development.C GSEA plots show depletion of mid-foetal (blue) and enrichment of early-foetal (red) neural development-related gene sets among genes deregulated in Myt1l-mutant brains at E18.5.NES, normalised enrichment score; FDR, false discovery rate.D Overlap of genes up-or downregulated upon Myt1l mutation in mouse cortex at E18.5 and P22 and genes that are up-or downregulated in ASD patient brains determined by GeneOverlap.For RNA-Seq analysis n ≥ 5 E18.5, n ≥ 4 P0 for (+/+, black), (+/−, teal) and (−/−, yellow), respectively; n = 6 P22, n ≥ 3 adult for (+/+) and (+/−), respectively.E Open field test showed that Myt1l (+/−) mutants covered more distance and spent more time in centre regions compared to control animals at P23; n = 25 for (+/+) and n = 29 for (+/−) from three independent cohorts.F Social chamber experiments showed male-specific decrease of time Myt1l (+/−) mutants spent to explore a novel mouse compared to control mice at the age of one month; n = 10 for (+/+) and (+/−) from three independent cohorts.Bar graphs show mean values, data points from individual animals are displayed, error bars = SEM, Mann-Whitney test **p < 0.01, ***p < 0.001, ****p < 0.0001, NS not significant.

Fig. 4
Fig.4Heterozygous MYT1L deletion delays neurogenesis in a WNT-and NOTCH-dependent manner and causes deregulation of ASDlinked genes in human neurons.A Schematic of conditional heterozygous MYT1L deletion during transcription factor-mediated induced human neurogenesis.B Western blot quantification of cells 7 days (1 week) after induction of neurogenesis normalised to control.Representative Western blot images using indicated antibodies are shown; n = 3. C Deregulated genes in MYT1L-mutant neurons 43 days (6 weeks) after induction of neurogenesis normalised to control exhibited significant overlap with genes linked to epilepsy, schizophrenia, and ASD determined by Fisher's exact test.D Volcano plot of differentially expressed genes in MYT1L-depleted human neurons after maturation for one week compared to isogenic controls.Highlighted are genes that were down-(blue) or upregulated (red) upon MYT1L depletion with absolute log2 fold change > 0.2 and p-adj < 0.1.E The MYT1L DNA binding motif AAAGTT was significantly enriched at upregulated genes in panel D. F Selected deregulated genes upon MYT1L depletion are displayed as fold change down-(blue) or upregulated (red) compared to isogenic control and number of MYT1L motifs in respective promoters is shown.B-E displays data for representative clone 1. G Ingenuity Pathway Analysis (IPA) of differentially-expressed genes in MYT1L-depleted human induced neurons.The state of activation (red) or inhibition (blue) of a pathway or biological function is represented by a z-score (right-tailed Fisher's exact test).Results are displayed for two clones, 1 and 6 weeks after transcription factor-mediated induced human neurogenesis.n = 4 (clone 1) or n = 5 (clone 2) for (+/fl) and (+/−), respectively.H Reduced TUJ1 protein levels in MYT1L-depleted induced human neurons at day 7 can be restored by WNT and NOTCH inhibition via XAV939 and DAPT with n = 4 (clone 1).Bar graphs show mean values, data points from individual biological replicates are displayed, error bars = SEM, unpaired t-test in panel B and H, **p < 0.01, ***p < 0.001.

Fig. 5
Fig. 5 Electrophysiological hyperactivity in MYT1L-deficient human and mouse neurons in culture and brain slices.A Increased spontaneous neuronal network activity measured by multi-electrode array (MEA) in MYT1L-deficient (+/−; teal) neurons compared to control (+/fl, black) 6 weeks after induced neurogenesis.Representative raster plots and spike and network spike quantifications are shown.B Network hyperactivity measured by MEA for control (+/+, black) and mutant Myt1l (+/−; teal and −/−; yellow) hippocampal neurons in culture at day in vitro 11 (DIV11).Representative raster plots and quantifications are shown.C Increased spontaneous excitatory postsynaptic currents (sEPSCs) of CA1 pyramidal neurons in acute mouse brain slices.Representative traces, quantification and cumulative distributions are shown, bar graphs display mean values with number of patched cells or MEA wells from indicated biological replicates, error bars = SEM, Mann-Whitney-test in panel A and C, One-way ANOVA in panel B, *p < 0.05, ***p < 0.001, ****p < 0.0001.

Fig. 6
Fig. 6 Electrophysiological and behaviour phenotypes caused by MYT1L deficiency can be normalised by genetic and pharmacological intervention.A Deregulated non-neuronal MYT1L target genes in MYT1L-deficient human neurons (week 6) and primary mouse neurons at day in vitro 11 (DIV11) include the cardiac sodium channel SCN5A.Deregulation of gene expression based on RNA-Seq is displayed as fold change compared to control, n = 4 (human) for (+/fl) and (+/−), n = 3 (mouse) for (+/+) and (+/-), respectively.Tissue expression data from GTEx portal is displayed as z-score.B Schematic timeline for the genetic and pharmacological rescue experiments in mouse primary hippocampal and induced human neuronal cultures.C Significant downregulation of MYT1L target genes Scn5a and Hes1 in MYT1L-deficient mouse primary hippocampal cultures at DIV11 upon Myt1l overexpression (OE) at DIV3 compared to GFP OE determined by qRT-PCR.Graphs show column scatter plots with the median displayed; n ≥ 8. D Network hyperactivity in mouse primary neurons deficient of MYT1L can be rescued at DIV11 by overexpression of Myt1l in postmitotic neurons at DIV3.E Knockdown of Scn5a by shRNA expression in postmitotic neurons at DIV3 reduced spiking frequency of Myt1l (+/−) and (−/−) neurons.F Hyperactivity in mouse primary neurons deficient of MYT1L can also be rescued by acute application of 10 μM lamotrigine at DIV11.G SCN5A knockdown with shRNA in human induced neurons resulted in reduced spiking frequency.H Increased spike firing upon MYT1L mutation was normalised towards control by acute application of 10 μM lamotrigine in induced human neurons at week 6.I Open field tests showed that Myt1l (+/−) mutants covered more distance in centre regions compared to control animals at P23 and acute application of 20 mg/ kg lamotrigine can normalise these hyperactivity phenotypes.Vehicle: n = 23 for (+/+) and n = 29 for (+/−), lamotrigine: n = 19 for (+/+) and n = 23 for (+/−) data from three independent cohorts.J Schematic summary of the key roles of MYT1L during early and postnatal stages of development and defects upon MYT1L loss that can be compensated in part by genetic and pharmacologic intervention at both stages.Bar graphs display mean values with number of MEA wells from indicated biological replicates or data points from individual animals, error bars = SEM, Mann-Whitney test in panel C, Two-way ANOVA in panel D-I, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, NS not significant.