SINEUP RNA rescues molecular phenotypes associated with CHD8 suppression in autism spectrum disorder model systems

Abstract

Loss-of-function mutations in the chromodomain helicase DNA-binding 8 (CHD8) gene are strongly associated with autism spectrum disorders (ASDs). Indeed, the reduction of CHD8 causes transcriptional, epigenetic, and cellular phenotypic changes correlated to disease, which can be monitored in assessing new therapeutic approaches. SINEUPs are a functional class of natural and synthetic antisense long non-coding RNAs able to stimulate the translation of sense target mRNA, with no effect on transcription. Here, we employed synthetic SINEUP-CHD8 targeting the first and third AUG of the CHD8 coding sequence to efficiently stimulate endogenous CHD8 protein production. SINEUP-CHD8 were effective in cells with reduced levels of the target protein and in patient-derived fibroblasts with CHD8 mutations. Functionally, SINEUP-CHD8 were able to revert molecular phenotypes associated with CHD8 suppression, i.e., genome-wide transcriptional dysregulation, and the reduction of H3K36me3 levels. Strikingly, in chd8-morpholino-treated and ENU mutant zebrafish embryos, SINEUP-chd8 injection confirmed the ability of SINEUP RNA to rescue the chd8-suppression-induced macrocephaly phenotype and neuronal hyperproliferation. Thus, SINEUP-CHD8 molecule(s) represent a proof-of-concept toward the development of an RNA-based therapy for neurodevelopmental syndromes with implications for, and beyond ASD, and relevant to genetic disorders caused by protein haploinsufficiency.

Graphical abstract

CHD8 de novo mutations are linked to autism spectrum disorders (ASDs) via protein haploinsufficiency. Di Leva and colleagues employ SINEUP-RNA to boost CHD8 translation, restoring CHD8 protein levels, correcting transcriptional/epigenetic defects in human cells and neuroanatomical phenotypes in zebrafish. This study provides a proof-of-concept for SINEUP-based neurodevelopmental therapies.

Introduction

Among the genes associated with autism spectrum disorder (ASD), the chromodomain helicase DNA-binding 8 (CHD8) is one of the strongest ASD risk factors, independently validated in multiple studies.^1,2,3,4,5 A total of 51.5% of the identified mutations in CHD8 are disruptive (156 out of 301 variants), leading to protein loss of function.^2,3,5,6,7 Initial works explored the effects of CHD8 downregulation in vitro and in vivo to link cellular and molecular phenotypes to the reported aberrant brain development. CHD8 has pleiotropic functions, such as (1) transcriptional regulation due to the interaction with TP53,⁸ histone H1,⁹ and β-catenin¹⁰; (2) epigenetic remodeling, through its binding to MLL1 and WAR complexes^11,12; (3) control of cell proliferation^1,13,14; and (4) modulation of alternative splicing.^15,16,17,18

In vitro analyses of CHD8 reduction in human neuronal progenitor cells (NPCs) and cerebral organoids, or systemic heterozygous knockout in mice, caused cell-type-specific transcriptional dysregulation of a variety of functionally distinct ASD-associated genes and of a subset of genes implicated in cell-cycle regulation, cell proliferation, cancer progression, and alterations in electrophysiological properties. Moreover, it produced a significant reduction of trimethylated histone 3 at the residue lysine 36 (H3K36me3), correlated with genome-wide changes in alternative splicing patterns.^{14,18,19,20,21,22} In vivo, in mice and in zebrafish, Chd8/chd8 suppression correlates with increased head size as a result of the expansion of the forebrain/midbrain, caused by hyperproliferation of neuronal progenitors, along with an impairment of gastrointestinal motility and enteric nervous system.^1,14,21,23 Importantly, a closer examination of 15 ASD children carrying CHD8 mutations revealed phenotypic characteristics such as evident macrocephaly and gastrointestinal disturbances, clearly recapitulated in the zebrafish and mouse model systems.^{24,25,26,27,28,29}

Since most CHD8 mutations are disruptive and lead to protein haploinsufficiency, any approach able to specifically increase CHD8 protein expression may rescue the aberrant phenotypes and represent a new path for therapeutic intervention. Here, in the context of a neurodevelopmental syndrome, we test SINEUP, antisense long non-coding RNAs (lncRNAs) reported to augment, in a specific and controlled manner, the expression of target proteins, thus directly modulating the translation of their protein-coding sense mRNA counterpart.³⁰ SINEUP are modular lncRNAs, whose function is based on two domains: the binding domain (BD), the region that overlaps in opposite orientation with the 5′ of the target gene, providing binding specificity, and the effector domain (ED), containing the inverted short interspersed nuclear element (SINE) B2 element (invSINEB2), which mediates the upregulation of translation.³⁰ Thanks to this modular structure, it is now possible to design specific BDs targeting an mRNA of interest, allowing increasing protein synthesis of potentially every transcript without affecting the mRNA level.^{31,32,33,34,35,36} SINEUP molecules have been successfully tested in several cellular model systems,^{30,31,34,35,36,37} in vivo in aquatic³³ and mammalian³² animal models where they are able to increase target mRNA translation^37,38 in a physiological range. Notably, SINEUP-Gdnf (glial cell-derived neurotrophic factor) delivered into the striatum of a Parkinson’s disease mouse model leads to a specific and persistent increase of the target protein without the side effect of ectopic expression.³² Because of these advantages, SINEUP represents a promising therapeutic tool for diseases caused by haploinsufficiency, such as ASD with CHD8-inactivating gene mutations. This is especially important for genetic neurodevelopmental disorders and ASD in particular, where few and poorly effective drugs are available, most targeting only ASD-associated behavioral symptoms.^39,40 Thus, SINEUP RNA-based technology—thanks to its ability to stimulate protein production within a physiological range and only in cells/body districts where the transcript of interest is usually expressed^{34,35,38,41,42}—can be considered as an advanced tool for therapeutic intervention.

Here, we provide clear evidence that SINEUP-CHD8 is able to increase protein translation of endogenous CHD8 in human-induced neural progenitor cells and patient-derived fibroblasts with aberrantly reduced levels of the protein. Two different canonical SINEUP molecules (BD = −40/+4 bp around AUGs) directed against the translation initiation site (TIS) and against an in-frame, internal AUG were successfully used. SINEUP-CHD8 molecules were able to recover transcriptional dysregulation and epigenetic changes caused by CHD8 knockdown (KD) in vitro, whereas in vivo delivery of SINEUP-chd8 into one- to two-cell stage fertilized zebrafish eggs significantly rescued previously described macrocephaly in chd8 morphant and chd8^sa19827 heterozygous mutant larvae.^1,14,43

Results

Design, optimization, and in vitro testing of SINEUP-CHD8

To synthetize an effective SINEUP-CHD8 able to rescue physiological amounts of endogenous CHD8 protein levels in patients’ cells and in disease animal models, we designed a series of BDs targeting CHD8 mRNAs. First, we identified two target regions in the long (GenBank: NM_001170629) and short (GenBank: NM_020920) isoforms of CHD8 transcripts (Figures 1A and 1B) to design BDs overlapping the sense transcript. BDs with the canonical length (−40/+4 bp),^30,35,38 spanning the AUG of the TIS in the long isoform (SINEUP-CHD8_001) and in the short isoform (SINEUP-CHD8_002) (Figure 1B) were designed and cloned in antisense orientation into pDUAL_EGFP vector.

Figure 1.

SINEUP-CHD8/Chd8 stimulates protein production in human hiNPCs and mouse primary neuronal cells

(A) Schematic representation of SINEUP modular structure and its localization on CHD8 target transcript. BD, binding domain, overlapping in antisense orientation with the target mRNA; ED, effector domain, recruiting the target mRNA to polysomes to induce its translation. Two different SINEUP molecules have been designed which target either the translational initiation site (1^stAUG, 1^st Met) or one AUG in the CDS (int AUG). (B) Schematic representation of SINEUP-CHD8 (SINEUP_001/006, SINEUP_002, SINEUP_003/007) showing their location within CHD8 human isoforms, short GenBank: NM_020920 and long GenBank: NM_001170629, respectively. SINEUP_001 and _006 target the translational initiation site (1^stAUG) of the long isoform, SINEUP_002 targets the 1^stAUG of the short isoform, while SINEUP_003 and _007 recognize one in-frame AUG (Int AUG), which is common to both isoforms. (C) Representative western blots (WBs) for SINEUP_CHD8 delivery via electroporation or lentiviral transduction to a hiNPC CHD8 knockdown model obtained by Sh-driven CHD8 suppression (Sh4-CHD8) from Sugathan et al.¹⁴ Sh4-CHD8 cells were electroporated (left) or transduced (right) with lentiviral vectors expressing SINEUP control (pDUAL EGFP or pAIB-Empty) or the different SINEUP molecules targeting CHD8 (SINEUP_001, SINEUP_003, or the SINEUP_001 + 003 administered together). (D) Quantification of CHD8 protein levels from WB experiments (C) normalized on HSP90 and the respective control. HSP90 loading control—although smaller in size compared with CHD8 protein—carefully mirrors CHD8 protein levels in WB experiments, at high and low exposure, showing comparable correlation between the quantity of lysate and the chemiluminescence intensity with R2 coefficients above 0.95. (E) Schematic representation of SINEUP-Chd8 (SINEUP_01/02, SINEUP_03, and SINEUP_05) showing their target location within the Chd8 mouse ortholog (GenBank: NM_201637.3). (F) Representative WBs for SINUEP-Chd8 delivery via transfection to P19 cells. P19 cells were transfected with either SINEUP control vector (empty) or different SINEUPs targeting Chd8 (SINEUP_01, SINEUP_02, SINEUP_03, or SINEUP_05). (G) Quantification of Chd8 protein levels from WB experiments (F) normalized on β-actin and the empty control. (H) Primary cortical neuronal culture experimental setup from wild-type embryonic day 18.5 (E18.5) embryos (top) and representative images of neuronal cultures at 13 days in vitro (DIV) after AAV9 viral transduction with SINEUP_03 (bottom). (I) Western blots for SINEUP_03 delivery via viral transduction to primary cortical neurons. Primary neurons were transduced with SINEUP control vector (empty) or SINEUP_03 targeting Chd8. (J) Quantification of Chd8 protein levels from WB experiments, data normalized on Hsp90 (I). Data are presented as scatter dot blot, mean ± SEM with one-way ANOVA (parametric data) (G), or Kruskal-Wallis (non-parametric data) (D) significance testing followed by post-hoc tests for multiple testing correction; two-sided t-test in (J); significance levels: ∗∗∗∗p ≤ 0.0001, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p < 0.05; ns, not significant. Scale bar, 50 μm (H).

Although the majority of the previously designed SINEUPs target the TIS, a SINEUP against the internal in-frame AUG of the Frataxin transcript was proven to be effective in upregulating Frataxin protein.²⁸ Thus, we designed a third SINEUP targeting an in-frame internal AUG common to both isoforms (SINEUP-CHD8_003; Figures 1A and 1B). Finally, considering that shorter BDs can more easily reach target sites located in RNA structures not fully accessible due to secondary structures,^34,44,45 we designed and cloned shorter BDs (−14/+4 versus −40/+4; Figure 1B), still targeting the same regions in CHD8 transcripts: SINEUP-CHD8_06 (shorter BD for TIS in long isoform, GenBank: NM_001170629) and SINEUP-CHD8_007 (shorter BD for the internal common AUG).

Since packed secondary structures can impede the correct sense-antisense pairing at the BD region, the first 800 nt in both CHD8 transcript isoforms were closely inspected.⁴⁶ The MFold prediction tool (MFold web server: http://www.unafold.org) showed highly organized secondary structures rich in GC content at the 5′ UTR of both CHD8 isoforms. However, the complexity of the structures dropped in the vicinity of the TIS, with both RNA isoforms quite accessible and AUGs more exposed or only partially included in a hairpin loop (Figures S1A and S1B). Similarly, the internal, in-frame AUG in the long isoform was located in a partially accessible region (Figures S1A and S1B). Structural findings for the internal in-frame AUG codons were reinforced by in vivo click-selective 2-hydroxyl acylation and profiling experiment (icSHAPE) data obtained from mouse embryonic stem cells,⁴⁷ where this region is accessible and open to be targeted by SINEUP-BDs (Figures S1C–S1F). On the contrary, the two TIS surrounding regions did not present any signal from icSHAPE data, suggesting potential roadblocks for SINEUP binding. To gain information about possible different transcription start sites (TSSs) usage for the long and short CHD8 isoforms, we interrogated the ZENBU genome browser (ZEMBU genome browser: https://fantom.gsc.riken.jp/zenbu/⁴⁸) using the FANTOM5 CAGE collection.⁴⁹ We found that transcripts starting from the TSS upstream the TIS in the long isoform are generally much more represented than the ones originating from the TSS of the short isoform (Figures S1G and S1H). In particular, tissues and cell lines of relevance for nervous system development and function (induced pluripotent stem cell [iPSC]-derived neurons, neural stem cells, and newborn cerebellum) always showed expression from the TSS in the long isoform, whereas the short isoform was not expressed at all (Figures S1I and S1J).⁴⁹ In light of these considerations, SINEUP-CHD8_002, targeting the TIS of the CHD8 short isoform, was used as negative control to test SINEUP specificity.

SINEUP-CHD8s stimulate protein production in hiNPCs with reduced levels of CHD8 protein

As relevant model systems for this study,^{14,19,22,50,51,52} we employed iPSC-derived neuronal progenitor cell lines (hiNPC GM8330-8) and CHD8 KD clones. The latter were previously transduced with lentivirus to express short hairpin (Sh4) RNA targeting the CHD8 coding sequence to mimic CHD8 haploinsufficiency, or control shRNA targeting green fluorescent protein (GFP, used as negative control (Figures S2A–S2C)¹⁴).

At first, we delivered SINEUP_GFP, targeting GFP TIS, using a previously described plasmid, reported to increase GFP translation in in vitro and in vivo systems.^30,35,38 SINEUP_GFP and empty pDUAL_GFP vectors were electroporated into the hiNPC GM8330-8 parental cell line. Imaging and molecular analyses revealed a statistically significant increase in functionally active GFP protein production upon SINEUP administration (Figures S2D–S2F). Notably, the increase in GFP protein level was obtained without significant GFP transcript changes (Figures S2G), proving that hiNPCs are expressing the molecular machinery needed for SINEUPs to act as translational enhancers.

Then, to assess SINEUP-CHD8 efficacy in a cellular system expressing endogenous wild-type CHD8 levels, we initially delivered SINEUP-CHD8_001 (TIS, long isoform), SINEUP-CHD8_002 (TIS, short isoform), and SINEUP-CHD8_003 (common internal AUG) into hiNPC GM8330-8 cells by electroporation. We analyzed CHD8 protein levels in the presence or absence of SINEUP RNAs by western blot (WB) and transcript levels by RT-qPCR (Figures S3A–S3D). We observed a trend toward a positive protein increase with both SINEUP molecules targeting the long isoform (FC = 1.3 for SINEUP-CHD8_001 and FC = 1.7 for SINEUP-CHD8_003) (Figures S3A and S3B), with SINEUP-CHD8_003 reaching statistical significance. Consistent with the fact that CHD8 short transcript is only minimally expressed in our model system,^14,53 SINEUP-CHD8_002 (targeting the TIS in the short CHD8 isoform) did not elicit any effect when delivered into the cells (Figures S3A and S3B). As expected, no CHD8 transcript increase was detected when SINEUP RNA was highly expressed (Figures S3C and S3D).

Although SINEUP-CHD8 administration elicited a modest effect in cells already expressing physiological levels of this chromatin regulator (Figures S3A–S3C), ASD patients with CHD8 haploinsufficiency present reduced CHD8 protein levels. To mimic the disease condition, we tested SINEUP-CHD8 efficacy in the Sh4-CHD8 hiNPC line, previously transduced with Sh4 RNA, targeting the 24th exon of the CHD8 coding region.¹⁴ As shown in Figures S2A–S2C, Sh4-CHD8-mediated KD led to roughly 50% decrease in RNA and protein expression levels.

In this CHD8-KD model system, the delivery of SINEUP-CHD8_001 or SINEUP-CHD8_003 by electroporation resulted in a clear upregulation of CHD8 protein translation (Figures 1C and 1D). A statistically significant increase (1.4-fold) in CHD8 protein level was detected with SINEUP-CHD8_001, while administration of SINEUP-CHD8_003 led to a more substantial stimulation in protein production (1.8-fold). The observed increase in protein production is in line with the previously reported properties of SINEUP molecules.³⁶ The SINEUP mechanism was further supported by RT-qPCR experiments in which no change in CHD8 RNA levels was detected when SINEUP was overexpressed (Figures S3E and S3F).

To test whether a stronger or more reproducible effect in CHD8 protein stimulation could be obtained by co-administration of SINEUP molecules, we electroporated equal amounts of SINEUP-CHD8_001 and SINEUP-CHD8_003 in the Sh4-CHD8 hiNPC line. Although an increase in CHD8 translation was detected (1.4-fold, similar to the one described following administration of a single SINEUP molecule), no synergistic effect was observed by the contemporary delivery of two SINEUP RNAs (Figures 1C, 1D, S3E, and S3F, SINEUP-CHD8_001 + 003).

Considering some variability in the electroporation results, we moved to a lentivirus-mediated delivery of SINEUP-CHD8s. We transduced Sh4-CHD8 hiNPC with SINEUP-CHD8_001 and SINEUP-CHD8_003 after subcloning into a lentiviral expression vector. Similarly, to what was observed in the electroporation experiments, the efficiency of the SINEUP-CHD8_001 molecule was lower (FC = 1.25) compared with SINEUP-CHD8_003 (FC = 1.81, Figures 1C and 1D). Importantly, SINEUP-CHD8_003 upregulation was significant and stable across experiments, strengthening the discovery that targeting the internal in-frame AUG led to a more robust upregulation of CHD8 (Figures 1C and 1D). In all the experiments, no perturbation of CHD8 transcription was observed when SINEUP was present (Figures S3E and S3F).

Finally, we tested shorter SINEUP molecules. We successfully delivered the two short SINEUP (SINEUP-CHD8_006 and SINEUP-CHD8_007) using the optimized conditions into the Sh4-CHD8 hiNPC cell line. Although the experiment was repeated several times with reproducible results (n = 6), shorter SINEUP-CHD8 molecules were not able to increase CHD8 translation (Figures S3G and S3H). Similarly, and as expected, no increase of target gene transcription was reported when SINEUPs molecules were administered to cells (Figures S3I and S3J).

SINEUP-Chd8 molecules can increase Chd8 protein production in mouse model systems

To test the potential of SINEUP molecules in a second mammalian model system relevant for translational studies, we designed four SINEUPs with BDs targeting the mouse Chd8 transcript ortholog (GenBank: NM_201637.3). SINEUP-Chd8_01 and 02 were designed to overlap the first AUG (_01: −40/+4 bp canonical/standard length BD, _02: −14/+4 bp short BD), SINEUP-Chd8_03 the second, internal, AUG with a −40/+4 bp canonical BD, and SINEUP-Chd8_05 the third, internal, AUG with a −40/+4 bp canonical BD, respectively (Figure 1E). After cloning into a pAAV vector, the mouse P19 cell line was transfected with the individual expression plasmids. In line with previous observations in human neuronal progenitors, WB analysis revealed a modest, yet significant, increase in Chd8 protein levels in cells transfected with SINEUP-Chd8_02 and SINEUP-Chd8_03 plasmids (Figures 1F and 1G), whereas Chd8 mRNA levels remained unchanged (Figure S4A) upon robust SINEUP transfection (Figure S4B). Then, to assess SINEUP functionality in developing primary neurons, AAV9 viral particles were produced for SINEUP-Chd8_03, the most effective one, to transduce mouse primary cortical cultures (Figure 1H). To that end, cortical tissues from a total of 10 embryonic day 18.5 (E18.5) embryos were dissected, dissociated, and primary neurons were cultured for 8 days before AAV9 transduction. Five days later the cells were harvested and WB analysis revealed a significant stimulation of Chd8 protein production in SINEUP-Chd8_03 transduced primary neurons when compared with the empty control (Figures 1H–1J), while Chd8 mRNA levels remained unaffected by AAV9-mediated SINEUP expression (Figures S4C and S4D). Taken together, these data further support SINEUP functionality also in the murine background and show the feasibility to perform rescue experiments with the SINEUP-Chd8_03 in vivo using AAV9 viral vectors.

SINEUP-CHD8 molecules are functional to rescue cellular and molecular phenotypes due to CHD8 suppression

Since the original description of CHD8 as a chromatin remodeler, CHD8 depletion was reported to dysregulate the transcription of several genes^14,19 and induce reduction of H3K36me3 histone modification.¹⁷ Thus, we asked whether validated SINEUPs-CHD8 could rescue cellular and molecular phenotypes caused by CHD8 suppression. To this end, we resorted to the already described Sh4-CHD8 and Sh-GFP hiNPC from Sugathan et al.¹⁴ and transduced these cells with SINEUP-CHD8_001, SINEUP-CHD8_003, and empty lentiviral vectors. To have a global overview of the rescue efficacy, we performed genome-wide transcriptomic analyses using RNA sequencing (RNA-seq) to track changes elicited by CHD8 suppression and then rescued by SINEUPs-CHD8 treatment. At first, we validated the previously reported specific reduction of CHD8 transcript levels,¹⁴ with no significant alterations in other CHD family members (Figure S5A). We also confirmed a general enrichment of GO terms and KEGG pathways related to “cell-matrix adhesion,” “proteoglycans,” and “proliferation regulation” among the genes upregulated upon CHD8 suppression, whereas GO terms and pathways related to “nervous system development,” “synaptic organization,” and “axon guidance” were over-represented among the downregulated genes (Figure S5B). Then, we studied the effects on transcription of viral delivery of the control vector into Sh4-CHD8 cells. Although the transcriptomic landscape responded to the viral infection, the global transcriptional signature was preserved, with more than 60% (61% downregulated and 67% upregulated differentially expressed genes [DEGs]) of genes shared between Sh4-CHD8 and Sh4-CHD8 treated with SINEUP empty vector (Sh4-CHD8 pAIB-Empty, Figure S5C). Intriguingly, among the 2,477 genes that show concordant transcriptional dysregulation (significant p < 0.05 and −0.3 < LogFC > 0.3) after CHD8 suppression (Sh4-CHD8 versus Sh-GFP and Sh4-CHD8-empty versus Sh-GFP), transduction with SINEUP-CHD8_001 and SINEUP-CHD8_003 was able to restore the initial expression levels of 415 and 554 genes, respectively (Figures 2A–2E and S5C). In accordance with its higher efficiency (see Figures 1C and 1D), SINEUP-CHD8_003 rescued a higher number (22%) of dysregulated targets compared with SINEUP-CHD8_001 (16%), but the vast majority of them (341/415, 82% of SINEUP-CHD8_001 and 341/554, 61% of SINEUP-CHD8_003) were indeed rescued by both molecules (Figure 2C). Specifically, the rescued target genes—mainly upregulated following CHD8 suppression—were implicated in “cell-matrix adhesion,” “regulation of proliferation,” “nervous system development,” and the “regulation of gene expression” GO terms and pathways (Figure 2E). Notably, 21 genes that responded to SINEUP treatment and normalized their expression levels upon CHD8 protein increase are listed as ASD risk genes in the SFARI gene database, highlighting the possibility that SINEUP-CHD8 might modulate ASD-relevant transcriptional networks (Figure 2D; Table S1). As expected, SINEUP-CHD8 administration did not elicit any transcriptional change in CHD8 (Figure S5D).

Figure 2.

SINEUP-CHD8 molecules rescue cellular and molecular phenotypes associated with CHD8 suppression

(A and B) Heatmaps of deregulated genes upon CHD8 knockdown in Sh4-CHD8 and Sh4-CHD8 hNPCs cells transduced with lentiviral particles expressing the pAIB-Empty control vector, rescued through SINEUP_001-mediated (A) and SINEUP_003-mediated (B) CHD8 protein translation induction (n = 4 replicates per condition, all DEGs first filtered for adjusted p value < 0.05 and −0.3 < log2FC > 0.3 as indicated by dark gray arrows, resulting in a total of N = 415 and N = 554 responding DEGs, respectively. (C) Area proportional Venn diagram demonstrating overlap of number of DEGs and percent of overlap between rescued transcripts (at p_adj < 0.05 and −0.3 < log2FC > 0.3) in Sh4-CHD8 knockdown cells. (D) Selection of ASD-risk genes found among the N = 341 shared DEGs responding to SINEUP_001 and SINEUP_003 expression in Sh4-CHD8 hiNPCs (C) and their SFARI gene score. (E) Selected terms from KEGG 2021 (teal, top), GO-term biological process 2023 (aquamarine, middle), and Reactome 2022 (green, bottom) gene list annotation as performed by the Enrichr tool^54,55,56 on the N = 341 shared DEGs represented in (C). (F) CHD8 regulates H3K36 tri-methylation levels of cells chromatin (scheme, top), WB visualization of H3K36me3 levels in acidic Histone extracts upon hiNPC transduction (bottom). (G) Scheme of targeted LC-MS/MS approach to quantify H3K36me3 levels in histone extracts upon hiNPC transduction. A parallel reaction monitoring (PRM) method was established to target the trimethylated form of KSAPSTGGVKKPHR (H3.3_27–40) peptide. Skyline (version 24.1) was used to analyze and normalize peptide abundance changes (top). Relative abundance of H3K36me3 levels as obtained by targeted LC-MS/MS experiments (bottom) (n = 3 replicates per condition, data are presented as scatter dot blot ± SEM; one-way ANOVA followed by Sidak’s test for multiple testing correction, significance level ∗∗p ≤ 0.01, ∗p ≤ 0.05; ns, not significant).

To gain additional evidence on the functional effectiveness, we tested whether SINEUP-CHD8 could rescue the recently discovered phenotype of a reduction of histone H3K36me3 linked to CHD8 suppression¹⁷ (Figure 2F, top). Again Sh-GFP and Sh4-CHD8 hiNPCs were transduced with lentiviral particles with pAIB Empty, SINEUP-CHD8_001, and SINEUP-CHD8_003 vectors. Acidic histone extracts were prepared and subjected to targeted nano LC-MS/MS to quantify the trimethylated form of the KSAPSTGGVKKPHR (H3.3_27–40) peptide. As presented in Figures 2F and 2G, lentiviral delivery of SINEUP-CHD8_001 and, to a lesser extent, SINEUP-CHD8_003, was able to elicit a partial/significant increase of H3K36me3 levels as also additionally visualized by WB experiments with H3K36me3 antibody on the same histones’ lysates (Figure 2F, bottom).

SINEUP-CHD8 stimulates protein production in CHD8 mutation carrier-derived fibroblasts

The ability of SINEUP-CHD8 to increase CHD8 protein levels in hiNPC haploinsufficient cells and to rescue the molecular phenotypes associated to this condition, prompted us to evaluate their therapeutic potential in fibroblasts carrying de novo truncating mutations in the CHD8 locus (Figure 3A). Specifically, two lines were employed, one with a frameshift mutation (c.6307_6310del) and one with a stop-gained plus a frameshift mutation (c.2485dupA) within the same locus. Both CHD8 individuals’ derived fibroblasts presented reduced levels of transcript and protein levels when compared with controls (Figures S4E–S4G), confirming the expected haploinsufficiency for truncating mutations. SINEUP-CHD8 viral administration in the two patient-derived fibroblast lines, led to a statistically significant increase in CHD8 protein production in the range of 1.2- to 2.1-fold when compared with the control (Figures 3B and 3C). Interestingly, only SINEUP-CHD8_001 led to a statistically significant recovery of CHD8 protein in c.2485dupA cells, (Figure 3C, left), while both SINEUP molecules promoted CHD8 production in c.6307_6310del, with SINEUP-CHD8_003 being more effective (Figure 3C, right). In line with SINEUP functionality, CHD8 mRNA expression was unchanged when SINEUPs were administered (Figures 3D and 3E).

Figure 3.

SINEUP-CHD8 stimulates protein production in patient-derived CHD8 fibroblasts

(A) Schematic representation of the CHD8 variants present in the fibroblast cell lines used in this study. The main CHD8 domains are depicted as colored boxes along the CHD8 protein (CHROMO domains, DEXDc helicase domain, HELC-SANT domain, BRK domain). (B) Representative WBs on patient-derived fibroblasts harboring c.2485dup (left) and c.6307_6310del (right) mutations, transduced with SINEUP_001/003 or with the control vector (pAIB-Empty). (C) Bar graph representing fold change quantification of CHD8 protein level for WB experiments (B) normalized on HSP90. (D and E) Bar graph reporting RT-qPCR quantification of CHD8 transcript expression (D) or SINEUP expression (E) upon SINEUP-CHD8_001/003 or control vector transduction. NONO was used as a stable housekeeping gene for normalization. Data presented as mean normalized relative expression 2^(–ΔΔCt) ± SEM. n(WB) = 4–5 replicates per condition, n(RT-qPCR) = 3 replicates per condition; data are presented as scatter dot blot ± SEM, Kruskal-Wallis (WB) and one-way ANOVA (RT-qPCR) followed by post hoc multiple testing correction, significance level ∗∗p ≤ 0.01, ∗p ≤ 0.05; ns, not significant.

SINEUP-chd8 rescues chd8 haploinsufficiency-driven macrocephaly in zebrafish models

To test SINEUP efficacy in an in vivo model of ASD, we employed zebrafish model systems (Danio rerio).⁵⁷ This approach was suggested by previous reports demonstrating SINEUP efficacy in Medaka fish.³³ Initially, we confirmed chd8 expression at all developmental stages from blastula to 5 days post fertilization (dpf) in the AB/Tü wild-type strain by RT-qPCR (Figure S6A). We could detect highest chd8 transcript levels at blastula and gastrula stages, while decreasing at later stages, in line with previous reports from the ATLAS RNA-seq database (RNA-seq database: http://www.ebi.ac.uk/gxa/home). Then, to detect the presence of single-nucleotide polymorphisms (SNPs) within the prospective SINEUP annealing regions that could possibly hamper their efficiency, we amplified and sequenced the regions corresponding to the chd8 TIS and the chd8 internal AUG. No SNPs were found in either Tü/Tü or AB/Tü strains in the amplified regions (data not shown), thus confirming the annotated sequence as reported in Genome Browser (GenBank: NM_001347671).⁵⁸ In addition, we interrogated CAGE data from different zebrafish developmental stages (i.e., two cells, dome/30% epiboly, 14 somites, and prim-6) to gauge information about possible different TSS usage, and therefore the expression of other endogenous transcript isoforms to consider for SINEUP design.⁵⁹ CAGE data confirmed that the TSS reported for GenBank: NM_001347671 was ubiquitously expressed in all the tested conditions. However, an alternative TSS could be hypothesized since a low-to-moderate CAGE signal was observed starting from exon 3. Given the low expression of this alternative TSS and the fact that an experimental confirmation of its existence was not available, we selected the regions of interest (TIS and internal AUG) in the zebrafish chd8 GenBank: NM_001347671 isoform (Figure 4A), to design two chd8 BDs (SINEUP_004 and SINEUP_005). Similarly to the human RNA sequences, we performed chd8 RNA secondary structure analysis through MFold tool (MFold web server: http://www.unafold.org). The TIS of chd8 is localized in a more compacted stem structure, while the internal AUG falls into a hairpin, more open arrangement, predicting a better accessibility for this site (data not shown). To exclude off-target effects, we computationally tested whether SINEUP-chd8_004 and SINEUP-chd8_005 sequences could bind with significant strength to other transcripts, different from chd8. The results showed that SINEUP molecules were only partially (less than 50%) overlapping with other transcripts (20/44 nt with E value > 0.2) (Table S2), therefore suggesting target specificity. SINEUP-chd8_004 and SINEUP-chd8_005 were then cloned into pCS2+link vector and their sequences verified by Sanger sequencing.

Figure 4.

SINEUP-chd8 rescues chd8 suppression-driven macrocephaly in chd8 mutant zebrafish models

(A) Schematic representation of chd8 SINEUP-chd8_004 and _005 showing their location within chd8 zebrafish isoform GenBank: NM_001347671 (top) and enlargement of the first 9 exons of chd8 in zebrafish with the localization of SINEUP-chd8_004/005 and the two morpholinos used (MO3 targeting exon 7, while MO4 targeting exon 8) (bottom). (B) Graphical representation of the experimental setup on zebrafish embryos. (C) Bar graph represents percentage of larvae showing macrocephaly (calculated as eye distance >12% in respect to wild type, light gray) or normocephaly (dark gray). Measurement of eye distance was performed both in wild-type Tü/Tü or AB/Tü strains of zebrafish larvae injected with 8 ng of chd8-MO3, 8 ng of chd8-MO4 with (+) or without (−) ∼200 pg of SINEUP-chd8_004 or _005; non-injected zebrafish larvae were used as control. Scramble (S) MO or SINEUP-chd8_004/005 alone were used as negative control (mean ± SEM are shown). Fisher-Yates exact test was performed. ∗p ≤ 0.05. n ≥ 25 embryos/condition. (D) Representative images of chd8^p.Glu223∗ heterozygous zebrafish larvae head (4.2 dpf), dorsal views, injected with control or SINEUP-chd8 (left) and scatter dot blot reporting eye distance in wild-type (AB chd8^+/+) or chd8^p.Glu223∗-injected larvae (right). (E) Representative images of whole-mount PHH3 immunostaining on 2 dpf chd8^p.Glu223∗ embryos injected with control or SINEUP-chd8_005 (left), and scatter dot plot showing the number of PHH3-positive cells in the head of 2 dpf chd8p.Glu223∗ heterozygous zebrafish embryos upon treatment (data were normalized to the number of PHH3-positive cells in the head of the 2 dpf wild-type (AB chd8^+/+, right). (F) Representative lateral views of the hindgut area of chd8^p.Glu223∗ zebrafish larvae (4.2 dpf) injected with control or SINEUP-chd8_004 or SINEUP-chd8_005 (left) and stained with anti-HuC/D, and scatter dot plot showing the number of HuC/D-positive cells in SINEUP_005- or SINEUP_004-injected larvae and control-injected larvae (right). Data are presented as scatter dot blot of mean ± SEM. One-way ANOVA was performed followed by Sidak’s post-hoc test for multiple testing correction (D and G); two-sided t test in (E); significance levels: ∗∗∗∗p < 0.0001, ∗p < 0.05; ns, not significant. Scale bars, 100 μm (D and F), 50 μm (E). A, anterior; P, posterior; e, eye; y, yolk.

To mimic the CHD8 haploinsufficiency described in ASD individuals, carrying de novo mutations in CHD8, we injected zebrafish embryos with two different morpholino oligonucleotides (MOs) previously used successfully to reduce chd8 levels^1,14 by disrupting the splice site donors at exon 7 and exon 8, respectively. The two different MOs were used at the same dosage (8 ng) as tested previously.^1,14 At 4.2 dpf, fish larvae were anesthetized, collected, and fixed in 4% paraformaldehyde (Figure 4B). As anticipated, both MOs caused the expected aberrant splicing (Figures S6B–S6D) and a concomitant reduction of WT chd8 transcript (Figure S6D, primers chd8 int6), while a previously described general increase in total chd8 transcript expression was detected using primers on exons 31/32, away from the region of MO targeting (Figure S6D, primer chd8 31–32).^1,14

As expected,^1,14 MO injection induced a macrocephaly phenotype as assessed by eye distance measurements at three different positions of the fish head (Figures 4B and 4C). The macrocephaly phenotype was defined by an eye distance exceeding by at least 12% the average distance in control embryos (described previously in Sugathan et al.¹⁴). We then calculated the percentage of larvae with macrocephaly or with head size in the normal range, for all conditions. As expected, MO3 and MO4 increased the percentage of macrocephaly (43% and 78%, respectively), when injected with 8 ng dosage compared with control embryos (not-injected or scramble-injected [S]) (Figure 4C).^1,14

Finally, we proceeded with in vitro transcription of capped SINEUP-chd8_004 and SINEUP-chd8_005, which were injected (100 and 200 pg) into one- to two-cell stage zebrafish fertilized eggs, with or without chd8 MOs (Figure 4C). Importantly, MOs sites in the chd8 RNA sequence were different and not overlapping with the target sites of the SINEUP molecules, thus excluding a possible interference with SINEUP functionality (Figure 4A). Although lower doses (∼100 pg) of SINEUP-chd8_005 and SINEUP-chd8_004 molecules were not able to rescue chd8-MO-induced macrocephaly (data not shown), ∼200 pg of SINEUP-chd8_004 and SINEUP-chd8_005 were effective in reducing MO4-induced macrocephaly (from 72% to 31% with SINEUP-chd8_005 and from 72% to 52% with SINEUP-chd8_004) (Figure 4C). Taken together, SINEUP-chd8_004, directed against the chd8 TIS, seems to be less efficient compared with SINEUP-chd8_005 directed against the internal AUG, similar to our in vitro results obtained with hiNPCs. On the contrary, when we co-injected MO3 and SINEUP-chd8_004 or SINEUP-chd8_005, a moderate trend toward improvement in the macrocephaly phenotype was observed, but it did not reach statistical significance (Figure 4C). Finally, we confirmed that both SINEUP-chd8_004 and SINEUP-chd8_005 did not induce macrocephaly when injected alone (Figure 4C). Importantly, the rescue effect observed by injecting SINEUP-chd8 molecules did not elicit a change in chd8 mRNA expression levels (Figure S6E). We also confirmed that SINEUP molecules were correctly delivered and expressed in the zebrafish larvae (RT-qPCR on 4.2 dpf larvae, Figure S6E).

To further test the efficacy of SINEUP-chd8_004 and SINEUP-chd8_005 in a stable ENU (N-ethyl-N-nitrosourea)-induced chd8 mutant zebrafish line, we utilized our previously described chd8^p.Glu223∗ that, when in heterozygosity, exhibits chd8 haploinsufficiency, macrocephaly due to increased cell proliferation, and a reduced number of enteric neurons.⁴³ Coherently with previous experiments, we injected 200 pg/egg of the SINEUP_control, SINEUP-chd8_004, or SINEUP-chd8_005 into the cell at one-cell stage and performed analysis at 2 or 4.2 dpf (Figure 4B). First, injection of SINEUP_control molecules (not-targeting control with a BD directed against human CHD8 short isoform, not annealing in the zebrafish genome) did not affect the macrocephaly phenotype (eye distance, 199 μm in the mutant SINEUP_control injected larvae versus 153 μm in the wild-type control) typical of chd8 heterozygous mutants (Figure 4D). Conversely, injection of SINEUP-chd8_005 molecules, but not of SINEUP-chd8_004, led to a complete rescue of the macrocephaly phenotype at 4.2 dpf (Figures S6F and 4D), consistent with aforementioned data on the chd8 morphants. Notably, and as reported previously,^1,14,43 chd8 haploinsufficiency-associated macrocephaly was accompanied by increased cell proliferation on 2 dpf embryos. Therefore, to explore whether the rescue of the macrocephaly phenotype by SINEUP-chd8_005 injection was preceded by a change in proliferation rate, we counted the number of proliferative cells at 2 dpf in injected embryos using phospho-histone H3 immunostaining (Figure 4E). Strikingly, zebrafish embryos injected with SINEUP-chd8_005 revealed a significant decrease in the number of proliferative cells compared with embryos injected with the SINEUP_control (Figure 4E), suggesting that SINEUP-chd8_005 is able to rescue the proper balance of cell proliferation during early development. However, neither the injection of SINEUP-chd8_005 nor SINEUP-chd8_004 corrected the lower count of enteric neurons in the intestine observed in chd8 heterozygous mutant larvae⁴³ (Figures S6G 4F, and 4G).

Taken together, the cumulative evidence derived from transient and stable zebrafish models for chd8 haploinsufficiency highlights the capacity of SINEUP molecules to effectively modulate cell proliferation at early developmental stages to rescue chd8 haploinsufficiency-associated macrocephaly in vivo.

Discussion

CHD8 mutations have been recurrently and independently validated as risk factors for ASD. More than 50% of these mutations cause an insufficient quantity of CHD8 protein production.^2,4,60 Although several different studies aimed to unveil the molecular mechanisms underlying this condition, no effective therapies are currently available.^39,61 In recent years, several therapies based on nucleic acids have been developed to treat human diseases, with some of them already on the market and many more in clinical trials.⁶² RNAs such as small inhibitory RNAs and anti-sense oligonucleotides (ASOs) aim to interfere with pathogenic genes by inhibiting their expression.⁶³ In contrast, diseases similar to CHD8-associated ASD can benefit from drugs that augment transcription and/or translation of the affected gene. To this purpose, several gene-specific transcriptional activating RNAs and non-degradative ASOs have been developed to stimulate both in vitro and in vivo the production of the target gene.^64,65 Similarly, SINEUP molecules share the ability to increase, in a physiological and controlled way, the translation of a target transcript.^30,41 Their modular structure allows synthetic design of an overlapping region (BD) to target transcripts of interest without changing the overall structure of the original lncRNA. Over competing technologies, SINEUPs have three major advantages: (1) they modulate translation of target mRNAs without introducing stable genomic changes into target cells, (2) their induction of selected protein is typically in a more physiological range (2-fold) than most conventional gene replacement strategies, avoiding the unwanted side effects that have halted many clinical trials in the past, (3) they do not act on cells that do not express the target mRNA avoiding ectopic overexpression of the therapeutic protein in cells that normally do not contain it.^30,41,42

In this study, we could show that SINEUP-CHD8 are functioning in vitro to increase CHD8 protein translation in model systems presenting CHD8 reduced levels. Importantly, in human neuronal progenitors, we showed that SINEUP targeting CHD8 not only effectively increase protein production (on average by 1.5-fold), but also substantially rescue CHD8-suppression aberrant expression of dysregulated genes and restore reduction in H3K36me3 deposition at actively transcribed genes.¹⁷ Here, a short-hairpin RNA-based CHD8 suppression cellular model was employed, which exhibited a robust response to CHD8-SINEUP treatment. Parallel investigations are currently underway to evaluate the application of SINEUP technology to CRISPR-edited human cell lines.

Moving forward, to provide further support to the effectiveness of SINEUP-Chd8 molecular tool in another mammalian model system, we successfully tested SINEUP-Chd8 molecules targeting a mouse Chd8 ortholog. These results are particularly relevant for the translatability of SINEUP technology for acquiring preclinical data with in vivo testing via behavioral neuroscience and functional brain imaging.

The active SINEUPs have a BD of 44 nt—of the “canonical,” standard length, described in the initial SINEUP studies^30,32,34,35—that perfectly pair the sense transcript both on the first and the internal in-frame AUG of the long CHD8 isoform. Consistent with its negligible expression levels in hiNPC, SINEUP-CHD8 targeting the short CHD8 isoform did not elicit any significant increase in protein production. Thus, SINEUP-CHD8 directed toward the short, not-expressed CHD8 isoform offered a specific non-targeting control, whereas furtherly supporting the specificity of SINEUP molecules to only target transcripts actively transcribed in the model system of choice. In addition, testing SINEUPs with shortened BD (18 bp) did not lead to any statistically significant effect. Therefore, at least in the context of CHD8 haploinsufficiency and differently from BDs targeting other mRNA species, shorter SINEUP molecules seem less efficient compared with the canonical, long molecules. Although not fully dissected, structural analysis on SINEUP molecules revealed the importance of a “core domain” of 88 bases in the central region of the ED for which little nucleotide changes/deletion can diminish functional efficacy.^45,66,67 In fact, other RNA molecules, such as riboswitches, are known to be able to fold into different structures, but only few of them retain functionality.⁶⁸ Thus, in our molecular and cellular context only the canonical long SINEUP-CHD8 might acquire the ideal secondary structure when bound to the CHD8 transcript.

To better grasp the functional characteristics of SINEUP-CHD8, we wondered whether administering multiple SINEUP molecules at the same time would promote protein upregulation efficacy. No synergistic effect was observed when two SINEUP-CHD8 molecules were administered together in hiNPCs, although maintaining the same final concentration of a single treatment. Since a dose-dependent relationship was observed in previous experiments (a higher SINEUP molar ratio led to coherently higher protein production³⁵), we cannot exclude that substantially escalating (doubling or more) the molar ratio could ameliorate the combined functional effect.

Crucially, for possible therapeutic transition of the SINEUP-CHD8 molecules, here we showed that the newly produced CHD8 protein retains unaltered functionality as proven by its ability to rescue genome-wide transcriptional changes caused by CHD8 haploinsufficiency and epigenetic phenotypes related to reduced levels of H3K36me3. Interestingly, the observed functional rescues are generally in line with the reported, different efficiencies for SINEUP-CHD8; i.e., the most efficient SINEUP-CHD8 targeting the internal AUG is indeed the molecule eliciting better rescue at the translational level. Nevertheless, some variability in the rescue efficiency is noted and might be linked to the type of analysis conducted (transcription versus histone modification analysis), but also whether the measured phenotypes could be directly connected to CHD8 protein levels or rather indirectly linked to downstream disruption of other molecular players. Accordingly, more than 80% of the rescued transcriptional targets—indeed shared between samples treated with the two different SINEUP-CHD8 molecules—are especially enriched for genes directly bound by CHD8 on their promoter and upregulated following CHD8 suppression. However, at best, SINEUP-CHD8 treatments in hiNPC could revert no more than 20% of the transcriptional changes elicited by CHD8 suppression. Recent studies report distinct, non-linear effects of Chd8 dosage during cortical development and X chromosome inactivation.^23,69 As such, different SINEUP-CHD8 molecules would need to be tested for their ability to increase CHD8/Chd8 protein at different levels, in order, then, to obtain a stronger transcriptional rescue. Of note, genome-wide transcriptional and epigenetic phenotypes can be solidly measured even without a specific sorting of the SINEUP-CHD8 transduced hiNPC: although possibly modest, a dilutive effect elicited by the not-transduced cells cannot be ruled out.

We then tested SINEUP-CHD8 molecules in patient-derived fibroblasts harboring inactivating CHD8 mutations. We noted different efficiency in SINEUP activity between the two lines—potentially ascribable to diverse levels of CHD8 transcript expression (wild type and mutant) and/or CHD8 protein levels. In fact, at least in humans, cells with lower CHD8 transcript and protein levels display better SINEUP functionality. This consideration is in line with other observations reporting negligible SINEUP functionality for highly expressed targets (i.e., ACTB) and furtherly mirrors what was also described for different translation activators, such as CRISPRa technology.^70,71 However, the data collected so far are promising and confirm the ability of SINEUP-CHD8 molecules to rescue protein haploinsufficiency in human cells directly derived from ASD patients. These results represent another step forward toward the therapeutic applicability of this molecular tool in the context of ASD.

We moved to test the SINEUP-chd8 molecules in vivo using zebrafish. Importantly, several molecular and phenotypic characteristics described in ASD patients harboring CHD8 mutations such as increase in brain size (macrocephaly) and gastrointestinal defects can be recapitulated in zebrafish following chd8-morpholino administration.^1,14,43 Here, we first selected and injected two splice blocking chd8-MOs that have been previously characterized, and we recapitulated the already reported macrocephaly.¹⁴ Similarly to that observed in iNPCs, SINEUP-chd8 directed against the internal AUG was more efficient compared with the one against the TIS, thus suggesting a possible peculiar SINEUP-CHD8/chd8 secondary structure formed by the sense-antisense pairing that might be instrumental for functionality. Strikingly, SINEUP-chd8 and, specifically, the one directed against the internal AUG, confirmed its clear effectiveness in a stable ENU induced zebrafish mutant line,⁴³ in which injection of SINEUP-chd8_005 efficiently rescued neuronal hyperproliferation and brain macrocephaly. In contrast, the reduced number of enteric neurons associated with chd8-suppression was insensitive to SINEUP-chd8. We speculate that the failure to rescue the intestinal phenotype may be due to the continued division and migration of enteric neuron precursors, the neural crest cells. As these cells proliferate and migrate to colonize the gut, the SINEUP-chd8 content in individual cells could be rapidly diluted, reducing it to levels insufficient for effective rescue.

Although most of the reported effective SINEUPs are targeting the TIS, our data strongly indicate that SINEUP-CHD8 against the internal in-frame AUG are more efficient. Data from both in vitro (human iNPC, mouse cultures) and in vivo zebrafish models support this conclusion. These considerations are also in line with icSHAPE results displaying a more accessible structure for the region surrounding the internal in-frame AUG, suggesting that a more relaxed secondary structure can be, at least in part, the reason for the higher efficacy of SINEUP-CHD8. Indeed, at least another example of SINEUP targeting internal AUG has been reported. This is the SINEUP targeting the Frataxin gene, which showed efficient upregulation of the Frataxin protein.³¹ These results suggest that target mRNA can be loaded on the ribosome by an internal bait that does not interfere with translational initiation. In general, a retrospective study to integrate information about target transcript folding/structure with downstream target engagement would be advisable, comprehensively analyzing many different transcripts (many of which have been already targeted by SINEUP), their levels of basal expression, the length and location of various BDs, and possible modifications of the target transcript mRNA (i.e., m6A methylation). This will be a crucial resource to assist everyone in the design of effective SINEUP BDs.

In conclusion, our data provide experimental evidence toward the development of a state-of-the-art type of RNA-based therapy for neurodevelopmental disorders, with implications beyond CHD8 haploinsufficiency and ASD, and with potential impact for a large repertory of currently incurable, neurologic conditions caused by reduced levels of target protein. Thereby, timing of delivery remains a crucial issue: when should SINEUP molecules be delivered to rescue the phenotype? CHD8 plays a pivotal role during the early phases of human development; thus, the appropriate dosage and developmental window of intervention need to be properly explored. While in recently described Chd8^+/− mouse models,^{15,19,20,21,22,23,72} in utero (E13.5) SINEUP-Chd8 administration could be proposed to possibly rescue disease-associated phenotypes, in human patients this raises a series of serious ethical and technical concerns. However, a recent study in mice⁷³ reported a strong imbalance in excitatory/inhibitory transmission occurring at a key developmental window—the juvenile period (from postnatal days 14 to 20). These data offer a unique angle of intervention, suggesting that the delivery of SINEUP-Chd8 post-birth might still be efficient in alleviating important ASD phenotypes. While further studies are needed to address these questions and provide experimental evidence to promote the clinical development for this unmet medical need, this approach overcomes challenges associated with the lack of prenatal diagnosis for CHD8 mutations and ethical concerns for in utero SINEUP administration.

Materials and methods

Cell culture

Human iPSC-derived NPC line GM8330-8, and Sh4-CHD8 and Sh-GFP lines, previously generated by lentiviral delivery of shRNAs targeting CHD8 and GFP coding sequences, respectively, were kindly provided by the laboratory of Dr. Stephen Haggarty (Massachusetts General Hospital and Harvard Medical School, Boston, MA).^14,74 Cells were cultured on poly-L-ornithine hydrobromide (20 μg/mL, Sigma)/laminin (3 μg/mL, Life Technologies)-coated plates in hiNPC medium (70% v/v DMEM (Life Technologies) completed with 30% v/v HAM F12 (Euroclone), 2% v/v B27 (Life Technologies), 1% v/v penicillin-streptomycin solution (Life Technologies), and 1% v/v L-glutamine (Corning), and supplemented with EGF (20 ng/mL, Sigma), bFGF (20 ng/mL, R&D), and heparin (5 μg/mL, Sigma). Semi-confluent monolayers of hiNPCs were maintained in 5% CO₂, in a 37°C humidified incubator.

Fibroblasts from healthy individual GM03652 (Coriell) were kindly provided by the laboratory of Dr. Gemma Louise Carvill (Northwestern University, Feinberg School of Medicine, Chicago, IL). Patient-derived fibroblasts, TR0000002 and TR0000028, harboring de novo CHD8 mutations, were collected in accordance with University of Washington IRB protocol HSD#42744 and kindly provided by Dr. Raphael Bernier’s laboratory (Center on Human Development and Disability, University of Washington, Seattle, WA). Human embryonic kidney HEK293T and human fibroblasts were cultured in DMEM (Life Technologies) completed with 10% v/v FBS (Gibco), 1% v/v L-glutamine (Life Technologies), and 1% v/v penicillin-streptomycin solution (Life Technologies). Semi-confluent monolayers of cells were maintained in 5% CO₂, in a 37°C humidified incubator.

SINEUP-CHD8 design and cloning

pEGFP-C2 (Clontech, here pDUAL_EGFP) and SINEUP targeting EGFP (here SINEUP_EGFP) were described previously by Carrieri et al.³⁰ SINEUP targeting human CHD8 and zebrafish chd8 (BD) were cloned into pDUAL_EGFP and into pCS2 vectors, respectively, already containing the inverted SINEB2 element (ED) in their backbones. The BDs were created by selecting −40/+4 bp (canonical SINEUP) and −14/+4 bp (short SINEUP) regions overlapping the TIS and the AUG for the internal, in-frame AUG of both human CHD8 and zebrafish chd8. Specifically (1) the TIS of the long (GenBank: NM_001170629) and short (GenBank: NM_020920) CHD8 isoforms (SINEUP_001, SINEUP_006 for the human GenBank: NM_001170629 and SINEUP_002 for the human GenBank: NM_020920); (2) the internal, in-frame AUG common to both human isoforms (SINEUP_003 and SINEUP_007); (3) the TIS (SINEUP_004) of the zebrafish chd8 isoform (GenBank: NM_001347671); and (4) the internal, in-frame AUG targeting zebrafish isoform (SINEUP_005) were designed. Full-sequence details of designed SINEUP BDs are included in Table S3. BDs were synthesized (Eurofins) and cloned in inverted orientation into pDUAL_EGFP or pCS2 upstream of the ED using T4 DNA ligase (Invitrogen), following the manufacturer’s instructions. Ligations were transformed into DH5α cultured in LB broth at 37°C supplemented with kanamycin (40 μg/mL) for plasmid selection. Plasmids were purified using the PureYield Plasmid Maxiprep System (Promega) following the manufacturer’s instructions. Ethanol precipitation (sodium acetate 3 M and ice-cold 100% ethanol) was performed to concentrate plasmid. After a centrifugation at 13,000 rpm for 30 min at 4°C, DNA pellets were washed twice with 75% ethanol and re-suspended in water. Plasmids were analyzed by restriction map and sequencing to confirm the presence of the correct BD in the correct orientation. The MFOLD web server (MFold web server: http://unafold.rna.albany.edu/?q=mfold) was used to predict CHD8 RNA secondary structures (Figures S1A and S1B). Analysis of alternative TSSs for the CHD8 gene was performed using the ZENBU genome browser (ZEMBU genome browser: https://fantom.gsc.riken.jp/zenbu/), as described previously (Figures S1C–S1F).^48,49 SINEUP off-target prediction was performed using Basic Local Alignment Search Tool (BLASTN search tool: https://blast.ncbi.nlm.nih.gov/Blast.cgi) against the human and zebrafish cDNA database (Ensembl repository GRCh38 and GRCz11, respectively). Results with an E score below 1 were selected and are listed in Table S2, if multiple transcripts for the same gene were obtained, only the most relevant isoform was considered in the table.

SINEUP-CHD8 viral vector production

SINEUP_001 and SINEUP_003 BD and ED regions were subcloned into pAIB viral vector under SSFV promoter. Three plasmids were used for viral preparation: pAIB containing SINEUP of interest, psPAX2 for viral packaging, and pHDMG VSV-G envelope protein carrier. The different plasmids were transformed into DH5α cultured in LB broth at 37°C supplemented with antibiotics for plasmid selection. Plasmids were purified using the PureYield Plasmid Maxiprep System (Promega) following the manufacturer’s instructions. Plasmids were analyzed by restriction mapping and sequencing to confirm the presence of the correct SINEUP in the correct orientation.

Lentiviral particles were produced in HEK293T cells at 40% confluency in a 100 mm disk. Before adding the plasmid for viral vector production, the medium was changed with DMEM supplemented with 10% FBS and 1% L-glutamine (without antibiotics). A solution for each SINEUP was created by mixing in 1 mL final volume of Opti-mem (Gibco): 50 μL of PEI (Sigma), 10 μg of pAIB vector, 7.5 μg of psPAX2 vector, and 2.5 μg of pHDMG vector. The mix was then vortexed and incubated for 10 min at room temperature before adding it to HEK293T, which was finally incubated ON at 37°C. The day after the medium was changed with complete medium with 4% v/v penicillin-streptomycin solution (Life Technologies). Cells were again incubated at 37°C for 48 h before collection of the medium containing the produced lentiviral vectors (MVV). The MVV was centrifuged for 5 min at 500 × g and the supernatant was filtered using a 0.40 PES filter (Sartorius) and aliquoted for downstream experiments. Vector titers were measured as reverse transcriptase units (RTU) by the SG-PERT method.⁷⁵

Cell electroporation

In brief, pDUAL_GFP containing SINEUP_001, SINEUP_002, SINEUP_003, SINEUP_006, SINEUP_007, SINEUP_GFP, or pDUAL_EGFP empty vector were transiently delivered into hiNPC GM8330-8 or Sh4-CHD8 lines. A total of 5 × 10⁶ cells and 5 μg of plasmid was electroporated using the Nucleofector Device (Lonza) with A033. Cells were maintained in culture for 24 or 48 h and then collected for RNA and protein extraction.

Viral vector transduction

For transduction experiments, hiNPCs and patient-derived fibroblasts were plated at 60% confluency in medium without antibiotics. Cells were treated with 1.5 RTU of lentiviral vectors in the presence of 4 ng/mL of polybrene (Sigma). The medium was changed with complete medium supplemented with 1% v/v penicillin-streptomycin solution (Life Technologies) after incubation ON at 37°C and cells were collected after 48 h of incubation for protein and RNA quantification.

Cell fixation and microscope analysis

After 24 or 48 h, cells electroporated with pDUAL_EGFP with or without SINEUP_GFP were fixed with 4% PFA and stained with Hoechst 33342, 1:20,000 (Life Technology) for further analysis using Operetta CLS High-Content Analysis System (PerkinElmer). For every image, 22 visual fields were taken. The GFP fluorescence of each field was measured when the nuclei stain Hoechst 33342 was detected. The different field intensities were then averaged together to obtain the fluorescence for each condition.

Mouse SINEUP-Chd8 design and testing

Mini SINEUP-Chd8 molecules were designed to target the Chd8 mRNA GenBank: NM_201637.3 transcript at the following positions: SINEUP-Chd8_01 targeting −40/+4 of the first AUG, sequence: CCATCTTGGTAGAGTAATAAAGGGTATTTGTTGAAGTTCCGGGG; SINEUP-Chd8_02 targeting −14/+4 of the first AUG, sequence: CCATCTTGGTAGAGTAAT; SINEUP-Chd8_03 targeting −40/+4 of the second AUG, sequence: CCATGATGGGGTCTGCCATCTTGGTAGAGTAATAAAGGGTATTT; and SINEUP-Chd8_05 targeting −40/+4 of the third AUG, sequence: TCATCTGATCCAAGGAGTCCAGAGAGCTTGGCAGTCCAAGTGCT; SINEUPs were cloned in a pAVV9 vector under the CAG promoter.³²

P19 cells were cultured according to standard cell culture conditions and transfected at 80% confluency with the SINEUP-Chd8 vectors using Dharmafect in Opti-MEM. Cells were harvested after 24 h post-transfection for further analysis. AAV9 viral particles were generated as described previously.³²

Primary neuronal culture from mouse embryonic cortical tissue

All animal experiments were conducted to minimize pain and discomfort, under the approved Institutional Animal Care and Use Committee protocol of the Italian Ministry of Health (project authorization no. 1075/2020-PR, revised as 3FAF3.48.EXT.52). The mouse primary neuronal culture experiments were performed according to Beaudoin et al.,⁷⁶ with few modifications. In brief, pregnant C57BL6/J female mice were sacrificed through cervical dislocation, E18.5 embryos were removed from the uterus, placed in ice-cold 1× PBS supplemented with 1× penicillin-streptomycin and decapitated with sharp scissors. Dissection of developing forebrain tissues and removal of meninges was performed rapidly under a stereomicroscope in HBSS with 1× penicillin-streptomycin, tissue was mechanically fragmented and washed 2× with fresh dissection buffer (tissue fragments were allowed to settle through sedimentation before buffer changes). Tissue was dissociated using 500 μL of 0.25% trypsin-EDTA in a 37°C water bath for no longer than 15–20 min. Then, DNase 1 was added to the dissociation buffer and incubated for an additional 5 min at room temperature. Dissociation medium was carefully removed, the tissue clumps washed 2× in fresh dissection buffer and 2× in seeding medium (DMEM supplemented with 2 mM L-glutamine, 100 U/mL penicillin-streptomycin, and 10% FBS). Then, using fire-polished Pasteur pipettes, the tissue pieces were triturated 8–10 times until tissue pieces disappeared. Cell suspension was filtered through a 70-μm cell strainer to obtain single cells. Cell aliquots were counted using Trypan blue to (1) determine viability of the cells and (2) cell numbers, and cells were seeded in seeding medium on PLO/laminin-coated six-well tissue culture plates for biochemical analysis (20 × 10⁴ cells per well). After 4 h, cells were checked for attachment to the substrate and seeding medium exchanged with neural maintenance medium (Neurobasal medium, Gibco, no. 21103049, supplemented with B27, GlutaMax, and penicillin-streptomycin). Fifty percent of the medium was then exchanged every 3–4 days for culture maintenance. At 8 days in vitro (DIV) cells were transduced with SINEUP-Chd8_03 and control AAV9 viral particles (5 × 10⁴ MOI), maintained in culture, and checked for robust GFP reporter expression before collection for protein and RNA extractions at DIV 13.

Husbandry, zebrafish lines, SINEUP-RNAs, morpholinos injection, eye distance measurements, whole-mount staining, and imaging

Zebrafish (Danio rerio) strains, were maintained in a semi recirculating water system (Techniplast) at pH 7.5 ± 0.5, conductivity 200–600 μS, and temperature 27.5°C ± 1°C, on a 14/10-h light/dark cycle with 30 min sunset/dusk effect. Tü, AB/Tü, or the AB strains (obtained from the European Zebrafish Resource Center) were used as wild-type control lines for this study. The mutant line chd8^sa19827, carrying the mutation c.C667T (p.Glu223∗), was obtained from the European Zebrafish Resource Center (EZRC, no. 24433). This 50% AB/50% TL mutant line was then crossed several times with AB strain from EZRC to obtain the chd8 mutation under a pure AB background. Zebrafish were fed with dried food (SAFE) and freshly hatched Artemia salina larvae (Ocean Nutrition). Embryos were raised in E3 medium at 28.5°C, under constant darkness. Developmental stages and genotypes of zebrafish larvae are indicated in the text and figures. Zebrafish embryos and larvae were used within 5 dpf in this study. Zebrafish larvae at this stage are not independently fed and are thus not subjected to Italian legislation (D.Lgs. no. 26/2014). For zebrafish larvae, both males and females were used since the sex can only be determined at age 2 months. All experiments performed with the stable chd8^sa19827 mutant line were carried out according to the guidelines of the Ethics Committee of IGBMC and ethical approval was obtained from the French Ministry of Higher Education and Research under APAFIS no. 15025-2018041616344504. To genotype the mutant line chd8^sa19827, samples were digested in 50 mM NaOH at 95°C, neutralized using 1 M of Tris-HCL (pH 7), and subjected to PCR, primers used: 5′-GTCAGACTCAAGTGCTGCAG-3′ and 5′-GACACTTTGGTCGGAT-3ʹ. PCR products were digested using RsaI restriction enzyme to discriminate heterozygous chd8^{+/p.Glu223∗} mutants from chd8^+/+ wild types since the sa19827 mutation leads to the disruption of the RsaI restriction site.

Two different antisense MOs were obtained from Gene Tools.^1,14 chd8_MO3 (5′-GAGAATGGAATCATAACTTACTTGA-3′) and chd8_MO4 (5′-GCAAATGTGCAAGCAAGTAACACCT-3′) were directed against the splice site of exons 7 and 8, respectively (Figures 4A and S6B). Scrambled_MO (length 25 nucleotides) was used as a standard control, since it does not recognize any genetic sequence in Danio rerio. Morpholinos were stocked in aliquots with a concentration of 20 μg. Before use, they were diluted in phenol red and H₂O to different final doses (8 and 4 ng) to be tested for efficacy and toxicity.

RNAs of SINEUP-chd8_004 and SINEUP-chd8_005 were in vitro transcribed from pCS2+ plasmids using the SP6 mMessage mMachine kit (Ambion, cat. no. AM1340) according to the manufacturer’s instructions. Briefly, five reactions with 1 μg of NotI-linearized pCS2+ plasmid carrying SINEUP sequences were used for in vitro transcription. After purification using Micron Centrifugal filter device (Millipore, cat. MRCF0R030), ∼100 or 200 pg of SINEUP_chd8 molecules were injected into the yolk of fertilized eggs at one- to four-cell stage (MO experiments) or injected into the cell at one-cell stage of fertilized stable chd8 mutant eggs. Morpholinos alone, SINEUP-chd8_004/005 RNA, or a combination of both were microinjected using the Eppendorf Femtojet Microinjector. Embryos (2 dpf) or 4.2 dpf larvae were fixed with PFA 4% overnight at 4°C for PHH3 staining or head size measurements and HuC/D staining, respectively (Figure 4B). For head size measurement, PFA-fixed larvae were washed 3 times for 5 min with PBS 1×, bright-field dorsal views were acquired and eye distance was measured. For PHH3 and HuC/D whole-mount immunostainings, PFA-fixed 2 dpf embryos (PHH3 staining) or 4.2 dpf larvae (HuC/D staining) were stored in 100% methanol at −20°C for at least 2 h. Samples were then rehydrated in increasing amounts of PBS 1×. PFA-fixed embryos were washed in IF buffer (0.1% Tween 20, 1% BSA in PBS 1×) for 10 min and then incubated in the blocking buffer (10% FBS, 1% BSA in PBS 1×) for 1 h. After two washes of 10 min in IF buffer, embryos were incubated with 1:750 anti-phospho-histone H3 (ser10)-R (sc-8656-R, Santa Cruz) or 1:1,000 anti-HuC/D (A21271, Invitrogen) diluted in blocking solution, and incubated overnight at 4°C. After two washes in IF buffer for 10 min each, embryos were incubated in the secondary antibody solution, 1:1,000 Alexa Fluor donkey anti-rabbit IgG and Alexa Fluor goat anti-mouse IgG (A21207, A11001, Invitrogen), in blocking solution, for 1 h at room temperature. After three washes in PBS-Tween 0.1%, lateral views of the head (for PHH3) and the intestine (for HuC/D) were acquired using the MacroFluo Orca Flash (Leica) system. Head sizes were measured by the distance between the eyes of the 4.2 dpf zebrafish larvae using ImageJ or Photoshop software. Numbers of stained cells were counted in defined areas of the head for PHH3 or along the intestine for HuC/D using ImageJ software.

RT-qPCR

Total RNA was extracted from cells and whole embryos of zebrafish using TRIzol reagent (Invitrogen) following the manufacturer’s instructions. To remove DNA contamination, RNA was treated with DNase I (Invitrogen) and RNase inhibitor SUPERase (Invitrogen), and finally purified with an RNeasy Mini Kit (QIAGEN) following the manufacturer’s instructions. The extracted RNA was retro-transcribed using an iScript cDNA Synthesis Kit (Bio-Rad) following the manufacturer’s instructions. For SNP characterization in chd8 gene 5′ UTR, DNase I-treated RNA was retro-transcribed as reported previously and amplified using green master-mix (Invitrogen) with the chd8_Fw1 (5′-CACTGGATATCACTCTTTCTTTGC-3′) and chd8_Rv (5′-GTGGTGTGTCATCAAAGAGGTC-3′) primers. Purified amplicon was then subjected to Sanger sequencing. For RT-qPCR experiments, 1 μg of RNA was retro-transcribed, cDNA was diluted 1:15 and analyzed with the iTaq Universal SYBR Green Supermix (Bio-Rad) protocol. Primers for EGFP and SINEUP were described previously.³⁰ Human NONO and TBP were used as housekeeping genes (HKG). The sequences of all primers used in this study are included in Table S3. Relative mRNA expression level was calculated with the ΔΔCt method and the fold change presented as 2^–ΔΔCt.⁷⁷

Genome-wide transcriptomic analyses and RNA-seq

Raw reads obtained from Sugathan et al.¹⁴ for corresponding samples (Sh-GFP, Sh-GFP2, and Sh4-CHD8) were used to calculate DEGs on GEN-CODE v.26 transcripts.

RNA-seq libraries from Sh-GFP, Sh4-CHD8, Sh4-CHD8 infected with pAIB-Empty, with pAIB-SINEUP-CHD8_001 and pAIB-SINEUP-CHD8_003 were obtained from total RNA isolated using TRIzol (Invitrogen) and treated with DNase (Ambion). RNA quality was assessed by TapeStation (RNA integrity number equivalent from 9.9 to 10). A stranded Illumina library was prepared according to the Illumina Stranded mRNA Prep manufacturer’s protocol starting from 100 ng of purified total RNA. Samples were barcoded using the IDT for Illumina RNA UD Indexes, Ligation kit (Illumina). The library quality was checked using the DNA 1000 Kit and the 2100 Bioanalyzer instrument (Agilent Technologies). The 100-bp paired-end sequencing was performed using the Novaseq 6000 System (Illumina) at the Genomics Facility at the Italian Institute of Technology (IIT, Genova, Italy), obtaining ∼50–60 million reads per sample. DEGs were identified using the DESeq2 R package.⁷⁸ The ∗.readsPerGene files produced by STAR were used as input, excluding all genes with less than three counts per sample from the analysis.⁷⁸ Pathways analysis was performed using the Enrichr gene set enrichment analysis tool (Enrichr enrichment tool: https://maayanlab.cloud/Enrichr/).^54,55,56

Protein extraction and WB analysis

Total proteins were extracted from cells using RIPA Buffer (Sigma) supplemented with halt protease and phosphatase inhibitor cocktail (Life Technology). Samples were sonicated using Q700 (Qsonica) instruments and centrifuged at 12,000 × g for 20 min at 4°C to remove DNA pellet. Proteins were quantified by Pierce BCA Protein Assay Kit (Life Technology). For WB, protein samples were separated by 4%–12% Bis-Tris Protein Gels (ThermoFisher) and transferred on Amersham Protran 0.45-μm nitrocellulose (GE Healthcare) membrane. Membranes were blocked with 5% w/v not-fat dried milk (NFDM) and incubated with the following primary antibodies: anti-GFP, 1:20,000 (Diatech, cat. no. 632380), anti-β-tubulin, 1:10,000 (Santa Cruz, cat. no. 53140), anti-CHD8/Chd8 1:1,000 (Novus Biologicals, cat. nos. 10060417 or A301-224A), anti-HSP90, 1:5,000 (Cell Signaling Technology, cat. no. 4874). Proteins were detected using horseradish peroxidase-conjugated secondary antibodies anti-mouse IgG 1:10,000 (Euroclone) or anti-rabbit IgG 1:10,000 (Euroclone), and revealed with ECL Select WB detection reagent (GE Healthcare). Signal quantification was performed with ImageJ-lab software (Bio-Rad).

Histone extraction and WB analysis

To quantify H3K36me3, cells were resuspended in hypotonic buffer (10 mM HEPES [pH 8], 10 mM KCl, 0.1 mM MgCl, 0.1 mM DTT, 0.1 mM EDTA) supplemented with halt protease and phosphatase inhibitor cocktail (Life Technologies). After a brief incubation on ice, cells were centrifuged at 5,000 rpm for 10 min at 4°C to remove the supernatant-containing cytosolic fraction. Nuclei in the pellet were resuspended in HCl 0.2 N, put in rotation overnight at 4°C, and centrifuged at 4,000 rpm for 10 min. Supernatant was collected and proteins were quantified using Bradford assay (Sigma). For WB, protein samples were separated by 15% Tris-glycine acrylamide gel and transferred on to Amersham Protran 0.45-μm PVDF (GE Healthcare) membrane. Membranes were blocked with 5% w/v NFDM and incubated with the following primary antibodies: H3K36me3 (Abcam, no. AB9050) and H3 (Cell Signaling Technology, no. 4499S) diluted 1:2,000 in 5% NFDM. Proteins were detected using horseradish peroxidase-conjugated secondary antibody anti-rabbit IgG 1:10,000 (Euroclone), and image acquisition and analysis were performed as described previously.

Nano-LC-MS/MS for H3K36me3 analysis

Histone extracts were precipitated with 33% trichloroacetic acid overnight and dissolved in 40 μL 100 mM ammonium bicarbonate (NH₄HCO₃). The samples were then subjected to chemical derivatization: briefly, the propionylation mix was freshly prepared by mixing 4 μL of propionic anhydride with 10 μL of acetonitrile and added to the histone samples to derivatized histone lysine side chains. The pH was adjusted with 8 μL of ammonium hydroxide to prevent acidification of the reaction solution. After two rounds of histone propionylation, histones were resuspended in 100 mM NH₄HCO₃, digested overnight at 37°C with trypsin (enzyme/sample ratio of 1:20), and vacuum centrifuged to quench digestion. Histone peptides were resuspended in 100 mM ammonium bicarbonate and the derivatization procedure was repeated to cap peptide N-termini. After drying in a SpeedVac, the samples were desalted by C18 stage-tip, lyophilized, and resuspended in 15 μL of 0.1% formic acid for LC-MS/MS analysis. Peptide mixture were analyzed using an EASY nLC 1200n HPLC system (Thermo Scientific) and separated on a 30-cm C18 reversed-phase column (inner diameter 75 μm, 1.7 μm particle size, MSWIL, the Netherlands) using a two-component mobile phase (A = 0.1% formic acid in water, and B = 0.1% formic acid in acetonitrile). The gradient was as follows: from 2% to 34% B over 45 min, from 34% to 90% B in 5 min, and 90% B for 10 min at a flow rate of 200 nL/min.

The Orbitrap Fusion electrospray ionization source mass spectrometer (Thermo Scientific) was operated in parallel reaction monitor (PRM) mode. The trimethylated form of the KSAPSTGGVKKPHR (H3.3_27–40) peptide was targeted in the PRM assay with the following modifications: propionyl[N-term]+112.052, trimethyl[K36]+42.047 (me3), and propionyl[K37]+56.026 (un). The 830.4807 (z = 2) precursor ion was selected in the quadrupole, fragmented in the higher-energy collisional dissociation cell (HCD, collision energy of 30%), and the resulting fragment ions were analyzed in the Orbitrap mass analyzer at high resolution (120,000 FWHM, mass range of 150–1500, AGC 5 × 10e5). Full MS/MS spectra were acquired for the selected precursor, allowing simultaneous monitoring of all fragment ions. Extracted ion chromatograms for all transitions were inspected using the Xcalibur Qual Browser (Thermo Scientific). QCloud⁷⁹ was used for all acquisitions to control long-term instrumental performance during the project, using quality control standards.

Data processing and quantification were performed using Skyline software (v.24.1, MacCoss Lab Software), where the area under the curve (AUC) of the fragment ions corresponding to the five most intense fragment ions from the peptide were extracted and quantified. Normalization of the AUC by the total ion current was performed to account for variations in sampling volumes across conditions.

Statistical analysis

Statistical analyses were performed using Microsoft Excel 2016 and GraphPad Prism10. When applicable, D′Agostino-Pearson normality testing was used to evaluate data distribution, means, and standard deviations of the data. Parametric data were analyzed for significance using unpaired two-tailed t tests, one-way ANOVA with Sidak’s post-hoc multiple testing correction, using ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001 for significance, and presented as bar graphs, scatter dot plots, and mean ± standard deviation (SD) or standard error of the mean (SEM) on n ≥ 3 replicates. For zebrafish morpholinos experiments, the Fisher test was used. For datasets comparing conditions with non-normal distributions (e.g., CHD8 protein quantification, Figure 1D) Kruskal-Wallis nonparametric test was used, followed by Dunn’s multiple testing correction.

Data and code availability

All RNA-seq datasets were submitted to the GEO repository and are fully accessible to readers under the accession GSE281179. The MS proteomics data were deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD056482 and 10.6019/PXD056482.

Acknowledgments

We are indebted to the Biagioli, Gustincich, Espinoza, and Golzio lab members for insightful scientific discussions. This work was supported by Department CIBIO Institutional funding to M.B. and S.C., the Young Investigator award by the Brain and Behavior Research Foundation (BBRF) to M.B. M.B., S.E., and S.G. are recipients of the Simons Foundation Autism Research Initiative-2022 Genomics of ASD: Pathways to Biological Convergence and Genetic Therapies no. 1018191. Y.B. was supported by the University of Trento Strategic Project TRAIN (Trentino Autism Initiative: https://projects.unitn.it/train). The MS and Proteomics core facility of CIBIO is supported by the European Regional Development Fund (ERDF) 2014-2020. We are grateful to the Genomics Facility, Istituto Italiano di Tecnologia (IIT), Genova, Italy for RNA-seq libraries preparation and sequencing. This work was funded by Agence Nationale de la Recherche under the projects (JCJC-ANR-17-CE12-0006) (to C.G.). This work of the Interdisciplinary Thematic Institute IMCBio+, as part of the ITI 2021–2028 program of the University of Strasbourg, CNRS and Inserm, was supported by IdEx Unistra (ANR-10-IDEX-0002), and by SFRI-STRAT’US project (ANR-20-SFRI-0012) and EUR IMCBio (ANR-17-EURE-0023) under the framework of the France 2030 Program. C.G. is a permanent INSERM investigator. M.V.L. is a doctoral fellow supported by EUR IMCBio funds. M.M. is a doctoral fellow supported by a PhD fellowship from Fondation ARD/Fondation de France. We thank the IGBMC Zebrafish Facility, in particular Sandrine Geschier.

Author contributions

F.D.L., M.A., J.M., S.G., S.E., and M.B. conceptualized and designed the study. F.D.L., M.A., and S.S. performed in vitro SINEUP experiments. F.D.L., M.E.C., and A.M. performed the in vivo experiments on MO-treated zebrafish embryos. M.M., M.V.L., and C.G. performed zebrafish experiments on the mutant stable chd8 line. S.C. supervised the zebrafish work. C.B. performed the mouse SINEUP work. G.Z. assisted with the overexpression, transcript, and protein quantification experiments. M.P. performed RNA-seq data analysis. E.D. supervised and assisted with RNA-seq data analysis. R.B. provided human fibroblasts with CHD8 disruptive mutations. G.R. provided AAV9 SINEUP-CHD8 viral particles. S.Z. was instrumental for the initial design, cloning, and assistance with the project start. S.Z. deceased in October 2019. Y.B. and S.C. contributed expertise and helpful discussion. F.D.L., M.A., S.S., J.M., and M.B. wrote the article. M.B., S.E., J.M., and S.G. supervised the project. All authors revised, read, and approved the submitted version.

Declaration of interests

M.B., S.G., F.D.L., and M.A. are named inventors in a patent issued by the Italian patent office no. 102022000011546 deposited on May 31, 2022, and PCT/IT2023/050128 deposited on 5 May, 2023 (WO2023233437A1).