Cell-type-resolved NRXN1 isoforms across human brain tissues and hiPSC organoids

Cell-type-resolved NRXN1 isoforms in human brain and hiPSC cortical organoids | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Cell-type-resolved NRXN1 isoforms in human brain and hiPSC cortical organoids Lei Cao , Yu Fan , Sadaf Ghorbani , Jessica Mariani , Yanchun Zhang , Michael B. Fernando , Jaroslav Bendl , John Fullard , Susana Isabel Ramos , Edward A. Mead , Nicola A.L. Hall , Gintaras Deikus , Kristin G Beaumont , Bohan Zhu , View ORCID Profile Sai Ma , Robert Sebra , Nadejda Tsankova , Panos Roussos , View ORCID Profile Kristen J. Brennand , Gang Fang doi: https://doi.org/10.1101/2025.11.11.687875 Lei Cao 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yu Fan 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sadaf Ghorbani 2 Department of Psychiatry, Yale School of Medicine , New Haven, CT, 06520 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jessica Mariani 2 Department of Psychiatry, Yale School of Medicine , New Haven, CT, 06520 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Yanchun Zhang 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Michael B. Fernando 3 Nash Family Department of Neuroscience, Icahn School of Medicine at Mount Sinai , New York, NY, 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jaroslav Bendl 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 4 Center for Disease Neurogenomics; & Friedman Brain Institute; & Department of Psychiatry , New York, NY, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site John Fullard 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 4 Center for Disease Neurogenomics; & Friedman Brain Institute; & Department of Psychiatry , New York, NY, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Susana Isabel Ramos 5 Department of Pathology , Molecular, and Cell-Based Medicine, Icahn School of Medicine at Mount Sinai , New York, NY, 10029, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Edward A. Mead 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Nicola A.L. Hall 6 Department of Psychiatry, University of Oxford, Warneford Hospital , Oxford, OX3 7JX, UK ; 7 Oxford Health NHS Foundation Trust, Warneford Hospital , Oxford, OX3 7JX, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site Gintaras Deikus 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kristin G Beaumont 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Bohan Zhu 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sai Ma 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sai Ma Robert Sebra 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Nadejda Tsankova 5 Department of Pathology , Molecular, and Cell-Based Medicine, Icahn School of Medicine at Mount Sinai , New York, NY, 10029, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Panos Roussos 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 4 Center for Disease Neurogenomics; & Friedman Brain Institute; & Department of Psychiatry , New York, NY, USA 8 Center for Precision Medicine and Translational Therapeutics, Mental Illness Research Education and Clinical Center (MIRECC), James J. Peters VA Medical Center , New York, 10468, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kristen J. Brennand 2 Department of Psychiatry, Yale School of Medicine , New Haven, CT, 06520 Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Kristen J. Brennand For correspondence: kristen.brennand{at}yale.edu gang.fang{at}mssm.edu Gang Fang 1 Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai , New York, NY 10029 Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: kristen.brennand{at}yale.edu gang.fang{at}mssm.edu Abstract Full Text Info/History Metrics Supplementary material Preview PDF ABSTRACT NRXN1 undergoes extensive alternative splicing that generates a highly diverse repertoire of isoforms, diversifying protein–protein interactions, shaping synaptic specialization, and contributing to neuropsychiatric disease when disrupted. However, the splicing landscape of NRXN1 across distinct human brain cell types remains poorly defined. Because NRXN1 expression level is relatively low in adult brains and human induced pluripotent stem cell (hiPSC) derived neurons, single-cell long-read cDNA sequencing often provides insufficient coverage to capture its full isoform diversity, particularly across heterogeneous cell populations. To address this gap, we developed an integrative sequencing strategy that combines single-cell transcriptomics, targeted enrichment of NRXN1 transcripts, and long-read sequencing. This approach reveals a comprehensive catalog of cell-type resolved NRXN1 isoforms across adult and fetal human postmortem brains as well as hiPSC-derived cortical organoids. From the adult prefrontal cortex (PFC) region, our analyses reveal distinct splicing programs across interneuron subtypes, pyramidal neurons, and glial lineages. Comparisons between prenatal and adult brains indicate that NRXN1 isoform profiles are established during early development and remain stable throughout neuronal maturation. In hiPSC organoid models, splicing profiles partially mirror both developmental and mature brain patterns, with a subset of splice sites exhibiting cell type – and stage-specific divergence. In the cerebellum of an autism case and in organoids derived from schizophrenia patients, carrying non-recurrent heterozygous NRXN1 deletions, we identify disrupted isoform expression and mutant NRXN1α isoforms and characterize their distribution across diverse cell types. Using this integrated long-read sequencing framework in patient-derived organoids, we further show that antisense oligonucleotide (ASO) targeting of mutant splice junctions reduces gain-of-function NRXN1 isoforms and reshapes splicing patterns in a cell type–specific manner, providing a platform for evaluating ASO efficiency. Together, these findings help understand the cell type–specific splicing landscape of NRXN1 and establish a framework for decoding cell type-specific isoform diversity and assessing transcript-targeted therapies for genes with low expression levels. INTRODUCTION Alternative splicing (AS) is a key post-transcriptional process in the human brain that generates diverse mRNA isoforms, contributing to the brain’s functional complexity 1 – 4 . Presynaptic cell-adhesion molecules known as neurexins ( NRXN1-3 ) stand out as canonical examples of AS-driven synaptic diversity, playing central roles in synapse formation, specification, and plasticity 5 – 9 . The NRXN1 gene generates an exceptional number of isoforms through six major alternative splice sites (SS1-6) 10 – 12 , which are essential for fine-tuning synaptic properties and diversification. For instance, alternative splicing at splice site 4 (SS4) is known to reshape ligand binding 13 – 16 , while differences at SS3 in hippocampal GABAergic interneurons, specifically parvalbumin-positive ( PV +) and cholecystokinin-positive ( CCK +) subtypes, have been linked to distinct presynaptic calcium channel usage and modulatory receptors 17 , 18 . Aberrant alternative splicing due to NRXN1 +/− deletions disrupt the excitatory–inhibitory balance by weakening excitatory transmission in glutamatergic neurons and enhancing inhibitory transmission in GABAergic neurons 19 , underscoring how precise isoform regulation is critical for maintaining synaptic function. A complete, cell-type-resolved isoform catalog of NRXN1 remained elusive, particularly due to the technological limitations for detection of the large number of neurexin isoforms and their extensive cellular diversity in the human brain. Methods such as patch pipette-based qPCR 20 and next generation sequencing 20 can examine various insertions at alternatively spliced segments but do not assess the combinatorial nature of splicing events on a single mRNA molecule and lack the throughput to capture the NRXN1 splicing patterns in large populations of diverse cells. Long-read platforms, coupled with single-cell RNA barcoding, represent a major development to study splicing of complex genes such as NRXN1 in a cell type-specific manner. However, due to the low expression level of NRXN1 , these transcriptomic approaches provide insufficient depth to comprehensively assess the large repertoire of NRXN1 isoforms. To overcome these technical barriers, we developed an integrative sequencing strategy that combines (1) probe-based long-read targeted-seq (long-read Capture-seq) 21 , 22 with (2) single-cell/single-nucleus RNA isoform sequencing for in-depth NRXN1 isoform profiling, complemented with (3) an optimized long-read RACE-seq 23 , 24 (Rapid Amplification of cDNA Ends) based enrichment module to focus on specific mutant isoforms of interest. In the core idea of this integrative strategy, single-cell transcriptomics is used to provide cell-type deconvolution, while targeted capture is used to enrich NRXN1 transcripts for in-depth analysis; long read sequencing is then employed to achieve the crucial full-length isoform resolution. The use of both probe-based targeting and RACE-seq (PCR-based) enrichment complement each other in recovering the diversity of NRXN1 isoforms (the strength of probes) and detecting rare or low-abundance mutant isoforms (the strength of RACE-seq). Using this integrative strategy, we mapped a comprehensive catalog of cell-type-specific NRXN1 isoforms across diverse neuronal and glial populations. From the PFC region of the human adult brain, we revealed distinct splicing profiles across GABAergic interneuron, pyramidal neurons and glial lineages. Interneurons derived from the same developmental origin exhibited similar AS patterns, whereas those derived from different origins showed divergent profiles. Analyses of the AS pattern in fetal brain suggest that NRXN1 splicing profiles are defined in early development and remain largely consistent during neural maturation. Using hiPSC cortical organoids derived from patients with NRXN1 +/− deletions and controls, we observed that glial progenitor cells and upper-layer excitatory neurons (Ex-L2/3) are enriched for the mutant isoforms, and we found that the organoids partially recapitulate the NRXN1 splicing patterns of both developing and adult brains. In the postmortem brain tissue from an autistic patient case carrying a heterozygous NRXN1 deletion, we found that the mutant isoforms transcribed from deletion allele were enriched in molecular layer interneurons (MLI1/2) and astroglia cells. These findings provide insights into the complex alternative splicing of NRXN1 across brain cell types, along with a generally applicable framework for decoding cell type-specific isoform diversity for important genes with low gene expression levels. RESULTS An integrated strategy reveals a cell-type resolved catalog of NRXN1 isoforms across human brain and hiPSC cortical organoids Single-cell long-read transcriptomics provides a unique ability to resolve full-length NRXN1 isoforms across cell types, but whole transcriptomic profiling does not provide sufficient depth of NRXN1 due to its very low expression level. From the published human postmortem brain single-nuclei Iso-seq data 25 , NRXN1 reads constituted only 0.09-0.2% of the total reads across different cell types, underscoring the need for a targeted enrichment strategy ( Fig. 1a , upper panel ). To maximize isoform recovery while preserving cell-type resolution, we used a targeted sequencing strategy with two complementary components ( Fig. 1b ). To generate a comprehensive isoform catalog, we designed capture probes tiled across all NRXN1 exons at 1× density, which achieved a ∼36-96-fold enrichment of NRXN1 reads over the non-targeted sequencing method ( Fig. 1a , lower panel ). We built on Tequila-seq, which uses isothermal amplification of capture probes from low-cost oligonucleotide templates 21 to reduce per-reaction costs by two to three orders of magnitude while maintaining accurate transcript quantification across gene panels of varying sizes. To recover cell-type specificity for isoforms of interest, particularly the mutant isoforms transcribed from the allele carrying exonic deletions, we adapted the RACE-seq (Rapid Amplification of cDNA Ends) method from short-read and optimized for long-read sequencing to achieve a more sensitive detection of low-abundance isoforms. To link isoforms to specific cell types, we matched the UMIs and cell barcodes on the NRXN1 reads (from NRXN1 -targeted long-read sequencing data) with the UMIs from the cell cluster information by short-read single-cell RNA-seq data ( Methods ). This design leveraged the sequencing depth and accuracy of short reads together with full-length isoform resolution by long reads, enabling precise assignment of complex NRXN1 isoforms to distinct cell types. Download figure Open in new tab Figure 1. A targeting strategy reveals cell-type catalog of NRXN1 isoforms. (a) Upper panel Bar graph showing the percentage of NRXN1 reads detected by long-read Capture-seq (orange) and non-targeted iso-seq (blue) across major brain cell classes. GABA, GABAergic neurons; Glu, glutamatergic neurons; Astro, astrocytes; Micro, microglia; Oligo, oligodendrocytes; OPC, oligodendrocyte progenitor cells. Lower panel: Fold change of NRXN1 -mapped reads obtained by long-read Capture-seq relative to the non-targeted single-cell Iso-seq. (b) Schematic illustrating the comprehensive characterization of cell-type-specific NRXN1 isoforms using the targeting strategy. Full-length single-cell/nuclei cDNA is divided into two parts: one processed for a short-read library using Tn5 transposase and sequenced to identify cell types; the other subjected to probe-based long-read Capture-seq and RACE-seq to detect NRXN1 isoforms with matched barcodes and UMIs (see Methods). (c) NRXN1 isoforms identified by single-cell long-read Capture-seq across samples. Each row represents a unique isoform, categorized as α, α-mutant, β, γ, other known, or novel. Ensembl-annotated isoforms are labeled with ‘ENST’ IDs. The heatmap shows presence (blue) or absence (white) across adult prefrontal cortex, fetal neocortex, and human cortical organoids. (d) PCA biplot of NRXN1 splicing profiles in GABAergic and glutamatergic neurons across samples. Principal component analysis of mean inclusion fractions for exons at 6 splicing sites (SS1-6). Arrows indicate SS event loadings with arrow direction showing which sample groups have higher inclusion of corresponding events. The arrow length reflects the magnitude of contribution to variance. A schematic of NRXN1 gene structure is shown, with splice sites (SS1–SS6) and alternatively spliced exons at each site highlighted in blue. We applied this workflow to single-cell or single-nucleus cDNA from multiple sources, including the PFC of adult brains, gestational cortical samples and cerebellar tissue from an NRXN1 +/– autism patient. In addition, cortical organoids derived from schizophrenia patients harboring unique NRXN1 +/– deletions were profiled to assess the impact of these variants on splicing regulation. After isoform identification and quality filtering, using long-read Capture-seq, we identified 50 distinct NRXN1 isoforms across all datasets, including 27 previously annotated (five α, five β, five γ, and twelve known others) and 23 novel isoforms not represented in Ensembl 26 ( Extended Fig. 1a ). Cell barcode matching revealed 25 annotated and 19 novel isoforms expressed in glutamatergic and GABAergic neurons across samples ( Fig. 1c ). Expression of these novel isoforms was validated across our in-house datasets and confirmed using independent long-read transcriptomes from postmortem human brains reported by others 25 , 27 using miniQuant-based quantitative analysis 28 ( Extended Fig. 1b ). Among all the NRXN1 isoforms, alpha isoforms constitute the largest and most structurally complex class. We applied long-read RACE-seq, a more sensitive method to single-cell or single-nucleus cDNA of all samples and identified 103 alpha isoforms including 57 that have been reported previously ( Extended Fig. 1c ). We compared exon inclusion across NRXN1 splice sites (SS1–SS6) in GABAergic and excitatory neurons across all samples and performed PCA. Exons 3 and 4 at SS1 and exon 7 at SS2 show consistently higher inclusion in glutamatergic neurons, whereas exon 21 at SS4 is enriched in GABAergic neurons in adult PFC. Notably, GABAergic neurons in cortical organoids display a distinct splicing profile driven by exon 23 at SS5 ( Fig. 1d ) . NRXN1 exhibits distinct splicing patterns across cell types in adult human brains To understand NRXN1 alternative splicing patterns across mature neurons in the human brain, we applied the probe-based long-read Capture-seq on the single-nucleus cDNA from the PFC region of 61 adult donors 27 ( Fig. 2a, b ; Extended Fig. 2 ). We generated 2.43M NRXN1 reads, among which 1.13M had 10x single cell barcodes, and 1.03M reads sharing the barcodes with short-read data. By integrating the barcode information from short– and long-read data, a total of 125,199 cells were identified across all donors. From the long-read Capture-seq of the single nuclei cDNA pools, we identified 46 isoforms with confident cell barcodes and UMIs matched to short-read data, including 6 alpha isoforms not annotated but reported previously, and 14 isoforms not annotated in Ensembl 26 but later validated by our in-house brain dataset ( Fig. 1c , Extended Fig. 1 ). Download figure Open in new tab Figure 2. NRXN1 expression repertoires exhibit distinct splicing patterns across cell types in human brains. (a) Workflow overview of long-read Capture-seq applied to single-nucleus cDNA from adult PFC. (b) UMAP visualization of putative cell types in the adult PFC. Colored clusters represent neuronal and glial subtypes. (c) Normalized NRXN1 splicing events at six splicing sites (SS1–SS6) in GABAergic interneurons (top), glutamatergic excitatory neurons (middle), and glial lineages (bottom). SS1–3 represents exon 3 inclusion/ exclusion at splicing site, similar for SS1-4, SS2-7, etc. Each bar represents a cell class. Within GABAergic interneurons, PROX1 + - INs are colored in green, LHX6 + - INs in red, and LAMP5⁺ LHX6 +- INs in orange. Exon inclusion (“splice-in”) is shown upward and exclusion (“splice-out”) downward. Data represent mean ± SEM. A schematic of NRXN1 gene structure is shown, with splice sites (SS1–SS6) and alternatively spliced exons at each site highlighted in blue. (d) PCA plot based on NRXN1 splicing patterns across splice sites SS1–SS6, showing that PROX1⁺ –INs, LHX6⁺ –INs, and PCs form distinct, separately clustered classes. Colors correspond to cell types as in (c): PROX1⁺- INs (green), LHX6⁺ -INs (red), LAMP5⁺ LHX6 + -GABAergic neurons (orange), glutamatergic neurons (blue), and glial lineages (gray). (e) The difference in mean exon inclusion fraction between GABA and Glu neuronal cell types (Δ = mean − mean) at 6 splicing sites. Positive values indicate higher inclusion in GABA neurons; negative values indicate higher inclusion in Glu neurons. Significance labels denote BH-adjusted p -values from Welch’s t-tests (** p < 0.01; * p < 0.05; ns, not significant). The brain comprises a great diversity of cell types, and studying neurons with shared developmental markers offers a means to uncover how NRXN1 alternative splicing contributes to this cellular and functional diversity. Based on developmental markers 27 , neurons were categorized as Neurod2-expressing pyramidal cells 28 (PC), LHX6 -expressing interneurons 29 ( LHX6 + -INs originated from medial ganglionic eminence, MGE), PROX1 -expressing interneurons 30 ( PROX1 + -INs originated from caudal ganglionic eminence, CGE), and LAMP5 + LHX6 + GABAergic neurons ( PROX1 + labeled as LAMP5 ). LHX6 + -INs can be further subclassified into PVALB + and SST + INs, while PROX1 + -INs are divided into LAMP5 + , SNCG + , and VIP + INs. GABAergic INs are a highly heterogeneous collection of cell types, characterized by their unique spatial and temporal capabilities to influence neuronal circuits 31 – 33 . We observed that GABAergic interneurons showed the greatest NRXN1 splicing diversity ( Fig. 2c , top ). Splicing patterns were more variable at SS1, SS3, and SS4, whereas SS2 and SS6 were invariably excluded and SS5 was consistently included. Within the PROX1 + -INs, the LAMP5 + , SNCG + and VIP + subtypes share similar patterns across all 6 alternative splicing sites, while PVALB + and SST + , subclasses of LHX6 + -INs, differed at SS1. We also identified a LAMP5 + LHX6 + subclass that shared splicing patterns with the PROX1 + -INs and SST + of LHX6 + – INs ( Fig. 2c ). In PC neurons ( Fig. 2c , middle ), we found: (1) across cortical layers, exons at SS1, SS3 and SS5 were dominantly spliced in (“in”; upward bars), while exons at ASS6 and SS4 are dominantly spliced out; (2) layer-specific differences at SS2, where exon 7 was often spliced out in PCs of layers L2/3-IT, L5-IT, and L6-IT-Car3, but spliced-in into most PCs of layers L4-IT, L5/6-NP, L6B, L6-CT, and L6-IT. Among glial cells (astrocytes, oligodendrocyte progenitor cells (OPCs), oligodendrocytes), exon 17 at SS6 was consistently spliced out, whereas exon 23 at SS5 was spliced in. Oligodendrocytes showed spliced-in for exon1 and exon5 at SS1, which were dominantly spliced-out in astrocytes and OPCs. At ASS4, exon 21 was included in astrocytes but not in OPCs or oligodendrocytes, consistent with prior studies showing that astrocytic neurexins splice SS4 at high ratios compared to excitatory neuronal neurexins 34 , 35 . Building on the striking patterns of NRXN1 alternative splicing observed in the human brain, we conducted principal component analysis (PCA) of splicing levels across cell types, which separately clustered PROX1 + -INs, LHX6 + -INs and PCs classes ( Fig. 2d ). To identify NRXN1 splicing events distinguishing glutamatergic (Glu; n = 9) and GABAergic (GABA; n = 4) neurons, we compared mean exon inclusion across 6 splice-site events using Welch’s t-tests with Benjamini–Hochberg correction. SS2-7 and SS3-12 exhibited higher inclusion in Glu neurons, whereas SS4-21 was preferentially included in GABA neurons ( Fig. 2e ). Overall, these findings suggest that NRXN1 splicing profiles exhibit substantial differences across various neuronal populations and glial lineages. Within GABAergic neurons, interneurons derived from the CGE ( PROX1 + -INs subtypes) are closely clustered by the splicing patterns, yet different from INs originating from the MGE ( LHX6 + -INs subtypes). NRXN1 splicing profiles are conserved between fetal and adult neurons within defined neurogenic lineages To determine whether the NRXN1 splicing patterns of the adult brain are defined as early as in the human developing neocortex, we generated a snRNA-seq dataset from two unfixed snap-frozen postmortem neocortex samples obtained from the second and third trimesters of gestation (24 and 33 gestational weeks, gw). After quality control, we obtained the transcriptomic profile of 52,426 nuclei, with a median of 2,910 genes, 6,156 UMIs from each nucleus. We identified 10 primary cell types from the single cell transcriptome: five neuronal clusters, including excitatory neuronal subpopulations expressing layer-specific cortical marker genes (EX-L2/3, EX-L4/5/6); inhibitory neuronal subpopulations (IN-CGE, IN-MGE and IN-Reelin); and glial cell populations, radial glia and astrocyte (RG/AC) clusters, astrocytes, OPCs, transient glial intermediate progenitor cell (gIPC) population, microglia and FLT1 + blood vessel cells (BVCs) 36 ( Fig. 3a ). We further validated our cell annotations by comparing top differentially expressed genes (DEGs) per cell cluster to published canonical lineage and proliferation markers 36 , 37 ( Fig. 3c , Supplementary Table 3 ). Next, we analyzed how cell type proportion varied between prenatal neocortex and adult PFC. The 33 gw neocortex sample had more glial, INs-MGE and INs-CGE, whereas the 23 gw sample had more INs-Reelin and excitatory neurons EX-L4/5/6 ( Fig. 3b ). Neuronal populations constituted most cells in the fetal neocortex, whereas glial transient intermediate progenitor cells were exclusively present in fetal neocortex and subsequently disappeared or transformed along with the cortical development in adult PFC, which is consistent with the stages of neurogenesis in human cerebral cortex 38 , 39 ( Extended Fig. 3 ). Download figure Open in new tab Figure 3. NRXN1 splicing profiles remain consistent across cells sharing a neurogenic origin during neural maturation. (a) UMAP visualization of putative cell types in the neocortex from two fetal brains. Left: merged UMAP of combined datasets; Right: separate UMAPs showing neocortex samples from 23 gw and 33 gw. Colored clusters represent neuronal and glial subtypes in the fetal neocortex. Abbreviations: gIPC, glutamatergic intermediate progenitor cells; RG/AC, radial glia/astrocyte lineage; Ex, excitatory neurons; IN, inhibitory neurons; BVC, blood vessel cells; MG, microglia. (b) The proportions of each cell class in 23 gw and 33 gw neocortex samples. Colored blocks indicate neuronal and glial subtypes. (c) Heatmap of marker genes expression used for cell-type annotation. Each row represents a marker gene and each column a cell class; darker red indicates higher expression. Colored blocks indicate neuronal and glial subtypes in fetal neocortex. (d) Normalized NRXN1 splicing events at SS1–SS6 in PROX1 ⁺-INs (green), LHX6 ⁺-INs (red), glutamatergic excitatory neurons (blue), and glial lineages (gray). A schematic of NRXN1 gene structure is shown, with splice sites (SS1–SS6) and alternatively spliced exons at each site highlighted in blue. Exon inclusion (“splice-in”) is shown upward and exclusion (“splice-out”) downward. Data represent mean ± SEM. (e) Pearson’s correlation of NRXN1 exon inclusion levels across splicing sites (SS1–SS6) between adult PFC and prenatal neocortex in LHX6 ⁺-INs (left), PROX1 ⁺-INs(middle), and PC neurons (right). To determine NRXN1 splicing patterns by cell type in the fetal neocortex, we applied the probe-based long-read Capture-seq on the single nuclei cDNA. Based on regional marker genes, exon 23 at SS5 was predominantly spliced-in across all cell types, whereas exons at other splice sites showed variable inclusion. In most (>60%) PCs across all layers, exons at SS1 and SS3 were consistently spliced in, exon 21 at SS4 was spliced out, and exon 17 at SS6 showed variable usage. In contrast, exon 7 at SS2 was differentially spliced between upper-layer (L2/3) and deeper-layer (L4/5/6) PCs. Similarly, in GABAergic interneurons (INs), exons at SS1 and SS3 were largely spliced in, while exons at SS2, SS4, and SS6 were spliced out. Notably, the proportion of IN- LHX6⁺ cells was higher than that of IN- PROX1 ⁺ cells at SS3 (spliced in) and SS4 (spliced out). The glial cells were variably spliced at SS1, SS2, SS6 and SS4, but consistently spliced-in at SS3 and SS5 ( Fig. 3d ). We next compared NRXN1 splicing in matched neuronal populations between prenatal neocortex and adult PFC. Exon inclusion level across 6 splicing sites were highly correlated between prenatal neocortex and adult PFC for LHX6 + -INs ( Pearson’s r = 0.82, p = 0.01), PROX1 + -INs ( Pearson’s r = 0.87, p = 0.005), PCs ( Pearson’s r = 0.87, p = 0.005) and glial populations (OPC, Pearson’s r = 0.9, p = 0.003; RG/AC, Pearson’s r = 0.82, p = 0.01) ( Fig. 3e , Extended Fig. 3 ). These findings suggest that NRXN1 splicing profiles are defined in early development and remain largely consistent during neural maturation. NRXN1⁺/⁻ splicing profiles in human cortical organoids recapitulate developmental and mature features and identify cell types expressing the mutant isoforms From hiPSC-derived two-dimensional (2D) neurons with a 3′- NRXN1 deletion, we previously uncovered dozens of mutant isoforms exclusively from the deletion-carrying allele (mainly affecting alpha and beta isoforms) 12 . Later, we demonstrated that this exonic deletions at the 3’ site disrupted normal NRXN1 splicing and isoform expression, leading to distinct phenotypes in induced glutamatergic (iGLUTs) and GABAergic (iGABAs) neurons, primarily reflected in altered synaptic frequency 19 , 40 . Building on these observations, we sought to explore how the deletions affect the splicing events across cell types in a more complex neurodevelopmental system. We generated human cortical organoids (hCOs) from a control line and schizophrenia probands with the 3’deletion (3’-del) in parallel. hCOs were guided toward the forebrain 41 ( Methods, Extended Fig. 4 ). To profile cell types in the hCOs, we performed single cell RNA-seq on hCOs from two 3’-del cases (line 581 and line 641) and one control line (line 690) at day 150 with the 10x NextGEM 5pv2 kit, when neural activity is well characterized 42 – 44 . After quality control data processing steps, we merged high-quality scRNA-seq data from 3 cortical organoid samples into a core dataset of 27,325 cells. Following unsupervised clustering, we identified and annotated major cell types using an extensive curated list of known marker genes 19 , 41 , which consisted of transit-amplifying cell (TAC), neuronal intermediate progenitor cells (nIPCs) giving rise to distinct subpopulations of glutamatergic excitatory neurons (Ex-L2/3, Ex-L4/5/6) and GABAergic inhibitory neurons (INs, IN- SST ,IN-Reelin); gIPCs leading to the subpopulations of radial glia and astrocyte clusters (RG/AC); as well as BVCs ( Fig. 4a ). We further validated our cell annotations by comparing gene expression signatures of our scRNAseq cell clusters to published brain organoid datasets as reference ( Fig. 4b , Supplementary Table 4 ). We then compared cortical organoids derived from the control hiPSC line with the prenatal neocortex in terms of cell-type composition. We found gIPCs (OR = 2.06, Fisher’s test ) and Ex-L2/3 (OR = 1.67, Fisher’s test ) were significantly enriched in hCOs from the 3’-del lines than in control lines, while cell number of IN-Reelin were significantly low in 3’-del lines (OR = 0.03, Fisher’s test ) ( Fig. 4c ). Download figure Open in new tab Figure 4. NRXN1⁺/⁻ human cortical organoids (hCOs) splicing profiles recapitulate developmental and mature features and identify cell types expressing the mutant isoforms. (a) Left: UMAP of putative cell types in hCOs, with colored clusters representing neuronal and glial subtypes. Right: Proportions of each cell class in organoids derived from control line 690 or 3′-del lines (581 and 641). Abbreviations: gIPC, glutamatergic intermediate progenitor cells; nIPC, neuronal intermediate progenitor cells; TAC, transient amplifying cells; RG/AC, radial glia/astrocyte lineage; Ex, excitatory neurons; IN, inhibitory neurons; BVC, blood vessel cells. (b) Heatmap of marker gene expression used for cell-type annotation. Each row represents a marker gene and each column a cell type; darker red indicates higher expression. (c) Bar graph showing enrichment of specific cell types in 3′-del organoids compared with control lines. (d) Normalized NRXN1 splicing events at SS1–SS6 in all cell classes from control hCOs (upper) and two 3′-del hCOs (lower). Each splicing graph is accompanied by a schematic of NRXN1 gene structure with red shading indicating 3′-del–associated regions; alternatively spliced exons at each splice site are highlighted in blue. Exon inclusion is shown upward (“splice-in”), exclusion downward (“splice-out”). Data represent mean ± SEM. (e) Pearson’s correlation of NRXN1 exon inclusion levels across splicing sites (SS1–SS6) between 3′-del and control hCOs. (f) Left: Schematic of NRXN1α isoform structures identified by long-read RACE-seq, along with a schematic of NRXN1 gene structure. Red shading indicates regions associated with the 3′ deletion, and alternatively spliced exons at each splice site are highlighted in blue. Each row represents a unique NRXN1α isoform, with blue exons indicating inclusion and blanks indicating exclusion. Right: Presence or absence of each NRXN1α isoform across 3′-del and control hCOs (green = present; blank = absent). (g) The composition of NRXN1 isoforms identified by long-read RACE-seq within each cell type across 3′-del and the control hCOs. (h) Exon inclusion levels at SS1–SS6 in GABAergic INs (top), glutamatergic excitatory neurons (middle), and RG/AC (bottom) across hCOs (blue), fetal neocortex (orange), and adult PFC (green). Statistical comparisons were performed using one-way ANOVA (*, P < 0.05; **, P < 0.01; ***, P < 0.001).Data represent mean ± SEM. To better understand the greater splicing patterns of NRXN1 across cortical organoid cell types, we applied the long-read Capture-seq on the full-length single-cell cDNA. From the control lines and those with the 3’-deletion, we generated 317,521 NRXN1 reads, of which 144,450 with 10x single cell barcodes and 120,873 reads sharing the barcodes with short-read data. Integrating the barcode information across short– and long-read data, we identified 23,187 cells from all cortical organoid samples ( Extended Fig. 4 ). Using these targeted single cell data, we first analyzed the NRXN1 AS patterns in excitatory neurons, inhibitory neurons and non-neuronal cells for both 3’-del lines and control lines at the advanced maturation stage. Within the glutamatergic excitatory neurons (the major population in hCOs), NRXN1 splicing patterns were highly similar between the control line and 3’del lines, except for differences in exon inclusion level at SS1 (exon 3) and SS3 (exon 12) ( Fig. 4d,e ). The ∼136-kb deletion impacting SS4 and SS5 (exons 21-23) 12 , 19 slightly reduced exon inclusion at SS4 in the 3’-del lines ( Fig. 4e ). As the deletion in the 3’-region of NRXN1 generates mutant alpha and mutant beta isoforms, we sought to dissect which cell types are enriched for these mutant isoforms. As some of the mutant isoforms are low in abundance and the long-read Capture-seq works better with abundant isoforms, we employed long-read 5’RACE-seq to resolve alpha isoforms with single cell barcode information from the 5’GEX single cell cDNA (Methods). We identified 21 α-isoforms across all cortical organoids, of which 18 were not detected by long-read Capture-seq, including six mutant isoforms from the two 3′-del NRXN1 ⁺/⁻ derived organoids ( Fig. 4f ). In addition, we identified seven β-isoforms, including one mutant β-isoform ( Extended Fig. 4 ). Compared with the control line, the mutant α-isoforms were expressed widely across all cell types in both lines with 3′- NRXN1 ⁺/⁻ (except for DLX1/2 -INs in 641). In contrast, the mutant β-isoform was detected in fewer cell clusters, showing consistent expression in excitatory neurons in both 3′- NRXN1 ⁺ /⁻ lines, and exclusively in TACs in line 581 and RG/AC cells in line 641 ( Fig. 4g ). The composition of NRXN1 isoform by cell type demonstrated that mutant isoforms from the allele bearing 3’-deletion are expressed in all cell types, contributing to altered synaptic activity in excitatory neurons and DLX1/2 -INs. With NRXN1 splicing profiles available across multiple developmental stages, we next asked which stage the control hCOs most closely resemble in terms of NRXN1 splicing– the fetal neocortex or the adult PFC. We therefore compared the exon inclusion level across splicing sites (SS1-6) for NRXN1 in three major cell classes that appear in all samples, i.e., GABAergic inhibitory neurons, excitatory neurons, and RG/AC ( Fig. 4h ). In GABAergic neurons, exon inclusion levels were highly correlated between organoids and the fetal neocortex (Pearson’s r = 0.97, p = 3.82E-04 ) and between cortical organoids and the adult PFC (Pearson’s r = 0.88, p = 0.008 ), although exons at SS2, SS3, and SS5 displayed differential splicing. Similarly, significant agreement was observed between cortical organoids and the fetal neocortex in glutamatergic excitatory (Glu-EX; Pearson’s r = 0.73, p = 0.04 ) and radial glia/astrocyte (RG/AC; Pearson’s r = 0.78, p = 0.04 ) populations, and between organoids and the adult PFC in Glu-EX (Pearson’s r = 0.88, p = 0.004 ) and RG/AC (Pearson’s r = 0.92, p = 0.003 ) populations ( Extended Fig. 4 ). Together, these findings indicate that cortical organoid splicing profiles partially recapitulate both developmental and mature splicing patterns, with a subset of splice sites showing cell type – and stage-specific divergence ( Fig. 4h , Extended Fig. 4 ). Mutant NRXN1 isoforms are enriched in MLIs in ASD cerebellum with exonic deletion NRXN1 deletions are also strong risk factors for autism spectrum disorder (ASD) 45 , 46 . The cerebellum shows consistent structural and functional alterations in ASD, and NRXN1α isoform complexity is significantly different in the cerebellum compared to the cerebral cortex. To understand NRXN1 isoform diversity in human brain with NRXN1 heterozygous deletion, we applied our probe-based long-read Capture-seq to the postmortem cerebellum of an ASD patient with NRXN1 +/- (encompassing exon 9-13) ( Extended Fig. 5a ). We performed single-nucleus RNA-seq on this sample and obtained 9,689 high-quality nuclei. Unsupervised clustering and annotation identified eight cerebellar cell type classes: excitatory granule cells; inhibitory neurons (Purkinje cells; two types of molecular layer interneurons MLI1, MLI2; and Purkinje-layer interneurons); astrocytes; Bergmann glia; oligodendrocytes; and OPCs ( Fig. 5a ). These annotations matched expected marker gene expression 47 ( Extended Fig. 5b, Supplementary Table 5 ). Download figure Open in new tab Figure 5. A patient-specific exonic deletion drives cell-type–specific enrichment of mutant NRXN1 isoforms in the ASD cerebellum. (a) Left: UMAP of putative cell types in the cerebellum, with colored clusters representing neuronal and non-neuronal subtypes. Right: Cell proportion of each cell class in the cerebellum of ASD patient ( NRXN1 -del) and the control. (b) Schematic of NRXN1 isoform structures identified by long-read capture-seq, with each row representing a unique NRXN1 isoform. The NRXN1 ⁺/⁻ specific mutant isoforms are highlighted in red. (c) Expression level of NRXN1 isoforms (expression level as TPM) from cerebellum of control or NRXN1 -del. The red box highlights mutant isoforms specific to the deletion-carrying allele. (d) The proportion of cells in each cell class expressing any NRXN1 isoforms, wild-type isoforms( NRXN1 -WT) and mutant isoforms ( NRXN1 -MT). (e) Normalized NRXN1 splicing events at SS1–SS6 for Granule and MLI1/2 cells from NRXN1 +/- ( NRXN1 -del) and control (Br1446) cerebellum. Schematic splice graph of NRXN1 depicting splice sites (SS1–SS6), with red shading marking regions impacted by the exonic deletion allele. Alternatively spliced exons at each splice site are highlighted in blue. Exon inclusion is shown upward (“inclusion”) and exclusion downward (“exclusion”). Data represent mean ± SEM. (f) Aggregate enrichment of GO terms within each biological theme across cell types by down-regulated (blue) and up-regulated (red) DEGs, excluding themes implicated by only one term and nonsignificant cell types. (g) Overrepresentation of synaptic compartment genes [SynGO database 65 ] within cell type–specific DEGs. *P < 0.05, **P < 0.01, ***P <0.001, Hypergeometric test, FDR < 0.05. When applying the probe-based long-read Capture-seq to the single nuclei cDNA of the cerebellum, we identified 9 NRXN1 isoforms with cell barcodes matched to short-read data, including two mutant isoforms associated with the exon 9-13 deletion ( Fig. 5b,c ) . Among NRXN1 reads with cell barcodes matched to short-read data, the mutant isoforms 3312 and 4131 were exclusively detected in NRXN1 -deletion cerebellum, with transcript 3312 being the most abundant and 4131 less abundant, together accounting for 28.34% of reads ( Fig. 5c ). When assigning isoforms to cell types, we found the two mutant isoforms were significantly enriched in MLI1 (Fisher’s Exact test, p =5.72E-18), MLI2 (Fisher’s Exact test, p =0.04), and astroglia cells (Fisher’s Exact test, p =51.49E-13) compared to wild-type NRXN1 isoforms ( Fig. 5d ). The enrichment of mutant isoforms in MLIs, key GABAergic regulators of Purkinje cell output, links these mutations to ASD-associated E/I imbalance, suggesting impaired inhibitory circuitry as a primary site of cerebellar vulnerability 48 – 50 . Next, we examined NRXN1 splicing across six alternative splice sites in these cell types. In the ASD case with NRXN1 deletion, more than 70% of granule cells showed exon inclusion at SS1, SS3, SS4, and SS5, with moderate inclusion at SS2 (>50%) and predominant exon exclusion at SS6. The two MLI subtypes exhibited distinct splicing patterns, particularly at SS1 and SS2. In controls, NRXN1 splicing varied across cell types, most prominently at SS2–SS4. Comparative analysis revealed broadly similar splicing patterns between ASD and control in granule cells and MLI1 cells, with high exon inclusion at SS1, SS3, SS4, and SS5, variable inclusion at SS2, and predominant exclusion at SS6. In contrast, subtype-specific differences were observed in MLI2 cells (SS1 and SS2), PLI cells (SS3), and Purkinje cells (SS4) ( Fig. 5e , Extended Fig. 5c ). To gain insight into how exonic deletion of NRXN1 affects the DEG networks in the cerebellum potentially associated with ASD, we performed differential expression (DE) analysis for each cell type between control and NRXN1 -del (see Materials and Methods for details). we then performed Gene Ontology analysis on the up– and downregulated genes per cell type and grouped significantly enriched terms into 20 biological themes ( Extended Fig. 5d , Supplementary Table 6,7 ). Biological themes related to synapse organization and neuron projection development were associated with DEG sets within excitatory and inhibitory neuronal populations ( Fig. 5f ). Given the prevalence of pathways related to synaptic structure and function, we further assessed enrichment of synaptic compartment genes using the SynGO database. The top-level SynGO category “synapse” was significantly enriched (FDR < 0.05, Hypergeometric test) across all neuronal cell types ( Fig. 5g ). Cell-type–resolved evaluation of ASO inhibition of GOF NRXN1 isoforms in cortical organoids Antisense oligonucleotides (ASOs), recently used to treat several neurological diseases 51 , provide a strategy to selectively target disease-associated mutant isoforms by modulating specific RNA splice sites. In our recent work, we tested ASO efficacy in 2D iGLUT neurons with bulk RNA-seq, but this approach lacked cellular complexity and cell-type resolution 19 . We hypothesized that ASO efficacy and impact are cell-type specific, particularly in inhibitory neuronal populations enriched for mutant isoforms. Here, we applied the long-read single cell targeted-seq approach to evaluate ASO in three-dimensional (3D) human cortical organoids in a cell type–resolved manner. We adopted the same ASO design that target the exon 20–24 mutant splice junction and applied ASO treatment (10 μM) to the organoids derived from 641 line (3’-del) for 96 hours (see Methods; Extended Fig. 6a, b). We collected the ASO-treated organoids for single-cell cDNA synthesis, using half of the single-cell cDNA for NRXN1 isoform profiling with both 5’RACR-seq and long-read Capture-seq, while the other half used for cell type annotation by single-cell RNA-seq. From the NRXN1-targeted long-read sequencing data, we found ASO-Aplus decreased total NRXN1-MT isoforms by approximately 51% in all cells ( Extended Fig. 6b,c ). Differential splicing analysis confirmed a reduction in GOF splicing ( Fig. 6a , ΔPSI (percent spliced index) = −0.23), increased exon23 inclusion at SS5 (ΔPSI = 0.313) and decreased exon 21 inclusion at SS4 (ΔPSI = −0.154), accompanied by a significant shift from α to β isoforms ( Extended Fig. 6c , p < 0.001, Fisher’s exact test). Download figure Open in new tab Figure 6. Evaluation framework for cell-type–resolved ASO inhibition of GOF NRXN1 isoforms in cortical organoids. (a) Left: UMAP of putative cell types in ASO-treated hCOs derived from line 641, with colored clusters representing neuronal and glial subtypes. Right: Proportions of each cell class in line 641 treated with ASO-NT or ASO-Aplus. (b) Splicegraphs displaying the splicing of the MT exon 20→24 cluster in 641 hCOs treated with ASO-Aplus (upper) or ASO-NT(lower), compared via a Dirichlet-multinomial generalized linear model. (c) Cell-type–specific mapping of NRXN1 reads (containing exons 20–24) following ASO-NT or ASO-Aplus treatment. Color coding of cell types corresponds to the scheme used in the UMAP ( Fig. 6a ). (d) Percentage of wildtype or mutant NRXN1 isoforms detected by long-read RACE-seq across cell types in 641 hCOs with ASO treatment. A significant enrichment of the wildtype isoform in nIPCs is observed in the 641-Aplus condition (OR = 0.015, p = 1.53 × 10⁻⁷, BH-adjusted p = 6.13 × 10⁻⁷, Fisher’s Exact test). (e) Normalized NRXN1 splicing events at SS1–SS6 for neuronal (EXs and INs) and non-neuronal (RG/AC, nIPC) cells in 641 hCOs treated by ASO-Aplus treated (left) or ASO-NT (right). Exon inclusion is shown upward (“splice-in”) and exclusion downward (“splice-out”). Data represent mean ± SEM. To understand which cell types are mostly perturbed by ASO targeting of the mutant isoforms in the hCOs, we first analyzed single-cell data from the ASO-treated organoids for cell type annotation ( Fig. 6b , Extended Fig. 6d ). We did not observe significant difference for cell composition between the organoids treated with ASO-NT and ASO-Aplus ( Fig. 6b ). Single-cell RNA-seq profiling revealed robust downregulated DEGs in IN and RG/AC cells after ASO-Aplus treatment, enriched for regulation of translation and RNA splicing reported by previous study 52 ( Extended Fig. 6e,f and Supplementary Table 8 ). Integration with cell type–resolved single-cell RNA-seq data revealed that the proportion of NRXN1 reads containing exons 20–24 was markedly reduced in RG/AC cells and IN cells following ASO-Aplus treatment ( Fig. 6c ). Most importantly, ASO-Aplus significantly reduced proportion of NRXN1 -MT reads in nIPCs ( Fig. 6d ; OR = 0.015, p = 1.53 × 10⁻⁷, BH-adjusted p = 6.13 × 10⁻⁷, Fisher’s Exact test) and showed decreased trend in excitatory neurons, not in RG/AC, indicating cell-type–specific differences in the uptake efficiency of ASOs delivered via gymnosis and/or their knockdown efficacy within each cell type. The activity of ASO-Aplus is consistent with established cell-type differences in ASO efficacy in the central nervous system, where neuronal populations show greater target engagement and higher transfection 53 , 54 . The decreased proportion of RG/AC in ASO-treated conditions without accompanying NRXN1 -MT knockdown in those cells suggests a selective cytotoxic or antiproliferative effect on RG/AC rather than on-target silencing 55 . Splicing analysis revealed that exon 21 (SS4) was reversely spliced across all cell types, whereas exon 23 (SS5) remained consistently included following ASO-Aplus compared to ASO-NT treatment. In RG/AC cells, exons at SS1, SS3, and SS4 also exhibited reversed splicing, while splicing patterns at SS2, SS5, and SS6 were largely unchanged between ASO-Aplus and ASO-NT conditions ( Fig. 6e ). Notably, neither wild-type transcripts containing exons 20–24 nor the mutant were detected in INs ( Fig. 6c ). This likely reflects partial overlap of the ASO-Aplus sequence with exons 20 and 24, resulting in unintended targeting of wild-type NRXN1 in addition to NRXN1 -MT derived from the 3′ deletion allele ( Extended Fig. 6b ). Altogether, we characterized alterations in MT vs WT NRXN1 transcripts and splicing events across cell types following ASO treatment using long-read Capture-seq and long-read RACE-seq in hCOs derived from a 3′-del patient. These results highlight the combined use of long-read Capture-seq and RACE-seq as a powerful toolkit for comprehensive evaluation of ASO-induced changes in transcript repertoires and splicing patterns across diverse cell types. DISCUSSION We applied an integrative isoform sequencing strategy combining long-read Capture-seq and long-read RACE-seq to generate a cell-type–resolved catalog of NRXN1 isoform diversity across human cortical organoids, prenatal cortex, adult PFC, and cerebellum. NRXN1 exhibited pronounced cell type– and subtype-specific splicing patterns. In human cortical organoids and adult PFC, NRXN1 splicing showed clear cell-type specificity, whereas analysis of cerebellar tissue from an ASD case revealed deletion-associated isoforms enriched in molecular-layer interneurons and astroglia. Using this framework, we further demonstrated that ASOs targeting the mutant splice junction promote degradation of mutant isoforms in specific cell types within patient-derived organoids. Together, these results establish a platform for resolving NRXN1 alternative splicing at cell-type resolution and for evaluating ASO-mediated transcript perturbations across cellular contexts. Our strategy integrates probe-based enrichment, single-cell barcoding, and long-read sequencing to overcome key challenges in detecting low-abundance, full-length NRXN1 transcripts. Targeted capture increased on-target recovery by approximately 50-fold compared with untargeted long-read approaches while preserving single-cell resolution. This workflow links full-length isoforms to their cellular identities, enabling quantitative mapping of NRXN1 splicing across neuronal and glial populations and connecting genetic variation to functional transcriptomic phenotypes. Glutamatergic neurons, the predominant cortical population, demonstrated highly concordant of NRXN1 splicing patterns between prenatal neocortex and adult PFC. Likewise, glutamatergic neurons in iPSC-derived cortical organoids recapitulated splicing profiles observed in the prenatal cortex, suggesting that NRXN1 +/– organoids capture key developmental splicing programs. GABAergic interneurons in the PFC exhibited substantial molecular diversity, reflected in distinct NRXN1 splicing patterns. Although PVALB -expressing and SST -expressing interneurons both originate from the MGE ( LHX6 +), they showed subtype-specific exon usage consistent with their known functional divergence 56 , 57 . Notably, NRXN1 splicing programs appear to be established early in development and remain largely stable, as evidenced by the strong concordance of exon inclusion levels in PROX1 + interneurons between the prenatal neocortex and adult PFC 20 . Analysis of an NRXN1 +/– cerebellum from an ASD patient revealed unique mutant isoforms carrying an exon 9–13 deletion that were enriched in molecular-layer interneurons and astroglia, whereas granule cells predominantly expressed wild-type isoforms. Compared to a neurotypical cerebellum, NRXN1 intergenic deletion was associated with cell-type–specific alterations in gene expression networks. Notably, pathways related to synapse organization and neuron projection development were enriched in both excitatory and inhibitory neuronal populations. Consistent with this, SynGO analysis showed significant enrichment of synaptic compartment genes across all neuronal cell types, indicating convergent disruptions in synaptic structure and function. 13 , 59 These findings support previous reports that NRXN1 deletions perturb cortical and cerebellar circuits through cell-type–specific mechanisms, consistent with circuit dysfunction observed in ASD and schizophrenia 58 – 60 . Given that NRXN1 isoforms determine synaptic partner specificity, the isoform atlas generated here provides a foundation for isoform-targeted therapeutic strategies, including ASO-based modulation of pathogenic splicing. Long-read Capture-seq and RACE-seq provide complementary approaches for transcript isoform characterization. Capture-seq uses probe-based enrichment of single-cell barcoded cDNA 61 to detect diverse isoforms, including alternative transcription start and termination sites, across multiple cell types while preserving relative transcript abundance. However, this method preferentially captures abundant transcripts and shows reduced sensitivity for low-abundance or long α-isoforms. In contrast, RACE-seq selectively amplifies transcripts from defined 5′ or 3′ ends, enabling accurate reconstruction of full-length isoforms, including those spanning deletion junctions. Although RACE-seq is less quantitative and has lower throughput, it provides higher sensitivity for rare transcripts. Applying this framework to assess ASO-induced transcript perturbations in organoids yielded fewer than 1,000 cells following treatment with either ASO-Aplus (827 cells) or control ASO-NC (668 cells) ( Supplementary Table 2 ). The limited recovery likely reflects ASO-associated toxicity ( Extended Fig. 6a,b ) and reduced cell viability in day-222 organoids 62 , 63 . Long-read Capture-seq profiling of these samples provided high-resolution transcriptome-wide splicing information; however, sequencing depth was insufficient to comprehensively analyze splicing at sites SS4 and SS5 due to the limited number of recovered cells. In this context, 5′ RACE-seq complemented Capture-seq by enabling detection of mutant isoforms derived from the deletion allele when expression levels were low. Some limitations should be noted. Both Capture-seq and RACE-seq remain constrained by RNA integrity and read length, particularly in postmortem tissue. In addition, scRNA-seq requires tissue dissociation, which disrupts cell–cell interactions and introduces dissociation bias, limiting the ability to assess how NRXN1 isoform perturbations reshape local cellular networks in situ. Future integration with spatial transcriptomics could help preserve tissue architecture and resolve cell–cell interactions. Another limitation of this study is the small number of patient-derived samples. Specifically, we analyzed two organoids’ lines carrying 3′- NRXN1 deletions, along with one ASD cerebellum sample harboring an NRXN1 deletion and one control. This limited sample size reflects both the rarity and non-recurrent nature of NRXN1 exonic deletions in the currently available biobank 3 , 64 . However, building on this study, we are in collaboration with a broad research network of NRXN1 deletion and will expand the research to additional patient-derived samples in future studies. With the promise of ASO based gene therapy targeting RNA splicing, resolving transcript diversity at cell-type resolution is essential for understanding the delivery and knockdown efficiency. Our targeted long-read sequencing approach enables high-resolution characterization of isoform usage and splicing shifts across distinct cell populations following ASO treatment, providing a cell-type–specific readout of isoform knockdown efficiency. Overall, this study demonstrates that targeted long-read sequencing offers a framework for comprehensive characterization of low-abundance disease-associated transcript isoforms with cell type specificity, with broad application to study natural isoform variations and the evaluation and optimization of ASO-based interventions targeting pathogenic splicing events. METHODS Data Availability Raw long-read sequencing data have been deposited in the Sequence Read Archive under accession number PRJNA1447486. Processed and raw single-cell RNA sequencing (scRNA-seq) data, including gene–cell count matrices and associated metadata, have been deposited in the Gene Expression Omnibus under accession number GSE327718 (to be updated upon release). Acquisition and processing of human post-mortem samples The MSSM cohort single nuclei cDNA The sequencing data and full-length single-nucleus cDNA libraries of 61 individuals from the Mount Sinai School of Medicine cohort were generated from the previous publication 27 . The leftover barcoded full length single-nucleus cDNA was used as input for the current study. To obtain 1-1.5 µg of cDNA for the hybridization step, 5-10 ng of the surplus barcoded 10x cDNA was amplified in two 50 µl reactions, which consist of 25 µL 2X PrimeSTAR GXL premix (Takara, Cat# R051A), 1 µL partial read 1 primer (10 µM, 5’-CTACACGACGCTCTTCCGATCT-3’), 1 µL partial TSO primer (10 µM, 5’-AAGCAGTGGTATCAACGCAGAG-3’). The amplification program is set as follows: 98 °C for 30 seconds, 8 cycles of 98 °C for 10 seconds, 65 °C for 15 seconds and 68 °C for 8 minutes, 1 cycle of 68 °C for 10 minutes. The concentration of PCR products was quantified by Qubit, and additional cycles were added if the yield was lower than 500 ng per reaction. The PCR products were cleaned up with 0.6× AMPure XP beads. NRXN1 +/- cerebellum of human post-mortem brain Postmortem human cerebellum of the ASD case with a heterozygous NRXN1 +/- mutation were collected and consented through the Department of Pathology at Western Michigan University Homer Stryker MD School of Medicine under WCG IRB protocol #20111080 and the other samples were consented through the National Institute of Mental Health Intramural Research Program under NIH protocol #90-M-0142, and was acquired by LIBD via material transfer agreement. Fetal brain Two human gestational cortical tissue samples (24 gw and 33 gw) were obtained from previous study 36 in accordance with the policies and regulations at the Icahn School of Medicine at Mount Sinai and its Institutional Review Board. Single cell/nucleus full-length barcode cDNA 60mg of the frozen post-mortem tissues were submitted to Yale Center for Genome Analysis for single nuclei isolation. Briefly, single nuclei were automatically extracted from the frozen tissue using the Singulator system, which includes washing nuclei suspension in buffer supplemented with sucrose, and nuclei FACS sorting. The isolated nuclei were then processed using the 10x Genomics 5’GEX V2 to create single nuclei cDNA and matched short-read RNA-seq libraries per manufacturer’s instructions. Patient-derived organoids Human induced pluripotent stem cell (hiPSC) lines used were validated using methods as previously described. Genome-wide SNP genotyping was performed using Illumina genome-wide GSAMD-24v2-0 SNP microarray at the Children’s Hospital of Philadelphia. Cultures were tested for Mycoplasma contamination and maintained Mycoplasma-free. A total of 3 hiPS cell lines derived from fibroblasts collected from 1 healthy subject (690) and two 3’-del lines from two patients (581 & 641) were used for experiments. Approval for this study was obtained from the Stanford IRB panel, and informed consent was obtained from all subjects. Stem cell maintenance Human iPSCs were obtained as previously described in Flaherty et al., 2019. Under sterile conditions, 6-well plates (Corning 3516) were coated with 60 ug/mL Geltrex matrix (Thermo Fisher Scientific A1413302) diluted in Dulbecco’s Modified Eagle Medium (DMEM, Gibco 11965-092) per well and incubated for at least 30 min at 37 °C. All hiPSC lines were seeded on coated plates and maintained in StemFlex (StemFlexTM Basal Medium (Gibco A3349301) supplemented with StemFlexTM Supplement (Gibco A3349201)) at 37 °C. After 4 days, hiPSC lines were passaged by removing the media washing cells with PBS. Then cells were incubated in 0.5 µM EDTA in PBS (Invitrogen AM9261) for 4 min at 37 °C. After removing EDTA, colonies were broken into smaller pieces by triturating using a p1000 pipette and StemFlex. It is worth noting that very small colonies have a very low survival chance and should be supplemented with 50nM chroman (CHR; MedChemExpress, Cat. No. HY-15392). After dissociation, colonies were distributed in new coated plates containing StemFlex in a ratio of 1:6 (250K cell per well of a 6 well plate). A complete media change was performed the following day. Complete media changes were performed as needed or every other day. Cortical organoid generation Cultures of hiPSC were dissociated at 70% confluency and checked for signs of differentiation. Mature undifferentiated cells were washed with PBS, followed by incubation with Accutase (STEMCELL™ Technologies, Cat. No. 07920 and CHR for 10-20 min to achieve complete detachment. Cells were gently triturated to achieve a single-cell suspension. The resulting cell suspension was transferred into DMEM (Gibco, Cat. No. 11960044) and centrifuged at 500 × g for 4 min. Cells were resuspended in StemFlex™ medium (Thermo Fisher Scientific) supplemented with CHR and diluted at a density of 1.5 × 10^6 cells/mL. AggreWell™ plates (STEMCELL Technologies) were prepared with anti-adherent solution according to the manufacturer’s instructions. A volume of 2 mL of the hiPSC suspension was seeded per well to promote spheroid formation overnight (defined as day 0). Neural induction was subsequently performed following the human forebrain spheroid protocol described by Sloan et al 66 . In brief, medium was replaced daily until day 15, every other day from days 15–42, and every 4 days from day 43 onward. On day 1, embryoid bodies were dislodged and transferred to ultra-low adhesion 10 cm dishes containing Neuronal Induction Media (StemFlex™ medium without StemFlex™ supplement), SMAD inhibitors SB431542 (10 µM; R&D Systems/Tocris Bioscience,1614), LDN193189 (100 nM; STEMCELL Technologies, 1614), XAV (100nM; STEMCELL Technologies,72672). Plates were placed on an orbital shaker at 53 rpm and maintained at the same speed until day 6. From days 6–25, Neuronal Differentiation Media (Neurobasal-A (Thermo Fisher Scientific), 1x Antibiotic-Antimycotic (Gibco,15240062), 1x GlutaMAX™(Gibco, Cat. No. 5050061), and 1x B-27 supplement minus vitamin A (Gibco, Cat. No. 2587010) was used supplemented with recombinant human FGF2/bFGF (20 ng/mL; R&D Systems, Cat. No. 233-FB-01M) and recombinant human EGF (20 ng/mL; R&D Systems, Cat. No. 236-EG). Between days 25–42, Neuronal Differentiation Media was supplemented with brain-derived neurotrophic factor (BDNF; 100 nM; R&D Systems, Cat No.248-BD-025) and neurotrophin-3 (NT-3; 100 nM; PeproTech, Cat No.450-03). From day 42 onwards, organoids were maintained in Neuronal Differentiation Media. Cryoprotection and immunocytochemistry Organoids were washed with PBS and fixed in 4% paraformaldehyde (PFA)/phosphate buffered saline (PBS) for 3 to 4 hours overnight at 4°C. Organoids were then washed in PBS and transferred to 25% sucrose/H 2 O for 48 hours. Organoids were embedded in gelatin/sucrose embedding solution by dissolving 7.5 g gelatin in 100 ml 10% sucrose solution at 37°C to obtain a 7.5% gelatin/10%sucrose embedding solution. For immunofluorescence staining, 14-16 μm-thick sections were washed with PBS and blocked in 10% donkey serum, and 0.1% Triton-X in PBS for 1 h at room temperature followed by primary (overnight, 4 °C) and secondary antibody (1 h at room temperature, Jackson ImmunoResearch or ThermoFisher Scientific) incubation. Slides were then mounted (VECTASHIELD, Vector Labs) and imaged on a Zeiss microscope with an apotome module equipped with ZEN 3.3 (ZEN pro) software. Primary antibody list:. Primary Antibody list: BCL11B/CTIP2 (rat, 1:500, Abcam), FOXG1 (rabbit, 1:200, Abcam), GAD1/GAD67 (mouse, 1:1000, Millipore), HuC/D (mouse, 1:500, Invitrogen), KI67 (rabbit, 1:500, Vector Labs), PAX6 (mouse, 1:200, BD Bioscience), SOX1 (goat, 1:20, R&D Systems), TBR1 (rabbit, 1:1000, Abcam). Antisense oligonucleotide treatment A single HPLC-purified antisense oligonucleotide (ASO) targeting the mutant exon 20/24 splice junction was designed and synthesized by IDT. The ASO contains a phosphorothioate-modified backbone with incorporated locked nucleic acid (LNA) nucleotides (Affinity Plus, Aplus). A non-targeting ASO (ASO-NT) was used as a matched control in all experiments. The sequence of ASO-Aplus as +G*+G*+T*T*G*G*C*T*G*C*A*G*G*+G*+T*+A; and the sequence of ASO-NT as A*A*C*A*C*G*T*C*T*A*T*A*C*G*C. To optimize delivery conditions and dosage, human cortical organoids (hCOs) derived from line 641 (3′- NRXN1 ⁺/⁻) at day 222 were treated with 1 μM or 10 μM ASO-Aplus or ASO-NT, delivered either via Lipofectamine RNAiMAX 67 or gymnosis, and incubated for 96 h prior to qPCR validation or single-cell and RNA collection. Expression levels of mutant and wild-type NRXN1 transcripts were quantified by qPCR to assess ASO knockdown efficiency. Single-cell dissociation and RNA-seq library preparation Organoids with optimal morphology from each line were selected in days 25, 60, and 120. Five organoids from each line were pooled at each time point and transferred to a 5 cm plate dish and washed with PBS. Organoids were chopped into small pieces using a sharp sterile blade. 300 ul of Accumacs was added on top of the pieces without scattering them. Pieces were collected by a cut tip and were transferred into a 1.5 ml tube. The tubes were incubated at 37°C and 800 g rotation on a thermal shaker for 20 min. After 20 min, cell suspension was very gently pipetted up to 5 times with a p 1000 tip, to facilitate the cell dissociation. 500 ul fresh Accumacs was added into the suspension and incubated again for 20 min. Cell suspension was pipetted up to 5 times more and passed through a filter in a 15 ml sterile tube containing 1 ml of HEPES buffer. The filter was washed with 1 ml of HEPES buffer. The single cell suspension in a 15 ml tube was centrifuged at 300 g for 4 min to collect cells in a pellet. After removing the supernatant, the pellet was resuspended in 200 ul HEPES buffer and cells were counted using a hemocytometer. The suspension was diluted to reach 1200 cell per ml and was submitted to Yale Genomics Core facility for cDNA library prep using 10x 5’ NEXT GEM V2 according to the manufacturer’s instructions. NGS sequencing of single cell/ nucleus cDNA The prepared 10X Genomics single cell/nucleus libraries were submitted to Yale Center for Genome Analysis and sequenced on an Illumina NovaSeq sequencer with 150-bp paired-end sequencing on an S4 flow cell. For each sample, ∼10,000 cells were sequenced at a minimum depth of 50,000 reads per cell. Probe-panel design IDT Lockdown probes (Integrated DNA Technologies) are 120-nt long oligos that are biotinylated at their 5’ ends and were designed using the xGen Panel Design Tool ( https://www.idtdna.com/pages/tools/xgen-hyb-panel-design-tool ) to ensure robust coverage of the targeted regions. Briefly, a customized fasta file containing all Ensembl 26 -annotated exons (across all isoforms) of target genes, including UTRs was uploaded as input file. The probe sequence list was generated at the 1x probe tilling density using “Homo sapiens (Human) NCBI GRCh37.p13 (hg19)” as the reference (Supplementary table 1). The probe panel was further synthesized employing TEQUILA-seq 21 . First, the probe panel pool of 150-nt long oligos, each comprising a 3’ end 30-nt universal primer binding sequence (5’-CGAAGAGCCCTATAGTGAGTCGTATTAGAA-3’) followed by a 120-nucleotide target-specific probe, was synthesized with Twist Bioscience (Supplementary table 1). Then the oligo pools were amplified and biotin-labeled using nickase-induced linear strand displacement amplification as described previously. Finally, synthesized probes were purified with 1.8× volumes of AMPure XP beads and quantified by NanoDrop 2000 Spectrophotometer. Long-read Capture-seq on single cell/ nucleus cDNA All hybridization and capture steps were performed following the IDT “Hybridization capture of DNA libraries using xGen Lockdown probes and reagents” protocol. Briefly, 1-1.5 µg of amplified 10x cDNA together with 1 µL of each blocking oligos (1 mM, Supplementary Table1 ) were dried using a speed vac (Thermo, Cat# SPD140P1). Next, the dried sample were re-hydrated with 8.5 μL of 2× Hybridization Buffer, 2.7 μl Hybridization Buffer Enhancer (xGen IDT 1080577), and 3.8 μL nuclease-free water, followed by denaturation at 95 °C for 10 min. The denatured mixture was then incubated with 200 ng of TEQUILA probes at 65 °C for 16 h. Next, the capture step was performed by adding 100 μL of washed M-270 streptavidin beads (Invitrogen, #65306) to the mixture, followed by incubation at 65 °C for 45 minutes with vortexing every 10 minutes to ensure thorough mixing. After capture, the mixture was then immediately washed with high-temperature and room temperature buffers and resuspended in 40 μL nuclease-free water. The streptavidin bead-captured cDNA was amplified using PrimeSTAR GXL premix (Takara, Cat# R051A) and 10x cDNA amplification primers by incubating at 95 °C for 3 min, followed by 12 cycles of (98 °C for 20 s, 65 °C for 15 s, 68 °C for 8 min), with a final extension at 68 °C for 8 min. Post-capture PCR products were purified using 0.7× volumes of AMPure XP beads, and quantified by the Qubit dsDNA High Sensitivity assay and size checked by Agilent TapeStation. Gene-targeted single cell/ nucleus RACE-PCR To sequence the NRXN1α isoforms and 10x single cell barcode from one long-read, the modified RACE-PCR was employed to amplify the target from the 10x Genomics 3’GEX or 5’GEX cDNA. 5’RACE was used for cDNA generated by 10x 5’GEX kit, while 3’RACE was used for cDNA generated by 10x 3’GEX kit. PCR primer design As the expression level of NRXN1α is relatively low in adult brains, nested PCR would be needed to achieve enough molecules. The gene-specific primers (GSPs) should meet the criteria: 23-28 nt long to ensure specific annealing, with 50%-70% GC, a T m 65°C (best to have a T m >70°C) and be specific to the target gene ( Supplementary Table 1 ). Universal primers: The universal primer for first –round PCR should be the partial read1 sequence from 10x cDNA. To increase sensitivity for the second-round PCR, a sequence of “AATGATACGGCGACCACCGAGATCT” was added to the 5’ of the read1 ( Supplementary Table 1 ), forming the second-round universal PCR primer (UP2) ( Supplementary Table 1 ). Protocol The input of 1-50 ng of 10x 5’GEX or 3’GEX cDNA was used for the RACE-PCR. The first round of RACE-PCR is carried out with 1 µL partial Read1 primer (10 µM) and 4 µL NRXN1 –specific primer 1 (10 µM) in a 50 µl-reaction, which consists of additional 25 µL 2X PrimeSTAR GXL premix (Takara, Cat# R051A), 1-50 ng of full-length single cell cDNA by 10x 5’GEX or 3’GEX kit. Choose the PCR program based on the GSP T m . Program 1: when GSP T m >70°C, run 5 cycles of 94°C for 30s, 72°C for 8min; 5 cycles of 94°C for 30s, 70°C for 30s and 72°C for 8min; 15-20 cycles of 94°C for 30s, 68°C for 30s and 72°C for 8min. Program 2: when T m = 60-70°C, run 15-20 cycles of 94°C for 30s, 68°C for 30s and 72°C for 8min. The concentration of PCR products was quantified by Qubit, and additional cycles can be added if the concentration of PCR product is lower than 10 ng/µL. After the PCR, 3 µL of products were analyzed with 1% agarose/ GelRed gel (biotium, cat# 41002), while the rest products were purified using 0.7× volumes of AMPure XP beads and eluted with 30 µL nuclease-free water. If the primary PCR fails to show the distinct band(s) of interest or produces a smear, a second-round of RACE-PCR should be performed. The second-round of RACE-PCR is performed with 2.5 µL of UP2 (10 µM) and 2.5 µL NRXN1 -specific primer 2 (10 µM) in a 50 µl-reaction together with 25 µL 2X PrimeSTAR GXL premix and 10 µL of the purified first-round RACE-PCR products. The PCR program follows the program 2 and runs for 10-15 cycles based the final concentration of PCR products. After the PCR, 3 µL of products were analyzed with 1% agarose/ GelRed gel for size check, which is followed by the purification using 0.7× volumes of AMPure XP beads. Library preparation for Oxford Nanopore sequencing Oxford Nanopore SQK-NBD114.24 was used for library preparation. Amplified cDNA from post-capture of RACE-PCR, approximately 200 ng of PCR products from each sample were used for library preparation according to the ONT SQK-NBD114.24 protocol. Briefly, cDNA products were end-repaired and dA-tailed with NEBNext Ultra II End Repair/dA-Tailing Module by incubating at 20 °C for 30 min and 65 °C for 30 min. The cDNA was then purified with 1× volume of AMPure XP beads and eluted in 10 μl of nuclease-free water. Native barcode ligation was then performed using 10 μl of end-prepped DNA, 2.5 μl of native barcodes, and 12.5 μl of Blunt/TA Ligase Master Mix (NEB, #M0367) by incubating at room temperature for 30 min. After the ligation step, 2.5 μl of EDTA was added to each tube to inactivate the ligase. Barcoded cDNA samples were then pooled and purified using 0.4× volume of AMPure XP beads and eluted in 31 μl of nuclease-free water. Next, ligation of Native Adapter was performed with 30 μl of pooled barcoded samples, 5 μl of native adapter, 10 μl of 5× NEBNext Quick ligation reaction buffer, and 5 μl of NEBNext Quick T4 DNA ligase at room temperature for 30 min, followed by the bead’s purification using 0.45× volume of AMPure XP beads and twice wash with short fragment buffer. The libraries were quantified with Qubit, and 50 fmol of final library was sequenced with a R10.4.1 MinION or PromethION flow cell for 72 hours. Single-cell RNA-seq processing and analysis Raw 10x Genomics single-cell RNA-seq data were processed using Cell Ranger (v7.2.0) with the GRCh38 reference genome to generate gene-cell UMI count matrices. Downstream analyses were performed in Seurat (v5.0.3) 68 under R (v4.4.0). Cells with fewer than 200 detected genes, more than 6,000 detected genes, or mitochondrial transcript fractions exceeding 20% were excluded. Potential doublets were identified and removed using DoubletFinder 69 . See the sequencing summary in the Supplementary Table 2 . Post-quality control, gene expression values were normalized by scaling each cell’s total expression to 10,000 counts and log-transforming the results. Data were then processed using SCTransform 68 , which performs regularized negative binomial regression to normalize and correct for sequencing depth and mitochondrial content. Highly variable genes were selected for PCA, and dimensionality reduction was performed using UMAP. Clustering was conducted using a shared nearest neighbor (SNN) graph and the Louvain algorithm to identify transcriptionally distinct populations. Cell type annotation was guided by canonical marker gene expression and cross-referenced with public human brain (or tissue-specific) single-cell catalogs. Differential expression analyses between groups or clusters were performed using the Wilcoxon rank-sum test with Benjamini-Hochberg correction for multiple testing (FDR < 0.05). Following dimensionality reduction and clustering, known marker genes were used to guide the assignment of biological cell identities. Each cluster was annotated based on the expression of established cell-type-specific markers reported in previous studies. GO Biological Process (BP) enrichment analysis was performed for each cell type using differentially expressed genes (DEGs) identified between ASD and control samples. Up– and downregulated DEGs were analyzed separately. Enrichment was tested using clusterProfiler (v4.12.6) with a background of all genes tested in that cell type (18,903 genes). Significant GO terms were retained at FDR < 0.05 (Benjamini–Hochberg correction). For each cell type and direction, up to the top 200 most significant terms were carried forward for downstream clustering. To reduce redundancy among significant GO BP terms and identify broad biological themes, semantic similarity clustering was performed using the rrvgo R package. For each direction (up/down), a pairwise semantic similarity matrix was computed across all significant GO terms pooled from all cell types. Terms were clustered into representative parent terms using reduceSimMatrix with a similarity threshold of 0.7. Each GO term was scored as the maximum −log₁₀(FDR) observed across all cell types. The resulting clusters were summarised into biological themes, and the top 20 themes ranked by the number of cell types with significant signal are displayed. For each cell type × theme combination, the cell in the heatmap represents the maximum −log₁₀(FDR) across all member GO terms within that cluster. Enrichment of DEGs within synaptic compartments was assessed using the SynGO database (v1.3). Three cellular component categories were tested: Synapse, Presynapse, and Postsynapse. For each cell type and direction (up/down), a one-sided hypergeometric test. Multiple testing correction was applied using the Benjamini–Hochberg method across all cell type × compartment combinations within each direction. Enrichments with FDR < 0.05 were considered significant. Long-read isoform sequencing analysis Full-length transcript isoforms were characterized using long-read RNA sequencing (Oxford Nanopore or PacBio). Raw reads were basecalled using dorado (v0.4.1). Reads were quality-filtered to retain those with a minimum mean Phred quality score of Q10 and minimum read length of 200 bp, and adapter sequences were trimmed prior to alignment. Processed reads were aligned to the human reference genome (GRCh38) using minimap2 (v2.28) with cDNA splice – aware parameters (-ax splice –-secondary=no –C5). Alignments were coordinate-sorted and indexed using SAMtools (v1.21), and only primary alignments with a minimum mapping quality of MAPQ ≥ 10 were retained for downstream analysis. Isoform reconstruction and quantification were performed using IsoQuant (v3.2) with GENCODE v43 as reference annotation. IsoQuant 70 clusters reads into high-confidence transcript models, classifying each as known, novel in catalog (NIC), or novel not in catalog (NNC) based on splice junction composition. Only isoforms supported by at least three full-length reads were retained. Differential splicing analysis was performed using LeafCutter 71 . Briefly, splice junctions were extracted from BAM files using RegTools, with a minimum anchor length of 8 bp and intron size ranging from 50 to 500,000 bp. Junction files from all samples were combined and clustered into intron groups using the LeafCutter clustering pipeline, requiring a minimum of 50 supporting reads per cluster. Integration of long-read isoform data with short-read single-cell transcriptomes Integration between long-read and short-read data was performed at both the gene and isoform levels. Cell barcode and UMI assignment from Nanopore reads was accomplished using BLAZE (v1.4.0) 72 , which accurately detects 10x cellular barcodes and unique molecular identifiers (UMIs) directly from Nanopore cDNA reads. BLAZE localizes barcode sequences within read adapter contexts, compares candidate sequences against the 10x whitelist, and assigns valid barcodes with up to one mismatch tolerance to each read. This enabled unambiguous association of long reads with individual cell identities and facilitated construction of a cell × isoform count matrix for isoform-level single-cell analysis. NRXN1 Isoform Quantification using miniQuant For the BA4 samples, miniQuant 73 was applied in hybrid mode, integrating both short-read and long-read RNA-seq generated in-house to improve isoform-level quantification. For the FC and hippocampus datasets 25 , 74 , where only long-read RNA-seq data were available, miniQuant was applied in long-read-only mode to estimate isoform abundances. In miniQuant , long-read cDNA sequencing provides high-confidence isoform structures and reduces ambiguity in assigning reads to specific isoforms, whereas short-read sequencing improves quantification accuracy for low-abundance transcripts by reducing sampling error. miniQuant ’s gene-specific weighting model adaptively balances these complementary data sources based on isoform structural complexity (K-value), transcript abundance, and sequencing depth, producing optimized isoform abundance estimates. Isoform expression levels were reported as transcripts per million (TPM) to allow direct comparison across samples. To ensure quantification reliability, we filtered out low-confidence isoforms following the criteria described in the original miniQuant framework, requiring both sufficient long-read support and a minimum expression threshold. Only isoforms with TPM ≥ 1 in at least one sample were retained for downstream analyses. Data visualization and statistical analysis All visualizations, including UMAPs, dot plots, violin plots, and bar plots, were generated in R or Python using the ggplot2 , matplotlib , and seaborn libraries. For continuous variables, such as comparisons of AS event inclusion levels or isoform expression differences, we applied Welch’s t-test , which does not assume equal variances between groups. AUTHOR CONTRIBUTIONS LC, YF and GF designed the methods. LC and YF performed targeted long-read sequencing and conducted data analysis. SG generated cortical organoids, and JM performed ASO treatments. YZ assisted with bioinformatic analyses. MBF designed the ASO. JB, JF, and PR provided single-nuclei RNA sequencing data from adult post-mortem brains and contributed to data analysis. SIR and NT provided post-mortem fetal brain tissues. NH provided technical assistance with Capture-seq. GD, KGB, and RS assisted with long-read sequencing. BZ and SM provided technical assistance with single-nuclei cDNA preparation. EAM contributed to data analysis and interpretation. LC, YF, SG, KJB and GF wrote the manuscript with input from all authors. KJB and GF supervised the research. COMPETING INTERESTS The authors declare no competing interests. ACKNOWLEDGEMENTS We thank Daniel Weinberger and the Lieber Institute for Brain Development at Johns Hopkins School of Medicine, for sharing post-mortem brain tissues. We thank Feng Wang and Yi Xing (Children’s Hospital of Philadelphia) for technical assistance with TEQUILA-seq. This work is supported by the National Institute of Mental Health grants RO1 MH125579 (GF and KJB) and RO1 MH121074 (KJB and GF), and U01MH116442 (PR) and R01MH125246 (PR). Research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under Award Number 1S10OD030363-01A1. N.A.L.H. was supported by the National Institute for Health and Care Research (NIHR) Oxford Health Biomedical Research Centre. This work was also supported in part through the computational resources and staff expertise provided by the Department of Scientific Computing at the Icahn School of Medicine at Mount Sinai. Funder Information Declared National Institute of Mental Health , RO1 MH125579 , RO1 MH121074 , U01MH116442 , R01MH125246 National Institute of General Medical Sciences , 1S10OD030363-01A1 Footnotes In this revision, we expanded the figure set to better clarify the cell-type-specific splicing landscape of NRXN1 and its functional implications. In Figure 1, we added two new panels that summarize the distribution of NRXN1 isoforms across cell types detected by Capture-seq and present a PCA highlighting the major splicing events that distinguish glutamatergic and GABAergic neurons across samples. To further support these findings, we included an additional panel in Figure 2 that summarizes the key splicing events underlying differences between these neuronal populations in the adult PFC. In Figure 4, we added two panels that summarize cell-type-specific biological themes derived from DEGs, along with SynGO enrichment, comparing control and NRXN1 deletion samples. Finally, we introduced a new Figure 6 that presents an evaluation framework for cell-typ-resolved ASO inhibition of gain-of-function NRXN1 isoforms in cortical organoids, providing a conceptual and analytical basis for assessing isoform-targeted therapeutic strategies. REFERENCES 1. ↵ Fagnani , M. et al. Functional coordination of alternative splicing in the mammalian central nervous system . Genome Biol . 8 , R108 ( 2007 ). OpenUrl CrossRef PubMed 2. Raj , B. & Blencowe , B. J . Alternative splicing in the mammalian nervous system: recent insights into mechanisms and functional roles . Neuron 87 , 14 – 27 ( 2015 ). OpenUrl CrossRef PubMed 3. ↵ Furlanis , E. , Traunmüller , L. , Fucile , G. & Scheiffele , P . Landscape of ribosome-engaged transcript isoforms reveals extensive neuronal-cell-class-specific alternative splicing programs . Nat. Neurosci . 22 , 1709 – 1717 ( 2019 ). OpenUrl CrossRef PubMed 4. ↵ Lipscombe , D. & Soto , E. J. L . Alternative splicing of neuronal genes: new mechanisms and new therapies . Curr. Opin. Neurobiol . 57 , 26 – 31 ( 2019 ). OpenUrl CrossRef PubMed 5. ↵ Gomez , A. M. , Traunmüller , L. & Scheiffele , P . Neurexins: molecular codes for shaping neuronal synapses . Nature Reviews Neuroscience vol. 22 137 – 151 Preprint at doi: 10.1038/s41583-020-00415-7 ( 2021 ). OpenUrl CrossRef PubMed 6. De Wit , J. & Ghosh , A . Specification of synaptic connectivity by cell surface interactions . Nat. Rev. Neurosci . 17 , 4 ( 2016 ). OpenUrl CrossRef PubMed 7. Südhof , T. C . Synaptic neurexin complexes: a molecular code for the logic of neural circuits . Cell 171 , 745 – 769 ( 2017 ). OpenUrl CrossRef PubMed 8. Harkin , L. F. et al. Neurexins 1-3 each have a distinct pattern of expression in the early developing human cerebral cortex . Cerebral Cortex 27 , 216 – 232 ( 2017 ). OpenUrl CrossRef PubMed 9. ↵ The role of neurexins and neuroligins in the formation, maturation, and function of vertebrate synapses . 10. ↵ Schreiner , D. et al. Targeted Combinatorial Alternative Splicing Generates Brain Region-Specific Repertoires of Neurexins . Neuron 84 , 386 – 398 ( 2014 ). OpenUrl CrossRef PubMed 11. Treutlein , B. , Gokce , O. , Quake , S. R. & Südhof , T. C . Cartography of neurexin alternative splicing mapped by single-molecule long-read mRNA sequencing . Proc. Natl. Acad. Sci. U. S. A . 111 , ( 2014 ). 12. ↵ Flaherty , E. et al. Neuronal impact of patient-specific aberrant NRXN1α splicing . Nat. Genet . 51 , 1679 – 1690 ( 2019 ). OpenUrl CrossRef PubMed 13. ↵ Uemura , T. et al. Trans-synaptic interaction of GluRδ2 and Neurexin through Cbln1 mediates synapse formation in the cerebellum . Cell 141 , 1068 – 1079 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 14. Siddiqui , T. J. et al. An LRRTM4-HSPG complex mediates excitatory synapse development on dentate gyrus granule cells . Neuron 79 , 680 – 695 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 15. Boucard , A. A. , Chubykin , A. A. , Comoletti , D. , Taylor , P. & Südhof , T. C . A splice code for trans-synaptic cell adhesion mediated by binding of neuroligin 1 to α-and β-neurexins . Neuron 48 , 229 – 236 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 16. ↵ Ko , J. , Fuccillo , M. V , Malenka , R. C. & Südhof , T. C . LRRTM2 functions as a neurexin ligand in promoting excitatory synapse formation . Neuron 64 , 791 – 798 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 17. ↵ Fuccillo , M. V et al. Single-cell mRNA profiling reveals cell-type-specific expression of neurexin isoforms . Neuron 87 , 326 – 340 ( 2015 ). OpenUrl CrossRef PubMed 18. ↵ Freund , T. F. & Katona , I. Perisomatic inhibition . Neuron 56 , 33 – 42 ( 2007 ). OpenUrl CrossRef PubMed Web of Science 19. ↵ Fernando , M. B. et al. Phenotypic complexities of rare heterozygous neurexin-1 deletions . Nature 1 – 11 ( 2025 ). 20. ↵ Lukacsovich , D. et al. Single-cell RNA-seq reveals developmental origins and ontogenetic stability of neurexin alternative splicing profiles . Cell Rep . 27 , 3752 – 3759 ( 2019 ). OpenUrl CrossRef PubMed 21. ↵ Wang , F. et al. TEQUILA-seq: a versatile and low-cost method for targeted long-read RNA sequencing . Nat. Commun . 14 , ( 2023 ). 22. ↵ Ray , T. A. et al. Comprehensive identification of mRNA isoforms reveals the diversity of neural cell-surface molecules with roles in retinal development and disease . Nat. Commun . 11 , ( 2020 ). 23. ↵ Ohara , O. , Dorit , R. L. & Gilbert , W . One-sided polymerase chain reaction: the amplification of cDNA . Proceedings of the National Academy of Sciences 86 , 5673 – 5677 ( 1989 ). OpenUrl Abstract / FREE Full Text 24. ↵ Frohman , M. A. , Dush , M. K. & Martin , G. R . Rapid production of full-length cDNAs from rare transcripts: amplification using a single gene-specific oligonucleotide primer . Proceedings of the National Academy of Sciences 85 , 8998 – 9002 ( 1988 ). OpenUrl Abstract / FREE Full Text 25. ↵ Hardwick , S. A. et al. Single-nuclei isoform RNA sequencing unlocks barcoded exon connectivity in frozen brain tissue . Nat. Biotechnol . 40 , 1082 – 1092 ( 2022 ). OpenUrl CrossRef PubMed 26. ↵ Dyer , S. C. et al. Ensembl 2025 . Nucleic Acids Res . 53 , D948 – D957 ( 2025 ). OpenUrl CrossRef PubMed 27. ↵ Ruzicka , W. B. et al. Single-cell multi-cohort dissection of the schizophrenia transcriptome . Science (1979) . 384 , eadg5136 ( 2024 ). OpenUrl CrossRef PubMed 28. ↵ Telley , L. et al. Sequential transcriptional waves direct the differentiation of newborn neurons in the mouse neocortex . Science (1979). 351 , 1443 – 1446 ( 2016 ). OpenUrl Abstract / FREE Full Text 29. ↵ Liodis , P. et al. Lhx6 activity is required for the normal migration and specification of cortical interneuron subtypes . Journal of Neuroscience 27 , 3078 – 3089 ( 2007 ). OpenUrl Abstract / FREE Full Text 30. ↵ Miyoshi , G. et al. Prox1 regulates the subtype-specific development of caudal ganglionic eminence-derived GABAergic cortical interneurons . Journal of Neuroscience 35 , 12869 – 12889 ( 2015 ). OpenUrl Abstract / FREE Full Text 31. ↵ Tremblay , R. , Lee , S. & Rudy , B . GABAergic interneurons in the neocortex: from cellular properties to circuits . Neuron 91 , 260 – 292 ( 2016 ). OpenUrl CrossRef PubMed 32. Cardin , J. A . Inhibitory interneurons regulate temporal precision and correlations in cortical circuits . Trends Neurosci . 41 , 689 – 700 ( 2018 ). OpenUrl CrossRef PubMed 33. ↵ Huang , Z. J. & Paul , A . The diversity of GABAergic neurons and neural communication elements . Nat. Rev. Neurosci . 20 , 563 – 572 ( 2019 ). OpenUrl PubMed 34. ↵ Liakath-Ali , K. & Südhof , T. C . The Perils of Navigating Activity-Dependent Alternative Splicing of Neurexins . Front. Mol. Neurosci . 14 , 659681 . Preprint at ( 2021 ). OpenUrl CrossRef PubMed 35. ↵ Baxter , P. S. , Dando , O. & Hardingham , G. E . Differential splicing choices made by neurons and astrocytes and their importance when investigating signal-dependent alternative splicing in neural cells . Front. Mol. Neurosci . 16 , 1214439 ( 2023 ). OpenUrl PubMed 36. ↵ Ramos , S. I. et al. An atlas of late prenatal human neurodevelopment resolved by single-nucleus transcriptomics . Nat. Commun . 13 , 7671 ( 2022 ). OpenUrl CrossRef PubMed 37. ↵ Wang , L. et al. Molecular and cellular dynamics of the developing human neocortex . Nature 1 – 10 ( 2025 ). 38. ↵ Zhu , K. et al. Multi-omic profiling of the developing human cerebral cortex at the single-cell level . Sci. Adv . 9 , eadg3754 ( 2023 ). OpenUrl CrossRef PubMed 39. ↵ Eze , U. C. , Bhaduri , A. , Haeussler , M. , Nowakowski , T. J. & Kriegstein , A. R . Single-cell atlas of early human brain development highlights heterogeneity of human neuroepithelial cells and early radial glia . Nat. Neurosci . 24 , 584 – 594 ( 2021 ). OpenUrl CrossRef PubMed 40. ↵ Sebastian , R. et al. Schizophrenia-associated NRXN1 deletions induce developmental-timing-and cell-type-specific vulnerabilities in human brain organoids . Nat. Commun . 14 , 3770 ( 2023 ). OpenUrl CrossRef PubMed 41. ↵ Birey , F. et al. Assembly of functionally integrated human forebrain spheroids . Nature 545 , 54 – 59 ( 2017 ). OpenUrl CrossRef PubMed 42. ↵ Trujillo , C. A. et al. Complex oscillatory waves emerging from cortical organoids model early human brain network development . Cell Stem Cell 25 , 558 – 569 ( 2019 ). OpenUrl CrossRef PubMed 43. Fair , S. R. et al. Electrophysiological maturation of cerebral organoids correlates with dynamic morphological and cellular development . Stem Cell Reports 15 , 855 – 868 ( 2020 ). OpenUrl PubMed 44. ↵ Park , Y. , et al. Modulation of neuronal activity in cortical organoids with bioelectronic delivery of ions and neurotransmitters . Cell Reports Methods 4 , ( 2024 ). 45. ↵ Matsunami , N. et al. Identification of rare recurrent copy number variants in high-risk autism families and their prevalence in a large ASD population . PLoS One 8 , e52239 ( 2013 ). OpenUrl CrossRef PubMed 46. ↵ Kim , H.-G. et al. Disruption of neurexin 1 associated with autism spectrum disorder . The American Journal of Human Genetics 82 , 199 – 207 ( 2008 ). OpenUrl CrossRef PubMed Web of Science 47. ↵ Kozareva , V. et al. A transcriptomic atlas of mouse cerebellar cortex comprehensively defines cell types . Nature 598 , 214 – 219 ( 2021 ). OpenUrl CrossRef PubMed 48. ↵ Kim , J. & Augustine , G. J . Molecular layer interneurons: key elements of cerebellar network computation and behavior . Neuroscience 462 , 22 – 35 ( 2021 ). OpenUrl CrossRef PubMed 49. Brown , A. M. et al. Molecular layer interneurons shape the spike activity of cerebellar Purkinje cells . Sci. Rep . 9 , 1742 ( 2019 ). OpenUrl CrossRef PubMed 50. ↵ Wisden , W. , Murray , A. J. , McClure , C. J. & Wulff , P . Studying cerebellar circuits by remote control of selected neuronal types with GABA-A receptors . Front. Mol. Neurosci . 2 , 808 ( 2009 ). OpenUrl 51. ↵ Brunet de Courssou , J.-B. , Durr , A. , Adams , D. , Corvol , J.-C. & Mariani , L.-L. Antisense therapies in neurological diseases . Brain 145 , 816 – 831 ( 2022 ). OpenUrl CrossRef PubMed 52. ↵ Ottesen , E. W. , Luo , D. , Singh , N. N. & Singh , R. N . High concentration of an ISS-N1-targeting antisense oligonucleotide causes massive perturbation of the transcriptome . Int. J. Mol. Sci . 22 , 8378 ( 2021 ). OpenUrl PubMed 53. ↵ Mortberg , M. A. et al. A single-cell map of antisense oligonucleotide activity in the brain . Nucleic Acids Res . 51 , 7109 – 7124 ( 2023 ). OpenUrl CrossRef PubMed 54. ↵ Buijsen , R. A. M. et al. Calcium-enhanced medium-based delivery of splice modulating antisense oligonucleotides in 2D and 3D hiPSC-derived neuronal models . Biomedicines 12 , 1933 ( 2024 ). OpenUrl PubMed 55. ↵ Hjelmgren , L. et al. Dysregulation of the DNA damage response by phosphorothioate antisense oligonucleotides . Nat. Commun . ( 2026 ). 56. ↵ Lim , L. , Mi , D. , Llorca , A. & Marín , O . Development and functional diversification of cortical interneurons . Neuron 100 , 294 – 313 ( 2018 ). OpenUrl CrossRef PubMed 57. ↵ Nguyen , T.-M. et al. An alternative splicing switch shapes neurexin repertoires in principal neurons versus interneurons in the mouse hippocampus . Elife 5 , e22757 ( 2016 ). OpenUrl CrossRef PubMed 58. ↵ Biswas , M. S. , Roy , S. K. , Hasan , R. & Pk , M. M. U . The crucial role of the cerebellum in autism spectrum disorder: neuroimaging, neurobiological, and anatomical insights . Health Sci. Rep . 7 , e2233 ( 2024 ). OpenUrl 59. ↵ van der Heijden , M. E. , Gill , J. S. & Sillitoe , R. V . Abnormal cerebellar development in autism spectrum disorders . Dev. Neurosci . 43 , 181 – 190 ( 2021 ). OpenUrl CrossRef PubMed 60. ↵ D’Mello , A. M. & Stoodley , C. J . Cerebro-cerebellar circuits in autism spectrum disorder . Front. Neurosci . 9 , 408 ( 2015 ). 61. ↵ Peng , H. et al. Single-cell Rapid Capture Hybridization sequencing reliably detects isoform usage and coding mutations in targeted genes . Genome Res . 35 , 942 – 955 ( 2025 ). OpenUrl Abstract / FREE Full Text 62. ↵ Richter , D. et al. The age of the hominin fossils from Jebel Irhoud, Morocco, and the origins of the Middle Stone Age . Nature 546 , 293 – 296 ( 2017 ). OpenUrl CrossRef PubMed 63. ↵ Bhaduri , A. et al. Cell stress in cortical organoids impairs molecular subtype specification . Nature 578 , 142 – 148 ( 2020 ). OpenUrl CrossRef PubMed 64. ↵ Lowther , C. et al. Molecular characterization of NRXN1 deletions from 19,263 clinical microarray cases identifies exons important for neurodevelopmental disease expression . Genetics in Medicine 19 , 53 – 61 ( 2017 ). OpenUrl CrossRef PubMed 65. ↵ Koopmans , F. et al. SynGO: an evidence-based, expert-curated knowledge base for the synapse . Neuron 103 , 217 – 234 ( 2019 ). OpenUrl CrossRef PubMed 66. ↵ Sloan , S. A. , Andersen , J. , Pașca , A. M. , Birey , F. & Pașca , S. P. Generation and assembly of human brain region–specific three-dimensional cultures . Nat. Protoc . 13 , 2062 – 2085 ( 2018 ). OpenUrl CrossRef PubMed 67. ↵ Chen , X. et al. Antisense oligonucleotide therapeutic approach for Timothy syndrome . Nature 628 , 818 – 825 ( 2024 ). OpenUrl CrossRef PubMed 68. ↵ Hao , Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis . Nat. Biotechnol . 42 , 293 – 304 ( 2024 ). OpenUrl CrossRef PubMed 69. ↵ McGinnis , C. S. , Murrow , L. M. & Gartner , Z. J . DoubletFinder: doublet detection in single-cell RNA sequencing data using artificial nearest neighbors . Cell Syst . 8 , 329 – 337 ( 2019 ). OpenUrl PubMed 70. ↵ Prjibelski , A. D. et al. Accurate isoform discovery with IsoQuant using long reads . Nat. Biotechnol . 41 , 915 – 918 ( 2023 ). OpenUrl CrossRef PubMed 71. ↵ Li , Y. I. et al. Annotation-free quantification of RNA splicing using LeafCutter . Nat. Genet . 50 , 151 – 158 ( 2018 ). OpenUrl CrossRef PubMed 72. ↵ You , Y. et al. Identification of cell barcodes from long-read single-cell RNA-seq with BLAZE . Genome Biol . 24 , 66 ( 2023 ). 73. ↵ Li , H. et al. Improving gene isoform quantification with miniQuant . Nat. Biotechnol . 1 – 13 ( 2025 ). 74. ↵ Joglekar , A. et al. Single-cell long-read sequencing-based mapping reveals specialized splicing patterns in developing and adult mouse and human brain . Nat. Neurosci . 27 , 1051 – 1063 ( 2024 ). OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted April 17, 2026. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Cell-type-resolved NRXN1 isoforms in human brain and hiPSC cortical organoids Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Cell-type-resolved NRXN1 isoforms in human brain and hiPSC cortical organoids Lei Cao , Yu Fan , Sadaf Ghorbani , Jessica Mariani , Yanchun Zhang , Michael B. Fernando , Jaroslav Bendl , John Fullard , Susana Isabel Ramos , Edward A. Mead , Nicola A.L. Hall , Gintaras Deikus , Kristin G Beaumont , Bohan Zhu , Sai Ma , Robert Sebra , Nadejda Tsankova , Panos Roussos , Kristen J. Brennand , Gang Fang bioRxiv 2025.11.11.687875; doi: https://doi.org/10.1101/2025.11.11.687875 Share This Article: Copy Citation Tools Cell-type-resolved NRXN1 isoforms in human brain and hiPSC cortical organoids Lei Cao , Yu Fan , Sadaf Ghorbani , Jessica Mariani , Yanchun Zhang , Michael B. Fernando , Jaroslav Bendl , John Fullard , Susana Isabel Ramos , Edward A. Mead , Nicola A.L. Hall , Gintaras Deikus , Kristin G Beaumont , Bohan Zhu , Sai Ma , Robert Sebra , Nadejda Tsankova , Panos Roussos , Kristen J. Brennand , Gang Fang bioRxiv 2025.11.11.687875; doi: https://doi.org/10.1101/2025.11.11.687875 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17697) Bioengineering (13894) Bioinformatics (41951) Biophysics (21455) Cancer Biology (18592) Cell Biology (25507) Clinical Trials (138) Developmental Biology (13380) Ecology (19903) Epidemiology (2067) Evolutionary Biology (24321) Genetics (15610) Genomics (22509) Immunology (17737) Microbiology (40398) Molecular Biology (17182) Neuroscience (88618) Paleontology (667) Pathology (2833) Pharmacology and Toxicology (4825) Physiology (7641) Plant Biology (15158) Scientific Communication and Education (2046) Synthetic Biology (4296) Systems Biology (9825) Zoology (2271)