Single cell ATACseq of mouse developing cranial motor neurons, spinal motor neurons, and surrounding neuronal tissue at e10.5 and e11.5.

Although Mendelian disorders are overwhelmingly attributed to protein-coding pathogenic variants, a majority of unsolved cases do not harbor obvious causal pathogenic variants in the coding sequence, suggesting a potential non-coding etiology. However, classification of pathogenicity in non-coding sequence remains prohibitive due to a vastly increased search space and the lack of a standardized rubric for interpretation. Here, we present an integrated single cell multiomic framework to nominate pathogenic non-coding variants for the congenital cranial dysinnervation disorders (CCDDs). The CCDDs are Mendelian neurodevelopmental disorders that result from aberrant development of cranial motor neurons in the embryonic brainstem. We created a non-coding reference atlas of single cell chromatin accessibility profiles for 86,089 embryonic mouse cranial motor neurons (cMNs). We found that high-quality single cell ATAC-seq (scATAC) profiles alone were a strong predictor of enhancement (64% in vivo validation rate). To further aid in interpretation, we integrated single cell histone modification and gene expression information to distinguish individual enhancers and their cognate genes. Relatively subtle differences in cellular composition of input data often led to substantial differences in predicted enhancer strength, cognate gene, and tissue of activity. Next, we mapped candidate non-coding variants from 899 whole genome sequences from 270 CCDD pedigrees to the murine cMN-specific regulatory elements and trained a machine learning classifier to accurately predict the functional effects of patient variants within these elements. We then performed high coverage scATACseq and site-specific footprinting analysis on an allelic series of CRISPR-humanised mice to validate our machine learning predictions and render important clues to the mode of pathogenicity. Finally, we performed peak- and gene-centric allelic aggregation to nominate non-coding variants, including those regulating MN1 and EBF3, respectively. Altogether this work extends non-coding variant analysis to Mendelian disease and presents a generalizable framework for nominating novel non-coding variants in other rare disorders.Fluorescence-assisted microdissection was performed to collect samples cMN3/4, cMN7, and sMN from Isl1MN:GFP mice and likewise to collect samples of cMN6, cMN12, and sMN from Hb9:GFP mice, each at both e10.5 and e11.5. Nuclei were isolated in accordance with Low Cell Input Nuclei Isolation guidelines provided by ‘Demonstrated Protocol – Nuclei Isolation for Single Cell ATAC Sequencing Rev A (Protocol #CG000169) from 10x Genomics. performed scATAC transposition, droplet formation, and library construction as described in protocol CG000168 using v1 reagents (10x Genomics). scATAC libraries were sequenced on the Illumina NextSeq 500 system using standard Illumina chemistry. Paired inserts were minimum 2 x 34 bp in length excluding indices, and libraries were distributed to achieve an estimated coverage of ≥ 25,000 read pairs per cell in accordance with 10x Genomics guidelines. bcl outputs from NextSeq 500 were processed using 10x Cell Ranger ATAC mkfastq pipeline (https://www.10xgenomics.com/support/software/cell-ranger/) for demultiplexing and fastq file generation. ATAC fragment counts were tabulated using the Cell Ranger count function using default parameters. .bam and/or fragment files were processed for peak calling and differential accessibility analysis using R ArchR package (https://github.com/GreenleafLab/ArchR).

尽管孟德尔疾病绝大多数归因于编码致病变异，然而大量未解病例并未在编码序列中携带明显的因果致病变异，这表明可能存在非编码的病因。然而，由于搜索空间的大幅增加和缺乏标准化的解释准则，对非编码序列致病性的分类仍然具有挑战性。在此，我们提出了一种集成的单细胞多组学框架，用于提名先天性颅神经功能失调症（CCDDs）的致病非编码变异。CCDDs是一种由胚胎脑干颅神经运动神经元异常发育引起的孟德尔神经发育疾病。我们构建了一个包含86,089个胚胎小鼠颅神经运动神经元（cMNs）的单细胞染色质可及性图谱的非编码参考图谱。我们发现，仅高质的单细胞ATAC-seq（scATAC）图谱本身就是一个强有力的预测增强因子（体内验证率为64%）。为了进一步辅助解释，我们将单细胞组蛋白修饰和基因表达信息整合起来，以区分单个增强子和它们相应的基因。输入数据的细胞组成相对细微的差异，往往导致了预测的增强子强度、相应基因和组织活动的显著差异。接着，我们将来自270个CCDD家系的899个全基因组序列中的候选非编码变异映射到小鼠cMN特异性的调控元件上，并训练了一个机器学习分类器，以准确预测患者变异在这些元件中的功能效应。然后，我们对一系列CRISPR人类化小鼠进行了高覆盖率的scATAC-seq和位点特异性足迹分析，以验证我们的机器学习预测，并提供有关致病机制的重要线索。最后，我们进行了峰值和基因中心的等位基因聚合，以提名非编码变异，包括调节MN1和EBF3的变异。总之，这项工作将非编码变异分析扩展到孟德尔疾病，并提出了一个可推广的框架，用于在其他罕见疾病中提名新的非编码变异。通过荧光辅助显微切割，从Isl1MN:GFP小鼠中收集cMN3/4、cMN7和sMN样本，以及从Hb9:GFP小鼠中收集cMN6、cMN12和sMN样本，每个样本均在胚胎发育的e10.5和e11.5阶段进行。根据10x Genomics提供的“演示协议——单细胞ATAC测序核苷酸分离Rev A（协议编号CG000169）”中的低细胞输入核苷酸分离指南进行核分离。按照CG000168协议使用v1试剂（10x Genomics）进行了scATAC转座、滴液形成和文库构建。在Illumina NextSeq 500系统上使用标准Illumina化学进行scATAC文库测序。配对插入片段的长度至少为2 x 34 bp，不包括索引，文库分布以达到每个细胞估计覆盖≥ 25,000个读取对的目标，符合10x Genomics指南。NextSeq 500的bcl输出通过10x Cell Ranger ATAC mkfastq管道进行去复用和fastq文件生成。使用Cell Ranger count功能（默认参数）对ATAC片段计数进行汇编。.bam和/或片段文件使用R ArchR包（https://github.com/GreenleafLab/ArchR）进行峰调用和差异可及性分析。