Massively parallel screens to identify splice disruptive variants in human disease genes

Splicing is a critical step in mRNA maturation with roles in gene regulation and proteome diversification. Splice disruptive variants (SDVs) are implicated in diverse human diseases, and 10-33% of exonic variants may disrupt splicing. However, identifying SDVs remains challenging due to the degeneracy and redundancy of the underlying sequence code. Experimental splicing measurements from patient cells or mini-gene assays can detect SDVs but have traditionally been low-throughput. Massively parallel reporter assays (MPSAs) systematically measure splicing impacts at scale and could clarify variant pathogenicity and inform models of splicing regulation. In this assay, complex barcoded libraries of mutant exons are synthesized, cloned into minigene constructs, and transfected into human cells. Splicing outcomes of each mutation are quantified en masse using targeted RNA-seq of minigene-derived transcripts, and I developed custom python package to process the resulting data. In Chapter 2, I apply this assay to the pituitary transcription factor gene <em>POU1F1</em> (in collaboration Dr. Sally Camper’s lab). Mutations in <em>POU1F1</em> cause combined pituitary hormone deficiency (CPHD), a clinically and genetically heterogenous disorder with prevalence ~1:4000. We targeted exon 2, which has two alterative isoforms (alpha and beta) using competing splice acceptors that encode mutually antagonistic proteins. We measured the splicing effects of 1,070 SNVs across the exon and surrounding introns and identified 96 SDVs - 14 of which were synonymous substitutions. Our measurements were concordant with six nearby heterozygous missense and synonymous variants seen in unrelated hypopituitarism patients. This map identifies a putative splice silencer motif that represses the use of the normally lowly expressed beta isoform. In Chapter 3, I apply a MPSA to a critical developmental renal transcription factor gene, <em>WT1</em> (in collaboration with clinical nephrologist Dr. Jen Lai Yee). Mutations in <em>WT1 </em>are implicated in nephrotic syndrome and sexual differentiation phenotypes. I focus on exon 9 which is alternatively spliced at competing donor sites resulting in two isoforms (KTS+ and KTS-). KTS+ and KTS- are normally expressed in ~2:1 ratio, but perturbation of the ratio can lead to Frasier’s syndrome – a rare nephrotic syndrome. We tested 518 SNVs for splicing defects and identified 8 known Frasier’s Syndrome variants as well as 16 additional variants that similarly lowered the KTS ratio. We also detected 19 variants increasing the KTS ratio - two of which have been observed in patients with sexual differentiation phenotypes. Although MPSAs can measure splicing effects of hundreds of variants simultaneously, the current scale of variant discovery via exome and genome sequencing demands efficient and accurate computational approaches to identify splice disruptive variants genome-wide. To evaluate the state of the art within contemporary splice prediction algorithms, in Chapter 4 I employed the results of five high throughput splicing assays and one literature curated variant set. A unique advantage of MPSAs over typical training and validation datasets is that they avoid bias towards essential splice site variants. I found the latest deep learning tools, SpliceAI and Pangolin, were most concordant with the measured splicing effects. However, all tools showed less agreement with exonic splicing outcomes compared to intronic. Some tools’ predictions, like SpliceAI’s, were sensitive to specified annotation files. Thus, there is still room for improvement within the next generation of splice prediction algorithms which future MPSA studies may facilitate.剪接（splicing）是mRNA成熟过程中的关键步骤，参与基因调控与蛋白质组多样性构建。剪接干扰变异（splice disruptive variants, SDVs）与多种人类疾病相关，约10%~33%的外显子变异可能干扰剪接过程。然而，由于底层序列编码存在简并性与冗余性，识别SDVs仍颇具挑战。基于患者细胞或迷你基因实验的剪接检测方法可检出SDVs，但传统上通量较低。大规模平行报告基因检测（massively parallel reporter assays, MPSAs）能够系统性、规模化地量化剪接影响，有助于阐明变异的致病性，并为剪接调控模型提供支撑。本实验中，研究人员首先合成带有复合条形码标记的突变外显子文库，将其克隆至迷你基因构建体中并转染人类细胞。通过对迷你基因来源的转录本进行靶向RNA测序（RNA-seq），可批量定量所有突变的剪接结果；本人还开发了定制化Python包以处理后续产生的数据。在第2章中，本人与Sally Camper博士实验室合作，将该实验方法应用于垂体转录因子基因<em>POU1F1</em>。<em>POU1F1</em>的突变会引发联合垂体激素缺乏症（combined pituitary hormone deficiency, CPHD）——这是一种临床与遗传异质性疾病，患病率约为1:4000。我们针对外显子2展开研究，该外显子存在两种可变剪接异构体（α与β），其竞争性剪接接受位点编码相互拮抗的蛋白质。我们定量检测了外显子及其侧翼内含子区域共1070个单核苷酸变异（single nucleotide variants, SNVs）的剪接效应，最终鉴定出96个SDVs，其中14个为同义替换。我们的检测结果与6个在无关垂体功能减退症患者中发现的邻近杂合错义变异及同义变异高度一致。通过该变异图谱，我们鉴定出一个推定的剪接沉默子基序，该基序可抑制通常低表达的β异构体的剪接使用。在第3章中，本人与临床肾病学家Jen Lai Yee博士合作，将MPSA应用于关键的发育相关肾脏转录因子基因<em>WT1</em>。<em>WT1</em>的突变与肾病综合征及性别分化表型相关。本研究聚焦外显子9，该外显子在竞争性剪接供体位点发生可变剪接，产生两种异构体（KTS+与KTS-）。正常情况下，KTS+与KTS-的表达比例约为2:1，但若该比例失衡则会引发弗雷泽综合征（Frasier syndrome）——一种罕见的肾病综合征。我们针对518个SNVs的剪接缺陷进行检测，鉴定出8个已知的弗雷泽综合征相关变异，以及16个同样会降低KTS比例的新增变异。此外，我们还检测到19个会升高KTS比例的变异，其中2个曾在存在性别分化表型的患者中被发现。尽管MPSA可同时检测数百个变异的剪接效应，但当前通过外显子组与基因组测序开展的变异发现规模，亟需高效且精准的计算方法以在全基因组范围内鉴定SDVs。为评估当代剪接预测算法的现有水平，第4章中本人采用了5项高通量剪接检测的结果与1组经文献人工整理的变异集。与常规训练与验证数据集相比，MPSA的独特优势在于其不会偏向于必需剪接位点变异。研究发现，最新的深度学习工具SpliceAI与Pangolin与实测剪接效应的一致性最高。然而，与内含子区域的剪接结果相比，所有工具对外显子区域剪接结果的预测一致性均较低。部分工具的预测结果（如SpliceAI）会受指定注释文件的影响。因此，下一代剪接预测算法仍有改进空间，而未来的MPSA研究将助力此类算法的优化。