Disruption of RNA Splicing Is an Important Contributor to Congenital Hypopituitarism and Other Human Genetic Diseases

Genetic diagnosis is valuable for predicting disease risk and prognosis. Exome sequencing can resolve approximately one-third of cases, depending on disease type and patient selection criteria. The difficulty in identifying pathogenic variants among many rare variants is a major challenge. Computational predictors exist, but additional evidence is required for accurate variant interpretation. Functional studies are one option but are rarely completed on a time scale useful for individuals’ diagnosis. As a result, most protein-coding variants are of uncertain significance.

Congenital hypopituitarism is a genetically heterogeneous disorder involving > 30 genes, and many are dominant with incomplete penetrance [reviewed in (1)]. These issues make it challenging to identify the genetic basis of pituitary dysfunction, and >75% of cases are not solved by exome sequencing. Mutations in POU1F1 cause 2-8% of hypopituitarism cases. Affected individuals have deficiencies in GH, TSH, and prolactin, because POU1F1 transactivates those hormone genes. Akiba and colleagues analyzed the mechanism whereby an intronic c.142-83 A > G variant causes dominant pituitary hormone deficiency characterized by lack of GH, normal TSH with thyroid hormone insufficiency, and variable pituitary size (2). They demonstrated that this variant affects splicing and is pathogenic in mice.

Across human genetic disorders, up to one-third of pathogenic variants act by disrupting splicing (3), often leading to frameshifted or prematurely truncated proteins. Some splice disruptive variants are easily recognized, such those disrupting canonical GU/AG donor and acceptor dinucleotides at the ends of most introns. Programs like SpliceAI are increasingly utilized to predict the effect of genetic variation on RNA splicing. However, many sequence features that contribute to specifying proper splicing have less defined sequence requirements, making them more challenging to identify based on sequence alone. These include intronic positions outside the canonical splice sites, the branchpoint used to form the lariat structure during splicing, and enhancer and silencer motifs that influence splice site usage. Variants in the GH gene were an early demonstration of disease caused by disruption of exonic splice enhancer sequences [reviewed in (4)]. More than 80 diseases are caused by synonymous variants in coding sequences that alter splicing or mRNA regulation, stability, or translation (5).

Cell culture assays are an important tool for evaluating the functional consequences of genetic variants, but care must be taken in their interpretation as they may not fully reflect biological processes in vivo. Human organoids and mouse models of human disease can be particularly useful for studying disease mechanisms and pathogenesis and for assessing the efficacy of therapeutic approaches. A mouse model of the POU1F1 variant c.142-83 A > G was generated, and the homozygous mutant mice phenocopy the heterozygous children's severe growth insufficiency and pituitary hypoplasia (2). No GH positive cells were detected, and there was loss of GH and prolactin expression, with no major differences in other hormone-producing cell types. Two isoforms of POU1F1, alpha and beta, are generated by alternative splice acceptor usage, in a ratio of 99:1. This ratio was different in homozygous c.142-83 A > G mutant mice; the predominant isoforms were exon skipping and beta (79:1). These results mirror a saturation mutagenesis screen of the human exon (4).

Pou1f1 beta contains an insertion of 23 amino acids within the transactivation domain, and it can repress the activity of the normally predominant alpha isoform. Previously, 6 mutations within the beta isoform-encoding exon were shown to increase the usage of the beta isoform splice acceptor and cause dominant hypopituitarism with incomplete penetrance. These variants may inactivate an exonic splice silencer. Skipping exon 2 is predicted to produce a protein with an in-frame deletion of amino acids 48-97 within the transactivation domain. This deletion protein has markedly reduced activity on the POU1F1 and PRL promoters, and no detectable activity on the GH promoter (2). It does not appear to interfere with the function of the alpha isoform. Thus, there are multiple mechanisms whereby POU1F1 splice disruption can cause disease.

The c.142-83 A > G, variant is 5 nucleotides away from the beta isoform splice acceptor (named c.142-5A > G relative to beta). The intronic POU1F1 c.142-83 A > G variant causes hypopituitarism by destroying a critical branch point (2). This branchpoint is 83 nucleotides upstream of the Pou1f1 alpha acceptor, much farther than the median of ∼30 nucleotides, but more distal branchpoints may influence splice site selection at alternatively spliced exons (6), in this case favoring Pou1f1 alpha, the next acceptor AG at a viable distance downstream.

An analysis of A > G changes in 8 predicted branch points in a cell culture splicing assay revealed that only c.142-83 A > G caused exon 2 skipping, beta usage, and loss of the alpha isoform (2). While proving lariat formation requires other approaches, it is notable that a branchpoint motif prediction algorithm, BPP, ranks c.142-83 as the highest scoring branchpoint upstream of the alpha acceptor site. If branch points and splicing regulatory elements could be predicted more accurately, more cases of disease could be diagnosed effectively by genetic testing without the need for extensive functional analyses.

Analysis of patient transcriptomes improves patient diagnosis by ∼35% because it identifies splice altering variants in exonic and deep intronic regions that are missed by other approaches (7). This approach is obviously not feasible for disorders affecting tissues that cannot readily be biopsied. This limitation can be overcome if highly tissue-specific transcripts, like POU1F1, can be detected at trace levels in peripheral blood (8). This approach may be effective for detecting altered splicing for other genes suspected to cause disease or, potentially, to implicate regulatory variants that could ablate expression entirely.

High throughput cataloging of all possible variant effects in multiplex assays can generate beneficial tables for interpretation. These methods enable the creation of “function maps” containing measurements for all possible variants within a targeted gene, such that when novel variants are observed, their measured effects can be looked up. This, and improvements in prediction of splice disruptive variants, would enhance genetic diagnosis for patients.

Funding

We thank the National Institutes of Health: R01HD097096 (S. A. C.), R03 DE031037 (S. A. C., J. O. K.), R01GM129123 (J. O. K).

Disclosures

J. O. K. is a scientific advisor to MyOme, Inc. The authors declare no other competing interests.