The next-generation sequencing technology is revealing an increasing number of variations, and bioinformatics predictions and functional studies are crucial for the accurate classification of these variants. Here, we focused on potentially spliceogenic PKHD1 intronic variants reported in the ClinVar database, and checked 12 variants to dissect splicing effects using pSPL3 vector in HEK293T and Hela cells. Remarkably, a considerable percentage of tested variants (11/12, 91.67%) impaired splicing, underlining the specificity of our selection criteria. Variants presented similar splicing patterns and slightly different proportions of transcripts in two cell lines, which may be attributed to differences in splicing factor expression.

According to the prediction software and their location, eight variants disrupted the canonical 3’ and 5’ splice sites (c.53-5T > A, c.281 + 3 A > G, c.2716–3 C > G, c.5751 + 3 A > G, c.6332 + 3 A > C, c.7109 + 3 A > C, c.8302 + 5G > A, c.11310 + 3 A > G), five created novel active splice sites(c.281 + 3 A > G, c.1694–10 C > G, c.6332 + 3 A > C, c.8441–5 A > G, c.11175-10T > G). Notably, the wild-type (WT) minigene of Exon 62 generated a transcript of 413 bp (IVS62-ins14) coexist with the canonical transcript which might reflect a complex alternative splicing mechanism of PKHD1.

Functional studies of variants provide key information for their clinical interpretation. FPC undergoes a complicated pattern of Notch-like proteolytic processing. The latest research revealed that C-terminal cleavage product ICD15 translocate to mitochondria, thereby enhancing mitochondrial respiration in renal epithelial cells [8]. In this research, the predicted FPC encoded by aberrant transcripts resulting from eight variants of c.281 + 3A > G, c.2716-3C > G, c.5751 + 3A > G, c.6332 + 3A > C, c.7109 + 3A > C, c.8441-5A > G, c.11175-10T > G and c.11310 + 3A > G would be truncated at different position of extracellular regions, resulting in the loss of transmembrane domains and cytoplasmic domains. These predicted proteins have lost their basic structures and are unable to recognize signals and undergo cleavage and processing, ultimately impairing mitochondrial function and leading to a severe phenotype. Qiu LR had reported that the patients with compound heterozygous mutations of c.5751 + 3A > G and EX6_11DEL can present with renal cysts and reduced kidney function [9], which is consistent with our results. Although the remaining seven variants were listed as ‘LP’ or ‘VUS’ in the Clinvar database, our findings indicate that they are ‘detrimental’ variants. Moreover, the surveillance systems may recognize premature termination codons and recruit enzymes to degrade the aberrant mRNAs, which requires further study.

Variant c.53-5T > A causes a partial loss of IPT1 domain which may bind to undetermined ligands, modulating oligomerization of the protein and mediating adhesive interaction [10]. Consistent with “LB” presented in Clinvar database, the abnormal transcripts accounted for only a small proportion, which may be associated with a milder phenotype. Unlike the “VUS” presented in Clinvar database, the FPC caused by c.8302 + 5G > A would loss the part of G8 domains. It is speculated that the G8 domains may be related to scaffolding, cell signaling and adhesion. Thus, the above function of the mutant protein may be weakened. While it is true that variant c.1694–10 C > G will impact splicing, the resulting alteration is not necessarily pathogenic. Insertions (c.1693_1694insTGCATCCAG) are multiples of three base pairs (9 bp) and thus do not cause frameshifting.

The results predicted by SpliceAI and BDGP are basically consistent with minigene results, except for the variant c.2593-39G > T. BDGP demonstrated that there is an alternative splice site (tgacAGaatt, score:0.13) upstream of the acceptor splice site of exon 25(score:0.24), and the variants c.2593-39G > T increased the score of the alternative site from 0.13 to 0.51. However, the results of minigene assays were negative, which may be attributed to limitations of bioinformatics tools and the differential sequence context in regulating splicing.

It is worth noting that the true splicing effect of these variants needs to be confirmed by analyzing the mRNA from the patient. In addition, the functional activities and the cell surface expression of these mutant FPCs need further research.

In summary, our studies were very helpful in reclassification according to ACMG criteria and developing novel targeted gene therapy. Antisense oligonucleotide (ASO) therapy brings new hope for the treatment of ARPKD, which has made pleasing progress in neurological disorders. A research group has successfully designed three ASOs targeting the PKHD1 gene variant c.2141-3T > C, and they confirmed that the ASOs can block the abnormal pseudo exon insertion [6]. For this study, it is expected that ASO will be designed to block pseudoexon insertion associated with variant c.11310 + 3 A > G. Though beyond the scope of this study, future efforts are needed in this direction.