A Novel AR Splice-Site Variant Causing Complete Androgen Insensitivity Syndrome: Transcript Confirmation and Lessons for Splice Prediction

Technical Summary: A Novel AR Splice-Site Variant Causing Complete Androgen Insensitivity Syndrome

Problem Statement
Androgen insensitivity syndrome (AIS) is an X-linked condition caused by pathogenic variants in the androgen receptor (AR) gene, manifesting as a spectrum from complete (CAIS) to partial (PAIS) and mild (MAIS) forms. While missense variants are the most commonly reported cause, disruption of pre-mRNA splicing represents a significant yet under-recognized mechanism. For many AR splice-site variants, pathogenicity is inferred primarily from variant position and in silico predictions, with direct experimental confirmation at the transcript level remaining limited. Furthermore, the reliability of single-predictor tools for non-substitution variants (e.g., duplications) at canonical splice sites is uncertain.

Methodology
The study employed a combination of clinical evaluation, genomic sequencing, and functional transcript analysis to characterize a novel AR variant.

• Clinical Cohort: A 46,XY individual with CAIS was identified following a family history of primary amenorrhea and a 46,XY karyotype in a sibling. Clinical workup included pelvic ultrasound, karyotyping, and hCG stimulation tests.
• Genomic Analysis: Sanger sequencing of all eight coding exons and flanking intronic regions of AR (NM_000044.6) was performed on genomic DNA from peripheral blood leukocytes.
• Transcript Analysis: To confirm the functional consequence, total RNA was extracted, reverse-transcribed, and a cDNA fragment spanning exons 5–8 was amplified and sequenced.
• Literature Curation and In Silico Comparison: The authors curated 14 previously reported AR splice-site variants associated with AIS from public databases (Androgen Receptor Gene Mutations Database, ClinVar, HGMD, PubMed). These were compared regarding clinical phenotypes and in silico splice predictions using SpliceAI and Human Splicing Finder (HSF).

Key Results

• Novel Variant Identification: The patient was found to carry a novel canonical acceptor variant, c.2319-1G>T, in intron 5 of the AR gene. The mother was confirmed as a carrier.
• Functional Confirmation: RT-PCR and cDNA sequencing demonstrated complete skipping of exon 6. This results in an in-frame deletion of 44 amino acids within the ligand-binding domain, providing a molecular explanation for the complete loss of receptor function and the CAIS phenotype.
• Genotype-Phenotype Correlation: Among the 15 total variants analyzed (14 previously reported + 1 novel), 14 (93%) were associated with CAIS. Only one variant (c.2319-1_2332del) was associated with PAIS. This suggests a strong class-level association between canonical splice-site disruption and the complete phenotype, contrasting with the more variable genotype-phenotype correlations often seen in missense variants.
• Predictor Discordance: In silico analysis revealed a critical limitation in relying on single prediction tools. While SpliceAI assigned high delta scores to all single-nucleotide substitution variants, it assigned a score of 0.00 (no effect) to the canonical donor-site duplication c.1616+1dup. Conversely, Human Splicing Finder correctly predicted donor-site loss for this duplication.

Significance and Claims
The paper claims that these findings expand the spectrum of functionally confirmed AR splice-site variants causing CAIS. The authors emphasize that transcript analysis serves as an important reference standard for assessing novel or uncertain splice-associated AR variants, particularly when in silico tools may fail.

Specifically, the study highlights that:

1. Canonical splice-site disruptions in AR appear to define a molecular class with a consistent tendency toward the CAIS phenotype.
2. Reliance on a single splice predictor (such as SpliceAI) is insufficient, particularly for insertions, duplications, or deletions at canonical splice positions, as these may yield false-negative scores.
3. Transcript analysis should be prioritized when evaluating non-substitution variants at canonical splice positions to avoid misinterpretation of pathogenicity.

The authors conclude that while in silico predictors are useful for prioritization, they should not be used in isolation, and functional transcript assessment remains essential for definitive variant classification in the context of AIS.