J Allergy Clin Immunol. 2025 Jan;155(1):166-175. doi: 10.1016/j.jaci.2024.08.014

PMID: 39182630

Ines Santisteban, Francisco X Arredondo-Vega, Pawan Bali, Busra Dalgic, Hyun Ho Lee, Minsoo Kim, Jake Hermanson, Teresa K Tarrant, Michael S Hershfield

Highlights: This study systematically assessed the pathogenic potential of 46 rare ADA missense variants by expressing them in an Escherichia coli system. The researchers established a "Genotype Category" (GC) scoring system based on expressed enzymatic activity, which proved superior to the AI-based AlphaMissense tool in predicting the pathogenicity of milder variants. The study highlights that pathogenicity is a continuum rather than a binary state. Furthermore, it identifies red blood cell (RBC) dAXP levels at diagnosis as a critical prognostic marker, particularly for patients with "indeterminate" phenotypes identified through newborn screening.

Background: Adenosine deaminase (ADA) deficiency is known to cause Severe Combined Immunodeficiency (SCID). However, the disease spectrum is broad, ranging from fatal SCID to delayed-onset, late-onset, and even partial deficiency in healthy individuals. With the advent of newborn screening (TMS and TREC) and genomic sequencing, clinicians are increasingly identifying asymptomatic infants with "indeterminate" phenotypes and ADA variants of unknown significance (VUS). Current definitions of "pathogenic" or "benign" are often insufficient for these ambiguous cases.

Objective and methods: The study aimed to define the relationship between ADA genotypes, expressed enzyme activity, metabolic accumulation (RBC dAXP), and clinical phenotype. 46 rare missense variants were expressed in the ADA-deficient E. coli strain SØ3834. Variants were classified into Expression Groups (I-IV) based on their activity relative to wild-type (WT) ADA. These were correlated with clinical data from 58 reference patients. Additionally, the study benchmarked its functional grading system against AlphaMissense and used a minigene assay to investigate splicing defects in exon 9.

Material and Methods: ADA missense variants were introduced into plasmid pZC11 and expressed in E. coli. Enzymatic activity was measured using a radiochemical assay and normalized to total protein. Patients were categorized into phenotypic groups (SCID, Delayed onset, Late onset, Partial, Indeterminate). RBC dAXP levels were measured using HPLC. The AlphaMissense (AM) tool was used to score variants, and these scores were compared with the functional results derived from the E. coli expression system.

Results: The expressed activity of the 46 variants ranged from ~0.001% to ~70% of WT activity. A strong inverse correlation was found between total expressed ADA activity (txADA) and disease severity/RBC dAXP levels. Patients with Genotype Category I (GC-I, <0.05% activity) universally presented with SCID and high dAXP. In contrast, GC-IV genotypes were associated with milder or indeterminate phenotypes. Crucially, the functional GC scoring system outperformed AlphaMissense; the AI tool incorrectly predicted several mild (Group IV) variants as "likely pathogenic" and failed to identify pathogenic splicing variants in codon 282.

Keywords: Adenosine deaminase deficiency, SCID, genotype-phenotype correlation, deoxyadenosine nucleotides, AlphaMissense, splicing