To determine the utility of single-stranded conformation polymorphism (SSCP) analysis for mutation screening in the BTK (Bruton's tyrosine kinase) gene, we investigated 56 X-linked agammaglobulinemia (XLA) families. To obtain genotype/ phenotype correlations, predicted protein aberrations were correlated with the clinical course of the disease. This study included 56 patients with XLA, with or without a positive family history, who were diagnosed on the basis of their clinical features, low peripheral B-cell count, and low immunoglobulin levels. Ten patients with isolated hypogammaglobulinemia and 50 healthy males served as controls. SSCP analysis was performed for the entire BTK gene, including the exon-intron boundaries and the promoter region. Structural implications of the missense mutations were investigated by molecular modeling, and the functional consequences of some mutations also were evaluated by in vitro kinase assays and Western blot analysis. We report the largest series of patients with XLA to date. All but 5 of the 56 index patients with XLA screened with SSCP analysis showed BTK gene abnormalities, and in 2 of the 5 SSCP-negative patients, no BTK protein was found by Western blot analysis. There were 51 mutations, including 37 novel ones, distributed across the entire gene. This report contains the first promoter mutation as well as 14 novel missense mutations with the first ones described for the Tec homology domain and the glycine-rich motif in the SH1 domain. Each index patient had a different mutation, except for four mutations, each in two unrelated individuals. This result supports the strong tendency for private mutations in this disease. No mutations were found in the controls. Our results demonstrate that molecular genetic testing by SSCP analysis provides an accurate tool for the definitive diagnosis of XLA and the discrimination of borderline cases, such as certain hypogammaglobulinemia or common variable immunodeficiency patients with overlapping clinical features. Genotype/ phenotype correlations are not currently possible, making prediction of the clinical course based on molecular genetic data infeasible.