Relationship of splice site information content to severity of hemophilia A and B. Y.v. Kodolitsch¹, I. Schalwat¹, S.R. Sjojanovsky², E. Berger¹, P.K. Rogan². 1) Dept Cardiol, Univ Hosp Eppendorf, 20246 Hamburg, Germany; 2) Sect Mol Genet & Mol Med, Children's Mercy Hosp, Univ Missouri-Kansas City Sch Med 64108.
   Protein reduction and bleeding symptoms vary widely among different mutations in the factor VIII (F8C) or factor IX (F9) genes. Predicting disease severity by mutation analysis may be useful in titrating F8C and F9 levels in replacement therapy. Individual information-based models (R_i) of splicing have been used to predict severity of F8C and F9 splicing mutations, many of which are located within splice sites or create novel splice junctions (71 of 584). Mutant sites with R_i values <2.4 bits or with >100 fold reduction in strength (DR_i>6.6) are inactivated and usually produce severe phenotypes, whereas leaky mutations with decreased splicing have R_i values >2.4 bits, <100 fold decrease in splice site strength, and produce milder symptoms. Of 61 distinct splice variants (n=71), loss of splice site recognition was predicted for 43, decreased splicing for 15, and unaltered R_i values for 3 variants. Cryptic splicing was predicted for 12 of these mutations. The predictions were compared with measures of protein activity and clinical severity. Factor VIII and IX protein activities <1% of normal were graded severe, moderate was between 1 and 5%, and mild was >5%. R_i analysis accurately predicted reduction in protein activity for 50 mutations (70%). Reduction in activity was overestimated for 13 mutations predicted as severe, and was underestimated for 5 mutations predicted to be leaky. There was concordance (18/18) between mutations predicted to be severe and clinical severity based on frequency of bleeding symptoms, but only 5 of 8 leaky mutations were clinically mild, an overall accuracy of 88%. For cryptic splicing mutations, differences between the strengths of cryptic and mutant natural sites diminished residual splicing at the natural site. When these differences were considered, the predictions corresponded to the observed reduction in protein activity for 15 of 18 mutations (10/18 if the cryptic site was not considered). In summary, R_i analyses correctly discriminated 74% of severely affected patients carrying splicing mutations from those with milder disease.