The crystal model of the complex of the somatically mutated anti-p-azophenylarsonate (Ars) Ab 36-71 F(ab) with phenylarsonate reveals that six residues (Asn35, Trp47, Tyr50, Ser99, and Tyr106 in the H chain and Arg96 in the L chain) contact hapten. Further study of this model suggested that H chain Phe108, which forms the base of the combining cavity, also affects Ars binding. We predicted that Trp with a bulkier aromatic side chain might be accommodated in this position and increase Ars affinity. The substitution of Phe by Trp using in vitro mutagenesis at position 108 enhanced affinity 10-fold in the germline-encoded Ab 36-65. However, the same mutation in Ab 36-71 abolished the binding. Phe108 was then mutated to different amino acids in both Abs. The results indicated that except for the Trp substitution in 36-65, all other substitutions at position 108 decrease or abolish Ars binding in both Abs. It was shown previously that the 200-fold difference in affinity between 36-65 and 36-71 could be reproduced by changing only three VH amino acids. Because the mutation of Phe108 to Trp has never been observed during in vivo affinity maturation, we constructed mutants of 36-65 in which Trp108 was combined with one or more of the "favorable" mutations of 36-71, to determine whether the mutations were additive. The results indicate that it is possible to maintain an affinity significantly higher than wild-type by such combined mutations. Thus, the failure to observe Trp108 in vivo is not due to structural idiosyncrasy, but may simply be due to codon usage at Phe108 in the germline sequence. Such limited "adaptability" of a germline sequence indicates that it is possible to achieve higher affinity Abs through protein engineering via routes that are constrained during in vivo selection.