USE OF A NEURAL-NETWORK SECONDARY STRUCTURE PREDICTION TO DEFINE TARGETS FOR MUTAGENESIS OF HERPES-SIMPLEX VIRUS GLYCOPROTEIN-B

Herpes simplex virus glycoprotein B (HSV gB) is essential for penetrat ion of virus into cells, for cell-to-cell spread of virus, and for cel l-cell fusion. Every member of the family Herpesviridae has a gB homol og, underlining its importance. The antigenic structure of gB has been studied extensively, but little is known about which regions of the p rotein are important for its roles in virus entry and spread. In contr ast to successes with other HSV glycoproteins, attempts to map functio nal domains of gB by insertion mutagenesis have been largely frustrate d by the misfolding of most mutants. The present study shows that this problem can be overcome by targeting mutations to the loop regions th at connect alpha-helices and beta-strands, avoiding the helices and st rands themselves. The positions of loops in the primary sequence were predicted by the PHD neural network procedure, using a multiple sequen ce alignment of 19 alphaherpesvirus gB sequences as input. Comparison of the prediction with a panel of insertion mutants showed that all mu tants with insertions in predicted alpha-helices or beta-strands faile d to fold correctly and consequently had no activity in virus entry; i n contrast, half the mutants with insertions in predicted loops were a ble to fold correctly. There are 27 predicted loops of four or more re sidues in gB; targeting of mutations to these regions will minimize th e number of misfolded mutants and maximize the likelihood of identifyi ng functional domains of the protein. (C) 1998 Elsevier Science B.V. A ll rights reserved.