Transcription antitermination in phages and P22 uses N proteins that bind to similar boxB RNA hairpins in regulated transcripts. In contrast to the N-boxB interaction, the P22 N-boxB interaction has not been extensively studied. A nuclear magnetic resonance structure of the P22 N peptide boxB left complex and limited mutagenesis have been reported but do not reveal a consensus sequence for boxB. We have used a plasmidbased antitermination system to screen boxBs with random loops and to test boxB mutants. We find that P22 N requires boxB to have a GNRA-like loop with no simple requirements on the remaining sequences in the loop or stem. U:A or A:U base pairs are strongly preferred adjacent to the loop and appear to modulate N binding in cooperation with the loop and distal stem. A few GNRA-like hexaloops have moderate activity. Some boxB mutants bind P22 and N, indicating that the requirements imposed on boxB by P22 N overlap those imposed by N. Point mutations can dramatically alter boxB specificity between P22 and N. A boxB specific for P22 N can be mutated to N specificity by a series of single mutations via a bifunctional intermediate, as predicted by neutral theories of evolution.and P22 share a closely related system of regulating the expression of early lytic genes by allowing transcription past terminators in the P left and P right operons (47). The recognition of the boxB RNA hairpins of nut (N-utilization) sites by the binding domains of the viral N proteins (Fig. 1A, B, and C) initiates the assembly of an antitermination complex. This complex contains N, host factor NusA, RNA polymerase, and other host factors bound to the viral nut site boxB and the nonhairpin boxA, allowing transcription to proceed through downstream transcription termination signals (32). The four wild-type nut sites of and P22 are similar in sequence but differ even between boxB left and boxB right in the same virus. Notably, both P22 boxBs have a C in the loop where both boxBs have an A, P22 boxB stems are longer than those of by 1 base pair, and P22 boxB right appears to have a noncanonical C:C base pair. boxBs bind noncognate N peptides poorly in vitro (2,8,44). Likewise, noncognate N-nut interactions do not function in vivo (14, 28), and noncognate N proteins do not rescue N-deficient viruses (10).The details of the N-boxB interaction have been revealed by extensive genetic and biochemical work (47) and by solution state nuclear magnetic resonance (NMR) structures of the arginine-rich domain of the N protein bound to boxB left (29) and boxB right (39) RNA. Genetic and biochemical studies of the P22 interaction are less complete (13, 44) but are supported by the solution state NMR structure of the arginine-rich domain of the N protein bound to P22 boxB right (5).and P22 boxBs bind their N peptides as hairpins in which 4 of 5 bases adopt a GNRA-like tetraloop structure (Fig. 2). Tetraloops are frequently found in RNAs serving structural roles as stable caps to stems and as motifs recognized by proteins; GNRA tetraloops are noted ...