The human immunodeficiency virus type 1 Tat protein binds to an RNA stem-loop structure called TAR which is present at the 5' end of all human immunodeficiency virus type 1 transcripts. This binding is centered on a bulge within the stem of TAR and is an essential step in the trans-activation process which results in a dramatic increase in viral gene expression. By analysis of a series of TAR derivatives produced by transcription or direct chemical synthesis, we determined the structural and chemical requirements for Tat binding. Tat binds well to structures which have a bulge of two to at least five unpaired bases bounded on both sides by a double-stranded RNA stem. This apparent flexibility in bulge size is in contrast to an absolute requirement for an unpaired uridine (U) in the 5'-most position of the bulge (+23). Substitution of the U with either natural bases or chemical analogs demonstrated that the imido group at the N-3 position and, possibly, the carbonyl group at the C-4 position of U are critical for Tat binding. Cytosine (C), which differs from U at only these positions, is not an acceptable substitute. Furthermore, methylation at N-3 abolishes binding. While methylation of U at the C-5 position has little effect on binding, fluorination reduces it, possibly because of its effects on relative tautomer stability at the N-3 and C-4 positions. Thus, we have identified key moieties in the U residue that are of importance for the binding of Tat to TAR RNA. We hypothesize that the invariant U is involved in hydrogen bond interactions with either another part of TAR or the TAR-binding domain in Tat.