hypertranslocation ͉ hybrid ͉ stability ͉ topological lock T ranscription lies at the heart of cellular gene expression. Unlike replication, transcription is not distributive; that is, any dissociation of an elongation complex is a terminal event. Consequently, elongation complexes must be highly stable while transcribing at up to several hundred bases per second. Recent studies demonstrate, however, that elongation is not a uniform process, with reasonably long-lived, sequence-dependent pauses occurring stochastically (1-4). Indeed, the elongation phase is well known to be subject to regulation through attenuation: sequence-dependent signals leading to pause, arrest, slippage, or termination (5). Elongating RNA polymerases must hold the DNA template and the RNA product tightly enough to be highly stable through nonterminating pauses, yet they must be able to dissociate the complex in response to specific sequences in the DNA. Understanding this interplay requires understanding mechanisms of dissociation.In classic rho-independent or ''intrinsic'' termination, dissociation is thought to occur as the polymerase slows in response to sequence, and structure begins to form concurrently in the nascent transcript (6-8). A run of encoded U's has been proposed both to slow transcription and to weaken the RNA-DNA hybrid. Recently, it has been proposed that, rather than direct dissociation of the nascent transcript from a complex halted at the primary termination site (with or without allosteric assistance), dissociation occurs primarily from forward-translocated states (9-11). In forwardtranslocated (hypertranslocated) states, the hybrid is shortened, weakening the binding of the RNA to the complex. Additionally, shortening of the hybrid may also serve to remove a topological locking of the RNA onto the template strand. As illustrated in Fig. 1, dissociation of the RNA transcript would be substantially faster from forward translocated states than from the initially paused state. Although forward-translocated states are expected to be energetically disfavored relative to the initial on-pathway state and so would be expected to exist at lower populations, the difference in the kinetics of dissociation could compensate such that the dominant dissociative pathway proceeds via forward translocated states.The model shown in Fig. 1 predicts certain behaviors. In general, forward translocation requires melting of the DNA duplex downstream and dissociation of the most upstream base(s) in the hybrid. These costs are partially offset by the energy gained by collapse of a base pair in the DNA at the upstream edge of the bubble. Thus, one would expect that altering the balance between these effects could shift the distribution toward more forward-translocated states and therefore toward or away from dissociation. As demonstrated previously, a direct crosslink between the DNA strands downstream of the halt site should prevent forward translocation and is in fact seen to reduce termination from an intrinsic terminator (9).We propose t...