Human (HIV-1) and simian (SIV) immunodeficiency virus fusion with the host cell is promoted by the receptor-triggered refolding of the gp41 envelope protein into a stable trimer-of-hairpins structure that brings viral and cellular membranes into close proximity. The core of this hairpin structure is a six-helix bundle in which an inner homotrimeric coiled coil is buttressed by three antiparallel outer HR2 helices. We have used stopped-flow circular dichroism spectroscopy to characterize the unfolding and refolding kinetics of the six-helix bundle using the HIV-1 and SIV N34(L6)C28 polypeptides. In each case, the time-course of ellipticity changes in refolding experiments is well described by a simple two-state model involving the native trimer and the unfolded monomers. The unfolding free energy of the HIV-1 and SIV trimers and their urea dependence calculated from kinetic data are in very good agreement with data measured directly by isothermal unfolding experiments. Thus, formation of the gp41 six-helix bundle structure involves no detectable population of stable, partly folded intermediates. Folding of HIV-1 N34(L6)C28 is five orders of magnitudes faster than folding of its SIV counterpart in aqueous buffer: k(on),(HIV-1)=1.3 x 10(15)M(-2)s(-1) versus k(on),(SIV)=1.1 x 10(10)M(-2)s(-1). The unfolding rates are similar: k(off),(HIV-1)=1.1 x 10(-5)s(-1) versus k(off),(SIV=)5.7 x 10(-4)s(-1). Kinetic m-values indicate that the transition state for folding of the HIV-1 protein is significantly more compact than the transition state of the SIV protein. Replacement of a single SIV threonine by isoleucine corresponding to position 573 in the HIV-1 sequence significantly stabilizes the protein and renders the folding rate close to that of the HIV-1 protein yet without making the transition state of the mutant as compact as that of the HIV-1 protein. Therefore, the overall reduction of surface exposure in the high-energy transition state seems not to account for different folding rates. While the available biological evidence suggests that refolding of the gp41 protein is slow, our study implies that structural elements outside the trimer-of-hairpins limit the rate of HIV-1 fusion kinetics.