The thermal denaturation of immobile four-way DNA ("Holliday-") junctions with 17 base pair arms was studied via fluorescence spectroscopic measurements. Two arms of the molecule were labeled at the 5'-end with fluorescein and tetramethylrhodamine, respectively. Melting was monitored by the fluorescence intensity of the dyes, the fluorescence anisotropy of tetramethylrhodamine, and Forster resonance energy transfer (FRET) between fluorescein and rhodamine. To fit the thermal denaturation curves of the four-way junctions, two basic thermodynamic models were tested: (1) all-or-none transitions assuming a molecularity of one, two, or four and (2) a statistical "zipper" model. The all-or-none models correspond to reaction mechanisms assuming that the cooperative melting unit (that is, the structure changing from complete helix to complete coil) consists of (1) one arm, (2) two neighboring arms (which have one continuous strand common to the two arms), or (3) all four arms. In each case, the melting of the cooperative unit takes place in a single step. The tetramolecular reaction model (four-arm melting) yielded unrealistically low van't Hoff enthalpy and entropy values, whereas the monomolecular model (one-arm melting) resulted in a poor fit to the experimental data. The all-or-none bimolecular (two neighboring arm model) fit gave intermediate standard enthalpy change (Delta H) values between those expected for the melting of a duplex with a total length between the helix lengths of one and two arms (17 and 34 base pairs). Simulations according to the zipper model fit the experimental curves best when the length of the simulated duplex was assumed to be 34 base pairs, the length of a single strand. This suggests that the most important parameter determining the melting behavior of the molecule is the end-to-end distance of the strands (34 bases) rather than the length of the individual arms (17 base pairs) and that the equilibrium concentration of partially denatured intermediate states has to be taken into account. These findings are in good agreement with results obtained for three-way DNA junctions (Stuhmeier, F.; Lilley, D. M.; Clegg, R. M. Biochemistry 1997, 36, 13539). An interesting result is that the extent-of-melting curves derived from the fluorescence intensity and anisotropy nearly agree, whereas the curve derived from the FRET data shows a change prior to the melting. This may be an indication of a conformational change leaving the doublestranded structure intact but changing the end-to-end distance of the different arms in a way consistent with the transition to the extended square configuration (Clegg, R. M.; Murchie, A. I.; Lilley, D. M. Biophys. J. 1994, 66, 99) of this branched molecule.