The synthesis of a divalent ATP analogue [3,3'-dithiobis[3'(2')-O-[6-(propionylamino)hexanoyl]adenosine 5'-triphosphate] (bis22ATP)] is described in which two molecules of ATP are linked via esterification of their 3'(2')-hydroxyls to the linear dicarboxylic acid 3,3'-dithiobis[N-(5-carboxypentyl)-propionamide] [[HO2C(CH2)5NHC(O)(CH2)2S-]2]. This linkage introduces 22 atoms (a maximum of approximately 2.8 nm) between the ribose oxygens of two ATP molecules. Myosin subfragment 1 (SF1) or heavy meromyosin (HMM) readily cleave bis22ATP to bis22ADP. Upon subsequent addition of excess vanadate ion, both enzymes are rapidly inactivated by formation of a stable vanadate-bis22ADP complex at the active site. By adjustment of the reaction conditions, dimers of SF1 or HMM, both cross-linked with bis22ADP-vanadate, could be prepared. Dimers of SF1 could be separated from monomers by sucrose gradient centrifugation but not by gel filtration. These observations imply that the average Stokes radius of the dimer approximates that of the monomer, a result predicted only for monomers linked approximately side by side. Conversely, dimers of HMM were separated from HMM monomers by gel filtration, reflecting an increase in their Stokes radii. This increase, however, prevented resolution of HMM dimers from monomers by sucrose gradient centrifugation. These results and the molecular dimensions of bis22ATP suggest that the 3'-(2')-hydroxyl of ATP is no more than 1.3 nm from the surface of myosin and suggest further in the simplest interpretation that the active site is most likely located near the middle of the heads of myosin. Analytical sedimentation velocity experiments were performed in order to compare the sedimentation coefficient (s0(20),w) of the SF1 dimer formed by cross-linking to values predicted from ellipsoidal models of the dimer. The observed s0(20),w of the dimer was much closer to the range predicted for a side-to-side arrangement of SF1 monomers than the range predicted for two monomers linked end to end, a result consistent with the active site location suggested above. During the course of these experiments, unmodified SF1 was used as a control, and its sedimentation behavior was reexamined. We have corroborated the finding that the s0(20),w displayed by SF1 can be affected to a limited extent by the particular experimental parameters employed during centrifugation [Morel, J. E., & Garrigos, M. (1982) Biochemistry 21, 2679-2686].(ABSTRACT TRUNCATED AT 400 WORDS)