The crystal structure of the beta(2) homodimer of Vibrio harveyi lucif
erase has been determined to 2.5 Angstrom resolution by molecular repl
acement. Crystals were grown serendipitously using the alpha beta form
of the enzyme. The subunits of the homodimer share considerable struc
tural homology to the beta subunit of the alpha beta luciferase hetero
dimer. The four C-terminal residues that are disordered in the alpha b
eta structure are fully resolved in our structure. Four peptide bonds
have been flipped relative to their orientations in the beta subunit o
f the alpha beta structure. The dimer interface of the homodimer is sm
aller than the interface of the heterodimer in terms of buried surface
area and number of hydrogen bonds and salt links. Inspection of the s
ubunits of our structure suggests that FMNH(2) cannot bind to the beta
(2) enzyme at the site that has been proposed for the alpha beta enzym
e. However, we do uncover a potential FMNH(2) binding pocket in the di
mer interface, and we model FMN into this site. This proposed flavin b
inding motif is consistent with several lines of biochemical and struc
tural evidence and leads to several conclusions. First, only one FMNH(
2) binds per homodimer. Second, we predict that reduced FAD and ribofl
avin should be poor substrates for beta(2). Third, the reduced activit
y of beta(2) compared to alpha beta is due to solvent exposure of the
isoalloxazine ring in the beta(2) active site. Finally, we raise the q
uestion of whether our proposed flavin binding site could also be the
binding site for flavin in the alpha beta enzyme.