Previous work indicates that one subunit of the AraC protein dimer bin
ds to a DNA target araI(1), of 17 base-pairs. We systematically substi
tuted every base-pair in a synthetic araI(1) target with the three pos
sible alternatives and then tested binding of a araI(1) and of these 5
1 DNA targets to AraC by quantitative gel shift analysis in the presen
ce of L-arabinose. We found that every substitution of the underlined
bases reduces AraC binding tenfold or more: 5' TAGCATTTTTATCCATA 3'. S
ubstitutions at other bases have little or no effect. In the absence o
f L-arabinose we observed a sixfold reduction of binding of AraC to ar
aI(1). We have designated the 5' AGC sequence the A-box and the 5' TCC
ATA sequence the B-box. We synthesised DNA targets containing either t
wo A or two B-boxes with the natural araI(1)-I-2 spacing. Wild-type Ar
aC binds both targets in the presence of L-arabinose in a gel shift ex
periment. In the absence of L-arabinose, AraC binds only to the double
B-box. We then tested various AraC mutant proteins in the same way. S
208A and H212A bind to the double B-box but not to the double A-box in
the presence or absence of L-arabinose. D256A binds to the double A-b
ox, but not to the double B-box, in the presence of L-arabinose but no
t in its absence. The implications of these results for the mechanism
of AraC induction by L-arabinose are discussed. (C) 1996 Academic Pres
s Limited