120 Pingoud, Alves, and Geigerthe recognition site is used by the enzyme to differentiate between spe-
cific and nonspecific DNA (118). For example, EcoRV needs the
presence of the thymin methyl groups in its recognition sequence
-GATATC- for a productive interaction, whereas EcoRI accepts
oligodeoxynucleotides as substrates that contain deoxyuridine in place
of thymidine in the recognition sequence -GAATTC- (83, 87,119-121).
Studies with modified oligonucleotides have also shown that
isoschizomers do not necessarily have the same structural require-
ments for DNA cleavage (89,107,122). Furthermore, these studies have
demonstrated that some restriction enzymes fail to cleave DNA modi-
fied by methylation at a position in the recognition sequence other than
the one modified by the corresponding modification methyltransferase.
For example, EcoRV does not cleave DNA containing the sequences
-Gm6ATATC - or -GATm6ATC - (87), although only the former one is
formed in vivo by the action of the EcoRV methylase (123).
The most detailed insight into the understanding of sequence speci-
ficity has been provided by the X-ray structure analysis of an EcoRI-
oligodeoxynucleotide complex (56). This complex was crystallized in
the absence of Mg 2÷, the essential cofactor for catalysis and, therefore,
does not represent a bona fide enzyme substrate complex. Neverthe-
less, the structure analysis provides sufficiently detailed information
to propose models for the mechanism of recognition. According to the
suggestion of McClarin et al. (56), the specificity of EcoRI is based on
12 precisely directed hydrogen bonds, six from each of the two iden-
tical subunits to the recognition sequence -GAATTC-: Arg200 is sup-
posed to form two hydrogen bonds to guanine; Glu144 and Arg145
supposedly interact with the two adenine residues via two hydrogen
bonds each (see note added in proof at end of chapter). Site-directed
mutagenesis experiments, however, in which these amino acid resi-
dues of EcoRI have been replaced suggest that the cleavage specificity
of EcoRI is determined by factors in addition to direct hydrogen bond-
ing to the bases of the recognition sequence (82,85,124-126). Since
the precise geometry of the phosphodiester backbone is determined by
the sequence, also backbone contacts may contribute to specificity
(127). It seems that sequence discrimination by EcoRI is redundant.
Presumably this is true for restriction enzymes in general, since only
an overdetermined discrimination is fail safe.