A restriction enzyme (or restriction endonuclease) is an enzyme that cuts double-stranded DNA. The enzyme makes two incisions, one through each of the phosphate backbones of the double helix without damaging the bases. The chemical bonds that the enzymes cleave can be reformed by other enzymes known as ligases, so that restriction fragments carved from different chromosomes or genes can be spliced together, provided their ends are complementary (more below). Many of the precedures of molecular biology and genetic engineering rely on restriction enzymes.
Sites of cleavage
Rather than cutting DNA indiscriminately, a restiction enzyme cuts only double-helical segments that contain a particular nucleotide sequence, and it makes its incisions only within that sequence--known as a "recognition sequence"--always in the same way.
Some enzymes make strand incisions immediately opposite one another, producing "blunt end" DNA fragments. Most enzymes make slightly staggered incisions, resulting in "sticky ends", out of which one strand protrudes. There are three known evolutionary lineages of restriction enzyme, which each cleave DNA by a different mechanism.
Fragment complementarity and splicing
Because recognition sequences differ between restriction enzymes, the length and the exact sequence of a sticky-end "overhang", as well as whether it is the 5' or the 3' strand that overhangs, depends on which enzyme produced it. Base-pairing between overhangs with complementary sequences enables two fragments to join or "to be spliced," which they tend to do spontaneously in a test tube. Thus, a sticky-end fragment will readily reunite with the fragment from which it was originally cleaved, but it will also attach to any other fragment generated by the same restriction enzyme; because cuts made by a given type of enzyme always produce identical ends with identical sequences. These rules enable molecular biologists to anticipate which fragments will join and how they will join--and to choose enzymes to produce fragments they can splice. This knowledge represents more or less the essence of genetic engineering.
Restriction enzymes as tools
Recognition sequences typically are only four to twelve nucleotides long. Because there are only so many ways to arrange the four nucleotides--A,C,G and T--into a four or eight or twelve nucleotide sequence, recognition sequences tend to "crop up" by chance in any long sequence. Furthermore, restriction enzymes specific to hundreds of distinct sequences have been identified and synthesized for sale to laboratories. As a result, potential "restriction sites" appear in almost any gene or chromosome. Meanwhile, the sequences of all artificial plasmids include a "linker" that contains dozens of restriction enzyme recognition sequences within a very short segment of DNA. So no matter the context in which a gene naturally appears, there is probably a pair of restriction enzymes that can snip it out, and which will produce ends that enable the gene to be spliced into a plasmid (i.e. which will enable what molecular biologists call "cloning" of the gene).
Recognition sequences are palindromic
While recognition sequences vary widely, all of them are palindromic; that is, the sequence on one strand reads the same in the opposite direction on the complementary strand.
Enzyme Source Recognition Sequence Cut
EcoRI Escherichia coli 5'GAATTC
Examples
3'CTTAAG
BamHI Bacillus amyloliquefaciens 5'GGATCC 5'---G GATCC---3'
3'CCTAGG 3'---CCTAG G---5'
HindIII Haemophilus influenzae 5'AAGCTT
3'TTCGAA
MstII Microcoleus species 5'CCTNAGG
3'GGANTCC
TaqI Thermus aquaticus 5'TCGA
3'AGCT
NotI Nocardia otitidis 5'GCGGCCGC
3'CGCCGGCG
AluI* Arthrobacter luteus 5'AGCT 5'---AG CT---3'
3'TCGA 3'---TC GA---5'