MABEL: Mutagenesis with blunt-end ligation
This is broadly similar to the Stratagene QuickChange protocol but simpler and easier. So far it seems to be working pretty reliably.
1. Design two divergent non-overlapping primers, one forward and one reverse, centred on the offending restriction site. One of them obviously must include the base which is being mutated, preferably at or near the 5' end so that the mismatch does not affect annealing too much. The primers can be quite short (17 bases or so) since they don't overlap or include any non-complementary tails. The 5' ends of the primers must be adjacent to each other (on opposite strands, of course) so that the PCR product will include every base of the vector plus insert.
2. Perform PCR using a proof-reading polymerase. I recommend Kod from Novagen (Merck); it is as accurate as Pfu and 4 times faster, which is handy since this procedure requires you to amplify the entire vector plus insert. Use a very minimal amount of the plasmid as a template.
3. Check the PCR product on a gel. It does not matter if there are mutiple products, since any short wrong PCR products will not be recovered.
4. Purify the PCR product from the reaction mixture.
5. Set up a ligation as follows:
- 6 microlitres water
- 1 microlitre PCR product
- 1 microlitre T4 DNA ligase buffer (with ATP)
- 1 microlitre T4 DNA ligase (Promega)
- 1 microlitre T4 polynucleotide kinase (Promega)
TOTAL 10 microlitres. Incubate at 16 C overnight.
This is the cunning part - the polynucleotide kinase will work fine in the ligase buffer, which contains ATP anyway since the ligase needs it. The PNK will add phosphate to the 5' ends of the blunt PCR product generated by the proof-reading enzyme, making it ligatable, and since the DNA concentration is low, self-ligation will be favoured over intermolecular ligation. When you transform E. coli with the reaction mixture, most of your clones (about 5 out of 6 in our experiments so far) should be the mutated product - just miniprep 3 or 4, and do a quick restriction digest and gel to check that the site has gone. Of course, you should then confirm by sequencing that the site has been cleanly removed with no reading frame shifts at that position.