Rules of thumb for sequencing cloned PCR products

[Dom Spinella]

Mark:
Your question inspired me to do a litle research.  My lab has been 
cloning PCR products (RT-PCR) derived from ligand binding domains of 
single-chain receptors (cytokine receptors, growth factor receptors, 
etc.).  The PCR products are typically 600 -1600 bp in length.  Our 
standard is to pick 5 clones from each PCR product and sequence them 
all.  To date we have cloned nearly three dozen PCR products in this 
way -- more than 150 templates sequenced.  Because the sequences are 
already published in the literature, we can compare our sequences to 
the published ones to look for Taq misincorporations.

What we have found is that, on average, we need to sequence 3-4 clones 
to find one "clean" one. The "mutations" are typically different in 
each incorrect clone which is our clue that we are truly dealing with 
a Taq misincorporation. Occasionally, we find that all five of our 
clones has the same base substitution relative to the published 
sequence.  In this case, we ascribe the "mutation" to an allelic 
variant, so we don't count it as a PCR error.  Another thing we have 
noticed is that PCR "error" rate varies from template to template.  
Some of our targets are completely clean in every clone, while others 
have a different mutation (sometimes multiple mutations) in each of 
the five clones we sequence.

The bottom line is that I suspect that your rule of thumb estimate of 
the Taq error rate is probably a little low -- and I certainly 
wouldn't trust a single clone to provide an accurate sequence.  One 
thing you can do (assuming that you are cloning only to obtain 
sequence information) is to avoid cloning altogether, and sequence 
directly from your PCR product.  In this way you get a "consensus" 
sequence which is much more likely to be accurate.  The ABI protocols 
are very efficient at providing robust sequencing data from uncloned 
PCR products.

Hope this helps.  Regards.
D.G. Spinella, Ph.D.