Bitesize Bio is shiny!

I just discovered Bitesize Bio, an awesome site full of discussions of common molecular-bio lab questions and problems. It has an RSS feed but is also set up with categories and menus for non-blog-style browsing. The ~5 posts I've read have all been informative, well-written, and interesting. I got a couple about interesting new techniques, a couple about theoretical questions and their practical consequences, and a couple about being a good grad student or a good mentor. It's really shiny, especially for a new lab member like me who wants to know the reasoning behind the magical incantations we sometimes do. "Wash with 0.75mL Buffer PE"? What does Buffer PE even do? ...OK, that's a cheating example -- Buffer PE is a proprietary mix from a commercial kit (but I'm told that after you add ethanol to it, as you must, it's just an improvement on straight ethanol).

For example, I just read an article about touchdown PCR, a neat hack on traditional PCR.

Sometimes in a PCR reaction you'll have trouble with the primers binding in incorrect places, because there happens to be some random sequence in your sample that's sorta-kinda complementary to your primers. Then you get nonspecific amplification of random crap, which can drown the gene fragment you actually wanted to amplify. What do you do?

You can calculate the optimal annealing/melting temperature (Tm) of your primers, and make sure to do your annealing step at that temperature. But random salts and stuff in your reaction can affect the Tm, so the calculated Tm is only an approximation. And at any temperature reasonably close to the Tm, even if it's not optimum, some annealing will happen. Maybe not much, but some. That's thermodynamics for you.

The idea of touchdown PCR is that there's a sweet-spot temperature where, statistically, it's too hot for nonspecific annealing, but just cool enough that the correct annealing can happen. You can't know exactly where this temperature is -- and you wouldn't want to run your whole reaction at that temperature anyway, because you'd only get a small amount of primer annealing and your yield would be low. So instead, for your first cycles you use an annealing temperature significantly higher than the calculated Tm for your primers (about 10 C higher). Then, in subsequent cycles, you gradually lower the annealing temperature. So in early cycles, only a few primers will anneal, and they'll almost all anneal to your actual sequence of interest, and not to random other sequences that are close-but-a-little-off. By the end of the cycle, you've amplified your correct sequence by a little bit relative to incorrect sequences. Run through several more cycles, and by the time you get to a "standard" annealing temperature where nonspecific annealing can happen, you've hopefully amplified the correct sequence quite a bit already, so nonspecific annealing becomes less of a problem because there are just more copies of the correct sequence available for the primers to bind to.

When I read about that, I was blown over. It's such a clever thermodynamics hack! -- it takes advantage of the fact that chemical reactions (DNA base-pairing or anything else) have fuzzy, stochastic behavior. For a given reaction, there isn't a sharp cutoff temperature where it goes from "too hot to react" suddenly to "OK now reaction proceeds fully". It's fuzzy. If it's too hot, a few molecules will react. Get a bit closer to the optimal temperature, and more molecules go. If you have two competing reactions with slightly different optimal temperatures (like specific and nonspecific annealing!), it's like having two bell curves overlapping, centered at slightly different values. You can find a value where one bell curve is acceptably high and the other is very low -- and then once you've amplified your chosen sequence N-fold, the game changes and you don't have to worry about the incorrect sequences nearly as much. It's like magic!

References (from original post):
1. Roux KH. Genome Research. 1995. 4: S185-194.
2. Mattick JS et al. Nature Protocols. 2008. 3(9). 1452 - 1456