June 08, 2016 at 05:22AM

Today I Learned: 1) Ligase Cycling Reaction (LCR) is a cloning method that's sort of like Gibson cloning but without the whole recombinase part -- it's all done with DNA and ligase. It works by "stapling" two strands of DNA together, blunt end, using a short oligo bridging the two. I think the best way to grok LCR is probably to look at the abstract figure here (http://ift.tt/1t8c5Wv), but I'm not sure if that's behind a paywall or not, and I'd like to take a crack at explaining it in words. So. Here's how it works. Start with two pieces of (double stranded) DNA that you want to glue together, call them A and B. Add a short single stranded DNA oligo that matches A on the 5' end and B on the 3' end (or vice versa). Mix these with some DNA ligase. Cycle pretty much just like normal PCR -- heat to denature A and B into single-stranded DNA, cool down to re-anneal the DNA, then raise the temperature a little bit for efficient ligation (with a thermostable ligase). When you anneal, some of the A and B strands are going to hybridize to the bridging oligo instead of to their full-size reverse complements. If a strand of A and a strand of B hybridize to the same bridge, then the ligase can glue them together, giving you the full-size DNA strand you want. Keep cycling -- now the one full-size fragment can bridge the reverse complement strand, giving you a nice blunt-end ligation. I'm pretty sure ligase is more expensive than Gibson mix, so you might not want to use this if you're already doing Gibson cloning. Then again, you don't need giant oligos for LCR, so maybe it's cheaper sometimes...? According to the authors of the paper linked above, LCR and Gibson have virtually identical accuracy and efficiency when ligating a small number of pieces, but LCR is good for much larger assemblies (12+ pieces). Note: not to be confused with Ligation Chain Reaction (also LCR), which is a variant of PCR where the "primers" are just the front and back havles of the target, and you just ligate together instead of actually doing any replication. 2) Important news for molecular biologists out there using Phusion polymerase -- don't! According to every source I can find, Q5 polymerase is strictly better. It runs faster, it's twice as accurate, it handles high GC content better, and it's oh-so-slightly cheaper. Just do Q5. 3) ...a little more about V(D)J recombination, a key (and awesome) part of the adaptive immune system. This one's a long one. A little background -- the adaptive immune system works by producing lots (millions? I'm guessing tens of millions to hundreds of millions) of B cells which each produce a different, randomly antibody. By expressing bajillions of different antibodies, the body is likely to have at least one antibody for any foreign invader. When a B cell happens to produce an antibody that binds to an intruder, it undergoes crazy-fast replication and starts producing a ton of the antibody, which either directly kills the foreign thing, targets it for clearance by more active cells, or both. Here's the problem -- the immune system needs to produce millions and millions of different antibody variants, but it doesn't have millions and millions of genes. Furthermore, every B cell has to produce a random antibody, distinct from its sisters, but consistent over time. How? The answer, as you might be able to guess from the pedagological format of this particular TIL, is V(D)J recombination. Here's the one sentence version of V(D)J recombination: Antibody protein genes have a bunch of copies of slightly different variants of a few different domains, and each B cell randomly splices together a few of these at the DNA level to make a unique combination of domains with unique binding properties. This is really unusual because it's a *DNA-level* modification -- for the most part, Eukaryotes don't like editing their DNA except during reproduction. It tends to lead to cancer, for one thing (more on that at the end). But it's important to have a functioning adaptive immune system, too, so it's worth it in this particular instance. Today I learned a bit more about *how* V(D)J recombination works. Perhaps unsurprisingly for a recombination process, it involves lots of recombination enzymes. Specifically, V(D)J recombination is performed by a complex complex (as in, a hard-to-understand collection, not a typo) of enzymes called VDJ recombinase. These are enzyme names I can get behind. Anyway. VJD includes, among other things, one of the enzymes involved in non-homologous end joining repair, some nucleases, at least one DNA ligase, and two genes called RAG 1 and RAG 2 which target V(D)J recombination. The RAGs recognize a pair of DNA motifs 7 and 9 nucleotides long, separated by a constant-length sequence, which tells VDJ recombinase where to clip. When it does, it clips in a really weird way that produces a blunt end on the side of the cut with the coding antibody subunit while ligating *together* the two strands on the other side of the cut, converting that piece of DNA into a hairpin. Recognition domains are found at the end of each coding subdomain; by randomly binding to and cutting at two domains, VDJ recombinase can splice together two random domains. The coding regions are ligated together (presumably with that non-homologous-end-joining enzyme), and the bit in the middle is spliced together in a *single-stranded circle* of loose DNA, which is pretty unusual. There's some evidence that these circles can sometimes re-integrate into the genome, which is a good way to get cancer... way to go, immune system.... Speaking of getting cancer, here's something to consider. As far as we know, VDJ recombinase recognizes its targets using a 7-nucleotide recognition site and a 9-nucleotide recognition site. That's effectively a 16 bp recognition site, which means it should occur, on average, once every 4^16 ~ 4 billion base pairs. Now that's not particularly common (once or twice in a human genome, assuming a perfectly random human genome), but the recognition site isn't perfect. It's a bit fuzzy. We would expect sequences that are *close enough* to the recognition sequence to initiate V(D)J recombination to occur much more frequently. In fact, a quick (read: not that quick) check shows that there are about 250 occurrences of sequences within one base pair mismatch of the VDJ recognition sequence. So how does VDJ recombinase avoid snipping the chromosome where it really shouldn't? The answer may be "it doesn't". Today I also learned that cancers tend to be heavily enriched for active immune system genes. That's thought to largely be due to the fact that the immune system does try to kill the cancer, so cancers tend to have a lot of healthy immune cells mixed in with them. Ready for a bit of pure speculation? Perhaps cancers are more likely to arise when immune system lineage genes are active, because VDJ recombination can screw up genomes in addition to its usual function. I'll bet Virginia Rutten can give a more complete explanation of V(D)J recombination, and/or correct any misunderstandings I've unintentionally propagated, so hopefully she's watching if you have any questions!