With the recent Church and Zhang papers in science showing the feasibility and reasonably high efficiency of using CRISPR (i.e. Cas9 + RNA) in human cells, it seems like CRISPRs could emerge as the newest alternative for generating DSBs/genome engineering/gene therapy. I'm curious though - how good are the current alternatives (TALENs, ZFNs, homing meganucleases, RNAi) in terms of efficiency, specificity, toxicity, and "ease-of-use" (iirc, TALENs are awfully hard to synthesize because of their repetitive domains)?

I think the consensus (at least what Church thinks, I saw him talk on Tuesday) is that:
-Homing meganucleases are enormous and old-school
-ZFNs are compact (28:1 for input DNA:target) but obviously hard to target, so they may have specialized applications
-TALENs can be synthesized solid-phase, but they're pretty huge (I think it was either 100 or 1k:1). The modularity is an improvement over ZFNs, which is why they were in fashion
-CRISPR are of course the best (1:1), although the extra machinery may be tough to deliver. Also, I'm curious as to the packing of the RNA--I feel like that might pose a problem for delivery?

In the end, an adenovirus can only carry about 4k, so CRISPRs are the most valuable, I think...until the next big thing.

I haven't studied enough about CRISPRs just yet, but the big questions for me revolve around mechanism, especially given their dependency on RNA secondary/tertiary structure. How do these things work?

Interestingly, there are crystal structures for a bunch of the Cas(1-3) proteins, and an EM structure for Cascade, which I'd like to look into.

I do think the delivery issue is the primary one. None of the Cas/Cascade mechanisms are found in eukaryotes, and it takes quite a bit of work to get them there.

Re: RNA - You can simply throw your desired RNA sequence behind a U6 RNApolIII promoter and stick that on the same vector as your Cas9. The real advantage of CRISPR here is the multiplexability (and ease of "customizing") - it's (theoretically) easy to do a "one-pot" (or more like one plasmid) reaction to edit multiples sites in parallel (how multiplexable these RNAs are remains to be seen - Feng Zhang shows only 2 RNA edits). One question for RNA targeting though is whether to use the full precrRNA-tracrRNA array or simply 1 chimeric RNA molecule that contains both; Zhang's paper shows that using the full complex enables much higher efficiencies whereas Church's (and Doudna's MS) seem to indicate that one long chimeric RNA is sufficient.

Re: Eukaryotes, specifically in terms of "None of the Cas/Cascade mechanisms are found in eukaryotes, and it takes quite a bit of work to get them there" - that's the beauty of Cas9; in this one protein, you have the helicase, DNA/RNA binding, and nuclease all in a single package, whereas in the other systems, you have multiple Cas proteins serving individual roles. Also, I think Keith Joung is about to publish a Cas9 in zebrafish paper; there are some folks at Northwestern playing around with Cas9 in yeast systems as well. Obviously, it remains to be seen if there might be toxicity (Church seems to think not), or how high the efficiency levels one can achieve, but it looks really promising.

I think there are really 2 outstanding things to solve:
1. PAM-limitations: Cas9 - and other Cas systems - can target almost any generic sequence using RNA but there's one constraint - PAM (protospacer adjacent motif), which is small nucleotide motif that dictates where Cas can bind, in this case, Cas9 can bind any sequence that precedes NGG.

It's actually quite amazing how much specificity this 2-base GG PAM motif confers [sciencemag.org I'm really curious which domains on Cas9 mediate this PAM binding. Would be interesting to do some site-directed mutagenesis (or even PACE-style experiments) to engineer Cas9 variants to be fully unbiased to target any arbitrary sequence (even though that would kill your added PAM specificity).

2. Specificity: Currently, Cas9 is limited to 12-bp off the RNA + GG PAM motif, so effectively ~14bp of specificity. At least for human genome modification, this is so-so - Church shows this enables unique targeting of ~40.5% [arep.med.harvard.edu.

As comparison, Sangamo's current Phase II HIV drug SB-728 has 24-bp specificity (chimeric FokI with 2 pairs of 4 zinc fingers).

Can you similarly create Cas9-chimeras? (Church mentioned this to me, but it seems difficult given how big/multifunctional Cas9 is, compared to the modularity of ZFs/TALEs - maybe you could fuse 2 Cas9's together and knockout the nuclease activity on one of them and use two guide RNA molecules to target a 24-bp sequence; seems kinda ghetto though...)

Three more Cas9 papers published in Nature Biotech yesterday: one in zebrafish [nature.com ], one in bacteria (S. pneumoniae, E. coli) [nature.com, and another one in human cells (scooped...) nature.com