Bench philosophy: Genome Editing for Everybody
by Steven Buckingham, Labtimes 01/2014
For years, genome editing has been a tedious and tough business that, especially in mammalian cells, requires a high level of frustration tolerance. With the new CRISPR-Cas system, however, it is suddenly stress-free.
Editing the genome got even easier over the last year with the arrival of a new technique: CRISPR-Cas. CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeats, and it operates in conjunction with Cas, a DNA-cutting enzyme. The reason for the excitement is that it allows vastly simpler and more convenient DNA editing than ever before. Since its announcement about this time last year, it has provoked a deluge of publications, so intense is the excitement.
Of course, there have been a few genome-editing methods around for a little while now. They all follow the same basic strategy: use a restriction endonuclease to cut the DNA and provide the DNA-repair machinery with your own template that gets stitched in to fix the cut. What sets the various genome editing approaches apart is how you get the ligase to cut where you want it to. Two popular choices were Zinc Finger Nucleases (ZFNs) and, more recently, TALENs (Transcription Activator-Like Effector endoNucleases). ZFNs are artificial enzymes that combine a general endonuclease with a protein that directs the enzyme to the site you want to edit. TALENs are similar, except that you fuse the endonuclease with a TAL-effector binding domain: a stretch of amino acids with a highly conserved sequence, except for a couple of residues near the middle that recognise specific nucleotides.
To understand why CRISPR is so much easier than either of these, we need to look at how it works. The CRISPR mechanism is a sort of bacterial immune system. The CRISPR locus on the bacterial genome consists of a series of short, conserved repeat sequences separated by highly variable sections called spacers. When a bacterium identifies suspect foreign DNA, such as that from a phage, the invading DNA is cut up into regularly-sized pieces and somehow incorporated into the spacers in the CRISPR locus. That way, the bacterium “remembers” invading DNA for future reference, inviting comparisons with the vertebrate immune system.
There are three slightly different types of CRISPR found in different types of bacteria but they all work in similar ways. In the best-studied form of the system (type II), the CRISPR spacers are used as a template to make small RNA strands (called crRNAs) that form the basis of a retaliatory strike on the foreign DNA. For the attack to work, the crRNAs depend on two other elements being present: the Cas endonuclease that does the DNA cutting and a “trans-acting RNA” (trRNA) that adapts the crRNA to the Cas.
The crRNA binds to the matching strand on the foreign DNA and guides the Cas endonuclease to the site. The Cas has multiple DNA-cutting enzyme domains, one of which cuts one strand while the other unzips the DNA and cuts the opposite strand. There is only one small catch (which can actually be a feature, rather than a bug – see below): the Cas only binds if the crRNA complementary sequence is present, along with a “protospacer associated motif” (PAM) right next to it.
So much for the native mechanism. To adapt this to the job of genome engineering, you introduce the CRISPR components into your target cells. Once again, nature has provided us with yet another boon. Although the bacterial system needs all three enzymes (crRNA, trRNA and Cas), in the lab, cells are perfectly happy if you fuse the two RNAs together in tandem to produce “single guide RNA” (sgRNA).
Once you have made your double-stranded break (DSB), you have to decide how you are going to hijack the repair mechanism to your own ends. There are two main alternatives: Non-Homologous End Joining (NHEJ) or Homologous Recombination (HR). NHEJ leaves the repair machinery to do its job unmolested, knowing that it does so in a noisy, error-prone way, accidentally introducing insertions or deletions. This puts the sequence out of frame, resulting in a loss of function.
For HR, you supply the oligonucleotide template for the sequence you want to insert, be it a long reporter gene or just a single mutation. However, you can usually only use HR on actively dividing cells. To get a bit inserted, you would use double-stranded DNA, whereas for point mutations you would use single-stranded DNA oligos.
A variant on the CRISPR technique uses Cas with an aspartate-to-alanine mutation (D10A), which only nicks the DNA (i.e. cuts just one strand), making it easier to insert a point mutation without the risk of an accidental insertion/deletion. Or you can be really smart and use a pair of sgRNAs designed to hit either strand – a pair of nicks reduces the danger of off-target effects, due to imperfect specificity in either of the sgRNAs.
Cas activity demands the presence of a PAM site immediately next to the crRNA-complementary sequence. At first sight, this might look like a disadvantage but in the case of the CRISPR-Cas system from Streptococcus pyogenes, the PAM sequencehappens to be just NGG, which statistically will occur about once every eight base pairs, so this is not really much of a restriction. Systems from other species can be longer, which again appears to be a disadvantage but it does mean there will be fewer off-target effects, because of the lower probability of accidental binding of the Cas complex.
CRISPR can be used at all levels: you can transform cells in culture by transfecting them with message-encoding Cas and the sgRNA, then use genotyping or phenotyping to pick out clonal cell lines. Or you can use it to generate mutant animals by injecting constructs into the germ line.
So what makes CRISPR so much handier than older methods? One can sum up most of the answer in one word: modularity. If you want to edit a genome using Zinc finger or TALENs, you have to engineer a new fusion enzyme for every editing task. With CRISPR on the other hand, all you have to adjust for each task is the crRNA and that comes down to simple plasmid construction. It all comes from having two separate components.
The modularity brings lots of other advantages, too. For one, it means it is easy to multiplex – edit different sites at the same time. All you have to do is to supply a crRNA plasmid for each change you want to make, along with the Cas, of course. We will leave the reader to exercise their imagination but for starters, you can insert a mutation at one site and a reporter gene at another, for instance.
But the power of CRISPR goes beyond the convenience of its modularity. It opens up editing on species that are recalcitrant to RNAi, such as zebrafish. It also has very high germline transmission – reaching 100% in Drosophila and zebrafish.
So, given the convenience and versatility offered by the CRISPR-Cas system, it comes as little surprise that some very active imaginations have already been at work. Several labs have come up with some ingenious adaptations. Quite a few start with an enzymatically inactive Cas – “dead Cas” that still binds to the right sequence but does not cut the DNA. So instead, you can fuse the Cas with an effector, which turns on gene expression – so-called CRISPR-on. A sort of inverted RNAi.
Another adaptation also involves using dead Cas – but this time you turn genes off by designing your sgRNA, so that the Cas complex occupies (and so blocks) the promoter of a gene or, more efficiently, fuses the dead Cas with a transcription repression domain.
One group showed that you can insert loxP sites into the mouse genome to get conditional expression (Yang et al. 2013, Cell 154: 1370-1379). Admittedly, you can do this with Zinc Finger Nucleases but it is a lot more work that way. And Robin Ketteler at University College London, UK has pointed out that the technologies successfully developed for the construction of shRNA libraries for high-throughput, genome-wide screens could easily be adapted for similar screens based on CRISPR (Frontiers in Genetics 2013, 4: 1-6). He was right. As I write, the week’s issue of Nature (9 January 2014) contains two reports of genomic screens using CRISPR.
But are there any potential pitfalls with CRISPR? Yes. First, there is the ever-present hazard of off-target effects. These could be disastrous: cell death resulting from spurious indels would, at least, alert the experimenter that an innocent gene has been hit but more subtle transformations could introduce stealthy errors into the results. And you have a bit of optimising to do when it comes to the concentrations – the amount transfected, as well as the ratio between Cas and sgRNA, has to be carefully titrated to avoid off-target effects. Having said that, when Rudolf Jaenisch’s group at the Whitehead Institute for Biomedical Research made their conditional loxP mice, even though they multiplexed, they found little evidence of off-target mutations (Yang et al., Cell, 9;153(4):910-8). Ann Ran et al. report that more than three mismatches between the crRNA and the DNA are not tolerated, although other reports put the figure as high as six (Nature Protocols 8: 2281-2308).
So clearly, the technique is still too new on the scene for its downsides to be fully assessed. But it is clear that in CRISPR we have a significant new tool in our editing toolbox. Is it a replacement or an add-in? Ask me in five years’ time.
Last Changed: 05.02.2014