A non-rational approach to optimized targeting of CRISPR-Cas9 complexes

Abstract

In this issue of Cell Chemical Biology, Bush et al.¹ report an in vitro selection method for optimizing CRISPR-Cas9 single-guide RNAs. This approach may be useful in targeting previously intractable genomic sequences. The results also provide insights into which positions in single-guide RNAs are most amenable to modification.

The discovery of CRISPR (clustered regularly interspaced short palindromic repeats)-Cas systems, notably CRISPR-Cas9 from group A Streptococcus pyogenes bacteria, is undoubtedly the premier genetic engineering breakthrough of our generation.² These molecular machines are so tantalizing for many reasons. First, as their name indicates, their genomic loci were first noted simply as odd repetitive sequence patterns in the DNA of microorganisms.³ Second, they turned out to be totally unexpected microbial adaptive immune systems, complete with elaborate storage mechanisms to “remember” past pathogens.⁴ Finally, they are readily harnessed for targeted cleavage of double-stranded DNA (dsDNA) in living cells, triggering DNA repair that can be helpfully mutagenic.^5–7

The ribonucleoprotein DNA cleavage complex in natural CRISPR-Cas systems is assembled through the incompletely understood actions of an intricate set of structural and enzymatic factors encoded at CRISPR loci, accompanied by even more intricate mechanisms for selecting and storing historical targets.⁸ Genetic engineering applications typically simplify the cleavage system to a single endonuclease protein (e.g., Cas9), and the artificial fusion of the essential ~75-nucleotide tracrRNA and the 39–42-nucleotide variable-sequence crRNA “spacer” (targeting) RNA, forming a convenient single-guide RNA (sgRNA).⁵

The CRISPR-Cas9 machine uses free energy released by the binding of a 3–5 base-pair Protospacer Adjacent Motif (PAM) sequence in dsDNA to pry apart the duplex just enough to test the pairing of the terminal crRNA bases with the target DNA strand. Successful pairing propagates further base pairing without additional energy input because RNA-DNA base pairs replace DNA-DNA base pairs as hybridization occurs. A sufficient sequence match cues cleavage of both target DNA strands. Interestingly, the cleaved CRISPR-Cas9-DNA complex persists for a prolonged period.⁹

Although targeted DNA cleavage by CRISPR-Cas9 is remarkable, the efficiency of the engineered implementation is plagued by inconsistency, even with advanced tools to predict strong target sites in DNA. Bush et al.¹ aim at one possible factor responsible for CRISPR-Cas9 inefficiency at certain DNA sequences: unproductive intramolecular base pairing interactions between the tracrRNA and crRNA domains within the sgRNA (Figure 1). These authors hypothesize that such interactions might be minimized by properly adjusting the sequence of the tracrRNA domain, preserving its key interactions with the Cas9 protein while relieving unwanted crRNA interactions that thwart efficient DNA hybridization. Perhaps a different optimal tracrRNA domain exists for each crRNA domain.

Figure 1. sgRNAs are prone to unfavorable intramolecular interactions that inhibit proper loading into Cas9.

(A) Proper folding of crRNA domains (blue), tracrRNA domains (green), and engineered tetraloop (red) of the sgRNA allows Cas9-sgRNA ribonucleoprotein (RNP) complex formation with a PAM (purple) and DNA target (yellow).

(B) Unproductive base pairing within a sgRNA inhibits RNP-DNA complex formation.

Rather than rational engineering of such solutions, Bush et al.¹ engage the power of non-rational, unbiased genetic selection in their adaptation of the powerful SELEX (systematic evolution of ligands by exponential enrichment) technique to create a protocol they term BLADE (binding and ligand activated directed evolution). In the key tactic of this approach, a vast degenerate library of sgRNAs is complexed with Cas9 and challenged with in vitro cleavage of a dsDNA target known to be readily cleavable. Successful cleavage by the Cas9 nuclease frees a new DNA 5′ terminus while the entire Cas9-sgRNA-dsDNA complex remains intact. The authors cleverly exploit this convenient post-cleavage stability to reward sgRNAs that enabled target cleavage by biotinylating the revealed 5′ DNA termini and partitioning the successful complexes using streptavidin affinity (Figure 2). Isolated sgRNAs then become templates to regenerate pools of successful sgRNAs using PCR and in vitro transcription. Subsequent selection rounds ensue to allow further optimization prior to deep sequencing.

Figure 2. BLADE SELEX.

(A) A vast sgRNA library contains fixed regions (black) for primer binding and a central 60-nucleotide randomized region biased toward the S. pyogenes wild-type tracrRNA sequence (various colors).

(B) The sgRNA library competes in tests of Cas9 binding, target DNA binding, and DNA cleavage.

(C) Successfully cleaved DNAs are tagged for selective capture of bound Cas9-sgRNA RNPs.

(D) Recovered sgRNAs are amplified to regenerate an enriched sgRNA library. Rounds of selection enhance enrichment of successful sgRNAs.

(E) Libraries are sequenced to determine which sgRNA sequences are successful and warrant validation.

In principle, BLADE rewards performance in four consecutive molecular events: proper sgRNA folding, sgRNA loading into Cas9, ribonucleoprotein complex loading onto target dsDNA, and Cas9-mediated targeted DNA cleavage. In fact, the authors first select for individual steps. They initially perform two rounds of SELEX based only on successful sgRNA loading into Cas9, capturing fruitful protein-RNA interactions by filtration. The third selection round, which incorporates the sgRNA binding step already required in rounds 1–2, adds the further constraint of Cas9-sgRNA-dsDNA complex formation. Finally, rounds 4–5 integrate the full nuclease-activity-based approach described above.

Bush et al.¹ show that BLADE enables selection of new sgRNAs that reduce intramolecular entanglements, generating target cleavage comparable to the best-known efficiencies at the test site. What about improvements beyond that? Rather surprisingly, the authors find that some efficient alternative tracrRNA “cassette” sequences identified by BLADE can be combined with unselected crRNA sequences to generate improved cleavage efficiency at other target sites. A small collection of such alternative “cassettes” may therefore prove useful in improving CRISPR-Cas9 cleavage efficiency for difficult DNA targets of interest.

Given that the authors’ initial selection steps rewarded behavior that is prerequisite for the cleavage step harnessed in rounds 4–5, it is possible that these preliminary selection rounds were redundant and not ultimately needed for successful selection. It remains to be seen whether BLADE SELEX can be simplified by truncating the stepwise approach described here.

It is intriguing to consider how natural CRISPR-Cas systems contend with potential unproductive RNA-RNA interactions between their separate tracrRNAs and crRNAs. Sequential installation of these RNAs into the Cas endonuclease during the natural assembly process might prevent such unproductive interactions, but there is evidence that the RNA components are not separately installed.¹⁰ Prior study does not rule out, however, that components of natural CRISPR-Cas systems act as RNA chaperones to prevent unhealthy entanglements.

There is fascinating evidence that natural CRISPR systems only “remember” DNA targets that can be successfully targeted by CAS cleavage because successful target cleavage by Cas9 is required for integration of new spacer sequences (primed adaptation). Interestingly, complete matching of a spacer sequence with an appropriately positioned PAM results in lower efficiency of primed adaptation, as if implying to the host that no more spacers are needed to target this pathogen.⁸ The implication is that any crRNA-tracrRNA combination prone to misfolding would not be utilized by microbial immune systems of this kind. If a tracrRNA cannot properly assemble into Cas9 with a spacer (crRNA), the crRNA will simply not be processed. In such cases, microorganisms do not resort to BLADE; they just move on and are content to target more amenable pathogen sequences. Genetic engineers focused on specific target sequences do not have this luxury. Thus, BLADE and other creative approaches for optimizing engineered sgRNAs may become valuable gene editing tools.