Nucleosome breathing and remodeling constrain CRISPR-Cas9 function

1The recent development of CRISPR (clustered regularly interspaced short palindromic repeats) systems, particularly the type II CRISPR-Cas9 mechanism from Streptomyces pyogenes, as an artificial tool for genome engineering, gene regulation, and live imaging is a remarkable achievement with profound impact in a wide variety of research fields and applications. Despite its successful adoption across numerous eukaryotic organisms, relatively few details are known of the mechanism by which bacterial CRISPR-Cas9 systems operate in eukaryotic cells.

2CRISPR-Cas9 originated in bacteria, where genomic DNA generally consists of supercoiled circular molecules associated with nucleoid-associated proteins. In contrast, eukaryotic chromosomes are linear, packaged with histone octamers into nucleosomes, and further organized into higher-order structures. The packaging of DNA into nucleosomes generally inhibits the binding of sequence specific DNA binding factors. In the simplest model, nucleosomes would analogously inhibit Cas9 action. Further, in eukaryotes ATP-dependent chromatin remodelers reposition, remove, or restructure nucleosomes to regulate the access of DNA binding factors. It can therefore be imagined that the action of remodelers also regulates the action of Cas9 on nucleosomes.

3To quantitatively test the above models we performed biochemical studies to measure Cas9 activity on nucleosomes assembled with native and artificial nucleosome positioning sequences. We find that the combination of nucleosome breathing, by which DNA transiently disengages from the histone octamer, and the action of chromatin remodeling enzymes allow Cas9 to act on nucleosomal DNA with rates comparable to naked DNA. The results provide a biochemical explanation for the efficacy of Cas9 in eukaryotic cells.

4To determine if a nucleosome inhibits the ability of Cas9 to scan, recognize, and cleave sgRNA-directed DNA targets, we performed in vitro Cas9 cleavage assays using mononucleosomes (single nucleosomes on short dsDNA molecules) reconstituted using the Widom 601 positioning sequence with 80 base pairs of flanking DNA on both sides (referred to as 601 80/80 particles, Figure 1A). The 601 sequence is an artificially derived sequence with high affinity for the histone octamer and has proved a valuable tool for assembling well positioning nucleosomes for biochemical studies. Using sgRNAs targeting the nucleosomal dyad, entry/exit sites, and flanking DNA, we measured the rates of Cas9 cleavage with naked 601 DNA and the 601 80/80 particles. Targeting the DNA flanking the nucleosome showed cleavage rates comparable to those of naked DNA. Cleavage rates at entry/exit sites of the nucleosome were much lower compared to naked DNA (~23–28x decrease cleavage rate vs. DNA alone) (Figure 1B,C). Targeting near the nucleosomal dyad resulted in further inhibition of cutting by Cas9 (~1000x decrease vs. DNA alone) (Figure 1C,D). Previous work has shown that nucleosomal DNA transiently disengages from the histone octamer, a process termed as nucleosomal DNA unpeeling or breathing. The equilibrium for DNA unpeeling gets progressively more unfavorable the closer the DNA site gets to the dyad. The nucleosome-mediated inhibition of Cas9 activity is more pronounced near the dyad suggesting that Cas9 cleavage occurs on DNA that is transiently disengaged from the histone octamer.

5Nucleosomes block the ability of Cas9 to cleave DNA, but it is unclear at which step of Cas9 activity this occurs. Cas9 recognizes DNA target sites in a two-step process that begins with binding to the DNA protospacer adjacent motif (PAM, in this case 'NGG') through its C-terminal PAM-interacting region, followed by sequential melting of the DNA double strand and annealing of the sgRNA guide segment to the unwound target DNA strand (Figure 1—figure supplement 1A). Complete annealing of the 20-nt guide RNA to the target DNA is required to drive a progressive conformational transformation that authorizes Cas9 to simultaneously cleave both DNA strands. Given this order of events, it is conceivable that nucleosomes can interfere with any of the steps preceding and including DNA cleavage.

6To identify the point at which nucleosomes disrupt Cas9 function, we assessed binding of nuclease-dead Cas9 (dCas9) to mononucleosomal particles by an electrophoretic mobility shift assay. We performed dCas9 binding assays using 601 0/0 nucleosomal particles which are devoid of naked DNA overhangs. Binding of dCas9 pre-loaded with core targeting sgRNA with 601 0/0 nucleosomes is undetectable whereas binding to naked DNA control is still robust (Figure 1—figure supplement 1B). The presence of super shifts in the gel migration pattern suggests that multiple dCas9 molcules are engaging the same DNA substrate molecule. We investigated this further and determined that, in our binding assay, the highly transient dCas9 binding to PAMs within target DNA is observable as super shifts, likely due to a combination of a high number of PAMs on the target DNA (23 NGG PAMs present in 601 0/0 sequence) and common caging effects of gel binding assays. The absence of a super shift binding pattern with 0/0 nucleosomes (Figure 1—figure supplement 1B, right) suggests that dCas9 cannot stably interact with PAMs located on nucleosomes, in a manner consistent with a recently published study.

7The artificial Widom 601 is an atypically strong nucleosome positioning sequence that shows ~100-fold less breathing dynamics compared to physiological nucleosome positioning sequences, such as the 5S rRNA gene. To determine how Cas9 contends with nucleosomes assembled on this physiological positioning sequence, we performed cleavage experiments with nucleosomes assembled from 5S rRNA gene sequences from Xenopus borealis (Figure 2A). Cas9-mediated cleavage at sites near the entry/exit of the nucleosome is substantially enhanced (700–fold) with 5S nucleosomes compared to 601 particles (Figure 2B–D). In the context of 601, cutting at this site is 1000-fold slower than in naked DNA. In contrast, with 5S nucleosomes, cutting at the comparable site is only 1.5-fold slower than in naked DNA. However, Cas9 cleavage near the dyad is inhibited to a similar extent on both 5S and 601 nucleosomes, showing that the 5S-specific enhancement of Cas9 activity does not extend all the way to the nucleosomal dyad. These results support our interpretation that nucleosomal DNA breathing substantially enhances Cas9 binding to nucleosomes and demonstrate that nucleosomal DNA sequence, through its influence on nucleosome stability, can regulate Cas9 activity over a large dynamic range.

8We next investigated whether chromatin remodeling could enhance Cas9 activity towards chromatin substrates. Nucleosome positioning in vivo is strongly dependent on ATP-dependent chromatin remodelers, which are capable of loading, repositioning, and removing nucleosomes from the chromatin fiber. To measure how chromatin remodelers can influence Cas9 activity, we performed experiments where we pre-incubated 601 nucleosomes with remodeler enzymes prior to Cas9-mediated cleavage. For our experiments with the human ISWI-family remodeler SNF2h, we used asymmetric nucleosomes that possess flanking DNA only on the entry side (601 80/0 particles). When incubated with 601 80/0 particles, SNF2h promotes sliding of the nucleosome towards the center of the DNA molecule (Figure 3A–B, Figure 3—figure supplement 1). We then performed in vitro cleavage experiments where 80/0 particles, pre-remodeled with SNF2h, were incubated with Cas9:sgRNA complex with its target site located within the nucleosome exit region. Remodeling 80/0 nucleosomes by SNF2h resulted in a strong enhancement of Cas9 cleavage to ~34%, showing that SNF2h slides the nucleosome enough to improve Cas9’s accessibility to the target site and that Cas9 is now able to bind and cleave at a higher rate (Figure 3A–D).

9We also performed this experiment by simultaneously adding SNF2h and Cas9 and found a similar rate enhancement (Figure 3—figure supplement 2).

10While the ISWI remodeler SNF2h predominantly slides nucleosomes, remodelers from the SWI/SNF class have multiple outcomes, which include generation of DNA loops and eviction of the histone octamer in addition to nucleosome sliding. To determine if the types of remodeled products generated influence Cas9 activity, we performed similar experiments using 601 80/80 particles and the yeast chromatin remodeler RSC. We find that RSC activity also dramatically enhances cleavage on 601 80/80 nucleosomes when Cas9 is targeted to the entry site, negating most of the inhibitory influence of the nucleosome on Cas9 (Figure 3E–F). These results demonstrate that two different classes of chromatin remodeling enzymes can significantly enhance Cas9 access to DNA targets normally obscured by nucleosomes.

11Here we demonstrate, using detailed biochemical studies with a variety of nucleosomal templates, that (i) the intrinsic stability of the histone-DNA interactions, (ii) the location of the target site within the nucleosomes (nucleosome positioning), and (iii) the action of chromatin remodeling enzymes play critical roles in regulating the activity of S. pyogenes Cas9. Below we discuss the implications of our results.

12Nucleosomes have been shown to inhibit the action of DNA binding factors. Recent work using nucleosomes assembled on the 601 sequence has led to the qualitatively similar conclusion that nucleosomes are refractory for Cas9 action. The comparison here between Cas9 action on 601 nucleosomes vs. nucleosomes assembled on the native 5S sequence suggests a more refined model for how nucleosomes regulate Cas9 action. We find that Cas9 sites near the entry/exit sites of 5S nucleosomes are cleaved ~700-fold better than the corresponding sites within 601 nucleosomes. Given that DNA breathing occurs at least 100-fold more in 5S nucleosomes than 601 nucleosomes we propose that Cas9 gains access to nucleosomal DNA when the DNA is transiently unpeeled from the histone octamer. This model also explains why sites closer to the entry/exit sites are cut more readily by Cas9 than sites near the dyad. This is because DNA unpeeling up to the dyad is substantially less favored (100-fold) for both the 601 and 5S nucleosomes than DNA unpeeling near their respective entry/exit sites.

13In vivo, as in vitro, the precise position of nucleosomes can greatly affect DNA factor binding. Chromatin remodeling enzymes can move nucleosomes away or towards the factor binding sites to respectively enhance or inhibit factor binding. We find that Cas9 activity can also benefit from chromatin remodeling to access nucleosomal DNA, as evidenced by the strong enhancements of Cas9 cleavage resulting from the action of the chromatin remodelers SNF2h and RSC. These two remodelers produce distinct nucleosomal arrangements yet still substantially alleviate nucleosome-mediated occlusion of Cas9 activity.

14In combination, our data lead to a comprehensive model that reconciles both biochemical evidence and in vivo observations to explain how Cas9 is able to access nucleosomal DNA in live cells (Figure 3I). In vivo, the majority of nucleosomes are not located on strong positioning sequences, and therefore may be permissive to Cas9 binding, especially at target sites that are readily accessible by DNA unpeeling. Chromatin remodeling activities can further provide diverse mechanisms to potentiate Cas9 activity at sites located close to the nucleosomal dyad or within more strongly positioned nucleosomes, which would otherwise be refractory to Cas9 action. We hypothesize that the combination of spontaneous DNA unpeeling and remodeling contributes to the widespread success of CRISPR-Cas9 in eukaryotic cells.

15Interestingly, most applications of CRISPR-Cas9 in vivo have focused on genome engineering of protein-coding genes and other functional genomic elements associated with gene expression, which are typically associated with high rates of nucleosome remodeling. It is also conceivable that Cas9 can temporarily gain access to less accessible regions of the genome during specific points of cell cycle (e.g. DNA replication), leading to sufficient DNA cleavage events to promote NHEJ-mediated mutagenesis or HDR-mediated DNA integration at appreciable rates. Recent studies on Cas9’s behavior by single molecule imaging have also demonstrated that Cas9 favors more accessible euchromatin regions but is not completely excluded from transcriptionally silent heterochromatin. For other CRISPR applications that require stable binding of nuclease-deficient dCas9 to DNA, such as transcriptional regulation and live-cell imaging with fluorescent dCas9, even modest nucleosome phasing could have a dramatic impact. For example, the +1 nucleosome in RNA pol II-transcribed genes is strongly positioned with phasing that dissipates gradually with each following nucleosome. Several high resolution studies conducted in parallel to our work have established that the +1 nucleosome and resulting nucleosome phasing can exert a strong influence on dCas9’s DNA-binding ability for transcriptional regulation, but the effect is less striking on genome editing with Cas9.

16Our observations suggest that sgRNA design strategies that avoid targeting near the dyad of strongly phased nucleosomes are likely to be more successful than current methods. Large scale nucleosome positioning or DNA accessibility maps are now readily available and can inform CRISPR sgRNA design in order to avoid targeting regions of low accessibility. Alternatively, whole cell chromatin de-condensation or de-repression using chromatin factor drugs such as HDAC or DNA methyltransferase inhibitors may be an alternative and attractive strategy for improving CRISPR-Cas9 activity towards densely compact regions of chromatin.

17Wild-type Streptococcus pyogenes Cas9 and catalytically-inactive Cas9 (dCas9) containing D10A and H840A mutations were individually cloned into a custom pET-based expression vector encoding an N-terminal 6xHis-tag followed by a small ubiquitin-related modifier (SUMO) fusion tag and a Ulp1 protease cleavage site. Recombinant Cas9 variants were then expressed in Escherichia coli strain BL21 (DE3) (Novagen) and further purified to homogeneity as previously described.

18Single guide RNAs (sgRNAs) were prepared by in vitro run-off transcription using recombinant His-tagged T7 RNA polymerase and PCR product templates. Briefly, the DNA templates containing a T7 promoter, a 20-nt target DNA sequence (listed in Table 1) and an optimal 78-nt sgRNA scaffold were PCR amplified using Phusion Polymerase (NEB) according to manufacturer’s protocol. The following PCR products were used directly as DNA templates for in vitro RNA synthesis in 1x transcription buffer (30 mM Tris-HCl pH 8.1, 20 mM MgCl2, 2 mM spermidine, 10 mM DTT, 0.1% Triton X-100, 5 mM each NTP, and 100 μg mL-1 T7 RNA polymerase). After incubation at 37°C for 4–8 hr, the reactions were further treated with RNase-free DNase I (Promega) at 37°C for 30 min to remove the DNA templates. The synthesized sgRNAs were purified by Ambion MEGAclear kit and eluted into DEPC-treated H2O for downstream experiments.

19Gradient salt dialysis was used to assemble mono-nucleosomes on DNA templates containing the 147 bp long 601 or the 5S positioning sequence from X. borealis (listed in Table 2), and labeled with fluorescein on the 5’ upstream end. Histones and histone octamers were prepared as previously described.

20Cleavage assays were conducted as previously described with the following modifications. Cas9:sgRNA complexes were reconstituted by incubating Cas9 and sgRNA for 10 min at 37°C. Reactions contained 5 nM fluorescein labeled DNA or nucleosomes and 100 nM Cas9:sgRNA. In combined cleavage and remodeling experiments, 25 nM SNF2h or 3 nM RSC was first incubated with 5 nM naked DNA or nucleosomes for 45 min at 37°C. Cleavage assays were carried out in reaction buffer (20 mM Tris-HCl pH 7.5, 70 mM KCl, 5 mM MgCl2, 5% Glycerol, and 1 mM DTT) at 25°C. For SNF2h and RSC remodeling experiments, 0.2 mM ATP was also added. For RSC remodeling experiments, 1 mM MgCl2 was used. Time points were quenched using stop buffer (20 mM Tris-HCl pH 7.5, 70 mM EDTA, 2% SDS, 20% glycerol, and 0.2 mg/mL xylene cyanol and bromophenol blue). Proteins were digested with 3 mg/mL of Proteinase K and incubated at 50°C for 20 min. Samples were resolved on 1x TBE, 10% Polyacrylamide gels for 4 hr at 140 V before visualizing using a Typhoon scanner (GE Healthcare) and quantifying with Image J. For band quantification, background intensity was first subtracted after averaging the intensity of three areas. For cleavage gels, fraction uncleaved was determined by measuring the intensity of the uncleaved band compared to the total intensity for the lane. Similarly, fraction unbound was determined by measuring the intensity of the unbound band compared to the total intensity for the lane.

21All experiments were performed in triplicate. Experiment variability is presented as the standard error of the mean, calculated by the standard deviation divided by the square root of N.

22Propagation of error for Rates of Cleavage on Nucleosomes to Rates of Cleavage on DNA was calculated as follows:

23Data were fit on Graphpad Prism using a standard one phase decay model:

24where Y is the fraction of uncleaved DNA, Y0 is the value of Y at time = 0, k is the observed rate constant (min-1) and t is time (min).

25dCas9 and a 2x molar ratio of sgRNA were incubated for 10 min at 37°C. Various concentrations of dCas9:sgRNA complex were incubated with 20 nM naked DNA or nucleosomes in binding buffer (20 mM Tris-HCl pH 7.5, 100 mM KCl, 5 mM MgCl2, 5% Glycerol, 1 mM DTT, and 0.02% NP-40). Samples were incubated at room temperature for 1 hr before being run on native 0.5X TBE 6% polyacrylamide gels, visualized on a Typhoon scanner, and quantified using ImageJ. Fraction unbound was measured as the intensity of all unbound species divided by the total intensity. Fraction unbound was then converted to fraction bound:

26and binding curves were fit with: