Peptide Nucleic Acid-Mediated Regulation of CRISPR-Cas9 Specificity

Abstract

Although CRISPR-Cas9 gene therapies have proven to be a powerful tool across many applications, improvements are necessary to increase the specificity of this technology. Cas9 cutting in off-target sites remains an issue that limits CRISPR’s application in human-based therapies. Treatment of autosomal dominant diseases also remains a challenge when mutant alleles differ from the wild-type sequence by only one base pair. Here, we utilize synthetic peptide nucleic acids (PNAs) that bind selected spacer sequences in the guide RNA (gRNA) to increase Cas9 specificity up to 10-fold. We interrogate variations in PNA length, binding position, and degree of homology with the gRNA. Our findings reveal that PNAs bound in the region distal to the protospacer adjacent motif (PAM) site effectively enhance specificity in both on-target/off-target and allele-specific scenarios. In addition, we demonstrate that introducing deliberate mismatches between PNAs bound in the PAM-proximal region of the gRNA can modulate Cas9 activity in an allele-specific manner. These advancements hold promise for addressing current limitations and expanding the therapeutic potential of CRISPR technology.

Introduction

CRISPR-Cas9 is a powerful gene-editing tool that has been adapted across a wide range of applications.^1,2 There is a large potential for CRISPR systems in therapeutics targeting various genetic disorders, such as a recent study showing in vivo Cas9 editing that resulted in partial restoration of hearing in mice with monogenic and digenic dominant hearing loss.³ However, CRISPR-based therapies have been shown to induce cleavage in off-target sites,^4,5 presenting a challenge to their use in human-based therapies. As a result, methods to improve Cas9 specificity would be valuable to increase the safety and efficacy of these therapies.

One particular application where specificity is required is in the treatment of autosomal dominant diseases, where the therapeutic goal would be to knock out the mutant allele while preserving the wild-type allele, which may differ by just a single base pair. In cases where the autosomal dominant missense mutations form a novel protospacer adjacent motif (PAM), CRISPR-based therapies are generally able to successfully discriminate between alleles and specifically disrupt the mutant allele via the nonhomologous end joining repair pathway.^6,7 However, when the single nucleotide mutation falls in the guide sequence, current CRISPR-Cas9 systems struggle to differentiate between alleles.⁸ An example of this are mutations that cause dominant negative TGFBI corneal dystrophies. While there have been many mutations that have been identified to cause corneal dystrophy, the five most common are as follows: R124H, R124C, R124L, R555W, and R555Q.^8,9 All five of these mutations are single-base pair substitutions and result in different types of corneal dystrophy. For example, R124H causes granular corneal dystrophy type 2, while R124L results in Reis–Bücklers corneal dystrophy.^8,9 Overall, the disease outcome is debilitating, and while the ultimate treatment is a corneal transplant, recurrence of the disease often necessitates subsequent corneal transplants.⁹ With a severe limit on both treatment options and preventative measures, corneal dystrophy represents a disease where patients would benefit from a therapeutic gene-editing option.

Various methods have been employed in an attempt to increase Cas9 specificity, including the use of small molecules and proteins to bind and inhibit Cas proteins.^10–12 One example of this involves using a cell-permeable small molecule to activate Cas9 nucleases by inserting a 4-hydroxytamoxifen-responsive intein at various positions in the Cas9, which resulted in an average sixfold higher specificity compared with wild-type Cas9.¹⁰ It has also been shown that an anti-CRISPR DNA mimic, AcrIIA4, reduces off-target editing of Cas9 when delivery of the molecule is timed correctly.¹¹ Although the use of these molecules has had success, they are associated with a number of drawbacks. Many of these techniques are designed to be compatible with only one Cas protein and are therefore unable to be used universally.^13,14 They also have associated toxicity and the potential for immunogenicity due to protein expression.^13,14 Although nucleic acid inhibitors have also been explored, they lack sufficient sequence specificity due to their binding of components of the PAM site rather than the spacer sequence.^15,16

In this work, we address these issues through the use of peptide nucleic acids (PNAs). These are charge neutral DNA analogs with a polyamide backbone¹⁷ (Fig. 1A). They bind with high affinity to DNA and RNA¹⁸ and have a wide variety of applications, including gene editing,^19–22 regulating gene expression,^23,24 and regulating transcriptional and translational processes.^25–27 PNAs, by themselves, have been shown to achieve gene editing when paired with donor DNAs and delivered to cells via polymeric nanoparticles.^19–21 In a mouse model of β-thalassemia, triplex-forming tail-clamp PNAs designed to bind to genomic DNA target sites, delivered along with donor DNAs, mediated b-globin gene editing in bone marrow cells.¹⁹ Similarly, when used in a cystic fibrosis model, PNA and donor DNA combination treatment was able to mediate CFTR gene editing in mice.²⁰ However, because of the broad implementation of CRISPR technology for gene editing in clinical and nonclinical applications, we have evaluated the potential for PNAs to be used as a tool to regulate CRISPR/Cas9 activity by exploiting the robust RNA binding (rather than DNA binding) properties of PNAs. As a first step, previous work from our group established that PNAs designed to bind to the spacer region of guide RNAs (gRNA), or “antispacer PNAs,” are able to modulate Cas9 activity when bound to the gRNA in Cas9 Ribonucleoprotein (RNP)²⁸ (Fig. 1B).

FIG. 1.

Impact of varying length and guide RNA binding positions of antispacer peptide nucleic acids (PNAs) to regulate Cas9 activity in a BFP gene target in human cells. (A) Comparative chemical structures of DNA and PNA oligomers. (B) Schematic of PNAs effect on Cas9 activity. (C) Schematic depicting PNAs length and binding position to the sgRNA. (D) Indel formation in the BFP gene in a K562-BFP cell line following treatment with Cas9 RNPs complexed with protospacer adjacent motif distal PNAs of various lengths. Two-tailed t-tests were performed comparing PNA conditions with the no-PNA control. (E) Indel formation in a K562-BFP cell line following treatment with Cas9 RNPs complexed with 6mer PNAs bound at different regions across the guide sequence. Two-tailed t-tests were performed comparing the PNA conditions with the no-PNA control.

Here, we further characterize and refine design principles to guide the application of antispacer PNAs to modulate Cas9 specificity. The results show that varying length, binding position, and degree of homology between the PNA and gRNA can all be used to manipulate Cas9 activity to varying degrees. We demonstrate the versatility of PNAs in multiple gene-editing backgrounds, such as reducing Cas9 cutting at off-target sites as well as increasing specificity in an allele-specific context.

Materials and Methods

Cell lines

K562-BFP

K562 reporter cell lines were acquired from the laboratory of Jacob Corn. The cells endogenously express Blue Fluorescent Protein (BFP) due to a lentiviral-inserted BFP gene (Addgene: #111092).²⁹

U2OS-RFP/GFP TGFBI R124

Target sequences from the TGFBI gene were designed as complementary oligos with restriction site-compatible overhangs, annealed in vitro, and ligated into the pRGS reporter vector digested with EcoR1 and BamH1³⁰ (RV01, PNA Bio). Target sequences were designed such that the downstream Green Fluorescent Protein (GFP) start codons remain out-of-frame, with each GFP allele having a different frameshift. U2OS reporter cell lines were then created by cloning the reporter sequences into lentiviral vectors under the EF1α promoter and subsequent transduction with Blasticidin selection (K496000, Invitrogen). Stably transduced Blasticidin resistant U2OS cells were then sorted using a BD FACS AriaII cell sorter to isolate a subpopulation of Red Fluorescent Protein (RFP)-positive and GFP-negative cells. Vectors and resulting cell lines were sequenced by Sanger sequencing to confirm the presence of each reporter element and the target sequence. U2OS cell lines were periodically sorted to ensure similar and robust RFP+GFP− expression.

Cell culture

K562-BFP cells (CCL-243, ATCC) were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS, Life Technologies). U2OS cells (HTB-96, ATCC) were maintained in McCoy’s 5A medium supplemented with 10% FBS (Life Technologies).

Cas9 RNP with PNA formulations and nucleofections

50 pmol of single guide RNA (sgRNA) (Alt-R^TM CRISPR-Cas9 sgRNA, IDT) and 50 pmol of PNA were annealed at 37°C for 15 min in a 4.5 μL reaction in 1 × 3.1 Buffer (B7203S, NEB). sgRNA/PNA was then complexed with 45 pmol of SpCas9 (CP02, PNA Bio) at room temperature for 10 min in a 6 μL reaction. Immediately before nucleofection, 100 pmol of Alt-R^TM Cas9 Elec Enhancer (10007805, IDT) was spiked into the Cas9 RNP/PNA solution. A full list of gRNA sequences is listed in Supplementary Table S1.

To confirm the annealing of PNAs to sgRNAs, a gel shift assay was performed on a 5% polyacrylamide gel. Carboxy tetramethylrhodamine (TAMRA)-labeled PNAs were annealed to the appropriate sgRNA and then complexed with Cas9 before being run in 1× TBE Buffer at 10 mA and imaged using a Cy3 filter, followed by staining in SYBR gold (S11494, Invitrogen) and imaging using a standard ultraviolet (UV) filter.

For nucleofections, 1 × 10⁶ K562-BFP cells or 5 × 10⁵ U2OS cells were resuspended in Lonza cell line solution (V4XC-2024/V4XC-1024, Lonza). 6.5 μL Cas9 RNP/PNA solution was added to the cells and then nucleofected using a Lonza 4D-Nucleofector X unit. Treated cells were then seeded in 2 mL of complete media and incubated for 72 h, 7 days, or 10 days, prior to analysis.

For Cas9/RNA pretreatment, 50 pmol of sgRNA (Alt-R^TM CRISPR-Cas9 sgRNA, IDT) and 45 pmol of SpCas9 (CP02, PNA Bio) were complexed at room temperature for 10 min in a 5 μL reaction. Immediately before nucleofection, 100 pmol of Alt-R^TM Cas9 Elec Enhancer (10007805, IDT) was spiked into the Cas9 RNP solution. 1 × 10⁶ K562-BFP cells were resuspended in Lonza cell line solution V4XC-2024. 5.5 μL Cas9 RNP solution was added to the cells and then nucleofected using a Lonza 4D-Nuclofector X unit. Cells were then allowed to recover in 1.25 mL of complete media for 2 h. Then 50 pmol of PNA was similarly nucleofected into cells using Lonza cell line kits and the 4D-Nucloefector. Cells were then seeded in 2 mL of complete media and incubated for 72 h.

PNA synthesis and purification

Manual PNA synthesis

PNA oligomers were synthesized manually on 10% l-lysine-loaded 4-methylbenzhydrylamine resin (RMB-10450PI, Peptides International), using a standard Boc chemistry process. All Boc-aeg-PNA monomers were purchased from ASM Research Chemicals GmbH. A colorimetric Kaiser test was performed throughout synthesis to ensure proper deprotection and coupling of monomers. Completed PNAs were cleaved from the resin at room temperature. This process was completed using the following cocktail mix: meta-cresol (m-cresol): thioanisole: trifluoromethanesulfonic acid: trifluoroacetic acid (TFA) (1: 1: 2: 6) for 30 min (×2). The resulting crude PNA was then precipitated with cold ether and purified using reverse phase–high-performance liquid chromatography (RP-HPLC) [5–95% acetonitrile (ACN)/water/0.1% TFA gradient]. The HPLC program setup consisted of the following: Waters 2998 Photodiode Array Detector, Waters 2545 Quaternary Gradient Module, and Waters 2707 Autosampler.

Automated PNA synthesis

In some cases, a Biotage Initiator + Alstra microwave peptide synthesizer was used to synthesize PNA oligomers. Sequences were assembled automatically on 10% l-lysine-loaded Rink Amide ChemMatrix Resin (727768-5G, Sigma Aldrich) using a standard Fmoc chemistry process. Fmoc-aeg-PNA monomers were purchased from PNA Bio Inc. After synthesis, PNAs were cleaved from the resin at room temperature using the following cocktail mix: water: triisopropylsilane: TFA (1: 1: 38) for 60 min (×1). The resulting crude PNA was then precipitated with cold ether and purified using RP-HPLC [5–95% ACN/water/0.1% TFA gradient]. The HPLC program setup consisted of the following: Waters 2998 Photodiode Array Detector, Waters 2545 Quaternary Gradient Module, and Waters 2707 Autosampler.

Some PNAs were also purchased directly from PNA Bio Inc. PNA sequences and sources are summarized in Supplementary Table S2. All PNAs were synthesized with one l-lysine (K) residue on both the C- and N-termini in order to increase solubility and binding affinity. All PNA stock solutions were prepared using nanopure water, and the concentrations were determined using a ThermoScientific^TM NanoDrop^TM One^C microvolume spectrophotometer using the following extinction coefficients: 13,700M⁻¹cm⁻¹ (A), 6,600M⁻¹cm⁻¹ (C), 11,700M⁻¹cm⁻¹ (G), and 8,600M⁻¹cm⁻¹ (T).

Targeted amplicon sequencing

Genomic DNA was purified from 1 × 10⁶ cells using the Promega ReliaPrep gDNA Tissue Miniprep System (A2052) and eluted in 50 μL of water. 100 ng of gDNA was then used for library prep with an Ampliseq for Illumina Library Plus kit (20019101, Illumina). Library prep used a custom primer pool, which was designed to amplify a predetermined panel of genomic targets. Libraries were indexed using AmpliSeq CD Indexes (20019105, Illumina) and pooled and then loaded into a mid-output (300 cycles, Illumina, FC-420-1004) cartridge. Paired-end sequencing was then carried out on an Illumina Miniseq instrument. Quality analysis was carried out on generated FASTQ files (Basespace FASTQC), after which they were analyzed for indel frequency using the Cas9-Analyzer assessment tool (parameters: comparison range (R)—both ends of the full amplicon sequence, minimum frequency (n) = 1).

For R124H U2OS reporter cell lines, primers were designed to bind in the RFP motif and GFP motif regions flanking the target site. This region was then amplified from 250 ng of purified gDNA by Phusion High-Fidelity Polymerase (M0530S, NEB) and purified using a Monarch nucleic acid purification kit (T1030S, NEB). Library prep used 150 ng of purified amplicon input with an Illumina DNA Prep kit (20025519, Illumina). These libraries were then indexed using Illumina DNA/RNA UD Indexes (20091654, Illumina) and then sequenced and analyzed as above. Primers are listed in Supplementary Table S3.

Flow cytometry

K562-BFP cells

Cells were spun down and resuspended in 400 μL Phosphate-buffered saline (PBS) 15 min prior to analysis, filtering through a cell strainer-capped tube. Samples were then analyzed for BFP and GFP fluorescence using a Cytoflex LX instrument (Beckman Coulter) with PacificBlue and FITC lasers, respectively.

U2OS-RFP/GFP TGFBI R124

Cells were detached from plates using Trypsin-Ethylenediaminetetraacetic acid (EDTA) (0.05%) with phenol red (25300120, ThermoFisher) at 37°C for 5 min. Cells were then spun down and resuspended in 400 μL PBS 15 min prior to analysis, filtering through a cell strainer-capped tube. Samples were then analyzed for RFP and GFP fluorescence using a Cytoflex LX instrument (Beckman Coulter) with PE and FITC lasers, respectively.

All gating was done using FloJo v10.8.0 software. Background fluorescence was determined by measuring a mock-nucleofected cell line in triplicate.

Cell viability assay

5 × 10⁵ U2OS cells were nucleofected with specified doses of Cas9 RNP and PNA using a Lonza 4D-Nucleofector. Cells were then serially diluted and seeded at 1,500 cells/well into opaque-walled 96-well plates and incubated at 37°C for 72 h. Cell viability was then measured in-plate using the Cell Titer-Glo Luminescent Cell Viability Assay (G7570, Promega). Luminescence was measured with a Synergy H1 Multi-Mode Microplate Reader (Biotek) and then graphed as percent viability relative to a mock-nucleofected control.

UV spectroscopy thermal melting analysis

20mer gRNA was mixed with PNA or DNA (2 μM of each strand) in 10 mM sodium phosphate buffer at pH 7.4. Guide sequences and DNA oligos are listed in Supplementary Table S3. UV melting experiments were performed using an Agilent Cary 100 UV–Vis spectrophotometer. UV melting spectra were recorded after every 1°C temperature change by recording the absorbance at 260 nm from 95°C to 25°C to 95°C, with a heating/cooling ramp rate of 1°C/min. The heating and cooling curves were overlapped to confirm reversible denaturation. Spectra were plotted using Prism 10 (v10.1.2) software. Origin 2024 software was used to generate the first derivative plots of the melting curves to determine the melting temperature for each duplex.

Statistics

Graphing and normalization were performed for each data set using Prism 10 (v10.1.2) software. Negative values resulting from normalization of data were reported as zero. Two-tailed t-tests were performed for each data set using Microsoft Excel (v2331).

Results

Characterizing the impact of length and binding position of antispacer PNAs on Cas9 activity

In order to optimize antispacer PNAs modulation of Cas9 activity, the length of the PNAs and binding position within the gRNA spacer sequence were systematically evaluated. To begin, a range of antispacer PNAs were synthesized to bind the PAM distal region of a gRNA targeting the BFP gene, ranging in length from 6 nucleotides to 16 (Fig. 1C). These PAM distal PNAs were annealed to the sgRNA and then complexed with Cas9. A gel shift assay was performed on a selection of these PNA/Cas9 RNP complexes to ensure complex formation and binding (Supplementary Fig. S1). These complexes were then nucleofected into a K562-BFP cell line.²⁹ Seventy-two hours later, flow cytometry was performed to measure the BFP drop-out and thus determine Cas9 cutting activity in the BFP target gene (Fig. 1D) (representative primary data showing the gating are presented in Supplementary Fig. S2). Shorter PAM distal PNAs, ranging from 6 bp to 12 bp, had little effect on Cas9 activity; however, the 7–20 14mer PAM distal PNA reduced Cas9 activity by 55% (P < 0.01) and the 5–20 16mer PAM distal PNA reduced Cas9 activity by 98% (P < 0.0001) (Fig. 1D). This indicates that a PAM distal PNA can bind over 60% of the gRNA region with minimal effect on Cas9 activity, but binding 70% will cause partial inhibition, while binding 80% or more of the gRNA region will completely inhibit Cas9 activity.

To determine whether this reduction in Cas9 activity was due to solely PNA length, or whether the extension of the binding position into the PAM-proximal gRNA region was also relevant, short 6mer PNAs were synthesized, which were bound in the middle and PAM-proximal regions of the same BFP gRNA (Fig. 1C). Analysis of BFP knockout by flow cytometry showed that the PAM-proximal 1–6 6mer PNA inhibited Cas9 activity by 96% (P < 0.0001) (Fig. 1E). The 5–10 6mer PNA reduced Cas9 activity by 27% (P < 0.01) and the 9–14 6mer PNA did not significantly reduce Cas9 activity (P = 0.08). We saw that the editing levels persisted up to 10 days (Supplementary Fig. S3), with the BFP 7–20 14mer PNA/Cas9 RNP condition actually yielding a slight increase in indel readout over the time course. This slight increase is consistent with our proposed mechanism, since the BFP 7–20 14mer PNA/Cas9 RNP is partially active for target site cleavage (see Fig. 1D), and this continued partial activity over time can lead to a gradual accumulation of additional edited alleles. Taken together, these results indicate that both the length and binding position of PNAs in the gRNA spacer sequence contribute to modulating Cas9 cutting activity.

PAM distal antispacer PNAs increase the ratio of Cas9 on-target to off-target activity in the FANCF gene

We next sought to test whether varying length and binding positions of antispacer PNAs would improve the on-target specificity of Cas9 activity, which has been defined here as the ratio of on-target to off-target editing rates. Previous work has shown that PAM distal 10mer PNAs increased the specificity ratio in a FANCF genomic target by over twofold, while a PAM-proximal 10mer PNA completely inhibited Cas9 activity in both on- and off-target contexts.²⁸ We propose that a PNA bound to the PAM-proximal region of the gRNA prevents R-loop formation and thus Cas9 cutting, while PNAs bound to the PAM distal region allow R-loop formation to occur, but with an energetic penalty, which makes lower affinity off-target sites unfavorable to Cas9 cutting (Fig. 2A).

FIG. 2.

Effect of antispacer peptide nucleic acids (PNAs) on relative on-target versus off-target activity of Cas9 targeting of the FANCF gene in human cells. (A) Schematic of PNAs effect on Cas9 R-loop formation. (B) FANCF gene on-(gray) and off-target (gray stripes) relative editing percentages by Cas9 RNPs complexed with the indicated antispacer PNAs along with the resulting specificity ratios (white). N/A indicates that the specificity ratio could not be calculated because off-target editing was undetectable. Two-tailed t-tests were performed comparing the PNA treatments with the no-PNA control group.

Here, we further explored this observation by annealing various antispacer PNAs to a FANCF gRNA and then complexing it with Cas9. The RNP was then nucleofected into U2OS cells. After 72 h, gDNA was extracted, and Illumina sequencing was performed to determine on- and off-target editing percentages (Fig. 2B). Consistent with previous results,²⁸ 1–6 and 1–10 PAM-proximal PNAs completely inhibited Cas9 cutting activity in both on- and off-target contexts (P < 0.0001).

We found that the 7–20 14mer PNA partially reduced on-target Cas9 activity by 45% (P < 0.01) but reduced off-target Cas9 activity to an even greater extent (89% reduction; P < 0.0001) (Fig. 2B). Therefore, this PNA resulted in the largest increase in specificity in the FANCF context, increasing the specificity ratio to over 80 to 1, a 5.3-fold increase in specificity compared with the no-PNA control (Fig. 2B).

The 5–10 6mer and 9–20 12mer PNAs both performed similarly and resulted in the second highest specificity ratios. The 5–10 6mer PNA reduced on-target Cas9 activity by 6% (P < 0.01), while the off-target Cas9 activity was reduced by 61% (P < 0.01) (Fig. 2B). This resulted in a 2.4-fold increase in the specificity ratio compared with the control. Similarly, the 9–20 12mer PNA reduced on-target Cas9 activity by 7% (P < 0.05) and reduced off-target Cas9 activity by 63% (P < 0.001) (Fig. 2B). This resulted in a 2.5-fold increase in the specificity ratio.

The 9–14 6mer PNA had a more moderate effect on the specificity ratio, only increasing the specificity by 1.5 times compared to the no-PNA control, while the 11–20 10mer did not significantly alter on- and off-target Cas9 activity and thus did not alter the specificity ratio (Fig. 2B).

Taken together, the results for the FANCF gene target show that the 7–20 14mer PNA, 9–20 12mer PNA, and 5–10 6mer PNA are all able to increase on-target specificity. While the 7–20 14mer PNA increased the specificity ratio by the largest amount, it also reduced the absolute level of the on-target Cas9 activity the most. This indicates that depending on the application, different PNAs may be more desirable than others, and the system can be tuned depending on the gene target.

PAM distal antispacer PNAs promote increased specificity ratios in an allele-specific context

Next, we wanted to evaluate whether antispacer PNAs would be applicable in an allele-specific context. To determine this, we created cell lines in a U2OS cell background with a dual RFP and GFP reporter system³⁰ (Fig. 3A), incorporating between the reporter genes a region of the TGFBI gene that harbors mutations associated with the autosomal dominant genetic disease, corneal dystrophy.⁸ From 5’ to 3’, the reporter system has mRFP constitutively expressed, followed by the TGFBI sequences, and then two eGFP motifs, which are out of frame by one or two base pairs, respectively (Fig. 3A). We designed this dual GFP reporter cassette to thereby capture two out of three classes of indels (all the ones that would shift the reading frame by either 1 or 2 bp). This reporter system will not detect indels that retain the reading frame (which contains a premature stop codon) or that would result in another in-frame stop codon. Hence, two thirds of the frameshifts should shift one or the other of the two eGFP alleles into the frame, thus capturing indel events 2/3 of the time. Edited cells will then express both RFP and GFP, which can be discerned by flow cytometry (representative gating in Supplementary Fig. S2). RFP levels do not vary upon treatment with Cas9 RNP or PNA and serve as a control (Supplementary Fig. S4). Cell lines were created with different versions of the TGFBI domain with either the wild-type or disease-associated alleles: the wild-type TGFBI R124, mutant TGFBI R124L, and mutant TGFBI R124H. The wild-type sequence differs from the mutant alleles by only one base pair (Supplementary Table S1). Consistent levels of RFP expression were confirmed for all reporter cell lines by flow cytometry to rule out variations in construct expression levels (Supplementary Fig. S4).

FIG. 3.

Peptide nucleic acids (PNAs) improve the specificity of Cas9 targeting in an allele-specific manner. (A) Schematic of an RFP/GFP reporter system used to determine the specificity of the cutting activity of Cas9 in human cells. (B) Relative editing percentages in the TGFBI R124L mutant allele (gray) and the TGFBI R124 WT allele (gray stripes) by Cas9 RNPs complexed with the indicated antispacer PNAs along with the resulting specificity ratios (white). Two-tailed t-tests were performed comparing the PNA conditions with the no-PNA control. (C). Relative editing percentages in the TGFBI R124H mutant allele (gray) and the TGFBI R124 WT allele (gray stripes) by Cas9 RNPs complexed with the indicated antispacer PNAs along with the resulting specificity ratios (white). N/A indicates that the specificity ratio could not be calculated because editing in the R124 WT cells was undetectable. Two-tailed t-tests were performed comparing the PNA conditions with the no-PNA control.

gRNAs were designed to be specific for each of the mutant alleles (R124L or R124H) with the intent of mediating targeted knockout, and these were tested in cells containing reporter constructs with either the respective mutant allele or the wild-type allele. We synthesized a series of antispacer PNAs to bind to the spacer regions of these guides to potentially modulate their activity, varying the length and binding position similar to the experiments above. The sgRNAs were then annealed to the PNAs and complexed with Cas9. Cells were nucleofected with the Cas9 RNPs, and flow cytometry was performed 72 h later to quantify indel formation in the respective targets.

When the PNAs are tested with the gRNA targeting R124L in the TGFBI R124L target cells, we see that the 1–6 6mer PNA completely inhibits Cas9 activity, while the 5–10 and 9–14 6mer PNAs have little effect on Cas9 activity in the R124L mutant allele (Fig. 3B). They also did not significantly affect Cas9 activity in the R124 WT allele and thus had little overall change in the specificity ratio (Fig. 3B). The 11–20 10mer and 9–20 12mer PNAs had a moderate effect on Cas9 activity in the R124L mutant allele (P < 0.05 and P < 0.05, respectively) and the R124 wild-type allele (P < 0.01 and P < 0.001, respectively) (Fig. 3B). Therefore, these PNAs resulted in only moderate increases in the specificity ratio as compared with the no-PNA control. In contrast, the 7–20 14mer PNA reduces Cas9 activity against the mutant allele by 39% (P < 0.001) and by 94% against the wild-type allele (P < 0.01) (Fig. 3B). Hence, the 7–20 14mer PNA resulted in a specificity ratio of almost 40 to 1, a 10-fold increase in specificity compared with the no-PNA control group.

PNAs were also tested in combination with gRNAs targeting the R124H allele versus the wild-type allele containing reporter cells. In this context, we again see complete inhibition of Cas9 activity in both the wild-type and mutant alleles with the 1–6 6mer PNA (P < 0.01 and P < 0.001, respectively) (Fig. 3C). We also continue to see no significant reduction in Cas9 activity in both the wild-type and mutant alleles when the 5–10 6mer (P = 0.48 and P < 0.05, respectively) and 9–14 6mer (P = 0.28 and P = 0.20, respectively) PNAs are annealed to the Cas9 RNP. When the 11–20 10mer PNA was tested, we saw a 30% reduction in Cas9 activity in the R124H allele (P < 0.05) and a 59% decrease in Cas9 activity in the WT allele (P < 0.05) (Fig. 3C). This resulted in a moderate increase of 1.7 in the specificity ratio. The 9–20 12mer PNA had a 41% reduction in Cas9 activity in the R124H allele (P < 0.01) and an 80% reduction in Cas9 activity in the wild-type allele (P < 0.001), resulting in a threefold increase in specificity ratio. Finally, when the 7–20 14mer PNA was tested, we saw an 87% reduction in Cas9 activity in the R124H mutant allele (P < 0.001) and a 97% reduction in Cas9 activity in the wild-type allele (P < 0.01) (Fig. 3C). This resulted in a specificity ratio of 17 to 1, a fourfold increase compared with the control. These results show that the 7–20 14mer PNAs consistently result in the largest increase in specificity ratios in two different allele-specific contexts.

PAM-proximal antispacer PNAs with key mismatches exhibit modulation of Cas9 activity in human cells

Antispacer PNAs bound to the PAM-proximal region of the gRNA have been shown to completely inhibit Cas9 activity; however, we wanted to further explore whether this proximal region could still be exploited to modulate Cas9 activity by introducing key mismatches between the PNA and the gRNA. We hypothesized that while a PNA with perfect homology to the PAM-proximal region would prevent R-loop formation and thus prevent Cas9 activation, a mismatch might allow for R-loop formation but increase the ΔG, thus favoring binding to higher affinity sites (Fig. 4A). To probe this hypothesis, we synthesized 1–6 6mer PNAs with single mismatches at each position of the PNA for a BFP target (Fig. 4B). We then used the K562-BFP cells previously described to determine how Cas9 cutting activity was affected by the presence of mismatches.

FIG. 4.

Introducing key mismatches into short protospacer adjacent motif (PAM)-proximal peptide nucleic acids (PNAs) to modulate allele-specific targeting. (A) Schematic of mismatched PAM-proximal PNAs effect on Cas9 R-loop formation. (B) Schematic depicting mismatch positions in antispacer PNAs relative to the sgRNA. (C) Indel formation in a K562-BFP cell line by Cas9 RNPs complexed with the indicated antispacer PAM-proximal 6mer PNAs with a single mismatch between the PNA and gRNA. Two-tailed t-tests were performed comparing mismatched PNAs with the 1–6 6mer control PNA. (D) PNA binding positions relative to the gRNA sequence with perfect homology (1–6 6mer PNA) indicated by a green nucleotide and a single mismatch (1–6 m3 6mer PNA) indicated by a red nucleotide. Relative editing percentages in the TGFBI R124H mutant allele (gray) and the TGFBI R124 WT allele (gray stripes) along with the resulting specificity ratio (white). N/A indicates that the specificity ratio could not be calculated because editing in the R124 WT cells was undetectable. Two-tailed t-tests were performed comparing the PNA conditions with the no-PNA control. (E) PNA binding positions relative to the gRNA sequence with perfect homology (1–6 6mer PNA) indicated by a green nucleotide and a single mismatch (1–6 m3 6mer PNA) indicated by a red nucleotide. Relative editing percentages in the TGFBI R124 WT allele (gray) and the TGFBI R124H allele (gray stripes) along with the specificity ratio (white). Two-tailed t-tests were performed comparing the PNA conditions with the no-PNA control.

Mismatches at positions 1 and 6 had no significant difference on Cas9 activity (P = 0.07 and P = 0.35, respectively) as compared to cells treated with a Cas9 RNP annealed with a perfectly homologous 1–6 6mer PNA (Fig. 4C). Mismatches at positions 2 and 4 had less of a reduction in Cas9 activity as compared with the control, with 25% (P < 0.05) and 22% (P < 0.01) indel formation, respectively (Fig. 4C). Finally, mismatches at positions 3 and 5 allowed the most Cas9 activity as compared with the control. A mismatch at position 3 resulted in 48% indel formation (P < 0.01), while a mismatch at position 5 resulted in 45% indel formation (P < 0.01) (Fig. 4C). These results indicate that the PAM-proximal region of the gRNA may be used as another system to modulate Cas9 activity when the PNA contains selected mismatches to the gRNA sequence.

PAM-proximal antispacer PNAs with key mismatches promote increased specificity ratios in an allele-specific context

We next wanted to test PNAs with mismatches in the PAM-proximal region in an allele-specific targeting context. In the TGFBI R124 target, the single base pair mutations between alleles fall within the proximal region of the gRNA. Therefore, there were naturally occurring mismatches between position 3 of the gRNA and the 1–6 6mer PNAs that were synthesized for each of the sequences. These PNAs were then tested using the method previously described.

When the R124H gRNA was used, we saw complete inhibition of Cas9 cutting activity in both the mutant (P < 0.001) and wild-type alleles by the 1–6 6mer PNA (P < 0.01) (Fig. 4D), and so there was no meaningful effect on specificity. However, with the 1–6 6mer PNA with a mismatch at position 3 (1–6 m3 6mer), we saw differential effects on Cas9 activity in the mutant (P = 0.79) and wild-type alleles (P = 0.13) when the PNA was annealed with the Cas9 RNP (Fig. 4D). This condition thus showed a moderate increase in the specificity ratio when compared with the no-PNA control.

When the R124 WT gRNA was used, we again saw almost complete inhibition of Cas9 activity by the 1–6 6mer PNA in both the wild-type (P < 0.01) and R124H alleles (P < 0.01) (Fig. 4E). When a 1–6 6mer PNA with a mismatch at position 3 was used (1–6 m3 6mer), we saw a 64% reduction in Cas9 activity in the wild-type allele (P < 0.05) and an 88% reduction in Cas9 activity in the R124H allele (P < 0.01) (Fig. 4E). This resulted in a 3.1-fold increase in specificity. These data show that mismatches between the PNA and the gRNA in the PAM-proximal region are a feasible alternative strategy to modulate Cas9 activity and thus make PNAs further tunable depending on the gene target and application.

Additional characterization of PNA modulation of Cas9 RNP activity

While most of the data throughout this article has been reported as GFP fluorescence readouts based on flow cytometry, we also wanted to confirm that editing was occurring at the genomic level at levels comparable to the fluorescence measurements. To this end, R124H U2OS cells were treated with PNA/Cas9 RNP and, 72 h later, cells were harvested for not only flow cytometry but also gDNA extraction for sequencing. Illumina sequencing was performed, and the results were compared to the fluorescence readout (Supplementary Fig. S5). We found, as expected, that the sequencing method detected higher frequencies, consistent with the ability of the sequencing to detect not only frameshift mutations but also mutations that retain the reading frame. Importantly, however, the proportional differences between experimental samples were preserved, as there was a good correlation of the editing frequencies as measured by sequencing compared with GFP expression. Interestingly, all the conditions tested had peaks at deletions that were −1 and −9 in length and insertions that were +1 in length. Since the −9 deletion would keep the sequence in frame, these editing events would not have been captured with our flow cytometry readout.

We also wanted to determine if treatment with the PNA/Cas9 RNP complexes resulted in any cytotoxicity in the cells. We had previously tested similar PNA/Cas9 RNP reagents for impact on cell viability in the K562-BFP cell line and found no evidence of cytotoxicity.²⁸ However, we were also interested in testing the impact of the PNA/Cas9 RNP complexes on R124 U2OS cell viability. R124H cells were treated with PNA/Cas9 RNP and then, 72 h later, a cell titer glo assay was performed to determine cell viability relative to a mock-nucleofected control. We found no significant toxicity in cells treated with PNA/Cas9 RNPs (Supplementary Fig. S6).

We were also interested in determining if the different levels of Cas9 modulation achieved with different length PNAs could be correlated with the Tm of the PNA/RNA duplexes. To test this, we combined PNA and the appropriate 20mer gRNA sequence (Supplementary Table S3). The absorbance was then measured from 25 to 95°C, and the first derivative of the curve was taken to determine the Tm of each PNA (Supplementary Fig. S7). These melt curves were performed for most of the PNAs used in the K562-BFP system as well as the R124 system, and all melting temperatures are reported in Supplementary Table S4. We found that as the length of the PNA increased, so did the melting temperature, with the shorter homologous 6mer PNAs having melting temperatures ranging from 40 to 65°C, while the longer PNAs, which were generally more effective at increasing the specificity ratio of the system, had higher melting temperatures ranging from 82 to 95°C. The longest PNA tested was a 16mer, which completely blocked Cas9 activity and had a first derivative peak of 95°C.

We also performed the same thermal annealing experiments described above for corresponding DNA/RNA duplexes in the R124 system (Supplementary Fig. S7). When comparing the difference between melting temperatures of the PNA/RNA versus DNA/RNA complexes, we found that as the length increased, so did the delta. The shorter 6mer PNAs/DNAs that were homologous to the gRNA had deltas ranging from 11 to 17°C, while the longer 10, 12, and 14mer sequences had deltas ranging from 46 to 50°C (Supplementary Fig. S7, Supplementary Table S4). Since the PAM distal 14mer PNAs typically showed the greatest improvement in specificity, these results provide a target range of differential melting temperatures as a starting point for designing reagents.

Finally, in our prior publication,²⁸ we performed a series of time-course experiments in which Cas9 RNPs were first transfected into cells without PNAs, followed by separate transfection of the PNAs. These prior experiments showed that some PNA-mediated inhibition of Cas9 activity occurred even when transfected separately, although the effectiveness decreased as the time between transfections increased.²⁸ We were interested in determining how the new PNAs tested in this article would impact Cas9 activity when a sequential treatment was implemented. We treated K562-BFP cells with Cas9 RNPs and then transfected PNAs 2 h later. While we see that Cas9 activity is in some cases still inhibited by the PNAs, the effectiveness is much less robust than what is seen with preannealing the PNA to the sgRNA (Supplementary Fig. S8). These results indicate that while posttreatment of PNA still results in the modulation of Cas9 activity, preannealing the PNA to the sgRNA results in a more robust effect.

Discussion

Here, we show that antispacer PNAs are able to modulate Cas9 cutting activity when annealed to the gRNA sequence of the Cas9 RNP complex. We demonstrate that varying the length and binding positions of PNAs within the gRNA region can have different effects on Cas9 activity. For the targets tested, we show that a 7–20 14mer PNA has the greatest impact on Cas9 specificity, inhibiting cleavage of off-target sites by a significantly greater degree than in on-target sites. While this results in the 7–20 14mer PNAs increasing on-target specificity by the largest amounts, it should be noted that a limitation of these 14mer PNAs is the reduction of on-target editing. Many of the 14mer PNAs tested in this article show at least a 50% reduction in on-target Cas9 activity, which may be suboptimal when considering use in potential therapeutic treatments. Therefore, it may be necessary to consider multiple doses of the PNA/Cas9 RNP complex to achieve similar levels of on-target editing as seen with an unmodified Cas9 RNP. However, the reduction of off-target editing achieved with the use of the 14mer PNAs may make the reduction in on-target Cas9 activity worthwhile when considering the balance between efficacy and safety in the clinic and may prove necessary in order to accomplish an allele-specific effect that is otherwise not feasible for standard technology.

We also show that 9–20 12mer PNAs maintain on-target Cas9 activity while reducing off-target activity. While the resulting increase in specificity is not as robust as what is seen in the 7–20 14mer PNAs, the phenomenon observed may prove useful in gene targets when maximizing on-target Cas9 activity is of utmost importance.

Finally, while 1–6 6mer PNAs homologous to the gRNA sequence completely inhibit Cas9 activity, introducing key mismatches between the PNA and gRNA sequence allows for some Cas9 activity compared with controls. We found that selective mismatches can increase the specificity in allele-specific contexts and thus provide another avenue for Cas9 modulation. This may be useful in targets where mutations lie close to the PAM site or when PAM distal PNAs are less feasible.

Although this article has established trends for PNAs designed for the antispacer region of the gRNA, general guidelines for optimizing PNA design have not been fully established. UV spectroscopy and thermal melting experiments were conducted to examine the melting temperatures of various PNAs. As expected, longer PNAs had a higher melting temperature and delta when compared with the corresponding DNA/RNA duplexes. The most effective PNAs for mediating a differential effect on specificity (typically the 14mers) had melting temperatures in the range of 85–90°C and deltas in the range of 45–50°C. While these results provide initial guidelines for reagent design, further work will be helpful to develop more concrete rules for PNA design and represent a limitation of the current study.

It should also be noted that shorter PNA designs may have a higher chance of off-target binding to other RNAs within cells. Care should be taken when designing PNA sequences to minimize the potential for off-target binding effects. Preannealing the PNA to the sgRNA, where possible, depending on the application and as performed here, may reduce the possibility of off-target events as compared to treating the cells with separate doses of PNA and Cas9 RNP. PNA binding to off-target sites should be explored further in future work in order to continue optimizing the PNA/Cas9 RNP system for potential therapeutic uses.

Other parameters regarding PNA chemistry should also be explored to further reduce the limitations of this system. Sidechain substitutions at the gamma (γ) position in the PNA backbone may be beneficial to explore in future studies. It has been shown that when certain chemical sidechains, such as a polyethylene glycol (mini-PEG) or a hydroxymethyl group (serine), are added at the γ position, the resulting PNAs have increased binding affinity and are preorganized into a helical structure.^19,31 γ-PNAs therefore have the potential to improve specificity ratios even further by modifying the binding affinity of the PNAs to the sgRNA.

Taken together, these results indicate that the antispacer PNA approach is highly tunable and can be adjusted and applied to a variety of targets and applications. We have found that PNAs are nontoxic to cells used in this study (Supplementary Fig. S6), can be used to reduce Cas9 cutting in off-target sites, and are also applicable in allele-specific contexts. This indicates that PNAs may be useful in the treatment of autosomal dominant diseases, such as TGFBI corneal dystrophy, in which knockout of the mutant alleles is needed, but preservation of the wild-type alleles is still required for function. While we have focused on examining PNAs annealed to sgRNAs throughout this article, antispacer PNAs can also be used as a pretreatment²⁸ or posttreatment²⁸ (Supplementary Fig. S8) to Cas9 RNP, further illustrating their versatility. They can theoretically be applied to other Cas systems as well as a wide range of disease models and genetic targets.

Data Availability Statement

All data generated or analyzed during this study are included in this published article (and its supplementary information files, including source data). There are no restrictions on data availability.

Author Disclosure Statement

P.M.G. is a founder of and consultant for Cybrexa Therapeutics and Gennao Bio, is a consultant to pHLIP Inc. and to Immunome, and holds equity in Patrys Ltd. None of these companies have any relationship to the work in this article. N.G.E. and P.M.G. are inventors on US patent application no. 63/197,879 submitted by and assigned to Yale University, which covers compositions and methods for CRISPR-Cas9 modification using PNAs.

Funding Information

This work was supported by grants from the NIH (U01AI145965 and R35CA197574 to P.M.G.) and by an award from the Blavatnik Fund for Innovation at Yale to P.M.G.