Programmable site-specific DNA double-strand breaks via PNA-assisted prokaryotic Argonautes

Abstract Programmable site-specific nucleases promise to unlock myriad applications in basic biology research, biotechnology and gene therapy. Gene-editing systems have revolutionized our ability to engineer genomes across diverse eukaryotic species. However, key challenges, including delivery, specificity and targeting organellar genomes, pose barriers to translational applications. Here, we use peptide nucleic acids (PNAs) to facilitate precise DNA strand invasion and unwinding, enabling prokaryotic Argonaute (pAgo) proteins to specifically bind displaced single-stranded DNA and introduce site-specific double-strand breaks (DSBs) independent of the target sequence. We named this technology PNA-assisted pAgo editing (PNP editing) and determined key parameters for designing PNP editors to efficiently generate programable site-specific DSBs. Our design allows the simultaneous use of multiple PNP editors to generate multiple site-specific DSBs, thereby informing design considerations for potential in vitro and in vivo applications, including genome editing.


INTRODUCTION
Members of the Argonaute protein family have been identified in all domains of life ( 1 , 2 ).Among prokaryotes, 32% of Archaea and 9% of Eubacteria harbor genes that encode members of the Argonaute superfamily ( 3 ).Prokaryotic Argonaute (pAgo) proteins function as an innate immune system to fend off invading genetic elements from bacteriophages and conjugati v e plasmids ( 4 , 5 ).pAgos are classified as long pAgo, short pAgo and PIWI-RE proteins ( 6 ).The long-A pAgos have characteristic N, PAZ (PIWI / Argonaute / Zwille), MID (middle) and PIWI (P-element-induced wimpy testes) domains.The MID and PAZ domains mediate the binding of the 5 and 3 ends, respecti v ely, of a guide nucleic acid molecule ( 7 ).In contrast to eukaryotic Argonautes that use a single-guide RNA, most described pAgos rely on small single-stranded guide DN A (gDN A) molecules to target DNA sequences.The PIWI domain mediates the catalytic activity of pAgo and possesses an RNase-H-like fold with a DEDX catalytic motif mediating the endonuclease activity, in most cases by slicing the target sequence between nucleotides 10 and 11 in relation to the 5 end of the guide molecule ( 8 ).
How pAgos acquire their guides and elicit their function in vivo has onl y partiall y been elucidated.Many pAgos possess guide-independent activity that is suspected to be involved in guide acquisition from highly expressed or repetiti v e regions of target plasmids ( 9 ).Importantly, pAgos lack intrinsic helicase activity and are thus unable to unwind target double-stranded DN A (dsDN A) on their own and possess only one nuclease domain ( 3 ).Ther efor e, the generation of double-strand breaks (DSBs) requires the binding of two pAgo complexes, on the upper and lower strands of the target DNA sequence.
Many studies have attempted to harness and de v elop pAgos as programmable DNA editors for genome-editing applica tions.Initial a ttempts were limited to pAgos from hyperthermophiles such as Pyrococcus furiosus and Thermus thermophilus to generate dsDNA breaks due to temperature dena tura tion ( 10 , 11 ).With the discovery of pAgos from mesophilic organisms such as Clostridium butyricum , Limnothrix rosea and Kurthia massiliensis , their applications remained limited to regions of supercoiled DNA substrates with low GC content (12)(13)(14).Various attempts have been made to overcome this challenge, such as deploying singlestranded DN A (ssDN A)-binding proteins and combining CbAgo with a RecBC helicase from Esc heric hia coli ( 15 ).Whether these strategies can be successfully applied to eukaryotic cells in vivo remains to be explored.Indeed, the lack of robust pAgo activity at physiological temperatures precluded their translational applications in gene editing and biotechnology and limited their use to bacterial cells ( 16 , 17 ).
Peptide nucleic acids (PNAs) are synthetic oligonucleotide analogs characterized by a neutrally charged 2-aminoethyl glycine backbone replacing the canonical sugar-phosphate backbone ( 18 ).As a result of their neutral charge, PNAs can bind to complementary DNA or RNA with high affinity and specificity, resulting in hybrids that ar e mor e stable than their naturally occurring RNA or DNA counterparts ( 19 , 20 ).Additionally, their modified backbone renders PNAs stable and resistant to cleavage by proteases and nucleases ( 21 , 22 ).Owing to their remar kab le properties, PNAs have been employed for diverse biotechnolo gical a pplications, such as inhibition of PCR, transcription, translation and design of improved fluorescence in situ hybridization probes ( 21 , 23 ).
The potential use of PNAs for genome editing relies on the ability of PNAs to form triplexes with genomic DNA in a site-specific manner ( 24 , 25 ).PNAs have been deployed in gene editing to correct genetic diseases (26)(27)(28), and they have been used along with DNA oligomers to carry inf ormation f or homolo gous recombination.PN As were hypothesized to invade target DNA, generating distortions in DNA helix that is recognized and resolved by the nucleotide excision repair pathway, leading to stimulated recombination in the presence of externally provided DNA repair template.For example, PNAs modified at gamma position ( ␥ PNAs) and donor DNA oligomers have been delivered intravenousl y via pol y(lacticco -gl ycolic acid) nanoparticles into a ␤-thalassemia mouse model for the amelioration of this genetic disease ( 29 ).This PNA treatment led to a correction of the β-globin gene ( 30 ).PNAs hold much promise for gene therapy; howe v er, the low efficiency of gene correction limits their applications in gene editing ( 30 , 31 ).
Here, we harnessed the features of PNA molecules and pAgos and de v eloped PNA-assisted pAgo editors (PNP editors) to generate programmable site-specific DSBs independently of the GC content and DNA form at ambient temperatures.We harnessed the power of PNAs to mediate targeted and specific DNA invasion and the programmability of guided pAgos to bind to any ssDNA sequence of interest specifically and with high efficiency to generate targeted DSBs.Our work shows that PNA-assisted pAgo can generate single and multiple site-specific DSBs and targeted modifications for in vitro and in vivo applications.The PNP editor technology holds great promise for genome editing and clinical applications in gene therapy.

Design and synthesis of different ␥PNA and ␥tcPNA molecules
␥ PN A and ␥ -substituted tail-clamp PN A ( ␥ tcPN A) molecules were designed based on previous reports ( 23 , 29 ) and custom synthesized by PANAGENE Inc. based on general PNA synthesis guidelines ( https://www.pnabio.com/support/PN A Tool.htm ).All ␥ PN As and ␥ tcPNAs were ␥modified with alanine molecules and three lysine moieties to facilitate solubility and invasion efficiency (Supplementary Table S2).

Cloning of ␥PNA invading target regions into pMRS and pUC19 plasmids
␥ PNA invading target inserts wer e pur chased as forward and re v erse oligonucleotides from Integrated DNA Technologies Inc., with overhangs corresponding to BamHI and EcoRI restriction sites for cloning in pMRS and pUC19 vectors (Supplementary File S2A and B).All target sequences are listed in Supplementary Table S3.Oligonucleotides were independently phosphorylated using T4 PNK (Promega) before annealing.Plasmids were digested using BamHI-HF and EcoRI-HF in 1 × CutSmart buffer (NEB) for 8 h.Phosphorylated dsDNA inserts were mixed with 10 ng of digested plasmid purified from agarose gel in 3:1 molar ratio and incubated with T4 DNA Ligase (Promega) in 1 × T4 DNA Ligase Reaction Buffer (Promega; catalog # M1801) for 3 h at 23 • C, followed by 2 h at 16 and 4 • C until transformation of E. coli competent cells (Thermo Scientific, One Shot ™ TOP10 Chemically Competent E. coli , C404010) with 2 l of the ligation reaction using a heat shock method.Transformed cells were plated on agar plates containing 50 g / ml kanamycin (pMRS) or 100 g / ml ampicillin (pUC19).Plasmids were isolated from bacterial liquid cultures using a QIAprep Spin Miniprep Kit (Qiagen, 27106) and adjusted to 100 ng / l for downstream applications.All clones were confirmed by Sanger sequencing using the primers listed in Supplementary Table S4.

PNA invasion and mobility shift assay
PNA invasion tests were performed in 10 l reaction volume consisting of 1 × MOPS buffer (20 mM MOPS, pH 7.0, 5 mM CH 3 COONa and 1 mM EDTA), 50 nM linear dsDNA target ( ∼350 bp) and 2 M PNA.Target dsDNA was generated by PCR amplification from pMRS plasmid using primers flanking the PNA target regions (1442 and 1557, Supplementary Table S4).Reactions were incubated overnight at 37 • C, mixed with 5 × Nov e x Hi-Density TBE Sample Buffer (Invitrogen, LC6678) and run on a Nov e x ™ 6% nati v e TBE gel at 180 V for 1 h.Gel was stained with 1 × SYBR Gold (Invitrogen, S11494) for < 10 min and visualized using a Molecular Imager Gel Doc XR+ System (Bio-Rad).

Argonaute in vitro cleavage assay of circular and linear DNAs
All circular plasmid DNA targets were linearized by restriction digest with the desired restriction enzyme; the resulting linear products were purified with a QIAquick ® Gel Extraction Kit (28706) as per the manufacturer's guidelines.Prior to pAgo protein cleavage, DNA targets were incubated with the two PNA molecules for 45 min (for circular DNA) or overnight (for linear DNA) at 37 • C. PNA invasion r eactions wer e conducted in 10 l reaction volume containing 200 ng of DNA template (pMRS target; ∼5.3 kb, corresponding to a pproximatel y 6 nM), 1 × MOPS buffer, PNA1 (final concentration of 100 nM) and PNA2 (final concentration of 100 nM).Further, pAgo-mediated cleavage was achie v ed in two steps: In the first step, pAgo was loaded with two different guides independently in half reaction 1 and half reaction 2 at 37 • C for 15 min.All guide sequences used in this study are listed in Supplementary Table S5.Each half reaction contained 1 × final pAgo reaction buffer [1 × composition: 10 mM HEPES-NaOH, pH 7.0, 100 mM NaCl, 1 mM MnCl 2 and 5% (v / v) glycerol], 1 M guide and 1 M recombinant pAgo in the total volume of 8 l, unless stated otherwise.In the second step after guide loading, the two half reactions (8 l each) were combined in one tube, to which 4 l of PNA-invaded or non-invaded plasmid DNA was added (80 ng target DNA final, corresponding to approximately 1.2 nM).The total volume of the cleavage reaction was 20 l.Each cleavage reaction was incubated at 37 • C for 60 min, with the lid temperature set to 39 • C. In the case of circular plasmid DNA, following the pAgo cleavage reactions, 2 l of 10 × CutSmart and 1 l of the appropriate restriction enzyme (10 units / 1 l) were added to the reaction and incubated at 37 • C for 30 min.Next, 1 l of proteinase K (Invitro gen; catalo g # 25530049) was added, followed by incubation at 37 • C for 30 min.Four microliters of 6 × Gel Loading Dye Purple (NEB; catalog # B7024S) was added to each sample, which was then loaded onto a 0.9% (w / v) agarose gel with GelRed ® and electrophoresed for 1 h and 30 min at 145 V. Finally, the gel was visualized using a FluorChemQ Gel Doc system.All r eactions wer e set up a t room tempera ture.The table with detailed protocol can be found in Supplementary Data.

pAgo cleavage site identification
For the determination of the cleavage site, ␥ PNA5 + ␥ PNA6 binding target region was cloned into the pMRS plasmid at BamHI and EcoRI restriction sites such that the regions inv aded b y ␥ PN A5 + ␥ PN A6 were located on the opposite strand of the primer binding site, since ␥ PNAs are known to inhibit strand elongation.Cleavage of targets ␥ PNA5 + ␥ PNA6 in pMRS linearized by BsrGI restriction digest was performed as described above; the products were separated from 1% (w / v) agarose gels.The bands corresponding to cleavage products were excised and purified from the agarose gel using a QIAquick ® Gel Extraction Kit (28706) and subsequently sequenced by Sanger sequencing using specific primers (Supplementary Table S4).Cleavage sites were determined by identifying read termination sites.

Ov ervie w and working principle of the PNP editors
pAgo proteins lack intrinsic helicase activity and cannot unwind dsDNA targets to cleave them, limiting their adoption for biotechnological applications such as genome editing.The ability of different PNA molecules to invade ds-DN A in a highl y specific manner prompted us to investigate whether this property might be exploited to assist mesophilic pAgos in performing site-specific DSBs ( 32 ).Addition of functional groups to the ␥ -position of PNAs pre-organizes the helix, facilitating invasion of the canonical B-DNA f orm.We theref ore tested the invasion efficiency of different ␥ PNAs and confirmed that these molecules can invade DNA substrates, albeit with different efficiency (Figure 1 A-C).
Encouraged by this result, we designed our concept based on the simultaneous use of two ␥ PN A molecules ca pable of invading DNA substrates, in close proximity and on opposing DNA strands.The resulting invasion into the DNA helix w ould mak e a segment of displaced ssDNA available for pAgo enzymatic activity (Figure 1 D).Simultaneous nicking on opposing DNA strands by the two pAgo-guide complexes should generate DSBs that could be harnessed for di v erse in vitro and in vivo genome engineering applications (Figure 1 D).

pAgos generate site-specific DSBs on DNA substrates independently of GC content and DNA form
To test the above concept, we designed two ␥ PNA molecules ( ␥ PNA1 and ␥ PNA3) to invade specific DNA se-quences cloned in the pMRS plasmid on opposite strands, separated by a single-nucleotide spacer (Figure 1 E).We produced and purified recombinant CbAgo and KmAgo from E. coli .We then loaded these proteins with corresponding 5 -phosphorylated gDNAs (5 P-gDNAs) and incubated the resulting pAgo complexes with intact plasmid substrate or a plasmid substra te tha t had been invaded by ␥ PNA1 and ␥ PNA3.We determined that non-invaded circular plasmid is nicked by CbAgo and KmAgo in a guideindependent manner without generating DSBs (Figure 1 F).In contrast, we detected the generation of guide-dependent DSBs on PNA-invaded plasmid substra tes, indica ti v e of doub le-nicking acti vities that r esulted in the r elease of a DNA fragment of the expected size following restriction digestion (Figure 1 F).
We also tested the double-nicking activity and generation of specific DSBs on linear DNA substrates; to this end, we linearized the pMRS plasmid by BsrGI restriction digest and used the resulting linear DNA molecule as substrate (non-invaded or invaded with ␥ PNA1 and ␥ PNA3) for the same pAgo cleavage assay as above.We observed that CbAgo and KmAgo can specifically bind to linear DNA substra tes tha t had been invaded with the two ␥ PNAs to generate DSBs, as evidenced by the release of the band of the expected size (Figure 1 G).Non-invaded DNA substrates did not show any band release, suggesting that ␥ PNA invasion is a necessary pr er equisite for pAgo binding and cleavage (Figure 1 G).
CbAgo and KmAgo nuclease activity on supercoiled DNA substra tes in vitr o is limited to regions of low GC content at physiological temperatur es, and decr eases inversely proportionally with increasing GC content, indicating that strand unwinding is one of the limiting factors for harnessing the power of pAgo for biotechnological applications ( 13 , 14 , 33 ).We asked whether the barrier to DNA unwinding presented by high GC content in the target sequence might be overcome by employing our concept.Accordingly, we assessed the induction of DSB formation in non-invaded circular and linear DNA targets at regions with increasing GC content ranging from 12% to 75%.In agreement with previous studies, we detected some catalytic activity for CbAgo and KmAgo on circular plasmids with target sequences of low GC content.Specifically, we established that CbAgo exhibits catalytic activity on DNA sequences with GC contents of 13% and 22%, whereas KmAgo exhibited catalytic activities on DNA sequences with GC contents of 31% and below (Figure 2 A).Howe v er, neither CbAgo nor KmAgo showed any catalytic activity on linear DNA substrates, e v en ov er low-GC-content regions (Figure 2 B).Notably, CbAgo and KmAgo demonstrated robust catalytic activities on DNA sequences with GC contents of 50% and 75% if DNA had been previously inv aded b y the corresponding PNAs, indicating that GC content and DNA form are not limiting factors for our PNP editors, thus unlocking the activities of pAgos on all DNA substra tes a t arbitrary sequences.

Effect of different ␥PN A combinations, ␥PN A length and different types of PNAs on PNP editor activity
To test the versatility and programmability of the PNP editor concept, we designed two additional ␥ PNA molecules,  Linearized, non-invaded pMRS plasmid was incubated with pairs of pAgo-gDNA complexes targeting regions with GC content of 13%, 22%, 27%, 31%, 35%, 50% and 75% (lanes 1-7) at 37 • C for 1 h.Linearized plasmid was then invaded by ␥ PNA1 and ␥ PNA3 at regions with 55% and 75% GC content, respecti v ely, and incubated with the respecti v e pAgo targeting complexes at 37 • C for 1 h (lane 8).Non-invaded pAgo-mediated cleavage (lane 9) and SacI-digested (lane 10) samples were used as control reactions.Expected cleavage product sizes are listed on top of each gel lane.Lane M r epr esents the 1-kb plus DNA ladder.
cloned their respecti v e targets in various combinations ) in the pUC19 plasmid and assessed the formation of DSBs (Supplementary Figure S1A).We established that all ␥ PNA combinations result in pAgo-mediated site-specific DNA cleavage, albeit with variable efficiencies that are likely due to varying invasion efficiencies of different ␥ PNA molecules (Supplementary Figure S1B).In particular, the ␥ PNA1 + ␥ PNA3 and ␥ PNA3 + ␥ PNA4 combinations displayed the highest efficiency for genera ting DSBs.Notably, the ef ficiency of DSB genera tion by CbAgo and KmAgo followed the same pa ttern, indica ting that ␥ PNA invasion efficiency is the limiting step in the reaction.
We also investigated whether a specific PNA length is required for proper function of PNP editors.For this pur-pose, we synthesized a range of truncated ␥ PNA1 and ␥ PNA3 molecules ranging from 10 to 20 nt in length.We invaded the linearized pMRS vector containing the cloned target regions with these shorter ␥ PNA molecules.We also preloaded pAgo with a 16-nt-long gDNA and performed cleavage reactions (Supplementary Figure S2A and B).We detected cleavage products onl y w hen ␥ PN A1 and ␥ PN A3 were 20 nt in length, suggesting that both CbAgo and KmAgo r equir e the str etch of PN A-displaced ssDN A to be longer than the corresponding guide for optimal activity.To confirm this observation, we employed different ␥ PNAs and guides of equal length to cleave the linearized pMRS plasmid.We observed weak activity only when employing 20-nt-long PNAs and corresponding guides of the same length and no activity from shorter pairs of PNAs and guides (Supplementary Figure S2C and D).We specula te tha t this activity might be attributed to a larger segment of e xposed ssDNA availab le for pAgo activity generated specifically by the longest ␥ PNA employed.
We set out to expand the PNP concept by using ␥ tcPNA f or DNA in vasion.␥ tcPNA f orms a triplex structure with target DNA via both Watson-Crick and Hoogsteen base pairing.Owing to this property, ␥ tcPNA forms exceptionally stable invasion products that can be harnessed for use with PNP editors.We designed two ␥ tcPNA molecules and constructed target plasmids containing their complementary DNA sequences.The mobility shift assay showed that ␥ tcPN A can efficientl y invade dsDNA templates (Supplementary Figure S3A).Cleavage assays also re v ealed efficient activity of a linear dsDNA target inv aded b y ␥ tcPNA1 and ␥ tcPNA2 with both CbAgo and KmAgo, whereas the non-invaded target exhibited no cleavage (Supplementary Figure S3B and C).These results demonstrate that DNA unwinding can be achie v ed by triple x-forming PNAs, without steric hinderance affecting pAgo binding and cleavage.

Multiple x ed, site-specific generation of DSBs using PNP editors on plasmid DNA
To demonstra te tha t PNP editors can sim ultaneousl y induce multiple DSBs, we constructed a pMRS plasmid containing two regions each targeted by a different pair of PNP editors (Figure 3 A).We performed the assay using CbAgo and KmAgo following sim ultaneous ␥ PN A invasion at two differ ent r egions on cir cular and linear plasmid DNA.We observed the release of a 820-bp fragment and all expected bands in both cases, indicating that all four ␥ PNAs invaded the DN A substrate, w hich facilitated pAgo binding and cleavage.In contrast, targeting the PNP editors to a noninvaded DNA template failed to release any fragment (Figure 3 B and C).

Effect of spacer length and guide orientation on PNP editor activity
Our concept depends on PNA invasion of sequences at opposite DNA strands, leading to the displacement of short stretches of ssDNA that serve as substrates for pAgo recognition and cleavage (Figure 4 A and B).Therefore, spacer length between the two ␥ PNA-invaded strands and the orientation of the two guides may be key parameters for the design of efficient PNP editors (Figure 4 C).We tested PNP editor efficiency on targets with different spacer lengths (1, 3, 5, 6, 10, 15, 20 or 30 nt) guided by pAgos with gDNAs in two distinct orientations (inward and outward) (Figure 4 D and E).CbAgo activity increased with longer spacers, reaching a peak on targets with 10-nt spacers when using outward-facing guides.In contrast, we noticed an overall lower activity for CbAgo with inward-facing guides, as we detected slightly higher activity on targets with shorter spacers (Figure 4 E).KmAgo showed higher cleavage efficiency with shorter spacers, peaking at 1-nt spacer length (Figure 4 D and E).Unlike CbAgo, KmAgo activity was independent of guide orientation, possibly reflecting structural and steric differences in target DNA binding.Finally, in all cases, target plasmids with spacer lengths of at least 15 nt did not show any cleavage (Figure 4 D and E).

pAgo guide r equir ements f or efficient gener ation of DSBs
gDNA length is a key factor for robust pAgo catalytic activity.Ther efor e, we determined the best-suited length for gDNA to achie v e optimal performance of CbAgo and KmAgo on dsDNA templates inv aded b y ␥ PNAs of 20 nt in length.To this end, we used 5 P-gDNAs of different lengths and performed catalytic activity assays on linearized pMRS plasmid containing ␥ PNA1 and ␥ PNA3 target regions.We determined that a guide length of 16 nt is optimal for CbAgo and KmAgo catalytic activity, resulting in the generation of DSBs evidenced by the release of a band of the expected size (Figure 5 A).This finding is consistent with reported optimal guide length r equir ements for CbAgo and KmAgo activity on ssDNA ( 12 , 33 ).In conclusion, both pAgos perform optimally with 16-nt-long guides, but KmAgo demonstrated higher flexibility by employing guides with a broader range of lengths than CbAgo.
pAgos were shown to mostly use 5 -phosphorylated short DNA molecules to direct their activity in vitro and in vivo .We asked whether the 5 modification of gDNA molecules would affect pAgo activity on their PNA-invaded dsDNA substrates.Moreover, we wanted to explore whether CbAgo and KmAgo might bind to PT-guides for potential application in living organisms to pre v ent guide degradation by nucleases.First, we preloaded CbAgo and KmAgo with guides harboring phosphor othioate gr oups at different positions and tested their activity on ssDNA targets.We observ ed good acti vity in all cases; howe v er, in the case of both Agos, we observed a modest decrease in activity as the number of PT modifications in the guide increased, with the most marked reduction when using fully phosphorothioated guides.Introduction of small number of PT modifica tions a t the 5 and 3 guide ends had little effect on CbAgo, but led to a more pronounced decrease in activity of KmAgo when 5 modified guides were used (Supplementary Figure S4A-C).We also performed cleavage assay on inv aded and non-inv aded linearized plasmid DNA substrates using different types of guide molecules.We loaded pAgos with 5 -OH DNA guides , 5 -phosphorylated PT-guides , RN A guides and 5 P-DN A guides.Consistent with the data on ssDNA, we determined tha t dif ferent guide modifications significantly affect cleavage activity of CbAgo and KmAgo.Both pAgos pr eferr ed 5 P-DNA guides, as expected (Figure 5 B).Our data indicated that both pAgos can employ PT-guides with only a modest reduction in cleavage efficiency.Moreover, only KmAgo could employ 5 -OH DNA guides, but with markedly lower efficiency compared to 5 -phosphorylated guides.As previously reported, CbAgo and KmAgo could bind to 5 P-RNA guides to target DNA, although with substantially lower activity (Figure 5 B).These results demonstrate versatile modes to guide PNP editors.Ne v ertheless, our data confirm that 5 P-DNA guides ar e pr eferr ed for optimal activity, with phosphorothioate bonds in the backbone offering a potentially more viable strategy to guide PNP editors in the context of cellular environments.

PNP editor efficiency at dsDNA flanking regions
We speculated that each ␥ PNA might expose a wider region of ssDNA next to the invasion site and render it targetable by PNP editors.We also wondered whether targeting the non-invaded region flanking the invaded strand with only one ␥ PNA would result in noticeable cleavage of dsDNA in a circular plasmid.We thus designed guides targeting both flanks of invaded pUC19 plasmid as well as guides targeting invaded strands (Supplementary Figure S5A).We observed that pAgos directed against the double-stranded regions flanking the ␥ PNA invasion sites as well as invaded DNA strand fail to generate cleavage products.As previous experiments showed, CbAgo and KmAgo mediated cleavage only when directed to free ssDNA strands exposed by ␥ PNA invasion (Supplementary Figure S5B).
To expand the range of usable gDNAs and elucidate the effect of neighboring dsDNA sequences, we designed guides that can target regions spanning the unwound regions in both inward and outward orientations and tested them on linearized pMRS plasmid.We defined inward orientation as 5 ends of guides facing each other and outward orientation as 5 ends facing away (Supplementary Figures S6A and S7A).When using outward-facing guides, we detected cleavage with guides that contained up to 11 nt in the dsDNA sequence flanking the ␥ PNA invasion (Supplementary Figure S6B).We obtained similar results with inward-facing guides, where the flanking region was located between two invaded regions.In this case, we observed cleavage bands with guides up to 10 nt (with CbAgo) or up to 8 nt (with KmAgo) in the dsDNA region (Supplementary Figure S7B).This experiment revealed that pAgos exhibit catalytic activity even when guides target partiall y unwound DN A, likel y due to the limited ability of gDNA to displace neighboring dsDNA regions.Howe v er, a pre-e xisting ssDNA r egion corr esponding to roughly half the length of the gDN A m ust be present to result in activity.

Effect of temper atur e on ␥PNA invasion and pAgo activity
Robust ␥ PN A invasion and catal ytic activity of the pAgo proteins at physiological temperatures are necessary to harness the power of PNP editors for potential genome-editing applications in various organisms.To elucidate the optimal temperatur e conditions r equir ed f or ␥ PNA in vasion and pAgo activity, we used the pMRS plasmid linearized by BsrGI digest and invaded with ␥ PNA5 and ␥ PNA6 at differ ent temperatur es.To deter mine the perfor mance of pAgos at differ ent temperatur es, we performed ␥ PNA invasion at 37 • C and incubated the target with CbAgo or KmAgo at temper atures r anging fr om 20 to 45 • C. Both pr oteins mediated robust catalytic activities at temperatures as low as 20 • C (Figure 6 A).Both proteins demonstrated cleavage at all temperature conditions, e v en displaying increasing acti vity at ele vated temperatures (Figure 6 A).␥ PNA invasion is critical to expose the ssDNA strand for pAgo activity; hence, it is important to determine whether it occurs at physiolo gicall y r elevant temperatur es.Ther efor e, we performed an overnight invasion of linearized pMRS plasmid with ␥ PNA5 and ␥ PNA6 a t tempera tures ranging from 20 to 45 • C and conducted pAgo cleavage at 37 • C to ascertain strand invasion.We observed that ␥ PNA invasion occurs more efficiently at higher temperatures, effecti v ely starting at 25 • C (Figure 6 B).Finally, we tested both ␥ PNA5 and ␥ PNA6 invasion and pAgo catalytic activity at the same temperature, which we varied from 20 to 45 • C. When both steps were performed at lower temperatures, we detected a cumulati v e decrease in overall activity.Nevertheless, ␥ PNA invasion and pAgo catalytic activity remained efficient at 30 • C and improved when performed at elevated temperatur es (Figur e 6 C).

Determining the pAgo cleavage site in ␥PNA-invaded ds-DNA
Pre vious studies hav e shown that most pAgos cleave DNA targets between nucleotides 10 and 11 from the 5 end of the guide, with some notable exceptions ( 33 , 34 ).Because the precise site of cleavage is an important consideration in genome engineering and related applications, we determined whether the cleavage site is conserved in the PNP editor concept.We thus identified the cleavage site catalyzed by CbAgo and KmAgo in DNA substrates ( 13 , 14 ).Accordingly, we performed a cleavage assay on BsrGI-linearized plasmid that had been invaded with ␥ PNA5 and ␥ PNA6, followed by gel purification of the released bands, which we subjected to Sanger sequencing (Figure 6 D).We established that both proteins cut the displaced DNA strand of the PN A-invaded DN A substra tes a t their canonical site located between nucleotides 10 and 11 from the 5 end of the gDNA.We also observed that PNA invasion does not affect the position of pAgo cleavage, making PNP editors amenable to generating site-specific staggered end DSBs (Figure 6 E).

Effect of gDNA mismatches on the activity of PNP editors
gDNA ar chitectur e can be divided into several segments depending on their position from the 5 end ( 13 ).In most studied pAgos, the first nucleotide at the 5 end is the anchoring nucleotide and binds to the MID domain binding pocket.The seed region is composed of 8 nt from the 5 end of the guide, mediates target recognition and is stabilized by the C-terminal lobe.The central part and the 3 end bind to the PAZ domain ( 3 ).Se v eral reports hav e shown that gDNA mismatches can be tolerated depending on their number and position.We tested the extent to which mismatches in differ ent r egions would compromise the PNP editor activity.We first introduced single-and double-nucleotide mutations in different parts of the guide molecule and in the anchor, seed, central, supplementary and tail regions (Figure 7 A and B).Single-nucleotide mismatches were largely tolerated across the entire gDNA sequence.Howe v er, doub lenucleotide mismatches had a larger effect and compromised pAgo activity most severely when located in the central and supplementary regions.Notably, KmAgo activity was completely abrogated with double-nucleotide mutations in the central guide region (Figure 7 C).To determine the number of guide mismatches r equir ed to completely abolish pAgo activity, we tested up to four mismatches in the gDNA (Supplementary Figure S8A and B).
Our results consistently showed that the cleavage efficiency of CbAgo and KmAgo gradually decreases with more misma tches and tha t misma tches in the and supplementary regions close to the 3 end of the guide have a greater effect on activity than those closer to the 5 end (Supplementary Figure S8B).We observed complete loss of activity with CbAgo when the gDNA contained four mismatches regardless of their position.In contrast, KmAgo tolera ted misma tches a t the 5 end of the guide, e v en with four mismatches.

Effect of ␥PNA activity of PNP editors
The PNP concept relies on a PNA that specifically invades two opposing DNA strands to allow pAgo activity.Since cleavage specificity is paramount for applications in genome editing, we explored the effects of ␥ PNA mismatches on the activity of PNP editors.We introduced mismatches on DNA substrates corresponding to the 3 , 5 , central and nonconsecuti v e regions of the ␥ PNA sequence and tested their influence on pAgo activity.PNP editors in principle possess two layers of specificity resulting from specific PNA invasion and specific gDNA target r ecognition.Ther efor e, we tested specificity using guides that were perfectly complementary to the region with mismatched ␥ PNAs and separa tely with misma tched guides and misma tched invading ␥ PNA (Supplementary Figures S9A and S10A).We introduced four types of misma tches, positioned a t the 5 end and 3 ends consisting of 1-, 2-, 3-, 4-, 5-, 7-, 9-or 11-nt mismatches.Additionally, we introduced up to four consecuti v e mismatches or two, four, six or eight nonconsecuti v e mismatches in the central region of the target DNA.We determined that the position and the number of mismatches between the target and ␥ PNA have a decisi v e influence on PNP editor cleavage efficiency.We observed that ␥ PNAs containing misma tches a t their 5 and 3 termini still retain sufficient ability to invade target DNA, as evidenced by the band release seen on agarose gels, e v en when up to fiv e mismatched nucleotides at the 3 end and se v en mismatched nucleotides at the 5 end were introduced.In sharp contrast, introducing mismatches in the central region of the target resulted in a marked drop of pAgo activity, especially when using CbAgo.With both pAgos, 4-nt mismatches in the center of the target resulted in complete loss of cleavage activity, indicating the inability of ␥ PNA to invade the target region.We detected the most pronounced effect, howe v er, w hen ␥ PN A was directed to invade regions containing nonconsecuti v e mismatches in the center of the target.We only observed a weak band in samples with KmAgo targeted to a target containing two mismatched nucleotides (Supplementary Figure S9).In the second set of experiments, we performed cleavage assays using gDNAs that contained the same number of mismatches and at the same positions relati v e to the ␥ PNA.We observed that mismatches in the PNA r egions and corr esponding gDN A synergisticall y compromise the activity of both proteins.We detected pAgo activity only with one and two ␥ PNA and gDNA nucleotide misma tches a t the 3 end, and one misma tched nucleotide a t the 5 end when using either pAgo.Only KmAgo showed activity when targeted to central mismatched regions and only with one mismatched nucleotide.Additionally, all nonconsecuti v e mismatches in the central region completely abrogated activity in all cases (Supplementary Figure S10).Finally, we performed time course assay to determine the time r equir ed f or PNA to in vade different f orms of dsDNA.We invaded circular and linearized plasmid DNA at 37 • C for increasing periods of time, followed by Argonaute cleavage for 1.5 h.Our results indicated that circular and linear dsDN A can be efficientl y invaded in as little as 10 min or 2 h, respecti v ely (Supplementary Figur e S11).Inter estingly, in the case of circular plasmid, we observed that extended invasion time only resulted in a modest increase in PNP editor-mediated cleavage efficiency, regardless of whether CbAgo or KmAgo was employed.Con versely, f or linear ds-DNA, we noted that cleavage efficiency reached its peak after 8 h of PNA incubation but substantially declined after 16 h.

DISCUSSION
The genome editing field has benefited from technologies harnessing natural molecular mechanisms combined with innovations in bioengineering.Zinc finger nucleases, transcription activator-like effector nucleases (TALEN) and cluster ed r egularly interspaced short palindromic r epeats (CRISPR) have been used for genome-editing applications in di v erse eukaryotic species ( 35 ).These technologies hav e their advantages and limitations, the latter especially in the translational potential of these genome engineering platforms due to challenges in cargo deli v ery , specificity , toxicity and imm uno genicity.Here, we introduced PNP editors, a novel pla tform tha t combines the targeted DNA strand invasion ability of PNA molecules with the DNAguided programmability of pAgos to enable programmable generation of site-specific DSBs.The inability of pAgos to mediate DSBs on linear DNA of arbitrary GC content and physiological temperatures is attributed to their lack of intrinsic helicase activity.The application of various pAgos has so far been limited to manipulation of ss-DNA, circular (supercoiled) dsDNA with low GC content, cleavage of circular dsDNA at elevated nonphysiological temperatures or enhancement of in vivo recombination in bacterial cells ( 9 , 14 ).We aimed to unlock the potential of pAgos for in vivo applications, including genome editing.pAgo activity at physiological temperatures is a pr er equisite for their use in genome editing and other in vivo applications.Se v eral studies have identified different pAgos from mesophilic bacteria with activity at a broad range of temperatures from 20 to 60 • C. For example, CbAgo, KmAgo and LrAgo proteins have been identified and characterized ( 12-15 , 33 ).Although mesophilic pAgos such as KmAgo, CbAgo and LrAgo are known to be programmable with DNA guides to cut dsDNA sequences, such directed activity was very low against GC-rich sequences at 37 • C, with efficient cleavage only shown at 55 • C and at AT-rich regions.The activity of pAgos at physiological temperatures on linear dsDNA sequences of arbitrary GC content would unlock the potential of these proteins for in vitro and in vivo genome-editing applications.Although combining CbAgo with a RecBC helicase from E. coli was recently shown to mediate DSB generation in linear dsDNA, this strategy is yet to be explored for in vivo applications in eukaryotic cells ( 15 ).We addressed the absence of intrinsic helicase activity characteristic of pAgos by employing PNA molecules for targeted, site-specific invasion of any DNA substrate.
PN A molecules were initiall y introduced due to their ability to invade DNA and form anomalous PN A-DN A structures that trigger the cellular repair machinery to resolve this structure as a recombination v ent.Although this concept has been applied for over three decades, the efficacy of PNA-media ted genome modifica tion r emains low, r endering it impractical for clinical applications in gene therapy ( 24 , 30 , 36 ).Targeted PNA invasion of DNA ne v ertheless offers clear advantages, including simplicity, resistance to cellular nucleases and potential to unwind B-DNA helices at specific sites for recruitment of various DNA modifying enzymes.Furthermore, various synthetic DN A analo gs that can invade DNA substrates may be used for gene editing and be incorporated into the PNP editor scheme.Examples include locked nucleic acid and Zorro oligonucleotides that can bind to both strands of the DN A sim ultaneousl y, w hich could facilitate in vitro and in vivo genome editing ( 37 ).
In this work, we coupled site-specific PNA invasion into target DNA to generate a displaced ssDNA strand that can be used as a substrate for guided pAgo nucleases.We showed that both pAgos tested here were capable of targeting DNA substrates irrespecti v e of their GC content and DN A form at physiolo gical temperatur es, thus over coming a major limitation in harnessing the power of pAgos in genome editing.We demonstra ted tha t a DSB generated only when pAgos were directed to displaced ssDNA regions and not when they targeted invaded strands or when the pAgo target region was flanking invasion sites.
The length and type of PNA is an important consideration when designing PNP editors.Here, we showed that PNA length should be greater than the length of the guide for optimal pAgo activity.Guide length of 16 nt in combination with 20-nt-long PNA led to the most efficient cleavage.Moreover, both ␥ PN A and ␥ tcPN A molecules efficiently invaded target DNA and showed similar effecti v eness in facilitating pAgo cleavage.This observation indica ted tha t any PNA molecule tha t can ef ficiently unwind ds-DNA w ould lik ely facilitate robust pAgo cleavage.We also showed that spacer length between 1 and 10 nt between two PNA invasion sites was optimal to generate a DSB in vitro , thus informing the optimal PNP editor design for in vivo use.We established that protein orientation, as defined by guide orientation, did not influence KmAgo activity and that the protein showed better performance with shorter spacers.In contrast, CbAgo displayed overall lower activity and opposite spacer length pr efer ence with guides in the inward orientation and optimal activity with 10-nt spacers in the outward orientation.Another major limitation is the activity of pAgos at physiological temperatures.Importantly, both PNA molecules and pAgos performed well at, and below, physiological temperatures, indicating the potential for their use in vivo in various eukaryotic organisms such as zebrafish ( Danio rerio ), the nematode Caenorhabditis elegans , etc.
The canonical pAgo cleavage site was reported to shift under certain conditions, such as when using guides of shorter length ( 33 ).We confirmed that the position of nicking sites on PNA-invaded DNA was not altered due to the presence of a PNA and was located between nucleotides 10 and 11 from the 5 end of the guide.Further characterization of gDNA r equir ements r e v ealed tha t 1-nt misma tches in the gDNA were well tolera ted a t any position within the guide.Howe v er, introducing two or more mismatched nucleotides into the gDNA re v ealed that central and supplementary regions close to the 3 end were especially sensiti v e to mismatches, leading to e v entual complete abrogation of activity with four mismatched nucleotides.Our results are in agreement with previous reports, where Argonautes are directed to ssDNA regions ( 12-14 , 33 ).
The specificity of PNP editors depends on two e v ents: site-specific PNA invasion into DNA and DNA-guided cleavage of the displaced ssDNA strand by pAgo.Our in vitro PNA specificity assays demonstrated that PNA-target misma tches a t the 3 and 5 ends w ere w ell tolerated up to 5 and 7 nt, respecti v ely.This result indicated that PNA molecules can invade targets of partial complementarity with reasonable efficiency, as long as they are located at the PNA terminus.Howe v er, invasion efficiency was impaired if the mismatched region was located in the center of the PN A-DN A duplex and further decreased for nonconsecuti v e mismatches.DNA guides employed by pAgos ensure a secondary layer of specificity in our concept.The activity of PNP editors dropped substantially when we used gDNAs tha t were misma tched a t the precise positions corresponding to the PNA invasion site, demonstra ting tha t any activity at nonspecific sites in the genome would be extremely inef ficient.In addition, DSB genera tion in vivo r equir es PNA invasion and pAgo cleavage to happen within a narrow window, further reducing the chances of off-target activity.
This w ork pro vides an additional alternati v e to other genome-editing technologies, including CRISPR / Casmedia ted genome modifica tion, and may help overcome key challenges for CRISPR / Cas-based clinical applications.The di v ersity of the pAgo protein family and the practicality of PNA synthesis open up myriad applications for gene therapy and for targeted genome editing in di v erse species.Mor eover, ther e ar e several advantages of using PNP editors for genome editing.For instance, unlike Cas proteins with PAM, PNP editors do not r equir e any sequence motif as a pr er equisite for binding and cleavage.PNP editors employ the pAgo enzyme, which possesses one RNase-H-like fold in the PIWI domain and can generate only a single nick.Ther efor e, the generation of DSBs r equir es two PNA binding e v ents and the positioning of two guided pAgo complexes in close proximity, minimizing the chances for any off-target activity.PNP editors employ PN A oligonucleotide analo gs and pAgo proteins of smaller size compared to Cas9; therefore, packaging and deli v ery into target cells is expected to be more efficient compared to the CRISPR / Cas systems.At present, organellar DNA can be modified only using TALENs and base editors fused to TALEs due to inefficient transport of crRNA (CRISPR RNA) through organellar membranes.A PNP editor programmed with short DNA guides that carry a much smaller charge than crRNA and PNAs decorated with mitochondrial targeting signal could potentially be used to edit organellar genomes.
Although these advantages make PNP editors a powerful genome-editing technology, key issues warrant future research and focus.These include the specificity of PNA invasion, solubility and activity at high ionic strength and high salt concentrations present in the context of cellular environments.Deli v ery of PNA into cells and the localization of PNA to the cell nucleus pose a challenge, although significant progress has been made by utilizing miniPEG modification and nanoparticles ( 29 ).Furthermore, detailed analysis of imm uno genicity and cellular toxicity is expected to provide detailed design guidelines for the use of PNAs for gene editing in vivo in a clinical context ( 29 , 38-40 ).A couple of features of pAgo proteins may complicate their use in genome editing: (i) the nonspecific guide-independent cleavage and chopping activity r equir ed for guide acquisition in bacterial hosts; and (2) nonspecific loading of pAgo proteins with endogenous microRNA or degraded DNA fragments within the cells and short half-life of externally provided gDNA molecules.These issues may be overcome by using a guide-preloaded pAgo or by providing PT-modified gDNAs, followed by deli v ery of preformed pAgo-gDNA complexes by lipofection or nucleofection.The chromatin state may help protect the genome from chopping activity of guide-free pAgo proteins due to the presence of histones.
We hope that our work may inspire the de v elopment of next-genera tion PNA molecules tha t exhibit robust invasion and entry into the cell nucleus and maintain high specificity and high on-target invasion efficiency at low concentrations.Due to the versatility of PNP editors, we envision that current work will expand the use of other pAgo proteins capable of pro grammable DN A binding for genome-editing a pplications.Finall y, we envision the therapeutic value of this PNP editor technology, as the system is highly specific and the size of the complexes is much smaller than that of CRISPR cargoes, with potentially improved specificity and imm uno genicity and lower cellular toxicity.

Figur e 1 .
Figur e 1. PN A invasion and pAgo-mediated cleavage of dsDNA molecules.( A ) Structure comparison of DN A and PN A molecules.PN A modifications were added a t ␥ position.( B ) Schema tic illustra tion of ␥ PN A invasion of dsDN A. ( C ) Invasions of different ␥ PN A molecules ( ␥ PN A1, ␥ PN A2, ␥ PN A3 and ␥ PNA4) into a specific dsDNA tar get sho wn by a mobility shift assay.PCR amplicons (50 nM final) containing the target region specific for each ␥ PNA were invaded overnight with 2 M ␥ PNA and resolved on native 6% TBE polyacrylamide gel.( D ) Ov ervie w of the PNP concept.The ability of ␥ PN A to specificall y invade target dsDN A can be exploited to facilita te cleavage media ted by gDN A-loaded pAgo of any DN A by cleaving the displaced strand opposite from the ␥ PNA invasion site to generate DSBs.( E ) Schematic diagram of pAgo-mediated cleavage of the pMRS-␥ PNA1 + ␥ PNA3 plasmid inv aded b y ␥ PN A1 and ␥ PN A3. ( F ) Gel images showing the pAgo-mediated cleavage of circular dsDNA invaded by ␥ PNA1 and ␥ PNA3.Upper and lower gels r epr esent CbAgo and KmAgo cleavage, respecti v el y. ␥ PN A1 and ␥ PN A3 were used to invade high-GC-content target regions (75% and 55%, respecti v ely) cloned in the pMRS-␥ PNA1 + ␥ PNA3 plasmid in close proximity to a SacI site.The bands corresponding to anticipated cleavage products were observed only in the reaction containing the invaded plasmid and the two gDNAs following the BsrGI digest (lane 7).( G ) Gel images showing the pAgo-mediated cleavage of the target region in BsrGI-linearized dsDNA invaded by ␥ PNA1 and ␥ PNA3.Upper and lower gels r epr esent CbAgo and KmAgo cleavage, respecti v el y. ␥ PN A1 and ␥ PN A3 wer e used to invade high-GC-content target r egions (75% and 55%, r especti v el y) in pMRS-␥ PN A1 + ␥ PNA3 plasmid linearized with BsrGI, in close proximity to a SacI site.The bands corresponding to expected cleavage products were observed in the reaction containing invaded linear plasmid and the two pAgo-gDNA complexes (lane 7).Lane M represents the 1-kb plus DNA ladder.

Figure 2 .
Figure 2. Effect of GC content on cleavage of circular and linear dsDNA.( A ) Effect of GC content on cleavage of a circular plasmid.Upper and lower gels r epr esent CbAgo-and KmAgo-mediated cleavage, respecti v el y.Non-invaded pMRS-␥ PN A1 + ␥ PN A3 circular plasmid was incubated with pairs of pAgo-gDNA complex es targeting r egions with GC contents of 13%, 22%, 27%, 31%, 35%, 50% or 75% (lanes 1-7) at 37 • C for 1 h.The same plasmid was then inv aded b y ␥ PNA1 and ␥ PNA3 at regions with GC contents of 55% and 75%, respecti v ely, and incubated with their respecti v e pAgo-gDNA complexes at 37 • C for 1 h (lane 8).Following pAgo cleavage reaction, plasmids were digested with SacI or BsrGI depending on the position of the target r egion r elati v e to the restriction enzyme site.Non-inv aded pAgo cleav age (lane 9) and SacI + BsrGI-digested (lane 10) samples are included as control reactions.( B ) Effect of GC content on cleavage of a linearized plasmid.Upper and lower gels r epr esent CbAgo-and KmAgo-mediated cleavage, respecti v ely.Linearized, non-invaded pMRS plasmid was incubated with pairs of pAgo-gDNA complexes targeting regions with GC content of 13%, 22%, 27%, 31%, 35%, 50% and 75% (lanes 1-7) at 37 • C for 1 h.Linearized plasmid was then invaded by ␥ PNA1 and ␥ PNA3 at regions with 55% and 75% GC content, respecti v ely, and incubated with the respecti v e pAgo targeting complexes at 37 • C for 1 h (lane 8).Non-invaded pAgo-mediated cleavage (lane 9) and SacI-digested (lane 10) samples were used as control reactions.Expected cleavage product sizes are listed on top of each gel lane.Lane M r epr esents the 1-kb plus DNA ladder.

Figure 3 .
Figure 3. Multiplexed cleavage of circular plasmid containing two target sites with two pairs of ␥ PNAs and pAgo-guide complexes.( A ) Schematic diagram of the two target regions in the circular and AseI-linearized pMRS multiplexing plasmid.Target region 1 contains the ␥ PNA1 + ␥ PNA3 binding sequences; target region 2 contains the ␥ PNA5 + ␥ PNA6 binding sequences.The two target regions are separated by 820 bp (EGFP sequence).( B ) Representati v e gel images showing the multiplex cleavage of circular plasmid with two pairs of pAgo-gDNA complexes.Upper and lower gels r epr esent CbAgo-and KmAgo-mediated cleavage, respecti v el y.The plasmid was initiall y invaded with two pairs of ␥ PN A molecules ( ␥ PN A1 + ␥ PN A3 and ␥ PN A5 + ␥ PN A6) and incubated with pAgo loaded with gDNA for 1 h at 37 • C. Non-invaded and invaded plasmids were incubated with the two pairs of pAgo-gDNA complexes (lanes 1 and 5, respecti v el y), one gDN A pair (lanes 2-3 and 6-7) or no gDNA (lanes 4 and 8).SacI + AgeI-digested (non-invaded) (lane 9) and SacI-linearized (non-invaded) (lane 10) samples were used as controls.Lane M represents the 1-kb plus DNA ladder.( C ) Multiplex cleavage of linear plasmid with two pairs of pAgo-gDNA complexes.Upper and lower gels r epr esent CbAgo-and KmAgo-mediated cleavage, respecti v ely.The AseIlinearized plasmid was invaded with two pairs of ␥ PNA molecules ( ␥ PNA1 + ␥ PNA3 and ␥ PNA5 + ␥ PNA6) overnight and incubated with a pair of pAgos (1.5 M final) loaded with gDNA for 2.5 h at 37 • C. Non-invaded and invaded plasmids were incubated with the two pairs of pAgo-gDNA complexes (lanes 1 and 5, respecti v el y), one gDN A pair (lanes 2-3 and 6-7) or no gDNA (lanes 4 and 8).AseI-, SacI-and AgeI-digested (non-invaded) (lanes 9-12) samples were used as controls to visualize positions of resulting bands.Lane M represents the 1-kb plus DNA ladder.

Figure 4 .Figure 5 .
Figure 4. Testing the effect of spacer length between ␥ PNA1 and ␥ PNA3 target regions in pMRS plasmid on cleavage ef ficiency.( A ) Schema tic diagram showing the ␥ PNA1 and ␥ PNA3 binding regions with guides facing outward.( B ) Schematic diagram showing the ␥ PNA1 and ␥ PNA3 binding regions with guides facing inward.RE1 and RE2 indicate the positions of the SacI restriction sites in different target plasmids.( C ) Table summarizing the different spacer sequences, lengths (1, 3, 5, 6, 10, 15, 20 and 30 bp) and the position of SacI on the target plasmid.( D ) Representati v e gel images showing pAgomediated (upper , CbAgo; lower , KmAgo) cleavage of DNA targets invaded by ␥ PNA1 and ␥ PNA3 (guides facing outward) containing varying spacer lengths between the two PNA invasion sites.All plasmids were linearized using BsrGI digestion and invaded overnight with ␥ PNA1 and ␥ PNA3 before being cleaved with a pair of pAgo-gDNA complexes (lanes 1-8).SacI + BsrGI-restricted (non-invaded) (lane 9) sample was used as size control.( E ) Representati v e gel images of pAgo-mediated (upper , CbAgo; lower , KmAgo) cleavage of DNA targets invaded by ␥ PNA1 and ␥ PNA3 (guides facing inward) containing different spacer lengths between the two PNA invasion sites.All plasmids were linearized using BsrGI digestion and invaded overnight with ␥ PNA1 and ␥ PNA3 before being cleaved with a pair of pAgo-gDNA complexes (lanes 1-8).A SacI + BsrGI-digested (non-invaded) (lane 9) sample was used as size control.Lane M r epr esents the 1-kb plus DNA ladder.Quantification values are shown as mean ± standard deviation (SD) ( n = 3).

Figure 6 .
Figure 6.Testing the effect of temperature on PNA inv asion, pAgo cleav and identification of pAgo cleavage sites.( A ) Representati v e gel images showing pAgo activity at differ ent temperatur es .To test the ability of pAgos to cleave the invaded target at a range of temperatures, linearized plasmid containing ␥ PNA5 + ␥ PNA6 binding regions was invaded with ␥ PNA5 and ␥ PNA6 at 37 • C overnight.The target was then incubated with a pair of pAgo complexes (upper gel, KmAgo) targeting the displaced ssDNA region at 20, 25, 30, 27, 40 or 45 • C for 1 h (lanes 1-6).( B ) Representati v e gel images showing invasion by ␥ PNA5 and ␥ PNA6 a t dif ferent tempera tures.Linearized plasmid containing ␥ PNA5 and ␥ PNA6 binding regions was invaded at 20, 25, 30, 27, 40 or 45 • C overnight, followed b y cleav age b y pAgo at 37 • C for 1 h (lanes 1-6).( C ) Representati v e gel images showing invasion by ␥ PNA5 and ␥ PNA6 and pAgo activity a t dif ferent tempera tures.Both the invasion step and pAgo cleavage were performed under the same varying temperature (20, 25, 30, 27, 40 or 45 • C) conditions.The invasion step was performed overnight, followed by pAgo (upper gel, CbAgo; lower gel, KmAgo) cleavage assa y f or 1 h (lanes 1-6).In all gels (A, B and C), samples in vaded with ␥ PN A5 and ␥ PN A6, non-invaded templates cleaved with pAgo (upper gel, CbAgo; lower gel, KmAgo) at 37 • C (lanes 7 and 8, respecti v ely) and AgeI-digested DNA (non-invaded) (lane 9) were included as control reactions.Lane M r epr esents the 1-kb plus DNA ladder.Quantification values are shown as mean ± SD ( n = 3).( D ) Schematic diagram showing inv asion b y ␥ PNA5 and ␥ PNA6, pAgo-guide complex binding sites and primer orientation for Sanger sequencing.( E ) Determination of pAgo cleavage site on a dsDNA template linearized by BsrGI digestion and invaded by ␥ PNA5 and ␥ PNA6.Confirmation of pAgo (upper gel, CbAgo; lower gel, KmAgo) cleavage site was performed on invaded pMRS-␥ PNA5 + ␥ PNA6 that had been linearized by BsrGI digestion.Following the cleavage assay, the bands corresponding to cleavage products were purified from the agarose gel and subjected to Sanger sequencing.Asterisks denote sequencing artifacts arising from AmpliTaq adding a non-templated 3 A upon reaching the end of a linear template.

Figure 7 .
Figure 7. Effect of mismatches at different positions on pAgo cleavage efficiency of BsrGI-linearized plasmid DNA.( A ) Table summarizing the differ ent r egions in the gDNA molecule.( B ) Table summarizing the mismatched guides 1 and 2, highlighting the region, type and number of mismatched nucleotides.( C ) Representati v e gel images showing the pAgo cleavage of pMRS-␥ PNA1 + ␥ PNA3 plasmid linearized by BsrGI digestion and invaded by ␥ PNA1 and ␥ PNA3 using different mismatched guides.Invaded plasmid was incubated with CbAgo (upper gel) or KmAgo (lower gel) preloaded with different gDNAs containing 1-or 2-nt mismatches at different positions in the pAgo guide architecture for 1 h at 37 • C (lanes 1-8).Reactions of samples invaded by ␥ PNA1 and ␥ PNA3 with specific guides and nonspecific guides (lanes 9 and 10, respecti v ely) were included.Non-invaded samples but incubated with specific guides (lane 11) and SacI-digested samples (non-invaded) (lane 12) were included as control reactions.Lane M represents the 1-kb plus DNA ladder.Quantification values are shown as mean ± SD ( n = 3).