Systematic analysis of factors that improve HDR efficiency in CRISPR/Cas9 1 technique 2

The bacterial CRISPR/Cas9 system has a proven to be an efficient tool for genetic 16 manipulation in various organisms, but the efficiency of sequence replacement by 17 homologous direct repair (HDR) is substantially lower than random creation of indels. 18 Many studies focused on improving the efficiency of HDR using double sgRNA, cell 19 synchronization cycle and the delivery of ssODN with a rational design. 20 In the present study, we tested and compared the combination of these three methods 21 to improve HDR efficiency. To our tests, we chosen the TNFα gene (NM_000594) for 22 its crucial role in a variety of biological processes and diseases. 23 Our results showed a dramatically increases of HDR efficiency from undetectable HDR 24 event to 39% of HDR efficiency and provide a new strategy to facilitate CRISPR/Cas9- 25 mediated human genome targeting. Furthermore, we showed that TNFα gene could be edited with CRISPR/Cas9 27 methodology, an opportunity to safely correct, in the future, the specific mutations of 28 each patient.


Introduction 31
In the last decade, the use of the novel CRISPR-associated endonuclease Cas9 protein 32 has been implemented for analytical and therapeutic approaches, in a broad spectrum 33 of cell types and model organisms (1)(2). After the introduction of this technique, the 34 creation of a knock-in and knock-out gene has become as simple, rapid, and economical 35 5 one cycle of 95°C for 5 min, 35 cycles of 98°C for 20 s, 61°C for 15 s and 72°C for 30 135 s, and one cycle of 72°C for 1 min. The PCR products were analyzed on 1,3% agarose 136 gel containing Midori Green Xtra (NIPPON Genetics Europe, Dueren, Germany). The 137 concentration of PCR DNA was quantitated based on the band intensity relative to a 138 DNA standard using the software Image Lab (Bio-Rad, Hercules, CA). About 200 ng 139 of PCR DNA was used for T7 endonuclease I and SmaI analyses. 140 141

T7-Endonuclease I Assay 142
The widely used T7-endonucelase I assay targets and digests hetero-duplexes formed 143 by hybridization of mutant WT strands resulting in two smaller fragments, and this 144 method was performed to assess sgRNA-specific activity. After transfection, cells were 145 incubated for 48 hr. The cells were then pelleted, and the lysis performed using QIAamp 146 DNA Mini Kit (Qiagen). The PCR products were denatured and then reannealed using 147 the following program: 95° C for 5 min, ramp down to 85°C at 2°C/s, and ramp down 148 to 25°C at 0.1 C/s. Immediately after the reannealing step, and the consequent 149 heteroduplex formation, 5 units of T7 endonuclease I (New England Biolabs, Ipswich, 150 MA) was added to the mix and incubated for 1 hr at 37°C. The product was resolved 151 on 1,3% agarose gel containing Midori Green Xtra (

Analysis of HDR by SmaI restriction digestion 171
The reaction consisted of 1.6 µg of PCR DNA and 20 units of SmaI enzyme in CutSmart 172 Buffer (NEB, Ipswich, MA). After 1hr of incubation at 37°C, the reaction was arrested 173 with heat inactivation at 65°C for 20 min. The product was resolved on 1.3% agarose. 174 The band intensity was quantitated using Image Lab. The percentage of HDR was 175 calculated using the following equation (b + c / a + b + c) × 100 for the single cut 176 strategy, (b + c / a + b + c + d) x 100 for the dual cut strategy. In the first equation, 'a' 177 is the band intensity of DNA substrate and 'b' and 'c' are the cleavage products. In the 178 second equation, 'a' is the band intensity of DNA substrate wild type, 'b' and 'c' are 179 the cleavage products, and 'd' is the deleted fragment in which both sgRNA worked 180 properly but there wasn't a KI event. To further confirm the presence of the edited 181 sequence, conventional Sanger sequencing was performed (Fig. S3A). 182 183

Off-target analysis 184
To predict the most likely off-target sites for the sgRNAs used to knock-down the TNFα 185 gene in this study, we used a public webserver: 186

Selection of specific RNA guides for DSB induction 214
For this purpose, we used the previously described Staphylococcus pyogenes nuclease, 215 that utilizes a human-codon optimized SpCas9 and a chimeric sgRNA expression 216 vector to direct efficient site-specific gene editing (21)(22). We designed two 20-nt long 217 sgRNA with a wide-spam of 473 nucleotide to guide Cas9 to introns 1 and 3 of the 218 TNFα gene (sgRNA1 and sgRNA2) (Fig. 1). Without the presence of donor DNA 219 containing the homology arms, cells repair the DNA primarily by NHEJ, leaving 220 insertion and/or deletion (indels). Considering the small introns size (»300-600bp) in 221 TNFα locus, indels can involve adjacent exons and may generate frameshift mutations 222 that knocking-out the TNFα gene. 223 In detail, in pX333 plasmid the two sgRNAs were cloned as single guide (pX333-224 sgRNA1; pX333-sgRNA2), in tandem combination (pX333-2sgRNA1/1; pX333-225 2sgRNA2/2) and the two different guides were cloned in the same vector (pX333-226 2sgRNA1/2). To asses if the plasmid with puromycin can helps for positive clone 227 selection, we cloned the two sgRNAs also in pX459 vector (pX459-sgRNA1 and 228 pX459-sgRNA2), a plasmid that carry the puromycin resistance. The seven constructs 229 thus obtained were further transfected into HEK293 cells and the puromycin antibiotic 230 was added for 48h only in cells transfected with pX459 vector. 231 To evaluate the efficiency of DNA cutting and NHJE repair after the use of sgRNA 232 guides, the genomic DNA was isolated from HEK293 cells and screened for the 233 presence of site-specific gene modification by PCR amplification and T7E1 234 endonuclease assay, of region around the target sites ( Fig.2A). 235 8 The results showed detectable bands in pX333-sgRNA2, pX333-sgRNA2/2, pX459-236 sgRNA2. The use of sgRNAs cloned in single or in tandem when co-expressed with 237 the SpCas9 nuclease was able to mediate gene modification with comparable level of 238 efficiency. No differences were finally detected between pX333 or pX459 plasmids 239 after puromycin selection, maybe due to high efficiency of transfection obtained in 240 HEK293 cells. Notably, HEK293 cells transfected with the 2sgRNA plasmid 241 (2sgRNA1/2) and not treated with T7E1 nuclease resulted in a full-length (FL) and in 242 short-edited (SE) amplicons confirming the expected deletion of the region between the 243 two selected protospacers ( Fig. 2A). 244 Moreover, the frequency of Indels in the cells transfected with all sgRNAs, were 245 measured by sequencing 10 PCR amplicons encompassing the target sites. As reported 246 in the figure 2B, the highest editing frequencies achieved were 50% and 75% using 247 sgRNA2 and 2sgRNA1/2 guides, respectively (Fig. 2B). The types of insertions and 248 deletions at this locus presented variable patterns of rearrangements of the coding 249 sequence, insertion from 1 to 10 nucleotides and deletion from 5 to 930 nucleotides. 250 Deletion of region between the 2 PAMs was observed in the cells transfected with 251 2sgRNA plasmid (2sgRNA1/2) (~480nts). 252 Furthermore, we observed the predominance of a precise junction between the two 253 DSBs when 2sgRNA1/2 was transfected, a mechanism already described (1)(12)(23). 254 These data further confirmed that the dual sgRNA (2sgRNA) is the most efficient 255 method for DNA excision in the endogenous locus. Based on these genomic results, we 256 selected the sgRNA2 and the 2sgRNA1/2 guides for the following HDR editing. 257 258

Homologous direct recombination efficiency (HDR) 259
With the aim to improve the HDR efficiency we chose the plasmids pX333-sgRNA2 260 and the pX333-2sgRNA1/2 that showed a highest degree of DSB in HEK293 261 transfection. Traditional HDR gene editing requires long homology arms to allows 262 proper and high-specificity recombination. The use of Cas9-gRNAs directed 263 recombination allows the use of much smaller homology arms (~90 bp to 700 bp) with 264 higher recombination rates than conventional HDR (1). 265 We decided to assess the efficiency of HDR by transfection of single (pX333-266 sgRNA2) and double sgRNA (pX333-2sgRNA1/2) coupled with a rational design of 267 ssODNs. The structure of ssODNs with asymmetric arms complementary to a non-268 target locus with long arm on the PAM-proximal side and short arm on the PAM-distal 269 side of the break, has been previously reported to induce highest HDR efficiency (14). 270 However, we decided to test this asymmetric donor for the TNFα locus by comparing 271 it with other possible ssODN structures on HDR efficiency. 272 Thus, we generated twelve ssODN molecules having different sequences overlap 273 on the 5' and 3' side of the break, specific for pX333-sgRNA2 and pX333-2sgRNA1/2 274 guides, and complementary to either target or non-target DNA strand (Fig.3A,5A). We 275 co-delivered these ssODNs in combination with single sgRNA and for the first time 276 with a couple sgRNAs (2sgRNA1/2). 277 To facilitate the selection of HDR events, we inserted a restriction site for the enzyme 278 SmaI in all the ssODNs. The pX333-sgRNA2 and pX333-2sgRNA1/2 plasmids and the 279 respectively six ssODNs were then co-transfected in HEK293 cells. In addition, we 280 introduced a nocodazole cells synchronization, described to improve the HDR (5). 281 Then, we included a fourth condition in which the cells were transfected three 282 consecutive times with the same donor and plasmid. 283 Since the edited sequences contain a newly acquired SmaI restriction site, the HDR 284 efficiency, can be easily detected using the digestion on PCR products obtained with 285 primers flanking the targeted locus. 286 287

Single cut and HDR efficiency 288
First, we determined the HDR efficiency with the co-delivery of pX333-sgRNA2 guide 289 (who induces single DSB) and six ssODN molecules (A-F). We compared the HDR 290 efficiency after SmaI digestion in single transfection events, triple transfection and with 291 nocodazole cells treatment. Notably, among the six ssODNs, only the donors A, C and 292 E showed HDR event (Fig.3B,C,D,E). The tree donors are all complementary to the 293 non-target DNA strand, and this observation is consistent with previous studies (14). 294 The donors A and C are able to induce HDR but only after nocodazole treatment though 295 with a low frequency of 5,4% and 3,9%, respectively. With triple transfections in 296 unsynchronized cells, donor A increase the HDR efficiency from not detectable to 297 10,5%, and donor C form non detectable to 9,3%. Donor E, after triple transfection, 298 increases the HDR efficiency of 1,4-fold. The highest HDR frequency achieved was 299

22.1% for donor E, with triple transfection and nocodazole cells treatment. 300
The graphic in Fig. S1A,B (supplementary figure) compares the HDR efficiency in 301 synchronized and unsynchronized cells in one and triple cell transfections, respectively. 302 The graphic in Fig. S1C,D instead, compares the HDR efficiency between single and 303 triple transfection in synchronized and unsynchronized cells. These data showed that 304 no significant differences have been detectable with nocodazole treatment in our 305 experimental condition, while the triple transfection increases the HDR efficiency, but 306 only for the ssODNs A, C and E. While the other three donors (B, D, F) characterized 307 to be complementary to target DNA strand didn't induce HDR event. 308 Notably, the donor E induces the highest HDR efficiency. It showed the same 309 structure described by Richardson et al. to be the best for the HDR, with asymmetric 310 arms complementary to a non-target locus with 90bp on the PAM-proximal side and 311 36bp extension arm on the PAM-distal side of the break (14). The HDR efficiency using 312 donor E increased up to 22%, (more than two-fold) with triple transfection in 313 synchronized cells (Fig.4). 314 315

Dual cut and DBS efficiency 316
Furthermore, we tested the HDR efficiency combining the use of dual 2sgRNA and 317 rational design of ssODNs. HEK293 cells have been co-transfected with px333-318 2gRNA1/2 guide (who induces double DSB) and six ssODNs (G-N) (Fig.5A). Next, 319 we determined systematically the effect on HDR efficiency in controls, in nocodazole 320 synchronized cells and triple transfection. 321 To note, as already here described, a simple PCR amplification before the T7E1 assay, 322 (that shows the DNA deletions caused from 2sgRNA1/2) is sufficient to assess the DSB 323 efficiency in various experimental conditions. Interestingly, the DSB efficiency 324 increases more than three-fold with triple transfection in synchronized cells (Fig.6A) 325 326

Dual cut and HDR efficiency 327
We further compared the HDR efficiency using px333sgRNA1/2 and the six ssODNs 328 (G-N) with SmaI digestion (Fig.5A). With the use of 2sgRNA1/2 all ssODNs designed 329 showed detectable SmaI digestion bands who indicates that HDR occurred in all 330 experimental condition used. The graphic in Fig.S2 (supplementary figure) showed the 331 effect of nocodazole and triple transfection on HDR efficiency using double sgRNA 332 guides. Again, the nocodazole doesn't have a significant effect on HDR; while the triple to 39% (Fig.5B,C,D,E). The graphic in figure 6B compares the HDR efficiency of 339 donor M in all experimental conditions used. The HDR after triple transfection in sync 340 cells, increases up to 1,8-fold. As already reported in literature, even in TNFα locus, the 341 use of 2sgRNA dramatically increase the HDR compared to single sgRNA more than 342 Collectively, the results demonstrated that the combination of 2sgRNA, asymmetric 344 donor and triple transfection, induce a dramatic increase of HDR, from undetectable to 345 39% HDR efficiency. 346 347

Off-targets analyses 348
To predict the most likely off-target sites for the sgRNAs used to edit the TNFα gene 349 in this study, we used a public webserver: 350 (https://eu.idtdna.com/site/order/designtool/index/CRISPR_PREDESIGN) able to 351 assess and prioritize potential CRISPR/Cas9 activity at off-target loci. The top three 352 potential off-target sites for each sgRNA were assessed by the T7E1 assay. None of the 353 loci analyzed showed detectable levels of off-target events (Fig.S3B). 354

355
Discussion 356 Here, we report a new and simple approach to enhancer genome engineering in human 357 cells. We compared the use of single sgRNA with coupled 2sgRNA to edit the TNFα 358 locus. According to data already reported, our results showed that the use of 2sgRNA 359 increases dramatically the DSB efficiency. Furthermore, the use of 2sgRNA creates a 360 precise DNA excision between each PAM sequence, a huge advantage over the 361 randomly sized indels created by single sgRNA transfection (Fig.2). In addition, the 362 deletions between the 2sgRNA can be easily identified by PCR amplification and 363 agarose electrophoresis, thus avoiding the T7E1 assay. Further, the results showed that 364 the DSB increases after triple transfection up to two-fold and, in combination with 365 nocodazole treatment increases up to three-fold (Fig.6). 366 As already known, even in TNFα locus, the use of 2sgRNA increases the HDR 367 efficiency more than double compared to transfection of single sgRNA. This could be 368 explained by the capability of double 2sgRNA to create a precise cut on the target site, 369 oscillating from time to time only for a few bases, without generating unpredictable 370 indels. Otherwise, a single-cut approach, induces truly extensive deletions reducing a 371

Fig.S1
A B C D