Highly multiplexed genome engineering using CRISPR/Cas9 gRNA arrays

Design and construction of gRNAs

A detailed protocol for the design and construction of guide RNA (gRNA) arrays is found in supplementary protocol 1, alongside ApE plasmid maps in supplementary files 1. gRNAs were designed to the desired region of each target gene using the CRISPR Design Program (http://crispr.mit.edu). Multiple gRNAs were chosen based on the highest ranked values determined by avoiding predicted off-targets. The gRNAs were ordered in oligonucleotide pairs, annealed, and ligated into pENTR221-U6 stuffer [25]. Briefly, oligonucleotide pairs were annealed and phosphorylated using T4 PNK (New England BioLabs) and 10X T4 Ligation Buffer (New England BioLabs) in a thermocycler with the following protocol: 37°C 30 minutes, 95°C 5 minutes and then ramped down to 25°C at 5°C/minute. pENTR221-U6 stuffer vector backbone [25] was digested with FastDigest BbsI (Fermentas), FastAP (Fermentas) and 10X Fast Digest Buffer and used for the ligation reaction. The digested pENTR221-U6 stuffer vector was ligated together with the phosphorylated and annealed oligo duplex (1:200 dilution) from the previous step using T4 DNA Ligase and Buffer (New England BioLabs). The ligation was incubated at room temperature for at least 1 hour and then transformed and mini-prepped using the GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific). Plasmids were analyzed by Sanger sequencing to confirm proper oligonucleotide ligation.

Validation of gRNAs

HEK293T cells were maintained in DMEM medium supplemented with 10% fetal bovine serum (FBS). 1x10⁵ cells of HEK293T cells were seeded in 24-well plate the day before transfection. Transfection was performed using Lipofectamine 2000 (Thermo Fisher Scientific), following the manufactures protocol. 500 ng of pT3.5-CAG-hCas9, 250 ng of pENTR221-U6-gRNA plasmid and 100ng of pGFP-MAX plasmid (Lonza), to assess transfection efficiency, were diluted in 75 μl of OptiMEM and 5 μl of Lipofectamine 2000 was diluted in 75 μl of OptiMEM and then the mixtures were combined. The complete mixture was incubated for 15 min before being added to cells in a drop wise fashion. After 16 hours, the media was changed to fresh DMEM medium containing 10% fetal bovine serum. Cells were incubated for 3 days at 37°C and then genomic DNA was collected using the GeneJET Genomic DNA Purification Kit (Thermo Fisher Scientific). Activity of the gRNAs was quantified by Surveyor nuclease assay, gel electrophoresis, and densitometry as described [25,26].

Vectors construction and LR clonase reactions

Plasmids containing the ccdB gene were maintained in One Shot® ccdB Survival™ 2 T1R bacteria (Invitrogen). All enzymes used in this study were purchased from New England BioLabs. All pENTR1 vectors were derived originally from pENTR221-GUS (Invitrogen). gBlocks (Integrated DNA Technologies) encoding the pGG and pACPT cassettes were ordered and constructed to contain attB1/2 sites. Single BP Clonase reactions were performed using manufacturer’s protocols with Clonase II enzyme mixes (Invitrogen). The pT3.5-CAG-Csy4-T2A-hCas9 and pT3.5-CAG-MS2-p65-HSF1-T2A-eGFP vectors were generated using a combination of restriction enzyme cloning and Gateway cloning. Briefly, the nucleotides encoding Pseudomonas aeruginosa Csy4 [5,27] flanked with attL1 and T2A sites (attL1-Csy4-T2A) with SbfΙ and BspHΙ restriction sites was ordered as a gBlock (Integrated DNA Technologies). The DNA fragment of Csy4-T2A was cloned into pENTR221-hCas9 digested with NcoI. pENTR221-Csy4-T2A-Cas9 was then cloned into pT3.5-CAG-DEST via Gateway LR Clonase reaction, following manufacturer’s instructions (Thermo Fisher Scientific). A gBlock encoding MS2-p65-HSF1-T2A-eGFP flanked by attB1/2 was used for BP Clonase reaction with pDONR221 to generate pENTR221-MS2-p65-HSF1-T2A-eGFP, which was subsequently used for LR Clonase reaction with pT3.5-CAG-DEST to generate pT3.5-CAG-MS2-p65-HSF1-T2A-eGFP.

Golden gate assembly of gRNA arrays

Single pGG vectors were generated using the same oligonucleotide ligation approach described above. Primer sequences can be found in S1 Table. Arrays were then generated using golden gate assembly. Briefly, 150 ng of each pGG-gRNA plasmid was combined with 150 ng of the appropriate pACPT vector, followed by addition of 1 uL of BsaI (New England BioLabs), 1 uL of T4 DNA Ligase and Buffer (New England BioLabs) and water to a total of 20 uL. Golden gate assembly was then carried out using the following thermocycling protocol: (37°C 5 minutes, 16°C 10 minutes) x10, 50°C 5 minutes, 80°C for 5 minutes and then cooled to 4°C. One uL of 25 mM ATP and 1 uL of Plasmid Safe (Epicentre) were then added to the reaction and incubated for 1 hour at 37°C. Assembled plasmids were then transformed into TOP10 bacteria (Thermo Fisher Scientific) and plated on spectinomycin selection plates with X-gal and mini-preps performed on white colonies using the GeneJET Plasmid Miniprep Kit (Life Technologies). The plasmids were then analyzed by Sanger sequencing to confirm proper gRNA array assembly.

Validation of gRNA arrays

HEK293T cells were maintained in DMEM medium supplemented with 10% fetal bovine serum (FBS). 1x10⁵ cells of HEK293T cells were seeded in 24-well plate the day before transfection. Transfection was performed using Lipofectamine 2000 (Thermo Fisher Scientific), following the manufactures protocol. 500 ng of pT3.5-CAG-Csy4-T2A-hCas9, 250 ng of pENTR221-U6-gRNA or pACPT array plasmid were diluted in 75 μl of OptiMEM and 5 μl of Lipofectamine 2000 was diluted in 75 μl of OptiMEM and then the mixtures were combined. The complete mixture was incubated for 15 min before being added to cells in a drop wise fashion. After 16 hours, the media was changed to fresh DMEM medium containing 10% fetal bovine serum. Cells were incubated for 3 days at 37°C and then genomic DNA was collected using the GeneJET Genomic DNA Purification Kit (Thermo Fisher Scientific). Activity of the gRNAs was quantified by a Surveyor nuclease digest, gel electrophoresis, and densitometry as described [16]. For gene activation experiments, 250 ng of pT3.5-CAG-Csy4-T2A-dCas9-VP64, 250 ng of pT3.5-CAG-MS2-p65-HSF1-2A-eGFP and 250 ng of pENTR221-U6-gRNA or pACPT array plasmid were transfected as above. Cells were incubated for 3 days at 37°C and then RNA was extracted using PureLink® RNA Mini Kit (Thermo Fisher Scientific) and then reverse-transcribed by Transcriptor First Strand cDNA Synthesis Kit (Roche).

Surveyor nuclease assay

Surveyor assays were performed as previously descried [18]. Briefly, after electroporation of CRISPR/Cas9 plasmids and incubation for 3 days genomic DNA was extracted using GeneJET Genomic DNA Purification Kit (Thermo Fisher Scientific), following manufacturer’s instructions. PCR amplicons were generated spanning the Cas9 binding site using Accuprime™ Taq HF (Invitrogen) using the following PCR cycle: initial denaturation at 95°C for 5 min; 40x (95°C for 30 sec, 55°C or 60°C for 30 sec, 68°C for 40 sec); final extension at 68°C for 2 min. PCR amplicons were denatures and annealed as follows: 95°C for 5 minutes, 95–85°C at -2°C/s, 85–25°C at -0.1°C/s, 4°C hold. Primer sequences can be found in S1 Table. Three microliters of the annealed amplicon was then diluted with 6 μL of 1x Accuprime PCR buffer and treated with 1 μL of Surveyor nuclease with 1 μL of enhancer (Thermo Scientific) at 42°C for 20 min. The reaction was then stopped by the addition of 3 μL of 15% Ficol-400 and 0.05% Orange G solution containing 1 mM EDTA and subsequently run on a standard 10% TBE gel. Percent gene modification was calculated using Image J software as described [25,26].

Q-RT-PCR analysis

RNA was extracted from fresh cell pellets using an RNeasy Mini Kit (Qiagen, Hilden, Germany) and reverse-transcribed by TaqMan® Reverse Transcription Reagents (Applied Biosystems, Foster, CA, USA). Real-time quantitative PCR (q-PCR) were performed using following specific TaqMan probes: ASCL1; Hs00269932_m1, MYOD1; Hs02330075_g1, HBG1/HBG2; Hs00361131_g1, IL1B; Hs01555410_m1, IL1R2; Hs01030384_m1 and ACTB Hs99999903_m1 as an internal control (Thermo Fisher Scientific). TaqMan® Gene Expression Master Mix (Thermo Fisher Scientific) was used with the CFX96™ Real-Time System). Data were normalized using beta-actin in each sample, by the 2-ΔΔCT method as previously described [28]. All experiments were performed in technical triplicate.

Golden gate assembly of gRNA arrays with up to 10 gRNAs

To simplify the generation of gRNA arrays we used the golden gate cloning system [24]. We used type IIS restriction enzyme sites (BsmBI) and over hangs previously published and validated for robust golden gate assembly of TALEN DNA binding domains [29]. We then designed and assembled cassettes for oligonucleotide ligation of protospacer sequences using a different type IIS restriction enzyme (BsaI) flanking a stuffer sequence. In addition, we included a Csy4 ribonuclease target sequence directly upstream of the target gRNA sequences such that after golden gate assembly each gRNA is directly flanked by the Csy4 target sequence (Fig 1A). Next, we designed a gRNA array acceptor plasmid (pACPT) containing a LacZ gene, for blue/white colony selection after golden gate assembly, flanked by appropriate BsmBI sites and an upstream U6 pol II promoter to drive expression of assembled gRNA arrays (Fig 1A). We also included a terminal Csy4 target sequence, such that the last gRNA is free of additional sequence when processed, and a poly T termination sequence. In order to produce a highly modular system for rapid and efficient cloning of the U6 driven gRNA array cassettes, attL1/2 sequences were included in pACPT for Gateway cloning (Fig 1A). We have termed this set of plasmids for oligonucleotide ligation pGG1-10 and the acceptor plasmids pACPT1-10 (Fig 1B). An example of the plasmids required for golden gate assembly of a 4 gRNA array and the structure of the final expression plasmid are shown in Fig 1C.

In order to validate our golden gate assembly system we ligated oligonucleotides encoding protospacer sequences targeting 10 genes we previously validated for CRISPR/Cas9 mediated DSB induction with an average editing frequency of ~22% and a range of 10–35% (1:GOSR1, 2:PPP2R2A, 3:CNTFR, 4:DMD, 5:ZBTB10, 6:KAT7, 7:SPPL3, 8:CCM2, 9:PRDX1, 10:TRIP12; Panel A in S1 Fig). We then performed golden gate assembly of arrays containing 3, 5, 7, or 10 of these gRNAs and validated the resultant plasmids by Sanger sequencing (Panel A in S2 Fig). As with the previously described golden gate assembly system [24], we observed sufficient white bacterial colonies upon X-gal staining and all white colonies sequenced as expected. These data demonstrate that we have developed a highly functional platform for the generation of gRNA arrays using our golden gate cloning platform.

Assessment of gene editing frequencies using gRNA arrays

In order to validate the functionality of our assembled gRNA arrays, we transfected HEK293T cells with gRNA arrays containing 3, 5, 7, or 10 gRNAs and a plasmid expressing Cas9 alone or Cas9 linked via a P2A element to the human codon optimized Pseudomonas aeruginosa Csy4 ribonuclease [30] (Fig 2A). Notably, we observed negligible editing without expression of Csy4 to process the array into individual gRNAs, confirming the necessity of Csy4 for array processing (Panel A in S3 Fig). The results of nuclease activity for the 10 gRNA array transfected with Cas9-P2A-Csy4 demonstrated detectable rates of editing with the first four gRNAs in the array and then the editing diminished to nearly undetectable levels at gRNA 7 and 8, but editing was again observed with the 9^th and 10^th gRNA (Fig 2B).

Interestingly, two previous reports using 2 or 4 gRNA arrays differed in the length of flanked Csy4 target sites used, 20bp or 28bp [31,32]. The larger 28bp Csy4 sequence contains a ‘handle’ region of 8bp that may be important for Csy4 processing [27]. However, there is no clear evidence demonstrating which flanking Csy4 target sequence is superior for optimal Csy4 processing. Thus, we generated an additional set of pGG1-10 and pACPT plasmids harboring the 28bp Csy4 target site and again assembled the 10 gRNA array expressed via the U6 pol III promoter. The gRNA array containing the 20bp Csy4 site produced slightly higher levels of gene editing at all 10 target sites, indicating the 20bp Csy4 site may be more efficiently cleaved by Csy4 than the 28bp sequence (Fig 2B). However, we still observed low editing overall and nearly undetectable editing at gRNA 7 and 8. This phenomenon was even less prominent with shorter 3 and 5 gRNA arrays (Panel A in S4 Fig). These data demonstrate that golden gate assembled gRNA arrays with an optimal 20bp Csy4 site can induce detectable gene editing at up to 10 target sites, but the editing is highly reduced after the initial 4 gRNAs in the array when using larger arrays.

Pol II promoter driven gRNA arrays produce enhanced gene editing frequencies

While a 20bp Csy4 cleavage site leads to slightly higher gene editing frequencies than a 28bp site, the gene editing frequencies are lower than individual U6-gRNA editing overall. We hypothesized that the use of the canonical U6 driven expression of the highly repetitive array may produce low abundance of the gRNA array transcript for two main reasons. Firstly, the lack of a polyA tract and G cap associated with pol III driven genes may reduce the stability of the transcript to RNAses, and secondly pol III promoters typically express shorter transcripts (~100–500 nt) such as tRNAs, 5S rRNAs or miRNAs, which might limit the expression of an entire array of 10 gRNAs in tandem [30]. Thus, we removed the U6 promoter from the pACPT1-10 plasmids (Panel A in S5 Fig) and again assembled an array of 10 gRNAs that were subsequently cloned into a vector containing the strong pol II CMV promoter with a poly adenylation sequence (Fig 3A). The array was transfected into HEK293T and demonstrated improved gene editing frequencies overall, with ~10%-21% gene editing across all gRNA targets, albeit still at lower frequencies than individual U6-gRNA editing (Fig 3B, green line vs blue dots). We then tested the 10 gRNA array expressed from the very strong intron containing pol II CAG promoter with a poly adenylation sequence (Fig 3A). Surprisingly, we found the rates of gene editing using the CAG promoter were significantly higher than the individual U6-gRNA editing, with only 3 gRNA target sites having slightly lower editing when expressed in CAG driven array (Fig 3B, purple line, p < 0.001). We further tested the effect of the promoter driving the expression of arrays containing 3, or 5 gRNAs and again observed optimal gene editing using the CAG promoter (Panel A in S4 Fig). These results demonstrate gRNA arrays expressed from strong pol II promoters with polyadenylation sequences enhance gene editing frequencies to levels as high or higher than observed with individual standard U6-gRNA plasmids.

gRNA arrays produce significantly higher gene editing over multiplexed plasmid delivery

One of the motivations for the development of a robust gRNA array system is to avoid the toxicity and low transfection efficiency associated with delivery of numerous U6-gRNA containing plasmids. Thus, we wanted to test our gRNA array plasmids against with standard multiplexed plasmid delivery. We first attempted to deliver 10 individual U6-gRNA plasmids along with a Cas9 expressing plasmid using 250 ng of each plasmid as we did with our gRNA array and Csy4/Cas9 vector (Fig 4A). However, this produced substantial toxicity and death of nearly all cells transfected, likely due to the use of a large amount of plasmid. Next, we reduced the amount of each of the 10 individual U6-gRNA plasmids to 25 ng along with Cas9 to equal the same amount (in micrograms) as our gRNA array and Csy4/Cas9 vector. This resulted in no obvious toxicity and analysis of gene editing efficiency at all 10 target sites was significantly higher using our gRNA array approach (Fig 4B). The average editing efficiency for 10 individual U6-gRNA plasmids were significantly lower (8.2%) compared to with the gRNA array (25.0%). These data demonstrate the gRNA array system is a superior approach to the use of numerous multiplexed U6-gRNA plasmids.

gRNA arrays can work with Pol II promoters in multiplex synergistic activation mediator (SAM) activation

In order to allow for multiplex gene activation using gRNA arrays, we generated a set of pGG1-10 vectors with gRNAs containing 2 MS2 binding sites (Panel A and B in S6 Fig). These gRNAs, driven by a strong pol II promoter, are compatible with the highly effective SAM activation system [4]. We used this system to generate a gRNA activation array containing 5 previously validated gRNAs used for gene activation [12] (Fig 5A). We then transfected HEK293T cells with individual U6-MS2-gRNAs, dCas9-VP64 and MS2:p65:HSF1 plasmids to assess standard gene activation with the SAM system using individual U6 gRNAs (Fig 5B). These activation results were then compared with gene activation in cells transfected with the MS2 gRNA array, Csy4/dCas9-VP64 and MS2:HSF1:p65 plasmids. We observed robust levels of gene activation and, therefore these results demonstrate that MS2 gRNA arrays are amenable to multiplex gene activation.