High efficiency genomic editing in Epstein-Barr virus-transformed lymphoblastoid B cells

While lymphoblastoid cell lines (LCLs) represent a valuable resource for population genetic studies, they are usually regarded as difficult for CRISPR-mediated genomic editing. It would be valuable to be able to take the results of their functional variant studies and test them in the same LCLs. We describe a protocol using a single-stranded donor oligonucleotide (ssODN) strategy for ‘scarless’ editing in LCLs. The protocol involves optimized transfection, flow cytometric sorting of transfected cells to single cells in multi-well plates and growth in conditioned, serum-rich medium, followed by characterization of the clones. Amplicon sequencing reveals the relative proportions of alleles with different editing events, with sequencing of DNA from clones showing the frequencies of events in individual cells. We find 12/60 (20%) of clones selected in this manner to have the desired ssODN-mediated recombination event. Long-range PCR of DNA at the edited locus and of RT-PCR products for the gene traversing the edited locus reveals 3/6 characterized clones (50%) to have large structural mutations of the region that are missed by sequencing just the edited site. The protocol does not require the use of lentiviruses or stable transfection, and makes LCLs a realistic cell type for consideration for CRISPR-mediated genomic targeting.


INTRODUCTION
Lymphoblastoid cell lines (LCLs) have been collected for decades as a self-renewing cellular resource from different individuals, initially focusing on multi-generational families to facilitate linkage mapping (1). In recent years, they have been collected as a resource for population 30 genetics studies, by the HapMap Consortium (2), the GEUVADIS Project (3), and the 1000 Genomes Project (4), among others. Repositories such as the Coriell Institute and the American Type Culture Collection have between them banked tens of thousands of LCLs, indicating the scale at which these resources are being generated. LCLs continue to be successfully exploited as a model system for understanding how DNA sequence variability 35 leads to individual differences in transcriptional regulatory variation. LCL panels have also proven useful in pharmacogenomic studies (5). Furthermore, individual LCLs have served as reference cell lines for a number of purposes. The 17 member, 3 generation CEPH pedigree 1463 has been the target of Illumina's Platinum Genomes sequencing, while the mother in the second generation of this family is the GM12878 cell line, a tier 1 ENCODE cell line (6) that 40 has been used in many novel assays, including some of ours (7), assays studying chromatin looping (8), and the Genome in a Bottle initiative (9). The position of LCLs in general and some lines in particular at the center of many modern molecular genomic discoveries is clear.
However, the discovery of functional variants in LCLs is rarely followed by genomic editing in the same cell type (10) unless lentiviral transfection is used to express Cas9 robustly (11). If 45 the locus containing the functional variant is very LCL-specific in its activity, it may be difficult to test the effect of DNA sequence polymorphism at the locus in a surrogate cell type.
Furthermore, if LCLs are to be used in CRISPR screens of the whole genome (12), an effective protocol for its use in this cell type is going to be needed. The most successful reported use to date of CRISPR in LCLs does not describe the efficiency of their strategy, making it difficult 50 to replicate (10). We therefore sought to test how efficiently homology-directed recombination (HDR) using a single-stranded donor oligonucleotide (ssODN) could be used for 'scarless', CRISPR-mediated editing in LCLs.

CRISPR/Cas9 editing yields a high proportion of successfully-edited clones 55
Our protocol overview is shown in Figure 1. The oligonucleotides used for this study are provide in Table 1. The protocol details are in the Methods section. We started with 4 × 10 6 LCL cells and 33.3 µg of plasmid for transfection, observing this to be accompanied by ~40% cell death. Overnight incubation allowed the GFP to be expressed in transfected cells, allowing fluorescence-activated cell sorting (FACS) to deliver individual cells into 96 well plates. 60 We plated over 2,000 individual cells and allowed the cells to grow in conditioned, FBS-rich medium for 2 weeks. Almost one-quarter of the plated individual cells were growing at this point. We re-plated a total of 480 cells, and chose the 60 fastest-growing cells colonies for screening. We performed PCR to amplify the targeted region, and then digested the PCR product from each clone with the StyI restriction endonuclease, as its recognition site, present 65 in the unedited cell line, should be disrupted by any genomic editing event (Figure 2a). This revealed 55 of the 60 clones to have had some sort of editing event at the locus, prompting the sequencing of these amplicons, and revealing that 12 of the 55 had apparent homozygosity for the desired homology-directed recombination (HDR) event.

Amplicon-seq reveals the distribution of different types of editing events
To gain a more comprehensive insight into the editing occurring in the cells, we performed amplicon sequencing of the pool of cells prior to the cloning step. We show the results of analysis of these data using CRISPResso (13) in Figure 2b. The amplicon PCR primers were designed to flank and avoid overlapping the ssODN sequence. The most common edit was 75 the desired HDR (25.96%), followed by the 1 bp deletion immediately upstream of the cut site representing non-homologous end-joining (NHEJ, 25.21%), with only 16.35% of the sequences representing unedited alleles. The remaining approximately one-third of alleles represented a wide range of events.
In Figure 3, we show the results of another experiment testing the effect of using the single-80 stranded oligonucleotide (ssODN) template. In Figure 3a, we show that the proportion of events with HDR without using the ssODN was zero, but rose to 0.233 when the ssODN was used. In Figure 3b, we separate events involving just HDR from those which also involve other types of NHEJ-mediated repair. By definition, the desired insertion and substitutions occur at 100% frequency in the HDR-only alleles, but when we add back the NHEJ events, 85 we see that they are most frequent immediately adjacent to the cut site, and fall off in frequency even within the first 10 bp from the cut site.

Testing for large rearrangements at the targeted locus
As it has recently been described that CRISPR-mediated genomic editing can induce large 90 structural rearrangements that would be missed when testing just the immediate <200 bp around the target site (14), we performed two types of screening to test for these events. One strategy was to perform long-range PCR in the 4 kb surrounding the locus, making sure neither primer was near the cut site or overlapped the ssODN sequence. As the targeted locus is within an intron of a gene that is transcribed over almost 200 kb of the genome, we also 95 designed RT-PCR primers, each amplifying ~1 kb fragments, spanning multiple exons and thus a larger region flanking the CRISPR target. We show the characteristics of the 5 cell samples tested in Figure (14), and demonstrating that the apparent homozygosity observed for the 7 bp deletion clone is instead likely to be due to hemizygosity at this locus.

DISCUSSION 110
As has been demonstrated in a prior report, LCLs can be edited using CRISPR and HDR to modify a locus to change from one to another human single nucleotide variant at a locus regulating local gene expression (10) . We take this observation further to reveal the critical steps in the protocol, not just the points of greatest inefficiencies, but also some surprising efficiencies. 115 We note that the transfection rate is highly sensitive to the number of cells used, requiring that very accurate cell counting be performed. We confirm some previous experimental design optimizations for improved HDR, including the design of ssODNs that do not hybridize with gRNAs or the template strand of transcribed genes, and the use of a double-stranded nuclease instead of nickase (15). By targeting the HDR-directed mutations to within the guide 120 RNA binding site, we reduce the re-cutting of the site by CRISPR to enhance scarless editing (16). Furthermore, we confirm that cuts close to the target edit site are more efficient (15,17).
Our study did not encompass testing for off-target effects. We attempted to guard against this by using a high-fidelty espCas9(1.1) that is created through mutations that affect its propensity to cleave when there are mismatches between the gRNa and the protospacer (18). The group 125 who designed this Cas9 did not observe any edits at common off-target sites when these were were assessed (18), which should apply in LCLs to the same extent as other cells.
There are some simple guidelines worth following when editing LCLs. The protocol outline shown in Figure 1 provides some guidance about cell numbers at each step. We had excessive cell loss during culture because the wells at the outside of the multi-well plate 130 evaporated more quickly. A trick for the future would be to keep those wells filled with water and use only the internal wells for cell culture. Plasmids need to be highly concentrated to avoid diluting nucleofection reagents. We anticipate that the efficiency of the protocol will improve markedly over time. With the introduction of ribonucleoprotein reagents for guide RNAs and Cas9, the use of plasmids will possibly diminish. All of the techniques we used are 135 potentially automated by liquid handling/cell culture systems, raising the possibility that LCL editing may be amenable to scaling to many cell lines at a time, potentially introducing the same variant in multiple genetic backgrounds to create a system for studying phenomena like epistasis, lending further value to this research workhorse cell type.

Cell line culture
The lymphoblastoid cell line (LCL) derived from a child within the CEPH Pedigree 1463 (GM12881) was purchased from the Coriell Institute and cultured in RPMI 1640 medium, supplemented with 15% fetal bovine serum (FBS, Benchmark), 100 IU/ml penicillin, and 100 145 µg/ml streptomycin (Life Technologies). Cells were kept in suspension in tissue culture flasks (NUNC, Thermo Scientific) at 37 °C in a 5% CO2 incubator and maintained between 2 x 10 5 and 8 x 10 5 cells/ml.

CRISPR/Cas9 editing and sorting
For transfection, cells were passaged at 3.5x10 5 48 hours and 24 hours before transfection.  Driven by evaporation, we recommend plating water in the peripheral wells of a 96-well plate to avoid disruption of cell growth by crystal formation. Conditioned medium was obtained from GM12881 cells, and cultured in 20% FBS RPMI 1640 for 24 hours. The medium was removed without disturbing cells at the bottom of the flask, centrifuged at 2,000 rpm, and the supernatant was filtered through a 0.3 micron sterile filter prior to use. When subsequent T7EI 185 assays were to be performed, cells were sorted directly into QuickExtract DNA extraction solution (Epicentre).

T7 endonuclease I assay (T7EI)
To verify editing in clones, in which an intronic enhancer of TBC1D4 was targeted, genomic 190 DNA was isolated from transfected and control cell pellets using QuickExtract DNA Extraction Solution (Epicentre) according to manufacturer's instructions. DNA was then concentrated by ethanol precipitation. The 1 kb region containing the gRNA targeted region was amplified with forward and reverse primers (Table S1)

Amplicon-seq generation and data analysis
Cell lines generated by CRISPR/Cas9 editing at a locus with or without a repair template were 205 assessed by amplicon-seq. A suspension of 10 4 cells sorted in QuickExtract DNA extraction solution (Epicentre) was used for DNA extraction according to manufacturer's instructions.
DNA was extracted by vortexing the cell suspension for 15 seconds, followed by incubation at 65°C for 6 minutes, an additional 15 second vortexing, and a final 2 minute incubation at 98°C. DNA was then concentrated by ethanol precipitation and submitted to an initial PCR with 210 locus-specific forward and reverse primers with portions of the Illumina TruSeq adapters on their 5' ends. PCR products were purified with DNA Clean and Concentrator-5 kit (Zymo), and a second round of PCR was performed on purified DNA with primers containing the remaining Illumina adapter along with a custom 6-nt index on the reverse primer. The amplicon libraries were then purified by gel extraction and sequencing using Illumina MiSeq technology, 250 bp 215 single end sequencing. The resulting data were analysed using CRISPResso (13).

RT-PCR of the TBC1D4 1 kb amplicon
Cell pellets were treated with QIAzol lysis reagent (Qiagen) and total RNA was isolated using the miRNAeasy kit (Qiagen) combined with on column DNAse (Qiagen) treatment according 220 to manufacturer's instructions. Synthesis of cDNA was performed with total RNA and SuperScript III First-Strand Synthesis System for RT-PCR (Life technologies) using oligo(dT)20 as primers. Subsequent PCR was conducted with primers designed using the NCBI Primer-BLAST web interface (20) (Table S1) and the Q5 hot start high fidelity polymerase master mix, according to the manufacturer's protocol. Samples were then analyzed on a 2% agarose gel. 225

T7-TBC1D4-FWD
GGCCACCATACCATCTTCACA T7-TBC1D4-REV ATTTGGCTCTGCTTGTAGCC   shown. Substantial cell death occurs following transfection, but the cloning steps were 315 relatively efficient, and the rate of recovery of edited clones was high. to which the guide RNA binds, is shown to be at the same site as is being targeted for editing by the ssODN, which then prevents the guide RNA from binding to cause further edits. The location of a StyI recognition motif present in the unedited DNA is shown, demonstrating how restriction enzyme digestion with StyI for this particular editing event can be used in screening 325 for editing events. In (b) we show the results of amplicon sequencing, and the relative frequencies of each type of editing event. The desired editing event is the most common event, followed by non-homologous end joining (NHEJ) with deletion of the single nucleotide immediately at the cut site.