Efficient CRISPR/Cas9-Mediated Versatile, Predictable, and Donor-Free Gene Knockout in Human Pluripotent Stem Cells

Summary

Loss-of-function studies in human pluripotent stem cells (hPSCs) require efficient methodologies for lesion of genes of interest. Here, we introduce a donor-free paired gRNA-guided CRISPR/Cas9 knockout strategy (paired-KO) for efficient and rapid gene ablation in hPSCs. Through paired-KO, we succeeded in targeting all genes of interest with high biallelic targeting efficiencies. More importantly, during paired-KO, the cleaved DNA was repaired mostly through direct end joining without insertions/deletions (precise ligation), and thus makes the lesion product predictable. The paired-KO remained highly efficient for one-step targeting of multiple genes and was also efficient for targeting of microRNA, while for long non-coding RNA over 8 kb, cleavage of a short fragment of the core promoter region was sufficient to eradicate downstream gene transcription. This work suggests that the paired-KO strategy is a simple and robust system for loss-of-function studies for both coding and non-coding genes in hPSCs.

Graphical Abstract

Highlights

• •Paired-KO represents an efficient gene ablation methodology in hPSCs
• •The cleavage sites and ligation pattern can be easily predicted in paired-KO
• •Paired-KO fits for loss-of-function studies for both coding and non-coding genes
• •Promoter removal is sufficient to eradicate downstream gene transcription

In this article, Zhang, Liu, Jin, and colleagues show that two adjacent double-strand breaks are preferentially repaired through precise ligation without insertions/deletions in human pluripotent stem cells. This strategy will thus serve as an efficient way to delete a gene fragment and represents a precise methodology for loss-of-function studies of both coding and non-coding genes in human pluripotent stem cells.

Introduction

The use of human pluripotent stem cells (hPSCs) in studying human biology and disease requires precise genome-targeting methodologies. While different from mouse cells, gene targeting in either human embryonic stem cells (hESCs) or human induced pluripotent stem cells (hiPSCs) is found to be technically difficult (Capecchi, 2005, Hockemeyer and Jaenisch, 2010, Zwaka and Thomson, 2003). This situation is now much improved with the development of the custom-engineered nucleases (CENs) (Hendriks et al., 2016). CENs are designed to introduce site-specific double-strand breaks (DSBs) within the genome, which trigger DNA repair and therefore facilitate genetic engineering (Hendriks et al., 2016, Hsu et al., 2014). To date three CENs have been developed, among which the type II clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) system is increasingly preferred, based on its high efficiency and less laboriousness compared with zinc-finger nucleases and transcription activator-like effector nucleases (Ding et al., 2013b, Hou et al., 2013).

The type II CRISPR/Cas9 system relies on three components as follows: the DNA nuclease Cas9, the CRISPR RNA (crRNA), and trans-activating crRNA (tracrRNA) (Jinek et al., 2012). To facilitate laboratory use, a single guide RNA (gRNA), which is encoded by a single vector plasmid (Cho et al., 2013, Cong et al., 2013, Mali et al., 2013), has now replaced the crRNA/tracrRNA duplex. The gRNA contains a 15- to 23-bp target DNA-matching sequence immediately upstream of a protospacer adjacent motif (PAM) (Fu et al., 2014, Hsu et al., 2013, Stemmer et al., 2015). The CRISPR/Cas9 derived from Streptococcus pyogenes is the most widely used system and has a 5′-NGG-3′ PAM (Cong et al., 2013, Mali et al., 2013). The 15- to 23-bp-long complementary sequence binds to the target genomic locus through strict Watson-Crick pairing (Fu et al., 2014, Hsu et al., 2013, Stemmer et al., 2015), where the gRNA directs Cas9 for DNA cleavage and therefore results in a DSB at a desired site (Jinek et al., 2012). A DSB introduced into the genome will initiate DNA repair through either error-prone non-homologous end joining (NHEJ) or homology-directed repair (HDR) in the presence of exogenous DNA templates (Geisinger et al., 2016, Heyer et al., 2010, Jasin and Rothstein, 2013, Lackner et al., 2015).

Both NHEJ and HDR have been used in hPSCs for gene editing (Byrne et al., 2015, Chen et al., 2014b, Ding et al., 2013a, Ding et al., 2013b, Genga et al., 2015, Hockemeyer et al., 2011, Hou et al., 2013, Kleinstiver et al., 2016, Slaymaker et al., 2016). For targeted gene knockout in hPSCs, NHEJ usually causes unpredicted insertions/deletions (indels), which makes the validation process technically troublesome and always generates unpredictable non-naturally existent proteins because of the frameshift introduced (Chen et al., 2015, Gonzalez et al., 2014). In addition, NHEJ cannot be applied for efficient knockout of non-coding RNAs. For HDR-mediated gene knockout in hPSCs, a recombination donor, which consists of the 5′ homology arm, an exon-replacing module and/or drug-resistant selection module, and the 3′ homology arm, is required for proper homologous recombination. Although outcome is predictable, the targeting efficiency of the HDR-based system is low and variable (Giudice and Trounson, 2008, Hockemeyer and Jaenisch, 2010, Mali and Cheng, 2012, Zwaka and Thomson, 2003). Moreover, the requirement for construction of the multimodule donor vector renders this system time consuming and labor intensive.

Here, we report an efficient paired gRNA-guided CRISPR/Cas9 knockout (paired-KO) method in hPSCs. Introducing a pair of gRNAs and Cas9 into hPSCs caused two DSBs, which were subsequently repaired with precise ligation without indels. Since each gRNA/Cas9 cleaves the targeting sequences between positions −3 and −4 upstream of the PAM sequence in a predictable manner, this paired gRNA strategy can serve as a powerful tool for coding gene knockout as long as the DSBs are designed shortly downstream of the ATG start codon and the frameshift triggers an immediate premature translation termination codon (PTC) downward. The paired-KO method also worked well for multiplexable gene targeting, and was equivalently efficient for lesion of microRNA (miRNA) and long non-coding RNA (lncRNA). We therefore conclude that the current paired-KO strategy is a rapid, robust, outcome-predictable, and multiplexable genome-editing methodology, which benefits loss-of-function studies of both coding and non-coding genes in hPSCs.

Results

Paired-KO Represents a Robust and Predictable Strategy for Gene Knockout in hPSCs

CRISPR/Cas9-introduced DSB mostly initiates NHEJ for DNA repair in the absence of recombination donors and is thus applied for gene lesion or mutation in cultured cells and animal models (Chen et al., 2015, Cong et al., 2013, Gonzalez et al., 2014, Shalem et al., 2014, Wang et al., 2014). It has been reported that introducing two DSBs could potentially increase the gene-targeting efficiency or induce genomic DNA-fragment deletion (Canver et al., 2014, Kang et al., 2015, Mandal et al., 2014, Zhang et al., 2015, Zheng et al., 2014). To ablate SMAD3 in hESCs, we designed two adjacent gRNAs in exon 3 of SMAD3 genome immediately downstream of the ATG start codon (Figure 1A). Both gRNAs and CAG promoter-driven Cas9-P2A-GFP vector were then co-electroporated into H9 hESCs together with a transient puromycin expression vector. Puromycin was added to the cells from day 2 to day 5 for a short period of selection, and drug-resistant clones were picked up and amplified for genomic DNA PCR analyses (Figure 1B). The predicted cleavage site of each complementary target region is between positions −3 and −4 relative to the PAM sequence, and for the two gRNAs designed for SMAD3 the interspace between these two cleavage sites is 44 bp long. We obtained 11 puromycin-resistant colonies, and genomic PCR analyses using primer sets flanking the two cleavage sites verified that seven colonies (7/11, 63.6%) showed biallelic genomic DNA deletion, while four (4/11, 36.4%) remained intact (Figure 1B). Western blotting confirmed that all seven colonies with genomic DNA deletion lacked SMAD3 expression, while all four uncut colonies expressed similar levels of endogenous SMAD3 (Figure 1B). We then performed Sanger sequencing for all seven cut PCR fragments, the results of which showed that in six colonies the cleavage sites were exactly between −3 and −4 bases upstream of both PAMs and that the cleaved blunt-ended genome DNAs were precisely ligated without indels (Figures 1C and S1A). One exception was colony #8, which showed one allele with indels after NHEJ (Figure S1A).

Figure 1.

Efficient SMAD3 Knockout in hPSCs through Paired-KO

(A) Schematic representation of the paired-KO strategy for SMAD3 knockout. The cleavage sites are pointed out by the scissors and arrows. The PAM sequences are in purple. The positions of the designed primers for genomic PCR are shown as arrows. PTC, premature translation termination codon.

(B) Eleven colonies retrieved from SMAD3 paired-KO in H9 hESCs. The top panel shows the genomic DNA PCR results and the lower panel shows western blotting results immunoblotted with the anti-SMAD2/3 antibody. Biallelically targeted colonies are labeled in red.

(C) Genomic DNA sequences before and after the paired gRNA-mediated cleavage and repair. A representative Sanger sequencing peak map verifies precise rejoining of the double blunt ends.

(D) qPCR data show decreased SMAD3 mRNA expression in the biallelically targeted colonies. Data are presented as mean ± SEM of three independent experiments. ^∗∗∗p < 0.001, unpaired two-tailed Student's t test.

(E and F) Efficient SMAD3 targeting in H1 hESCs (E) and hiPSCs (F) through paired-KO. Biallelically targeted colonies are labeled in red.

See also Figure S1.

We spent 3 days on SMAD3 gRNAs cloning, 10 days on colony formation after electroporation, and 5 days on cell expansion before genotyping analysis. Therefore, we obtained SMAD3 KO hESC lines within 20 days, which was much less laborious than our previous gene knockout strategy based on HDR (Chi et al., 2016). Moreover, two additional aspects make this paired-KO strategy even more fascinating. First, the biallelic gene-targeting efficiency was extremely high and there was no monoallelic gene deletion verified in this case. Second, different from results obtained from other mammalian cells, which frequently incorporated indels during the DNA repair process (Gonzalez et al., 2014), paired-KO hESCs showed predicted target-site cleavage and precise end joining without indels. We also designed another primer pair to extend the genomic DNA PCR fragment, whose results showed that the amplified 1,385-bp DNA products equally remained in all clones, suggesting that no large DNA fragments were deleted (Figures S1B and S1C). Therefore, paired-KO is more predictable and controllable than conventional strategies induced by single DSB-triggered NHEJ. Via careful gRNA selection in controlling the site and length of the DNA fragments designed for deletion, PTCs can be easily introduced, and the engineered mRNAs with PTCs targeted for degradation through the nonsense-mediated decay (NMD) pathway (Brogna and Wen, 2009, Fatscher et al., 2015, Lykke-Anderson and Jensen, 2015). Indeed, qPCR analysis showed 50% downregulation of SMAD3 mRNA in knockout cells compared with the wild-type counterpart (Figure 1D). Therefore, the current paired-KO strategy is more predictable and does not culminate in non-natural proteins, which may exhibit unexplainable biological functions.

To test whether the paired-KO strategy could be applied to other PSC lines, we performed the same experiment in H1 hESCs and hiPSCs (Hu et al., 2010). Again, the biallelic targeting efficiencies were 69.23% (9/13) and 66.67% (8/12) in H1 hESCs and hiPSCs, respectively (Figures 1E and 1F). Sanger sequencing data also confirmed that blunt ends were ligated without indels in both cell lines (Figure S1D).

Off-target effects are major concerns for applying CENs in genetic engineering. Introducing two gRNAs in the cells could possibly increase the off-target effects. We therefore analyzed the integrity of the potential off-target sites predicted by the online program designed by Zhang's laboratory (http://crispr.mit.edu/) through Sanger sequencing. As shown in Figures S1E and S1F, five principal suspected off-target sites for each gRNA remained intact in a randomly picked biallelically targeted colony. These data suggest that the off-target effect is not a major issue for paired-KO.

We next repeated this paired-KO paradigm by targeting CTNNB1 in hPSCs (Figure 2A). After electroporation and puromycin selection, 16 colonies were retrieved. Genomic DNA PCR results confirmed that six colonies were biallelically targeted, compared with only one monoallelic targeting colony and one mistargeted colony (Figure 2B). Western blotting and immunostaining experiments showed complete lack of CTNNB1 protein expression in the biallelically targeted lines (Figures 2C and 2D). Sanger sequencing and large DNA-fragment PCR amplification further revealed precise ligation of the double-cleaved blunt ends (Figures 2E and S2A–2D). Moreover, CTNNB1 mRNA level in biallelically targeted lines decreased up to 75%, further indicating an NMD response (Figure 2F). In addition, biallelic targeting of CTNNB1 in H1 hESCs and hiPSCs was also achieved with high efficiency (Figures 2G and 2H). Again, we did not find any off-target mutations in the predicted off-target sites (Figures S2E and S2F).

Figure 2.

Efficient CTNNB1 Knockout in hPSCs through Paired-KO

(A) Schematic representation of the paired-KO strategy for CTNNB1 knockout.

(B) Genomic DNA PCR results of 16 colonies retrieved from CTNNB1 paired-KO in H9 hESCs. Biallelically targeted colonies are labeled in red.

(C) Western blotting experiments confirm complete lack of CTNNB1 protein expression in biallelically targeted colonies.

(D) Confocal images show CTNNB1 and Nanog expression in wild-type (WT) and CTNNB1 knockout (KO) H9 hESCs by immunostaining. Scale bar, 50 μm.

(E) Genomic DNA sequences before and after the paired gRNA-mediated cleavage and repair. A representative Sanger sequencing peak map verifies precise rejoining of the double blunt ends.

(F) qPCR data show decreased CTNNB1 mRNA expression in the biallelically targeted colonies. Data are presented as mean ± SEM of three independent experiments. ^∗∗p < 0.01, unpaired two-tailed Student's t test.

(G and H) Efficient CTNNB1 targeting in H1 hESCs (G) and hiPSCs (H) through paired-KO. Biallelically targeted colonies are labeled in red.

See also Figure S2.

We also designed another ten pairs of gRNAs targeting cell-cycle regulators, glucose transporters, or transcription factors related to lineage development (Table 1 and Figure S3). With just one single attempt, we successfully obtained all desired knockout hESC lines. Among the total colonies counted, the percentage of lines showing at least one-allele gene targeting ranged from 33.33% to 95.35%, while the biallelic targeting efficiency ranged from 18.75% to 75.00% (Table 1). For each gene knockout, we sequenced two to three genomic DNA PCR-amplified fragments, and found that they were all precisely repaired (Figure S3). We screened all genes targeted for potential SNPs and found that MAF, GLUT3, and SHISA3 had both SNPs within a range of 5 kb upstream or 5 kb downstream of the cleavage sites in H9 hESCs. We then sequenced all MAF, GLUT3, and SHISA3 knockout colonies, the results of which showed that all colonies retained both upstream and downstream SNPs, indicating that no large genome fragments were deleted during the paired gRNA-mediated gene targeting.

Table 1.

Targeting Efficiencies of 14 Protein-Encoding Genes of Interest through Paired-KO in H9 hESCs

Gene	Clone Number	Targeted Number	Biallelically Targeted Number	Targeting Efficiency (%)	Biallelically Targeted Efficiency (%)
ADM	19	12	8	63.16	42.11
BHLHE40	17	11	7	64.71	41.18
DDIT4	26	24	13	92.31	50.00
GLUT3	16	8	3	50.00	18.75
SHISA3	19	17	10	89.47	52.63
CTNNB1	16	7	6	43.50	37.50
SMAD3	31	19	13	61.29	41.94
MAF	18	9	8	50.00	44.44
SP8	15	5	5	33.33	33.33
MEIS2	5	2	2	40.00	40.00
TP53	23	20	18	86.96	78.26
NF1	23	18	13	78.26	56.52
ASCL1	43	41	31	95.35	72.09
FOXG1	8	6	6	75.00	75.00

See also Figures S3 and S4.

One-Step Targeting of Multiple Genes with the Paired-KO Strategy

Given its high efficiency for biallelic gene targeting, we investigated whether the paired-KO strategy could be used for targeting of multiple genes simultaneously. NF1 and TP53 are two tumor-suppressor genes whose loss of function is closely related to many malignant tumors. We designed paired gRNAs targeting exon2 of NF1 and exon1 of TP53, respectively, which were all closely downstream of the ATG start codon (Figure 3A). All four gRNAs were electroporated into H9 hESCs with Cas9. After puromycin selection, we retrieved a total of 22 colonies. Genomic PCR results showed that for NF1, 13 out of 22 colonies showed biallelic DNA deletion, while 18 of 22 colonies had biallelic deletions for TP53 (Figure 3B). Among the 13 correctly targeted colonies for NF1, 11 were also biallelically targeted at the TP53 loci, demonstrating a 50% double-knockout efficiency (Figures 3B and 3F). TP53, but not NF1, is expressed in hESCs and, indeed, the biallelically targeted #4, #16, and #19 colonies showed no endogenous TP53 expression (Figure 3C). Sequencing results confirmed that the cleavage sites were just as expected, and the end-joining sites did not have any indels (Figure 3D). We then differentiated wild-type or double-knockout lines toward a neural fate for 25 days with a well-established protocol (Chi et al., 2016, Zhang et al., 2010), and qPCR analyses revealed a significant decay of both NF1 and TP53 mRNAs (Figure 3E).

Figure 3.

One-Step NF1 and TP53 Knockout in hESCs through Paired-KO

(A) Schematic representation of the paired-KO strategy for double knockout of NF1 and TP53 through one-step manipulation in hESCs.

(B) Genomic DNA PCR results of 22 colonies retrieved from H9 hESCs subjected to one-step targeting for both NF1 and TP53. Lines with both NF1 and TP53 biallelically targeted are labeled in red.

(C) Western blot shows complete TP53 knockout in the biallelically targeted lines.

(D) Genomic DNA sequences before and after the paired gRNA-mediated cleavage and repair in NF1 and TP53 loci. Sanger sequencing data of a double-knockout line confirm the precise ligations of the double blunt ends for both NF1 and TP53.

(E) qPCR data show decreased NF1 and TP53 mRNA expression in day-25 neural differentiation derivatives from the double-knockout lines. Data are presented as mean ± SEM of three independent experiments. ^∗∗p < 0.01, unpaired two-tailed Student's t test.

(F) Genome-targeting efficiency of NF1 and TP53.

Non-coding RNA Knockout via Paired-KO

Non-coding RNAs, including miRNAs and lncRNAs, play important roles in hPSC maintenance and lineage differentiation (Ivey and Srivastava, 2010, Jia et al., 2013, Pauli et al., 2011). Since the open-reading-frame frameshift mediated by the traditional NHEJ strategy is not sufficient to ablate the expression of a non-coding RNA, antisense and RNAi are widely used methods for downregulating their expression, although with still controversial efficacy. For gene targeting we first chose MIR1193, located at chromosome 14, whose biological functions remain to be investigated. Two gRNAs flanking MIR1193 loci were designed and then electroporated into H9 hESCs with Cas9 (Figure 4A). Drug-resistant colonies were then validated by genomic PCR. Our results showed that 4 out of 12 colonies showed biallelic deletions, suggesting that miRNAs can be knocked out in hESCs through the paired-KO strategy as efficiently as protein-encoding genes (Figure 4B).

Figure 4.

Targeting miRNA in hESCs through Paired-KO

(A) Schematic representation of the paired-KO strategy for MIR1193 knockout in hESCs.

(B) Genomic DNA PCR results show the monoallelic or biallelic deletion of MIR1193 in hESCs. Biallelically targeted colonies are labeled in red.

lncRNAs are defined as transcripts longer than 200 nucleotides without protein-coding potentials. The fragments we deleted for gene knockout described above via the paired-KO strategy were within 500 bp, so it remained unclear whether there is any tolerance for genomic DNA deletion related to the distance in between the double DSBs in hPSCs. MALAT1 is an lncRNA proposed to play a role in tumor metastasis, which is coded by just one exon longer than 8,000 bp (Gutschner et al., 2013). We designed five gRNAs whose target sites relative to the transcriptional start site (TSS) were about 0 bp (gRNA1), +500 bp (gRNA2), +1,000 bp (gRNA3), +3,000 bp (gRNA4), and +8,000 bp (gRNA5), respectively (Figure 5A). We then paired gRNA1 with each of the others and electroporated them in H9 hESCs to delete fragments of different length. gRNA1/2, gRNA1/3, and gRNA1/4 successfully yielded biallelic deletions with high efficiency, suggesting that the paired-KO strategy could be used to delete up to 3,000 bp of DNA fragments in the genome of hPSCs (Figures 5B–5D). However, in the gRNA1/5 pair designed to delete the entire lncRNA, only monoallelic targeting colonies were identified and no homozygous knockout colonies were successfully generated (Figure 5E).

Figure 5.

Targeting lncRNA in hESCs through Paired-KO

(A) Schematic representation of the paired-KO strategy for MALAT1 knockout in hESCs. Six gRNAs targeting different loci are designed and gRNA1 is paired with the remaining five gRNAs for deletion of desired fragments with different sizes. Forward and reverse primers are indicated by arrows.

(B) Genomic DNA PCR results show the monoallelic or biallelic deletion of the proximal 500-bp exon fragment of MALAT1 through gRNA1/2 combination. Biallelically targeted colonies are labeled in red.

(C) Genomic DNA PCR results show the monoallelic or biallelic deletion of the proximal 1,000-bp exon fragment of MALAT1 through gRNA1/3 combination. Biallelically targeted colonies are labeled in red.

(D) Genomic DNA PCR results show the monoallelic or biallelic deletion of the proximal 3,000-bp exon fragment of MALAT1 through gRNA1/4 combination. Biallelically targeted colonies are labeled in red.

(E) Genomic DNA PCR results show the paired gRNA/Cas9 yield monoallelic deletion of the entire ∼8,000-bp exon fragment of MALAT1 through gRNA1/5 combination.

(F) Sanger sequencing data show precise ligation in all four types of gRNA pairs with different deletion sizes.

(G) An inverse relationship between targeting efficiency and deletion size.

(H) Genomic DNA PCR results show the monoallelic or biallelic deletion of the 1,000-bp promoter region of MALAT1 through gRNA0/1 combination. Biallelically targeted colonies are labeled in red.

(I) qPCR results show the mRNA expression levels of MALAT1 in wild-type, heterozygous, or homozygous knockout colonies. Data are presented as mean ± SEM of three independent experiments. ^∗∗∗p < 0.001, unpaired two-tailed Student's t test.

Sanger sequencing further validated that in all four gRNA pairs, the double blunt ends were largely ligated with no indels, regardless of the length of deleted DNA fragments (Figure 5F). However, these results indicate an inverse relationship between the deletion size and deletion efficiency (Figure 5G). To fully eliminate MALAT1 expression, we constructed another gRNA (gRNA0), which targeted a locus 1,000 bp upstream of the TSS; this 1,000-bp-long DNA fragment belongs to the core promoter region of MALAT1 (Figure 5A). The gRNA0/1 pair was then electroporated into hESCs, and genomic PCR showed that among the six colonies chosen, three were biallelically targeted and one was monoallelically targeted (Figure 5H). qPCR analyses were then used to verify the mRNA expression levels of MALAT1 among all these six colonies. The colonies with biallelic 1,000-bp promoter deletion almost completely lacked mRNA expression, while the monoallelically targeted colony remained at half of the lncRNA gene expression (Figure 5I). These data suggest that promoter-based gene targeting may serve as an alternative method for efficient lesion of genes, especially large lncRNAs.

Discussion

Double gRNAs have been used to increase the gene-targeting efficiencies in either gene ablation or exogenous donor-based HDR (Chen et al., 2015, Hsu et al., 2014, Kang et al., 2015, Mandal et al., 2014, Zheng et al., 2014). It has also been reported that two adjacent DSBs could be used to create a deletion of the intervening fragment in both cultured cell lines and model organisms (Canver et al., 2014, Lee et al., 2010, Song et al., 2016, Zhang et al., 2015). In some reports, the two DSBs were repaired through NHEJ and indels were sometimes introduced (Canver et al., 2014, Chen et al., 2014a, Chen et al., 2014c, Ho et al., 2015, Lv et al., 2016, Zhou et al., 2014). In our current study, the paired-KO triggered DSBs are mostly ligated through direct end joining without indels in hPSCs. When Byrne et al. (2015) pioneered an experiment for replacing human THY1 gene with its mouse homolog through double gRNAs/Cas9-mediated cleavage in hiPSCs, they also found DNA-fragment deletion in between the two cleaved sites and rejoining of the blunt ends without indels. This precise NHEJ or precise gene deletion after paired gRNA-mediated DSBs has also been observed in HEK293 cells and hematopoietic stem cells (Geisinger et al., 2016, Ho et al., 2015, Mandal et al., 2014, Zheng et al., 2014). Given these predictable virtues of DNA cleavage and repair paradigms mediated by double gRNAs, the current paired-KO strategy is suitable for gene knockout in hPSCs without generating any non-natural proteins. Moreover, this paired gRNA strategy could also be used to delete a specific genomic fragment in hPSCs with designated cis-regulatory DNA element deletion, protein domain deletion, protein truncation, or mutation.

Another application of the paired-KO strategy is to ablate lncRNA expression in hPSCs. While >100-kb fragments have been successfully deleted in cell lines and animals (Canver et al., 2014, Chen et al., 2014c, Xiao et al., 2013, Zhang et al., 2015), we find that DNA fragments larger than 8 kb exhibit extremely low efficiency for gene targeting in hPSCs. Therefore, we offered two alternative ways for ablating the expression of lncRNA. For deletion of a short lncRNA or a small fragment within a large piece, direct DNA deletion through paired-KO works as efficiently as with coding genes, while for a large lncRNA itself, removal of the core promoter region may be more practical for complete loss-of-function analysis.

In summary, we report that the paired-KO strategy is an efficient, outcome-predictable, and multiplexable genome-editing methodology, suitable for loss-of-function studies of both coding and non-coding genes in hESCs and hiPSCs.

Experimental Procedures

hPSC Culture

hESCs (H9, WA09 or H1, WA01, passages 25–45, WiCell Agreement No. 14-W0377) were maintained on mouse embryonic fibroblast (MEF) feeder in a hESC culture medium as we previously described (Chi et al., 2016, Zhang et al., 2010). The components of the hESC culture medium (hESCM) are: DMEM/F12, 20% Knockout serum replacer, 1× minimal essential medium (MEM) non-essential amino acids solution, 1× L-glutamine solution, 0.1 mM β-mercaptoethanol, and 4 ng/mL fibroblast growth factor 2. Cells were passaged every 5 days through dispase (Gibco, 17105) digestion and the medium was changed every day. The hiPSCs line used in this study has been reported in our previous paper (Hu et al., 2010).

Neural Differentiation of hPSCs

The detailed procedure for neural differentiation of hPSCs was described previously (Zhang et al., 2001, Zhang et al., 2010). In brief, after dispase digestion, hPSCs were removed from the MEF layers and then pipetted up and down against the bottom of a 50-mL conical centrifuge tube to break up the hPSC clusters into 100- to 200-μM pieces. Within the first 4 days, hESCs were suspended in hESCM to form embryoid bodies and medium was then switched to the neural induction medium to induce the cells toward a neuroectoderm fate. At day 6 post differentiation, cell aggregates were plated down on a laminin-coated culture surface. At day 17, neural progenitor cells were detached from the plate and maintained in suspension to form neurospheres. The recipe for NIM is as follows: DMEM/F12, 1× N2 supplement, 1× MEM non-essential amino acids solution, and 2 μg/mL heparin.

Construction of the gRNA Plasmids

The blank gRNA vector with two BbsI restriction sites was as described previously (Chen et al., 2015). The gRNA vector was digested with BbsI, gel purified, and ligated to the annealed oligos containing targeting sequences of designed gRNAs. For the detailed gRNA sequences for each gene, see Supplemental Experimental Procedures.

Generation of Knockout hPSC Lines

hPSCs were cultured in hESCM with 1 mM Y-27632 (Calbiochem, Y27632), a Rho kinase inhibitor, for 3 hr prior to electroporation. Cells were then digested by trypsin for 3 min into single cells and rinsed with PBS twice. Cells (1 × 10⁷) were electroporated with appropriate combination of Cas9 plasmids (5 μg) (Addgene #44719), two gRNAs (5 μg), and CAG promoter-driven puromycin plasmid (5 μg) (Cui et al., 2016) in 200 μL of electroporation buffer using the Gene Pulser Xcell System (Bio-Rad) at 250 V and 500 μF in 0.4-cm cuvettes (Phenix Research Products). Cells were treated with puromycin (0.5 μg/mL) from day 2 to day 5 post electroporation. After puromycin selection MEF-conditioned hESCM was supplied, and drug-resistant colonies could be picked up for genotyping analyses after 10 days post electroporation. Primer sets for genomic DNA PCR are provided in Supplemental Experimental Procedures.

Immunofluorescence

Coverslip cultures were fixed in 4% paraformaldehyde at room temperature for 10 min, washed two times with PBS, and incubated for 1 hr in blocking buffer (PBS, 10% donkey serum, 0.1% Triton X-100). Cells were incubated with primary antibodies at 4°C overnight followed by three washes and stained with the fluorescently conjugated secondary antibodies (1:1,000; Jackson Laboratories) for 1 hr. Nuclei were counterstained with Hoechst 33,258 (Sigma D9542) for 5 min. Primary antibodies used in this study include the following: CTNNB1 (1:1,000; mouse immunoglobulin G [IgG], BD Biosciences 610154), and Nanog (1:2,000; goat IgG, R&D Systems AF1997).

Western Blotting

Cells were lysed in RIPA⁺ buffer and protein concentrations were analyzed by the BCA kit (Thermo Scientific). Fifteen micrograms of total proteins was resolved on 10% SDS-PAGE gel, transferred to nitrocellulose membranes, and blotted for SMAD2/3 (1:1,000; rabbit IgG, Cell Signaling Technology #8685), CTNNB1 (1:1,000; mouse IgG, BD Biosciences 610154), GAPDH (1:2,000; rabbit IgG, Santa Cruz Biotechnology SC25778), TP53 (1:1,000; rabbit IgG, Santa Cruz SC6243). The secondary antibodies were purchased from Jackson ImmunoResearch.

Statistical Analyses

Data were analyzed using Student's t test for comparison of independent means with pooled estimates of common variances.

Author Contributions

Z.X., L.L., J.G., and X. L. conceived the study. L.Z., H.Y., and S.L. performed most of the experiments. C.Z., X.X., C.L., F.B., F.Y., L.Y., M.L., W.Y., and Z.Q. helped to set up the CRISPR/Cas9-mediated gene-targeting system in hPSCs. Z.Q. helped with the data analyses and proofreading of the manuscript. L.Z., H.Y., S.L., L.L., and X.Z. wrote the manuscript.

Published: September 1, 2016