CRISPRi-mediated gene knockdown is highly specific to the target gene

To test the specificity of CRISPRi in human cells, we used RNA-seq to quantify the transcriptome of GFP+ HEK293 cells expressing dCas9-KRAB and either a negative control sgRNA (sgGAL4–1) or an sgRNA targeting GFP (sgGFP-NT1). Our results show that CRISPRi is highly specific as GFP is the only gene that is significantly suppressed by the GFP-targeting sgRNA (Figure 2 and Figure S3A). If we average the data from two independent biological replicates, no gene other than GFP changes by more than 1.5 fold. For the few genes where we measured changes in expression, we find no evidence for the presence of a near match to the sgRNA seed sequence suggesting that the changes observed are due to a small level of noise in RNA-seq measurements. Furthermore, we observed few changes in gene expression in the negative control sample, suggesting that the negative control sgRNA triggers minimal off-target effects (Figure 2 and Figure S3A). We also found no evidence for significant on- or off-target effects due to expression of either sgGAL4 or sgGFP-NT1 alone (Figure S3B). Finally, expression of dCas9-KRAB resulted in minimal changes in gene expression (Figure S3C).

CRISPRi can silence endogenous human genes

As there can be important differences in how endogenous and reporter genes are transcribed, we tested whether CRISPRi can silence endogenous human genes. We designed sgRNAs to target genes encoding cell surface transmembrane proteins, enabling us to use flow cytometry with directly conjugated fluorescent antibodies to quantify gene expression at the single-cell level. We cloned 10 sgRNAs each for the transferrin receptor (CD71) and C-X-C chemokine receptor type 4 (CXCR4), which were designed to span a 500-bp region upstream or downstream of the transcription start site, targeting either template or non-template strand (Table S1). Each sgRNA was transfected into HeLa cells stably expressing dCas9 or dCas9-KRAB. Cells were dissociated and stained for CD71 and CXCR4 expression 3 days after transfection. In this experiment we found that both dCas9 and dCas9-KRAB proteins can knock down endogenous expression of CD71 and CXCR4 (Figure 3A–B), with 3 out of 10 sgRNAs for each gene showing significant repression (60–80% repression). Interestingly, a strong correlation exists between the degree of knockdown for dCas9 and dCas9-KRAB for endogenous genes, suggesting that sgRNA binding to DNA or local chromatin structure may be limiting factors which dictate gene knockdown (Figure S4A). Further work is required to systematically analyze why recruitment of a KRAB domain has an effect in some contexts but not others.

We also tested whether stable lentivirus-mediated expression of both the sgRNA and dCas9 or dCas9-KRAB could repress endogenous gene expression. We transduced HeLa cells expressing dCas9 or dCas9-KRAB with three sgRNAs targeting CD71 or CXCR4 and 7 or 8 days later measured the amount of CD71 or CXCR4 expressed. Our data show CD71 and CXCR4 can be efficiently and stably knocked down (Figure 3C–D), providing evidence that CRISPRi can be used like RNAi to suppress endogenous gene expression.

A valuable extension of CRISPRi is the capability for multiplexed gene regulation in human cells. To test whether we can suppress the expression of multiple genes, we transfected sgRNAs targeting CD71, CXCR4 or both into HeLa cells stably expressing dCas9-KRAB. Cells were dissociated and stained for CD71 and CXCR4 expression 5 days after transfection. In this experiment we found that we can simultaneously knock down endogenous expression of CD71 and CXCR4 (Figure 3E). This result was independently confirmed by qPCR (Figure S4B).

CRISPRi is a general method for silencing transcription of endogenous genes in eukaryotes

To test if CRISPRi can be used for transcription repression in other eukaryotes, we examined whether dCas9, alone or fused to a transcriptional repressor, will effectively silence gene expression in Saccharomyces cerevisiae. We cloned dCas9 with 2 NLS into a vector under control of the TDH3 promoter (Figure S4C). Using this same scheme, we also fused dCas9 to Mxi1, a mammalian transcriptional repressor domain reported to interact with the histone deacetylase Sin3 homolog in yeast (Figure S4C) (Harper et al., 1996; Kasten et al., 1996; Schreiber-Agus et al., 1995). Both dCas9 fusion proteins were transformed into a strain carrying a TEF1-GFP fusion expressed from the endogenous locus (Huh et al., 2003). We expressed an sgRNA that targets the endogenous TEF1 locus under control of the Pol III SNR52 promoter (Figure S4C) (Dicarlo et al., 2013). With dCas9 alone, we observed 18-fold repression, which is increased to 53-fold with addition of the Mxi1 domain (Figure 3F). As CRISPRi shows robust endogenous gene silencing in yeast, it could also prove a useful tool for studying organisms with limited genetic tools, such as the pathogenic yeast Candida albicans.

We found that both the Mxi1 and KRAB domain strongly suppress transcription when targeted to some regions of DNA but not to others. Additionally, some genes may require recruitment of repressive complexes other than Kap1/HP1 or Sin3. Ongoing work will define what regions of a gene are most effective for dCas9 fusion mediated gene silencing. However, the remarkable success in using RNA-guided Cas9 to enable DNA cutting for many genes in distinct organisms argues that Cas9 can bind to target DNA sites in a broad range of genes or regulatory regions (Cho et al., 2013; Cong et al., 2013; Dicarlo et al., 2013; Gratz et al., 2013; Hwang et al., 2013; Jiang et al., 2013; Jinek et al., 2012, 2013; Mali et al., 2013; Wang et al., 2013). Taken together with the present study, these findings suggest that CRISPRi will be a general method for efficiently and specifically regulating transcription in many eukaryotes.

Mapping the function of enhancers and chromosome topology with dCas9

Our results provide evidence that dCas9 can target enhancers, introns, and other noncoding elements to map the regulatory functions of these elements on transcription. This could enable sophisticated high throughput mapping of enhancers at endogenous loci. In theory, using inducible dimerization systems, dCas9 could tether a distant enhancer close to the promoter of a gene to interrogate how chromosome conformation regulates transcription. Recent work has shown that during neural differentiation, CCCTC binding factor, cohesion and the mediator complex modulate chromosome topology to control development (Millau and Gaudreau, 2011; Phillips-Cremins et al., 2013). dCas9 could enable studying these processes in a much more dynamic and programmable manner by artificially creating different genomic conformations and measuring how these modulate transcription and development.

CRISPRi efficiently and specifically represses transcription in human cells

In our experiments, we observed robust gene knockdown of both reporter and endogenous genes (5~15-fold repression in human and 50-fold in yeast). Our method uses common lentiviral constructs to express both dCas9 and sgRNAs to achieve stable long term gene knockdown. This is comparable to the efficiency of existing gene knockdown techniques such as RNAi and TALE proteins. While we do not yet know what fraction of genes can be silenced by a dCas9 fusion protein, a significant portion of the genome should be targetable. In human cells, previous work has demonstrated that local chromatin context is a crucial determinant of KRAB mediated silencing. For example, recruitment of the KRAB domain to adjacent genomic regions can have no effect or can completely silence transcription (Groner et al., 2010). The biology underlying this observation remains to be characterized and CRISPRi could provide a tool to map chromatin structure and the function of chromatin remodeling complexes. Nonetheless, an exon trapping strategy in which Tet binding sites were randomly inserted in the genome allowing recruitment of a Tetracycline repressor-KRAB fusion protein revealed that ~80% of genes tested can be silenced by KRAB-mediated recruitment of Kap1 and HP1 proteins (Groner et al., 2010). Ongoing work will characterize additional fusion partners to understand which transcription repressor complexes are most useful for CRISPRi in different cell types.

Our RNA-seq results show that CRISPRi is highly specific with minimal off-target effects for two distinct sgRNA sequences. By contrast, genome editing by the catalytically active Cas9 nuclease can be accompanied by off-target genomic alterations which occur at loci where near cognate DNA seed matches exist (Fu et al., 2013). The specificity of CRISPRi remains to be determined across a broad range of sgRNA sequences. Nonetheless, it may be substantially easier to specifically modulate transcription as only a small fraction of the human genome is transcribed or required for controlling transcription thus binding of catalytically inactive dCas9, and it is derivatives, at near cognate sites will often not impact transcription.

Future optimization for CRISPRi technology

The observation that the sgRNA is a limiting factor for CRISPRi function suggests that improvements to RNA stability or loading into dCas9 could improve both CRISPRi and Cas9 mediated genome editing in human cells (Jinek et al., 2013). While we are able to target dCas9 to both endogenous and reporter genes, not all sgRNA targeting constructs work efficiently. Specific sgRNA sequence composition may improve sgRNA stability, loading into dCas9, or binding to DNA. Alternately, dCas9 binding may be strongly dictated by the local chromatin structure of the target sequence. The recently published genome-wide maps of chromatin structure can inform where one targets dCas9 for both transcription repression and activation (Thurman et al., 2012). High-throughput analysis of sgRNAs that are functional will discriminate between sequence requirements and local chromatin states to determine rules for dCas9 targeting and function. While the experiments described here have examined a set of fusion partners and fusion design strategies, future work will be required to refine and improve CRISPRi fusion proteins with different effectors for optimal performance. Finally, as additional Cas9 proteins are characterized, the use of parallel orthogonal dCas9 proteins with cognate sgRNAs will enable controlling multiple genes differentially (Chylinski et al., 2013; Zhang et al., 2013).

Plasmid design and construction for CRISPRi in human cells

The sequence encoding mammalian codon optimized Streptococcus pyogenes dCas9 (DNA 2.0) was fused with two C-terminal SV40 nuclear localization sequences (NLS) and tagBFP alone or with the KRAB, CSD, WRPW, VP64, or p65AD domains. Using Gibson cloning we cloned these fusion proteins into MSCV-Puro (Clontech) or pHR (Addgene). sgRNAs were expressed using a lentiviral U6 based expression vector derived from pSico which co-expresses mCherry-T2A–Puro from a CMV promoter. The sgRNA expression plasmids were cloned by inserting annealed oligos into the lentiviral U6 based expression vector that was digested by BstXI and XhoI. To detect indel formation in GFP in human cells, we purified genomic DNA from cells expressing dCas9-KRAB and either a negative control sgRNA or sgGFP-NT1 that had been grown in culture for 3 weeks. We PCR amplified the SV40 promoter and GFP, and sequenced the PCR products.

Cell culture, DNA transfections, viral production and fluorescence measurements for CRISPRi in human cells

HEK293 and HeLa cells were maintained in Dulbecco’s modified eagle medium (DMEM) in 10 % FBS, 2 mM glutamine, 100 units/mL streptomycin and 100 µg/mL penicillin. Lentivirus and retrovirus were produced by transfecting HEK293 or amphotropic Phoenix packaging cell lines with standard packaging vectors. HEK293 or HeLa cell lines were generated by transducing cells with a MSCV retrovirus expressing GFP from the SV40 promoter or lentivirus expressing dCas9, dCas9-KRAB from an SFFV promoter, or the sgRNA from a U6 promoter or a Gal4-GFP reporter. Pure populations of each stable cell line were sorted by flow cytometry using a BD FACS Aria2 for stable GFP, BFP or mCherry expression. For transient transfection experiments, reporter HEK293 cell lines were transfected using TransIT-LT1 transfection reagent (Mirus) with the manufacturers recommended protocol in 24 well plates. Cells were transfected with 0.5 µg of the dCas9 expression plasmid and 0.5 µg of the RNA expression plasmid, or if dCas9 is stably expressed with 1 µg of the RNA expression plasmid. Two sgRNA plasmids (0.75 µg each) were co-transfected for double knockdown experiments. For 6–7 day experiments, the cells are split at day 3. At 3 or 6 days following transfection or 7–14 days following transduction, cells were trypsinized to a single cell suspension, gated on mCherry positive population (top 30– 60% for transient transfection experiments) and GFP expression was analyzed. To analyze CXCR4 or CD71 expression cells were dissociated in 10 mM EDTA PBS and then stained in PBS/10% FBS for 1 hour at room temperature. Isotype control, CXCR4, and CD71antibodies for flow cytometry were purchased from MBL or eBioscience and used at 0.05 mg/mL. All flow cytometry analysis was performed using a LSR II flow cytometer (BD Biosciences).

Quantitative RT-PCR

Cells were harvested using trypsin (Invitrogen), and total RNA was isolated using the RNeasy Mini Kit (Qiagen), according to manufacturer’s instructions. RNA was converted to cDNA using AMV reverse transcriptase under standard conditions with oligo dT primers and RNasin (Promega). Quantitative PCR reactions were prepared with Go-Taq (Promega) and SYBR Green (Invitrogen), according to the manufacturer’s instructions. Reactions were run on a LightCycler thermal cycler (Roche). Primer sequences for CXCR4, CD71, and GAPDH are listed below.

• CXCR4: 5’ GAAGCTGTTGGCTGAAAAGG 3’ CTCACTGACGTTGGCAAAGA

• CD71: 5’ AAAATCCGGTGTAGGCACAG 3’ GCACTCCAACTGGCAAAGAT

• GAPDH: 5’ ACAGTCAGCCGCATCTTCTT 3’ ACGACCAAATCCGTTGACTC

The authors thank Martin Jinek for discussion and distribution of the dCas9 gene. The authors also thank Jason Park, Martin Kampmann, Mike Bassik, and Xin Xiong for lentiviral vectors, technical advice and helpful discussion. L.S.Q. acknowledges support from the UCSF Center for Systems and Synthetic Biology. This work was supported by NIH P50 GM102706 (J.S.W., J.A.D), NIH P50 GM081879 (L.S.Q., W.A.L., Z.L., E.H.W.), NIH U01 {"type":"entrez-nucleotide","attrs":{"text":"CA168370","term_id":"35089724","term_text":"CA168370"}}CA168370 (J.S.W), Howard Hughes Medical Institute (L.A.G., G.A.B., O.B., J.A.D., J.S.W., W.A.L.), NSF SynBERC EEC-0540879 (W.A.L.), NIGMS IMSD (S.E.T.), the Human Frontiers Scientific Program (N.S., L.M.), and a Ruth L. Kirschstein National Research Service Award (M.H.L.).