Doxycycline-Dependent Self-Inactivation of CRISPR-Cas9 to Temporally Regulate On- and Off-Target Editing

Abstract

Exome and deep sequencing of cells treated with a panel of lentiviral guide RNA demonstrate that both on- and off-target editing proceed in a time-dependent manner. Thus, methods to temporally control Cas9 activity would be beneficial. To address this need, we describe a "self-inactivating CRISPR (SiC)" system consisting of a single guide RNA that deactivates the Streptococcus pyogenes Cas9 nuclease in a doxycycline-dependent manner. This enables defined, temporal control of Cas9 activity in any cell type and also in vivo. Results show that SiC may enable a reduction in off-target editing, with less effect on on-target editing rates. This tool facilitates diverse applications including (1) the timed regulation of genetic knockouts in hard-to-transfect cells using lentivirus, including human leukocytes for the identification of glycogenes regulating leukocyte-endothelial cell adhesion; (2) genome-wide lentiviral sgRNA (single guide RNA) library applications where Cas9 activity is ablated after allowing pre-determined editing times. Thus, stable knockout cell pools are created for functional screens; and (3) temporal control of Cas9-mediated editing of myeloid and lymphoid cells in vivo, both in mouse peripheral blood and bone marrow. Overall, SiC enables temporal control of gene editing and may be applied in diverse application including studies that aim to reduce off-target genome editing.

Keywords: CRISPR; Cas9; doxycycline; exome sequencing; genome editing; genome-wide screen; hematopoietic stem cells; lentivirus; next generation sequencing; off target editing.

Graphical abstract

Figure 1

Self-Inactivating CRISPR-Cas9 System, SiC (A) Lentiviral vector with two constructs: SiC-V1 and SiC-V2. SgRNA editing Cas9 are numbered G1–G10, along with their site of action (top). Experiment schematic is shown at the bottom. (B) sgRNA G1–G10 were cloned individually into SiC-V2, and transiently transfected into HEK-Cas9dTomato cells, which stably express Cas9 and dTomato. Following 72 h with and without 2.5 μg/mL Dox, the Surveyor assay was performed. Percent of gene editing was quantified using densitometry. “U” indicates untransfected (wild-type) HEK cells. Arrows indicate indels. sgRNA G7 in the Rec1 lobe was most efficient at editing Cas9. Data are representative of 2–4 repeats for each sgRNA. (C) HEK-Cas9dTomato cells were either transiently transfected (left half of gel) or stably transduced (right half) to express SiC-V2-Cas9G7 (SiC-V2 with guide G7). The latter stably transduced cells are called “HEK-Cas9dTomato-Cas9G7. Dox concentration was varied over 72 h prior to Surveyor assay to monitor Cas9 editing. Cas9 editing is strictly Dox-dependent upon lentiviral transduction. Data are representative of 3–4 repeats. (D–F) Time course studies of Cas9 activity performed using “HEK-Cas9dTomato-Cas9G7” cells. dTomato reporter fluorescence measured using cytometry decreased gradually over 10 days in the presence of 1 μg/mL Dox (D). Genome editing rate was independent of Dox concentration in the range tested (0.5–10 μg/mL, E) Microscopy image shows >90% loss of dTomato signal after 10 days of Dox (F). Data are representative of >10 repeats. See also Figures S1 and S2.

Figure 2

Time-Dependent On- and Off-Target Editing Assessed Using Exome and Deep Sequencing (A) HEK-Cas9dTomato-Cas9G7 cells were transduced with a pool of five lentivirus carrying promiscuous sgRNA targeting human EMX1, C1GalT1C1, MGAT1, ST3Gal4, or ZSCAN2. Cells were red due to Cas9dTomato, cyan due to Cas9G7 sgRNA, which was co-expressed with a Cerulean reporter, and blue due to BFP co-expression by the virus carrying the promiscuous sgRNA. 1 μg/mL Dox was added either 18 h prior to viral transduction, at day 0 (0 day) or at day 2 (2 day) in order to self-inactivate Cas9 (A). Dox was absent in one case, and control cells remained untransduced (without Dox). 1 μg/mL puromycin was added to select for transduced cells starting at day 1 (1 day). Genomic DNA was purified at day 15 for exome sequencing (results in C and D). Selected on- and off-target sites were also amplified from genomic DNA, and indels were quantified using deep sequencing (results in E). (B) Cas9 expression (dTomato, red), presence of sgRNA-library (BFP, blue), and Cas9G7 sgRNA (Cerulean, Cyan) expression were monitored at various times using surrogate fluorescence reporters. Note that some decrease in dTomato fluorescence is observed in No Dox sample at day 15 as the cells were cultured to over-confluence, but there is no Dox editing in these cells based on deep-sequencing. (C and D) In exome sequencing runs, data are presented as mean ± SD for indel formation on 3 on-target genes (C1GalT1C1, ST3Gal4, ZSCAN2) and 32 putative off-target genes. Among the 32, 30 are exonic off-target edits and 2 are intronic. Here, both on-target and off-target editing increased if Dox addition was delayed. Low levels of off-target editing (0.27%) and no on-target editing (0%) was detected in the no-virus control sample, and thus the study design contains only low levels of basal noise. In general, off-target editing was ∼10-fold lower compared to on-target (C). (D) On-target editing percent for three genes based on triplicate runs at each time point. Here, whereas some genes were edited at early times, indels accumulated in others over a period of days (D). (E) In deep-sequencing runs, also, on- and off-target editing increased upon delaying Dox addition. Here, “-T” denotes on-target and “-OT” denotes off-target editing. Samples at each time were analyzed in triplicate as noted using red symbols. Samples with lower duration of Dox treatment are indicated using darker red symbols and these appear to the right in the individual plots. Untransduced cells and cells with earlier Dox addition time points appear to the left. A blue line links the mean “%indel values” of each of the samples, starting with “No virus, No Dox,” to earlier Dox to later, No Dox treatment samples. Percent indel was higher in some runs (left panel) compared to others (right panel), with no editing seen in EMX1-OT10. See also Figure S3.

Figure 3

SiC Reduces Off-Target Editing (A) HEK293T cells stably expressing vector SiC-V2 with Cas9-G7 were transduced with SiC-V1 carrying previously established promiscuous sgRNA targeting either EMX1 or VEGFA. Both guides have well-established on-target (T) and off-target (OT) editing sites. 48 h post transduction, the cells were divided into two groups with one group receiving 1 μg/mL Dox for another 13 days. Genomic DNA was prepared from these four types of cells at the end point and also control untransduced HEK293T-Cas9G7cells cultured with Dox. (B) 6-FAM labeled PCR products generated for EMX1 and VEGFA, on-target and off-target sites, on day 15 were resolved using capillary electrophoresis. Vertical red line indicates unedited PCR fragment from control cells (top row). Numbers in individual panels indicate percent editing based on electrophoresis area-under-the-curve calculations. Off-target editing is reduced upon Dox addition (B, middle versus bottom row). Data are representative of duplicate runs.

Figure 4

ST3Gal4 Knockout in Human Leukocytes Abolishes Selectin-Dependent Leukocyte Rolling (A) HL-60s were stably transduced with SiC-V2-Cas9G7 virus to create Cerulean-positive cells. These were subsequently transduced with SiC-V1 virus carrying either scramble-sgRNA or ST3Gal4-sgRNA. Following sorting of dTomato⁺Cerulean⁺ cells at day 3, 1 μg/mL Dox was added from days 4 to 14. (B) The Surveyor assay monitored ST3Gal4 (top) and Cas9 (bottom) indels. (C) Fluorescence microscopy at day 14 measured dTomato reporter. Dox treatment resulted in Cas9 editing and loss of red fluorescence. (D) Flow cytometry measured cell surface sialyl Lewis-X expression (using mAb HECA-452), and also P-, E-, and L-selectin binding function (treatments same as C). Loss of ST3Gal4 activity reduced sLe^X expression and selectin binding (events in bottom-left quadrant of each sub-plot). (E) No Dox cells from (C) were perfused over selectin substrates in microfluidic flow cell at wall shear stress of 2 dynes/cm². Cells lacking ST3Gal4 displayed minimal interaction with selectin substrates in bright field. Controls used function blocking mAbs against P- (clone G1), E- (P2H3), and L- (DREG-56) selectin. See also Figure S4.

Figure 5

Use of Cas9G7 in Small- and Genome- Scale CRISPR Screens (A) SiC-dual-V2 contains two sgRNA, one of which targets Cas9 and other against a target gene. (B) HL-60s were transduced with SiC-V1-Scr virus to create Cas9dTomato cells (red). Titers were adjusted so that 30% of the cells were untransduced (dTomato⁻), as these serve as internal controls. These cells were then transduced with either Cerulean⁺ SiC-dual-V2-scramble or SiC-dual-V2-COSMC lentivirus to abolish O-glycan biosynthesis. 1 μg/mL Dox was added on either day 0, 1, 2, 3, or 4 to knockout Cas9. Dox was removed 2 days prior to cytometry analysis on day 9 in order to reduce Dox-induced autofluorescence. U, untransduced cells; -, transduced cells without Dox. (C) On day 9, cells cultured with Dox displayed decreased dTomato signal, indicating Cas9 inactivation (top panels). A portion of the SiC-dual-V2-COSMC transduced cells displayed VVA-FITC binding due to COSMC-knockout (arrow, bottom panel). (D) Dot plot comparing 1 μg/mL Dox treatment for scramble versus COSMC depict a population of COSMC edited cells at day 9, without residual Cas9/dTomato activity (blue arrow). (E) Cas9⁺ dTomato HL60s were transduced with genome-scale library with ∼90,000 sgRNA in a vector containing BFP and puromycin selection marker. 35% of the cells were BFP-positive corresponding to ∼1 sgRNA/cell (left panel, day 2). BFP⁺ cells were selected by addition of 1 μg/mL puromycin at day 2 (middle panel, day 4). Cas9G7 sgRNA was electroporated on day 4 to inactivate Cas9. A BFP⁺ dTomato⁻ population was observed on day 14 (blue arrow). Data in (A)–(D) are representative of three repeats.

Figure 6

SiC Vector Application in Mouse (A) HSPCs isolated from transgenic Cas9-EGFP mouse bone marrow were transduced with SiC-V2-Scr (control) or SiC-V2-Cas9G7 at ∼40%–45% efficiency (based on CFP signal). 1 μg/mL Dox was added to a portion of the cells on day 2. (B) Cytometry analysis at day 7 shows 25% Cas9 editing (i.e., low GFP signal) upon Dox treatment of SiC-V2-Cas9G7 transduced cells. (C and D) Donor Cas9-EGFP mouse HSPCs (CD45.2⁺) transduced with SiC-V2-Cas9G7 overnight were transplanted into recipient B6.SJL (CD45.1⁺) mice. Animals received standard (n = 5) or Dox chow (n = 6). (C) Flow cytometry analysis was performed on peripheral blood cells 1 to 3 weeks post-transplant. Representative dot plot of CD45.2⁺ CD11b⁺ peripheral blood cells 1 week post-transplant (left). Percent GFP⁻ CD45.2⁺ CD11b⁺ cells at different times in mice fed with standard or Dox chow (right). (D) Bone marrow was analyzed at 4 weeks. Representative dot plot showing appearance of GFP⁻ CD45.2⁺ CD11b⁺/Gr-1⁺ granulocytes upon Dox treatment (left). GFP⁻ CD45.2⁺ LSK SLAM, CD11b/Gr-1, and B220/CD3 cells in mice fed with standard or Dox chow (right). *p ≤ 0.05; **p ≤ 0.01 (Student’s unpaired, two-tailed t tests). See also Figure S5.