Cell-cell contact-induced gene editing/activation in mammalian cells using a synNotch-CRISPR/Cas9 system

The CRISPR system has been widely used for genome manipulation in various cells, tissues and whole organisms. Although an increasing variety of inducible CRISPR systems have been exploited for a variety of applications, such as chemical switch (Zetsche et al., 2015), photo switch(Shao et al., 2018) and solution ligand switch (Baeumler et al., 2017; Kipniss et al., 2017; Schwarz et al., 2017) systems, a cell-cell interaction inducible system is absent. The synthetic Notch (synNotch) receptor is a recently developed cell-cell contact sensing platform, which contains a customized extracellular sensor module, a transmembrane core domain of native Notch, and a customized intracellular responder module (Morsut et al., 2016). Because of its extraordinary flexibility in terms of customizable sensing/response behaviors, the synNotch receptor serves as a powerful tool for cell engineering (Roybal et al., 2016a; Roybal et al., 2016b; He et al., 2017). In the current study, we combined the synNotch receptor with the CRISPR/Cas9 system to develop a cell-cell interaction inducible gene regulated tool. Since the previously reported synNotch receptor was based on mouse Notch1 (M1) (Morsut et al., 2016), we tried to develop other synNotch receptors using different Notch family members from several species, including human (H), mouse (M), drosophila (Fly) and zebrafish (Z), with antiCD19-ScFv/mCherry as a sensor/responder module (Fig. 1A). We found that the new synNotch receptors demonstrated better activation than the M1 synNotch (especially for Z3, 78.0% ± 9.8% of cells were activated with a 158.4 ± 19.7-fold change), while showing a background ranging from 0.5% to 45% (Figs. 1B, 1C and S1A). M4 system was not activated when treated with CD19+ Cells (Figs. 1B and S1A), which was consistent with the previous report that Notch4 does not signal in response to ligand but inhibits signaling from the Notch1 receptor (James et al., 2014). To decrease the background noise, P2A-Gal4KRAB or P2A-Gal4 was added downstream of Gal4-VP64 (Fig. S1B). Gal4KRAB completely blocked activation (Fig. S1C), and Gal4 dramatically decreased the background noise (M2, ∼10%; all the others, <5%) but simultaneously attenuated activation significantly (all, <15%) (Fig. S1D). It has been reported that EGF (epidermal growth factor) repeats can prevent the constitutive activation of Notch (Sakamoto et al., 2005). Therefore, we included an extra EGF repeat on the extracellular domain between the antiCD19 ScFv and the Notch core domain (Fig. S2A). By including an extra EGF, the background in the Z1, Z2, and Z3 systems was decreased, and nearly eliminated in the Fly system (Fig. S2B). However, it also affected activation, as represented by the Z3 system with an efficiency of decreasing to approximately 41.2% when stimulated (Fig. S2B and S2C). To balance background and efficiency, we shortened the EGF repeat by half (eZ3), which remarkably increased the stimulation efficiency to 62.8% while maintaining a tolerable background (8.5%) (Fig. S2B and S2C). Next, we tested the ability of the synNotch receptors to respond to different sender cells and the response flexibility of the receiver cells. In addition to K562-CD19 cells, CD19transduced B16 melanoma cells (B16-CD19) with either high (B16-10) or low (B16-3) CD19 expression could activate mCherry expression in the M1 and H1 systems (Fig. S3A and S3B). However, mouse splenocytes expressing mouse CD19 failed to activate the M1 and H1 receptors, demonstrating the specificity of the synNotch systems (Fig. S3C and S3D). To test the flexibility of the response modules in the synNotch systems, we replaced the mCherry reporter with EGFP and SIRPɑ-Fc in the H1, Z3 and eZ3 systems (Fig. 1A). FACS and fluorescence microscopy analyses showed that EGFP was potently activated by the CD19+ cells (Figs. 1D, 1E and S4). High-affinity SIRPɑ-Fc (CV1hIgG4) holds great therapeutic potential and has entered a phase I clinical trial for solid tumor treatment (Weiskopf et al., 2013). SIRPɑ-Fc also could be secreted by the receiver cells triggered by CD19+ sender cells (Fig. 1F). To develop a cell-cell contact-induced gene editing system, we swapped the Gal4-VP64 domain into Cas9 (SpCas9). In HEK293T-dEGFP cells that stably express a fast-degradable EGFP variant, transfected synNotch-Cas9 and EGFP-targeting sgRNA could down-regulate EGFP even without stimulation by CD19+ cells, suggesting a

. All sgRNA and the Linker 7 and 8 sequences are listed in Table S2. All constructs were verified through Sanger sequencing.

Cell culture
The K562 and B16 sender cells were transduced to stably express human CD19 and a BFP reporter. The U2OS receiver cells were first transduced with the Gal4UAS-IRES-mCherry-pGK-BFP response element, then a single clone was infected with lentiviruses encoding synNotch receptors. For EGFP or SIRPɑ-Fc protein activation, the MCF7 receiver cells were transduced with both the Gal4UAS-EGFP or Gal4UAS-SIRPɑ response element and the synNotch receptor antiCD19-Notch-Gal4-VP64 lentivirus in a pool mix. In the synNotch-Cas9:p300 system, the Jurkat receiver cells underwent knock-in with the Gal4UAS/TRE-Cas9:p300 response element via HITI (Suzuki et al., 2016;Ma et al., 2018) and were then transduced with H1/Z3/Tet-eZ3 synNotch receptors to establish stable cell lines. The genotyping primers to identify the correct knock-in clones are listed in Table S3.

Lentivirus production
All lentiviruses were produced by co-transfecting the transfer plasmids, and package and envelope plasmids (pMD2.G and psPAX2) using the PEI reagent (Sigma, USA) in HEK293-FT cells plated in 6-well plates at approximately 75% confluency. Viral supernatants were collected 3 days after transfection and then filtered using a 0.45 μm filter (Millipore, USA). The supernatant was used for transduction immediately or kept at -80°C for long-term storage.
Activation of synNotch and synNotch-Cas9:p300 systems For synNotch system activation, 3×10 5 sender cells were co-cultured with receiver cells which were plated in 24-well plates a day before at a 1:1 ratio.
After approximately 48 h, the receiver cells were analyzed using flow cytometry (BD, USA). Fluorescence images were obtained using a confocal microscope (Nikon, Japan). For the synNotch-Cas9:p300 system, 3×10 5 B16 sender cells were plated in 24-well plates, and the next day, after electroporation with the targeting sgRNAs using the Amaxa Cell Line Nucleofector Kit V (VCA-1003, LONZA, Switzerland) in accordance with the manufacturer's instructions (2D, Program X-001), the Jurkat receiver cells were added to the B16 or B16-CD19 for approximately 2 d. Finally, aliquots of suspension of Jurkat cells were used for western blotting, T7E1, and q-PCR analyses.

T7E1 and TIDE analyses
For T7E1 analysis, the amplicons were purified, denatured at 95°C for 5 min and annealed in NEB Buffer 2 with a slow ramp down (approximately −2°C/min) to 4°C, then subjected to T7 endonuclease I (NEB, UK) digestion for 3 h at 37°C before loading on a 2% agarose gel. The primers for T7E1 to amplify the amplicons were listed in Table S3.
The Tracking of Indels by Decomposition (TIDE) method (Brinkman et al., 2014) was applied for analyzing the editing indels and determining their frequencies in a cell population using Sanger sequences of the PCR amplicons from control samples and testing samples. All the parameters were set to the default and the TIDE analysis tool is available online (http://tide.nki.nl). The primers for TIDE were as the same as that used for T7E1.

Quantitative real-time PCR
Total RNA from the receiver cells was isolated using Trizol Reagent (Thermo Fisher, USA) in accordance with the manufacturer's instructions. Total RNA (1 μg) was reverse transcribed into cDNA and then quantitative real-time PCR (SYBR Premix Ex Taq II, TAKARA, China) was performed using a LightCycler 96 System (Roche, Switzerland). Relative gene expression was calculated using the 2 -ΔΔCt method after normalizing to GAPDH expression. All the qPCR primers are listed in Table S3.

Western blotting
To detect the SIRPɑ-Fc secretory protein, sender cells and receiver cells were co-cultured in DMEM without FBS for approximately 2 d. Then the supernatant media was subjected to 5× SDS loading buffer without reducing agents and boiled for 5 minutes. The lysates were resolved through SDS/PAGE and transferred onto a nitrocellulose membrane which was blocked using 5% non-fat milk, and sequentially incubated with HRP-conjugated goat anti-human IgG secondary antibody (Bioss, China). The probed proteins were finally detected through chemiluminescence in accordance with the manufacturer's instructions (Pierce, USA).
To detect the Cas9:p300 protein, the co-cultured receiver cells were lysed in 2× SDS loading buffer and boiled for 10 min. The lysates were resolved through SDS/PAGE and transferred onto a nitrocellulose membrane which was blocked using 5% non-fat milk and sequentially incubated with primary antibodies (anti-flag, sigma, USA; anti-b-tubulin, Proteintech, China) and a HRP-conjugated horse anti-mouse IgG secondary antibody (CST, USA). All the probed proteins were finally detected through chemiluminescence in accordance with the manufacturer's instructions (Pierce, USA).

FACS analyses
All flow cytometry analyses were performed using FlowJo software (TreeStar, USA). To compare the activation level of the synNotch systems, sender cells were gated out using BFP, and the receiver cells were gated using mCherry or GFP. The activation efficiency was determined as the proportion of mCherry or GFP positive cells within the total receiver cells, and the activation fold change was the ratio of the mean fluorescence intensity between positive and negative background cells. Figure S1. Decreasing the basal activation of synNotch systems by adding Gal4KRAB or Gal4.    Figure S4. Activation of EGFP with H1, Z3 and eZ3 systems.

TIDE analyses (Tet-eZ3)
A B Figure S7. The editing efficiencies of PD-1 and CTLA4 in the synNotch-Cas9 systems using TIDE assay.         showed that the MYOD gene was activated with the synNotch-Tet-Cas9:p300 system (eZ3). One representative from three repeated experiments are shown.