Quadruple-editing of MAPK and PI3K pathways effectively blocks the progression of KRAS-mutated colorectal cancer cells

Zaozao Wang (  zaozao83630@bjmu.edu.cn ) Beijing Cancer Hospital Bin Kang BGI-Shenzhen: BGI Group Qianqian Gao BGI-Shenzhen: BGI Group Lei Huang BGI-Shenzhen: BGI Group Yingcong Fan Beijing Cancer Hospital Jianhong Yu Beijing Cancer Hospital Jiabo Di Beijing Cancer Hospital Beihai Jiang Beijing Cancer Hospital Feng Gao BGI-Shenzhen: BGI Group Dan Wang BGI-Shenzhen: BGI Group Haixi Sun BGI-Shenzhen: BGI Group Ying Gu BGI-Shenzhen: BGI Group Jian Li Beijing Cancer Hospital Xiangqian Su Beijing Cancer Hospital

inhibition of MAPK and PI3K pathways is required for the complete inhibition of KRAS signaling and tumor progression. However, the overlapping toxicities limit the clinical activities of combined therapy with currently available inhibitors, such as MEK inhibitor (Pimasertib) and PI3K/MTOR inhibitor (Voxtalisib) [7], and MEK inhibitor (AZD6244) and AKT inhibitor (MK2206) [8]. Therefore, development of a novel strategy targeting MAPK and PI3K pathways is highly essential for the treatment of KRASmutated CRC cells.
CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) system has been reported as a potent strategy for e cient gene depletion in vitro and in vivo based on double-strand break (DSB) repair mechanism at target sites [9]. Moreover, multiplex genome engineering enables simultaneous editing of several sites within the mammalian genome by encoding multiple guide sequences into a single CRISPR array [10,11]. Recent studies demonstrated that therapeutic strategies based upon multiplex genome editing was powerful to ght against hematopoietic malignancies, which indicated its broad applications in cancer treatment [12,13]. Compared with the most popular enzyme SpCas9, SaCas9 is remarkably smaller in size and is more appropriate for in vivo editing, and has been previously used to rescue vision loss in patients with Leber congenital amaurosis [14,15]. Furthermore, selective targeting of oncogenic mutations of KRAS, such as G12V/D or G12S by CRISPR/Cas9, has been reported to inhibit the proliferation of tumor cells [16,17]. However, to date, no study has assessed the effects of multiple targeting of KRAS and downstream signaling pathways.
One bottleneck undermining the application of genome-editing techniques to treat cancer is the lack of an in vivo e cient and safe delivery method. The human adenovirus serotype 5 (ADV5) is widely used in gene therapy of cancer due to its high infection e ciency and high expression levels of therapeutic genes [18]. However, the current method of ADV5 administration is only local injection because of the nonspeci c tissue tropism and toxicity, especially in liver, under a systemic delivery. A previous research demonstrated that engineered proteins might be promising to overcome this challenge. For instance, an adaptor protein, consisting of a single-chain variable fragment (scFv) antibody against human epidermal growth factor receptor 2 (HER2), has shown to retarget ADV5 to the HER2-positive breast cancer cells and enhance the targeting speci city [19]. Besides, a scFv antibody against the hexon protein on ADV5 surface could attenuate its interaction with coagulation factor X (FX), and reduce its off-targeting tissue tropism as well [20].
In the present study, we constructed a CRISPR/SaCas9 system, simultaneously depleting four components of MAPK and PI3K pathways, such as KRAS, MEK1, PIK3CA, and MTOR. In addition, we developed two engineered proteins, an adaptor and a protector, to facilitate the intravenous delivery of CRISPR system in ADV5 vector to the CRC cells over-expressing epithelial cell adhesion molecule (EpCAM). The quadruple-depletion of MAPK and PI3K pathways effectively inhibited the progression of KRAS-mutated CRC cells in vivo, and might be a novel therapeutic strategy for CRC.

CRC tumor samples
In the current study, seven tumor tissues collected from patients with CRC who underwent radical surgery at Peking University Cancer Hospital & Institute (Beijing, China) were used to isolate primary tumor cells and establish PDX models. This study was approved by the Ethics Committee of Peking University Cancer Hospital & Institute (Approval No. 2015KT71), and all participants signed the written informed consent forms prior to commencing the study.

Vector construction
The lentiviral vector of gene depletion, Lenti-CMV::SaCas9-2A-GFP;U6::BsaI-sgRNA, was constructed by the Gateway recombination reaction between the donor vector containing synthetic SaCas9-2A-GFP and sgRNA expressing elements anked by attL sequences and the destination vector containing attR sequences, pLEX_305 (#41390; Addgene, Watertown, MA, USA). The most e cient sgRNA was chosen from Benchling tool (https://benchling.com), and inserted into the BsaI sites after annealing. The sequences of sgRNAs and matched PAMs were summarized in Additional File 1: Table S1. The scrambled gRNA non-existed in human and mouse genome was cloned into non-target control vector (NT) (5' GGCACTACCAGAGCTACTCA 3'). The multiplexed vectors were constructed through Golden Gate ligation of corresponding sgRNA cassettes. The plasmids expressing adaptor and protector proteins were respectively constructed by cloning the synthesized coding sequences of adaptor (the ectodomain of CXADR, a phage T4 britin polypeptide and a humanized anti-EpCAM/MUC1 scFv), and protector (a humanized anti-hexon scFv and a phage T4 britin polypeptide) into pENTER plasmids using restriction sites of Asis and Xho . The cloning primers were presented in Additional File 1: Table S2. The adenoviral vector of gene depletion was constructed by the Gateway recombination reaction between the donor vector containing synthetic SaCas9-2A-GFP and sgRNA expressing elements anked by attL sequences and the destination vector pAdeno-MCMV containing attR sequences. The diagrams of lentiviral and adenoviral vectors were shown in Additional File 2: Figure S1A-1B.
Streptomycin (1:500; Beijing Solarbio Science & Technology Co., Ltd., Beijing, China). After dissociation, the cell suspensions were ltered through a 70-μm cell strainer (BD Biosciences, Franklin Lakes, NJ, USA) and erythrocytes were removed using 2 ml Red Blood Cell Lysis Buffer (TBD Sciences Inc., Waltham, MA, USA). The isolated tumor cells were stored in liquid nitrogen for the testing of cell viability.
Cell viability, proliferation, and Transwell assays Fifty thousand primary tumor cells were infected with lentivirus for gene depletion, and the cell viability was assessed by using CellTiter-Glo® 2.0 kit (Promega, Madison, WI, USA) after 48 h according to the manufacturer's instructions. MEK inhibitor (AZD6244) and PI3K/MTOR inhibitor (BEZ235) were purchased from Selleck Chemicals LLC (Houston, TX, USA). The IC50 of AZD6244 or BEZ235 in CRC cells was calculated using GraphPad software. The cell proliferation was assessed using a commercial Cell Counting Kit-8 CCK-8 (Dojindo Molecular Technologies, Inc., Rockville, MD, USA) according to the manufacturer's instructions. Cell migration assay was performed using a Boyden chamber that contained a polycarbonate lter with an 8-μm pore size (Costar Inc., New York, NY, USA). Serum-free DMEM was added to the upper chamber, while complete medium containing 10% fetal bovine serum (FBS) was added to the lower chamber. After transfection, 2 × 10 5 HCT116 and SW620 cells were separately seeded into the upper chamber and incubated for 48 h. The migrated cells were stained with crystal violet, and ve random elds were captured by a microscope. Cell invasion was similarly detected except that the upper chamber was pre-coated with Matrigel (5 mg/ml; BD Biosciences, Franklin Lakes, NJ, USA). All experiments were performed in triplicate.
Western blotting analysis CRC cells were harvested 48 h after transfection of CRISPR vectors or MEK as well as PI3K/MTOR inhibitor treatments. Proteins were extracted from cells or tumor tissues using RIPA lysis buffer containing complete protease inhibitor cocktail (Roche, Basel, Switzerland). The extracts were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene uoride (PVDF) membranes. Membranes were blocked with 5% non-fat milk or 5% bovine serum albumin (BSA), followed by incubation with primary antibodies against the following antigens: membranes were washed three times with TBST, incubated with horseradish peroxidase (HRP)conjugated goat anti-mouse IgG or goat anti-rabbit IgG secondary antibodies for 1 h at room temperature, and then washed with TBST three times. The immunoreactive bands on the membranes were visualized using an enhanced chemiluminescence detection kit (Millipore, Burlington, MA, USA). The ImageJ software was used for image processing of band intensity, and relative protein expression levels were normalized to β-actin.

IHC
Formalin-xed para n-embedded CDX and PDX tumor tissues were cut into 4-μm sections. After depara nization and rehydration, slices were placed in a citrate buffer solution (pH=6.0) for antigen retrieval, and underwent inhibition and blockage of endogenous peroxidase activity. Sections were incubated with primary antibodies overnight at 4 ℃, followed by processing with a polymer horseradish peroxidase detection system (PV-9000, Zhongshan Goldenbridge Biotechnology). The primary antibodies of KRAS, MEK1, PIK3CA and MTOR were the same as Western blotting, and the primary antibody for Ki67 was purchased from Abcam (#ab16667). The expression levels were evaluated by H-score method. Scoring was independently reviewed by two experienced pathologists.

Hematoxylin and eosin (H&E) staining
Formalin-xed para n-embedded CDX and PDX tumor tissues of vital organs, including liver, lung, kidney, and colon were cut into 4-μm sections. Slices were further processed and stained with H&E. Histopathology was reviewed by an experienced pathologist.
Protein expression and puri cation The adaptor proteins EpCAM and MUC1, the truncated adaptor protein losing scFv fragment, and the protector protein with His tags were respectively expressed in HEK293T cells. The proteins were extracted using RIPA buffer, puri ed with Ni-NTA beads (Byeotime), and rinsed with phosphate-buffered saline (PBS) and Amicon Centrifugal Filters (10kDa NMWL, Millipore).

FACS analysis
In order to detect GFP expression, cells harvested were resuspended in 0.22 µm-ltered Dulbecco's phosphate-buffered saline (DPBS), and triplicate measurements from independent samples were analyzed with the CytExpert 2.3 software (Beckman Coulter, Brea, CA, USA).
In vivo genome editing The adenovirus was pre-incubated with the adaptor and protector proteins (1.0×10 -7 pmol) at room temperature for 2 h, and injected into the tail veins of CDX and PDX mouse models for two rounds with a 12-day interval. The number of viral particles was 7.0×10 9 . The tumor size and weight were measured every 3 days.

Statistical analysis
Statistical analysis was carried out respectively by two-tailed Student's t-test and two-way analysis of variance (ANOVA) using SPSS 20.0 (IBM, Armonk, NY, USA) and GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA) software. The Kolmogorov-Smirnov test was employed to indicate whether samples were normally distributed. The statistical signi cance was presented as * 0.01≤P 0.05, ** 0.001≤P 0.01, *** P 0.001.

Results
Single and multiple-editing vectors were constructed to target MAPK and PI3K pathways In order to obtain an effective dual-inhibition of MAPK and PI3K pathways, four important components of MAPK and PI3K pathways (KRAS, MEK1, PIK3CA, and MTOR) were selected for depletion ( Figure 1A). The most e cient single guide RNAs (sgRNAs) of CRISPR/SaCas9 system for these four genes were selected by using Benchling tool (https://benchling.com). Moreover, the vectors required for single and multiple depletions were respectively constructed through Golden Gate ligation of corresponding sgRNA cassettes.
In addition to single-editing vectors, the multiple-editing vectors included double-depletion of KRAS and MEK1 (KM-KO), and PIK3CA and MTOR (PM-KO), as well as quadruple-depletion of the four genes (KMPM-KO) were constructed ( Figure 1B). According to the results of T7E1 assay of HEK293T cells, the vectors of selected single gRNAs (labeled as #1 in Figure 1C-1F) could edit the corresponding gene sites more e ciently compared with those carrying non-overlapping gRNAs targeting the same genes (labeled as #2 in Additional File 3: Figure S2). Moreover, all the vectors with double and quadruple gRNAs could e ciently edit the corresponding targets, with e ciencies ranging from 44.7% to 56.6% ( Figure 1G-1I).
The expression levels of four gRNAs and the corresponding mutation statuses of KRAS, MEK1, PIK3CA, and MTOR were respectively con rmed by quantitative reverse transcription-polymerase chain reaction (RT-qPCR) and Sanger sequencing in HEK293T cells with quadruple-depletion (KMPM-KO). The abundances of the four gRNAs were almost comparable at the mRNA level, and those could e ciently induce on-target mutations of the corresponding genes (Additional File 4: Figure S3 and Figure 1J). Therefore, the single-and multiple-editing vectors targeting MAPK and PI3K pathways were successfully constructed, and they could be used to investigate the editing effects of dual-pathway on CRC cells.  Figure 2D). Furthermore, quadruple-depletion of the four target genes (KMPM-KO) signi cantly down-regulated the phosphorylation levels of ERK, AKT, and S6K, which were similar to the effects of combining treatment of AZD6244 and BEZ235 ( Figure 2E).

Quadruple
These results revealed that inhibitory effect of gene editing on the corresponding signaling pathway was speci c. The inhibitory effects of double-editing were more noticeable than those of single-editing when a pathway was targeted. Quadruple-editing of KRAS, MEK1, PIK3CA, and MTOR could e ciently and speci cally inactivate both MAPK and PI3K pathways. Correspondingly, KMPM-KO also signi cantly inhibited the proliferation, migration, and invasion of HCT116 cells, with a greater inhibitory effect than single-and double-depletion ( Figure 2F-2G). Therefore, quadruple-editing of MAPK and PI3K pathways effectively inhibited the malignant phenotypes of CRC cells with oncogenic mutations of KRAS and PIK3CA.
Quadruple-editing of KRAS, MEK1, PIK3CA, and MTOR effectively blocked the compensated PI3K activation in KRAS-mutated CRC cells with MAPK suppression A previous research showed that RTK-dependent activation of PI3K pathway was a resistant mechanism against the suppression of MAPK pathway in CRC cells with KRAS mutation and wild-type PIK3CA ( Figure  1A) [5]. In order to indicate whether the gene editing of MAPK pathway can induce compensated PI3K activation, we rst determined the IC50 value of AZD6244 in SW620 cells with KRAS G12V mutation and wild-type PIK3CA. The IC50 value of AZD6244 in SW620 cells was 1.0 mM ( Figure 3A). Then, the activation status of MAPK and PI3K pathways in SW620 cells with AZD6244 treatment and single-and double-depletion of MAPK pathway (KRAS-KO, MEK1-KO, and KM-KO) was detected by Western blotting. As shown in Figure 3B and 3C, similar to AZD6244 treatment, the single-and double-depletion signi cantly reduced the phosphorylation level of ERK and enhanced the phosphorylation level of AKT.
The up-regulation of p-AKT by KM-KO was more noticeable than that by KRAS-KO and MEK1-KO ( Figure  3C). The above-mentioned ndings suggested that the gene editing of MAPK pathway also induced compensated PI3K activation in KRAS-mutated CRC cells. In contrast, single-and double-depletion of PIK3CA and MTOR (PIK3CA-KO, MTOR-KO, and PM-KO) remarkably down-regulated the expression levels of p-AKT and p-S6K, while did not affect the expression level of p-ERK. The down-regulated expression levels of p-AKT and p-S6K by PM-KO were more notable than those by PIK3CA-KO and MTOR-KO ( Figure  3D). Furthermore, quadruple-depletion of the four target genes (KMPM-KO) signi cantly down-regulated the expression levels of p-ERK, p-AKT and p-S6K ( Figure 3E). Thus, quadruple-editing of KRAS, MEK1, PIK3CA, and MTOR could e ciently block the compensated activation of PI3K pathway under MAPK suppression. Correspondingly, KMPM-KO also markedly inhibited the proliferation, migration, and invasion of SW620 cells, with a greater inhibitory effect than single-and double-depletion of either MAPK or PI3K pathway ( Figure 3F-3G). Therefore, compared with only suppressing MAPK pathway, quadrupleediting of MAPK and PI3K pathways could enhance the anti-tumor effects on KRAS-mutated CRC cells through inhibiting the compensated PI3K activation.
In  Table S3. These tumor cells were respectively infected with the lentivirus of quadruple-editing, MEK inhibitor (AZD6244), and PI3K/MTOR inhibitor (BEZ235). The inhibition rates of cell survival under various treatments compared with non-treated controls were presented in Figure 4A-4C and 4E-4H. In all the seven cases, quadruple-editing led to a more remarkable inhibition of cell survival than a single-treatment with AZD6244 or BEZ235, and showed similar or more signi cant inhibitory effects to those of combined therapies. Furthermore, Western blotting results indicated that quadruple-editing of KRAS, MEK1, PIK3CA, and MTOR could e ciently inactivate MAPK and PI3K pathways in primary CRC cells carrying KRAS/PIK3CA double mutations (CRC-P01) or KRAS single mutation (CRC-PDX01). The inhibitory effects of quadruple-editing were more noticeable than those of single-treatment with AZD6244 and BEZ235, and were similar or a little superior to those of combined therapies ( Figure 4D and 4I). The above-mentioned results suggested that quadruple-editing had prevalent anti-tumor effects on KRAS-mutated CRC cells.
The complex combining ADV5 and engineered proteins intravenously delivered CRISPR system to CRC with high e ciency and speci city The ADV5 is extensively utilized in gene therapy of cancer. However, the current administration of ADV5 is only local injection because of its off-target tissue tropism under a systemic delivery. Thus, two proteins, an adaptor and a protector, were engineered to make ADV5 as a proper vector for the intravenous delivery of CRISPR system to CRC. ADV5 infected host cells through a high-a nity interaction between the knob domain of the viral ber proteins and coxsackievirus and adenovirus receptor (CXADR) displayed on the target cell surface [22]. Therefore, the adaptor protein was composed of the ectodomain of ADV receptor CXADR of cells (ECXADR), and a humanized single-chain variable fragment (scFv) recognizing a certain surface marker on CRC cells, which could be linked by a phage T4 britin polypeptide. Adaptor protein could interact with the knob protein of ADV5 ber through ECXADR domain, and retarget ADV5 to CRC cells through scFv ( Figure 5A) [19]. In addition, a scFv antibody against the hexon protein on ADV5 surface was fused to a phage T4 britin polypeptide to form a protector protein, which could cover ADV5 to reduce the tissue off-targeting ( Figure 5A) [20]. The britin induced the trimerization of adaptor and protector proteins, and up-regulated their a nities with ADV5 ( Figure 5A) [19,20].
EpCAM is a transmembrane glycoprotein mediating Ca 2+ -independent homotypic cell-cell adhesion in epithelia. It has an extremely high rate of over-expression (≥80%) in CRC cells [23,24]. The results of RT-qPCR showed that compared with 293T cells, EpCAM was over-expressed in 4 human CRC cell lines, including HCT116, LOVO, SW480, and SW620, and in seven types of CRC cells isolated from primary and PDX tumors ( Figure 5B-5C). In contrast, Mucin 1, cell surface associated (MUC1), showed lower expression levels in CRC cell lines and primary tumor cells compared with EpCAM ( Figure 5B-5C).
Therefore, an adaptor protein targeting EpCAM was constructed to enhance the infection e ciency of ADV5 to CRC cells, and a MUC1 adaptor was also developed as a control to con rm the function of the adaptor protein.
In order to validate the function of adaptor protein in vitro, the green uorescent protein (GFP)-expressing ADV5 was pre-incubated with various amounts of EpCAM adaptor protein, or a truncated adaptor protein losing an anti-EpCAM scFv fragment and remaining only ECXADR domain as well as britin, and then, infected SW620 cells. The uorescence-activated cell sorting (FACS) analysis revealed that EpCAM adaptor protein gradually up-regulated the infection e ciency of ADV5 in SW620 cells when increasing amount of protein was pre-incubated with ADV5. However, the truncated adaptor protein gradually downregulated the infection e ciency of ADV5 in SW620 cells when the amount of protein increased (Additional File 5: Figure S4). The results indicated that EpCAM adaptor protein blocked the interaction between ADV5 and its receptor CXADR through ECXADR domain, and retargeted the virus to EpCAM protein on CRC cells by anti-EpCAM scFv. Afterwards, the GFP-expressing ADV5 was respectively preincubated with various amounts of EpCAM and MUC1 adaptor proteins, and infected SW620 cells. As displayed in Figure 5D, these two adaptor proteins could dramatically enhance the infection e ciency of ADV5 in SW620 cells. The infection e ciency up-regulated by EpCAM adaptor was more remarkable than that by MUC1 adaptor, which could be correlated with the difference of their expression levels in SW620 cells ( Figure 5B). The results indicated that the infection e ciency of the complex combining ADV5 and adaptor depended on expression level of a cell surface marker targeted by adaptor. Since EpCAM was extensively over-expressed in CRC cells, EpCAM adaptor might up-regulate the infection e ciency of ADV5 in the majority of CRC cases.
Furthermore, the functions of engineered proteins were validated in vivo. The ADV5 expressing SaCas9 (Additional File 2: Figure S1B) was intratumorally or intravenously injected into nude mice with SW620 cells-derived xenografts (SW620 CDX model), individually or via combination of EpCAM adaptor and protector proteins. According to the expression levels of SaCas9 in tissues collected 48 h after administration, the intratumorally delivered ADV5 was mainly enriched in tumor tissue. The intravenously delivered ADV5 had diverse tissue tropisms, especially towards liver. Adaptor protein signi cantly enhanced the tumor tropism of ADV5, and reduced the off-target tropisms towards the majority of organs except for colon where EpCAM was also over-expressed ( Figure 5E). The addition of protector protein signi cantly reduced the off-targeting tropisms in a variety of organs including colon, and enhanced the tumor tropism compared with adaptor only ( Figure 5E). Therefore, the combination of adaptor and protector could retarget ADV5 to CRC cells over-expressing EpCAM, and reduce its tissue off-targeting in vivo. The complex combining ADV5 and the two engineered proteins might be a proper vector to deliver CRISPR system intravenously.  Figure  6B, 6G). The expression level of SaCas9 in the tumor tissues of CRC-PDX01 mice was higher than that in HCT116-CDX mice, which might result from the higher expression level of EpCAM in CRC-PDX01 than that in HCT116 cells ( Figure 5B-5C, 6B, 6G). The tumor growth in CDX and PDX mice that received KMPM-KO was signi cantly blocked compared with those that were given NT ( Figure 6C, 6H). Besides, the nal tumor volume in CDX and PDX mice that received KMPM-KO was markedly reduced compared with those that were given NT ( Figure 6D, 6I). Furthermore, H&E staining of different tissues in CDX and PDX mice showed that the tumor tissues injected with KMPM-KO had a more signi cant cell necrosis compared with those that received NT ( Figure 6E, 6J). These results strongly suggested that quadruple-editing of MAPK and PI3K pathways blocked the progression of KRAS-mutated CRC cells in vivo. In addition, no injuries were observed by H&E staining in liver, lung, kidney, and colon of CDX and PDX mice that were given NT and KMPM-KO ( Figure 6E, 6J). No signi cant difference was noted in body weight of CDX and PDX mice that received NT and KMPM-KO APCs ( Figure 6F, 6K). Therefore, the EpCAM-targeting APC could be a speci c and safe vector for intravenous delivery of CRISPR system to CRC.

Quadruple-editing blocked MAPK and PI3K pathways in KRAS-mutated CRC cells in vivo
The results of T7E1 assay in the tumor tissues of CDX and PDX mice injected with NT and KMPM-KO showed that compared with NT samples, the genomic regions of the four target genes, including KRAS, MEK1, PIK3CA, and MTOR, were all e ciently edited in KMPM-KO samples ( Figure 7A, 7D). According to the results of immunohistochemistry (IHC) and Western blotting, the expression levels of the four target genes were signi cantly down-regulated in KMPM-KO samples compared with those in NT samples ( Figure 7B-7C, 7E-7F). The Ki67 staining indicated that the proliferation of tumor cells after quadrupledepletion was notably down-regulated compared with that in control cells ( Figure 7B, 7E). Furthermore, the phosphorylation levels of components of MAPK and PI3K pathways, such as ERK, AKT, and S6K, were remarkably down-regulated in KMPM-KO samples compared with those in NT samples ( Figure 7C, 7F).
The above-mentioned ndings indicated that quadruple-editing of the four target genes mediated by EpCAM-targeting APC e ciently blocked MAPK and PI3K pathways in KRAS-mutated CRC cells in vivo.

The status of on-target and off-target mutations in tumor tissues induced by quadruple-editing
Whole-exome sequencing (WES) was performed in the tumor tissues of HCT116 CDX mice that received NT and KMPM-KO to detect the mutation status induced by quadruple-editing. The sequencing depth was over 100×. Totally, seven types of on-target deletions were detected at genomic regions of the four target genes (KRAS, MEK1, PIK3CA, and MTOR). The total mutation frequencies of the four target genes were 34.9%, 80%, 40.4%, and 73.3%, respectively ( Figure 8A). Among 167 potential off-target loci of these genes predicted by Benchling tool, only one locus of MTOR was detected with a real off-target singlenucleotide variant (SNV) (Chr16, 28603692 C>T), whose mutation frequency was 1.68% ( Figure 8B). Furthermore, the total number of variants detected in NT and KMPM-KO samples was almost equal (24420 versus 24539). The number of insertion/deletions (INDELs) and single nucleotide variations (SNVs) detected at different genomic regions was also similar in NT and KMPM-KO samples ( Figure 8C-8D). Additionally, the majority of variants detected at different chromosomes were common ones shared by NT and KMPM-KO samples ( Figure 8E). These results strongly suggested the minor effect of quadruple-editing on the mutation status of CRC.
Altogether, the quadruple-depletion of KRAS, MEK1, PIK3CA, and MTOR intravenously delivered by the EpCAM-targeting APC blocked MAPK and PI3K pathways and the progression of KRAS-mutated CRC cells with high e ciency and speci city.

Discussion
The present study developed a quadruple-editing system for KRAS, MEK1, PIK3CA, and MTOR based on CRISPR/SaCas9, which could e ciently inhibit MAPK and PI3K pathways and signi cantly block the proliferation and invasion of CRC cells with diverse KRAS mutations or oncogenic mutations of KRAS and PIK3CA (Additional File 6: Figure S5). We also constructed two proteins, an adaptor and a protector, to form a complex with ADV5 and facilitate the intravenous delivery of CRISPR system to CRC cells overexpressing EpCAM. The quadruple-editing of MAPK and PI3K pathways delivered by the APC blocked the tumor progression of KRAS-mutated CDX and PDX models with high e cacy and speci city. As EpCAM showed to have high expression levels in CRC cases [24], our quadruple-editing system could cover the majority of KRAS-mutated CRC cells, highlighting its clinical signi cance for CRC treatment.
Although recent development of inhibitors against KRAS G12C and MEK had shed light on the treatment of a number of KRAS-mutated CRC cases, the diversity of KRAS variants and aberrant activation of PI3K pathway are still critical bottlenecks for the therapy of the majority of CRC patients. There are over ten types of oncogenic mutations in codon 12/13 of KRAS, such as G12D/V/C/S/A/R and G13D/C, and the majority of them have no promising inhibitors [25]. PI3K pathway could be activated by RTKs, compromising the inhibitory effect of MEK. The overlapping toxicities limit the clinical signi cance of combined therapies with inhibitors of MAPK and PI3K pathways [7,8].   [29][30][31][32].
Patients received combined regimens including BRAF and MEK inhibitors presented a prolonged overall survival than the control group in a BRAF V600E mutated CRC clinical trial [33]. For PI3K pathway, dual inhibitors of PI3K/MTOR displayed a stronger effect than MTOR inhibitor alone in not only hematological tumors but also solid tumors [34][35][36]. Furthermore, based on our data, simultaneously depleting KRAS/MEK1 and PIK3CA/MTOR respectively inhibited the signaling of MAPK and PI3K, more signi cantly than individual depletions, revealed by the extents of down-regulation of phosphorylated ERK and S6K ( Figure 2C-2D, 3C-3D). Depleting the signaling intermediates together provided better antitumor effects than depleting KRAS and PIK3CA only ( Figure 2F-2G, 3F-3G). Therefore, targeting two genes in one pathway is a better strategy than single targeting.
Here, we constructed an APC through combining ADV with an adaptor and a protector to overcome major challenges limiting systemic delivery of ADV. This two-protein construction e ciently retargeted ADV from its common receptor (CXADR) to EpCAM, a membrane protein over-expressing in CRC cells, and signi cantly reduced the tissue off-targeting of ADV, especially the liver tropism with over 10000 folds, which was better than the performance of some polymers such as polyethylene glycol (PEG) reported previously [20]. The proteomic studies had revealed protein markers in different types of cancer, including gastric cancer and pancreatic cancer. The EpCAM-targeting module of an adaptor protein was found exchangeable and could be replaced with scFv against membrane markers in gastric and pancreatic cancer cells, e.g. proteins of MUCIN family [37,38], accordingly forming speci c APC for gene therapies in these types of cancer.

Conclusions
In summary, CRISPR-mediated quadruple-depletion of KRAS, MEK1, PIK3CA, and MTOR intravenously delivered by ADV-protein complex could be a promising therapeutic option for KRAS-mutated CRC cells. In the future, the adaptor and sgRNA libraries will be established, and the genome editing therapy will have extensive applications in different types of cancer.

Declarations
Ethics approval and consent to participate

Availability of data and materials
The sequencing data reported in this study are available in the CNGB Nucleotide Sequence Archive. https://db.cngb.org/cnsa. Accession number CNP0000324.

Figure 5
Page 28/33 The complex combining ADV5 and engineered proteins intravenously delivered CRISPR system to CRC with high e ciency and speci city. (A). A diagram of ADV-protein complex delivering CRISPR system. The  proteins were normalized to that of -Actin, and presented below the blotting results. Figure 8