In vivo CRISPR screening protocol to identify metastasis mediators using iteratively selected mouse models

Summary

In vivo CRISPR screens uncover metastasis genes in native contexts, surpassing in vitro model limitations. Here, we present a protocol to identify metastasis-driving genes in ovarian cancer using an in vivo CRISPR screening technique. Key steps include single-guide RNA (sgRNA) library design and validation, lentiviral transduction, establishment of metastatic mouse models, tissue collection, sgRNA amplification for sequencing, bioinformatics-based candidate gene identification, and functional validation.

For complete details on the use and execution of this protocol, please refer to Wang et al.¹

Graphical abstract

Highlights

• •Detailed protocol for in vivo CRISPR screens to identify metastasis genes
• •Combines library design, in vivo screening, sequencing, and computational analysis
• •Examples for target identification and functional validation

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

In vivo CRISPR screens uncover metastasis genes in native contexts, surpassing in vitro model limitations. Here, we present a protocol to identify metastasis-driving genes in ovarian cancer using an in vivo CRISPR screening technique. Key steps include single-guide RNA (sgRNA) library design and validation, lentiviral transduction, establishment of metastatic mouse models, tissue collection, sgRNA amplification for sequencing, bioinformatics-based candidate gene identification, and functional validation.

Before you begin

While single-cell RNA-seq and spatial omics have advanced our understanding of cellular heterogeneity and tissue organization in disease, these methods remain largely descriptive. In contrast, high-throughput genetic screens directly connect gene function to phenotypic outcomes within pathologically relevant tissue contexts. In vivo CRISPR screens represent a powerful approach to identify genes critical for tumor progression, immune evasion, drug response, and metastasis, offering clinically actionable insights within living systems. However, technical challenges, including clonal heterogeneity, cell isolation difficulties, and recovery limitations, make in vivo CRISPR screening particularly demanding, especially for studying cross-organ metastasis.

This protocol provides a detailed description of a robust in vivo CRISPR screening system to identify genes essential for cancer metastasis. Essential procedures for this protocol include custom sgRNA library design and construction, in vivo CRISPR screen using an ovarian cancer distant metastasis mouse model, computational analysis for screening results, and functional validation of candidate genes (Figure 1). While we demonstrate this approach using a curated sgRNA library (derived from our prior in vivo screens) and an optimized ovarian cancer multi-route metastasis model, the protocol is adaptable to other custom libraries and metastasis models.

Figure 1.

Experimental design and timeline for in vivo CRISPR screening in a metastasis model, related to before you begin

The iteratively selected ovarian cancer metastasis model used in this protocol was established by intraperitoneal injection of ES-2-MC2-Hep cells, which exhibit high metastatic potential in both the liver and lung. These cells were isolated from the metastatic livers of nude mice previously injected intraperitoneally with ES-2-MC1-Hep cells, which, in turn, were derived from the metastatic livers of nude mice injected with ES-2 cells (Figure 2).

Figure 2.

Development of a high-efficiency ovarian cancer metastasis model using an iterative selection strategy, related to before you begin

This figure was reused from Wang et al. (2025).

(A) Schematic overview of the iterative in vivo selection strategy for isolating ES-2-derived distal metastatic variants in mice.

(B) Representative pseudocolor ex vivo BLI depicting liver and lung metastases in ES-2-MC2-Hep xenograft models.

(C) Bar graph depicting the metastatic penetrance of ES-2-MC2-Hep cells in the lung and liver, respectively. n = 13 from two independent batches of experiments.

Institutional permissions

All animal experiments were performed in compliance with ethical guidelines and approved by the Institutional Animal Care and Use Committee (IACUC) of Westlake University. Researchers must obtain appropriate institutional approval before implementing this protocol.

Preparation of STE (sodium-Tris-EDTA) buffer

To maximize genomic DNA (gDNA) extraction efficiency and prevent gDNA loss, an in-house high-salt precipitation method was employed for gDNA extraction from cells.¹ In this method, the STE buffer is used to facilitate thorough cell lysis while ensuring DNA integrity. Prepare STE buffer beforehand using the composition below and store STE buffer at 25°C.

Reagent	Final concentration	Amount
NaCl (1 M)	0.1 M	10 mL
Tris-HCl (pH=8.0, 1 M)	0.01 M	1 mL
EDTA (pH=8.0, 0.5 M)	1 mM	200 μL
ddH2O	N/A	88.8 mL
Total	N/A	100 mL

Software

Timing: 30 min

• 1.Install a conda virtual environment.

Check if conda is installed and available. If it does not return the installed version of conda, follow the installation guide at https://www.anaconda.com/ to install anaconda.

$ conda --version

• 2.Install and activate MAGeCK using conda.

$ conda install -c bioconda -c conda-forge mageck

$ source activate mageckenv

• 3.Install R and R packages MAGeCK-Flute, clusterprofiler, ggplot2.

$ R

> if (!requireNamespace("BiocManager", quietly = TRUE)

install.packages("BiocManager")

> BiocManager::install("MAGeCKFlute")

> BiocManager::install("clusterprofiler")

> BiocManager::install("ggplot2")

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-NMNAT1	Cell Signaling Technology (CST)	Cat# 98354S, 1:1,000 dilution
β-actin (8H10D10) mouse mAb	Cell Signaling Technology (CST)	Cat# 3700S, 1:5,000 dilution
IRDye 800CW goat-anti-rabbit	LI-COR	Cat# 926-32211, 1:10,000 dilution
IRDye 680RD donkey-anti-mouse	LI-COR	Cat# 926-68072, 1:10,000 dilution
Bacterial and virus strains
Endura electrocompetent cells	LGC (Lucigen)	Cat# 60242-2
Biological samples
Nude mouse livers and lungs	Human cancer cell line-derived xenograft models	N/A
Chemicals, peptides, and recombinant proteins
RPMI medium 1640 basic (1X)	Gibco	Cat# C11875500BT
DMEM basic (1X)	Gibco	Cat# C11995500BT
Opti-MEM (1X)	Gibco	Cat# 31985-070
Trypsin	Gibco	Cat# 25200
Penicillin-streptomycin	Gibco	Cat# 15140122
FBS	Cellmax	Cat# SA211.02
Puromycin	Gibco	Cat# A11138-03
D-luciferin, sodium salt	Yeasen	Cat# 40901ES03
NEBNext high-fidelity 2X PCR master mix	NEB	Cat# M0541L
BsmBI-v2	NEB	Cat# R0739L
NovoRec Plus one step PCR cloning kit	Novoprotein	Cat# NR005
Proteinase K	TIANGEN	Cat# U8507
Red cell lysis	Beyotime	Cat# C3702
cOmplete tablets EASYpack	Roche	Cat# 04693116001
Lipofectamine 2000	Invitrogen	Cat# 11668-019
Polybrene	MilliporeSigma	Cat# TR-1003-G
Critical commercial assays
Tumor dissociation kit (human)	Miltenyi Biotec	Cat# 130-095-929
Lenti-X GoStix Plus kit	Takara	Cat# 631280
QIAquick PCR purification kit	QIAGEN	Cat# 28104
QIAquick gel extraction kit	QIAGEN	Cat# 28706
QIAGEN plasmid plus maxi kit	QIAGEN	Cat# 12963
Qubit 1X dsDNA HS assay kit	Invitrogen	Cat# Q33231
SYBR Green I	Yeasen	Cat# 10222ES60
1 Kb DNA ladder	TransGen Biotech	Cat# BM201-01
Deposited data
Metabolic Mini-pool CRISPR in vivo screen data	Wang et al.1	GEO: GSE279723
Experimental models: Cell lines
Human clear-cell ovarian carcinoma cell: ES-2	ATCC	Cat# CRL-1978, RRID:CVCL_3509
HEK-293T	NICR	1101HUM-PUMC000010
Experimental models: Organisms/strains
Mouse: BALB/c nude	SHANGHAI SLAC LABORATORY ANIMAL	Female, 6–8 weeks
Oligonucleotides
Array-F	N/A	TAACTTGAAAGTATTTCGATTTCTTGGCTTT ATATATCTTGTGGAAAGGACGAAACACCG
Array-R	N/A	ACTTTTTCAAGTTGATAACGGACTAGCCTT ATTTTAACTTGCTATTTCTAGCTCTAAAAC
NGS-Lib-Fwd-1	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCC TACACGACGCTCTTCCGATCTTAAGTAGAGGCTTTA TATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-2	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCCT ACACGACGCTCTTCCGATCTATCATGCTTAGCTTTATA TATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-3	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTA CACGACGCTCTTCCGATCTGATGCACATCTGCTTTA TATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-4	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCCT ACACGACGCTCTTCCGATCTCGATTGCTCGACGCTTTA TATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-5	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTA CACGACGCTCTTCCGATCTTCGATAGCAATTCGCTTT ATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-6	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCCT ACACGACGCTCTTCCGATCTATCGATAGTTGCTTGCTT TATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-7	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCC CTACACGACGCTCTTCCGATCTGATCGATCCAGTTA GGCTTTATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-8	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCC TACACGACGCTCTTCCGATCTCGATCGATTTGAGCCT GCTTTATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-9	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCC TACACGACGCTCTTCCGATCTACGATCGATACACGAT CGCTTTATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Fwd-10	N/A	AATGATACGGCGACCACCGAGATCTACACTCTTTCCC TACACGACGCTCTTCCGATCTTACGATCGATGGTCCA GAGCTTTATATATCTTGTGGAAAGGACGAAACACC
NGS-Lib-Rev	N/A	CAAGCAGAAGACGGCATACGAGATNNNNNNNNGTG ACTGGAGTTCAGACGTG
sgNMNAT1#1	N/A	AATGGGTGGAAGTTGATACA
sgNMNAT1#2	N/A	ACACCATCAAGAGAAATTGG
sgNMNAT1#3	N/A	ACTCATTCCTGCCTATCACC
sgNMNAT1#4	N/A	TGAAGACATCACCCAAATCG
Recombinant DNA
pLenti-CRISPR-v2	Addgene	Cat# 52961
Human pooled mini-CRISPR library	Wang et al.1	N/A
pLP1	Beijing Zoman Biotechnology	Cat# ZK838
pLP2	Beijing Zoman Biotechnology	Cat# ZK839
pLP/VSVG	Beijing Zoman Biotechnology	Cat# ZK840
Software and algorithms
BioRender	BioRender	http://biorender.com/
Adobe Illustrator	Adobe	www.adobe.com
R v.4.2.3	R	https://www.r-project.org/
MAGeCK v.0.5.9.4	Li et al.2	https://sourceforge.net/p/mageck/wiki/
MAGeCK-Flute v.2.8.0	Wang et al.3	https://bioconductor.org/packages/release/bioc/html/MAGeCKFlute.html
clusterProfiler v.4.12.6	Yu et al.4	https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html
ggplot2 v.3.5.2	ISBN 978-3-319-24277-4	https://cran.r-project.org/web/packages/ggplot2/index.html
Other
gentleMACS C tube	Miltenyi Biotec	Cat# 130-093-237
gentleMACS Octo dissociator with heaters	Miltenyi Biotec	Cat# 130-096-427
Qubit 4	Invitrogen	Cat# Q33238
Gemini X2 electroporation system	BTX	Cat# 45-2040
Biospace Optima	Biospace Lab	Photon Imager Optima
Illumina HiSeq X	Illumina	N/A
NanoDrop spectrophotometer	Thermo Fisher Scientific	NanoDrop OneC

Step-by-step method details

Custom sgRNA library design and construction

Timing: 4 weeks

The protocol begins with custom sgRNA library construction, including 1) designing sgRNAs for target genes of interest; 2) synthesizing, amplifying, and purifying oligo pools; 3) cloning into lentiviral backbones; and 4) transformation and plasmid preparation. Quality control involves NGS verification of sgRNA distribution and mapping efficiency. This standardized 4-week workflow ensures representation of all designed guides while minimizing bias for downstream screening applications.

• 1.Prepare the target gene list of interest according to the purpose of the study.

Note: Public datasets such as GO (https://geneontology.org/) or KEGG (https://www.genome.jp/kegg/) databases could be used as references for gene annotations.

• 2.Select sgRNAs for each gene in the list using software tools including CRISPick (https://portals.broadinstitute.org/gppx/crispick/public) or assemble from other published libraries.

CRITICAL: To ensure the fidelity of screen results and eliminate off-target effects in CRISPR screens, it is recommended to include at least four sgRNAs per gene for in vivo screening experiments.

CRITICAL: To avoid sequence-specific biases and reduce off-target effects, it is crucial to maintain a balanced base composition (roughly equal proportions of A, T, C, and G nucleotides at each position) in the designed CRISPR screen library.

CRITICAL: After selecting targeting sgRNAs, include non-targeting sgRNAs as negative controls to assess screen performance. These control guides, which lack genomic targets, serve as critical benchmarks for distinguishing true biological effects from experimental noise. During analysis, valid hits will show significant enrichment or depletion of targeting sgRNAs in experimental versus control conditions, while non-targeting guides should maintain consistent representation across conditions.

Note: The trans-activating CRISPR RNA (tracrRNA) sequence varies in the pipeline of the Genetic Perurbation Platform (GPP) tool (https://portals.broadinstitute.org/gpp/public/, gtttv for the Chen version and gtttt for the Hsu version).^5,6

• 3.Synthesize the pooled sgRNA oligo library as a custom DNA array using a commercial high-throughput synthesis service (for instance, GeneWiz or GenScript).
• 4.
  PCR amplification of pooled oligo library for integration.

  • a.
    Prepare the PCR master mix using the components listed below:

    Reagent	Amount
    NEBNext High Fidelity PCR Master Mix, 2×	25 μL
    Pooled oligo library template from Step 2 (25 ng/ L)	1 μL
    Array-F	1.25 μL
    Array-R	1.25 μL
    ddH2O	21.5 μL
    Total	50 μL

    Note: The sequences of primers Array-F and Array-R were designed based on the U6 promoter and sgRNA scaffold regions of the pLentiCRISPR_v2 vector to facilitate subsequent homologous recombination.

  • b.
    Perform PCR programs using the following cycling conditions.

    Steps	Temperature	Time	Cycles
    Initial Denaturation	98°C	30 sec	1
    Denaturation	98°C	10 sec	25 cycles
    Annealing	63°C	10 sec
    Extension	72°C	15 sec
    Final extension	72°C	2 min	1
    Hold	4°C	Forever

    CRITICAL: PCR amplification should be limited to 25 cycles in a 50-μL reaction volume to maintain balanced representation of double-stranded DNA fragments.

  • c.
    Run the amplified DNA on a 2% (wt/vol) agarose gel in a TAE buffer with SYBR Green dye.

    Note: Perform electrophoresis until sufficient separation is achieved to clearly resolve target bands from potential primer-dimers.

  • d.
    Carefully excise the target band-containing gel fragment and purify the PCR products using the QIAGEN Gel Extraction Kit according to the manufacturer’s manual and use Qubit fluorometer to quantify the concentration of the DNA products.
• 5.
  Prepare the linearized backbone.

  • a.
    Digest the pLentiCRISPR-v2 plasmids with restriction enzyme Esp3I (BsmBI). Prepare the master mix using the following component table:

    Reagent	Amount
    FastDigest Buffer, 10×	2 μL
    Library Plasmid Backbone	1–2 μL
    FastDigest Esp3I (BsmBI)	1 μL
    FastAP Thermosensitive Alkaline phosphatase	1 μL
    DTT, 100 mM	0.2 μL
    ddH2O	13.8–14.8 μL
    Total	20 μL

  • b.
    Incubate the digestion mix at 37°C for 1–2 h.
  • c.
    Run the double-strand fragments on a 1.2% (wt/vol) agarose gel in TAE buffer with SYBR Green dye.
  • d.
    Carefully excise the target band-containing gel fragment and purify the PCR products using the QIAGEN Gel Extraction Kit according to the manufacturer’s manual and use Qubit fluorometer to quantify the concentration of the products.
• 6.
  Use HiFi DNA assembly mix to ligate the pooled insert with the backbone vector.

  • a.
    Prepare the master mix for HiFi reactions at 4°C according to the below table.

    Reagent	Amount
    NEB HiFi Assembly Master Mix, 2×	10 μL
    Digested pLentiCRISPR_v2 backbone from Step 5	330 ng
    sgRNA library insert from Step 4	50 ng
    ddH2O	X μL
    Total	20 μL

    Note: A parallel reaction with a control insert is recommended to assess assembly efficiency.

  • b.
    Incubate the HiFi assembly reaction at 50°C for 1 h after the preparation of the master mix is completed.

    Note: Extended incubation (up to 4 h) can enhance assembly efficiency in certain cases, but 8–12 h incubation should be avoided. To ensure comprehensive library coverage, scale up the HiFi assembly reaction volume to account for sgRNA overrepresentation. For a library containing 100,000 sgRNAs, perform one DNA assembly reaction, and scale the number of reactions proportionally based on the size of your custom sgRNA library.

• 7.
  Purify and concentrate the assembled plasmids.

  • a.
    Prepare the mix according to the following table:

    Reagent	Amount
    HiFi Assembly Reaction Product	20 μL
    Pre-cooled isopropanol	20 μL
    GlycoBlue Coprecipitant	0.2 μL
    NaCl solution, 5 M	0.4 μL
    Total	40.6 μL

  • b.
    Pipette and incubate the mix at −20°C for 1 h and centrifuge at >15,000 g for 20 min at 4°C to deposit the DNA.
  • c.
    Remove the supernatant and wash the DNA pellets with pre-cooled 80% ethanol (v/v) twice.

    Note: During washing steps, gently invert the tube to resuspend pellets rather than pipetting or vortexing, as mechanical disruption may compromise sample integrity.

  • d.
    Centrifuge at 12,000 g for 5 min at 4°C to pellet the DNA precipitate and remove the supernatant.
  • e.
    Carefully remove the residual ethanol and dry for 20 min. Resolve the DNA pellet in 5 μL of pre-warmed Ultrapure water (65°C) and use Qubit fluorometer to quantify the DNA concentration.
  • f.
    Dilute the library DNA to 50–100 ng/μL with Ultrapure water.

    CRITICAL: Pre-cooling the isopropanol and 80% ethanol at −20°C and incubating the mix at −20°C for precipitation is strongly recommended to increase the efficiency.

    CRITICAL: We recommend verifying assembly efficiency by gel electrophoresis, checking for complete disappearance of the pooled fragment band (Figure 3).

    Note: The precipitated DNA pellets should be blue if the GlycoBlue Coprecipitant is added. Without GlycoBlue Coprecipitant, white pellets are also apparent under the light.

• 8.
  Electroporate the library DNA prepared above using Endura ElectroCompetent cells according to the manufacturer’s manual.

  • a.
    Thoroughly chill 1.5 mL microcentrifuge tubes and electroporation cuvettes on ice before use.
  • b.
    Thaw Endura electrocompetent cells completely on ice, then aliquot 25 μL of cells into the pre-chilled 1.5 mL tubes.
  • c.
    Add 1 μL of library DNA to the 25 μL of cells while keeping the mixture on ice.
  • d.
    Gently transfer the 25 μL cell/DNA mixture into a pre-chilled electroporation cuvette, avoiding the introduction of air bubbles.
  • e.
    Conduct electroporation using the Gemini X2 Electroporation System (BTX, Cat# 45-2040) with the following settings: 10 μF, 600 Ohms, 1800 Volts, and a pulse duration of 10 milliseconds.
  • f.
    Immediately add 1 mL Recovery Buffer to the cuvette.
  • g.
    Transfer the cell suspension to a culture tube and incubate at 37°C with shaking at 225 rpm for 1 h.

CRITICAL: Scale up electroporation reactions proportionally to library size. As a guideline, perform one electroporation per 2,500–5,000 sgRNAs. For instance, a 7,000-sgRNA library requires at least two parallel electroporation reactions.

Note: The recommended library DNA counts for each electroporation reaction (25 μL system) range from 50 ng to 100 ng.

Note:Stbl4 ElectroCompetent cells are an alternative for DNA library amplification.

• 9.
  Prepare the counting plate and transfer the electroporated cells to large LB plates.

  • a.
    Prepare serial dilutions of electroporated cells:

    • i.
      Create 1,000× dilution by mixing 10 μL electroporated cells with 990 μL LB medium (mix thoroughly by pipetting).
    • ii.
      Prepare 10,000× dilution by adding 100 μL of the 1,000× dilution to 900 μL LB medium.
    • iii.
      Plate 100 μL aliquots from both dilutions on pre-warmed LB agar plates (100-mm Petri dish) as the counting plates to determine electroporation efficiency.
    • iv.
      Count the colony numbers on the 10,000 diluted plate and multiply 10,000 as the total colony numbers.
  • b.
    Transfer the remained liquid in the raw tube to pre-warmed LB plates (245 ∗ 245 mm) and spread them evenly.
  • c.
    Incubate all LB agar plates for 8–12 h at 37°C.

CRITICAL: Avoid complete drying of plates during plating, as this may compromise cell viability.

• 10.
  Collect E.coli cells from the plate and extract the library plasmids.

  • a.
    Calculate the electroporation efficiency by counting plates. Proceed with the following steps only if the electroporation coverage is over 500-fold. For example, at least 5 million colonies should be harvested for a library containing 10,000 sgRNAs.
  • b.
    Evenly add 15 mL LB medium onto the Large LB plate, then shake horizontally at 80 rpm for 15 min at 25°C. Collect the medium carefully.
  • c.
    Repeat Step 10b. Collect the medium into the same 50 mL tube. Centrifuge at 5,000 g for 10 min at 25°C.
  • d.
    Extract the library plasmids using QIAGEN Plasmid Plus Max Kit according to the manufacturer’s manual.

CRITICAL: Ensuring adequate coverage of the DNA library is crucial to the success of the screen.

Note: To preserve the diversity of the sgRNA library, it is essential to harvest colonies from plates. This approach ensures that each transformed colony derives from a single bacterium, minimizing the risk of biased representation. In contrast, liquid cultures—even under antibiotic selection—may inadvertently favor fast-dividing clones, leading to overgrowth and skewed library composition.

Note: We recommend loading ∼0.6 g of bacterial pellet or equivalent for each spin column.

• 11.
  Prepare samples for next-generation sequencing to quantify the relative abundances and distribution of sgRNAs in the library.

  • a.
    Use PCR to amplify the sgRNA target regions with sgRNA sequencing primers. Prepare the master mix and perform the PCR reaction using the following tables, respectively:

    Reagent	Amount
    KAPA HiFi HotStart ReadyMix, 2×	25 μL
    Template from Step 10d or Step 18h	50 ng (for plasmid library from Step 10d), 3–5 μg (for gDNA from Step 18h)
    NGS-Lib-Fwd	1 μL
    NGS-Lib-Rev	1 μL
    ddH2O	X μL
    Total	50 μL

    Steps	Temperature	Time	Cycles
    Initial Denaturation	95°C	5 min	1
    Denaturation	98°C	20 sec	25 cycles
    Annealing	60°C	15 sec
    Extension	72°C	15 sec
    Final extension	72°C	2 min	1
    Hold	4°C	forever

    CRITICAL: For parallel library preparation, incorporate unique barcodes in reverse primers to facilitate post-sequencing sample data demultiplexing.

  • b.
    Pool the PCR reactions and purify the PCR products using the QIAquick PCR Purification Kit according to the manufacturer’s directions. Quantify the DNA concentration by Qubit fluorometer.
  • c.
    Run the purified products on the 2% (w/v) agarose gel. Cast the gel to collect the bands ranging from 260 to 270 bp for purification by QIAquick Gel Extraction Kit according to the manufacturer’s directions.
  • d.
    Quantify the DNA concentration by Qubit fluorometer.
  • e.
    Perform sgRNA sequencing on an Illumina HiSeq X platform using 2×150 bp reads with P7 indexing. We recommend obtaining more than 100× coverage per sgRNA.
• 12.
  Analyzing the sgRNA distribution in the plasmid library using the MAGeCK algorithm.

  • a.
    Download the raw sequencing data (.fastq files).
  • b.
    Calculate the sgRNA counts by mageck-count using the following scripts. Optimal parameters and annotations are also listed below.

    $ mageck count -l library.txt -n plasmid.library –sample-label plasmid –fastq filename.R1.fastq.gz –fastq-2 filename.R2.fastq.gz –norm-method none

    Parameter	Explanation	Annotation
    -l	File containing the sgRNA IDs, sequences and target gene symbols, either in .csv or .txt format	–
    -n	Prefix of the output file	–
    --sample label	Sample name of each input	Separate the sample names by comma (,) when simultaneously inputting multiple libraries.
    --fastq	.fastq filename	Separate the sample by space; separate technical replicates by a comma (,) in the case of inputting multiple libraries simultaneously. We recommend using fastq-2 to read the reverse sequence.
    --norm-method	Method of normalization, including the “none” (no normalization), “median” (median normalization), “total” (normalization with total sgRNA counts), and “control” (normalization with provided control matrix specified by optional parameter --control-sgrna)	For quality control of the plasmid library, we recommend calculating the sgRNA counts without normalization. For calculating the sgRNA counts in samples, we recommend using the “median” or “control” parameters, depending on the number of non-targeting sgRNAs included in the library (use median normalization if < 50 non-targeting sgRNAs, and control normalization if ≥ 50 non-targeting sgRNAs).
    --control-sgrna (optional parameters)	File containing the non-targeting sgRNA IDs	Use this parameter only if the –norm-method specifies the “control”.

  • c.
    Determine the quality of the library using the parameters from the count_summary.txt, including the Mapping Reads/Total Reads, Gini Index, and Zero Counts.
  • d.
    Draw density plots or histograms to depict the distribution of sgRNA counts.

    CRITICAL: The quality of the library is vital. There are several essential standards to determine the quality of the plasmids library: 1) high mapping ratio (mapping reads / total reads) indicates the sound quality of next-generation sequencing (NGS) sample preparation; 2) low gini index and even distribution of density plots (or histograms) suggest the even distribution of sgRNAs in the pooled plasmids library; 3) low to no zero counts illustrate the tolerable loss of sgRNAs (Figure 3).

Figure 3.

Determining assembly efficiency by gel electrophoresis, related to step 7

sgRNA library insert (∼140 bp), digested pLentiCRISPR_v2 backbone (∼13,000 bp), and HiFi assembly reaction product were resolved on a 2% TAE agarose gel. Electrophoresis was performed at 120 V for 30 min in 1× TAE buffer, followed by SYBR green staining and UV visualization. 1 Kb DNA ladder were included for size validation.

In vivo CRISPR screen using an ovarian cancer distant metastasis mouse model

Timing: 7 weeks

The in vivo CRISPR screen involves generating lentiviral particles with the CRISPR library and determining viral titer for optimal infection. Cells are transduced at low MOI and injected into mice to model metastasis. Post-injection, metastatic tissues are harvested and gDNA is isolated, followed by PCR amplification and sequencing.

• 13.
  Generate lentiviral particles containing the CRISPR library for subsequent cell infection.

  • a.
    Plate 5 million HEK293-T cells into one 15-cm dish with 26 mL complete DMEM medium (10% FBS + 1% Pen/Strep).
  • b.
    Culture cells in a 37°C incubator with 5% CO₂ supplement to reach 50°C∼70% confluence. Then, 1 × 10⁷ HEK293-T cells are transfected with a pooled library (24 μg) and packaging constructs (8 μg each of pLP1, pLP2 and pLP/VSVG) at a 3:1 ratio using Lipofectamine 2000.
  • c.
    At 8 h post-transfection, replace the medium containing transfection reagent with fresh complete DMEM medium (10% FBS + 1% Pen/Strep).
  • d.
    Incubate HEK293-T cells for 48 h and collect the supernatant. Filter the supernatant with a 0.45-μm filter and store the lentiviruses at −80°C.
• 14.
  Determine the viral titer to optimize the infection condition and ensure efficient delivery of the CRISPR library.

  • a.
    Seed cells of interest at 20% confluence in 10-cm culture dishes and incubate for 24 h to ensure complete cell adhesion.
  • b.
    Perform preliminary lentiviral transductions using a viral dilution series supplemented with 8 μg/mL polybrene to establish optimal infection conditions for achieving a final multiplicity of infection (MOI) of 0.1–0.5.
  • c.
    Following an 8–12 h incubation period, refresh the complete DMEM medium (10% FBS + 1% Pen/Strep) to remove polybrene.
  • d.
    48 h post-infection, harvest cells, determine the initial cell count (n0) and initiate puromycin selection. Maintain cells by refreshing the medium every 48 h.
  • e.
    Following 3 days of selection, terminate the antibiotic treatment and quantify the surviving cells (nt).
  • f.
    Calculate the multiplicity of infection (MOI) using the formula: MOI = nt/n0.

CRITICAL: The cells should be infected at a relatively low MOI, often between 0.1 to 0.5. We chose 0.3 as the MOI used below.

• 15.Infect cells with the pooled library according to the previously tested infection rate. Repeat steps 14a–14e.

Note: We recommend infecting cells with lentivirus collected from different plates as infection replicates.

• 16.
  Inject transduced cells into mice via the appropriate route (e.g., intravenous, subcutaneous, or intraperitoneal) for disease modeling.

  • a.
    After 2–7 days of antibiotic selection (puromycin, in this case), cells were washed with PBS, digested with trypsin and harvested by the complete medium. Wash the cell pellets with PBS twice and resuspend them in PBS buffer.
  • b.
    Quantify and collect sufficient cells as a pre-injection group.
  • c.
    Dilute cells in PBS buffer to 1–10 million cells per milliliter, depending on the size of the customized library and the injection methods.

    CRITICAL: To ensure robust results, maintain pre-engraftment sgRNA coverage ≥100× (target 500×) for unbiased assessment, while limiting the cell number per injection to ≤1 × 10⁷.

  • d.
    Inject cells into mice using the proposed injection method.
• 17.
  Harvest target tissues from the animals and dissociate into single-cell suspensions for downstream analysis.

  • a.
    Starting 3–7 days post-injection, monitor tumors every 2–3 days using vernier calipers (physical dimensions) or bioluminescence imaging (luminescent models).
  • b.
    Euthanize mice when any of the following humane endpoints are reached: 1) Primary tumor diameter ≥ 20 mm; 2) > 20% body weight loss in 72 h; 3) severe clinical symptoms (lethargy, impaired mobility, or neurological deficits).
  • c.
    Aseptically harvest target tissues, place them in ice-cold PBS and mince tissues into <2 mm³ pieces using sterile instruments. Dissociate tissue fragments using the Miltenyi Tumor Dissociation Kit according to the manufacturer’s manual.

    Note: There are two different kits for dissociating mouse and human tumors, depending on the types of tumor cells engrafted to the mice. We recommend dissociating cells using gentleMACS with Heaters for maximum tissue dissociation efficiency.

  • d.
    Centrifuge cell suspensions at 300 g for 5 min at 4°C. Discard the supernatant and resuspend the cell pellets in 3–5 mL Red Lysis Buffer. After 5 min of incubation at 25°C, centrifuge the samples at 400–500 g for 5 min.
  • e.
    Aspirate the supernatant and retain the cell pellet for either cryopreservation or immediate culture.

    Optional: In certain experimental contexts, particularly metastasis studies, obtaining adequate cell quantities from tissue samples for downstream analysis presents significant challenges. To address this limitation, we recommend the following protocol: 1) dissociate target cells from tissue samples; 2) establish in vitro cultures supplemented with antibiotics for 3–5 days to prevent bacterial contamination. Importantly, to control for potential in vitro artifacts, parallel cultures of pre-injection cells should be maintained under identical conditions for subsequent comparative analysis.

• 18.
  Isolate gDNA from the dissociated tissues or cells to prepare for sgRNA detection and quantitation.

  • a.
    For lysis of 4 × 10⁶ cells, dissolve frozen cell pellets in 410 μL STE buffer (0.1 M NaCl, 0.01 M Tris-HCl, pH=8.0, 1 mM EDTA, pH=8.0), 90 μL of 10% SDS, and 10 μL of 10 mg/mL proteinase K, and incubate the samples at 65°C for 8–12 h.
  • b.
    Add 350 μL saturated NaCl solution (∼26% w/v, saturated in water at 20°C) to the lysed samples and vortex for thorough mixing.
  • c.
    Add 400 μL of chloroform to each sample and vortex for thorough mixing.
  • d.
    Centrifuge samples at 12,000 g for 20 min at 4°C. Next, carefully transfer the supernatant (∼800 μL) to new 1.5 mL tubes.
  • e.
    Add 400 μL of isopropanol to the supernatant and mix thoroughly, centrifuge the samples at 12,000 g for 20 min at 4°C.
  • f.
    Discard the supernatant and wash the gDNA pellets twice with ice-cold 75% ethanol. Place the tube in the fume hood to air-dry.
  • g.
    Redissolve gDNA pellets by adding 100–200 μL of pre-warmed (65°C) Ultrapure water once they transition from milky white to slightly translucent.
  • h.
    Quantify the DNA concentration using a Nanodrop Spectrophotometer.
• 19.Perform PCR to enrich the sgRNA-containing regions, preparing the samples for sequencing (see Step 11a).

CRITICAL: To compensate for underrepresentation of sgRNA-integrated regions in heterogeneous tissue gDNA, we recommend using 3–5 μg gDNA input per reaction (when available), and conducting 3–4 parallel PCR amplifications. This complementary approach enhances sgRNA detection sensitivity in complex tissue samples.

• 20.Prepare sequencing samples following the same steps at 11b–d.

Computational data analysis for screening results

Timing: 1 week

CRISPR screening, particularly in vivo CRISPR screening, allows researchers to identify key phenotypic genes from thousands of candidates, significantly accelerating the discovery of novel therapeutic targets and the development of targeted compounds. However, pinpointing functionally important genes through in vivo CRISPR screening presents several challenges: 1) biological variability arising from the complex in vivo environment; 2) integration difficulties due to the need to account for multiple sgRNAs; and 3) the off-target effects of the CRISPR-Cas9 system. To address these challenges, robust bioinformatic tools are essential for analyzing in vivo CRISPR screening data. Several algorithms, such as MAGeCK,^2,3 casTLE,⁷ and BAGEL,⁸ have been developed to facilitate sgRNA quantification and the identification of essential genes. Among these, we chose MAGeCK for downstream analysis due to its comprehensive software suite, user-friendly pipelines, and adaptability to diverse experimental conditions. Additionally, MAGeCK is frequently updated to keep pace with advancements in CRISPR screening technologies. Its reliability has been consistently validated through multiple studies, including those conducted by our team and other research groups.^9,10,11

• 21.Download the fastq or fastq.gz files and store them in the same path.
• 22.Calculate the sgRNA counts by mageck-count using the following scripts.

Note: Optimal parameters and annotations are also listed in the above table in step 12b.

$ mageck count -l mini_pool.csv -n minipool_control --norm-method control --control-sgrna sg_control.txt --sample-label pre_injection_1,pre_injection_2,pre_injection_3,pre_injection_4,s.c_1,s.c_2,s.c_3,s.c_4,liver_1,liver_2,liver_3,liver_4,lung_1,lung_2,lung_3,lung_4 --fastq HFL-1_S1_L001_R1_001.fastq.gz HFL-2_S12_L001_R1_001.fastq.gz HFL-3_S16_L001_R1_001.fastq.gz HFL-4_S17_L001_R1_001.fastq.gz HFL-9_S22_L001_R1_001.fastq.gz HFL-10_S2_L001_R1_001.fastq.gz HFL-11_S3_L001_R1_001.fastq.gz HFL-12_S4_L001_R1_001.fastq.gz HFL-13_S5_L001_R1_001.fastq.gz HFL-14_S6_L001_R1_001.fastq.gz HFL-15_S7_L001_R1_001.fastq.gz HFL-16_S8_L001_R1_001.fastq.gz HFL-17_S9_L001_R1_001.fastq.gz HFL-18_S10_L001_R1_001.fastq.gz HFL-19_S11_L001_R1_001.fastq.gz HFL-20_S13_L001_R1_001.fastq.gz --fastq-2 HFL-1_S1_L001_R2_001.fastq.gz HFL-2_S12_L001_R2_001.fastq.gz HFL-3_S16_L001_R2_001.fastq.gz HFL-4_S17_L001_R2_001.fastq.gz HFL-9_S22_L001_R2_001.fastq.gz HFL-10_S2_L001_R2_001.fastq.gz HFL-11_S3_L001_R2_001.fastq.gz HFL-12_S4_L001_R2_001.fastq.gz HFL-13_S5_L001_R2_001.fastq.gz HFL-14_S6_L001_R2_001.fastq.gz HFL-15_S7_L001_R2_001.fastq.gz HFL-16_S8_L001_R2_001.fastq.gz HFL-17_S9_L001_R2_001.fastq.gz HFL-18_S10_L001_R2_001.fastq.gz HFL-19_S11_L001_R2_001.fastq.gz HFL-20_S13_L001_R2_001.fastq.gz

CRITICAL: In Mageck, the “--norm-method (none,median,total,control)” option sets the method parameter for normalization, and default is “median”. If “control” is specified, the size factor will be estimated using control sgRNAs specified in --control-sgrna” option. We recommend normalizing sgRNA counts using non-targeting sgRNAs as a standard approach. However, median normalization can be a suitable alternative when the number of non-targeting sgRNAs is insufficient (<50) or their distribution is uneven.

CRITICAL: We recommend using the non-normalized sgRNA counts to determine the correlative relationship and the sgRNA distribution among samples.

• 23.Determine gene essentiality using the provided scripts with either RRA (Robust Rank Aggregation) or MLE (Maximum-Likelihood Estimation).

$ mageck test -k mini_pool.count_normalized.txt -n rra_test_liver1 --norm-method control --control-sgrna sg_control.txt -t liver_1 -c s.c_1

Or

$ mageck test -k mini_pool.count_normalized.txt -n rra_test_liver --norm-method control --control-sgrna sg_control.txt -t liver_1,liver_2,liver_3,liver_4 -c s.c_1,s.c_2,s.c_3,s.c_4 –paired

$ mageck mle -k mini_pool.count_normalized.txt -n metastasis -d design_matrix.txt --norm-method control --control-sgrna sg_control.txt

Parameter	Explanation	Annotation
-k	Input file name	Output file from the megeck count module.
-n	Prefix of the output file	-
-t	Sample names of treatment experiments	Separate the sample names by comma (,)
-c	Sample names of control group	Separate the sample names by comma (,)
--norm-method	Method of normalization, including the “none” (no normalization), “median” (median normalization), “total” (normalization with total sgRNA counts), “control” (normalization with provided control matrix specified by optional parameter --control-sgrna)	We recommend using the “median” or “control” parameters, according to the option selected in the mageck count module.
--control-sgrna (optional)	File containing the non-targeting sgRNA IDs	Use this parameter only if the –norm-method specifies the “control”.
--paired (optional)	Paired sample comparisons	We recommend using this parameter for paired comparison.

CRITICAL: The RRA and MLE modules employ distinct mathematical frameworks tailored for different scenarios. Generally, MLE outperforms RRA in analyses involving multiple sample comparisons, while both methods perform comparably well in simple treatment vs. control comparisons. Additionally, it is worth noting that RRA output files include differential results at both the gene and sgRNA levels, whereas MLE provides only gene-level differential data.

CRITICAL: When studying cell growth in vivo, it is important to account for and exclude the growth variation that occurs in vitro due to the culturing of cells over several days. This in vitro growth can introduce background noise that may confound the interpretation of in vivo growth dynamics. To address this, researchers can subtract the beta value (acquired in result table from MLE test) of in vitro growth from the corresponding in vivo growth values. This subtraction helps isolate the true in vivo growth signal by removing the contribution of in vitro artifacts.

• 24.
  Visualize the results by MAGeCK-Flute in R using the following codes (a. for RRA; b for MLE) (Figure 4).

  • a.
    RRA.

    > library(MAGeCKFlute)

    > library(clusterProfiler)

    > library(ggplot2)

    > setwd("filepath")

    > gene.df <- read.table("sample.matrix.control.gene_summary.txt", sep = "∖t", header = T)

    > sgrna.df <- read.table("sample.matrix.control.sgrna_summary.txt", sep = "∖t", header = T)

    > FluteRRA(gene.df, sgrna.df, proj="sample", incorporateDepmap = T, organism="hsa", outdir = "./")
  • b.
    MLE.

    > library(MAGeCKFlute)

    > library(clusterProfiler)

    > library(ggplot2)

    > setwd("filepath")

    > mle.df <- read.table("sample.matrix.mle.gene_summary.txt", sep = "∖t", header = T)

    > FluteMLE(file3, treatname="sample1", ctrlname="sample2", proj="sample", organism="hsa")

    Parameter	Explanation	Annotation
    -k	Input file name	Output file from the mageck count module
    -n	Prefix of the output file	-
    -d	Design matrix of conditions. Example table is represented as below.	Sample names in row, conditions in column. Use 1 or 0 to distinguish the conditions among samples. The sample names in the first column should match the column names in the count matrix. The baseline condition, represented by the second column in the design matrix with all values set to “1”, is essential for performing MLE module.
    --norm-method	Method of normalization, including the “none” (no normalization), “median” (median normalization), “total” (normalization with total sgRNA counts), “control” (normalization with provided control matrix specified by optional parameter --control-sgrna)	We recommend using the “median” or “control” parameters, according to the option selected in the mageck count module.
    --control-sgrna (optional)	File containing the non-targeting sgRNA IDs	Use this parameter only if the –norm-method specifies the “control”.

    Samples	Baseline	in.vivo_growth	liver_mets	lung_mets
    pre_injection_1	1	0	0	0
    s.c_1	1	1	0	0
    liver_1	1	0	1	0
    lung_1	0	0	0	1

    or

    Samples	Baseline	in.vivo_growth	liver_mets	lung_mets
    pre_injection_1	1	0	0	0
    pre_injection_2	1	0	0	0
    pre_injection_3	1	0	0	0
    pre_injection_4	1	0	0	0
    s.c_1	1	1	0	0
    s.c_2	1	1	0	0
    s.c_3	1	1	0	0
    s.c_4	1	1	0	0
    liver_1	1	0	1	0
    liver_2	1	0	1	0
    liver_3	1	0	1	0
    liver_4	1	0	1	0
    lung_1	1	0	0	1
    lung_2	1	0	0	1
    lung_3	1	0	0	1
    lung_4	1	0	0	1

    CRITICAL: If some parameters of MAGeCK-Flute require to be refined according to users’ experimental design, we recommend referring to the step-to-step manual (https://www.bioconductor.org/packages/release/bioc/vignettes/MAGeCKFlute/inst/doc/MAGeCKFlute.html).

    Note: For biological processes including subcutaneous tumor growth, standard replicate-based designs are sufficient to ensure reliable in vivo CRISPR screening results. However, for complex, multi-step processes—such as metastasis—sample consistency becomes a major challenge due to inherent variability in progression stages. To accurately identify key genes in such contexts, we recommend the following optimized workflow:

    • i.
      Employ multiple parallel cohorts (e.g., biologically adequate group sizes).
    • ii.
      Track tumor progression rigorously (e.g., bioluminescence imaging or tumor volumetry).
    • iii.
      Harvest tissues at matched disease stages (e.g., group mice with comparable metastasis burden at endpoint).
    • iv.
      Aggregate sgRNA counts (MAGeCK) within stage-matched groups; for low-input samples (e.g., disseminated metastases), pooled gDNA analysis is permissible.
    • v.
      Compare metastatic vs. primary tumor samples within stage-matched groups to minimize false negatives from progression-related heterogeneity.
    • vi.
      Integrate cross-group differential analyses and prioritize candidate genes based on biological relevance.

Figure 4.

Quality assessment of the custom sgRNA library, related to expected outcomes

This figure was reused from Wang et al. (2025).

(A) Lorenz curve analysis of sgRNA distribution in the custom sgRNA library.

(B) Composition analysis of targeting and non-targeting sgRNA in the custom sgRNA library.

Functional validation of candidate genes associated with the screening phenotype

Timing: 7–8 weeks

This step facilitates validation of the robustness of the screen and enables selection of high-confidence candidate genes for subsequent functional and mechanistic studies involving the screening phenotype. Our screening results identified NMNAT1 as a top 5 dependency gene during both hepatic and lung metastasis of ES-2-MC2-Hep cells. In this section, we use NMNAT1 as a representative example to detail the functional validation methods and results regarding its role in liver and lung metastasis.

• 25.
  Construct the plasmids inserted with individual sgRNAs targeting NMNAT1.

  • a.
    Top- and bottom-strand oligonucleotides corresponding to 4 individual NMNAT1-targeting sgRNAs are synthesized and individually cloned into the BsmBI-linearized pLentiCRISPR_v2 backbone using Gibson Assembly.

    Sequence name	Targeted gene	Strand	Species	Sequence (5′ – 3′)
    V2-sgNMNAT1#1-F	NMNAT1	Top	Homo sapiens	CACCGAATGGGTGGAAGTTGATACA
    V2-sgNMNAT1#2-F	NMNAT1	Top	Homo sapiens	CACCGACACCATCAAGAGAAATTGG
    V2-sgNMNAT1#3-F	NMNAT1	Top	Homo sapiens	CACCGACTCATTCCTGCCTATCACC
    V2-sgNMNAT1#4-F	NMNAT1	Top	Homo sapiens	CACCGTGAAGACATCACCCAAATCG
    V2-sgNMNAT1#1-R	NMNAT1	Bottom	Homo sapiens	AAACCGATTTGGGTGATGTCTTCAC
    V2-sgNMNAT1#2-R	NMNAT1	Bottom	Homo sapiens	AAACGGTGATAGGCAGGAATGAGTC
    V2-sgNMNAT1#3-R	NMNAT1	Bottom	Homo sapiens	AAACCCAATTTCTCTTGATGGTGTC
    V2-sgNMNAT1#4-R	NMNAT1	Bottom	Homo sapiens	AAACTGTATCAACTTCCACCCATTC

  • b.
    Transform the Gibson assembly reaction mixture into Stbl3 competent cells via heat shock, followed by bacterial expansion and plasmid maxipreparation (Maxiprep).
  • c.
    Verify the correct insertion of sgRNAs by sequencing using the U6 promoter forward primer.
• 26.
  Generate ES-2-MC2-Hep cell lines expressing sgRNAs targeting NMNAT1.

  • a.
    Transfect the HEK293-T cells (maintained in DMEM with 10% FBS) with the plasmid using Lipofectamine-2000.
  • b.
    After 48 h, collect the virus-containing supernatant and determine lentivirus titer using the Lenti-X GoStix Plus kit.
  • c.
    Perform viral transduction by incubating ES-2-MC2-Hep cells (20%–30% confluency in 6-well plates) with virus supernatant in 8 μg/mL polybrene-containing medium for 24 h.
  • d.
    After infection, replace the viral medium with fresh medium and culture for another 24 h before selecting stable transfectants with 1–2 μg/mL puromycin.
  • e.
    Expand ES-2-MC2-Hep cells expressing sgNMNAT1s following puromycin selection and validate NMNAT1 protein expression by western blot using an anti-NMNAT1 antibody (Cell Signaling Technology, Cat# 98354S, produced in rabbit, used at 1:1,000 dilution). β-ACTIN was used as a loading control.
• 27.Intraperitoneally implant 1 × 10⁶ ES-2-MC2-Hep cells expressing either control sgNC or sgNMNAT1 into nude mice.
• 28.After tumor growth and metastasis development, euthanize several tumor-bearing mice at least 14 days post-implantation and collect the lung and liver tissues for subsequent ex vivo bioluminescent imaging (BLI).
• 29.Acquire the BLI images using a BioSpace Optima small animal imaging system for 1 min and quantify the normalized photon counts.
• 30.To validate the role of NMNAT1 in ES-2-MC2-Hep cell metastasis, compare ex vivo BLI normalized photon counts from liver and lung tissues between the control sgNC group and sgNMNAT1 group (Figure 5).

Figure 5.

In vivo CRISPR screening with the custom sgRNA library uncovers the role of NMNAT1 in promoting ovarian cancer visceral metastasis, related to expected outcomes

(A) Bar graph comparing Gini coefficients quantifying sgRNA distribution equality in ES-2-MC2-Hep cells across experimental conditions: pre-injection, subcutaneous tumors, and liver/lung metastasis from the in vivo CRISPR screen.

(B) Cumulative frequency distribution of normalized sgRNA counts in ES-2-MC2-Hep cells across experimental conditions: pre-injection, subcutaneous tumors, and liver/lung metastasis from the in vivo CRISPR screen.

(C) Scatter plot showing the rank and log₂ fold-change of sgRNA in ES-2-MC2-Hep cells from liver (left) and lung (right) metastases relative to subcutaneous tumors in the in vivo CRISPR screen. Individual sgRNAs targeting NMNAT1 were highlighted.

Expected outcomes

The protocol is designed to generate a high-quality custom sgRNA library with balanced representation of all sgRNAs. Prior to the experiment, we anticipate successful PCR amplification and efficient insertion of sgRNA fragments into the plasmid backbone. The library should exhibit minimal bias, as confirmed by a Lorenz curve with an AUC near 0.5 and symmetrical histograms of sgRNA counts (Figures 4A and 4B). These metrics will ensure that the library is suitable for unbiased screening. During the in vivo CRISPR screen, we expect uniform sgRNA distribution across samples, as indicated by low Gini coefficients (Figures 5A and 5B). The screen should reveal significant depletion of sgRNAs targeting genes essential for metastasis, such as NMNAT1, in liver and lung metastases compared to subcutaneous tumors (Figure 5C). These findings would support the hypothesis that specific metabolic or regulatory genes play critical roles in ovarian cancer metastasis. For validation, we anticipate that knockout of top candidate genes (for example, NMNAT1) will impair metastatic potential in vivo. Specifically, NMNAT1-deficient cells should show reduced metastatic burden in liver and lung tissues, as measured by ex vivo BLI (Figures 6A and 6B). This outcome would confirm the screening results and establish a functional link between the identified genes and metastatic progression.

Figure 6.

Experimental validation of the metastasis-promoting function of NMNAT1 using the ES-2-MC2-Hep metastasis model, related to expected outcomes

This figure was reused from Wang et al. (2025).

(A) Immunoblot analysis of NMNAT1 knockout efficiency in ES-2-MC2-Hep cells expressing NMNAT1-targeting sgRNAs.

(B–C) Liver (B) and lung (C) metastatic burden quantification by ex vivo BLI in nude mice intraperitoneally injected with ES-2-MC2-Hep cells expressing NMNAT1-targeting sgRNAs. p value were calculated by a two-tailed, unpaired Student's test. Data are presented as box and whisker plots (n≥13).

Limitations

While this protocol aims to provide a versatile framework for various experimental scenarios, certain limitations warrant consideration due to technical constraints and research scope.

Microenvironmental complexity

In vivo screening faces inherent challenges including cell loss during hematogenous dissemination and limited library coverage in large-scale experiments. These constraints necessitate increased sample sizes and targeted secondary screens based on primary genome-wide data to ensure reliable results.

Protocol scope and extensions

The current protocol focuses on lentiviral transduction of cancer cells in vitro followed by in vivo implantation - a widely applicable approach for various cancer studies. For investigations involving primary cells or immune components, alternative delivery methods such as pooled adeno-associated virus (AAV) systems may be employed to study cancer initiation or tumor microenvironment interactions. While not covered in this protocol, these alternative approaches provide valuable complementary strategies for related research questions.

Troubleshooting

Problem 1

The major bottleneck in custom library preparation involves optimizing two critical efficiencies: (1) sgRNA fragment assembly and (2) bacterial electroporation. For large libraries, even modest reductions in either parameter dramatically decrease clone diversity and library complexity, limiting experimental utility.

Potential solution

Execute parallel assembly reactions to maximize integrated sgRNA libraries (Step 6), followed by DNA concentration (ethanol precipitation/commercial kits) (Step 7). Electroporate at 100–200 ng/μL using high-efficiency cells (>10⁹ cfu/μg), performing multiple reactions to ensure diversity (Step 8). Pool transformations to ensure adequate library representation and diversity.

Problem 2

A significant technical challenge of in vivo CRISPR screening is to obtain high-quality genomic DNA (gDNA) with minimal contamination from non-target cells. The cell contamination not only compromises gDNA purity but also interferes with PCR amplification during NGS library preparation, potentially skewing screening results. This challenge is particularly prominent when working with complex tissue samples containing heterogeneous cell populations.

Potential solution

To address this challenge, we recommend a multi-step purification approach: First, use specialized tissue dissociation kits combined with lysis buffers to remove hematopoietic cell contamination (Step 17d). For enhanced purity, implement fluorescence-activated cell sorting (FACS) using either pre-introduced fluorescent markers or cell surface-specific antibodies. As an alternative approach, consider short-term in vitro culture with antibiotic selection, though this method may introduce culture-induced artifacts (Step 17e). Regardless of the chosen method, always validate cell purity through marker analysis before proceeding with gDNA extraction to ensure screening reliability.

Problem 3

While we have established a streamlined pipeline for CRISPR screen data analysis here, the mathematical modeling required to integrate sgRNA differential effects and calculate gene essentiality remains inherently complex. This complexity is further confounded by the physiological intricacies of in vivo systems. Although MAGeCK provides robust analytical tools that address many computational challenges, its suitability for all experimental scenarios and biological questions warrants careful consideration.

Potential solution

For effective CRISPR screen analysis using mathematical modeling, we emphasize two priorities: (1) enhancing raw data quality through stringent experimental controls and (2) optimizing study design for MAGeCK compatibility. While alternative analytical methods exist for specialized cases, most researchers benefit most from robust data generation and proper experimental design. Complex applications may require collaboration with computational biologists to evaluate alternative approaches.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xi Wang (wangxi@westlake.edu.cn).

Technical contact

Technical questions on executing this protocol should be directed to and will be addressed by the technical contact, Yuqi Wang (wangyuqi@westlake.edu.cn).

Materials availability

This study did not generate new reagents.

Data and code availability

The datasets used in this paper have been listed in key resources table.

Acknowledgments

We thank the Westlake University Laboratory Animal Resource Center and Biomedical Research Core Facilities for technical support and Repugene (Hangzhou), Inc. for DNA-sequencing services. We thank members of the Wang and Zou labs for discussion and daily support.

This work was supported by the National Key Research and Development Program of China (2023YFA0914900), the “Pioneer” and “Leading Goose” R&D Program of Zhejiang (2024SSYS0036), the Westlake Laboratory of Life Sciences and Biomedicine (project number: 2024SSYS0033 and 2024SSYS0034, Y.Z. and X.W.), Westlake Education Foundation (Y.Z.), the Westlake University Research Center for Industries of the Future (Y.Z.), and the National Natural Science Foundation of China (project number: 82273257 and 32450793, Y.Z.).

Author contributions

Y.W. and M.H. performed the experiments. M.H. and X.W. performed the data analysis. X.W. and Y.Z. conceived and supervised the project. The authors co-wrote the manuscript.

Declaration of interests

Y.Z. is a consultant for Keen Therapeutics.