Parallel In Vivo Screening of Gene Knockout and Activation in the Mouse Mammary Gland

Summary

This protocol describes a scalable intraductal CRISPR screening approach enabling parallel in vivo gene knockout and activation studies in the mouse mammary gland.

Abstract

Forward genetics screens are routinely employed to perturb thousands of genetic elements in a pooled fashion with the goal of producing large-scale genotype-to-phenotype maps. While often carried out in cell culture systems, accumulating evidence supports that in vivo screens have the power to unveil new biology that cannot be recapitulated in vitro. However, the widespread application of this approach has been limited by two major challenges: a predominant focus on loss-of-function perturbations rather than gene activation and the significant technical hurdles of delivering complex genetic libraries to specific tissues in vivo. To overcome these challenges, we describe a simple and versatile intraductal injection strategy that enables efficient and rapid functional genomic screening in the mouse mammary gland, by generating tens of thousands of discrete epithelial clones. Furthermore, we provide all the details necessary for library generation, intraductal injection, screen deconvolution, and analysis of CRISPR-Knockout and Activation libraries for comprehensive in vivo screens. Using these tools, which we termed CRISPR-KOALA (Knockout and Activation Linked Assay), we have identified new tumor suppressors and oncogenes within the coding and non-coding genome in pooled libraries ranging from 46 loci to one-fifth of the genome. Importantly, this approach and analysis can be applied to other organs to study the biological function of any gene during homeostasis or disease.

Introduction

Functional genomics screens provide a powerful framework to interrogate gene function at scale¹. The development of CRISPR-Cas gene editing technologies has enabled the systematic perturbation of hundreds-to-thousands of genes in parallel, and pooled CRISPR-based screens in mammalian cell lines have become a cornerstone of modern cancer biology research². These in vitro approaches have identified essential genes³, tumor suppressors⁴, and synthetic lethal interactions⁵ across diverse cancer contexts. However, one central limitation is an inability to capture the complex biological features of tumors in vivo, including interactions with the immune system and stroma, nutrient gradients, and tissue architecture within the microenvironment². Consistent with this limitation, recent studies have demonstrated that many bona fide cancer driver genes fail to exert measurable effects in vitro, particularly when their function depends on microenvironmental cues that are absent in cell culture^6,7,8,9,10.

The mouse mammary gland provides a uniquely advantageous tissue context for in vivo functional genomics screening. Mammary epithelial cells are highly regenerative, undergo rounds of clonal expansion during estrous cycling, and are organized into a branched ductal network that can be readily accessed and genetically manipulated^11,12,13. These features have supported decades of mosaic genetic analyses and lineage-tracing experiments and establish the mammary gland as a tractable system for interrogating cell-intrinsic and microenvironment-dependent gene function during tissue homeostasis and disease^{14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33}. Importantly, genetic perturbations can be restricted to the epithelial compartment, enabling direct study of epithelial-specific drivers of transformation, cell competition, lineage fidelity, and tumor progression in an intact tissue context.

While several strategies have been developed to deliver genetic perturbations directly to the adult mammary gland in situ, significant technical hurdles remain for high-throughput applications. Adenovirus-mediated delivery of Cre recombinase has been widely used to induce somatic recombination of conditional alleles in the adult mammary gland^34,35,36,37, offering an efficient means to manipulate gene function without the complex germline breeding required for traditional genetically engineered mouse models (GEMMs). While high-titer adenovirus can also deliver CRISPR-Cas9 components to tissue-resident cells, these vectors do not integrate into the host genome^{38,39,40,41,42,43,44,45,46,47,48,49}. This severely limits the scale of screens to ~10–55 genes, as each gRNA target site must be sequenced individually using molecular inversion probes^{38,39,40,41,42,43,44,45,46,47}. Furthermore, current existing intraductal delivery methods enable direct access to the mammary ductal system but require either invasive surgical exposure of the mammary gland⁵⁰ or rely on the use of comparatively large Hamilton syringes that are cumbersome⁵¹. While effective for localized delivery of virus, these approaches are comparatively invasive, technically challenging, and not optimized for scalable delivery of pooled lentiviral libraries. Collectively, these studies demonstrate that in situ viral perturbations are feasible, but current tools do not support the stable, pooled, sequence-retrievable perturbations required for high-throughput in vivo CRISPR screening.

To address these limitations, we and others have turned to integrating lentiviral vectors, which enable stable and heritable genetic modifications. In contrast to cell culture systems, in vivo tissues impose strict constraints on the number of perturbed cells that can be generated within a single organ, rendering genome-wide screening impractical. Consequently, in vivo CRISPR screens necessitate the design of biologically informed, targeted libraries that balance coverage with feasibility. Furthermore, most in vivo CRISPR screens, reviewed here², have focused almost exclusively on lentiviral loss-of-function perturbations, despite increasing evidence that gene activation can reveal complementary and non-overlapping biological insights^52,53,54,55. We and others have previously demonstrated that integrating lentiviral CRISPR perturbations into the adult mammary epithelium enables stable, heritable genetic modification in situ and that they can be recovered and quantified from genomic DNA^{36,55,56,57,58}. Using this approach, we performed pooled in vivo CRISPR screens that identified new drivers of oncogenic transformation^36,55, establishing both the feasibility and biological relevance of lentiviral delivery and CRISPR screening. However, these studies also highlighted the need for a simple, standardized, and scalable protocol that could be broadly adopted and extended to include both loss- and gain-of-function screening models within the same gland.

Here, we describe a simple and versatile intraductal injection strategy that enables parallel in vivo screening of CRISPR-knockout and activation libraries, termed CRISPR-KOALA, within the adult mouse mammary gland⁵⁵. CRISPR-KOALA employs either a conventional 20-21 bp ‘single-guide’ (sg)RNA for CRISPR-KO capabilities or a 14 bp ‘dead-guide’ (dg)RNA for CRISPR-A. The dgRNA efficiently guides Cas9 to its target site but prevents Cas9 endonuclease activity^59,60,61, and thus allows the use of a single mouse expressing active Cas9 for both knockout and activation modalities. Introducing two PP7 hairpins into the dgRNA trcr enables the recruitment of a PP7 Coat Protein:P65:HSF1 transactivator⁶² and leads to efficient overexpression of genes in vivo⁵⁵ (Figure 1). By combining targeted gRNA library design with high-titer lentiviral production and non-surgical intraductal delivery using glass microcapillaries, this approach generates tens of thousands of spatially discrete epithelial clones while preserving tissue architecture and microenvironmental interactions. Importantly, lentiviral integration into the host genome enables the recovery and quantification of gRNAs via high-throughput sequencing, enabling pooled functional genomic screens in vivo. Together, this allows parallel or simultaneous (in the same cell) CRISPR-KO and CRISPR-A perturbations within the same mammary gland of a Cas9-expressing mouse.

We present a detailed step-by-step protocol encompassing gRNA library design, lentiviral production and concentration, intraductal delivery, library deconvolution, and data analysis (Figure 1). Together, this method provides a practical and scalable framework for in vivo CRISPR knockout and/or activation screening in the mouse mammary gland and can be adapted to study gene function during tissue homeostasis, different disease contexts, other tissues, and other species such as the rat.

Protocol

All animal procedures were performed in accordance with the guidelines of the Canadian Council on Animal Care and were approved by The Centre for Phenogenomics Animal Care Committee (protocol 26-0272H).

1. CRISPR library cloning

NOTE: This section describes pooled cloning of CRISPR knockout and CRISPR activation libraries from array-synthesized oligonucleotides into lentiviral backbones. We routinely generate libraries containing up to 4,000 gRNAs, which can be screened at 400X-1,000X coverage in 8-20 mice. The number of gRNAs and coverage can be adjusted as needed, which will affect the number of mice required to maintain statistical power. We recommend maintaining a minimum coverage of 400X, which can be sufficient for strong gain-of-function phenotypes.

1. Guide RNA library design
   1. Compile a list of all mouse genes to be screened within a single experiment for downstream CRISPR library design, maintaining separate target lists for CRISPR knockout (CRISPR-KO) and CRISPR activation (CRISPR-A) libraries.
      NOTE: CRISPR-KO and CRISPR-A libraries are generated separately and can be combined at the stage of lentiviral transduction for simultaneous screening.
   2. For CRISPR-KO libraries, select four to five single guide RNAs (sgRNAs) per gene using high-scoring prediction algorithms such as CRISPick^62,63, CHOPCHOP⁶⁴, or VBC Score⁶⁵.
      NOTE: VBC Score yields very high efficacy and on-target activity⁶⁶. Regardless of the tool, prioritize guides with high predicted on-target activity, minimal off-target potential, and targeting coding exons.
   3. For CRISPR-A libraries, select eight to ten guides per gene using CRISPick^62,63 or CHOPCHOP⁶⁴. For tools that allow user-specified guide lengths (such as CHOPCHOP), set the guide length to 14 bp prior to guide generation. If using a tool that only generates 20 bp guides, convert these sequences to dead-guide RNAs (dgRNAs) by retaining the 14 bp sequence immediately 5’ of the protospacer adjacent motif, which target promoter regions of genes.
      NOTE: dgRNAs are truncated guide sequences designed to enable CRISPR-based transcriptional activation at approximately 50-85% the effectiveness of full-length sgRNAs⁵⁵ without inducing DNA cleavage^59,60,61.
   4. Pool all sgRNAs or dgRNAs corresponding to a single experiment into separate lists to enable uniform cloning, viral production, and in vivo screening.
      NOTE: sgRNA and dgRNAs to be screened together will be pooled post-viral production.
   5. Generate pooled negative control libraries for CRISPR-KO and CRISPR-A screens, including non-targeting^55,67 and non-genic gRNA sequences⁶⁸, as previously described, to control for background effects independent of gene disruption or activation.
      ​NOTE: If desired, separate libraries of non-target and non-genic gRNAs can be made to create two independent negative control libraries.
2. Oligonucleotide pool design and amplification
   1. Design pooled oligonucleotides such that each gRNA contains the targeting sequence, restriction enzyme overhangs compatible with the destination lentiviral backbone, and library-specific primer sequences to enable selective amplification from a complex oligonucleotide pool (Figure 1, Table 1).
   2. Assign a unique primer pair to each gRNA library to allow independent amplification of individual libraries from a shared oligonucleotide pool. Unique primer pairs are listed in Table 1.
   3. Order pooled oligonucleotide libraries from a commercial array synthesis platform, ensuring equimolar representation of all gRNAs within each pool.
   4. Amplify individual gRNA libraries from the pooled oligonucleotide template using library-specific primers in the following reaction mixture. The reaction can be modified to adhere to the specific polymerase’s recommended composition. Prepare three reactions with a pooled oligonucleotide template and one no-template negative control reaction.
      4 ng Pooled oligonucleotide template
      2.5 μL 10 μM Forward library-specific primer
      2.5 μL 10 μM Reverse library-specific primer
      10 μL 5x Reaction buffer
      0.5 μL High-fidelity DNA polymerase
      X μL Nuclease-Free water
      50 μL Total volume
   5. Perform PCR amplification using the following cycling conditions, adjusting annealing temperature based on library-specific primer melting temperature:
      1 cycle of 98 °C for 30 s
      18 cycles of 98 °C for 10 s, 62–65 °C for 20 s, and 72 °C for 3 s
      1 cycle of 72 °C for 5 min
      NOTE: Do not exceed 18–19 amplification cycles to preserve equal gRNA representation within the library.
   6. Resolve PCR products on a 2.5% agarose gel at 120 V for approximately 45 min and excise the band corresponding to the expected gRNA insert size (approximately 100 bp, depending on primer design; Figure 2A).
   7. Purify PCR products using a gel extraction kit and elute in 20 μL of ultrapure water during the final elution step.
      NOTE: (PAUSE POINT) Purified gRNA inserts may be stored at -20 °C prior to cloning.
3. Preparation and digestion of lentiviral backbones
   1. Prepare lentiviral backbone plasmids containing a removable stuffer sequence for cloning gRNA inserts. For in vivo CRISPR-KO and CRISPR-A libraries use pLKO-Cre (Addgene Plasmid #158032 or 256774) and pXPR502-PPH-Cre (Addgene Plasmid #256776), respectively.
   2. Digest 5–10 μg of lentiviral backbone DNA in a total reaction volume of 200 μL using the appropriate restriction enzyme and buffer. Use Esp3I for the pLKO-Cre CRISPR-KO backbone and SapI for the pXPR502-PPH-Cre CRISPR-A backbone.
      5–10 μg Backbone Plasmid
      20 μL 10x Enzyme buffer
      5 μL Restriction enzyme
      X μL Nuclease-free water
      ​200 μL Total volume
   3. Incubate digestion reactions for 2–3 h at 37 °C.
   4. Add 2 μL of recombinant alkaline phosphatase to each digestion reaction and incubate for 1 h at 37 °C to dephosphorylate the digested backbone.
   5. Resolve digested products on a 1% agarose gel and excise the band corresponding to the linearized backbone (Figure 2B,C)
      NOTE: For both backbones, the stuffer fragment is visible as a distinct band at approximately 2000-2500 bp, depending on the vector.
   6. Purify the excised backbone DNA using a gel extraction kit and elute in a minimal volume of ultrapure water to maintain high DNA concentration (typically 20 μL).
      NOTE: (PAUSE POINT) Purified digested backbones may be stored at -20 °C prior to ligation.
4. Digestion and ligation of gRNA libraries
   1. Set up all-in-one digestion-ligation reactions using 750 ng of purified lentiviral backbone DNA, 43 ng of gRNA library insert DNA, 3 μL of restriction enzyme, 7.5 μL of T4 DNA ligase, 15 μL of 10 mM ATP, and the appropriate amount of reaction buffer and sterile distilled water to reach a total reaction volume of 150 μL.
      NOTE: Use the same restriction enzyme that was initially used in the plasmid backbone digestion step. A ratio of 750 ng of purified lentiviral backbone to 43 ng of library insert DNA yields an approximate vector-to-insert molar ratio of 1:7 when the recommended backbones are used.
   2. Divide each ligation mixture into three equal aliquots to increase transformation efficiency and library complexity.
   3. Prepare a single negative control reaction containing backbone DNA but no gRNA inserts to assess background ligation.
   4. Incubate digestion-ligation reactions overnight at 30 °C.
   5. The following morning, add 1 μL of the appropriate restriction enzyme to each reaction aliquot and incubate for 1 h at 37 °C to drive completion of the backbone digestion.
   6. Add 2 μL of 10 mM ATP and 2 μL of T4 DNA ligase to each reaction and incubate at 25 °C for 2–3 h to complete ligation.
   7. Pool the three individual digestion-ligation reactions per library, purify the reactions using a column-based oligonucleotide cleanup method, and elute each in 21 μL of ultrapure sterile distilled water. Elute the single purified negative control reaction in 7 μL of ultrapure sterile distilled water.
      NOTE: Do not elute ligation products in buffers containing salt or EDTA, as these significantly reduce bacterial electroporation efficiency. (PAUSE POINT) Purified ligation products may be stored at -20 °C prior to electroporation.
5. Library electroporation and expansion
   1. Pre-chill 0.1 cm electroporation cuvettes, microcentrifuge tubes containing 2 μL aliquots of purified ligation products, and 14 mL round-bottom polypropylene culture tubes on ice for at least 20 min prior to electroporation.
   2. Thaw electrocompetent bacterial cells on ice and gently mix 25 μL of cells with 2 μL of purified ligation product in the pre-chilled microcentrifuge tube by pipetting.
   3. Transfer the cell-DNA mixture to a pre-chilled 0.1 cm electroporation cuvette and electroporate at 1.8 kV, 25 μF, and 200–600 Ω.
      NOTE: A successful electroporation typically yields a time constant between 4.6 ms and 4.9 ms.
   4. Immediately add 1 mL of pre-warmed, 37 °C recovery medium provided with the electrocompetent cells to the cuvette, gently resuspend the cells, and transfer the suspension to a 14 mL round-bottom polypropylene tube.
   5. Incubate the recovered cells at 30 °C while shaking at 180 rpm for 90 min.
   6. Plate 10 μL of the recovered cells diluted in 90 μL of LB broth onto LB-agar plates containing ampicillin and incubate overnight at 30 °C to assess transformation efficiency and background ligation.
      NOTE: Successful cloning should result in hundreds of colonies on the plate with insert and few (> 30) colonies on the backbone-only negative control plate (Figure 3).
   7. Further validate the library by picking ~10 colonies for miniprep and Sanger sequencing forwards from the U6 promoter. Successfully cloned libraries will have < 90% of minipreps containing clear sequencing of a gRNA, without repeated guides across the 10 minipreps. Additionally, restriction enzyme digest the minipreps and the pooled plasmid library (step 1.5.10) and run on an agarose gel to check band sizes and ensure that the lentiviral constructs have not undergone aberrant recombination.
      NOTE: Library representation can be assessed by comparing colony counts from dilution plates to the total number of gRNAs in the library, targeting 1,000-fold coverage. Multiple serial dilutions may be used to aid in this calculation.
   8. If transformation efficiency is sufficient, plate the remaining recovered cells onto 15 cm LB-agar plates containing the appropriate antibiotic and incubate overnight at 30 °C to generate a dense lawn of colonies.
      NOTE: Plate 100 μL of recovered cells onto each of ten 15 cm LB agar plates to recover enough colonies.
   9. The following day, use a cell scraper to scrape bacterial colonies into 500 mL of LB broth with ampicillin to generate a single bacterial culture representing the cloned gRNA library.
   10. Expand the pooled bacterial culture at 30 °C with shaking for approximately 3 h and isolate plasmid DNA using a maxiprep kit.

2. Lentiviral production for in vivo screening

NOTE: For intraductal delivery and in vivo pooled screening, lentiviral preparations must be concentrated to achieve high titers (ideally between 1 x 10⁸ pfu/mL and 1 x 10⁹ pfu/mL) to ensure efficient transduction of mammary epithelial cells. The following protocol describes high-titer lentiviral production optimized for in vivo use.

1. Preparation of lentiviral producing cells
   1. Maintain HEK293TN cells in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, 1% Penicillin-Streptomycin, and culture cells at 37 °C with 5% CO₂.
   2. Passage cells every 2–3 days to maintain logarithmic growth and ensure that cells do not exceed 90% confluency. Use cells at passages <20 for optimal viral production.
   3. One day prior to transfection, coat 15 cm tissue culture plates with poly-L-lysine (0.02 mg/mL) for 1 h at room temperature. Prepare nine plates for one batch of viral preparation.
   4. Remove the poly-L-lysine solution and wash plates twice with phosphate-buffered saline.
   5. Seed 1.1 x 10⁷ HEK293TN cells onto each of nine poly-L-lysine-coated 15 cm tissue culture plates ensuring that cells reach ~90% confluency at the time of transfection.
      NOTE: For applications that do not require maximal viral titers, lentiviral production can be scaled down to fewer plates (e.g., three or six plates) with proportionally reduced yields. Scale down the reagents (plasmid DNA, PEI, and DMEM) for transfection accordingly.
2. PEI-based lentiviral transfection
   1. Prepare plasmid DNA for lentiviral production using a three-plasmid system consisting of 95 μg of the gRNA library plasmid, 95 μg of the packaging plasmid psPAX2 (Addgene Plasmid #12260), and 65 μg of the VSV-G envelope plasmid pMD2.G (Addgene Plasmid #12259).
   2. Combine the three plasmids for lentiviral production in 10.8 mL of serum-free DMEM and mix thoroughly by vortexing for approximately 10 s.
   3. Add 366 μL of 1 mg/mL polyethylenimine (PEI) dropwise to the diluted DNA solution, mix thoroughly by vortexing for approximately 10 s, and incubate the DNA-PEI complexes at room temperature for 10–30 min.
   4. During complex formation, aspirate culture medium from the HEK293TN plates and add 14 mL of serum-free DMEM.
      NOTE: Fetal bovine serum can inhibit plasmid uptake by the cells.
   5. Distribute 1.2 mL of the DNA-PEI complex dropwise throughout the plate, and rock to ensure even distribution.
   6. Incubate cells at 37 °C with 5% CO₂ for 6–8 h.
   7. Replace the medium on each plate with 20 mL pre-warmed DMEM supplemented with 10% fetal bovine serum (without 1% Penicillin-Streptomycin) and continue to culture the cells at 37 °C with 5% CO₂.
      NOTE: Overnight transfection is not recommended, as prolonged exposure to PEI and a lack of serum reduces viral yield.
3. Viral harvest and ultracentrifugation
   1. Prepare 200 mL of a sucrose cushion solution by combining 40 g of sucrose, 11.7 g of sodium chloride, 4 mL of 1 M HEPES (pH 7.4), and 1 mL of 0.5 M EDTA (pH 7.9) in an appropriate volume of sterile distilled water and filter-sterilize through a 0.45 μm low-protein-binding polyethersulfone filtration unit. This solution can be stored at 4 °C for several months.
   2. Collect the lentiviral supernatant 48–72 h post-transfection.
   3. Pass 10 mL of DMEM supplemented with 10% fetal bovine serum through the 0.45 μm low-protein-binding polyethersulfone filtration unit. Discard the flow through. Load the supernatant onto the filter and apply the vacuum.
      NOTE: If concentration cannot be performed immediately, store clarified viral supernatant at 4 °C overnight.
   4. Add 30 mL of clarified viral supernatant to an ultracentrifuge tube.
   5. Slowly underlay the clarified viral supernatant with 4 mL of sucrose cushion solution.
   6. Place the ultracentrifuge tube in a compatible adaptor for a swinging bucket rotor and precisely balance the tubes.
   7. Centrifuge the tubes at 104,000 × g for at least 2 h at 4 °C.
      NOTE: A swinging bucket rotor, such as the Beckman JS24.38 rotor, is required to properly pellet the virus through the sucrose cushion.
   8. Carefully decant the supernatant without disturbing the viral pellet, which may appear as a faint white or translucent film at the bottom of the tube.
   9. Allow the tubes to air dry for at least 10 min and use a laboratory wipe to remove as much residual liquid from inside the tube as possible without disturbing the pellet.
      NOTE: Proper drying of the tube is critical to achieving high viral concentrations, as it reduces the final recovery volume.
   10. Resuspend viral pellets in 20 μL of cold phosphate-buffered saline and carefully pipette up and down, being sure to avoid air bubbles.
   11. Incubate the tubes on ice for 2 h with gentle agitation to completely resuspend the viral pellets.
   12. Transfer the concentrated virus to a clean microfuge tube and clarify the solution by centrifugation at 2,000 × g for 5 min at 4 °C.
       NOTE: This spin removes debris and residual particulate that clog the injection needle during surgery.
   13. Aliquot the concentrated virus into single-use volumes to avoid repeated freeze-thaws and store at -80 °C.
4. Lentiviral titration using LSL-tdTomato mouse embryonic fibroblasts (MEFs)
   1. Plate 100,000 LSL-tdTomato MEFs (derived from Rosa-LSL-tdTomato [JAX# 007909]) in a 6-well plate one day prior to infection in DMEM supplemented with 10% fetal bovine serum, 1% Penicillin-Streptomycin, and culture the cells at 37 °C with 5% CO₂. Plate an extra well for cell counting.
      NOTE: LSL-tdTomato MEFs serve as a reporter for Cre within the lentivirus. Any suitable reporter system for viral titering of other backbones may be used.
   2. The following day, count the number of LSL-tdTomato MEFs from an extra well to quantify the number of cells present at the time of infection.
   3. Prepare serial dilutions of concentrated lentivirus encoding Cre recombinase in a total of 2 mL of Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum. Generally, 0.4 μL, 0.1 μL, and 0.025 μL of concentrated virus yield a clean dynamic linear range (Figure 4A–D).
   4. Add 2 mL of viral dilutions to each well of LSL-tdTomato MEFs and add 2 μL of 10 mg/mL polybrene to each.
   5. Infect MEFs by spinning at 400 × g for 90 min at 37 °C.
   6. Replace viral media with 2 mL of DMEM supplemented with 10% fetal bovine serum, 1% Penicillin-Streptomycin, and culture the cells at 37 °C with 5% CO₂ for 48–72 h.
   7. Quantify the tdTomato-positive cells by fluorescence microscopy or flow cytometry to determine the fraction of infected cells at each dilution.
   8. Calculate the functional titer based on the fraction of infected cells and the number of cells present at the time of infection.
      ​NOTE: CRISPR-KO and CRISPR-A viruses are produced, concentrated, and titrated separately. Functionally titers determined using LSL-tdTomato MEFs are used to mix CRISPR-KO and CRISPR-A viruses at a defined ratio prior to intraductal injection.
5. Quality assessment of plasmid and lentivirus libraries
   1. To ascertain that the gRNA libraries are cloned with sufficient depth/coverage and that all gRNAs are maintained during cloning and virus production, it is highly recommended to sequence the plasmid pool as well as the transduced MEFs.
   2. To assess lentiviral coverage, transfect MEFs with lentivirus at >1000× coverage based on calculated viral titer. It is not necessary to ensure single-infection of cells during this step.
   3. After 48–72 h, detach cells using trypsin or a cell scraper and pellet the cells. Extract the DNA using a commercial gDNA extraction kit.
   4. Sequence 1 µg of plasmid and 3 µg of extracted MEF DNA using the protocol described in Section 5. Deep sequencing of 5 million reads is sufficient for most custom libraries with <5000 gRNAs.
      NOTE: A strong correlation (R²> 0.6) between the reads of the plasmid pool and relatively equal representation of guides (>70% of guides within 1 log decade) with no individual guide dominating library counts indicates good quality (Figure 4E).

3. Intraductal lentiviral injection into the mouse mammary gland

1. Needle and virus dilution
   1. Pull glass microcapillary needles using a micropipette needle puller to generate fine-tipped capillaries suitable for intraductal injection of 9–10 μL volumes. Use a pair of fine forceps to break off the very tip of the needle.
      NOTE: The optimal location creates a needle size that is large enough to be loaded with liquid but thin enough to fit easily in the nipple opening.
   2. Dilute lentiviral preparations to a final concentration of 1 × 10⁷ pfu/mL in sterile phosphate-buffered saline supplemented with 0.05% (w/v) Fast Green FCF dye.
   3. If mixing CRISPR-KO and CRISPR-A viruses to screen them within the same complex pool, adjust the two libraries to an equal viral titer with sterile phosphate-buffered saline and mix them at a ratio to ensure an equal representation of gRNAs across both libraries.
2. Preparation of mice for intraductal injection
   1. Anesthetize an adult female mouse (8–20 weeks of age) under 4% isoflurane with 1 L/min oxygen and apply sterile ophthalmic lubricant to both eyes to prevent corneal drying during the procedure.
      NOTE: Intraductal injection is a non-invasive procedure and does not require analgesia. Use Rosa26-LSL-Cas9, H11-Cas9, or Rosa-LSL-tdTomato mice for these experiments.
   2. Secure the mouse in a supine position on a covered heating pad set to low to maintain body temperature and maintain anesthesia using 2% isoflurane with 1 L/min oxygen via nosecone throughout the procedure.
   3. Remove fur surrounding the third thoracic and fourth abdominal nipples using depilatory cream applied with a sterile cotton swab in a circular area of approximately 0.5 cm radius for 10–15 s. Immediately remove the cream with sterile gauze and sterile water to prevent skin irritation.
      NOTE: The third and fourth mammary glands are routinely the easiest to inject. Alternatively, the second and fifth glands may be prepared for injection in the same manner.
   4. Disinfect the skin surrounding each nipple by swabbing sequentially with 70% isopropyl alcohol and iodine using sterile gauze.
   5. Place the mouse under a dissecting microscope within a biosafety cabinet to allow for clear visualization of the nipples.
   6. Using fine forceps cleaned with 70% ethanol, gently remove any dead skin overlaying the nipple to expose the ductal opening.
      NOTE: This dead keratinized layer is loosely attached and can obstruct needle entry. Removal should not damage viable skin or underlying tissue.
3. Intraductal injection
   1. Load 8 μL of the diluted virus into a pulled glass microcapillary.
   2. Hold the nipple area with a pair of fine forceps, and gently insert the needle tip into the nipple opening (Figure 5A). An appropriately placed needle will enter the duct with minimal resistance. Slowly dispense the viral suspension into the mammary duct over the course of 5–10 s.
      NOTE: 8 μL of a 1 × 10⁷ pfu/mL lentivirus will yield approximately 30% infectivity of the luminal mammary epithelial cells and a 5% infectivity of the basal mammary epithelial cells, with negligible infection (< 0.5%) of the stroma³⁶. Viral titer can be adjusted to achieve higher or lower infectivity.
   3. Monitor the spread of Fast Green dye during the injection through the dissecting microscope. Confirm successful intraductal delivery by observing dye dispersion throughout the ductal network beneath the skin (Figure 5B).
   4. Identify failed injections by pooling of dye within the mammary fat pad adjacent to the nipple rather than along the ductal tree (Figure 5C).
   5. Repeat the injection procedure for the third thoracic and fourth abdominal mammary glands on both sides, injecting a total of four glands per mouse.
      NOTE: With experience, successful intraductal delivery is achieved in >95% of injections. Failed injections confined to the fat pad are not associated with adverse effects, and the remaining glands may still be used. Further validation of injections can be conducted through immunohistochemistry and flow cytometry (Figure 5D,E).
   6. Discontinue anesthesia and transfer the mouse to a clean recovery cage.
   7. Monitor the mouse until it is fully ambulatory and return the mouse to the designated biosafety housing once normal behavior is observed.
   8. Three days post-injection, perform a full cage change, discarding any housing and bedding as viral waste, and transfer the mouse to a clean cage that can be returned to standard housing.

4. Genomic DNA isolation from mammary glands and tumors

NOTE: Euthanize mice through CO₂ exposure or overdose of anesthetics followed by cervical dislocation, in accordance with animal care guidelines. Harvest mammary glands or mammary tumors at the desired timepoints or at humane endpoint according to regulatory guidelines. Infected cells have been harvested as early as 48 h post lentiviral infection^36,55 or as late as 5+ months post injection in wildtype mice³⁶, or >24 months in oncogenic backgrounds⁵⁵. The following steps should be done in a space cleaned with DNA Erase and with equipment that is free of plasmids containing sgRNA/dgRNA cassettes. Even trace amounts of plasmid DNA are sufficient to result in contamination of tissue samples, and subsequent deep sequencing PCRs and negative (no tissue) controls should be run to ensure clean preparations.

1. Tissue dissociation
   1. For mammary glands, digest each mammary gland in 2 mL of complete Epicult-B Medium (Mouse) supplemented with 0.2 mL of gentle collagenase/hyaluronidase in a 15 mL conical tube and incubate overnight at 37 °C.
   2. Pellet dissociated cells at 450 × g for 5 min at room temperature.
   3. Carefully transfer the upper fatty layer using a bovine serum albumin-coated pipette tip to a new 15 mL conical tube and discard the remaining supernatant.
      NOTE: The upper fatty layer at this step often contains residual epithelial organoids that have not been separated and remain associated with the adipose tissue. Processing this fraction in a separate tube greatly increases epithelial cell recovery.
   4. Wash both the fatty fraction and the cell pellet once with 5 mL of phosphate-buffered saline, centrifuge at 450 × g for 5 min at room temperature, and discard the supernatants.
   5. Combine the washed fatty fraction and the cell pellet from the same mammary gland into a single 15 mL conical tube containing 5 mL of phosphate-buffered saline.
   6. Pellet the combined material at 450 × g for 5 min and discard the supernatant.
   7. For mammary tumors, mince tissue into small chunks using sterile scissors or a scalpel and transfer directly to a 2 mL microfuge tube.
2. Genomic DNA isolation
   ​NOTE: Genomic DNA extraction can be done using the Qiagen DNEasy Blood and Tissue kit following manufacturer instructions or using the following phenol-chloroform protocol.
   1. Prepare DNA lysis buffer by combining 5 mL of 1 M Tris-HCl (pH 8.0), 1 mL of EDTA (pH 8.0), 25 mL of 5 M NaCl, and 50 mL of 10% SDS, and bring the final volume to 500 mL with sterile distilled water.
   2. Add 1 mL of DNA lysis buffer and 10 μL of Proteinase K (20 mg/mL) to each sample and mix thoroughly by inversion.
   3. Incubate samples at 56 °C overnight to allow complete lysis.
   4. Centrifuge lysates at 16,000 × g for 5 min at room temperature to pellet insoluble debris.
   5. Transfer the clarified supernatant to a new 2 mL microcentrifuge tube.
      CAUTION: Phenol:chloroform:isoamyl alcohol is toxic and corrosive. Perform all steps involving organic solvents in a chemical fume hood while wearing appropriate personal protective equipment.
   6. Add an equal volume of phenol:chloroform:isoamyl alcohol and mix vigorously by vortexing for approximately 15 s until a homogeneous mixture is formed.
   7. Centrifuge at 16,000 × g for 10 min at 4 °C to separate the organic and inorganic phases.
   8. Carefully transfer the upper aqueous phase to a new tube without disturbing the interphase.
      NOTE: Avoid aspirating material from the white interphase to minimize protein and RNA contamination.
   9. Add 0.7 volumes of isopropanol to the recovered aqueous phase and mix vigorously by vortexing for approximately 15 s.
   10. Centrifuge at 16,000 × g for 10 min at 4 °C.
   11. Carefully discard the supernatant and rinse the DNA pellet with 500 μL of 70% ethanol.
   12. Centrifuge at 16,000 × g for 10 min at 4 °C and completely remove the ethanol.
   13. Air-dry the DNA pellet at room temperature for 10 min.
   14. Resuspend the DNA pellet in 50 μL of sterile distilled water.
   15. Incubate samples at 37 °C for 1 h or at 4 °C overnight to facilitate the full resuspension of genomic DNA.
   16. Measure the DNA concentration and purity by nanodrop or fluorometric quantification.

5. Next-generation sequencing of CRISPR libraries

NOTE: Generate a reference sample to enable calculation of gRNA fold changes over time. Transduced MEFs (Section 2) or mammary glands (Section 4) harvested 3 days post-infection, as well as the original plasmid maxiprep, can be used to determine guide representation in the starting library and viral stock.

1. PCR amplification of gRNA cassettes from genomic DNA
   1. Use primer pairs listed in Table 2 to uniquely index samples, enabling multiplexing and demultiplexing of up to 306 samples after sequencing.
   2. Prepare PCR reactions using 1 µg of genomic DNA as template in a total reaction volume of 50 µL containing 25 µL of 2x Q5 Master Mix, 2.5 µL of 10 µM forward primer, and 2.5 µL of 10 µM reverse primer.
   3. Include a no-template control reaction in parallel to monitor for contamination.
   4. Perform PCR amplification using the following cycling conditions, adjusting annealing temperature based on library-specific primer melting temperature:
      1 cycle of 98 °C for 30 s
      5 cycles of 98 °C for 10 s, 68 °C (-1 °C/cycle) for 10 s, and 72 °C for 15 s
      20 cycles of 98 °C for 10 s, 63 °C for 30 s, and 72 °C for 15 s
      1 cycle of 72 °C for 2 min
   5. Resolve PCR products on a 2.5% agarose gel to verify the correct amplicon size.
   6. Confirm the presence of bands at approximately 306 bp for pLKO-based sgRNA libraries and 430 bp for pXPR502-based dgRNA libraries (Figure 6).
      NOTE: A clear no-template control ensures the purity of the reagents and confirms that the assay is free from cross-contamination.
   7. Purify PCR products using a suitable DNA purification method and quantify DNA concentration using fluorometric quantification.
2. Next-generation sequencing
   1. Determine the required sequencing depth based on expected library complexity and sample clonality.
      NOTE: For tumors suspected to be clonal, approximately 1 million reads per sample are sufficient to accurately quantify guide representation. To calculate sequencing depth from mammary glands, multiply the number of guides in the library by 1,000-fold coverage, then multiply by 6 pg of genomic DNA/cell to calculate the total mass of genomic DNA per sample that should be PCR amplified. The total number of reads required for sequencing is equal to the number of guides in the library times 1,000.
   2. Pool purified libraries from different samples at equimolar concentrations.
   3. Sequence pooled libraries using a short-read Illumina sequencing platform.
3. Data analysis
   1. Demultiplex raw sequencing FASTQ files using sample-specific indices.
   2. Trim reads to the gRNA protospacer sequence using Trimmomatic⁶⁹. For CRISPR-A samples, trim reads to 14 bp, and for CRISPR-KO, trim reads to 20 bp from the end of the forward primer.
   3. Align trimmed reads to the corresponding gRNA reference library using Bowtie⁷⁰. Use alignment parameters of -v 2 and -m 1 for CRISPR-KO and -v 2 and -m 2 for CRISPR-A and mixed pools.
      NOTE: Typically, over 90% of the reads are successfully aligned. Low alignment typically results from trimming errors, incorrect formatting of input files, or sample contamination during PCR.
   4. Quantify aligned reads for each gRNA using the count function in MAGeCK⁷² to generate final gRNA count tables for downstream analysis.

Results

Here, we provide a protocol for pooled in vivo CRISPR screening in the mouse mammary gland. This pipeline can be used end-to-end to design and amplify pooled gRNA libraries from oligonucleotide arrays, clone the gRNA libraries into lentiviral backbones, produce high-titer lentivirus, perform intraductal delivery of pooled lentiviral libraries into the adult mammary gland, and recover gRNAs for sequencing-based deconvolution of library representation (Figure 1).

Successful amplification and cloning of pooled gRNA libraries are first assessed by agarose gel electrophoresis. PCR amplification of gRNA libraries yields discrete bands indicating uniform amplification from pooled oligonucleotide chip arrays (Figure 2A). Restriction digestion of the pXPR502 PPH-Cre and pLKO Cre backbones results in clear separation of the linearized backbone from the stuffer fragment (Figure 2B,C). Efficient digestion and clean recovery of the backbone are critical for minimizing background ligation and achieving high library complexity during cloning.

An important consideration for the use of CRISPR-KOALA is the efficiency of dgRNAs in combination with live Cas9 compared with other conventional approaches that use full-length sgRNAs and dead Cas9. We have previously shown that dgRNAs with live Cas9 are sufficient to induce gene activation at approximately 50-85% the efficiency of full-length sgRNAs⁵⁵. To compensate for this reduction in activation and to overcome the potential of more inefficient dgRNAs as compared to sgRNAs, we aim to include twice as many dgRNAs per gene in our library design. As this methodology is limited to using live Cas9 for gene knockout, we reasoned that this level of activation was sufficient for CRISPR screening. To test this, we performed dgRNA-mediated gene activation of several endogenous genes, including Erbb2, Esr1, and Plgrkt, and achieved robust expression at both the mRNA and protein level⁵⁵. Importantly, we regularly achieved a 3- to 10-fold increase in target gene expression, which is often more physiologically relevant than the 100- to 1,000-fold increases observed with several conventional CRISPR-A systems.

Library complexity following ligation and electroporation is evaluated by plating serial dilutions of transformed bacteria. Successful library cloning is indicated by dense and evenly distributed colonies in ligation reactions compared with few colonies on the backbone-only negative control plates (Figure 3). Colony numbers on dilution plates are used to estimate library representation, aiming for greater than 1,000-fold coverage (e.g., 1,000 bacterial colonies per gRNA in the library). Insufficient coverage or high background on the negative control plates indicates suboptimal ligation and/or electroporation. In this case, these steps should be repeated prior to the large-scale expansion of the bacterial cultures.

For intraductal mammary gland injections and pooled in vivo screening, lentiviral preparations must be concentrated to achieve sufficiently high functional titers. High-titer viral production can be validated by functional titration using LSL-tdTomato mouse embryonic fibroblasts (MEFs, JAX# 007909), as both the pLKO and pXPR502 backbones highlighted in this protocol express Cre. However, alternative methods for other backbones may be used as necessary (e.g., using a different fluorescent reporter or antibiotic selection within the lentiviral backbone). Efficient lentiviral preparations show a clear, dose-dependent induction of tdTomato expression with a linear range, enabling calculation of functional titers (Figure 4A–D). For pooled and multiplexed CRISPR-KO and CRISPR-A screens, lentiviruses are produced, concentrated, and titered separately, and functional titers are used to normalize and mix the two lentiviral libraries at defined ratios prior to infection.

Correct and efficient delivery of pooled lentiviral libraries into the adult mammary gland is required to optimally infect the mammary epithelium. Proper placement of the glass microcapillary with the nipple to allow access to the mammary duct (Figure 5A) is crucial to spread the lentivirus through the ductal tree. A suitable-sized needle tip allows loading of the lentivirus within the needle itself while enabling entry into the mammary duct with minimal resistance. Successful injections can be seen by uniform spreading of Fast Green dye throughout the mammary ductal tree, visible externally within the glands immediately post-injection (Figure 5B). In contrast, failed injections are characterized by pooling of dye beneath the nipple, reflecting delivery into the mammary fat pad rather than the mammary duct (Figure 5C). Immunohistochemical staining of mammary tissue sections can be used to confirm infection of epithelial cells (Figure 5D), and flow cytometry can be used to quantify the proportion of luminal and basal epithelial cell infections (Figure 5E), both validating the successful targeting of mammary epithelial cells following intraductal delivery.

Once genomic DNA has been recovered, it is PCR amplified for next-generation deep sequencing to quantify which gRNAs have been enriched in a particular sample. Following PCR, products are separated by gel electrophoresis, which should yield lentiviral-specific products (Figure 1 and Figure 6). Occasional non-specific amplification may be observed, particularly in CRISPR-A samples (Figure 6B), although this product is usually greater than 1,000 bp and well separated from the product of interest; however, the presence of a strong band at the expected size indicates successful library recovery. No-template control reactions should yield no detectable band, and the presence of amplification in these controls suggests contamination, which will compromise downstream library quantification (Figure 7A). If this contamination is observed, each primer stock should be tested using water as a template to identify the source of contamination, with those primer(s) being replaced immediately. Purified amplicons are pooled at equimolar ratios and sequenced on a short-read Illumina sequencing platform for downstream guide quantification.

Sequencing of libraries provides a clear metric of the success of library cloning and viral production. Optimally produced and quality-controlled libraries show an even representation across gRNAs with no (or very few) guides absent from transduced cells (Figure 7B). For large libraries, the dropout of a small number of gRNAs will not have a strong impact on the performance of the library if the remaining gRNAs for each gene are well-represented.

Thus far, this screening approach has been predominantly used to assess the propensity for single clonal genetic perturbations to drive mammary epithelial cell transformation and tumor outgrowth in tumor-sensitized genetic backgrounds^36,55. Dissection and sequencing of an individual clonal tumor often reveals clonal enrichment of a single gRNA, which can be classified as a screen hit (Figure 7C). In some cases, tumors have two or more gRNAs present at approximately equal proportions. These can either be the result of double infection or two clones growing together into a single tumor mass. It is sometimes possible to disentangle these two possibilities by sampling multiple areas of the same tumor, however this degree of granularity is usually not required. True hits are easily separated from multiple infections by their prevalence across many tumors in the screen, while random bystanders will not re-occur across multiple tumors. Our standard approach for calling hits across a screen is to set a read count cutoff of 80%, wherein any gRNA that possesses at least 80% of the read counts within a given sample is called a hit. Quantifying the number of these hits across all the sequenced tumors will identify which genes have the strongest phenotypic effects.

This method can also be used on mammary glands prior to tumor formation by assessing gRNA abundances through tools such as MAGeCK⁷¹. In this setting, this approach works well in the context of enrichment screening. While testing for depletion is possible, it requires much higher coverage, and the resulting data may be less reliable, as it is statistically challenging to determine whether a gRNA was depleted or simply was not present at the onset of the screen. CRISPR-StAR offers a valuable alternative technology in this regard^9,10 but needs to be optimized for in vivo screening in GEMMs.

Together, these representative results provide quality-control benchmarks for each major step of the pooled in vivo CRISPR screening workflow, from library cloning and viral production to intraductal delivery and sequencing-based deconvolution.

Figure 1: Schematic representation of the in vivo screening pipeline for the mouse mammary gland. This protocol includes the design of gRNA library oligonucleotide chips, library PCR and cloning, lentiviral preparations, intraductal mammary gland injections, and subsequent downstream analysis to deconvolute hits. Please click here to view a larger version of this figure.

Figure 2: gRNA library amplification and cloning. Representative agarose gel images for (A) gRNA library PCR reactions, (B) digested pXPR502 PPH Cre backbone, and (C) digested pLKO Cre backbone. Please click here to view a larger version of this figure.

Figure 3: Representative images of serial dilution of bacterial pool transduced with the gRNA library following electroporation. Please click here to view a larger version of this figure.

Figure 4: Lentiviral titering. (A–D) Representative flow cytometry for the calculation of viral titers from an example (A and B) low titer and (C and D) high titer viral prep. In both cases, LSL-tdTomato MEFs were transduced with Cre-expressing viruses as indicated in Section 2.5. (E) Representative results of count comparison between deep sequenced plasmid and transfected cells showing one poorly represented library (left), with a large variation in counts and many low-count gRNAs, compared to a well-represented library with a tighter distribution. Please click here to view a larger version of this figure.

Figure 5: Intraductal injection of the mammary gland. (A) View of the glass needle inserted into the nipple. (B) Successful injection indicated by dye permeating through the tree as viewed from the exterior of the mouse (left) or in a mammary gland dissected immediately after injection (right). (C) Failed mammary injection, indicated by pooling of dye underneath the nipple. (D) Mammary sections stained with anti-GFP showing infected cells. (E) Representative flow cytometry plots showing GFP+ infected luminal and basal cells. Please click here to view a larger version of this figure.

Figure 6: Sequencing-based deconvolution of samples. Representative deep sequencing gels from (A) MEF samples and plasmid and (B) tumor samples. Note that the 1 kb band for some CRISPR-A samples is a non-specific band and should be disregarded. The constructs contributing to the sample are indicated above the wells (KO for pLKO, and A for pXPR502). Variation in band intensity is normal, but a complete absence of a band indicates a failed reaction. Please click here to view a larger version of this figure.

Figure 7: Library representation and representative results of tumor deep sequencing. (A) Mapping results for a successfully sequenced tumor in comparison to a contaminated deep sequencing reaction, in which many reads cannot be mapped to the library. (B) Representative results showing sequencing from a plasmid, a clonal tumor, and a region from the mammary gland. Please click here to view a larger version of this figure.

Table 1: Library design and primers.Please click here to download this file.

Table 2: List of deep sequencing primers.Please click here to download this file.

Discussion

A key consideration for pooled in vivo CRISPR screening is achieving sufficient library coverage while maintaining sparse, single-cell transduction to preserve clonal resolution². In the mouse mammary gland, the epithelial compartment conservatively consists of approximately 3.5 × 10⁵ cells per gland³⁶. At an average lentiviral transduction rate of 15%, which minimizes the frequency of multiple infections per cell, intraductal delivery results in approximately 5 × 10⁴ transduced epithelial cells per gland^36,55. Bilateral injection of both the third thoracic and fourth abdominal mammary gland therefore, yields ~2 × 10⁵ transduced epithelial cells per mouse^36,55. Based on these parameters, a library containing 4,000 gRNAs (equating to ~1,000 genes for CRISPR-KO or ~500 genes for CRISPR-A), requires approximately 4 × 10⁶ transduced cells to achieve robust representation of 1,000-fold per gRNA, corresponding to a cohort size of approximately 20 mice. If a strong gain-of-function phenotype is expected, lower coverage (e.g., 400×, or 8 mice) might be sufficient. Alternatively, a higher infection rate of e.g. 30-35% might also be feasible, especially for larger libraries. On the other hand, experiments focused on basal mammary epithelial cells may require higher lentiviral titers or larger numbers of animals to ensure sufficient coverage, whereas this lentiviral-mediated approach is likely not amenable to stromal perturbations due to the extremely low infectivity of these cells³⁶. This quantitative framework provides a practical basis for experimental design by balancing library complexity, animal usage, and clonal resolution, and can be adapted to other screening contexts by adjusting estimated target cell numbers and tissue-specific transduction efficiency.

While this protocol enables scalable pooled in vivo CRISPR screening, several limitations should be considered when designing experiments. First, effective library coverage is inherently constrained by tissue size and achievable transduction efficiency, necessitating targeted rather than genome-wide libraries for most in vivo applications². Variability in intraductal injection efficiency between glands and/or animals can further reduce efficient coverage and should be mitigated through careful injection technique, consistent viral titering, and pilot experiments to estimate infection rates at a range of viral titers. Despite these potential sources of variability, we have recently shown that CRISPR screens with a library of 4,000 gRNAs reach near saturation in 20 mice following library deconvolution⁵⁵.

To reduce the effective number of mice required to maintain such robust coverages, it is in practice feasible to inject the second thoracic and fifth inguinal mammary glands in addition to the third thoracic and fourth abdominal mammary gland, as highlighted in this protocol. While the nipples of these additional glands are readily observed under the dissection scope during the procedure, the first cervical mammary gland is often covered and obscured by the forelimbs that are affixed to the nosecone. As each mammary gland differs from one another, both within and between pairs⁷², the number of mammary epithelial cells in these additional glands would need to be empirically determined to calculate effective coverage. An important consideration, however, is that inducing mammary tumors across eight mammary glands of a single mouse may result in a total cumulative tumor burden endpoint prior to the overt enrichment of all gRNA-containing clones. Alternatively, exciting new approaches to reduce the coverages required for in vivo CRISPR screening, such as CRISPR-StAR (Stochastic Activation by Recombination)^9,10 may be combined with our CRISPR-KOALA technology. CRISPR-StAR introduces both a unique lineage trace for each gRNA and an internally controlled negative control, which has been shown to reduce screening coverage requirements in xenograft settings by over two orders of magnitude⁹. Whether these principles may be applied to our CRISPR-KOALA technology remains to be tested, but offers an exciting opportunity to continue to reduce the number of animals required per screen.

In addition to enabling pooled in vivo CRISPR screening, this protocol simplifies intraductal delivery relative to existing approaches. Prior methods for direct access to the mammary ductal system have relied on surgical exposure of the gland⁵⁰ or the use of Hamilton syringes for injection⁵¹ which can be invasive and technically more cumbersome. In contrast, the non-surgical intraductal injection strategy described here uses fine glass microcapillary needles, allowing rapid, reproducible delivery volumes of concentrated lentivirus with minimal tissue disruption^36,55. This streamlined approach facilitates efficient generation of large numbers of independently perturbed epithelial clones and is particularly well suited for pooled lentiviral delivery in screening applications, where maximum viral spread is important.

One major strength of this pipeline is that the framework is highly adaptable across cell lines and tissue contexts⁵⁵. The principles of targeted gRNA library design, sparse lentiviral transduction, and sequencing-based deconvolution are equivalent to those used in conventional in vitro CRISPR screens². However, here we provide an additional platform to perform multiplexed CRISPR-KO and CRISPR-A within the same screen, which can be leveraged in vitro at a genome-wide scale, where higher cell numbers permit increased library complexities and deeper coverages. The use of the dead guide RNAs for CRISPR-A enables the use of the same Cre-inducible or constitutive Cas9 mouse to perform CRISPR-A and CRISPR-KO. Additionally, these experiments can be performed within the same mammary gland or multiplexed within the same cell using constructs that express an sgRNA as well as a dgRNA⁵⁵. As the CRISPR screening protocol relies on stable lentiviral integration and barcode recovery, it is also compatible with and can be adapted for alternative perturbation modalities, such as CRISPR-Inhibition⁶², Cas12^73,74, and ORF-based screens⁷⁵. In addition to screening, the same pipeline can be used to perturb individual genes by cloning individual gRNAs, siRNAs, or ORFs and delivering them as individual perturbations in the same manner^{36,55,67,76,77,78,79}.

Additionally, it will be of great interest to adapt this technology to other species, particularly the rat⁸⁰. The larger size of the rat mammary gland would permit the use of significantly expanded libraries while maintaining high coverage and infection efficiency⁸⁰. Furthermore, rat mammary biology more closely recapitulates certain features of the human gland than the mouse^{80,81,82,83,84,85}; for instance, while mouse mammary tumors are predominantly hormone-independent⁸³, rat mammary tumors are frequently hormone-dependent^81,82,85.

In summary, this protocol provides a practical and reproducible approach for pooled in vivo CRISPR knockout and activation screening in the adult mouse mammary gland. By integrating targeted library design with high-titer lentiviral production, simplified non-surgical intraductal delivery, and sequencing-based deconvolution, it lowers the technical barriers to in vivo functional genomics and enables direct interrogation of gene function within native tissue environments. This framework should facilitate broader adoption of pooled in vivo screening approaches and support systematic investigation of context-dependent genetic regulators of tissue homeostasis and disease.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
0.45 micron filter	Sigma	S2HVU02RE
12k or 92k oligo chip	Customarray Inc. (Genscript)
15 cm cell culture plates	Corning	CLS353004
293FT	Invitrogen	R70007
293NT	Systems Biosciences	LV900A-1
Alkaline phosphatase	NEB	M0290L
Amplicillin	Fisher Scientific	BP1760-25
ATP	NEB	9804S
ATP	NEB	P0756S
Cutsmart buffer	NEB	B6004S
Deep sequencing (Next-Seq or Hi-Seq)	Illumina
DNAesy Blood and Tissue DNA extraction kit	Qiagen	69506
Dulbecco’s modified Eagle medium	Wisent	319-015-CL
Endura electrocompetent cells	Lucigen	60242-1
EpiCult-B Mouse Medium Kit	Stemcell Technologies	05610
Esp3I	NEB	R0734L
Fetal Bovine Serum	Wisent	090-150
Gel DNA-cleanup kit	Zymo Research	D4008
Gene Pulser/MicroPulser Electroporation Cuvettes	BioRad	1652089
Gentle Collagenase/Hyaluronidase	Stemcell Technologies	7919
H11-Cas9	Jackson Labratories	JAX#028239
High-Speed Centrifuge	Beckman Coulter	MLS-50
Kim-wipe	Kimberly-Clark	34155
LB Agar	Wisent Technologies	800-011-LG
Micropipette puller	Sutter Instrument	P97
Mini-prep plasmid Kit	Frogga Bio	PDH300
NEBuffer 3.1 (Buffer for BsmBI)	NEB	R0580L
Oligo Clean and Concentrator	Zymo Research	D4061
Oligo cleanup kit	Zymo research	D4060
PAGE purified illumina sequencing primer	IDT DNA
PCR Micropipettes	Drummond	5-000-1001-X10
PEI (polyethyleneimine)	Sigma	408727-100ML
Penicillin and Streptinomycin	Wisent	450-201-EL
pLKO-Cre	Addgene	158032
pMD2.G	Addgene	12259
Poly-L-Lysine	Sigma	P2636-100MG
psPAX2	Addgene	12260
pXPR502-PPH-Cre	Addgene	256776
Q5 Polymerase 2x Master mix	NEB	M0494L
Qubit Fluorometric Quantification	Invitrogen	Q33327
R26-LSL-Cas9-EGFP	Jackson Labratories	JAX#024857
R26-LSL-TdTomato mice	Jackson Labratories	JAX#007909
SapI	NEB	R0569L
T4 DNA ligase	NEB	M0202L
Ultra-centrifuge tubes	Beckman Coulter	344058
Vacuum Filter Units	Fisher Scientific	SCHVU02RE