There is a substantial need for scalable CRISPR-based genetic screening methods that can be applied in mammalian tissues in vivo while enablingcell-type-specific analysis. Here we developed an adeno-associated virus (AAV)-based CRISPR screening platform, CrAAVe-seq, that incorporates a Cre-sensitive sgRNA construct for pooled screening within targeted cell populations in mouse tissues. We used this approach to screen two large sgRNA libraries, which collectively target over 5,000 genes, in mouse brains and uncovered genes essential for neuronal survival, of which we validated Rabggta and Hspa5. We highlight the reproducibility and scalability of the platform and show that it is sufficiently sensitive for screening in a restricted subset of neurons. We systematically characterize the impact of sgRNA library size, mouse cohort size, the size of the targeted cell population, viral titer, and coinfection rate on screen performance to establish general guidelines for large-scale in vivo screens.

Animals

All mice were maintained according to the National Institutes of Health guidelines and all procedures used in this study were approved by the UCSF Institutional Animal Care and Use Committee. Mice were housed on a 12-h light/dark cycle at 22-25 °C, 50-60% humidity, and had food and water provided ad libitum. Mice were randomly assigned for the experimental groups at time of injection and both male and female mice were used. In accordance with approved protocol, mice were monitored post injection and if signs of distress appeared, mice were documented and euthanized promptly. The mice used in this study are LSL-dCas9-KRAB (LSL-CRISPRi) mice (B6;129S6-Gt(ROSA)26Sor^{tm2(CAG-cas9*/ZNF10*)Gers}/J, RRID: IMSR_JAX:033066)¹⁷ and dCas9-KRAB mice (B6.Cg-Igs2^{tm1(CAG-mCherry,-cas9/ZNF10*)Mtm}/J, RRID: IMSR_JAX:030000). A summary of the individual mice used for CRISPR screening and select in vivo experiments is provided in Supplementary Table 1.

Plasmids

The screening vector pAP215 is shown in Fig. 1a (fully annotated map on  Addgene plasmid # 217635). Details on cloning pAP215 are in Supplementary Methods. Additional plasmids in this study include pENN.AAV.hSyn.HI.eGFP-Cre.WPRE.SV40 (Addgene # 105540, a gift from James M. Wilson), pENN.AAV.CamKII.HI.GFP-Cre.WPRE.SV40 (Addgene # 105551, a gift from James M. Wilson), CN1851-rAAV-hI56i-minBglobin-iCre-4X2C-WPRE3-BGHpA (Addgene # 164450, a gift from The Allen Institute for Brain Science & Jonathan Ting) (PMID: 33789083), and pAAV-FLEX-GFP (Addgene plasmid # 28304, a gift from Edward Boyden). The NLS-mScarlet and NLS-mNeonGreen AAV plasmids were generated by restriction cloning, replacing the GFP sequence in plasmid CAG-NLS-GFP (Addgene # 104061, a gift from Viviana Gradinaru)¹⁸, and replacing the NLS sequence with one from the pMK1334 plasmid¹. 

sgRNA cloning

We transferred the sgRNA sequences from our pooled mCRISPRi-v2 sgRNAs, subpools M1-top 5 (targeting Kinases, Phosphatases, and Drug Targets) and M3-top 5 (targeting the proteostasis network)²² into the pAP215 plasmid backbone to create the pAP215-M1 and pAP215-M3 sgRNA libraries, respectively. The mouse chaperone targeting library was designed by selecting the mouse orthologs of a human chaperone targeting library that we previously developed³¹ and included 350 non-targeting control sgRNAs. Oligonucleotide pools were synthesized by Agilent, amplified by PCR, and cloned into the pAP215 backbone. Steps for library cloning is detailed in Supplementary Methods.

Individual sgRNAs were cloned in the pAP215 backbone digested with BstXI and Bpu1102I using annealed oligonucleotides (Integrated DNA Technologies.) with compatible overhangs. The protospacer sequences for the specific sgRNAs used in this study include sgCreb1 (GGCTGCGGCTCCTCAGTCGG), sgHspa5 (GAACACTGACCTGGACACTT’), sgRabggta (GCGGCGAACTCACCTGCTCA), and a non-targeting control (sgNTC) (GGATGCATAGATGAACGGATG). 

AAV packaging, purification, and titering

 To generate AAV for in vivo injections, two 15-cm dishes were each seeded with 1.5×10⁷ HEK 293T cells (ATCC, CRL-3216) in 25 ml DMEM complete medium: DMEM (Gibco, 11965-092) supplemented with 10% FBS (VWR, 89510, lot: 184B19), 1% penicillin-streptomycin (Gibco, 15140122), and 1% GlutaMAX (Gibco, 35050061). The next day, 20 µg of pAdDeltaF6 (Addgene # 112867, a gift from James M. Wilson), 7 µg of library plasmid, 7 µg of pUCmini-iCAP-PHP.eB (Addgene # 103005, a gift from Viviana Gradinaru)¹⁸, and 75 µl of 1 mg/ml polyethenylamine (PEI) (Linear, MW 25,000, Polysciences, 23966) were diluted into 4 ml of Opti-MEM (Gibco, 31985062), gently mixed, and incubated at room temperature for 10 min. The PEI/DNA transfection complex was then pipetted drop-wise onto the HEK 293T cells. After 24 hours, the media was replaced with 27 ml of fresh Opti-MEM.

72 hours after transfection, AAV precipitation was performed as previously described³², with modifications. Cold 5× AAV precipitation solution (40% polyethylene glycol (Sigma-Aldrich, 89510) and 2.5 M NaCl) was prepared. The cells and media were triturated and collected (~30 ml) into a 50 ml conical tube, followed by addition of 3 ml chloroform and vortexing for approximately 30 seconds. The homogenate was centrifuged at 3,000g for 5 min at room temperature, and the aqueous (top) phase was transferred to a new 50 ml conical tube and 5× AAV precipitation solution was added to a final 1× concentration, followed by incubation on ice for at least 1 hour. The solution was centrifuged at 3,000 × g for 30 min at 4°C. The supernatant was completely removed and the viral pellet was resuspended in 1 ml of 50 mM HEPES and 3 mM MgCl₂, and incubated with 1 µl DNase I (New England Biolabs, M0303S) and 10 µl RNase A (Thermo Scientific, EN0531) at 37°C for 15 min. An equal volume of chloroform was added, followed by vortexing for 15 sec, and centrifuged at 16,000g for 5 min at RT; this step was repeated once. Using 400 µl at a time, the aqueous phase was passed through a 0.5-ml Amicon Ultra Centrifugal Filter with a 100 kDa cutoff (Millipore, UFC510024) by 3 min of centrifugation at 14,000g, followed by buffer exchange twice with 1× DPBS. Titering was performed by quantitative RT-PCR as previously described³³ using primers (Integrated DNA Technologies) listed in Supplementary Table 4. This method of AAV production is available as a companion protocol on Protocols.io³⁴.

To prepare AAV for testing in primary neuronal cultures (for longitudinal imaging and qRT-PCR), HEK293T cells were seeded into a 6-well format containing 1.5 ml of DMEM complete media. The cells were transfected with 1 µg pAdDeltaF6, 350 ng pUCmini-iCAP-PHP.eB, and 350 ng of AAV transgene using PEI as above. Approximately 48 hours after transfection, the cells and media were collected in 2 ml microfuge tube, 200 µl of chloroform was added to each tube, vortexed for 15 sec, and centrifuged at 16,000 × g for 5 min at room temperature. The aqueous (top) phase was transferred to a new tube and AAV precipitation solution was added to 1× dilution, and incubated on ice for at least one hour. The precipitated AAV was centrifuged at 16,000 × g for 15 min at 4°C, the supernatant was removed, the pellet was resuspended in 100 µl of 1× PBS, and centrifuged again at 16,000 × g for 1 min to remove excess debris, and the supernatant (purified virus) was transferred to a new microfuge tube. 10 µl purified virus was used per well in primary neuronal cultures in a 24-well format.

Intracerebroventricular injection

Intracerebroventricular (ICV) injections were performed as previously described, with minor modifications³⁵. Briefly, neonatal mice were placed on a gauze-covered frozen cold pack and monitored for complete cryoanesthesia. The scalp was gently cleaned with an alcohol swab. AAVs were diluted in 1× PBS with 0.1% trypan blue into a 2 µl final volume per mouse and loaded into 10 µl syringe (Hamilton, 1701). The syringe was equipped with a 33-gauge beveled needle (Hamilton, 7803-05, 0.5 inches in length). The needle was inserted through the skull 2/5 of distance of the lambda suture to the eye and to a depth of 3 mm to target the left lateral ventricle. Following a one-time unilateral injection, the neonate was placed on a warming pad for recovery and returned to the parent cage. The number of viral particles injected in each mouse is listed in Supplementary Table 1.

sgRNA recovery and sequencing for CrAAVe-seq

Animals were euthanized using CO₂, and their whole brains were removed and stored at -80°C. The sex of the mice was recorded prior to euthanasia.

Initial protocol for sgRNA recovery

This protocol for episome recovery was used in the following figures: Fig. 1d, M1 library screen in Fig 3, the hSyn1-Cre screen in Fig 5d, the hSyn1-Cre versus no Cre screens in Extended Data Fig. 6, and the screens in Extended Data Fig. 7.  Each brain was placed in a PYREX 7 ml Dounce Homogenizer (Corning, 7722-7) with 2 ml of TRIzol (Invitrogen, 15596026) and thoroughly homogenized using the A pestle (0.0045 nominal clearance) for 10 or more strokes.  0.4 ml of chloroform was added, vigorously shaken for 30 seconds, and centrifuged at 12,000g for 15 min at 4°C. The aqueous phase (top) was collected and nucleic acids precipitated using 1 ml isopropanol, incubated on ice for 10 min, and centrifuged at 12,000g for 10 min at 4°C. The supernatant was discarded and the pellet was washed in 2 ml of 75% ethanol in DNase/RNase-free water and spun down at 7,500g for 5 min. The supernatant was then removed and the pellet was allowed to air dry for 10 mins, and then resuspended in 100 µl of DNase/RNase-free water and incubated with 1 µl of RNase A (Thermo Scientific, EN0531) at 37ºC overnight. The sample was then column purified by Zymo DNA Clean & Concentrator-25 kit (Zymo Research, D4033) and eluted in 50 µl of DNase/RNase-free water to yield recovered viral DNA.  The remaining RNAse-treated samples were considered recovered episomes for use in PCR below.

Optimized protocol for sgRNA recovery

An optimized protocol for episomal sgRNA recovery was used in the following figures: the M3 library screen in Fig. 3, the CaMKII-Cre screens in Fig 4., the hi56i-Cre screens in Fig. 5, and for the dPCR experiments in Fig 5. All steps in this protocol are the same as the above initial protocol except for two modifications. First, each brain was homogenized in 4 ml of TriZOL, phase separated using 0.4 ml chloroform, and the aqueous phase precipitated with 2 ml isopropanol, before resuspending in 100 µL of DNase/RNase-free water. Second, following overnight RNase A treatment as above, the sample was directly transferred to -20°C without column purification. The optimized protocol is available as a companion protocol on Protocols.io¹⁴.

PCR-amplification of sgRNAs and sequencing

The PCR was performed using Q5 High-Fidelity 2× Master Mix (NEB, M0492L). Each reaction contained 100 µL of recovered episomes, 110 µL of Q5 2× master mix, and 5.5 µL of each primer. For amplification of the AAV sgRNA libraries, the purified AAV was diluted 10-fold into H2O, and 1 µL of the diluted AAV was used as a template in a 100 µL PCR reaction. The reaction was distributed into PCR tubes at the maximum volume allowed by the PCR equipment. The following PCR cycling conditions were used: 98°C 30s, (98°C 30s, 60°C 15s, 72°C 15s) × 23 cycles, 72°C 10min. 

100 µL of each PCR reaction was purified using 1.1× SPRI beads (SPRIselect Beckman Coulter, B23317) and resuspended in 25 µL elution buffer (Machery Nagel, 740306). The purified products were pooled and sequenced on an Illumina HiSeq4000 at the UCSF Center for Advanced Technologies or on an Illumina NextSeq2000 and demultiplexed with Illumina Dragen BCL Convert. The amplification primers (with adapters) and custom sequencing primers (Integrated DNA Technologies) are listed in Supplementary Table 4. See Supplementary Information for details for digital PCR experiments.

Mouse cortical neuron primary cultures and immunocytochemistry

Neonates were briefly sanitized with 70% EtOH and decapitated using sharp scissors, and the brains were removed and placed into cold HBSS (Gibco, 14175095). The meninges were removed under a dissecting microscope, and the cortices were transferred to a 15-ml conical tube containing 10 ml of 0.25% Trypsin-EDTA (Gibco, 25200056) and incubated at 37°C for 30 min. The trypsin was removed and the brains were gently rinsed twice in 5 ml of DMEM complete media, followed by trituration of brains in 5 ml of DMEM complete media filtered through a 40 µm nylon cell strainer (Corning, 352340), and diluted into DMEM complete media in a volume as needed for plating. An equivalent of one brain was plated across each BioCoat Poly-D-Lysine 24-well TC-treated plate (Corning, 356414). The following day, day in vitro 1 (DIV1), the DMEM complete media was replaced with neuronal growth media composed of Neurobasal-A Medium (Gibco, 10888022), 1× B-27 Supplement minus vitamin A (Gibco, 12587010), GlutaMAX Supplement (Gibco, 35050079), and 1% penicillin-streptomycin (Gibco, 15140122). On DIV2, the cultures were further supplemented with cytarabine (AraC) to a final concentration of 200 µM (Thermo Scientific Chemicals, 449561000). The primary neuronal cultures were transduced with AAV on DIV4 and imaged starting 4 days after transduction. See Supplementary Information for details on RNA isolation, quantitative RT-PCR, live-cell imaging, and quantification of cell death. 

Mouse brain immunofluorescence staining

Whole brains were removed and fixed overnight at 4ºC in 4% paraformaldehyde (Electron Microscopy Sciences, 15710) diluted in 1× PBS. The following day, the fixative was replaced with 30% sucrose dissolved in 1× PBS for at least 48 hours. Fixed brains were blotting dry, cut down the midline with a razorblade, and mounting into a cryomold (Epredia, 2219) using OCT compound (Sakura Finetek, 4583). To snap freeze, cryomolds were partially submerged in a pool of 2-propanol cooled by a bed of dry ice. Brains were sectioned in the sagittal plane at 40 µm on a cryostat (Leica, CM1950) with a 34° MX35 Premier+ blade (Epredia, 3052835). The resulting brain sections were stored free-floating in 1× PBS + 0.05% NaN₃ at 4ºC. When ready for staining, representative brain sections were wasted three times in 1× PBS and incubated in a 24 well plate at room temperature for one hour in blocking buffer: 10% goat serum (Gibco, 16210064), 1% BSA (Sigma-Aldrich, A7906), and 0.3% Triton X-100 (Sigma-Aldrich, T8787) diluted in 1× PBS. The brain sections were incubated in primary antibodies diluted in blocking buffer overnight at 4ºC on a gentle shaker. The sections were washed three times in 1× PBS, then incubated in secondary antibodies for 2 hours at room temperature in the dark on a gentle shaker. Sections were washed three times in 1× PBS and moved to charged glass microscope slides (Fisher Scientific, 12-55015). After PBS was removed, Fluoromount-G with DAPI mountant (Invitrogen, 00-4959-52) was added, and a No. 1.5 coverslip (Globe Scientific, 1415-15) was applied. Slides were dried at room temperature in the dark overnight and sealed with nail polish. For experiments without DAPI, ProLong Gold Antifade mountant (Invitrogen, P10144) was used instead. For experiments with Hoechst instead of DAPI, sections were lastly incubated for 15 mins in Hoechst 33342 (BD Pharmingen, 561908) diluted 2 µg/ml in 1× PBS, then washed 3 times in 1× PBS before mounting using ProLong Gold mountant.

The following primary antibodies were used: rabbit anti-CREB (1:1,000 dilution, clone: 48H2, Cell Signaling Technologies, 9197), rabbit anti-SOX9 (1:2,000 dilution, polyclonal, EMD Millipore, AB5535), guinea pig anti-NeuN (1:500 dilution, polyclonal, Synaptic Systems, 266004), alpaca FluoTag-Q anti-TagFP nanobody (reacts to mTagBFP2 but not eGFP, 1:500 dilution, clone: 1H7, Alexa647 pre-conjugated, NanoTag Biotechnologies, N0501-AF647-L. The following secondary antibodies were used: goat anti-rabbit IgG Alexa Fluor 488 (1:1,000 dilution, Invitrogen A-11034), goat anti-rabbit IgG Alexa Fluor 568 (1:1,000 dilution, Invitrogen, A-11011), goat anti-rabbit IgG Alexa Fluor 647 (1:1,000 dilution, Invitrogen, A-21245), goat anti-guinea pig IgG Alexa Fluor 488 (1:1,000 dilution, Invitrogen, A-11073), goat anti-guinea pig IgG Alexa Fluor 647 (1:1,000 dilution, Invitrogen, A-21450). All secondary antibodies were highly cross-absorbed. See Supplementary Information for imaging parameters and quantification.

CRISPR Screen Analysis

Computational analysis of screen data was carried out using a newly developed bioinformatics pipeline which is publicly available (see Code availability section). Raw sequencing results were mapped to the M1 protospacer library using `sgcount` (Ref. ³⁶). Briefly, `sgcount` is a tool to match protospacers against a reference protospacer library with exact pattern matching. The resulting count matrices, containing guide and gene information along with count data for each sample, was used as input for subsequent analyses.

`crispr_screen` (Ref. ³⁷) was used to perform differential sgRNA abundance analysis and gene score aggregation analysis. `crispr_screen` is a reproduction of the original MAGeCK analysis but performs differential sgRNA analysis using a negative binomial as originally described in the study and not a truncated normal distribution as used in the current MAGeCK implementation.

In brief, sgRNA abundances are median normalized across samples then a weighted linear regression (weighted ordinary least squares) is used to fit the log-variance to the log mean of the control samples (representing sgRNA abundances in the AAV library). The fit variance and mean are then used to parameterize negative binomial distributions for each sgRNA and a survival function or cumulative distribution function is used to calculate a p-value for sgRNA under- and over-abundance. We excluded any sgRNAs that were represented with fewer than 100 reads across the control AAV samples.

To calculate a gene-level aggregated metric across sgRNAs of the same gene group we established a novel algorithm, geopagg. We performed the following operations on the under- and over-abundance p-values in parallel. First, the differential abundance p-values for sgRNAs were corrected for multiple hypothesis correction using the Benjamini-Hochberg correction procedure to calculate a false discovery rate (FDR) for each sgRNA. Next, FDRs for sgRNAs belonging to the same gene grouping were collected and sorted ascendingly. We then calculated a weighted geometric mean FDR ( ) for each gene ( ) across the FDRs ( ) for sgRNAs within the gene group ( ).

We calculated a weighted geometric mean to down-weight the relative impact of the first sgRNA within the group using a “Drop-First” weighting strategy. The first sgRNA (or top performing sgRNA) is down-weighted because we generally aim to capture genes with multiple high-performing sgRNAs. The weights for each gene grouping (  are defined as follows:

We also performed an aggregation of the log₂-fold-changes in abundance (a gene’s phenotype score) of each sgRNA ( ) within the gene group with an arithmetic mean:

We then created random groupings of non-targeting control sgRNAs, which we denote as the amalgam gene set ( ), to match the gene membership distribution of the input sgRNA library. This was performed by determining the membership size (number of sgRNAs) of each gene ( ) and sampling an equal amount of sgRNAs without replacement from the non-targeting controls. We next performed an identical calculation as above for each of the newly created amalgam genes.

We then calculated a ‘gene score’ for each gene and each amalgam gene within the dataset using the calculated weighted geometric mean of the FDR values ) and their phenotype score ( ).

We next sort the gene scores ( ) in an ascending order or a descending order for the under- and over-abundant tests respectively.

Finally, we calculate an empirical false discovery rate ( ) by stepping through the weighted geometric mean ( ) arrays and determining for each rank ( ) how many amalgam genes ( ) are preceding it. Because the true empirical false discovery rate will be zero for all genes preceding the first amalgam gene, we provide a non-zero score by constraining the reported false discovery rate to be the maximum of the empirical false discovery rate and the weighted geometric mean of that gene

This empirical false discovery is further constrained for explicit monotonicity by requiring the current score to be greater than or equal to the previous one.

The geopagg algorithm is performed for the sgRNA under- and over-abundant p-values in parallel and the final scores for each gene are reported as the most significant of the two tests.

151 genes in the M1 library and 129 genes in the M3 library are targeted at two different transcriptional start sites by different sets of sgRNAs. These sets were evaluated independently with a label of P1 and P2 (e.g. GeneA_P1 and GeneA_P2). In cases where only one set is significant and labeled on a heatmap or volcano plot, the P1 or P2 label is not shown, but this information is included in Supplementary Tables 2 and 3.

Details on the bootstrapping analysis is provided in Supplementary Information.

sgRNA library cloning

20 µg of the M1- or M3-CRISPRiv2 sgRNA library was digested with BstXI (Thermo Scientific, FD1024) and Bpu1102I (Thermo Scientific, FD0094). The guide-encoding inserts (84 bp) were resolved on a 4-20% Novex TBE gel (Invitrogen, EC62252BOX) and precipitated with GlycoBlue and sodium acetate. Inserts were washed with ethanol after precipitation and then eluted in DNase- and RNase-free water. 20 µg of the backbone vector, pAP215, was digested in parallel with BstXI and Bpu1102I, resolved on a 1% agarose gel, and purified from the gel (Zymo Research, D4001). The vectors and insert guides were annealed for 16 hrs overnight using T4 ligase (New England Biolabs, M0202L) at a 1:2 molar ratio of vector to insert, and then purified with sodium acetate and ethanol washing. After the final wash, a portion of the ligated library product was transformed into chemically competent E. coli (Takara, 636763) and 10 colonies were picked at random to ensure that each colony contained a unique sgRNA sequence. The remainder of the library product was electroporated into Mega-X competent cells (Invitrogen, C640003) and grown overnight, and a portion of the culture was plated to determine if a coverage of at least 250 colonies per guide was achieved, followed by growth of the remainder of the culture in 1 L of LB for 16 hrs and purification of the library using ZymoPURE II Plasmid Gigaprep Kit (Zymo Research, D4204).

Microscopy, image segmentation, and analysis

Slides containing brain sections were imaged using a Zeiss AxioScan.Z1 with a Zeiss Colibri 7 unit, ×20/0.8 NA objective lens, 5-30 ms exposure, 1×1 binning and 25-100% intensity using 425-nm, 495-nm, 570-nm and 655-nm lasers, running ZEN version 2.6 software. The images were imported into QuPath (version 0.4.2) for analysis⁴⁵. The raw CZI files are available on Dryad repository (see Data Availability).

To identify overlap between BFP, NeuN, and SOX9, a representative region of the cortex was outlined and the nuclei were segmented on the DAPI channel using the ‘Cell detection’ module without expansion of the nuclei to develop virtual cell boundaries. Classifiers were created to distinguish BFP⁺, NeuN⁺, and SOX9⁺ cells, and applied sequentially. Cells containing overlapping NeuN and SOX9 were considered to be neurons (as there was a low, but detectable signal in the SOX9 channel in all nuclei with this antibody) and only cells exclusively containing SOX9 were considered astrocytes. Similar segmentation on DAPI and sequential application of classifiers were used to examine overlap between nuclear BFP, mNeonGreen, and mScarlet signal in mice shown in Fig. 2d,e.

To evaluate CREB1 levels, a representative region was selected as indicated by specific brain regions, and the nuclei were segmented on the DAPI channel as above. The measurements for the segmented nuclei were exported. The mean fluorescence intensity for the anti-Creb1 channel was obtained selected by the top 200 nuclei of highest anti-mTagBFP2 fluorescence intensity. A representative region of brain stained with secondary antibodies only was selected to determine the background mean fluorescence intensity for that channel.  The same segmentation was performed in mice injected with FLEX-GFP, with the top 2% and bottom 2% of GFP⁺ or BFP⁺ nuclei examined for CREB1 mean fluorescence intensity.

For mouse primary neurons transduced with AAV, live imaging was performed every other day using an ImageXpress Micro Confocal HT.ai High-Content Imaging System (Molecular Devices). The imaging chamber was warmed to 37ºC and equilibrated with 5% CO₂. The system used an Andor Zyla 4.5 camera with a Plan Apo ×10/0.45NA objective lens, an 89 North LDI laser illumination unit, 10-500 ms exposure time, 1×1 binning, and 10% laser intensity using 405-nm, 475-nm, and 555-nm lasers, running MetaXpress (version 6.7.1.157). Resulting images were imported into Cell Profiler (version 4.2.1)⁴⁶ and analyzed using a custom pipeline. hSyn1-Cre-GFP+ nuclei were segmented using the ‘IdentifyPrimaryObjects’ module, with expected diameter 8-40 pixels, using an Adaptive threshold (size 50) and the Minimum Cross-Entropy method, with a 1.5 smoothing scale, 1.0 correction factor, and lower- and upper-bound threshold at 0.435 and 1, respectively. Segmented objects were exported, and counted in each field, then summed across all fields within a well to calculate the number of objects per well (n=29 fields per well, n=4 wells per condition), using a custom R script. This was repeated for each timepoint. Data was normalized to fluorescent intensity at day 8 (as before that day, fluorescence intensity increased linearly with time in all channels as cells manufactured fluorescent proteins) and percentage change was calculated for each well from day 8, for subsequent timepoints through day 16.

A similar protocol was used to analyze Rabggta knockdown data with some modifications. hSyn1-Cre-GFP⁺ nuclei were segmented using the ‘IdentifyPrimaryObjects’ module, with an expected diameter of 7-40 pixels, using an Adaptive threshold (size 50) and Minimum Cross-Entropy method, with a 1.3488 smoothing scale, 1.0 correction factor, and lower- and upper-bound threshold at 0.101 and 1, respectively. Segmented objects were exported and counted in each field, then summed across all fields within a well to calculate the number of objects per well (n = 29 fields per well, n = 3) using a custom R script. This was repeated for each timepoint. Data was normalized to fluorescent intensity at day 10 and percentage change was calculated for each well from day 10, for subsequent time points through day 26. Data was plotted using Prism GraphPad. Example images were created for Fig. 8e and 8j by importing into FIJI and applying the “red hot” LUT evenly across all images. This LUT is linear and covers the full range of the data.

Statistics and Reproducibility

No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications, as cited in the main text. Numbers of replicates are listed in each figure. In all figures that show a representative fluorescence micrograph, the experiments were repeated at least once to verify similar findings. No repeat measurements were made on the same samples. Data were assumed to be normally distributed except for instances within the `crispr_screen` pipeline where geopagg uses a negative binomial distribution to calculate p-value. The `crispr_screen` pipeline was used with FDR < 0.1 and controls for multiple comparisons using the Benjamini-Hochberg correction. For cell culture experiments, randomization was not performed because treatment groups of cells were derived from the same parent population of cells. Data collection and analysis were not performed blinded to the conditions of the experiments. No animals or data points were excluded from the relevant analyses. Major findings were validated using independent samples and orthogonal approaches.

Data Availability

All data are publicly available at the Dryad data repository (DOI: 10.5061/dryad.0k6djhb9t)³⁸. Data from the DepMap database used to generate Extended Data Fig. 6a is publicly accessible (https://depmap.org/portal). There are no restrictions on data availability.

Code Availability

For CRISPR screen analysis, we developed a highly efficient analysis toolkit called `sgcount` for sgRNA mapping and `crispr_screen` for differential gene abundance testing. The sgRNA mapping utility (`sgcount`, version 0.1.32) is also available on GitHub (https://github.com/noamteyssier/sgcount) and Zenodo (https://zenodo.org/doi/10.5281/zenodo.12774352)³⁶. The differential gene abundance tool (`crispr_screen`, version 0.2.8) is also available on GitHub  (https://github.com/noamteyssier/crispr_screen/) and Zenodo (https://zenodo.org/doi/10.5281/zenodo.12774208)³⁷. All bootstrapping analyses were performed using a custom python package (`rescreener`, version 0.1.0) available on GitHub (https://github.com/noamteyssier/bootstrap_analysis_invivo_crispr_screen).

The R notebooks for analysis are available at https://kampmannlab.ucsf.edu/article/scripts-vivo-screening-manuscript.

The CellProfiler pipelines are available on the Dryad data repository (DOI: 10.5061/dryad.0k6djhb9t)³⁸ and at https://kampmannlab.ucsf.edu/article/scripts-vivo-screening-manuscript.

For full methodological details, see the associated paper:

*Ramani B, *Rose IVL, Teyssier N, Pan A, Danner-Bocks S, Sanghal T, Yadanar L, Tian R, Ma K, Palop JJ, & Kampmann M. CRISPR screening by AAV episome-sequencing (CrAAVe-seq): a scalable cell type-specific in vivo platform uncovers neuronal essential genes. Nature Neuroscience (2025). 

References

31. Biebl, M. M. et al. NudC guides client transfer between the Hsp40/70 and Hsp90 chaperone systems. Mol Cell82, 555-569.e7 (2022).

32. Negrini, M., Wang, G., Heuer, A., Björklund, T. & Davidsson, M. AAV Production Everywhere: A Simple, Fast, and Reliable Protocol for In-house AAV Vector Production Based on Chloroform Extraction. Current Protocols in Neuroscience 93, e103 (2020).

33. Aurnhammer, C. et al. Universal real-time PCR for the detection and quantification of adeno-associated virus serotype 2-derived inverted terminal repeat sequences. Hum Gene Ther Methods 23, 18–28 (2012).

34. Rose, I. V. L., Ramani, B., Pan, A. & Kampmann, M. Protocol for AAV production for in vivo CRISPR screening in the mouse brain. Protocols.io (2025).

35. Kim, J.-Y., Grunke, S. D., Levites, Y., Golde, T. E. & Jankowsky, J. L. Intracerebroventricular Viral Injection of the Neonatal Mouse Brain for Persistent and Widespread Neuronal Transduction. JoVE (Journal of Visualized Experiments)e51863 (2014) doi:10.3791/51863.

36. Teyssier, N. sgcount. Zenodo https://doi.org/10.5281/zenodo.12774353 (2024).

37. Teyssier, N. crispr_screen. Zenodo https://doi.org/10.5281/zenodo.12774305 (2024).

38. Ramani, B. et al. Datasets from CRISPR screening by AAV episome-sequencing (CrAAVe-seq): a scalable cell type-specific in vivo platform identifies neuronal essential genes. Dryad https://doi.org/10.5061/dryad.0k6djhb9t (2025).

39. Schnepp, B. C., Jensen, R. L., Chen, C.-L., Johnson, P. R. & Clark, K. R. Characterization of Adeno-Associated Virus Genomes Isolated from Human Tissues. Journal of Virology 79, 14793–14803 (2005).

40. Penaud-Budloo, M. et al. Adeno-Associated Virus Vector Genomes Persist as Episomal Chromatin in Primate Muscle. Journal of Virology 82, 7875–7885 (2008).

41. Wang, J.-H., Gessler, D. J., Zhan, W., Gallagher, T. L. & Gao, G. Adeno-associated virus as a delivery vector for gene therapy of human diseases. Sig Transduct Target Ther 9, 1–33 (2024).

42. Yau, E. H. & Rana, T. M. Next-Generation Sequencing of Genome-Wide CRISPR Screens. Methods Mol Biol1712, 203–216 (2018).

43. Mathiowetz, A. J., Roberts, M. A., Morgens, D. W., Olzmann, J. A. & Li, Z. Protocol for performing pooled CRISPR-Cas9 loss-of-function screens. STAR Protoc 4, 102201 (2023).

44. Ye, L. et al. In vivo CRISPR screening in CD8 T cells with AAV-Sleeping Beauty hybrid vectors identifies membrane targets for improving immunotherapy for glioblastoma. Nat Biotechnol 37, 1302–1313 (2019).

45. Duan, Y. et al. The Clustered, Regularly Interspaced, Short Palindromic Repeats-associated Endonuclease 9 (CRISPR/Cas9)-created MDM2 T309G Mutation Enhances Vitreous-induced Expression of MDM2 and Proliferation and Survival of Cells. J Biol Chem 291, 16339–16347 (2016).

46. Kabadi, A. M., Ousterout, D. G., Hilton, I. B. & Gersbach, C. A. Multiplex CRISPR/Cas9-based genome engineering from a single lentiviral vector. Nucleic Acids Res 42, e147 (2014).

47. Levy, J. M. et al. Cytosine and adenine base editing of the brain, liver, retina, heart and skeletal muscle of mice via adeno-associated viruses. Nat Biomed Eng 4, 97–110 (2020).

48. Huijbers, I. J. et al. Using the GEMM-ESC strategy to study gene function in mouse models. Nat Protoc 10, 1755–1785 (2015).

49. Gilbert, L. A. et al. Genome-Scale CRISPR-Mediated Control of Gene Repression and Activation. Cell 159, 647–661 (2014).

50. Bankhead, P. et al. QuPath: Open source software for digital pathology image analysis. Sci Rep 7, 16878 (2017).

51. Stirling, D. R. et al. CellProfiler 4: improvements in speed, utility and usability. BMC Bioinformatics 22, 433 (2021).

52. Wang, G. et al. Mapping a functional cancer genome atlas of tumor suppressors in mouse liver using AAV-CRISPR-mediated direct in vivo screening. Sci Adv 4, eaao5508 (2018).