Scalable single-cell pooled CRISPR screens with conventional knockout vector libraries

23 Current methods for single-cell RNA profiling of pooled CRISPR screens are limited, either by 24 indirect capture of single guide RNAs (sgRNAs) or by custom modification of plasmid libraries. 25 Here, we present a direct sgRNA capture platform called Native sgRNA Capture and 26 sequencing (NSC-seq) that enables single-cell CRISPR screens using common knockout 27 plasmid libraries, facilitating genotype-phenotype mapping at multiple scales in vitro and in vivo. 28 Additionally, we characterize sgRNA expression in three whole-genome knockout libraries, 29 revealing a substantial subset of truncated (isoform) spacer reads. We provide this dataset as a 30

of expressed sgRNA isoforms that may potentially have compromised CRISPR gene 31 editing efficacy and precision. 32 33 34 35 36 37 38 39 40 41 42 43 44 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 2 Main text: 45 CRISPR screens systematically perturb targeted genes and assess functional outcomes 46 through phenotypic changes in mammalian cells, enabling unbiased discovery of gene 47 functions, regulatory network organization, and genotype-phenotype relationships [1, 2]. 48 Substantial progress has recently been made in implementing high-throughput CRISPR 49 screening at single-cell resolution, which utiliz es gene expression of i ndividual cells with 50 corresponding CRISPR-perturbed genes as phenotypic outputs. Many of such approaches, 51 such as Perturb-seq, utilize specialized vectors that allow the indirect capture of sgRNAs by 52 standard 3’-end single-cell RNA-seq (scRNA-seq) methods [3-6]. Direct 3’ sgRNA methods 53 have recently been reported but they also require custom modification of plasmid libraries [7, 8]. 54 The use of these specialized vectors limits the scale and flexibility of genetic screens, since they 55 are incompatible with existing genomes-scale knockout (KO) libraries [1, 9, 10]. Moreover, 56 specialized vectors are susceptible to sgRNA-barcode swapping events due to lentiviral 57 template switching [11]. 58 59 Here, we present a custom 3’ single-cell capture platform, called Native sgRNA Capture and 60 sequencing (NSC-seq), that enables flexible and multi-purpose single-cell CRISPR screening 61 using existing KO vector libraries. To capture non-polyadenylated sgRNAs [12], we designed an 62 inDrops-compatible capture sequence (CS) that binds to the canonical scaffold of gRNAs to 63 initiate direct reverse transcription (RT) of the captured sgRNA. A primer sequence was then 64 added to the 3’-end of the cDNA via template switching to facilitate downstream library 65 amplification (Fig. 1a; Supplemental methods, Supplemental Table 1). As proof of concept, we 66 assessed sgRNA detection efficiency and reproducibility through RNA-based readouts using a 67 human colorectal cancer (CRC) cell line SW620 with the Brunello library (Extended Data Fig. 68 1a-b). The NSC-seq approach exhibited comparable performance to traditional DNA-based 69 detection (Fig. 1b, Extended Data Fig. 1c). Importantly, a subset of sgRNAs exhibited 70 discordance between DNA- and RNA-based detection, with approximately 60% of these 71 sgRNAs shared between biological replicates (Extended Data Fig. 1d-e). This result suggests 72 that the discordance may not have resulted from random technical artifacts but from inherent 73 biases in the expression of different sgRNAs under the U6 promoter [13, 14]. The CS can be 74 used to capture a wide range of gRNAs with a similar scaffold sequence (sgRNAs, hgRNAs, 75 and stgRNAs) [15-17]. In a companion study, we used NSC-seq to directly capture hgRNAs, 76 enabling in vivo recording of lineage and temporal events during mammalian development and 77 tumorigenesis [18]. The capture efficiency of sgRNA at single-cell resolution using NSC-seq can 78 be as high as 95% as shown using the mouse breast epithelial cell line EpH4 with the Brie 79 library. With these performance parameters, we then performed single-cell pooled CRISPR 80 screens using five individual KO vector libraries to reveal effects of genetic perturbation on 81 transcriptomic changes [3] in vitro and in vivo (Fig. 1c). 82 83 We conducted a pooled single-cell fitness screen using the EpH4 cell line with the Brie library, 84 with the majority of captured cells having a single gRNA detected (Fig. 1d-e). While the wild-85 type (WT) cell line was homogeneous, CRISPR-perturbed cells displayed heterogeneity in 86 transcriptome space, reflecting the efficacy of the various genetic perturbations on changing 87 gene expression (Extended Data Fig. 2a-c). Our analysis revealed the enrichment of tumor 88 suppressor and/or apoptotic gene-targeting sgRNAs, including Bcl2l13 and Phactr4, to be 89 among the top 25 enriched sgRNAs (Extended Data Fig. 2d-e) [19, 20]. We also found that 90 NSC-seq detected the sgRNAs for all reported non-essential gene-targeting sgRNAs, whereas 91 sgRNAs targeting essential genes were not picked up as readily, given that KO of essential 92 genes leads to the elimination of targeted cells from the pool (Extended Data Fig. 2f) [21]. We 93 then conducted a more in-depth analysis using a mixed linear model [3], revealing seven 94 functional modules dominated by tumor suppressor and/or apoptotic genes (Extended Data Fig. 95 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 3 2g, Supplemental methods). Perturbed genes of the same pathway generated the same effects 96 on gene expression, and thus were organized into the same functional module. For instance, 97 Sqstm1, Msh2, Bclaf2, and Spata2 all belong to the “Programmed Cell Death” pathway 98 according to GO and were all part of module 7. Other genes, although known generally for their 99 antiproliferative effects, such as Cdkn1b and Bcl2l13, were found to operate through different 100 pathways. Our results reveal distinct pathways that affect cell survival and proliferation, and 101 facilitate the identification of the potential functions of less-characterized genes through 102 associations with the functional modules. 103 104 Next, we conducted a Trametinib resistance screen in the SW620 cells with the Brunello library 105 (Extended Data Fig. 3a, Supplemental methods) [22]. Trametinib-resistant cells were largely 106 found in a quiescent state (G1), a common mechanism for cancer cells to escape MEK1/2 107 inhibition (Extended Data Fig. 3b) [23, 24]. It has been reported that MEK inhibition reduces 108 cellular growth via multiple mechanisms, including the induction of MYT1 hypo-phosphorylation 109 [25] and increasing the levels of PDE5A [26]. Accordingly, enrichment of both MYT1- and 110 PDE5A-targeting sgRNAs were found amongst resistant cells. We further analyzed genetic 111 perturbations that modified drug resistance, revealing two regulatory modules with enrichment 112 of ubiquitin pathway or Rho pathway genes (Extended Data Fig. 3c) [27-29]. Differential gene 113 expression analysis revealed distinct metabolic gene enrichment, especially overexpression of 114 the glutathione metabolism gene GGT5 in the ubiquitin module (Extended Data Fig. 3d-f). We 115 also found overexpression of ERBB3 in resistant cells that resulted in PI3K/AKT activation due 116 to MEK inhibition-mediated negative feedback on ERBB receptors (Extended Data Fig. 3g) [30, 117 31]. Thus, our results reveal multi-factorial trametinib resistance mechanisms and a possible 118 actionable pathway through the PI3K/AKT pathway for KRAS mutant colorectal cancer cells to 119 escape MEK1/2 inhibition [32]. 120 121 We then performed a gene essentiality screen using the mouse colorectal cell line MC38 with a 122 custom sgRNA library to assess metabolic dependency of cancer cells (Extended Data Fig. 4a, 123 Supplemental methods, Supplemental table 2) [33]. After targeting components within the 124 glutamate pathway, we found that the surviving cells were highly proliferative, residing mostly in 125 G2M and S cell cycle phases. Despite similar proliferation kinetics, surviving cells displayed 126 heterogeneity, with distinct regulon activities characterizing Leiden clusters 0 and 2, (Extended 127 Data Fig. 4b-c). sgRNA enrichment analysis revealed five essential genes in the glutamate 128 pathway (Got2, Ppat, Gfpt1, Gclc, and Ctps), as their knockout led to the elimination of targeted 129 cells from the pool (Extended Data Fig. 4d). Among the non-essential genes, we revealed two 130 functional modules with distinct nucleotide metabolism gene enrichment, with glutamate 131 receptor Grik4 being upregulated in module 2 (Extended Data Fig. 4e). These results imply a 132 compensatory mechanism for increasing glutamate uptake to confer a fitness advantage to 133 cancer cells (Extended Data Fig. 4f-g) [34]. 134 135 Finally, we performed two in vivo pooled single-cell CRISPR screens using mouse primary 136 CD8+ cytotoxic T cells (CTLs). Perturbed CTLs were adoptively transferred into a MC38 tumor 137 xenograft model to assess CTL infiltration and proliferation in the tumor microenvironment 138 (TME) (Extended Data Fig. 5a-b, Supplemental table 2, Supplemental Methods) [35]. Isolation 139 of immune cells within the TME showed a predominant CTL and tumor-associated macrophage 140 (TAM) mixture, with CRISPR sgRNA only detected in CTLs, as expected (Extended Data Fig. 141 5c). The TME exhibited features of immune exhaustion, with CTLs expressing Pdcd1 and Ctla4 142 [36], and TAM expressing immunosuppressive genes such as Cd274 and Tgfbi (Extended Data 143 Fig. 5d) [37]. Targeting genes in the cell death pathway using this system confirmed previously 144 reported antiproliferative mechanisms, such as Tsc2 as one of the top enriched genes that 145 conferred a fitness advantage to CTLs in the TME (Extended Data Fig. 5e) [38]. The zinc 146 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 4 transporter Slc39a7 was found to be the most depleted gene, supporting its essential role in 147 lymphocyte development, as reported previously (Extended Data Fig. 5f) [39]. Overall, two 148 distinct modules in cell death pathway were revealed to affect fitness in the TME, one 149 modulating proliferation and the other modulating differentiation (Extended Data Fig. 5g-j). An 150 intrinsic apoptotic signature was found in the proliferative module, implicating a vicious cycle of 151 proliferation and apoptosis that leads to dysfunctional CTLs (Extended Data Fig. 5k) [40]. 152 153 We performed a similar in vivo CTL screen using a custom immune checkpoint pathway library 154 (Extended Data Fig. 6a-c, Supplemental table 2). Our analysis revealed Il2rb as one of the most 155 depleted genes in the screen, as expected (Extended Data Fig. 6d-e) [41-43]. Moreover, we 156 found that perturbation of stimulatory signal receptors (Icos and Cd80) fostered CTL depletion, 157 whereas perturbation of Icosl fostered CTL enrichment (Extended Data Fig. 6e) [44]. Ceacam1 158 perturbation showed a similar transcriptome to Ctla4 perturbation, suggesting parallel functions 159 for these genes in the immune checkpoint pathway (Extended Data Fig. 6f) [45]. Gene module 160 analysis revealed genes associated with suppressing CTL activation in regulatory module 1 161 owing to immune suppressive genes (Itgbl1, Fut4, Prkci, and Parp1), whereas module 3 was 162 characterized by T cell activation owing to T cell activation and differentiation-related genes 163 (Ms4a4b, Socs1, Cd81, and Cd200l1) (Extended Data Fig. 6g-i) [44, 46]. Overall, our results 164 demonstrate that NSC-seq can be used to dissect CTL gene network regulation that modulates 165 antitumor immunity [47]. 166 167 The foundational assumption in CRISPR screening is that the gRNAs expressed by individual 168 cells are functional. Consequently, any observed enrichment or depletion of a gRNA during the 169 screening process can be attributed to the effectiveness of a gRNA in perturbing the target 170 gene. Since NSC-seq directly captures gRNAs rather than proxy barcodes embedded in DNA 171 [3, 4], we assessed the quality of expressed gRNAs in cells directly. Surprisingly, many 172 instances of truncated spacer sequences, often missing bases from the 5’-end, were observed; 173 we termed these alternative sequences sgRNA isoforms (Fig. 2a, Extended Data Fig. 7a). It has 174 been reported that the U6 promoter displays a differential nucleotide preference for the 175 transcription start site (TSS) [13, 14], which we also observed as base bias in sgRNA isoforms. 176 For instance, we found ‘GG’ dinucleotide as the dominant first two bases of isoforms from the 177 Brie library (Extended Data Fig. 7b). We then used an orthogonal approach to confirm isoform 178 gRNA expression using terminal deoxynucleotide transferase (TdT) reactions (Extended Data 179 Fig. 7c). Here, we also observed truncated gRNA isoforms with a higher proportion of ‘GG’ 180 dinucleotides as the first two bases (Extended Data Fig. 7d-e). 181 182 We then characterized the prevalence of this phenomenon across three whole genome KO 183 screening libraries (Extended Data Fig. 8, Supplemental Table 3-6, Supplemental methods). We 184 found that a significant proportion (17-35%) of sgRNAs were expressed as both full length and 185 truncated isoforms with variable degrees of isoform reads (Fig. 2b, Extended Data Fig. 8). To 186 test whether sgRNA isoforms are functional, we assayed for their gene-editing efficiency and 187 found compromised efficiency for sgRNA isoforms (15 bp length) compared to their WT 188 counterparts (20 bp length) (Extended Data Fig. 9a-c) [48]. Furthermore, truncated isoforms 189 may have reduced specificity due to length-dependent complementarity, causing off-target 190 effects across the genome (Fig. 2c) [49-51]. Notably, we found that the Gab3-targeting sgRNA 191 isoform edits one of the four predicted off-target sites (Extended Data Fig.9d). Finally, we 192 assessed the effect of sgRNA isoforms on global transcriptional readouts and found a more 193 significant effect on the transcriptome conferred by sgRNA isoforms (which could have broader 194 off-targets) when compared with WT sgRNAs (Extended Data Fig. 9e). Thus, our data reveal 195 that truncated sgRNA isoforms exist in KO screening libraries which can potentially compromise 196 the quality of CRISPR screens [52, 53]. 197 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 5 198 In summary, the NSC-seq platform provides a framework for targeted capture of non-199 polyadenylated gRNAs. We showed that conventional KO screen libraries can be used for 200 pooled single-cell CRISPR screens via NSC-seq. As proof-of-principle, we provided 201 confirmatory evidence in multiple in vitro and in vivo single-cell screens that revealed new 202 insights into gene essentiality, gene function, and regulatory network modules in a variety of 203 applied biomedical contexts. More importantly, direct gRNA readouts provided by NSC-seq can 204 be used to assess the nucleic acid sequence of the sgRNA itself. Discordant/truncated sgRNAs 205 compromise the efficacy and specificity of gene editing, which, on a genome scale, can 206 detrimentally affect the quality of a screen. sgRNA isoforms can originate from multiple 207 mechanisms, including inherent sgRNA sequence biases, viral insertion position biases in the 208 genome, alternative TSS sequence biases, and degradation biases at the 5’-end of sgRNAs 209 that lead to differential stability. Irrespective of the underlying cause, truncated sgRNA isoforms 210 can lead to functional consequences in the downstream screening process. For genotype-211 phenotype mapping of rare cell types, direct sgRNA readout could be a better proxy than DNA 212 readout due to a higher copy number of sgRNAs per cell, which enhances detection. Our large-213 scale sgRNA expression characterization of three widely used whole genome KO libraries 214 provides reference datasets for truncated sgRNA profiles. Crosschecking across sgRNAs 215

datasets might help in the design of better sgRNAs that minimize false positive 216 discoveries. 217 218 Our multi-purpose platform is scalable and can be broadly applied beyond single-cell CRISPR 219 screens. In a companion study, we applied this platform for the simultaneous lineage and 220 temporal recording of mammalian development and precancer [18]. Further development of 221 NSC-seq can enable combinatorial genetic perturbations using KO plasmid libraries that can be 222 multiplexed with barcoding for simultaneous in vivo genetic perturbation and lineage tracking 223 [54, 55]. Overall, we envision that the NSC- seq platform will expand the application of CRISPR 224 approaches by facilitat ing genotype-phenotype mapping at scale with low cost and high 225 flexibility. 226 227

228 This publication is part of the HTAN (Human Tumor Atlas Network) consortium paper package. 229 The authors wish to thank the funding support by the HTAN grant U2CCA233291 (to R.J.C., 230 K.S.L.), TBEL U54CA274367 (to R.J.C., K.S. L.), R35CA197570 and P50CA236733 (to R.J.C.), 231 R01DK103831 (to K.S.L.). We thank members of the Lau and Coffey laboratories for assistance 232 in animal housing, cell culture, and single-cell data collection. VANTAGE sequencing core was 233 used for this study (P30CA068485). 1cellbio and RAN biotechnologies helped to synthesis the 234 custom hydrogel beads. Vanderbilt University submitted a U.S. patent application for NSC-seq 235 and M.I., R.J.C. and K.S.L. are listed as inventors. We use BioRender for drawing many 236 schematics in this study. We apologize in advance to those we have failed to acknowledge due 237 to space constraints. 238 239 Authors contribution: 240 Conceptualization, M.I. and K.S.L.; data curation, M.I., Y.Y., A.J.S., Y.X., P.M., E.F., B.C.G., 241 C.D.C., M.A.R.-S., and K.S.L.; formal analysis, M.I., Y.Y., M.A.R.-S.; investigation, M.I., R.J.C., 242 and K.S.L.; methodology, M.I., and K.S.L.; project administration, M.I.., A.J.S., R.J.C., and 243 K.S.L.; resources, M.I., Q.L., R.J.C., and K.S.L.; software, M.I., Y.Y., and K.S.L.; supervision, 244 M.I., W.P.T, I.J.M., J.C.R., R.J.C., M.J.S., and K.S.L.; validation, M.I. and Y.Y., and K.S.L.; 245 visualization, M.I. and Y.Y.; and K.S.L.; writing – original draft, M.I.; writing – reviewing and 246 editing, M.I., Y.Y., A.J.S., Y.X., E.F., W.D., B.C.G., P.M., C.D.C., M.A.R.-S., Q.L., W.P.T., 247 I.G.M., J.C.R., R.J.C., and K.S.L. 248 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 6 249 Conflict of interests: 250 J.C.R. is on the scientific advisory board of Sitryx Therapeutics. All other authors declare no 251 competing interests. 252 253 Data and code availability 254 All raw data generated in the present study have been deposited to the GEO with accession no. 255 *******. Reviewer token:*******. NSC-seq data analysis pipeline reported in GitHub: 256 https://github.com/Ken-Lau-Lab/NSC-seq. 257 258

259 1. Bock, C., et al., High-content CRISPR screening. Nat Rev Methods Primers, 2022. 2(1). 260 2. Shalem, O., et al., Genome-scale CRISPR-Cas9 knockout screening in human cells. 261 Science, 2014. 343(6166): p. 84-87. 262 3. Dixit, A., et al., Perturb-Seq: Dissecting Molecular Circuits with Scalable Single-Cell RNA 263 Profiling of Pooled Genetic Screens. Cell, 2016. 167(7): p. 1853-1866.e17. 264 4. Adamson, B., et al., A Multiplexed Single-Cell CRISPR Screening Platform Enables 265 Systematic Dissection of the Unfolded Protein Response. Cell, 2016. 167(7): p. 1867-266 1882.e21. 267 5. Datlinger, P., et al., Pooled CRISPR screening with single-cell transcriptome readout. 268 Nat Methods, 2017. 14(3): p. 297-301. 269 6. Jaitin, D.A., et al., Dissecting Immune Circuits by Linking CRISPR-Pooled Screens with 270 Single-Cell RNA-Seq. Cell, 2016. 167(7): p. 1883-1896.e15. 271 7. Replogle, J.M., et al., Combinatorial single-cell CRISPR screens by direct guide RNA 272 capture and targeted sequencing. Nat Biotechnol, 2020. 38(8): p. 954-961. 273 8. Wessels, H.H., et al., Efficient combinatorial targeting of RNA transcripts in single cells 274 with Cas13 RNA Perturb-seq. Nat Methods, 2023. 20(1): p. 86-94. 275 9. Sanson, K.R., et al., Optimized libraries for CRISPR-Cas9 genetic screens with multiple 276 modalities. Nat Commun, 2018. 9(1): p. 5416. 277 10. Hart, T., et al., Evaluation and Design of Genome-Wide CRISPR/SpCas9 Knockout 278 Screens. G3 (Bethesda), 2017. 7(8): p. 2719-2727. 279 11. Hill, A.J., et al., On the design of CRISPR-based single-cell molecular screens. Nat 280 Methods, 2018. 15(4): p. 271-274. 281 12. Jinek, M., et al., A programmable dual-RNA-guided DNA endonuclease in adaptive 282 bacterial immunity. Science, 2012. 337(6096): p. 816-21. 283 13. Gao, Z., et al., Mutation of nucleotides around the +1 position of type 3 polymerase III 284 promoters: The effect on transcriptional activity and start site usage. Transcription, 2017. 285 8(5): p. 275-287. 286 14. Ma, H., et al., Pol III Promoters to Express Small RNAs: Delineation of Transcription 287 Initiation. Mol Ther Nucleic Acids, 2014. 3(5): p. e161. 288 15. Perli, S.D., C.H. Cui, and T.K. Lu, Continuous genetic recording with self-targeting 289 CRISPR-Cas in human cells. Science, 2016. 353(6304). 290 16. Kalhor, R., P. Mali, and G.M. Church, Rapidly evolving homing CRISPR barcodes. Nat 291 Methods, 2017. 14(2): p. 195-200. 292 17. Kalhor, R., et al., Developmental barcoding of whole mouse via homing CRISPR. 293 Science, 2018. 361(6405). 294 18. Islam, M., et al., Temporal recording of mammalian development and precancer. 295 bioRxiv, 2023: p. 2023.12. 18.572260. 296 19. Solimini, N.L., et al., STOP gene Phactr4 is a tumor suppressor. Proc Natl Acad Sci U S 297 A, 2013. 110(5): p. E407-14. 298 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 7 20. Kataoka, T., et al., Bcl-rambo, a novel Bcl-2 homologue that induces apoptosis via its 299 unique C-terminal extension. J Biol Chem, 2001. 276(22): p. 19548-54. 300 21. Hart, T., et al., High-Resolution CRISPR Screens Reveal Fitness Genes and Genotype-301 Specific Cancer Liabilities. Cell, 2015. 163(6): p. 1515-26. 302 22. Wu, P.K. and J.I. Park, MEK1/2 Inhibitors: Molecular Activity and Resistance 303 Mechanisms. Semin Oncol, 2015. 42(6): p. 849-62. 304 23. Alarcón, T. and H.J. Jensen, Quiescence: a mechanism for escaping the effects of drug 305 on cell populations. J R Soc Interface, 2011. 8(54): p. 99-106. 306 24. Borst, P., Cancer drug pan-resistance: pumps, cancer stem cells, quiescence, epithelial 307 to mesenchymal transition, blocked cell death pathways, persisters or what? Open Biol, 308 2012. 2(5): p. 120066. 309 25. Villeneuve, J., et al., MEK1 inactivates Myt1 to regulate Golgi membrane fragmentation 310 and mitotic entry in mammalian cells. Embo j, 2013. 32(1): p. 72-85. 311 26. Arozarena, I., et al., Oncogenic BRAF induces melanoma cell invasion by 312 downregulating the cGMP-specific phosphodiesterase PDE5A. Cancer Cell, 2011. 19(1): 313 p. 45-57. 314 27. Narayanan, S., et al., Targeting the ubiquitin-proteasome pathway to overcome anti-315 cancer drug resistance. Drug Resist Updat, 2020. 48: p. 100663. 316 28. Deng, L., et al., The role of ubiquitination in tumorigenesis and targeted drug discovery. 317 Signal Transduct Target Ther, 2020. 5(1): p. 11. 318 29. Kreider-Letterman, G., N.M. Carr, and R. Garcia-Mata, Fixing the GAP: The role of 319 RhoGAPs in cancer. Eur J Cell Biol, 2022. 101(2): p. 151209. 320 30. Turke, A.B., et al., MEK inhibition leads to PI3K/AKT activation by relieving a negative 321 feedback on ERBB receptors. Cancer Res, 2012. 72(13): p. 3228-37. 322 31. He, Y., et al., Targeting PI3K/Akt signal transduction for cancer therapy. Signal 323 Transduct Target Ther, 2021. 6(1): p. 425. 324 32. Vander Velde, R., et al., Resistance to targeted therapies as a multifactorial, gradual 325 adaptation to inhibitor specific selective pressures. Nat Commun, 2020. 11(1): p. 2393. 326 33. Altman, B.J., Z.E. Stine, and C.V. Dang, From Krebs to clinic: glutamine metabolism to 327 cancer therapy. Nat Rev Cancer, 2016. 16(10): p. 619-34. 328 34. Reinfeld, B.I., et al., Cell-programmed nutrient partitioning in the tumour 329 microenvironment. Nature, 2021. 593(7858): p. 282-288. 330 35. Waldman, A.D., J.M. Fritz, and M.J. Lenardo, A guide to cancer immunotherapy: from T 331 cell basic science to clinical practice. Nat Rev Immunol, 2020. 20(11): p. 651-668. 332 36. Buchbinder, E.I. and A. Desai, CTLA-4 and PD-1 Pathways: Similarities, Differences, 333 and Implications of Their Inhibition. Am J Clin Oncol, 2016. 39(1): p. 98-106. 334 37. Gordon, S.R., et al., PD-1 expression by tumour-associated macrophages inhibits 335 phagocytosis and tumour immunity. Nature, 2017. 545(7655): p. 495-499. 336 38. Sugiura, A., et al., MTHFD2 is a metabolic checkpoint controlling effector and regulatory 337 T cell fate and function. Immunity, 2022. 55(1): p. 65-81.e9. 338 39. Anzilotti, C., et al., An essential role for the Zn(2+) transporter ZIP7 in B cell 339 development. Nat Immunol, 2019. 20(3): p. 350-361. 340 40. Horton, B.L., et al., Intratumoral CD8(+) T-cell Apoptosis Is a Major Component of T-cell 341 Dysfunction and Impedes Antitumor Immunity. Cancer Immunol Res, 2018. 6(1): p. 14-342 24. 343 41. Alderdice, M., et al., Evolutionary genetic algorithm identifies IL2RB as a potential 344 predictive biomarker for immune-checkpoint therapy in colorectal cancer. NAR Genom 345 Bioinform, 2021. 3(2): p. lqab016. 346 42. Fernandez, I.Z., et al., A novel human IL2RB mutation results in T and NK cell-driven 347 immune dysregulation. J Exp Med, 2019. 216(6): p. 1255-1267. 348 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 8 43. Chinen, T., et al., An essential role for the IL-2 receptor in T(reg) cell function. Nat 349 Immunol, 2016. 17(11): p. 1322-1333. 350 44. Raskov, H., et al., Cytotoxic CD8(+) T cells in cancer and cancer immunotherapy. Br J 351 Cancer, 2021. 124(2): p. 359-367. 352 45. Dankner, M., et al., CEACAM1 as a multi-purpose target for cancer immunotherapy. 353 Oncoimmunology, 2017. 6(7): p. e1328336. 354 46. Palmer, D.C. and N.P. Restifo, Suppressors of cytokine signaling (SOCS) in T cell 355 differentiation, maturation, and function. Trends Immunol, 2009. 30(12): p. 592-602. 356 47. Zhou, P., et al., Single-cell CRISPR screens in vivo map T cell fate regulomes in cancer. 357 Nature, 2023. 358 48. Liu, X., et al., Effects of sgRNA length and number on gene editing efficiency and 359 predicted mutations generated in rice. The Crop Journal, 2022. 10(2): p. 577-581. 360 49. Ren, X., et al., Enhanced specificity and efficiency of the CRISPR/Cas9 system with 361 optimized sgRNA parameters in Drosophila. Cell Rep, 2014. 9(3): p. 1151-62. 362 50. Fu, Y., et al., High-frequency off-target mutagenesis induced by CRISPR-Cas nucleases 363 in human cells. Nat Biotechnol, 2013. 31(9): p. 822-6. 364 51. Hsu, P.D., et al., DNA targeting specificity of RNA-guided Cas9 nucleases. Nat 365 Biotechnol, 2013. 31(9): p. 827-32. 366 52. Zhang, J.P., et al., Different Effects of sgRNA Length on CRISPR-mediated Gene 367 Knockout Efficiency. Sci Rep, 2016. 6: p. 28566. 368 53. Josephs, E.A., et al., Structure and specificity of the RNA-guided endonuclease Cas9 369 during DNA interrogation, target binding and cleavage. Nucleic Acids Res, 2015. 43(18): 370 p. 8924-41. 371 54. Santinha, A.J., et al., Transcriptional linkage analysis wi th in vivo AAV-Perturb-seq. 372 Nature, 2023. 373 55. Michlits, G., et al., CRISPR-UMI: single-cell lineage tracing of pooled CRISPR-Cas9 374 screens. Nat Methods, 2017. 14(12): p. 1191-1197. 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 9 398 399 400 401 402 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 1 0 Main Figures: 403 Fig. 1: Overview of single-cell pooled CRISPR screens using knockout libraries. 404 Fig. 2: Assessment of sgRNA expression. 405 406 407 Supplemental Figures: 408 Extended Data Fig. 1: Assessment of sgRNA detection efficiency and replicate reproducibility. 409 Extended Data Fig. 2: Analysis of a pooled single-cell gene fitness screen. 410 Extended Data Fig. 3: Analysis of trametinib resistant screen. 411 Extended Data Fig. 4: Analysis of glutamate pathway gene essentiality screen. 412 Extended Data Fig. 5: Analysis of cell death pathway screen in mouse primary T cells. 413 Extended Data Fig. 6: Analysis of immune checkpoint pathway screen in mouse primary T cells. 414 Extended Data Fig. 7: Truncated sgRNAs expression in cells. 415 Extended Data Fig. 8: Large scale assessment of isoform sgRNAs across three whole-genome 416 knockout libraries. 417 Extended Data Fig. 9: Gene editing efficiency of isoform sgRNAs. 418 419 420 Supplemental tables: 421 Supplemental table 1: Primer sequences. 422 Supplemental table 2: Custom sgRNA libraries. 423 Supplemental table 3: Brie sgRNA expression. 424 Supplemental table 4: Brunello rep A sgRNA expression. 425 Supplemental table 5: Brunello rep B sgRNA expression. 426 Supplemental table 6: GeCKOv2 sgRNA expression. 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 1 1 Main figures: 452 453 Fig. 1. Overview of single-cell pooled CRISPR screens using knockout libraries. (a) Single454 guide RNA (sgRNA) capture schematic for the NSC-seq platform. NSC- seq capture sequence455 (CS) anneals to 3’-end target sites of sgRNA scaffold. Here, CS contains cellular barcode456 (blue), UMI (green), and T7 promoter (brown) sequences . An additional sequence (gray) is457 added to the 3’-end of the cDNA during reverse transcription via template switching to enable458 downstream library amplification. ( b) sgRNA capture efficiency by NSC-seq assessed in a bulk459 experiment and compared with bulk DNA sequencing approach. ( c) Schematic of single- cell460 pooled CRISPR screens using NSC-seq platform. Five distinct single- cell screens were461 performed in this study using conventional knockout vector libraries (bottom). ( d) A462 representative UMAP plot showed cell with sgRNA (red) or without sgRNA (gray). ( e)463 Representative quantification of the number of sgRNAs detected per cell. Cells with 1 detected464 sgRNA were used for downstream analysis. 465 466 467 le ce de is le lk ell re A ) ed (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint 1 2 468 Fig. 2. Assessment of sgRNA expression. (a) Schematic of WT and truncated sgRNAs 469 (isoform). A representative sgRNA (Dkk1) read alignment and visualization using IGV (bottom) 470 from EpH4 cells with Brie library (see Extended Data Fig. 7). (b) Fraction of WT and isoform 471 sgRNAs expression across three whole-genome CRISPR knockout libraries from Addgene (see 472 Extended Data Fig. 8). (c) Predicted number of off-target sites across the mouse genome for 473 variable isoform length (see Extended Data Fig. 9). 474 475 476 477 478 479 480 481 482 483 e (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted February 2, 2024. ; https://doi.org/10.1101/2024.02.01.578192doi: bioRxiv preprint