TRECAN 252 No. of Pages 10

Opinion

Cancer CRISPR Screens In Vivo

Ryan D. Chow1,2,3 and Sidi Chen1,2,3,4,5,6,7,*

HighlightsIn vivo CRISPR screens enable high-throughput interrogation of complexprocesses in cancer.

Direct autochthonous models recapi-tulate human cancer by maintainingthe native microenvironment.

Direct in vivo CRISPR technologiescan empower patient-specific cancermodeling for precision medicine.

1Department of Genetics, YaleUniversity School of Medicine, NewHaven, CT, USA2Systems Biology Institute, YaleUniversity School of Medicine, WestHaven, CT, USA3Medical Scientist Training Program,Yale University School of Medicine,New Haven, CT, USA4Biological and Biomedical SciencesProgram, Yale University School ofMedicine, New Haven, CT, USA5Immunobiology Program, YaleUniversity School of Medicine, NewHaven, CT, USA6Comprehensive Cancer Center, YaleUniversity School of Medicine, NewHaven, CT, USA7Stem Cell Center, Yale UniversitySchool of Medicine, New Haven, CT,USA

*Correspondence:sidi.chen@yale.edu (S. Chen).

Clustered regularly interspaced short palindromic repeats (CRISPR) screeningis a powerful toolset for investigating diverse biological processes. MostCRISPR screens to date have been performed with in vitro cultures or cellulartransplant models. To interrogate cancer in animal models that more closelyrecapitulate the human disease, autochthonous direct in vivo CRISPR screenshave recently been developed that can identify causative drivers in the nativetissue microenvironment. By empowering multiplexed mutagenesis in fullyimmunocompetent animals, direct in vivo CRISPR screens enable the rapidgeneration of patient-specific avatars that can guide precision medicine. ThisOpinion article discusses the current status of in vivoCRISPR screens in cancerand offers perspectives on future applications.

CRISPR-Mediated Genome Editing Meets the Complexity of the CancerGenomeOver the past decade, tremendous efforts have been devoted to profiling patient cancers.Multi-institutional consortia such as The Cancer Genome Atlas have now profiled over 10 000tumors, generating petabytes of high-dimensional data that illuminate the complexities ofseveral cancer types [1]. Although the molecular portraits of cancer are now in higher resolutionthan ever before, the path to clinical translation for many cancer types remains largelyunexplored.

The central issue is that cancer genomics can inform what mutations are present, but haslimited power to indicate which ones are functionally important [2,3]. Individual tumors oftenhave hundreds, if not thousands, of molecular aberrations. While some of these alterations arefound in well-established oncogenes (see Glossary) and tumor suppressors, many novelaberrations occur in previously uncharacterized or even unannotated regions, making it difficultto reliably discern whether they are actually driving the progression of a given cancer. Further-more, different patients can present with unique combinations of these various mutations thatcan drastically influence a cancer’s growth pattern, tendency to metastasize, and susceptibilityto therapy.

CRISPR-mediated genome editing has become a powerful tool in cancer biology due to itsprogrammability and flexibility [4]. In addition to its canonical use for targeted gene knockouts[5,6], CRISPR has been reengineered for a variety of purposes including transcriptionalactivation [7,8], transcriptional repression [9], histone modification [10], base editing [11,12],DNA methylation [13–15], and genome architecture manipulation [16]. Following in the foot-steps of RNAi screens, the modularity of CRISPR has naturally lent itself to high-throughputscreening approaches [17]. By designing custom libraries of single-guide RNAs (sgRNAs),one can simultaneously screen large collections of genomic elements (coding genes, regulatoryelements such as enhancers, noncoding RNAs, or other noncoding features) in a givenbiological context [18]. Here we review the development and application of in vivo CRISPRscreens.

Trends in Cancer, Month Year, Vol. xx, No. yy https://doi.org/10.1016/j.trecan.2018.03.002 1© 2018 Elsevier Ltd All rights reserved.

TRECAN 252 No. of Pages 10

Glossarya

Cas9: an endonuclease that isdirected to specific sites in thegenome by CRISPR spacers, whereit induces double-stranded breaks inthe target DNA.Clustered regularly interspacedshort palindromic repeats(CRISPR): segments of prokaryoticDNA containing short repetitions ofbase sequences. Each repetition isfollowed by short segments of‘spacer DNA’ from previousexposures to a bacteriophage virusor plasmid. The CRISPR–Cas systemis a prokaryotic immune system thatconfers resistance to foreign geneticelements such as those present inplasmids and phages. Cas proteinsuse the CRISPR spacers torecognize and cut these exogenousgenetic elements in a manneranalogous to RNAi in eukaryoticorganisms. By delivering the Cas9nuclease and appropriate guideRNAs (gRNAs) into a cell, the cell’s

In Vivo CRISPR Screens to Interrogate the Complexities of CancerCRISPR screens were first developed by using pooled sgRNA libraries to target all annotatedgenes in the human genome [19,20]. High-throughput, and particularly genome-wide, CRISPRscreens combine the power of forward genetics and reverse genetics. Such approaches offeran unbiased yet precisely targeted method of identifying genes that contribute to a phenotype,such as cancer progression. Compared with random mutagenesis, CRISPR screens usecustomized sgRNA libraries, while simultaneously preserving the ‘randomness’ during selec-tion. Unlike RNAi screens, CRISPR generates precise mutations or complete knockoutsinstead of partial gene knockdown or silencing, which has been shown to facilitate higherbetween-construct concordance and lower off-target rates [21,22]. CRISPR screens have thusbecome the state-of-the-art approach for the discovery of genetic drivers and phenotypicmodulators in many biological contexts.

In cancer, CRISPR screens have been performed to identify genes involved in a wide variety ofprocesses [23], including regulators of drug resistance [24–28], synergistic and syntheticlethal interactions [29,30], regulators of PD-L1 expression [31], and essential genes [32–34](Figure 1). While in vitro studies are valuable for identifying cell-intrinsic properties of cancer cellsand potential therapeutic windows (Figure 2A, Key Figure), they cannot address problemsinvolving complex interactions between multiple cell types that reflect the bona fide nature ofcancer as an organ, instead of a collection of isolated tumor cells [35]. As highlighted by recentadvances in cancer immunotherapy, the microenvironmental milieu is constantly engaged inconversation with tumor cells, with significant clinical consequences [36–40]. To more faithfully

genome can be cut at a desiredlocation, allowing existing genes tobe removed and/or new ones added.CRISPR-associatedendonuclease in Prevotella andFrancisella 1 (Cpf1): also known asCas12a; a CRISPR-guidedendonuclease that can be utilized fortargeted genome editing in diversespecies. Unlike Cas9, Cpf1 does notrequire a tracrRNA for DNAcleavage, and has the capability toautonomously process crRNAarrays.CRISPR RNA (crRNA): the shortRNA sequence that directly guidesCRISPR nucleases to target sites inthe genome that is independent ofthe scaffold sequence present in full-length sgRNAs.dCas9: catalytically dead Cas9nuclease that cannot generatedouble-stranded breaks in DNA;commonly tethered to other proteinsto enable programmable targeting ofthe tethered enzymes.Double-knockout CRISPR screen:a CRISPR screen in which twogenes are simultaneouslyinterrogated rather than a singlegene.Driver mutation: a mutation that iscausally implicated in oncogenesis. Itconfers a growth advantage on the

Transforma on Metastasis

Drug resistance

Immune evasion

Inhibi onof apoptosis

Metabolicreprogramming

Genome instability

Angiogenesis

In vivo CRISPRscreening

Figure 1. Functional Cancer Genomics with In Vivo Clustered Regularly Interspaced Short PalindromicRepeats (CRISPR) Screening. In vivo CRISPR screening is a powerful, flexible tool to dissect important processes incancer. Genome-wide CRISPR screens have illuminated novel regulators of metastasis, malignant transformation,immune evasion, and drug resistance. Moving forward, in vivo CRISPR screens may improve our understanding of otherprocesses, such as angiogenesis, genome instability, and metabolic programming.

2 Trends in Cancer, Month Year, Vol. xx, No. yy

TRECAN 252 No. of Pages 10

cancer cell and is favored in themicroenvironment of the tissue inwhich the cancer arises.Genome editing: also known asgenome editing with engineerednucleases (GEEN); a type of geneticengineering in which DNA is inserted,deleted, or replaced in the genomeof an organism using engineerednucleases, or ‘molecular scissors’.Guide RNA (gRNA): a shortsynthetic RNA comprising a ‘scaffold’sequence necessary for Cas9binding and a user-defined �20-nucleotide ‘spacer’ or ‘targeting’sequence that defines the genomictarget to be modified.High-dimensional genetic screen:a CRISPR screen in which three ormore genes are simultaneouslyinterrogated, enabling theinvestigation of complex multigenicphenotypes.In vivo CRISPR screen: bygenerating libraries of sgRNAstargeting different genes, a CRISPRscreen can be performed to assessthe importance of these genestowards a given phenotype. With invivo CRISPR screens, the selectionphase occurs inside a livingorganism; for instance, a mouse.Oncogene: a gene that has thepotential to cause cancer. In tumorcells, oncogenes are often mutatedor expressed at high levels.Reporter gene: a gene thatresearchers attach to a regulatorysequence to facilitate readout ofgene regulatory activity. Commonexamples include GFP and fireflyluciferase.Single-guide RNA (sgRNA): a termgenerally interchangeable with gRNAin genome editing with the CRISPR–Cas9 system.Synthetic lethality: arises when acombination of mutations in two ormore genes leads to cell death,whereas a mutation in only one ofthese genes does not. In a syntheticlethal genetic screen, it is necessaryto begin with a mutation that doesnot kill the cell, although it mayconfer a phenotype (e.g., slowgrowth) and then systematically testother mutations at additional loci todetermine which ones conferlethality. Synthetic lethality indicatesfunctional relationships betweengenes.

recapitulate the human disease for enhanced translational accuracy, it is essential to investigatethese phenomena in vivo where cancers can develop in the context of a tissuemicroenvironment.

To date, virtually all in vivo CRISPR screens have aimed to investigate phenotypes of cancer,directly demonstrating the power and simplicity of this technology in oncology. The first in vivoCRISPR screen investigated annotated genes in the mammalian genome and their potential topromote tumor growth and metastasis when mutagenized [41]. In this study, the authorsintroduced a genome-wide sgRNA library into a non-metastatic cancer cell line. The pool-mutagenized cell library was then subcutaneously transplanted into the back skin of nude miceand monitored for metastasis to the lung. By sequencing the sgRNAs present in the metasta-ses, the authors identified and subsequently validated a panel of hits from the initial screen thatfunctionally drove lungmetastasis. Multiple in vivoCRISPR screens have since been performedto identify tumor suppressors [42–44], oncogenes [45], synthetically lethal genes [46], andregulators of cancer immunotherapy, in both two-cell type (2-CT) coculture systems [47] andtransplant tumor models [48], among others. A common thread in these studies is their two-step workflow: the sgRNA library is first introduced to cells in culture, followed by transplanta-tion into mice to assess phenotypes in vivo. After a selection phase (e.g., expression of areporter, outgrowth of a tumor, resistance to therapy), the cells are sequenced to identifyenriched and/or depleted sgRNAs (Figure 2B).

Limitations of Transplant-Based In Vivo CRISPR ScreensBy virtue of them having a microenvironment, compared with in vitro CRISPR screens, it isanticipated that in vivo studies more faithfully model carcinogenesis as it occurs in humans.However, transplant-based in vivo screens have significant limitations. First, introducing largenumbers of cancer cells into mice clearly does not resemble the normal process of tumori-genesis. Second, transplantations are commonly performed subcutaneously, rather thanorthotopically in the relevant organs. In addition, immunodeficient mice are often used topromote the grafting efficiency of the mutagenized cell library, limiting the applicability of thesemodels for investigation of cancer–immune interactions. Moreover, the engraftment efficiencyof transferred cells can vary from cell line to cell line and from host to host. Furthermore, thereexist multiple cellular bottlenecks of tumor evolution in vivo, such as the circulation limit duringthe metastasis cascade. Finally, the native organ microenvironment places numerous con-straints on transplanted cells even in orthotopic settings. These limitations can affect thepreclinical utility of cancer-cell transplant-based in vivo CRISPR screens.

Direct In Vivo CRISPR Screens to More Faithfully Recapitulate HumanCancersTo better model the natural context of cancer as it occurs in humans, the following features canprovide guiding principles for more accurate in vivo CRISPR screens in cancer: (i) the tumorsbeingmodeled derive from the endogenous target tissue; (ii) the immune system remains intact;and (iii) the corresponding tissue microenvironment is preserved. These challenges can beovercome by directly mutagenizing target tissues in vivo rather than using a transplantapproach.

A number of studies have demonstrated efficient CRISPR-mediated in vivo mutagenesisdirectly at the target organ site. In the liver, hydrodynamic injection of sgRNA-containingplasmids into the tail veins of Cas9 mice was sufficient to induce multiplexed mutagenesisin hepatocytes [49], while in the lung intratracheal delivery of sgRNA-carrying lentiviruses wasable to induce mutagenesis directly in Cas9-expressing lung epithelial cells [50]. Through the

Trends in Cancer, Month Year, Vol. xx, No. yy 3

TRECAN 252 No. of Pages 10

Tumor immunity: refers to theinteraction between the host immunesystem and a tumor. In differentcontexts, the immune system canhave opposing roles towards cancerprogression. Cancerimmunotherapies aim to tip thescales towards immune-mediatedelimination of the tumor.Tumor microenvironment: thestroma and supporting milieusurrounding the tumor, usuallycomprising fibroblasts, immune cells,and endothelial cells.Tumor suppressor gene: alsoknown as an antioncogene; a genethat protects a cell from one step onthe path to cancer. Mutations thatinactivate tumor suppressors willcontribute to cancer progression.

aCertain definitions are generalconcepts, modified from originaldefinitions.

use of adenoassociated viruses (AAVs), direct mutagenesis in the mouse brain was also shownto be feasible [51]. Because Cas9 and other genome-editing RNA-guided nucleases (RGNs)are large proteins, generation of Cas9 transgenic animals [52–54] simplifies the delivery ofCRISPR components and facilitates direct in vivo mutagenesis.

Extending on these direct in vivo CRISPR techniques, autochthonous CRISPR screens ofvarying scales have now been performed in a few organ systems (Figure 2C). Hydrodynamicinjection of plasmids with sgRNA and Cas9 expression cassettes flanked by Sleeping Beauty(SB) inverted repeats, along with a SB-transposase vector, has been utilized to screen tensgRNAs for their ability to induce CRISPR-mediated tumorigenesis in a KrasG12D-sensitizedmouse liver [55]. As another approach, lentiviral pools were utilized to screen 11 sgRNAs fortheir ability to drive lung tumorigenesis in Cre-driven Cas9 mice [56]. Finally, stereotaxic deliveryof an AAV sgRNA library directly into the mouse brain efficiently induced glioblastomas thatrecapitulate the pathologic features of the human disease [57], which enabled direct andquantitative assessment of the tumorigenicity of 280 sgRNAs targeting 56 genes. Of note, therelative mutant frequencies of these genes in the AAV-basedmodel significantly correlated withthe mutant frequencies observed in human glioblastoma cohorts. Delivery of AAV-CRISPRsgRNA pools intravenously followed by captured sequencing using customized probesets [57]or molecular inversion probe sequencing [58] allows autochthonous mapping of causativefunctional variants directly in the targeted genomic loci, as demonstrated in the brain [57] andliver [58]. Facilitated by the high efficiency of AAV-mediated transduction, various significantlyco-occurring pairs were identified, thereby pinpointing synergistic driver pairs in tumorigenesis.

Nevertheless, direct in vivo approaches have their share of limitations. Optimal technicalparameters that are easily achievable in in vitro or cell line transplant settings (i.e., ‘indirect’in vivo), such as library size, coverage, and multiplicity of infection (MOI), are much morechallenging with a direct in vivo approach. Therefore, the size of a sgRNA library must becontrolled to ensure adequate coverage in vivo. With an oversized library, random samplingerrors from low viral transduction rates will invariably lead to many spurious positive andnegative ‘hits’. Other key constraints include the number of tumor-originating cells in thenative organ, the accessibility of such cells due to the complexity of cellular organization inendogenous tissues, uncharacterized or unknown cell–cell interactions, the viral transductionefficiency, and immune rejection. To this end, it is worth noting that AAV-mediated approacheshave the major advantage of higher-titer, higher direct in vivo transduction efficiencies andprovoke minimal immune reactions [57], thus enabling larger CRISPR libraries to be screenedmore efficiently in an autochthonous setting. Since AAVs usually do not integrate into thegenome, the relative abundance of sgRNAs in tumors cannot be ascertained by sgRNAcassette sequencing as is commonly done for lentivirus-based screens. Instead, the AAV-mediated approach requires capture sequencing of the sgRNA target regions to functionallyextract the mutational signatures of the tumors.

Direct In Vivo CRISPR Pooled Mutagenesis Technologies for PrecisionMedicineWith further improvements and modifications, we envision that direct in vivo CRISPR screenscould be widely applied in multiple facets of precision medicine. Using the genomic informationfrom a patient’s tumor, a personalized CRISPR library could be readily designed to directlygenerate disease mimics, or cancers encompassing the same set of mutations, to investigatethe behaviors of those mutations in mice (Figure 3A). As a proof of principle, high-titer AAVlibraries have been successfully used to infect single cells at high MOI in vivo, leading togenetically complex tumors with several mutations [57,58]. This ‘mouse avatar’ approach

4 Trends in Cancer, Month Year, Vol. xx, No. yy

TRECAN 252 No. of Pages 10

Key Figure

Three Modes of Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) Screening

Len viral sgRNA pool

Len viral sgRNA pool

Infect cells in culture

Infect cells in culture

Select for phenotype

Transplant cells into mice

sgRNA sequencing

sgRNA sequencingof resultant tumors

Viral sgRNA pool Viral transduc ondirectly in vivo

Tumorigenesisand selec on

Capture sequencingof sgRNA target sites

Intratracheal

Intravenous

Intracranial

Drug resistance

Len

AAV

Cas9+

Cas9+

% o

f cel

ls

In vitro

Indirect in vivo (transplant based)

Direct in vivo (autochthonous)

(A)

(B)

(C)

(See figure legend on the bottom of the next page.)

Figure 2. (A) To perform an in vitro CRISPR screen, the desired single-guide RNA (sgRNA) library must first be cloned into expression vectors. Lentiviral vectors arecommonly used, as they can stably integrate into the host genome. After a selection phase to enrich for a desired phenotype, the sgRNA cassettes are amplified fromgenomic DNA and sequenced to identify the top candidate genes. (B) Indirect in vivo screens follow the same steps as in vitro studies, but the selection phase occurs ina recipient animal. Following transplantation of themutagenized cell pool into mice, different mutants will become enriched. In the case of a tumorigenesis screen, highlyabundant sgRNAs in the resultant tumors would be brought forward as candidate tumor suppressors. (C) For direct in vivo screens, CRISPRmutagenesis occurs at the

Trends in Cancer, Month Year, Vol. xx, No. yy 5

TRECAN 252 No. of Pages 10

Outstanding QuestionsDoes a corresponding CRISPR-engi-neered mouse avatar mimic the drugresponse of a genetically matchedtumor from a patient?

Can in vivo CRISPR screens be effec-tively leveraged for the study of com-plex genetic interactions in cancer?

How can direct in vivo CRISPRscreens be efficiently coupled withother high-throughput approachessuch as single-cell sequencing andhigh-content imaging to enable thedissection of highly heterogeneouscancers?

Can direct in vivo CRISPR screens beutilized to deduce mutations that areassociated with sensitivity to immuno-therapy as well as those conferringprimary or acquired resistance?

How can in vivo CRISPR screens beeffectively used in immune cells toidentify novel regulators of the tumormicroenvironment?

would enable robust enumeration of the tumorigenic potential of each mutation present in apatient’s tumor. More importantly, rapid generation of such disease mimics allows robusttherapeutic testing of approved or novel treatments in genetically matched tumors. This type ofapproach could then be utilized to predict the outcomes of treatments in patients andpotentially anticipate which specific tumors would exhibit sensitivity or resistance tochemo-, targeted, or immune therapies, and thus be leveraged for prioritization of therapies.

It was recently demonstrated that patient-derived tumor xenografts (PDXs) adopt evolutionarycourses divergent from the human primary tumor [59]. While this effect is likely to be due in largepart to intrinsic differences between mice and humans, a key contributing factor may lie in themethod, as xenografts generally require immunodeficient hosts for successful engraftment.The development and application of humanized mice [60] has helped to bridge this gap in part,particularly with regard to innate immunity, although complete humanization of the full immunesystem remains to be seen. By contrast, direct in vivomodels are designed to originate from theautochthonous tissue site in fully immunocompetent mice. For these reasons, we anticipatethat direct in vivoCRISPR screens will more faithfully recapitulate the behavior and evolutionarycourse of human cancers.

Intratumoral heterogeneity is increasingly being recognized as an important feature of humancancers, potentially contributing to drug resistance and relapse [61,62]. In addition to cell-typeheterogeneity and microenvironmental variations, intratumoral heterogeneity is often charac-terized by distinct subclonal mutation signatures that differentially influence cellular behavior[63]. To this end, AAV-mediated direct in vivo CRISPR screens are particularly well equipped tomodel the heterogeneity present in human cancers. As demonstrated in the brain and liver,AAV-CRISPR approaches can rapidly create genetically complex multiclonal tumors in indi-vidual mice [57,58]. This immense heterogeneity inherent to AAV-mediated CRISPR screenscould be exploited to investigate clonal dynamics in tumors, serving as a discovery platform forhow tumor subclones interact in controlled experimental settings in otherwise wild-typeorganisms.

Concluding RemarksRecent development and applications of in vivo CRISPR screens have showcased their powerfor unbiased identification of functional genetic elements. Particularly, direct in vivo CRISPRscreens offer a high-throughput strategy to interrogate candidate genes for their ability to drivetumorigenesis from the autochthonous tissue. Unlike in vitro or transplant-based approaches,direct in vivo CRISPR studies retain the endogenous tissue microenvironment and can beperformed in fully immunocompetent animals at high efficiency. Future studies will need toexplicitly evaluate whether direct in vivoCRISPRmodels better recapitulate human disease (seeOutstanding Questions). A key experiment would be to compare the in vivo evolutionarytrajectory of a primary patient tumor and the corresponding mouse avatar, with or withouttherapeutic selection pressure.

The design philosophy behind most CRISPR screens is to study the effects of single-genemutations on a desired phenotype. However, many biological processes are driven by inter-actions between multiple genes. In cancer, the precise combinations of mutations in individual

autochthonous target organ site instead of in culture. Lentiviral and adenoassociated virus (AAV) approaches have both been successfully used for multiplexed direct invivo mutagenesis. Intravenous, intracranial, and intratracheal viral injections can drive tumorigenesis from the liver, brain, and lung, respectively. Since AAVs do notintegrate into the genome, capture sequencing must be performed to read out the results of the screen (i.e., highly abundant indel variants).

6 Trends in Cancer, Month Year, Vol. xx, No. yy

TRECAN 252 No. of Pages 10

Mul -KO viral pool Direct in vivomul -KO screen

Mul -KO crRNAarray library

Cpf1 mul -KOcrRNA arrays

Genera on of complexautochthonous tumors

Single-cell analysis ofCRISPR-induced tumors

Capture sequencingand RNA sequencing

Genera on of complexautochthonous tumors

Mul plexed in situ hybridiza onfor sgRNAs on tumor sec ons

In situ clonal analysis andphenotype–genotype matching

Pa ent muta on signatures PersonalizedsgRNA library

Direct in vivomouse avatar

Priori za on oftherapeu c candidates

DNA

RNA

In situ clonal analysis of autochthonous tumors

Genera on and dissec on of tumor heterogeneity

Higher-dimensional screens

Personalized mouse avatars(A)

(B)

(C)

(D)

ControlDrug 1

Drug 2

Drug 3Time

Tum

pr b

urde

n

AAV

AAV

Double knockout

Triple knockout

n-tuple knockout

5ʹ

5ʹ

5ʹ

3ʹ

3ʹ

3ʹ

cr1

cr1 cr2

cr2 cr3

crNcr2cr1

DR

DR

DR

DR

DR

Microfluidic chip

(See figure legend on the bottom of the next page.)

Trends in Cancer, Month Year, Vol. xx, No. yy 7

TRECAN 252 No. of Pages 10

tumors can lead to strikingly diverse behaviors, influencing tumor aggressiveness and clinicalresponse. In this regard, in vivo CRISPR screens can be readily applied to identify phenotypicmodulators on specific sensitized backgrounds. For instance, direct in vivo CRISPR screenscould be used to identify factors that influence the metastatic properties of mutant Kras-driventumors. An exciting avenue for further work is the adaptation of CRISPR-associated endo-nuclease in Prevotella and Francisella 1 (Cpf1) [64,65] to in vivo cancer modeling. LikeCas9, the Cpf1 RGN can be used for precisely targeted genome editing, yet it has uniquemultiplexing capabilities due to its independence from a tracrRNA [64,65]. Cpf1 has just begunto emerge for higher-dimensional genetic screening of tumor growth and metastasis invivo [66] and remains to be more broadly applied for other aspects of cancer. The developmentof this technology would enable high-dimensional screening of different mutation combina-tions, potentially leading to the identification of novel synergistic or synthetically lethal geneticinteractions in cancer (Figure 3B).

Additionally, the application of dCas9-activator mice could enable the functional identificationof oncogenes directly in vivo, providing an orthogonal perspective to preexisting tumorsuppressor screens [67]. Similarly, mice engineered to express CRISPR-targeted base editorsmight offer an approach to screen specific point mutations that drive tumorigenesis from theautochthonous tissue. It would also be interesting to apply single-cell RNA-seq (scRNA-seq) totumors generated by direct in vivo CRISPR screens [68–70]. Together, these technologieswould enable the creation and subsequent dissection of cancer heterogeneity at ultrahighresolution (Figure 3C). To further provide spatial information, multiplexed fluorescent in situhybridization (MERFISH) [71] could potentially be adapted for the purpose of sequencingsgRNA pools directly in tissue sections. These data would allow matched comparisonsbetween CRISPR-induced mutational signatures and histopathological phenotypes(Figure 3D).

Finally, direct in vivo CRISPR screens can be readily applied for studies in cancer immunity, asthese models are fully functional in immunocompetent animals. A key question on the forefrontof immunooncology is to understandwhy only a fraction of patients respond to immunotherapy,such as checkpoint inhibitors. Multiplexed AAV-CRISPR screens could be applied in thecontext of checkpoint blockade to deduce which mutations are associated with sensitivityto immunotherapy, as well as those conferring primary or acquired resistance. With furthertechnological development, direct in vivo screens on primary immune cell populations may alsobecome feasible, allowing high-throughput interrogation of factors that regulate immuneresponses against tumors.

As these technologies continue to develop and mature, they can be adapted for personalizedcancer modeling and tumor driver profiling as well as the identification of novel and relevanttherapeutic targets. Emerging new technologies such as direct in vivo CRISPR screenscontinue to transform oncology discovery.

Figure 3. Applications and Extensions of Direct In Vivo Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) Screens. (A) Direct invivo CRISPR screens can be readily used to generate personalized mouse avatars. CRISPR libraries can be customized to the mutations present in a given patient’stumor. Following autochthonous mutagenesis, therapeutic candidates can be evaluated using the mouse avatars, informing clinical decision-making. (B) Higher-dimensional screens (i.e., double, triple, n-tuple knockouts) using CRISPR-associated endonuclease in Prevotella and Francisella 1 (CRISPR–Cpf1) offer an elegant,high-throughput approach to investigate genetic interactions. Such studies could uncover synergistic driver mutations and synthetically lethal combinations, whichmay help to inform patient prognostication and to identify novel therapeutic vulnerabilities. (C) Direct in vivo CRISPR mutagenesis drives the formation of geneticallycomplex, multiclonal tumors. Coupled with single-cell capture sequencing and RNA-seq, direct in vivo CRISPR screens offer a powerful approach for the generationand subsequent dissection of tumor heterogeneity. (D) In vivoCRISPR screens thus far have lacked spatial resolution, as single-guide RNA (sgRNA) sequencing and/orcapture sequencing is performed on genomic DNA extracted from dissociated cell suspensions. By combining direct in vivo CRISPR mutagenesis with multiplexed insitu hybridization, it may be feasible to perform clonal analysis and phenotype–genotype matching on tissue sections.

8 Trends in Cancer, Month Year, Vol. xx, No. yy

TRECAN 252 No. of Pages 10

AcknowledgmentsThe authors thank all members in the Chen laboratory for their support. S.C. is supported by the Yale SBI/Genetics Startup

Fund, Damon Runyon (DRG-2117-12; DFS-13-15), the Melanoma Research Alliance (412806, 16-003524), St Baldrick’s

Foundation (426685), the Breast Cancer Alliance, the Cancer Research Institute (CLIP), the AACR (499395), AACR (17-20-

01-CHEN), The Mary Kay Foundation (017-81), The V Foundation (V2017-022), the DoD (W81XWH-17-1-0235), and the

NIH/NCI (1U54CA209992, 5P50CA196530-A10805, 4P50CA121974-A08306). R.D.C. is supported by an NIH MSTP

training grant (T32GM007205).

References

1. Vogelstein, B. et al. (2013) Cancer genome landscapes. Science

339, 1546–1558

2. Chin, L. et al. (2011) Making sense of cancer genomic data.Genes Dev. 25, 534–555

3. Garraway, L.A. and Lander, E.S. (2013) Lessons from the cancergenome. Cell 153, 17–37

4. Hsu, P.D. et al. (2014) Development and applications of CRISPR–Cas9 for genome engineering. Cell 157, 1262–1278

5. Cong, L. et al. (2013) Multiplex genome engineering usingCRISPR/Cas systems. Science 339, 819–823

6. Jinek, M. et al. (2012) A programmable dual-RNA-guided DNAendonuclease in adaptive bacterial immunity. Science 337,816–821

7. Gilbert, L.A. et al. (2013) CRISPR-mediated modular RNA-guidedregulation of transcription in eukaryotes. Cell 154, 442–451

8. Qi, L.S. et al. (2013) Repurposing CRISPR as an RNA-guidedplatform for sequence-specific control of gene expression. Cell152, 1173–1183

9. Gilbert, L.A. et al. (2014) Genome-scale CRISPR-mediated con-trol of gene repression and activation. Cell 159, 647–661

10. Hilton, I.B. et al. (2015) Epigenome editing by a CRISPR–Cas9-based acetyltransferase activates genes from promoters andenhancers. Nat. Biotechnol. 33, 510–517

11. Komor, A.C. et al. (2016) Programmable editing of a target base ingenomic DNA without double-stranded DNA cleavage. Nature533, 420–424

12. Gaudelli, N.M. et al. (2017) Programmable base editing of A�T toG�C in genomic DNA without DNA cleavage. Nature 551,464–471

13. Liu, X.S. et al. (2016) Editing DNA methylation in the mammaliangenome. Cell 167, 233–247 e17

14. Morita, S. et al. (2016) Targeted DNA demethylation in vivo usingdCas9–peptide repeat and scFv–TET1 catalytic domain fusions.Nat. Biotechnol. 34, 1060–1065

15. Vojta, A. et al. (2016) Repurposing the CRISPR–Cas9 system fortargeted DNA methylation. Nucleic Acids Res. 44, 5615–5628

16. Morgan, S.L. et al. (2017) Manipulation of nuclear architecturethrough CRISPR-mediated chromosomal looping. Nat. Com-mun. 8, 15993

17. Shalem, O. et al. (2015) High-throughput functional genomicsusing CRISPR–Cas9. Nat. Rev. Genet. 16, 299–311

18. Montalbano, A. et al. (2017) High-throughput approaches topinpoint function within the noncoding genome. Mol. Cell 68,44–59

19. Shalem, O. et al. (2014) Genome-scale CRISPR–Cas9 knockoutscreening in human cells. Science 343, 84–87

20. Wang, T. et al. (2014) Genetic screens in human cells using theCRISPR–Cas9 system. Science 343, 80–84

21. Doench, J.G. et al. (2016) Optimized sgRNA design to maximizeactivity and minimize off-target effects of CRISPR–Cas9. Nat.Biotechnol. 34, 184–191

22. Evers, B. et al. (2016) CRISPR knockout screening outperformsshRNA and CRISPRi in identifying essential genes. Nat. Biotech-nol. 34, 631–633

23. Sánchez-Rivera, F.J. and Jacks, T. (2015) Applications of theCRISPR–Cas9 system in cancer biology. Nat. Rev. Cancer 15,387–395

24. Hou, P. et al. (2017) A genome-wide CRISPR screen identifiesgenes critical for resistance to FLT3 inhibitor AC220.Cancer Res.77, 4402–4413

25. Konermann, S. et al. (2015) Genome-scale transcriptional activa-tion by an engineered CRISPR–Cas9 complex. Nature 517,583–588

26. Kong, X. et al. (2017) Cancer drug addiction is relayed by anERK2-dependent phenotype switch. Nature 550, 270–274

27. Ruiz, S. et al. (2016) A genome-wide CRISPR screen identifiesCDC25A as a determinant of sensitivity to ATR inhibitors. Mol.Cell 62, 307–313

28. Sanjana, N.E. et al. (2016) High-resolution interrogation of func-tional elements in the noncoding genome. Science 353,1545–1549

29. Han, K. et al. (2017) Synergistic drug combinations for canceridentified in a CRISPR screen for pairwise genetic interactions.Nat. Biotechnol. 35, 463–474

30. Shen, J.P. et al. (2017) Combinatorial CRISPR–Cas9 screens forde novo mapping of genetic interactions. Nat. Methods 14,573–576

31. Burr, M.L. et al. (2017) CMTM6 maintains the expression ofPD-L1 and regulates anti-tumour immunity. Nature 549,101–105

32. Hart, T. et al. (2015) High-resolution CRISPR screens revealfitness genes and genotype-specific cancer liabilities. Cell 163,1515–1526

33. Wang, T. et al. (2015) Identification and characterization of essen-tial genes in the human genome. Science 350, 1096–1101

34. Wang, T. et al. (2017) Gene essentiality profiling reveals genenetworks and synthetic lethal interactions with oncogenic Ras.Cell 168, 890–903 e15

35. Quail, D. and Joyce, J. (2013) Microenvironmental regulationof tumor progression and metastasis. Nat. Med. 19,1423–1437

36. Herbst, R.S. et al. (2014) Predictive correlates of response to theanti-PD-L1 antibody MPDL3280A in cancer patients. Nature515, 563–567

37. Sharma, P. et al. (2017) Primary, adaptive, and acquired resis-tance to cancer immunotherapy. Cell 168, 707–723

38. Chen, D.S. and Mellman, I. (2013) Oncology meets immunology:the cancer–immunity cycle. Immunity 39, 1–10

39. Patel, S.P. and Kurzrock, R. (2015) PD-L1 expression as apredictive biomarker in cancer immunotherapy. Mol. CancerTher. 14, 847–856

40. Gordon, S.R. et al. (2017) PD-1 expression by tumour-associatedmacrophages inhibits phagocytosis and tumour immunity.Nature545, 495–499

41. Chen, S. et al. (2015) Genome-wide CRISPR screen in a mousemodel of tumor growth and metastasis. Cell 160, 1246–1260

42. Katigbak, A. et al. (2016) A CRISPR/Cas9 functional screenidentifies rare tumor suppressors. Sci. Rep. 6, 38968

43. Song, C.-Q. et al. (2017) Genome-wide CRISPR screen identifiesregulators of mitogen-activated protein kinase as suppressors ofliver tumors in mice. Gastroenterology 152, 1161–1173 e1

44. Kodama, M. et al. (2017) In vivo loss-of-function screens identifyKPNB1 as a new druggable oncogene in epithelial ovarian can-cer. Proc. Natl. Acad. Sci. U. S. A. 114, E7301–E7310

Trends in Cancer, Month Year, Vol. xx, No. yy 9

TRECAN 252 No. of Pages 10

45. Braun, C.J. et al. (2016) Versatile in vivo regulation of tumorphenotypes by dCas9-mediated transcriptional perturbation.Proc. Natl. Acad. Sci. U. S. A. 113, E3892–E3900

46. Yau, E.H. et al. (2017) Genome-wide CRISPR screen for essentialcell growthmediators in mutant KRAS colorectal cancers.CancerRes. 77, 6330–6339

47. Patel, S.J. et al. (2017) Identification of essential genes for cancerimmunotherapy. Nature 548, 537–542

48. Manguso, R.T. et al. (2017) In vivo CRISPR screening identifiesPtpn2 as a cancer immunotherapy target. Nature 547, 413–418

49. Xue, W. et al. (2014) CRISPR-mediated direct mutation of cancergenes in the mouse liver. Nature 514, 380–384

50. Sánchez-Rivera, F.J. et al. (2014) Rapid modelling of cooperatinggenetic events in cancer through somatic genome editing.Nature516, 428–431

51. Swiech, L. et al. (2015) In vivo interrogation of gene function in themammalian brain using CRISPR–Cas9.Nat. Biotechnol. 33, 102–106

52. Platt, R.J. et al. (2014) CRISPR–Cas9 knockin mice for genomeediting and cancer modeling. Cell 159, 440–455

53. Dow, L.E. et al. (2015) Inducible in vivo genome editing withCRISPR–Cas9. Nat. Biotechnol. 33, 390–394

54. Chu, V.T. et al. (2016) Efficient generation of Rosa26 knock-inmice using CRISPR/Cas9 in C57BL/6 zygotes. BMC Biotechnol.16, 4

55. Weber, J. et al. (2015) CRISPR/Cas9 somatic multiplex-muta-genesis for high-throughput functional cancer genomics in mice.Proc. Natl. Acad. Sci. U. S. A. 112, 13982–13987

56. Rogers, Z.N. et al. (2017) A quantitative and multiplexedapproach to uncover the fitness landscape of tumor suppressionin vivo. Nat. Methods 14, 737–742

57. Chow, R.D. et al. (2017) AAV-mediated direct in vivo CRISPRscreen identifies functional suppressors in glioblastoma. Nat.Neurosci. 20, 1329–1341

58. Wang, G. et al. (2018) Mapping a functional cancer genome atlasof tumor suppressors in mouse liver using AAV-CRISPR-medi-ated direct in vivo screening. Sci. Adv. 4, eaao5508

10 Trends in Cancer, Month Year, Vol. xx, No. yy

59. Ben-David, U. et al. (2017) Patient-derived xenografts undergomouse-specific tumor evolution. Nat. Genet. 49, 1567–1575

60. Rongvaux, A. et al. (2014) Development and function of humaninnate immune cells in a humanized mouse model. Nat. Biotech-nol. 32, 364–372

61. McGranahan, N. and Swanton, C. (2017) Clonal heterogeneityand tumor evolution: past, present, and the future. Cell 168, 613–628

62. Schmitt, M.W. et al. (2016) The influence of subclonal resistancemutations on targeted cancer therapy. Nat. Rev. Clin. Oncol. 13,335–347

63. McGranahan, N. and Swanton, C. (2015) Biological and thera-peutic impact of intratumor heterogeneity in cancer evolution.Cancer Cell 27, 15–26

64. Zetsche, B. et al. (2015) Cpf1 is a single RNA-guided endonu-clease of a class 2 CRISPR–Cas system. Cell 163, 759–771

65. Zetsche, B. et al. (2017) Multiplex gene editing by CRISPR–Cpf1using a single crRNA array. Nat. Biotechnol. 35, 31–34

66. Chow, R.D. et al. (2017) Mapping in vivo genetic interactomicsthrough Cpf1 crRNA array screening. bioRxiv Published onlineJune 21, 2017. http://dx.doi.org/10.1101/153486

67. Zhou, H. et al. (2018) In vivo simultaneous transcriptional activa-tion of multiple genes in the brain using CRISPR–dCas9-activatortransgenic mice. Nat. Neurosci. 21, 440–446

68. Dixit, A. et al. (2016) Perturb-seq: dissecting molecular circuitswith scalable single-cell RNA profiling of pooled genetic screens.Cell 167, 1853–1866 e17

69. Jaitin, D.A. et al. (2016) Dissecting immune circuits by linkingCRISPR-pooled screens with single-cell RNA-seq. Cell 167,1883–1896 e15

70. Datlinger, P. et al. (2017) Pooled CRISPR screening with single-cell transcriptome readout. Nat. Methods 14, 297–301

71. Chen, K.H. et al. (2015) RNA imaging. Spatially resolved, highlymultiplexed RNA profiling in single cells. Science 348, aaa6090