rna editing with crispr-cas13rna type 2) to transcripts in mammalian cells.this system, referred to...

RESEARCH ARTICLE◥

BIOTECHNOLOGY

RNA editing with CRISPR-Cas13David B. T. Cox,1,2,3,4,5,6* Jonathan S. Gootenberg,1,2,3,4,7* Omar O. Abudayyeh,1,2,3,4,6*Brian Franklin,1,2,3,4 Max J. Kellner,1,2,3,4 Julia Joung,1,2,3,4 Feng Zhang1,2,3,4†

Nucleic acid editing holds promise for treating genetic disease, particularly at the RNA level,where disease-relevant sequences can be rescued to yield functional protein products.TypeVICRISPR-Cas systems contain the programmable single-effector RNA-guided ribonucleaseCas13.We profiled typeVI systems in order to engineer a Cas13 ortholog capable of robustknockdown and demonstrated RNA editing by using catalytically inactive Cas13 (dCas13)to direct adenosine-to-inosine deaminase activity by ADAR2 (adenosine deaminase acting onRNA type 2) to transcripts in mammalian cells.This system, referred to as RNA Editing forProgrammable A to I Replacement (REPAIR), which has no strict sequence constraints, canbe used to edit full-length transcripts containing pathogenic mutations.We further engineeredthis system to create a high-specificity variant and minimized the system to facilitateviral delivery. REPAIR presents a promising RNA-editing platform with broad applicability forresearch, therapeutics, and biotechnology.

Precise nucleic acid–editing technologies arevaluable for studying cellular function andas novel therapeutics. Current editing tools,based on programmable nucleases such asthe prokaryotic CRISPR-associated nucle-

ases Cas9 (1–4) or Cpf1 (5), have been widelyadopted for mediating targeted DNA cleavage,which in turn drives targeted gene disruptionthrough nonhomologous end joining (NHEJ) orprecise gene editing through template-dependenthomology-directed repair (HDR) (6). NHEJ useshostmachineries that are active in both dividingand post-mitotic cells and provides efficient genedisruption by generating a mixture of insertionor deletion (indel) mutations that can lead toframe shifts in protein-coding genes. HDR, incontrast, ismediated by hostmachineries whoseexpression is largely limited to replicating cells.Accordingly, the development of gene-editingcapabilities for post-mitotic cells remains amajorchallenge. DNA base editors, consisting of a fu-sion between Cas9 nickase and cytidine deami-nase, can mediate efficient cytidine-to-uridineconversions within a target window and sub-stantially reduce the formation of double-strandbreak–induced indels (7, 8). However, the poten-tial targeting sites of DNAbase editors are limitedby the requirement of Cas9 for a protospacer ad-jacent motif (PAM) at the editing site (9). Here,we describe the development of a precise and

flexible RNA base editing technology using thetype VI CRISPR-associated RNA-guided ribonu-clease (RNase) Cas13 (10–13).Cas13 enzymes have two higher eukaryotes

and prokaryotes nucleotide-binding (HEPN)endoRNase domains that mediate precise RNAcleavage with a preference for targets with pro-tospacer flanking sites (PFSs) observed biochem-ically and in bacteria (10, 11). Three Cas13 proteinfamilies have been identified to date: Cas13a (pre-viously known as C2c2), Cas13b, and Cas13c (12, 13).We recently reported that Cas13a enzymes can beadapted as tools for nucleic acid detection (14) aswell as mammalian and plant cell RNA knockdownand transcript tracking (15), and observed that thebiochemical PFS was not required for RNA inter-ference with Cas13a (15). The programmable na-ture of Cas13 enzymes makes them an attractivestarting point to develop tools for RNA bindingand perturbation applications.The adenosine deaminase acting on RNA

(ADAR) family of enzymesmediates endogenousediting of transcripts via hydrolytic deaminationof adenosine to inosine, a nucleobase that isfunctionally equivalent to guanosine in transla-tion and splicing (16, 17). There are two function-al human ADAR orthologs, ADAR1 and ADAR2,which consist of N-terminal double-strandedRNA–binding domains and aC-terminal catalyticdeamination domain. Endogenous target sites ofADAR1 and ADAR2 contain substantial double-stranded identity, and the catalytic domains re-quire duplexed regions for efficient editing in vitroand in vivo (18, 19). The ADAR catalytic domain iscapable of deaminating target adenosines with-out any protein cofactors in vitro (20). ADAR1 hasbeen found to target mainly repetitive regions,whereas ADAR2 mainly targets nonrepetitivecoding regions (17). AlthoughADARproteins havepreferred motifs for editing that could restrictthe potential flexibility of targeting, hyperactive

mutants, such as ADAR2(E488Q) (21), relax se-quence constraints and increase adenosine-to-inosine editing rates. (Single-letter abbreviationsfor the amino acid residues are as follows: A, Ala;C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile;K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R,Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. Inthe mutants, other amino acids were substitutedat certain locations; for example, E488Q indicatesthat glutamic acid at position 488 was replacedby glutamine.) ADARs preferentially deaminateadenosines mispaired with cytidine bases in RNAduplexes (22), providing a promising opportunityfor precise base editing. Although previous ap-proaches have engineered targeted ADAR fusionsvia RNA guides (23–26), the specificity of theseapproaches has not been reported, and their re-spective targeting mechanisms rely on RNA-RNAhybridization without the assistance of proteinpartners that may enhance target recognitionand stringency.We assayed a subset of the family of Cas13 en-

zymes forRNAknockdownactivity inmammaliancells and identified the Cas13b ortholog fromPrevotella sp. P5-125 (PspCas13b) as the most effi-cient and specific for mammalian cell applica-tions. We then fused the ADAR2 deaminasedomain with the E488Qmutation (ADAR2DD) tocatalytically inactive PspCas13b anddemonstratedRNA editing for programmable A to I (G) replace-ment (REPAIR) of reporter and endogenous tran-scripts as well as disease-relevant mutations. Last,weuseda rationalmutagenesis scheme to improvethe specificity of dCas13b-ADAR2DD fusions in or-der to generateREPAIRv2withmore than919-foldhigher specificity.

Comprehensive characterization of Cas13family members in mammalian cells

Wepreviously developedLeptotrichiawadeiCas13a(LwaCas13a) for mammalian knockdown appli-cations, but it required a monomeric superfoldergreen fluorescent protein (msfGFP) stabilizationdomain for efficient knockdown, and although thespecificity was high, knockdown levels were notconsistently below 50% (15). We sought to iden-tify amore robust RNA-targeting CRISPR systemby characterizing a genetically diverse set of Cas13family members in order to assess their RNAknockdown activity inmammalian cells (Fig. 1A).We generated mammalian codon-optimized ver-sions of multiple Cas13 proteins, including 21orthologs of Cas13a, 15 of Cas13b, and seven ofCas13c, and cloned them into an expression vectorwith N- and C-terminal nuclear localization signal(NLS) sequences and a C-terminal msfGFP toenhance protein stability (table S1). To assayinterference in mammalian cells, we designed adual-reporter construct expressing the independentGaussia (Gluc) and Cypridina (Cluc) luciferasesunder separate promoters, allowing one luciferaseto function as a measure of Cas13 interferenceactivity and the other to serve as an internal con-trol. For each Cas13 ortholog, we designed PFS-compatible guide RNAs, using the Cas13b PFSmotifs derived from an ampicillin interferenceassay (fig. S1, table S2, and supplementary text)

RESEARCH

Cox et al., Science 358, 1019–1027 (2017) 24 November 2017 1 of 9

1Broad Institute of Massachusetts Institute of Technology(MIT) and Harvard, Cambridge, MA 02142, USA. 2McGovernInstitute for Brain Research, MIT, Cambridge, MA 02139,USA. 3Department of Brain and Cognitive Science, MIT,Cambridge, MA 02139, USA. 4Department of BiologicalEngineering, MIT, Cambridge, MA 02139, USA. 5Departmentof Biology, MIT, Cambridge, MA 02139, USA. 6Harvard-MITDivision of Health Sciences and Technology, MIT, Cambridge,MA 02139, USA. 7Department of Systems Biology, HarvardMedical School, Boston, MA 02115, USA.*These authors contributed equally to this work.†Corresponding author. Email: [email protected]

on January 27, 2020

http://science.sciencemag.org/

Dow

nloaded from


and the 3′H (not G) PFS from previous reports ofCas13a activity (10).We transfectedhumanembryonic kidney (HEK)

293FT cells with Cas13-expression, guide RNA,and reporter plasmids and then quantified levelsof Cas13 expression and the targetedGluc 48 hourslater (Fig. 1B and fig. S2A). Testing two guideRNAs for each Cas13 ortholog revealed a rangeof activity levels, including five Cas13b orthologswith similar or increased interference across bothguide RNAs relative to the recently characterizedLwaCas13a (Fig. 1B), andwe observed only aweakcorrelation between Cas13 expression and inter-ference activity (fig. S2, B to D). We selected thetop five Cas13b orthologs and the top two Cas13aorthologs for further engineering.We next tested Cas13-mediated knockdown of

Gluc without msfGFP to select orthologs that donot require stabilization domains for robust ac-tivity. We hypothesized that Cas13 activity couldbe affected by subcellular localization, as we pre-viously reported for optimization of LwaCas13a(15). Therefore, we tested the interference activityof the seven selected Cas13 orthologs C-terminallyfused to one of six different localization tagswithout msfGFP. Using the luciferase reporterassay, we identified the top three Cas13b designswith the highest level of interference activity:Cas13b fromPrevotella sp. P5-125 (PspCas13b) andCas13b fromPorphyromonas gulae (PguCas13b) C-terminally fused to the HIV Rev nuclear exportsequence (NES), and Cas13b from Riemerellaanatipestifer (RanCas13b) C-terminally fused to themitogen-activated protein kinase NES (fig. S3A).To further distinguish activity levels of the toporthologs,we compared the threeoptimizedCas13bconstructs with the optimal LwaCas13a-msfGFPfusionand to shorthairpin–mediatedRNA(shRNA)for their ability to knock down the endogenousKRAS (V-Ki-ras2 Kirsten rat sarcoma viral onco-gene homolog) transcript by using position-matched guides (fig. S3B). We observed thehighest levels of interference for PspCas13b (av-erage knockdown, 62.9%) and thus selected thisfor further comparison with LwaCas13a.To more rigorously define the activity of

PspCas13b and LwaCas13a, we designed position-matched guides tiling along both Gluc and Cluctranscripts and assayed their activity using ourluciferase reporter assay. We tested 93 and 20position-matched guides targeting Gluc and Cluc,respectively, and found that PspCas13b had con-sistently increased levels of knockdown relative toLwaCas13a (average of 92.3% for PspCas13b versus40.1% knockdown for LwaCas13a) (Fig. 1, C andD).

Specificity of Cas13 mammalianinterference activity

To characterize the interference specificities ofPspCas13b and LwaCas13a, we designed a plasmidlibrary of luciferase targets containing single mis-matches and double mismatches throughout thetarget sequence and the three flanking 5′ and 3′base pairs (fig. S3C). We transfected HEK293FTcells with either LwaCas13a or PspCas13b, a fixedguide RNA targeting the unmodified targetsequence, and the mismatched target library


Cas13a / Cas13c Cas1 Cas2

type VI-A/C CRISPR system

type VI-B CRISPR system

5' 3'spacer

5' 3'

type VI-A/C crRNA

type VI-B crRNA

Cas13b Csx27/Csx28

CRISPR array

direct repeat (DR)

0.0 0.3 0.6 0.9 1.2

0.0

0.3

0.6

0.9

1.2

guide 1 normalized RLU

guid

e 2

norm

aliz

ed R

LU

11 - P. gulae6 - R. anatipestifer

14 - P. gingivalis12 - Prevotella sp. P5-125

2 - P. intermediaL. wadei L. buccalis

Cas13aCas13bCas13c

norm

aliz

ed R

LU (

Clu

c)no

rmal

ized

RLU

(G

luc)

0 500 1000 1500position (nt)

LwaCas13a

PspCas13b

0 150 300 450

0.0

0.5

1.0

1.5

2.0 LwaCas13a

PspCas13b

position (nt)

0.0

0.5

1.0

1.5

2.0

log2(TPM+1) nontargeting guide

0 5 10 150

5

10

15

log 2(

TP

M+1

) ta

rget

ing

guid

e

log2(TPM+1) nontargeting guide

0 5 10 150

5

10

15

log 2(

TP

M+1

) ta

rget

ing

guid

e

0 50 100 150 200number of off-target transcripts

Cas13a

Cas13b

shRNA

(1)(0)

(180)

Gluc (shRNA)

Gluc (LwaCas13a)

Gluc (shRNA)

Gluc (PspCas13b)

Fig. 1. Characterization of a highly active Cas13b ortholog for RNA knockdown. (A) Schematicof stereotypical Cas13 loci and corresponding CRISPR RNA (crRNA) structure. (B) Evaluation of 19Cas13a, 15 Cas13b, and seven Cas13c orthologs for luciferase knockdown by use of two different guides.Orthologs with efficient knockdown by using both guides are labeled with their host organism name.Values are normalized to a nontargeting guide designed against the Escherichia coli LacZ transcript,withno homology to the human transcriptome. (C) PspCas13b and LwaCas13a knockdown activity (asmeasured by luciferase activity) by using tiling guides against Gluc. Values represent mean ± SEM.Nontargeting guide is the same as in (B). (D) PspCas13b and LwaCas13a knockdown activity (asmeasured by luciferase activity) by using tiling guides against Cluc. Values represent mean ± SEM.Nontargeting guide is the same as in (B). (E) Expression levels in log2[transcripts per million (TPM)+1]values of all genes detected in RNA-seq libraries of nontargeting control (x axis) compared withGluc-targeting condition (y axis) for LwaCas13a (red) and shRNA (black). Shown is the mean of threebiological replicates.The Gluc transcript data point is labeled. Nontargeting guide is the same as in (B).(F) Expression levels in log2[transcripts per million (TPM)+1] values of all genes detected in RNA-seqlibraries of nontargeting control (x axis) compared with Gluc-targeting condition (y axis) for PspCas13b(blue) and shRNA (black). Shown is the mean of three biological replicates.The Gluc transcript datapoint is labeled. Nontargeting guide is the same as in (B). (G) Number of significant off-targets from Glucknockdown for LwaCas13a, PspCas13b, and shRNA from the transcriptome-wide analysis in (E) and (F).

RESEARCH | RESEARCH ARTICLEon January 27, 2020


Dow

nloaded from


corresponding to the appropriate system. Wethen performed targeted RNA sequencing (RNA-seq) of uncleaved transcripts in order to quantifydepletion of mismatched target sequences. Wefound that LwaCas13a and PspCas13b had a cen-tral region that was relatively intolerant to singlemismatches, extending from base pairs 12 to 26for the PspCas13b target and 13 to 24 for theLwaCas13a target (fig. S3D). Double mismatcheswere even less tolerated than single mutations,with little knockdown activity observed over alarger window, extending from base pairs 12 to29 for PspCas13b and 8 to 27 for LwaCas13a intheir respective targets (fig. S3E). Additionally,because there are mismatches included in thethree nucleotides flanking the 5′ and 3′ ends ofthe target sequence, we could assess PFS con-straints on Cas13 knockdown activity. Sequenc-ing showed that almost all PFS combinationsallowed robust knockdown, indicating that aPFS constraint for interference in mammalian

cells likely does not exist for either enzyme tested.These results indicate that Cas13a and Cas13b dis-play similar sequence constraints and sensitiv-ities against mismatches.We next characterized the interference spec-

ificity of PspCas13b and LwaCas13a across themRNA fractionof the transcriptome.Weperformedtranscriptome-wide mRNA sequencing to detectsignificantdifferentially expressedgenes.LwaCas13aand PspCas13b demonstrated robust knockdownof Gluc (Fig. 1, E and F) and were highly specificcompared with a position-matched shRNA, whichshowed hundreds of off-targets (Fig. 1G), a findingconsistent with our previous characterization ofLwaCas13a specificity in mammalian cells (15).

Cas13-ADAR fusions enable targetedRNA editing

Given that PspCas13b achieved consistent, robust,and specific knockdown of mRNA inmammaliancells, we envisioned that it could be adapted as an

RNA-binding platform to recruit RNA-modifyingdomains, such as ADARDD, for programmableRNA editing. To engineer a PspCas13b lacking nu-clease activity (dPspCas13b, referred to as dCas13bhereafter), we mutated conserved catalytic resi-dues in the HEPN domains and observed loss ofluciferase RNA knockdown (fig. S4A). We hypoth-esized that a dCas13b-ADARDD fusion could berecruited by a guide RNA to target adenosines,with the hybridized RNA creating the requiredduplex substrate for ADAR activity (Fig. 2A). Toenhance target adenosine deamination rates, weintroduced two additional modifications to ourinitial RNA editing design:We introduced amis-matched cytidine opposite the target adenosine,which has been previously reported to increasedeamination frequency, and fused dCas13bwith thedeaminase domains of humanADAR1 or ADAR2containing hyperactivatingmutations in order toenhance catalytic activity [ADAR1DD(E1008Q) (27)or ADAR2DD(E488Q)] (21).


5048464442403836343230282624222018161412108642

NT

0 10 20 30

conversion oftarget A to I

A

dCas13bADARDD

target A specifiedby mismatching C

on guide RNA

5' 3'

guide RNA

target mRNA(Cluc W85X)

guide RNA

back

grou

nd s

ubtr

acte

d R

LU

mismatch distance mismatch distance

dCas13b-ADAR1DD(E1008Q) dCas13b-ADAR2DD(E488Q)

10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80

10

5

0

10

5

0

10

5

0

10

5

0

10

5

0

10

5

0

10

5

0

10

5

0

84 nt

70 nt

50 nt

30 nt

spacer length

Cluc W85X mutation (A45) A editing rate (%)

mis

mat

ch d

ista

nce

editi

ng ra

te (%

)

0

10

20

30

1 9 16 21 26 31 40 45 52 55 64 66 70 81 852 12 17 22 28 33 43 50 53 60 65 68 75 84 92

45A position

Fig. 2. Engineering dCas13b-ADAR fusions for RNA editing.(A) Schematic of RNA editing by dCas13b-ADARDD fusion proteins.dCas13b is fused to ADARDD, which naturally deaminates adenosinesto insosines in dsRNA. The crRNA specifies the target site by hybridizingto the bases surrounding the target adenosine, creating a dsRNA structurefor editing and recruiting the dCas13b-ADARDD fusion. A mismatched cytidinein the crRNA opposite the target adenosine enhances the editing reaction,promoting target adenosine deamination to inosine, a base that functionallymimics guanosine in many cellular reactions. (B) Schematic of Cypridinaluciferase W85X target and targeting guide design. Deamination of thetarget adenosine restores the stop codon to the wild-type tryptophan. Spacerlength is the region of the guide that contains homology to the targetsequence. Mismatch distance is the number of bases between the 3′ end

of the spacer and the mismatched cytidine. The cytidine mismatched baseis included as part of the mismatch distance calculation. (C) Quantificationof luciferase activity restoration for (left) dCas13b-ADAR1DD(E1008Q) and(right) dCas13b-ADAR2DD(E488Q) with tiling guides of length 30, 50, 70,or 84 nt. All guides with even mismatch distances are tested for each guidelength. Values are background-subtracted relative to a 30-nt nontargetingguide that is randomized with no sequence homology to the humantranscriptome. (D) Schematic of the sequencing window in which A-to-Iedits were assessed for Cypridina luciferase W85X. (E) Sequencingquantification of A-to-I editing for 50-nt guides targeting Cypridinaluciferase W85X. Blue triangle indicates the targeted adenosine. Foreach guide, the region of duplex RNA is outlined in red. Values representmean ± SEM. Nontargeting guide is the same as in (C).



Dow

nloaded from


To test the activity of dCas13b-ADARDD, wegenerated an RNA-editing reporter on Cluc byintroducing a nonsense mutation [W85X (UGG→UAG)], which could functionally be repaired tothe wild-type codon through A→I editing (Fig.2B) and then be detected as restoration of Clucluminescence.We evenly tiled guideswith spacers30, 50, 70, or 84 nucleotides (nt) long across thetarget adenosine so as to determine the optimalguide placement and design (Fig. 2C). We foundthat dCas13b-ADAR1DD(E1008Q) required lon-ger guides to repair the Cluc reporter, whereasdCas13b-ADAR2DD(E488Q) was functional withall guide lengths tested (Fig. 2C). We also foundthat the hyperactive E488Q mutation improvedediting efficiency as wild-type ADAR2DD displayedreduced luciferase restoration (fig. S4B). Fromthis demonstration of activity, we chose dCas13b-ADAR2DD(E488Q) for further characterization anddesignated this system RNA Editing for Program-mable A to I Replacement version 1 (REPAIRv1).To validate that restoration of luciferase activ-

ity was due to bona fide editing events, we di-rectly measured REPAIRv1-mediated editing ofCluc transcripts via reverse transcription andtargeted next-generation sequencing. We tested30- and 50-nt spacers around the target site andfound that both guide lengths resulted in the ex-

pected A-to-I edit, with 50-nt spacers achievinghigher editing percentages (Fig. 2, D and E, andfig. S4C). We also observed that 50-nt spacershad an increased propensity for editing at non-targeted adenosines within the sequencing win-dow, likely because of increased regions of duplexedRNA (Fig. 2E and fig. S4C).We next targeted an endogenous gene, PPIB.

We designed 50-nt spacers tiling PPIB and foundthat we could edit the PPIB transcript with up to28% editing efficiency (fig. S4D). To test whetherREPAIR could be further optimized, wemodifiedthe linker betweendCas13b andADAR2DD(E488Q)(fig. S4E and table S3) and found that linkerchoice modestly affected luciferase activity res-toration. Additionally, we tested the ability ofdCas13b and guide alone tomediate editing events,finding that the ADARDD is required for editing(fig. S5, A to D).

Defining the sequence parameters forRNA editing

Given that we could achieve precise RNA editingat a test site, we wanted to characterize the se-quence constraints for programming the systemagainst any RNA target in the transcriptome. Se-quence constraints could arise from dCas13b-targeting limitations, such as the PFS, or from

ADAR sequence preferences (28). To investigatePFS constraints on REPAIRv1, we designed aplasmid library that carryies a series of four ran-domized nucleotides at the 5′ end of a target siteon the Cluc transcript (Fig. 3A). We targeted thecenter adenosine within either a UAG or AACmotif and found that for both motifs, all PFSsdemonstrated detectable levels of RNA editing,with amajority of the PFSs having >50% editingat the target site (Fig. 3B). Next, we sought todetermine whether the ADAR2DD in REPAIRv1had any sequence constraints immediately flank-ing the targeted base, as has been reported pre-viously for ADAR2DD (28).We tested every possiblecombination of 5′ and 3′ flanking nucleotides di-rectly surrounding the target adenosine (Fig. 3C)and found that REPAIRv1 was capable of editingall motifs (Fig. 3D). Last, we analyzed whetherthe identity of the base opposite the target A inthe spacer sequence affected editing efficiency andfound that an A-C mismatch had the highest lu-ciferase restoration, in agreement with previousreports of ADAR2 activity, with A-G, A-U, andA-A having drastically reduced REPAIRv1 activ-ity (fig. S5E).

Correction of disease-relevant humanmutations using REPAIRv1

To demonstrate the broad applicability of theREPAIRv1 system for RNA editing in mamma-lian cells, we designed REPAIRv1 guides againsttwodisease-relevantmutations: 878G→A (AVPR2W293X) in X-linked nephrogenic diabetes insip-idus and 1517G→A (FANCCW506X) in Fanconianemia.We transfected expression constructs forcDNA of genes carrying these mutations intoHEK293FT cells and tested whether REPAIRv1could correct the mutations. Using guide RNAscontaining 50-nt spacers, wewere able to achieve35% correction ofAVPR2 and 23% correction ofFANCC (Fig. 4, A to D). We then tested the abil-ity of REPAIRv1 to correct 34 different disease-relevant G→A mutations (table S4) and foundthat we were able to achieve substantial editingat 33 sites with up to 28% editing efficiency (Fig.4E). The mutations we chose are only a fractionof the pathogenic G-to-A mutations (5739) in theClinVar database, which also includes an addi-tional 11,943 G-to-A variants (Fig. 4F and fig. S6).Because there are no strict sequence constraints(Fig. 3), REPAIRv1 is capable of potentially editingall of these disease-relevant mutations, espe-cially given that we observed editing regardlessof the target motif (Figs. 3C and 4G).Delivering the REPAIRv1 system to diseased

cells is a prerequisite for therapeutic use, and wetherefore sought to design REPAIRv1 constructsthat could be packaged into therapeutically rele-vant viral vectors, such as adeno-associated viral(AAV) vectors. AAV vectors have a packaging limitof 4.7 kb, which cannot accommodate the largesize of dCas13b-ADARDD [4473 base pairs (bp)]along with promoter and expression regulatoryelements. To reduce the size, we tested a varietyof N-terminal and C-terminal truncations ofdCas13 fused to ADAR2DD(E488Q) for RNA-editing activity. We found that all C-terminal


100

75

50

25

0

editi

ng r

ate

(%)

edit site motif

0

10

20

30

40

edit site motif

editi

ng r

ate

(%)

targeting guide RNAnon-targeting guide RNA

motif 3' base

mot

if 5'

bas

e

15 25 35editing rate (%)

identify PFS preference based onmotif associated with edited reads

read out ADAR editedmRNA by deep sequencing

5' 3'

randomizedPFS

Cypridina luciferase mRNA

Cas13 target siteeditingsite

5'

5'unedited reads

edited reads

or

Fig. 3. Measuring sequence flexibility for RNA editing by REPAIRv1. (A) Schematic of screenfor determining PFS preferences of RNA editing by REPAIRv1. A randomized PFS sequence iscloned 5′ to a target site for REPAIR editing. After exposure to REPAIR, deep sequencing ofreverse-transcribed RNA from the target site and PFS is used to associate edited reads with PFSsequences. (B) Distributions of RNA editing efficiencies for all 4-N PFS combinations at two differentediting sites. (C) Quantification of the percent editing of REPAIRv1 at ClucW85 across all possiblethree-base motifs. Values represent mean ± SEM. Nontargeting guide is the same as in Fig. 2C.(D) Heatmap of 5′ and 3′ base preferences of RNA editing at ClucW85 for all possible three-base motifs.



Dow

nloaded from


truncations tested were still functional and ableto restore luciferase signal (fig. S7), and the largesttruncation, C-terminal D984–1090 (total size ofthe fusion protein, 4152 bp) was small enough tofit within the packaging limit of AAV vectors.

Transcriptome-wide specificityof REPAIRv1

Although RNA knockdown with PspCas13b washighly specific in our luciferase tiling experiments,we observed off-target adenosine editing withinthe guide:target duplex (Fig. 2E). To see whetherthis was a widespread phenomenon, we tiled anendogenous transcript, KRAS, and measured thedegree of off-target editing near the target aden-

osine (Fig. 5A). We found that forKRAS, althoughthe on-target editing rate was 23%, there weremany sites around the target site that also haddetectable A-to-I edits (Fig. 5B).Because of the observed off-target editing with-

in the guide:target duplex, we initially evaluatedtranscriptome-wide off-targets by performingRNA-seq on all mRNAs with 12.5x coverage. Of all theediting sites across the transcriptome, the on-target editing site had the highest editing rate,with 89% A-to-I conversion. We also found thatthere was a substantial number of A-to-I off-target events, with 1732 off-targets in the target-ing guide condition and 925 off-targets in thenontargeting guide condition, with 828 off-targets

shared between the targeting and nontargetingguide conditions (Fig. 5, C and D). Given thehigh number of overlapping off-targets betweenthe targeting and nontargeting guide conditions,we reasoned that the off-targets may arise fromADARDD. To test this hypothesis, we repeatedthe Cluc-targeting experiment, this time com-paring transcriptome changes for REPAIRv1 witha targeting guide, REPAIRv1 with a nontargetingguide, REPAIRv1 alone, orADARDD(E488Q) alone(fig. S8). We found differentially expressed genesand off-target editing events in each condition(fig. S8, B and C). There was a high degree ofoverlap in the off-target editing events betweenADARDD(E488Q) and all REPAIRv1 off-target


SNPs (all

)

SNPs (G to

A)

path

ogen

ic

SNPs (all

)

path

ogen

ic

SNPs (G to

A)

0

20

40

60

80

100

Clin

Var

SN

Ps

(x 1

,000

)

0 200 400 600 800

0

10

20

30

40

50

number of ClinVar SNP with motifed

iting

rat

e (%

)

CAA

CAU

CAC

CAG AAU

AAC

UACUAU

AAA

AAG

GAA

UAG

GAC

GAG

GAUUAA

AVPR2 mRNA(878G>A, Trp293Ter) X-linked Nephrogenic diabetes insipidus

7 23 40 46 58 64

5' – – 3'

guid

e R

NA

3'

3'

3'

FANCC mRNA(1517G>A, Trp506Ter) Fanconi anemia, complementation group C

53 56 6256 28 33 38 45 4835

5' – – 3'

3'

3'

3'

– 5'

– 5'

– 5'

guid

e R

NA

editi

ng ra

te (%

)

10

20

30

40

50

IL2R

G F8

LDLR

CB

SH

BB

ALD

OB

DM

DS

MA

D4

BR

CA

2G

RIN

2AS

CN

9ATA

RD

BP

CF

TR

UB

E3A

SM

PD

1U

SH

2AM

EN

1C

8orf

37M

LH1

TS

C2

NF

1M

SH

6S

MN

1S

H3T

C2

DN

AH

5M

EC

P2

AD

GR

V1

AH

I1P

RK

NC

OL3

A1

BR

CA

1M

YB

PC

3A

PC

BM

PR

2

0

10

20

30

editi

ng r

ate

(%)

targeting guide RNAnon-targeting guide RNA

NT

mis

mat

ch d

ista

nce

35

AVPR2 G878A

100 20 30 40A editing rate (%)

37

33

7A

23A

40A

46A

58A

64A

46

FANCC G1517A

100 20 30 40

NT

mis

mat

ch d

ista

nce

35

32

37

5A

6A

28A

33A

35A

38A

45A

48A

53A

56A

62A

A editing rate (%)45

Fig. 4. Correction of disease-relevant mutations with REPAIRv1.(A) Schematic of target and guide design for targeting AVPR2 878G→A.(B) The 878G→A mutation (indicated by blue triangle) in AVPR2 iscorrected to varying levels by using REPAIRv1 with three different guidedesigns. For each guide, the region of duplex RNA is outlined in red.Values represent mean ± SEM. Nontargeting guide is the same as inFig. 2C. (C) Schematic of target and guide design for targeting FANCC1517G→A. (D) The 1517G→A mutation (indicated by blue triangle) inFANCC is corrected to varying levels by using REPAIRv1 with threedifferent guide designs. For each guide, the region of duplex RNA is

outlined in red. The heatmap scale bar is the same as in (B). Valuesrepresent mean ± SEM. Nontargeting guide is the same as in Fig. 2C.(E) Quantification of the percent editing of 34 different disease-relevantG→A mutations selected from ClinVar by using REPAIRv1. Nontargetingguide is the same as in Fig. 2C. (F) Analysis of all the possible G→Amutations that could be corrected by using REPAIR as annotatedin the ClinVar database. (G) The distribution of editing motifs for allG→A mutations in ClinVar is shown versus the editing efficiency byREPAIRv1 per motif as quantified on the Gluc transcript. Valuesrepresent mean ± SEM.



Dow

nloaded from


edits, supporting the hypothesis that REPAIR off-target edits are driven by dCas13b-independentADARDD(E488Q) editing events (fig. S8D).Next, we sought to compare two RNA-guided

ADAR systems that have beendescribed previously(fig. S9A). The first uses a fusion of ADAR2DD tothe small viral protein lambdaN (lN), which bindsto the BoxB-l RNAhairpin (24). A guideRNAwithdouble BoxB-l hairpins guides ADAR2DD(E488Q)to edit sites encoded in the guide RNA (25). Thesecond design uses full-length ADAR2 (ADAR2)and a guide RNAwith a hairpin that the double-strand RNA (dsRNA)–binding domains (dsRBDs)of ADAR2 recognize (23, 26). We analyzed theediting efficiency of these two systems comparedwith REPAIRv1 and found that the BoxB-ADAR2and full-length ADAR2 systems demonstrated 50and 34.5% editing rates, respectively, comparedwith the 89% editing rate achieved by REPAIRv1

(fig. S9, B to E). Additionally, the BoxB and full-length ADAR2 systems created 1814 and 66 ob-served off-targets, respectively, in the targetingguide conditions, compared with the 2111 off-targets in the REPAIRv1 targeting guide condition.All the conditions with the two ADAR2DD-basedsystems (REPAIRv1 and BoxB) showed a highpercentage of overlap in their off-targets, whereasthe full-length ADAR2 system had a largely dis-tinct set of off-targets (fig. S9F). The overlap inoff-targets between the targeting and nontarget-ing conditions and between REPAIRv1 and BoxBconditions suggests that ADAR2DD drives off-targets independent of dCas13 targeting (fig. S9F).

Improving specificity of REPAIR throughrational protein engineering

To improve the specificity of REPAIRv1, we usedstructure-guided protein engineering of ADAR2DD

(E488Q). Because of the guide-independent na-ture of the off-targets, we hypothesized that de-stabilizing ADAR2DD(E488Q)–RNA binding wouldselectively decrease off-target editing, but main-tain on-target editing because of increased lo-cal concentration from dCas13b tethering ofADAR2DD(E488Q) to the target site.Wemutatedresidues in ADAR2DD(E488Q) previously deter-mined to contact the duplex region of the tar-get RNA (Fig. 6A) (19). To assess efficiency andspecificity, we tested 17 single mutants withboth targeting and nontargeting guides, underthe assumption that background luciferase re-storation in the nontargeting condition wouldbe indicative of broader off-target activity. Wefound that mutations at the selected residueshad substantial effects on the luciferase activityfor targeting and nontargeting guides (Fig. 6,A and B, and fig. S10A). A majority of mutantseither significantly improved the luciferase ac-tivity for the targeting guide or increased theratio of targeting to nontargeting guide ac-tivity, which we termed the specificity score(Fig. 6, A and B).We selected a subset of these mutants (Fig.

6B) for transcriptome-wide specificity profil-ing by next-generation sequencing. As expected,off-targets measured from transcriptome-widesequencing correlated with our specificity score(fig. S10B) for mutants. We found that with theexception of ADAR2DD(E488Q/R455E), all se-quenced REPAIRv1 mutants could effective-ly edit the reporter transcript (Fig. 6C), withmany mutants showing reduction in the num-ber of off-targets (Fig. 6C and figs. S10C andS11). We further explored motifs surroundingoff-targets for the various specificity mutantsand found that REPAIRv1 and most of the en-gineered variants exhibited a strong 3′ G pref-erence for their off-target edits, which is inagreement with the characterized ADAR2motif(fig. S12A) (28).We focused on the mutant ADAR2DD(E488Q/

T375G)—because it had the highest percent edit-ing of the four mutants with the lowest numbersof transcriptome-wide off-targets—and termed itREPAIRv2. Compared with REPAIRv1, REPAIRv2exhibited increased specificity, with a reductionfrom 18,385 to 20 transcriptome-wide off-targetswith high-coverage sequencing (125x coverage,10 ng of REPAIR vector transfected) (Fig. 6D).In the region surrounding the targeted adeno-sine in Cluc, REPAIRv2 also had reduced off-target editing, visible in sequencing traces (Fig.6E). In motifs derived from the off-target sites,REPAIRv1 presented a strong preference toward3′ G but showed off-target edits for all motifs(fig. S12B); by contrast, REPAIRv2 only editedthe strongest off-target motifs (fig. S12C). Thedistribution of edits on transcripts was heavilyskewed for REPAIRv1, with highly edited geneshaving more than 60 edits (fig. S13A), whereasREPAIRv2 only edited one transcript (EEF1A1)multiple times (fig. S13B). REPAIRv1 off-targetedits were predicted to result in numerous vari-ants, including 1000 missense base changes(fig. S13C), with 93 events in genes related to


KRAS mRNA

Cluc (254 A>I)

1 6 11 16 21 G Cchromosome

0

20

40

60

80

100

editi

ng r

ate

(%)

editi

ng r

ate

(%)

REPAIRv1 + targeting guide (1732 off-targets)12.5X transcriptome coverage


0

20

40

60

80

100REPAIRv1 + non-targeting guide (925 off-targets)

12.5X transcriptome coverage

3'

3'

3'

3'

3'

3'

3'

– 5'

– 5'

– 5'

– 5'

– 5'

– 5'

– 5'

24 35 41 53 57 60 69 72 74

– 3'5' –

0 10 20 30

50

42

34

26

18

10

2

NT

mis

mat

ch d

ista

nce

KRAS target A

editi

ng ra

te (%

)

0

10

20

30

A editing rate (%)35

24A

25A

26A

35A

41A

53A

57A

58A

60A

69A

72A

74A

KRAS target A

Fig. 5. Characterizing specificity of REPAIRv1. (A) Schematic of KRAS target site and guide design.(B) Quantification of percent A-to-I editing for tiled KRAS-targeting guides. Editing percentages are shownfor the on-target (blue triangle) and neighboring adenosine sites. For each guide, the region of duplex RNA isoutlined in red. Values represent mean ± SEM. (C) Transcriptome-wide sites of significant RNA editing byREPAIRv1 (150 ng REPAIR vector transfected) with Cluc targeting guide. The on-target site Cluc site(254 A→I) is highlighted in orange. (D) Transcriptome-wide sites of significant RNA editing by REPAIRv1 (150 ngREPAIR vector transfected) with nontargeting guide. Nontargeting guide is the same as in Fig. 2C.



Dow

nloaded from


cancer processes (fig. S13D). In contrast, REPAIRv2only had six predicted missense changes (fig. S13E),none of which were in cancer-related genes (fig.S13F). Analysis of the sequence surrounding off-target edits for REPAIRv1 or -v2 did not reveal

homology to guide sequences, suggesting thatoff-targets are likely dCas13b-independent (fig.S14), which is consistent with the high overlapof off-targets between REPAIRv1 and the ADAR2deaminase domain (fig. S8D). To directly compare

REPAIRv2 with other programmable ADAR sys-tems, we repeated our Cluc-targeting experimentswith all systems at two different dosages of ADARvector, finding that REPAIRv2 had comparableon-target editing with that of BoxB and ADAR2


Fig. 6. Rational mutagenesis of ADAR2DD

to improve the specificity of REPAIRv1.(A) Quantification of luciferase signalrestoration (on-target score, red boxes)by various dCas13-ADAR2DD mutantsas well as their specificity score (blue boxes)plotted along a schematic of the contactsbetween key ADAR2 deaminase residuesand the dsRNA target (the target strandis shown in gray; the nontarget strandis shown in red). All deaminase mutationswere made on the dCas13-ADAR2DD(E488Q)background. The specificity score isdefined as the ratio of the luciferase signalbetween targeting guide and nontargetingguide conditions. [Schematic of ADAR2deaminase domain contacts with dsRNAis adapted from (20)] (B) Quantification ofluciferase signal restoration by variousdCas13-ADAR2 mutants versus theirspecificity score. Nontargeting guide isthe same as in Fig. 2C. (C) Quantificationof on-target editing and the numberof significant off-targets for eachdCas13-ADAR2DD(E488Q) mutant bytranscriptome-wide sequencing ofmRNAs. Values represent mean ± SEM.Nontargeting guide is the same as in Fig. 2C.(D) Transcriptome-wide sites of significantRNA editing with (top) REPAIRv1 and(bottom) REPAIRv2, with a guide targeting apretermination site in Cluc. The on-targetCluc site (254 A→I) is highlighted in orange.Ten nanograms of REPAIR vector wastransfected for each condition. (E) Repre-sentative RNA-seq reads surroundingthe on-target Cluc-editing site (254 A→I;blue triangle) highlighting the differences inoff-target editing between (top) REPAIRv1and (bottom) REPAIRv2. A-to-I editsare highlighted in red; sequencing errorsare highlighted in blue. Gaps reflect spacesbetween aligned reads. Nontargeting guide isthe same as in Fig. 2C. (F) RNA editingwith REPAIRv1 and REPAIRv2, with guidestargeting an out-of-frame UAG site in theendogenous KRAS and PPIB transcripts.The on-target editing fraction is shown as asideways bar chart on the right for eachcondition row. For each guide, the regionof duplex RNA is outlined in red. Valuesrepresent mean ± SEM. Nontargeting guideis the same as in Fig. 2C.

50010001500

specificity score

246

on-target score

R348R474

5' –

3' –

side-chain contact

backbone contact

R510 Q488R481T490R477 S495

R348E

V351L

R455

U C C C C A C A U U G

N

A C G U U C A

V351 C451 E396 T375 K376 S486

R474E R477E R481E

S486T

S495T R510E E488Q

N473 K475

N473D K475Q

specificity scoreon-target score

CC G G C A A G U

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

OHO

O

P

HOO

HO

HO

P P P P P P P P P P P P P P P P P P

O

HO P

OH O

P

OH O OH O OH O OH O OH O OH O OH O OH O OH

O OH

O OH O OH O OH O OH O OH O OH O OH O OH

U

HOO

O

P

OH

P P P P P P P P P P P P P P P P P

P

A G G G G U G U A AA

T375G SESG

R455

SAT490

5'

3'

0 2 4 60

500

1000

1500

2000T375G

T375S

N473D

R455E R510E

R348ET490A

V351L

R455G

on-target score (targeting RLU)

spec

ifici

ty s

core

(tar

getin

g R

LU /

non-

targ

etin

g R

LU)

E488Q

8 0255075100target A editing rate (%)

targ

etin

gno

n-ta

rget

ing

0 0.5 1 1.5 2

E488QT490AR510EN473DR455GR455ET375GT375SV351LR348EE488QT490AR510EN473DR455GR455ET375GT375SV351LR348E

off-target count (x1000)

REPAIRv2 + targeting guide (20 off-targets)125X transcriptome coverage

Cluc (254 A>I)

Cypridina luciferase mRNA (REPAIRv1 edited)

Cypridina luciferase mRNA (REPAIRv2 edited)

0

20

40

60

80

100

editi

ng r

ate

(%)

chromosome


A55 editing rate (%)

A46 editing rate (%)0 20 40

30

60

0 10 20 40

REPAIRv1 + KRAS guide 2

REPAIRv2 + KRAS guide 2REPAIRv1 + NT guide

REPAIRv2 + NT guideEGFP control

KRAS target 2

17 21 22 24 26 27 28 37 43 55 59 60 62 71 74

15 21 25 28 29 34 36 39 40 46 7342 56 74 75

REPAIRv1 + PPIB guide 1

REPAIRv2 + PPIB guide 1REPAIRv1 + NT guide

REPAIRv2 + NT guideEGFP control

PPIB target 1

editi

ng ra

te (%

)

0

10

30

20

40

50

editi

ng ra

te (%

)

0

10

30

20

40

Cluc (254 A>I)

REPAIRv1 + targeting guide (18,385 off-targets)125X transcriptome coverage

1 6 11 16 21 G C0

20

40

60

80

100

editi

ng r

ate

(%)





Dow

nloaded from


but with substantially fewer off-target editingevents at both dosages (fig S15). REPAIRv2 hadenhanced specificity compared with REPAIRv1at both dosages (fig. S15B), a finding that alsoextended to two guides targeting distinct siteson PPIB (fig. S16, A to D). It is also worth notingthat in general, the lower-dosage condition (10 ngREPAIR vector) had fewer off-targets than thatof the higher dosage condition (150 ng REPAIRvector) (fig. S5).To assess editing specificity with greater sen-

sitivity, we sequenced the low-dosage condition(10 ng of transfected DNA) of REPAIRv1 and v2at much higher sequencing depth (125x coverageof the transcriptome). Increased numbers of off-targets were found at higher sequencing depthscorresponding to detection of rarer off-targetevents (fig. S17). Furthermore, we speculated thatdifferent transcriptome states could also poten-tially alter the number of off-targeting events.Therefore, we tested REPAIRv2 activity in theosteosarcoma U2OS cell line, observing six andseven off-targets for the targeting and non-targeting guide, respectively (fig. S18).We targeted REPAIRv2 to endogenous genes

to test whether the specificity-enhancing muta-tions reduced nearby edits in target transcriptswhile maintaining high-efficiency on-target edit-ing. For guides targeting either KRAS or PPIB,we found that REPAIRv2 had no detectable off-target edits, unlike REPAIRv1, and could effec-tively edit the on-target adenosine at efficienciesof 27.1% (KRAS) or 13% (PPIB) (Fig. 6F). This spec-ificity extended to additional target sites, includingregions that demonstrate high levels of back-ground in nontargeting conditions for REPAIRv1,such as other KRAS or PPIB target sites (fig. S19).Overall, REPAIRv2 eliminated off-targets in du-plexed regions around the edited adenosine andshowed dramatically enhanced transcriptome-wide specificity.

Discussion

We showhere that the RNA-guidedRNA-targetingtype VI-B CRISPR effector Cas13b is capable ofhighly efficient and specific RNA knockdown,providing the basis for improved tools for inter-rogating essential genes and noncoding RNA aswell as controlling cellular processes at the tran-script level. Catalytically inactive Cas13b (dCas13b)retains programmable RNA-binding capability,which we leveraged here by fusing dCas13b tothe adenosine deaminase domain of ADAR2 toachieve precise A-to-I edits, a system we termREPAIRv1. Further engineering of the systemproduced REPAIRv2, which has dramaticallyhigher specificity than previously described RNA-editing platforms (25, 29) while maintaining highlevels of on-target efficacy.Although Cas13b exhibits high fidelity, our ini-

tial results with dCas13b-ADAR2DD(E488Q) fu-sions revealed a substantial number of off-targetRNA editing events. To address this, we used arationalmutagenesis strategy to vary theADAR2DDresidues that contact the RNA duplex, identify-ing a variant, ADAR2DD(E488Q/T375G), that iscapable of precise, efficient, and highly specific

editing when fused to dCas13b. Editing efficiencywith this variant was comparable with or betterthan that achieved with two currently availablesystems, BoxB-ADAR2DD(E488Q) or ADAR2 edit-ing. Moreover, the REPAIRv2 system created only20 observable off-targets in the whole transcrip-tome, which is at least an order of magnitudebetter than both alternative editing technologies.Although it is possible that ADAR could de-aminate adenosine bases on the DNA strand inRNA-DNA heteroduplexes (20), it is unlikely todo so in this case because Cas13b does not bindDNA efficiently and because REPAIR is cyto-plasmically localized. Additionally, the lack ofhomology of off-target sites to the guide sequenceand the strong overlap of off-targets with theADARDD(E488Q)–only condition suggest that off-targets are not mediated by off-target guidebinding. Deeper sequencing and novel inosineenrichment methods could further refine ourunderstanding of REPAIR specificity in the future.The REPAIR system offers many advantages

compared with other nucleic acid–editing tools.First, the exact target site can be encoded in theguide by placing a cytidine within the guideacross from the desired adenosine to create afavorable A-C mismatch ideal for ADAR-editingactivity. Second, Cas13 has no targeting sequenceconstraints, such as a PFS or PAM, and no motifpreference surrounding the target adenosine,allowing any adenosine in the transcriptome tobe potentially targeted with the REPAIR system.The lack of motif for ADAR editing, in contrastwith previous literature, is likely due to the in-creased local concentration of REPAIR at thetarget site owing to dCas13b binding. DNA baseeditors can target either the sense or antisensestrand, whereas the REPAIR system is limitedto transcribed sequences, constraining the totalnumber of possible editing sites. However, be-cause of the less constrained nature of target-ing with REPAIR, this system can effect moreedits within ClinVar (Fig. 4C) than Cas9-DNAbase editors. Third, the REPAIR system direct-ly deaminates target adenosines to inosines anddoes not rely on endogenous repair pathways togenerate desired editing outcomes. Therefore,REPAIR should be able to mediate efficient RNAediting even in post-mitotic cells such as neurons.Fourth, in contrast to DNA editing, RNA editingis transient and can be more easily reversed,allowing the potential for temporal control overediting outcomes. The transient nature of REPAIR-mediated edits will likely be useful for treatingdiseases caused by temporary changes in cellstate, such as local inflammation, and couldalso be used to treat disease by modifying thefunction of proteins involved in disease-relatedsignal transduction. For instance, REPAIR edit-ing would allow the recoding of some serine,threonine, and tyrosine residues that are thetargets of kinases (fig. S20). Phosphorylation ofthese residues in disease-relevant proteins af-fects disease progression for many disorders,includingAlzheimer’s disease andmultiple neuro-degenerative conditions (30). REPAIR might alsobe used to transiently or even chronically change

the sequence of expressed, risk-modifying G-to-Avariants so as to decrease the chance of enteringa disease state for patients. For instance, REPAIRcould be used to functionally mimic A-to-G allelesof IFIH1 that protect against autoimmune dis-orders such as type I diabetes, immunoglobulinA deficiency, psoriasis, and systemic lupus ery-thematosus (31, 32).The REPAIR system provides multiple op-

portunities for additional engineering. Cas13bpossesses pre–CRISPR-RNA (crRNA) process-ing activity (13), allowing for multiplex editingof multiple variants—any one of which alonemay not affect disease, but together might haveadditive effects and disease-modifying potential.Extension of our rational design approach, suchas combining promising mutations and directedevolution, could further increase the specificityandefficiencyof the system,whileunbiased screen-ing approaches could identify additional residuesfor improving REPAIR activity and specificity.Currently, the base conversions achievable by

REPAIR are limited to generating inosine fromadenosine; additional fusions of dCas13 with othercatalytic RNA editing domains, such as APOBEC(apolipoprotein B mRNA editing enzyme, cata-lytic polypeptide-like), could enable cytidine-to-uridine editing. Additionally, mutagenesis ofADAR could relax the substrate preference totarget cytidine, allowing for the enhanced spec-ificity conferred by the duplexed RNA substraterequirement to be exploited by C-to-U editors.Adenosine-to-inosine editing on DNA substratesmay also be possible with catalytically inactiveDNA-targeting CRISPR effectors, such as dCas9or dCpf1, either through formation of DNA-RNAheteroduplex targets (20) or mutagenesis of theADAR domain.We have demonstrated the use of the PspCas13b

enzyme as both an RNA-knockdown and RNA-editing tool. The dCas13b platform for program-mable RNA binding has many applications,including live transcript imaging, splicing mod-ification, targeted localization of transcripts,pulldown of RNA-binding proteins, and epi-transcriptomic modifications.We used dCas13to create REPAIR, adding to the existing suiteof nucleic acid–editing technologies. REPAIRprovides a new approach for treating geneticdisease or mimicking protective alleles and es-tablishes RNA editing as a useful tool for mod-ifying genetic function.

REFERENCES AND NOTES

1. P. D. Hsu, E. S. Lander, F. Zhang, Cell 157, 1262–1278(2014).

2. A. C. Komor, A. H. Badran, D. R. Liu, Cell 168, 20–36 (2017).3. L. Cong et al., Science 339, 819–823 (2013).4. P. Mali et al., Science 339, 823–826 (2013).5. B. Zetsche et al., Cell 163, 759–771 (2015).6. H. Kim, J. S. Kim, Nat. Rev. Genet. 15, 321–334 (2014).7. A. C. Komor, Y. B. Kim, M. S. Packer, J. A. Zuris, D. R. Liu,

Nature 533, 420–424 (2016).8. K. Nishida et al., Science 353, aaf8729 (2016).9. Y. B. Kim et al., Nat. Biotechnol. 35, 371–376 (2017).10. O. O. Abudayyeh et al., Science 353, aaf5573 (2016).11. S. Shmakov et al., Mol. Cell 60, 385–397 (2015).12. S. Shmakov et al., Nat. Rev. Microbiol. 15, 169–182 (2017).13. A. A. Smargon et al., Mol. Cell 65, 618–630.e7 (2017).14. J. S. Gootenberg et al., Science 356, 438–442 (2017).




Dow

nloaded from


15. O. O. Abudayyeh et al., Nature 550, 280–284 (2017).16. K. Nishikura, Annu. Rev. Biochem. 79, 321–349 (2010).17. M. H. Tan et al., Nature 550, 249–254 (2017).18. B. L. Bass, H. Weintraub, Cell 55, 1089–1098 (1988).19. M. M. Matthews et al., Nat. Struct. Mol. Biol. 23, 426–433 (2016).20. Y. Zheng, C. Lorenzo, P. A. Beal, Nucleic Acids Res. 45,

3369–3377 (2017).21. A. Kuttan, B. L. Bass, Proc. Natl. Acad. Sci. U.S.A. 109,

E3295–E3304 (2012).22. S. K. Wong, S. Sato, D. W. Lazinski, RNA 7, 846–858 (2001).23. M. Fukuda et al., Sci. Rep. 7, 41478 (2017).24. M. F. Montiel-Gonzalez, I. Vallecillo-Viejo, G. A. Yudowski,

J. J. Rosenthal, Proc. Natl. Acad. Sci. U.S.A. 110, 18285–18290(2013).

25. M. F. Montiel-González, I. C. Vallecillo-Viejo, J. J. Rosenthal,Nucleic Acids Res. 44, e157 (2016).

26. J. Wettengel, P. Reautschnig, S. Geisler, P. J. Kahle, T. Stafforst,Nucleic Acids Res. 45, 2797–2808 (2017).

27. Y. Wang, J. Havel, P. A. Beal, ACS Chem. Biol. 10, 2512–2519 (2015).28. K. A. Lehmann, B. L. Bass, Biochemistry 39, 12875–12884 (2000).29. T. Stafforst, M. F. Schneider, Angew. Chem. Int. Ed. Engl. 51,

11166–11169 (2012).

30. C. Ballatore, V. M. Lee, J. Q. Trojanowski, Nat. Rev. Neurosci. 8,663–672 (2007).

31. Y. Li et al., J. Invest. Dermatol. 130, 2768–2772 (2010).32. R. C. Ferreira et al., Nat. Genet. 42, 777–780 (2010).

ACKNOWLEDGMENTS

We thank R. Munshi, A. Baumann, A. Philippakis, and E. Banksfor computational guidance; C. Muus for sequencing assistance;A. Smargon for gene synthesis of a Cas13b orthologs 1–8;V. Verdine for assistance with RNA sequencing librarypreparation; and R. Macrae, R. Belliveau, and the entire F.Z.laboratory for discussions and support. D.B.T.C. is supported byaward T32GM007753 from the National Institute of GeneralMedical Sciences. The content is solely the responsibility of theauthors and does not necessarily represent the official views ofthe National Institute of General Medical Sciences or the NIH.O.O.A. is supported by a Paul and Daisy Soros Fellowship anda NIH F30 National Research Service Award 1F30-CA210382.F.Z. is a New York Stem Cell Foundation–Robertson Investigator.F.Z. is supported by NIH grants 1R01-HG009761, 1R01-MH110049,and 1DP1-HL141201; the Howard Hughes Medical Institute; theNew York Stem Cell Foundation; the Simons Foundation; the Paul

G. Allen Family Foundation; the Vallee Foundation; and J. andP. Poitras, R. Metcalfe, and D. Cheng. D.B.T.C., J.S.G., O.O.A.,and F.Z. are co-inventors on U.S. provisional patent applicationno. 62/525,181 relating to REPAIR and other applications relatingto CRISPR and CRISPR-Cas13 technology filed by the BroadInstitute. Sequencing data are available at Sequence Read Archiveunder BioProject accession no. PRJNA414340. The authors planto make the reagents widely available to the academic communitythrough Addgene and to provide software tools via the F.Z.laboratory website (www.genome-engineering.org) and GitHub(github.com/fengzhanglab).

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/358/6366/1019/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S20Tables S1 to S9References (33–38)

21 September 2017; accepted 16 October 2017Published online 25 October 201710.1126/science.aaq0180




Dow

nloaded from

https://github.com/fengzhanglab

http://www.sciencemag.org/content/358/6366/1019/suppl/DC1


RNA editing with CRISPR-Cas13David B. T. Cox, Jonathan S. Gootenberg, Omar O. Abudayyeh, Brian Franklin, Max J. Kellner, Julia Joung and Feng Zhang

originally published online October 25, 2017DOI: 10.1126/science.aaq0180 (6366), 1019-1027.358Science

, this issue p. 1019; see also p. 996Sciencecells.RNA knockdown and editing platform that allowed efficient and specific RNA depletion and correction in mammalian deaminase domain and used rational protein engineering to improve the resultant enzyme. These approaches yielded anspecific RNAs directly (see the Perspective by Yang and Chen). They fused Cas13b with the ADAR2 adenosine

took advantage of the ability of Cas13b, an effector from a type VI CRISPR-Cas system, to targetet al.disease. Cox Efficient and precise RNA editing to correct disease-relevant transcripts holds great promise for treating genetic

Precise transcriptome engineering

ARTICLE TOOLS http://science.sciencemag.org/content/358/6366/1019

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2017/10/24/science.aaq0180.DC1

CONTENTRELATED

http://stm.sciencemag.org/content/scitransmed/8/360/360ra134.fullhttp://stm.sciencemag.org/content/scitransmed/9/372/eaah3480.fullhttp://stm.sciencemag.org/content/scitransmed/9/411/eaan0820.fullhttp://science.sciencemag.org/content/sci/358/6362/432.full

REFERENCES

http://science.sciencemag.org/content/358/6366/1019#BIBLThis article cites 37 articles, 8 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2017, American Association for the Advancement of Science

on January 27, 2020


Dow

nloaded from

http://science.sciencemag.org/content/358/6366/1019

http://science.sciencemag.org/content/suppl/2017/10/24/science.aaq0180.DC1

http://science.sciencemag.org/content/sci/358/6362/432.full

http://stm.sciencemag.org/content/scitransmed/9/411/eaan0820.full

http://stm.sciencemag.org/content/scitransmed/9/372/eaah3480.full

http://stm.sciencemag.org/content/scitransmed/8/360/360ra134.full

http://science.sciencemag.org/content/358/6366/1019#BIBL

http://www.sciencemag.org/help/reprints-and-permissions

http://www.sciencemag.org/about/terms-service


rna editing with crispr-cas13rna type 2) to transcripts in mammalian cells.this system, referred to...

Documents