Near-infrared optogenetic engineering ofphotothermal nanoCRISPR for programmablegenome editingXiaohong Chena,1, Yuxuan Chena,1, Huhu Xina, Tao Wana, and Yuan Pinga,2

aCollege of Pharmaceutical Sciences, Zhejiang University, 310058 Hangzhou, China

Edited by Larry Zweifel, University of Washington, Seattle, WA, accepted by Editorial Board Member John A. Rogers December 9, 2019 (received for reviewJuly 17, 2019)

We herein report an optogenetically activatable CRISPR-Cas9nanosystem for programmable genome editing in the secondnear-infrared (NIR-II) optical window. The nanosystem, termednanoCRISPR, is composed of a cationic polymer-coated Au nanorod(APC) and Cas9 plasmid driven by a heat-inducible promoter. TheAPC not only serves as a carrier for intracellular plasmid deliverybut also can harvest external NIR-II photonic energy and convertit into local heat to induce the gene expression of the Cas9 endo-nuclease. Due to high transfection activity, the APC shows strongability to induce a significant level of disruption in different geno-mic loci upon optogenetic activation. Moreover, the precise controlof genome-editing activity can be simply programmed by finely tun-ing exposure time and irradiation time in vitro and in vivo and alsoenables editing at multiple time points, thus proving the sensitivityand inducibility of such an editing modality. The NIR-II optical fea-ture of nanoCRISPR enables therapeutic genome editing at deeptissue, by which treatment of deep tumor and rescue of fulminanthepatic failure are demonstrated as proof-of-concept therapeuticexamples. Importantly, this modality of optogenetic genome editingcan significantly minimize the off-target effect of CRISPR-Cas9 inmost potential off-target sites. The optogenetically activatableCRISPR-Cas9 nanosystem we have developed offers a useful toolto expand the current applications of CRISPR-Cas9, and also definesa programmable genome-editing strategy toward high precisionand spatial specificity.

CRISPR-Cas9 | gene delivery | spatiotemporal specificity |photoregulation | off target

The RNA-guided clustered, regularly interspaced, short palin-dromic repeats (CRISPR)-associated nuclease protein 9 (Cas9)

was originally an adaptive immune defense system, which manybacteria exploit to protect themselves from invading genetic ele-ments (1). It has been recently harnessed as an efficient tool forgenome editing in both single cells and whole organisms for a widerange of biomedical applications in biology, genetics, medicine,and so forth (2, 3). In principle, CRISPR-Cas9 is composed of asingle-guide RNA (sgRNA) for the identification of DNA targetsand a Cas9 endonuclease that can bind and process the recognizedDNA targets (4). CRISPR-Cas9–based genome-editing technologyoffers a powerful and reliable strategy for targeted modificationsof the genome, enabling the precise perturbation of virtually anygenomic sequence in living cells (2–5). Due to its genome-widespecificity and multiplexing capability, Cas9 and its variants haveshown great potential in the generation of loss-of-function animals(6), the correction of genetic disorders (7), functional genomescreening (8, 9), and the treatment of infectious diseases (10).Despite these exciting possibilities, the lack of temporal andspatial precision during the editing process has severely con-strained current CRISPR-Cas9 systems from complicated and di-verse genome-editing scenarios. Furthermore, off-target activityhas now become a major concern when the CRISPR-Cas9 systemis exploited for therapeutic purposes.

To improve the spatiotemporal specificity of Cas9-mediatedgenomic manipulation, recent efforts have been dedicated to thedevelopment of inducible CRISPR-Cas9 architectures to enablethe conditional control of Cas9 activity through either chemical(11–14) or optical (15, 16) means. By precisely limiting the timeof Cas9 function, off-target activity is also expected to be con-trolled by minimizing unwanted prolonged Cas9 activity (12, 16).Chemical methods mainly refer to the regulation of endonucleaseactivity of Cas9 through small molecule-triggered Cas9 bindingand self-splicing inteins (12, 13). Although a few examples havebeen illustrated for the temporal control of Cas9 activity (17–19),this strategy generally lacks spatial specificity and reversibility,making it difficult to be explored for in vivo studies. Furthermore,commonly used small molecules for chemical activation, such asrapamycin (14) and doxycycline (12, 13), may induce potentialcytotoxicity toward both edited and nonedited cells. As op-posed to chemical strategies, optical regulation of Cas9 func-tion is more favorable in terms of its noninvasiveness, spatialspecificity, and reversibility. In the past few years, several differentphotoactivatable systems have been adopted for the optical regu-lation of CRISPR-Cas9–based genome editing and transcrip-tional activation (20, 21). For example, a photoactivatable Cas9

Significance

The persistent Cas9 activity following on-target editing and thelack of spatial precision have severely constrained the currentCRISPR-Cas9 systems from complicated and diverse genome-editing contexts. Herein, we develop an optogenetically acti-vatable CRISPR-Cas9 nanosystem (termed nanoCRISPR), anddemonstrate that programmable and inducible genome editingcan be simply manipulated through photothermal regulation ofnanoCRISPR in the second near-infrared optical window. Thegenome-editing activity and spatial specificity can be preciselyprogrammed by photothermal activation of nanoCRISPR in vitroand in vivo, and the optogenetic regulation minimizes the off-target mutations of CRISPR-Cas9 in potential off-target sites. Thephotothermal nanoCRISPR offers a useful tool to expand thecurrent applications of CRISPR-Cas9 toward high precision andspatial specificity.

Author contributions: Y.P. designed research; X.C. and Y.C. performed research; X.C., Y.C.,H.X., T.W., and Y.P. analyzed data; and Y.P. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. L.Z. is a guest editor invited by theEditorial Board.

Published under the PNAS license.

Data Deposition: The datasets generated in this paper are available in the NCBI SequenceRead Archive (Bioproject ID PRJNA599254).1X.C. and Y.C. contributed equally to this work.2To whom correspondence may be addressed. Email: pingy@zju.edu.cn.

This article contains supporting information online at https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1912220117/-/DCSupplemental.

First published January 15, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.1912220117 PNAS | February 4, 2020 | vol. 117 | no. 5 | 2395–2405

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1

consisting of two split, deactivated Cas9 (dCas9) fragments andphotoinducible dimerization domains (magnets) was engineered toenable optogenetic control of CRISPR-Cas9 activity in humancells (16). Upon blue-light irradiation, the split Cas9 was fused tomagnet domains to recover its genome-editing activity, whichcould be simply switched off by extinguishing the irradiation. Morerecently, optogenetic anti-CRISPR variants comprising a powerfulCas9 inhibitor (hybrids of AcrIIA4) and a LOV2 photosensor wereengineered for the photoregulation of CRISPR-Cas9 activity(22). As photoirradiation enabled the release of dCas9 from theoptogenetic variant of AcrIIA4, the inhibited Cas9 activitycould be rapidly recovered to enable genome and epigenomeediting. Nevertheless, most optically controlled CRISPR-Cas9systems respond to photoactivation by blue light. This suggeststhese blue light-mediated activatable CRISPR-Cas9 systemsare not only difficult for deep-tissue penetration through turbidhuman tissues but also are potentially phototoxic in realisticgenome-editing applications. To address these issues, a far-redlight-mediated CRISPR-dCas9 device, which is built based onthe bacterial photoactivatable cyclic diguanylate monophosphate(c-di-GMP) synthase BphS and the c-di-GMP–responsive hybridtransactivator, has been recently developed for targeted epigeneticmodulation both in vitro and in vivo (23). The optical activationis based on the light-emitting diode array (400 to 730 nm),which affords moderate tissue penetration up to 5 mm (24).Most recently, a near-infrared upconversion-activated CRISPR-Cas9 nanoparticle system has been proposed for the optical con-trol of therapeutic gene editing toward cancer treatment (25).While the above studies revealed that infrared light is critical forthe regulation of genome editing and epigenome editing in vivo,precise CRISPR-Cas9 genome editing in a programmable, in-ducible manner has not been demonstrated yet, not to mentionthose for in vivo applications. In addition, off-target activity in-duced by light-controlled editing modalities still remains elusiveto date.We herein report a photoactivatable CRISPR-Cas9 nanosystem

for the optogenetic control of genome editing at the second near-infrared (NIR-II) optical window (1,000 to 1,700 nm). As shown inFig. 1, this CRISPR-Cas9 nanosystem is typically composed of acationic polymer-coated Au nanorod (APC) and the Cas9 plasmiddriven by a heat-inducible promoter, HSP70 (HSP-Cas9). Whereasthe cationic polymer is able to carry and deliver the plasmid intothe targeted cells, the Au nanorod serves as a photothermaltransducer to transform the harvested external light into intracel-lular local heat. As such, APC not only acts as the delivery carrierfor the plasmid delivery but also serves as an intracellular photo-thermal converter to trigger the transcription of Cas9 and sgRNA.By incorporating the expression vector with the Cas9 gene cloneddownstream of the heat-inducible HSP70 promoter, the elevatedlocal temperature subsequently offers a cue to promote the geneexpression of Cas9. Thus, Cas9 activity can be regulated by heat-induced gene expression and activated by photothermal signals.APC–plasmid is first internalized by the targeted cell throughcharge-mediated internalization, followed by the formation ofendosomes. After the endosomal escape, whereas the plasmidreleased from APC enters into the nucleus, APC is still retainedin the cytoplasm. Upon light irradiation at 1,064 nm, APC quicklygenerates localized heat in the intracellular microenvironment toinduce the transformation of the heat-shock factor (HSF) frominactive monomers to active trimers, which are capable of trans-locating into the nucleus. Then, the binding of the intranucleartrimers to the heat-shock element (HSE) of the HSP70 promoterresults in the activation of transcription (26). However, once thelight irradiation is switched off, the decreased temperature releasesthe bound trimer from the HSE, triggering the retransformation oftrimers back to monomers to inactivate the transcription process(27). Thus, APC acts as an optogenetic switch to regulate Cas9expression and activity with high spatial specificity. As NIR-II light

shows stronger tissue-penetration ability as compared with firstNIR (NIR-I) light (650 to 950 nm), the regulation of genomeediting in vivo is also afforded by APC through the optogeneticcontrol in the NIR-II optical window. As we found in our study,APC-mediated optogenetic activation and spatial control of geneexpression are demonstrated to direct Cas9 activity in a preciseand programmable manner and significantly reduce off-target ef-fects, thereby paving a safe way for in vivo therapeutic genomeediting and the spatial control of CRISPR-Cas9 in vitro and in vivo.

ResultsIn our study, the classic cetyltrimethylammonium bromide-mediatedsynthesis approach was used for the preparation of Au nanorods(ARs) (28, 29), and uniform ARs with an aspect ratio of 7.1(length 106.4 ± 14.1 nm; width 15.2 ± 3.3 nm) were obtained (SIAppendix, Fig. S1). Afterward, biocompatible polystyrene sulfonate(PSS), which acts as an interconnecting layer, was then coated onthe AR surface through electrostatic force to form PSS-coatedARs. Subsequently, β-cyclodextrin-polyethyleneimine, a cationicpolymer that has been well-demonstrated for the efficient trans-fection of plasmids both in vitro and in vivo (30), was assembled ontop of the PSS layer. The layer-by-layer assembly process to pre-pare APC was verified by zeta-potential analysis (Fig. 2A), wherethe final product APC showed a positive surface charge (+29.5 mV).ARs displayed a strong absorption in the NIR-II region, with anabsorption peak at ca. 1,070 nm (Fig. 2B). Noticeably, the as-sembly of polyelectrolytes on ARs barely affected the wavelengthof maximum absorption. Such an optical feature of NIR-II iscrucial for in vivo investigations. Upon continuous laser irradiationat 1,064 nm for 5 min, the temperature of the APC solutionquickly increased and achieved a plateau of 42 °C under a powerdensity at 0.33 W/cm2, as recorded by the infrared thermal camera(Fig. 2C). The maximum temperature generated by APC could befurther adjusted to 65 °C at a power density of 1.00 W/cm2. Therepeated heating and cooling of three cycles resulted in a similartemperature fluctuation (SI Appendix, Fig. S2) and laser irradia-tion merely changed the morphology of APC (SI Appendix, Fig.S3), thus demonstrating its good photothermal stability. As theoptimal temperature for the activation of the HSP70 promoter was∼42 °C (31), we also explored the irradiation mode that couldstabilize the temperature at this degree. By discontinuous irradi-ation, the temperature could be finely tuned to a narrow rangefrom 39.0 to 42.0 °C (SI Appendix, Fig. S4). Given the temperatureelevation would start from body temperature for in vivo activation,we explored fine temperature control starting from 37 °C by dis-continuous irradiation and found this irradiation approach couldlikewise control the temperature in an ideal range (41.5 to 42.0 °C)by slightly adjusting the discontinuous irradiation time (SI Ap-pendix, Fig. S5). In the meantime, high-angle annular dark-fieldscanning transmission electron microscopy (HAADF-STEM) andenergy-dispersive X-ray spectroscopy (EDS) mapping were per-formed to verify the layer-by-layer (LBL) structure of APC–HSP-Cas9 (Fig. 2D). The distribution of S, N, or P elements overlappedwell with the Au element. The LBL structure of APC–HSP-Cas9complexes was also confirmed by X-ray photoelectron spectros-copy (SI Appendix, Figs. S6 and S7). To demonstrate whether APCwas able to encapsulate the plasmid encoding Cas9, a gel elec-trophoresis assay was carried out. APC could completely inhibitplasmid DNA migration at an APC/plasmid weight ratio of 0.15,proving its excellent capability to condense and carry plasmidDNA for gene transfection (SI Appendix, Fig. S8). In the mean-time, bio-TEM images indicated that APC was primarily located inthe cytoplasm after GFP expression (Fig. 2E).In the current work, we constructed a Cas9-encoding plasmid

driven by an HSP70 promoter. The plasmid consists of a Cas9gene driven by the HSP70 promoter (SI Appendix, Table S1), anenhanced green fluorescent protein (EGFP) reporter, and a lu-ciferase reporter downstream of Cas9, all of which are separated by

2396 | www.pnas.org/cgi/doi/10.1073/pnas.1912220117 Chen et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

self-cleaving peptide P2A, followed by a segment of independentsgRNA sequence driven by the U6 promoter downstream of theluciferase reporter (SI Appendix, Fig. S9). Therefore, we firstchecked GFP expression after the intracellular delivery of APC–HSP-Cas9 complexes. As shown in Fig. 2F, very weak fluorescencegenerated from GFP was observed in the 293T cells without lasertreatment, implying low background activity. In sharp contrast,strong green fluorescence was observed after the light irradiationon the cells. Flow cytometry analysis indicated that after APC-mediated transfection and photothermal activation, the percent-age of GFP-positive cells reached more than 90% under laserirradiation, which is much higher than that from transfectionsupported by Lipofectamine 2000 (Lipo; 27.4% GFP-positivecells) or 25-kDa polyethyleneimine (PEI; 12.2% GFP-positivecells) at 42 °C (SI Appendix, Figs. S10 and S11). The high level ofgene expression was further corroborated by a luciferase reporterassay, where strong luciferase expression was detected when the

transfection was mediated by APC with laser irradiation (SIAppendix, Fig. S12). The incorporation of sgRNA cloneddownstream in the plasmid merely affected the transfection ac-tivity of APC (SI Appendix, Fig. S13). In the meantime, we foundthat the level of luciferase expression could be modulated by laserintensity (SI Appendix, Fig. S14) and irradiation time (SI Appendix,Fig. S15), implying the transgene expression level is preciselytunable. In order to elucidate the role of specific internalizationpathways, different inhibitors were added to the cell-culture me-dium before transfection in 293T cells (SI Appendix, Figs. S16 andS17). It was evident that the addition of methyl-β-cyclodextrinsignificantly reduced GFP expression, suggesting the internalizationof APC–plasmid complexes primarily follows caveolae-dependentendocytosis. Additionally, the inhibition of transfection activityby bafilomycin A1 suggested the strong buffering capacity ofAPC, which is critical to facilitate the endosomal escape of thedelivered plasmids. APC also showed high transfection activity

Fig. 1. Illustration of the optogenetic regulation of genome editing mediated by the photoactivatable CRISPR-Cas9 nanosystem. (A) Process of preparationof the APC–HSP complex. (B) Illustration of deep-tissue penetration by NIR-II light. (C) Intracellular delivery of APC–HSP-Cas9 complexes. (D) Mechanism ofinducible optogenetic regulation of Cas9 expression and genome editing.

Chen et al. PNAS | February 4, 2020 | vol. 117 | no. 5 | 2397

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1

toward different types of cell lines upon photothermal activation (SIAppendix, Fig. S18), and did not exhibit obvious cytotoxicity on 293Tcells up to a concentration of 1.35 μg/mL (SI Appendix, Fig. S19).Based on the above optimized results, we subsequently in-

vestigated whether optogenetic control of CRISPR-Cas9 ac-tivity could be manipulated through efficient transfection andphotothermal conversion by APC (Fig. 3A). We tested whetherAPC was capable of disrupting the EGFP gene in 293T cellsthat stably expressed EGFP (SI Appendix, Fig. S20). Upon theintracellular delivery of APC–HSP-Cas9 targeting EGFP, theintensity of GFP in 293T-EGFP cells decreased significantlywith the laser irradiation, suggesting the strong ability of APCto mediate the disruption of the EGFP gene. Nevertheless, thetreatment with APC–HSP-Cas9-sgEGFP without laser irradia-tion had negligible knockout effects. We also noticed that thetemperature elevation to 42 °C by photothermal irradiationneither adversely altered the secondary structure of Cas9 protein(SI Appendix, Fig. S21A) nor affected its nuclease activities (SIAppendix, Fig. S21B). Additionally, we found Cas9 expressionbecame strongest at 42 °C (SI Appendix, Fig. S22 A and B) andthe photothermal activation at this optimal temperature induceda minimum degree of cell death (SI Appendix, Fig. S22C). Tofurther validate the genome-editing efficiency, we studied theintracellular delivery of HSP-Cas9 plasmid targeting differentgenomic loci in the 293T cell line. Indels (insertions and deletions)detected by T7 endonuclease I (T7E1) digestion assays were car-ried out to evaluate the efficiency of genome editing at the targetedgenome sites. After the transfection and photothermal activa-tion, the bands from the digestion products of T7E1 distinguishing

indels in the double-stranded DNA were clearly detected from theuncut bands at the genomic locus of adeno-associated virus inte-gration site 1 (AAVS1). We noted that the editing efficiency isslightly dependent on the APC concentration, with the highestindel rate of 20.1% at the APC/plasmid weight ratio of 1:2. Asexpected, AAVS1 genome editing by Lipo and PEI resulted inindel rates of 8.9 and 3.3%, respectively, both of which werelower than that of APC-mediated genome editing (20.1% at theoptimal weight ratio of 0.5; Fig. 3B). Sanger sequencing confirmedthe mutations at the targeted loci, including base deletion, in-sertion, and substitution around the protospacer adjacent motif(PAM) (Fig. 3C and SI Appendix, Figs. S24 and S53), and deep-sequencing analysis showed that the mutation frequency was upto 44.6% when the transfection was mediated by APC. In themeantime, we further investigated whether the level of GPF ex-pression is synchronized with the Cas9-mediated genome disrup-tion. As expected, the level of GFP expression was well-correlatedwith the indel rate, suggesting the level of GFP expression couldreflect and estimate the indel rate (Fig. 3D). Similarly, by screeningdifferent sequences of sgRNA (SI Appendix, Tables S2 and S3),the optimized genome editing at the rhomboid family member 1(RHBDF1) locus mediated by APC showed an indel rate of 14.8%,which is more efficient than that of Lipo (6.5%) (SI Appendix, Fig.S23A) and was confirmed by Sanger sequencing (SI Appendix, Fig.S24). Deep-sequencing analysis indicates that sgRNA sequencedesign is also critical in affecting genome-editing activity, and thehighest indel rate was 15% when the optimal sgRNA sequence(sg4) was used (SI Appendix, Figs. S23B and S56). Furthermore,we examined whether the optogenetic control could likewise

Fig. 2. Characterization of APC and evaluation of transfection activity. (A) Zeta-potential analysis of Au nanorods, PSS-coated AR, and APC. Mean ± SD; n = 3.(B) Absorption spectrum of AR and APC. (C) Solution temperature of APC as a function of laser irradiation time. The laser wavelength was 1,064 nm. (C, Inset)Thermal image of a PBS solution at a laser power density of 0.33 W/cm2 (Left) and APC solution at a power density of 0.33 W/cm2 (Middle) and 1.00 W/cm2

(Right) at their respective maximum temperatures. (D) HAADF-STEM and EDS mapping of APC–HSP-Cas9. (Scale bar, 25 nm.) (E) Bio-TEM image of 293T cellsafter the transfection of APC–HSP-Cas9 complexes. The arrows show the presence of APC in the cytoplasm. (F) GFP expression mediated by APC–HSP-Cas9 with(+) or without (−) laser irradiation at 1,064 nm. Lipo- and PEI-mediated transfections at 42 °C were used as positive controls, whereas cells without anytreatment were used as a negative control. (Scale bar, 200 μm.)

2398 | www.pnas.org/cgi/doi/10.1073/pnas.1912220117 Chen et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

activate the multiplex genome editing. To this end, we deliveredtwo plasmids, both of which encoded a single but different sgRNAconstruct targeting AAVS1 and Plk1 (polo-like kinase 1), re-spectively. Both genomic loci showed an evident degree ofediting, with indel rates of 28.8% (AAVS1) and 37.0% (Plk1)(Fig. 3E and SI Appendix, Fig. S25). The transfection of theHSP-Cas9 plasmid with Lipofectamine followed by irradiationwas investigated as a control (SI Appendix, Fig. S26). Indel mu-tations were hardly detected in two genomic loci (AAVS1 and Plk1;SI Appendix, Fig. S26), suggesting the poor utility of Lipofectaminefor photothermal conversion. In sharp contrast, APC-mediatedtransfection followed by irradiation induces significant mutationsat both genomic loci. Since APC can well absorb NIR-II light thatcan afford deep-tissue penetration, we covered the cell-cultureplate with breast chicken tissue of different thicknesses and investi-gated whether the irradiation could still activate the genome editingin the transfected cells in the presence of tissue (Fig. 3F). Thoughthe increase of the tissue thickness reduced the genome-editing ac-tivity, the indel rate (6.6%) could still be detected in the presence of12-mm breast chicken tissue. This suggested that the increase in thetissue thickness impaired the penetration ability of NIR-II light,thereby affecting the photothermal conversion efficiency, as we havedemonstrated (SI Appendix, Fig. S27). Furthermore, the luciferase

expression was temperature-dependent and became evident whenthe temperature reached 39 °C, reaching the highest at 42 °C (SIAppendix, Fig. S28). Interestingly, such optogenetic activation alsoworks well for dCas9-mediated transcriptional activation of ex-ogenous genes. For example, when three plasmids (HSP-dCas9-SPH, U6-sgRNA, and miniCMV-mCherry) were cotransfected in293T cells, only very low basal fluorescence was observed beforeoptogenetic activation, due to the weak ability of miniCMV toinduce transcription. In sharp contrast, mCherry expressionbecame very strong after the transcriptional activation of theheat-shock promoter, suggesting its potential for heat-inducibletranscriptional activation (SI Appendix, Fig. S29). In the meantime,we investigated whether such a heat-shock approach affected cellcycles and induced potential apoptosis (SI Appendix, Fig. S30). Asexpected, cells treated with APC–HSP-Cas9 complexes with orwithout laser irradiation showed similar cell-cycle patterns as thosetreated with phosphate-buffered saline (PBS). In the meantime,cells transfected with APC–HSP-Cas9 complexes merely inducedany apoptosis, suggesting the biocompatibility of APC and thesafety of heat-shock optogenetic modality.While successfully establishing the above strong evidence of

optogenetic genome editing, we were curious about whether sucha modality could precisely control the degree of editing. We first

Fig. 3. Optogenetic activation of CRISPR-Cas9 genome editing by APC–HSP-Cas9. (A) Illustration of optogenetic activation mediated by APC. (B) Indelmutations of the AAVS1 locus of 293T cells transfected with APC–HSP-Cas9 complexes with or without laser irradiation. APC was complexed with HSP-Cas9 atdifferent weight ratios. Lipo and PEI were used as positive controls. (C) Deep-sequencing analysis of mutation frequency at the AAVS1 locus (optimal weightratio of 0.5) and Sanger sequencing results of T–A cloning from 293T cells (AAVS1) after APC-mediated transfection, followed by optogenetic activation. Thetarget sequences are marked in red. The PAM is underlined (black). Substitutions, insertions, and deletions are marked by red base sequences, underlining(red), and dotted lines, respectively. A library of genomic DNA pooled from the sample in triplicate was subjected to deep-sequencing analysis. (D) Analysis ofindel rate, as revealed by the grayscale density of cut bands from T7E1 results, after the transfection of APC–HSP-Cas9 targeting AAVS1 with or without lasertreatment (Left). The corresponding GFP expression was evaluated; 4, 5, 6, and 7 refer to APC/HSP-Cas9 weight ratios 0.4, 0.5, 0.6, and 0.7, respectively.Mean ± SD; n = 3. (E) Indel mutation from multiplex genome editing by HSP-Cas9 targeting Plk1 and AAVS1 with laser irradiation. (F) Illustration of culturedcells exposed to irradiation in the presence of a piece of chicken breast tissue (Left). Indel mutations of the AAVS1 locus from 293T cells transfected with APC–HSP-Cas9, followed by irradiation in the presence of breast chicken tissue of different thicknesses (Right). The arrowheads in B, E, and F show the cleaved DNAfragments of the target genome.

Chen et al. PNAS | February 4, 2020 | vol. 117 | no. 5 | 2399

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1

monitored the continuous bioluminescence (BL) intensity to re-flect the amount of Cas9 expression upon optogenetic activation,and the level of luciferase expression was studied as a function oftime. As reflected by SI Appendix, Fig. S31, after transfection for24 h, the BL intensity increased quickly following irradiation for5 min, suggesting a fast gene expression by optogenetic activation.Luciferase expression further increased and reached 3.3 × 105

relative light unit (RLU) per mg after optogenetic activation for 30min. The level of luciferase expression remained stable over aperiod of 24.5 h after the removal of irradiation. However, oncethe laser was switched on again, BL intensity rapidly increased upto 4.4 × 105 RLU per mg, and became stable upon the removal ofirradiation. The above information suggested photothermal con-trol of gene expression by APC is inducible and APC may serve asan optogenetic switch to regulate Cas9 expression and activity.Then, Western blot analysis demonstrated that the Cas9 expres-sion by nanoCRISPR was inducible, and the expression of Cas9protein can be exactly controlled by time of irradiation (Fig. 4Aand SI Appendix, Fig. S32). By controlling the time length of ir-radiation from 5 to 30 min, we found the indel range could beprecisely tuned from 3.4 to 31.4% through the T7E1 assay (Fig. 4Band SI Appendix, Fig. S33). Deep sequencing further indicated thatmutation frequency ranged from 29.1 to 45.4% (Fig. 4C and SIAppendix, Fig. S54), which was also reflected by GFP expressionwith different irradiation times (SI Appendix, Fig. S34). Further-more, the temporal control of genome-editing activity could alsobe simply realized by adjusting the number of irradiationepisodes (Fig. 4D). For example, when the irradiation was

conducted for only one time length (10 min), the resulted indelrate was 11.5%; however, the indel rate could be improved simplyby increasing the number of irradiation times to reach the expectedones (Fig. 4E and SI Appendix, Fig. S35). It is worthy to mentionthat such a stepwise, optogenetic activation modality is also stable.By analyzing the indel rate at different time points after genomeediting, we found the degree of editing was generally stable, irre-spective of the irradiation times and harvesting time points (Fig. 4F–H and SI Appendix, Fig. S36). Western blot assays were carriedout to examine the time course of Cas9 expression level aftertransfection (SI Appendix, Fig. S37). In general, the expressionlevel dropped with time from 24 to 72 h following the photo-thermal induction, implying the decreased Cas9 activity followingthe on-target genome editing by nanoCRISPR.To validate the potential of such an optogenetic control

modality in vivo, A549 cells transfected with APC–HSP-Cas9complexes were subcutaneously transplanted into the back ofBALB/c mice ex vivo. Following laser irradiation on the sub-cutaneously transfected cells, BL intensity gradually becamestrong along the increased irradiation time (Fig. 5A). We nextharvested these transplanted cells and checked whether theoptogenetic activation worked for in vivo genome editing. Asshown from T7E1 assay results (Fig. 5 B and C), indel ratesranging from 4.7 to 20.1% could also be well-manipulated bytuning the irradiation time, implying the controllable activationof Cas9 expression and programmable regulation of genome-editing activity in vivo. Moreover, we explored whether directin vivo transfection and optogenetic activation are possible. To

Fig. 4. APC-mediated optogenetic control of programmable genome editing in vitro. (A) Western blot analysis of Cas9 protein expression after the transfectionfollowed by the thermal induction at different irradiation times. (B) Indel mutations of the AAVS1 locus detected by T7E1 assay. 293T cells were first transfectedwith APC–HSP-Cas9 complexes, and then exposed to the laser irradiation from 5 to 30min. The indel mutations were evaluated 72 h after the irradiation. (C) Deep-sequencing analysis of mutation frequency at the AAVS1 locus. The experimental conditions were the same as those described in Fig. 3B. (D) Illustration of thetransfection, irradiation, and genome-editing processes. The irradiations were conducted at 8, 20, and 32 h after the transfection, and irradiation time was 10 mineach time. The indel analysis was conducted at 36, 48, 56, and 72 h. (E) Indel mutations of the AAVS1 locus detected by T7E1 assay. The irradiation was conductedat 8 h (Laser+), 8 and 20 h (Laser++), and 8, 20, and 32 h (Laser+++) after the transfection. The indel mutations were analyzed at 72 h after the transfection. (F–H)Indel mutations of the AAVS1 locus detected by T7E1 assay after exposure to irradiation for different times. The irradiation was conducted at 8 h (Laser+), 8 and20 h (Laser++), and 8, 20, and 32 h (Laser+++) after the transfection, and the cells were harvested at 36, 48, 56, and 72 h for indel analysis. The arrowheads in B, E,and H show the cleaved DNA fragments of the target genome.

2400 | www.pnas.org/cgi/doi/10.1073/pnas.1912220117 Chen et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

this end, APC–HSP-Cas9 complexes were first delivered into thehind limb of BALB/c mice via intramuscular injection, andoptogenetic activation was conducted after 8 h (Fig. 5D).Strikingly, strong in vivo bioluminescence was clearly detectedin the hind limb 40 h after optogenetic activation. Based on thesefindings, we further harvested and lysed the tissue from the muscleof the hind limb to evaluate the indel mutation of the edited cells.The indel rate reached 18.1% with laser irradiation; however, theindel was hardly detectable after the in vivo transfection of APC–HSP-Cas9 complexes without laser irradiation (Fig. 5 E and F).This further motivated us to explore whether optogenetic genomeediting could be manipulated in the deep tissue of local lesions.For this purpose, BALB/c nude mice bearing A549 xenograft tu-mors were first injected with APC–HSP-Cas9 complexes through

peritumoral injection, and irradiation was then carried out in thepresence of a piece of breast chicken tissue (5-mm thickness)covering the tumor position to simulate the deep-tissue condition(Fig. 5G). The presence of tissue slightly decreased the tumortemperature to 40.0 to 41.4 °C. Excitingly, strong BL intensity wasclearly observed over the tumor position, suggesting that moderatehyperthermia could well activate gene expression. In the mean-time, a significant level of genome editing was detected from boththe surface and deep layer of the tumor tissues, with an indel rateof 16.0 and 14.9%, respectively (Fig. 5 H and I). These resultsstrongly implied that such optogenetic control may be suitablefor regulating genome-editing activity in a deep-tissue environ-ment. Given that the elevated temperature improved the Cas9activity (32), such a heat-shock approach may also improve the

Subcutaneous injectionHSP-Cas9-Luciferase

After 24 h

A B

Marker

Lase

r-APC–HSP-Cas9

0 min

5 min

10 m

in

15 m

in

25 m

in

30 m

in

Indel (%) 4.7 7.8 15.915.9 20.1

C

1.0

2.0

3.0

4.0

D E

Muscle

K

0 min 5 min 10 min 15 min 25 min 30 min

0.5

1.0

1.5

2.0

2.5

0.5

1.0

1.5

2.0

2.5

55 ℃

30 ℃

40-41.4 ℃

I

Tumor

Indel (%) 16.0 14.9

Indel (%) 18.1

Marker

Saline

AR Skin la

yer

Lase

r-

Deep l

ayer

Marker

Saline

AR APC (+)

APC (-)

H

L

0.5

1.0

1.5

2.0

2.5

2.2

2.4

2.6

2.8

Muscle

J

APC–HSP-Cas9-Luciferase

Laser + ++ _Saline AR APCAPC

Laser + +_Saline APC APC–

HSP-Cas9APC–

CMV-Cas9

APC–HSP-Cas9-Luciferase

APC–HSP-Cas9-Luciferase

Liver

APC–HSP-dCas9-mCherry

Muscle

G

F

NIR-II

Tran

sfec

tion

NIR-II

NIR-II

NIR-II

NIR-II

Laser + ++ _Saline AR APCAPC

Laser + ++ _Saline AR APCAPC

APC (+)

APC (+)

AAVS1

AAVS1

AAVS1

Chicken breast tissue

Marker

Saline

Heart

Lung

Splenn

Indel (%) 9.9

Kidney

Liver

HSP (+)

_

Radiance( 107 p/sec/cm2/sr)+

Radiance

( 105 p/sec/cm

2/sr)

+

Radiance

( 105 p/sec/cm

2/sr)

+

Radiance

( 103 p/sec/cm

2/sr)

+

2.0

Radiance

( 105 p/sec/cm

2/sr)

+

NIR II

Fig. 5. APC-mediated optogenetic control of programmable genome editing in vivo. (A) Schematic illustration (Left). Transfection with APC–HSP-Cas9complexes, and then subcutaneous implantation. Whereas the right implanted position was exposed to irradiation from 5 to 30 min, the left position(nonirradiation) was used as the control. In vivo luciferase expression (Right). (B) Indel mutations detected by T7E1 assay. (C) Quantitative analysis of indelmutations. Mean ± SD; n = 3. (D) Schematic illustration (Left). APC–HSP-Cas9 complexes were subcutaneously injected into the muscle of the hind limb ofBALB/c mice, followed by irradiation at 1,064 nm. In vivo luciferase expression (Right). (E) Indel mutations detected by T7E1 assay. (F) Quantitative analysis ofindel mutations. Mean ± SD; n = 3. (G) Schematic illustration (Left). Tumor-bearing mice were administered APC–HSP-Cas9 complexes through peritumoralinjection, and the tumor (Right) was then exposed to irradiation for 30 min. In vivo luciferase expression (Middle). (H) Indel mutations detected by T7E1 assay.(I) Quantitative analysis of indel mutations. Mean ± SD; n = 3. (J) Evaluation of in vivo luciferase expression through intravenous injection, followed byoptogenetic activation in the liver. Schematic illustration (Left). BALB/c mice were injected with APC–HSP-Cas9 complexes via tail-vein injection, and the liverposition was exposed to irradiation for 30 min after the transfection. (K) Indel mutations from different tissues after the systemic administration of nanoCRISPRthrough intravenous injection, followed by injection with APC–HSP-Cas9 complexes via tail-vein injection, and the liver position was exposed to the irradiation for30 min after the transfection. (L) Optogenetic regulation of transcriptional activation of mCherry expression through in vivo transfection and optogenetic ac-tivation. The arrowheads in B, E, H, and K show the cleaved DNA fragments of the target genome.

Chen et al. PNAS | February 4, 2020 | vol. 117 | no. 5 | 2401

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1

editing capacity of the photothermal nanoCRISPR. We alsofound that the spatial specificity could be well-manipulated aswell through optogenetic regulation. The systemic administrationof APC–HSP-Cas9 to BALB/c mice by tail-vein injection resultedin strong luciferase expression in the liver that was exposed toirradiation, whereas the same treatment of mice with APC–CMV-Cas9 resulted in luciferase expression primarily in the lung (Fig. 5Jand SI Appendix, Fig. S38). Indel mutation (9.9%) was clearlydetected in the liver when the transfection was mediated by pho-tothermal nanoCRISPR (Fig. 5K), and such mutations were hardlyfound in any other organs including heart, spleen, lung, and kidney.As a control, indel mutations were also barely detected in the liverwhen the transfection of photothermal nanoCRISPR was con-ducted in the absence of NIR-II irradiation (SI Appendix, Fig.S39A). In sharp contrast, indel mutations were found in liver(3.8%), spleen (1.7%), and lung (0.7%) when the transfection wasmediated by APC–CMV-Cas9 (SI Appendix, Fig. S39B). In agree-ment with the results from transcriptional activation of exogenousgenes in vitro, the transcriptional activation of mCherry expressionwas also verified in vivo, when APC–plasmid complexes werecodelivered into the hind limb either through ex vivo transfectionor direct in vivo tissue transfection, followed by optogenetic acti-vation (Fig. 5L and SI Appendix, Figs. S40 and S41). Importantly,we demonstrated that the optogenetic activation through differentadministration approaches, including ex vivo transfection, directin vivo intramuscular administration, and systemic administration,merely induced toxicity in the major organs (heart, liver, spleen,lung, and kidney) after the irradiation (SI Appendix, Figs. S42–S44).The above results demonstrated that spatial and programmablegenome editing could also be safely achievable in vivo as well.We further investigated the optogenetic activation of photo-

thermal nanoCRISPR in vivo as proof-of-concept examples fortherapeutic genome editing. To this end, we first delivered Cas9plasmid with sgRNA targeting Plk1, a master regulator of mitosis(33), and activated expression after transfection in A549 cells.Indel analysis indicated that significant mutation was detected inthe targeted genomic locus, with an indel rate up to 41.5% whenthe optimized sgRNA targeting Plk1 was used (SI Appendix, Fig.S45A). The editing-induced mutation was also confirmed bySanger sequencing results, where significant deletions and inser-tions were detected at the targeted loci around the PAM (SI Ap-pendix, Fig. S24). Deep-sequencing analysis indicated that themutation frequency was up to 51.4% (SI Appendix, Figs. S45B andS55). Western blot analysis indicated that the level of Plk1 ex-pression was remarkably reduced after the transfection and acti-vation process (SI Appendix, Fig. S45C). These results in vitro wellestablish the fact that optogenetically regulated genome editingenables the efficient knockout of the target Plk1 gene. Based onthese results, we next investigated whether tumor growth could beeffectively inhibited on BALB/c nude mice bearing A549 xenografttumors by this therapeutic modality (Fig. 6A). After the peritu-moral injection of APC–HSP-Cas9 complexes and incubationfor 48 h, we noticed that the BL was still visible but becameweak after irradiation activation (Fig. 6B). In the meantime,T7E1 assay results suggested significant genome disruption inthe Plk1 site in the tumor tissue (Fig. 6B and SI Appendix, Fig.S47), and deep-sequencing analysis of a single library preparedfrom genomic DNA pooled from five mice also indicated sig-nificant mutation in the Plk1 locus (17.2%) (Fig. 6C and SIAppendix, Fig. S57). Therefore, the reduced expression wasprobably attributable to the presence of large amounts of ap-optotic cells induced by Plk1 disruption, leading to the poorability to express luciferase. In fact, the above speculation wasverified by the in vivo tumor-inhibition assay, where the tumor-bearing mice injected with APC–HSP-Cas9 targeting Plk1exhibited significant tumor regression after irradiation treatment.The tumor size also became much smaller in comparison with theinitial size before treatment (43.9 mm3 on average). In sharp

contrast, mice treated with the same formulation but withoutlaser treatment exhibited rapid tumor progression, reaching afinal tumor volume of 1,270 mm3 at 21 d (Fig. 6 D and E). As anindicator of systemic toxicity, we monitored body weight through atherapy session and noticed that a slight increase in body weightwas observed at the end of the treatment (20.6 g on average; Fig.6F). In the meantime, toxicity to major organs was investigated byhematoxylin and eosin (H&E) staining (SI Appendix, Fig. S46),and blood biochemistry was also evaluated to reflect liver andkidney index (SI Appendix, Fig. S48). As compared with the salinecontrol group (without laser treatment), the tumor slice showedthe fewest tumor cells combined with a significant degree ofnecrosis. Whereas H&E staining suggested that such a thera-peutic modality was generally safe and biocompatible in themajor organs, the function index of blood biochemistry furthervalidated that such an optogenetic treatment merely caused anydamage to the liver and kidney.Last, we explored whether the optogenetic activation of

photothermal nanoCRISPR could protect mice from fulminanthepatic failure. A wide range of liver diseases are associatedwith Fas-mediated apoptosis, and the inhibition of Fas geneexpression by RNA interference was shown to protect micefrom fulminant hepatic failure (34). Therefore, we speculatethat Cas9-mediated genome editing of Fas may protect fulmi-nant hepatic failure. Additionally, by controlling the irradiationof NIR-II light over the liver, we were able to improve thespatial specificity of genome editing exclusively in the liver.First, we found the optogenetic activation of photothermal nano-CRISPR with the HSP-Cas9 plasmid encoding sgRNA targetingFas (HSP-Cas9–Fas) induced an 18.0% indel mutation rate inHepa1-6 cells (SI Appendix, Fig. S49A), which was confirmedby Sanger sequencing (SI Appendix, Fig. S49B). Since galactose-decorated nanoparticles have been widely explored to improveliver-targeting ability by interacting with asialoglycoprotein recep-tors on the membrane of hepatocytes (35), we further deliveredgalactose-modified APC–HSP-Cas9–Fas complexes by tail-veininjection and examined whether such a prophylactic measurecould protect mice from concanavalin A (Con A)-induced ful-minant hepatic failure (Fig. 6G). The mice treated withgalactose-modified APC–HSP-Cas9–Fas significantly reducedhyperemia (Fig. 6 H and I), in contrast with the ones without ir-radiation (APC [−]) or treated with scramble sgRNA (APC/NT[+]). In the meantime, blood biochemistry indicated the opto-genetic activation of nanoCRISPR in the liver could effectivelyreduce AST (aspartate transaminase) and ALT (alanine amino-transferase) levels (303 and 217 U/L, respectively), which wereclose to that of mice without any treatment (Fig. 6 J and K). Thein vivo editing was validated by the T7E1 and Sanger sequencingresults, where significant genome disruption in the Fas site (6.7%indel mutation) was found in the liver tissue (Fig. 6L). Hemato-logical indicators were also evaluated after the systemic adminis-tration of APC to reflect the liver and kidney index in vivo (SIAppendix, Fig. S50). We monitored the mice for up to 14 d, andfound the systemic administration of APC at the therapeuticdose (12.5 μg/mL) merely induced any liver and kidney toxicityduring the observation period.To analyze the off-target effects generated by this editing

modality, we used an off-target searching tool, Cas-OFFinder(36), to estimate the potential off-target sites, and further carriedout both Sanger sequencing and deep sequencing to evaluatewhether genomic mutations could be detected in the estimatedoff-target sites (SI Appendix, Tables S4 and S5). First, Sangersequencing analysis proved that the sequence that was suspected tooff-target disruption displayed the same intact sequence as thewild-type one without any treatment (SI Appendix, Fig. S51). Wefurther carried out deep sequencing to quantitatively analyze theoff-target mutation at different time points following photothermaltransfection of nanoCRISPR. APC-mediated transfection of Cas9

2402 | www.pnas.org/cgi/doi/10.1073/pnas.1912220117 Chen et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

Tumor

NIR-II2.8

3.0

3.2

3.4

3.6

Indel (%) 28.3

Marker

Saline

APC (-)A B C

D

AR APC (+)

E FSaline (-)

Saline (+)

AR (+)

APC (-)

APC (+)

Plk1

G H I

J K LMark

er

Saline

APC/NT (+

)

APC (-)

APC (+)

Contro

l

Contro

l (+)

Indel (%) 6.7

Control Control (+) Saline

APC (-) APC/NT (+) APC (+)

Control Control (+) Saline

APC (-) APC/NT (+) APC (+)

Treatment of deep tumor

Rescue of fulminant hepatic failure

APC–HSP-Cas9-Luciferase

NIR-II Galactose-modified APC Systemic

adminstration

Liver

0 50 100

No Mutation Mutation

Radiance( 103 p/sec/cm2/sr)+

Fig. 6. APC-mediated optogenetic activation for therapeutic genome editing in vivo. (A) Illustration of optogenetic activation for in vivo cancer therapy.APC–HSP-Cas9 complexes were administered through peritumoral injection, followed by the exposure of the tumor to irradiation. (B) In vivo luciferaseexpression in the tumor tissue. The left tumor was exposed to laser irradiation, and the right one, without laser exposure, was used as the control. T7E1 assayof indel mutations of Plk1 in the tumor tissue. (C) Analysis of in vivo Plk1 gene mutation efficiency through deep sequencing and Sanger sequencing. Themutation frequency was determined from a single deep-sequencing library prepared from genomic DNA pooled from five mice. (D) Tumor growth curve afterthe transfection of APC–HSP-Cas9 complexes in the tumor tissue, followed by exposure to irradiation. The treatment was carried out twice a week, andcontinued for 3 wk. Mean ± SD; n = 5 (two-way ANOVA with a Bonferroni post hoc test, ****P < 0.0001). (E) Images of dissected tumor tissues from tumor-bearing BALB/c mice with different treatments. (F) The body-weight change during the treatment. (G) Illustration of optogenetic activation of nanoCRISPRin vivo to rescue mice from fulminant hepatic failure. Mice were injected intravenous through the tail vein with APC–HSP-Cas9–Fas complexes, followed byirradiation over the liver position. (H) Images of dissected livers from mice with fulminant hepatic failure. (I) H&E staining of liver slices from mice 10 d afterthe treatment. The regions in the dotted lines denote accumulation of blood cells. (Scale bar, 50 μm.) (J and K) Serum AST and ALT from a mouse injected withsaline, APC–HSP-Cas9-NT with irradiation (NT, nontargeted scramble sgRNA), and APC–HSP-Cas9–Fas (with or without irradiation) 10 d after treatment.Mean ± SD; n = 4 (one-way ANOVA with a Tukey post hoc test, ****P < 0.0001). (L) Indel mutation analysis of the Fas gene by T7E1 assay and mutation Sangersequencing after the optogenetic activation of nanoCRISPR in the liver of mice with fulminant hepatic failure. The arrowheads in B and L show the cleavedDNA fragments of the target genome.

Chen et al. PNAS | February 4, 2020 | vol. 117 | no. 5 | 2403

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1

plasmid driven by the cytomegalovirus (CMV) promoter, aconstitutive promoter, was used as a control. As revealed bydeep-sequencing analysis, the transfection of APC–CMV-Cas9complexes only resulted in moderate gene disruption at the on-target site of AAVS1 2 d after the transfection, with an indelmutation frequency of 16.6%. The degree of on-target muta-tion increased a further 4 d following the transfection, reachingan indel mutation frequency of 27.6%. After 6 d of transfection,the on-target mutation rate decreased to 15.2% (SI Appendix,Fig. S52A and Table S6). In comparison with CMV-Cas9, themutation frequency induced by photothermal transfection of HSP-Cas9 was at least two times higher 2 d after the transfection,suggesting high on-target genome-editing activity. Note that on-target mutation frequency dropped significantly 4 d following thephotothermal transfection of HSP-Cas9, and further decreased to20.0% 6 d after the transfection. The decreased on-target muta-tion might be partially due to the delayed cell division in themutated daughter population, where 53BP1 nuclear bodiesfunction as a key regulator to restrain the replication of dis-rupted genomic loci until late S phase (37). In addition, de-tectable off-target mutations were found throughout the wholeexperimental window, which were more frequent and signifi-cant over the photothermal transfection of HSP-Cas9 in general,though the latter also resulted in traceable off-target editing. Themean specificity ratio [defined as the ratio of on-target activityto off-target activity (38)] of HSP-Cas9–induced genome edit-ing was remarkably higher than that of CMV-Cas9–inducedediting during this time window (SI Appendix, Fig. S52B). Theseresults collectively demonstrated that optogenetically activatablenanoCRISPR may minimize off-target effects through optogeneticcontrol of Cas9 expression to reduce prolonged Cas9 activity.

DiscussionRemote activation with noninvasive NIR light has been extensivelyexploited in a wide range of biomedical applications, such asmicroRNA detection (39), brain stimulation (40), modulation ofgene expression (31), and immunomodulation (41), largely owingto the low photocytotoxicity and deep-tissue penetration capabilityof NIR light. In comparison with NIR-I light, NIR-II light hasbeen validated to afford deeper tissue penetration, a key challengepreventing many optogenetic control strategies from in vivo in-vestigation. Although a few types of organic (42, 43) and inorganicnanomaterials (28) that absorb NIR-II light have been devel-oped, AR was selected as a building block for APC largely due toits high photothermal conversion rate and photothermal stability(44). Furthermore, the AR not only converts the external photonicenergy into intracellular local heat but also serves as a templatewhere the cationic polymers are assembled for the subsequentencapsulation of large plasmids. As an unconventional finding, theassembly of PC over AR surprisingly resulted in far more effi-cient condensation of the Cas9 plasmid in comparison with PCalone (30), probably owing to the higher aspect ratio of goldnanostructure being more favorable for entangling the plasmid.This feature of AR may also facilitate APC to enter the nucleus bypassive fusion, as high-aspect ratio nanoparticles, such as nanorodsand nanoworms, were previously demonstrated to be superior tospherical ones with identical surface chemistries in terms of nu-clear entry (45). It is also noteworthy that the incorporation of theCas9 plasmid and photothermal transducer into the same carrierensures the delivery of two payloads into the same cell population,thereby maximizing the sensitivity and efficiency of optogeneticactivation of the Cas9 transcription. As opposed to other light-activated editing or transcriptional regulation systems that requirethe cotransfection of multiple plasmids by commercial transfectionagents (16, 22, 23), nanoCRISPR is a relatively straightforwardand simple system that only contains an “all-in-one” plasmid and aphotothermal nanocarrier, both of which are easy to construct.The efficient transfection and photothermal conversion enabled

by APC ensure the successful induction of Cas9 expression as wellas the precise control of Cas9 nuclease activity.Heat-induced transcription of genes encoding a major heat-

shock protein (HSP70) is a cytoprotective mechanism which awide variety of cells exploit to protect themselves from heat shockand other deleterious stresses (46). HSP70 promoters are regu-lated by cytosolic HSF, which becomes active in response tomoderate hyperthermia (41.5 to 42.0 °C) to induce the expressionof downstream heat-shock proteins that are critical for cellulardefense (47). The heat-responsive HSP70 promoters have beenpreviously explored for the spatial and temporal control of geneexpression through photothermal effects (48, 49). In the currentstudy, we constructed a Cas9-encoding plasmid driven by an HSP70promoter, which indeed serves as a photothermal switch to regulateCas9 transcription by sensing the surrounding temperature. Asthe temperature could be finely tuned by controlling the irradi-ation time length and is closely correlated with Cas9 expressionand activity, we are therefore capable of programming the de-gree of editing simply by adjusting exposure time and irradiationtime. Hence, this editing modality may be applicable in the con-text where CRISPR-Cas9 activity is required to fulfill editing mis-sions at multiple time points. For example, the optogeneticallyactivatable CRISPR-Cas9 nanosystem may serve as an ideal plat-form for inducible editing at multiple time points that is requiredfor CRISPR-Cas9 barcode editing to trace lineage information ofdifferent cells during development and disease (50). For manyother applications, CRISPR-Cas9 activity should be inhibited fol-lowing the on-target editing, and prolonged activity may otherwisecause undesired side effects. For instance, restriction of Cas9 ac-tivity to a narrow temporal window is critical in germ-line editing, asthe persistent Cas9 activity following the initial rounds of mitosiscontributes to mosaicism (51). Furthermore, nanoCRISPR mayserve as an inducible genome-editing platform to precisely controlthe dose of Cas9, as a number of side effects, such as genotoxicity,immunogenic response, and chromosomal translocations, are as-sociated with elevated levels of Cas9. Collectively, our system alsoprovides a robust method to diminish Cas9 activity after certainediting events simply by switching off the light.Spatial specificity of CRISPR-Cas9 is essential for many po-

tential therapeutic purposes in that Cas9 activity in ancillary tissuesmay give rise to safety risks. As a proof-of-concept study, wedemonstrated that the spatial control of CRISPR-Cas9 activity canwell be manipulated in the liver exclusively through optogeneticactivation after the systematic administration. Thanks to the NIR-II–absorbing feature of APC, spatial optogenetic control is vali-dated to be realizable in deep tissue, which opens an avenue forbroader in vivo investigations. In addition, the current findings alsosuggest that precise control of Cas9 activity by light is important todiminish off-target effects and other potential genotoxicities (52).In addition to 53BP1-mediated delayed cell division, the decreasedoff-target effects may also be attributed to the shortened exposuretime of the genome to Cas9/sgRNA through optogenetic control,which minimized the tolerable mismatches between sgRNA andgenomic loci bearing similar sequences (53). Our future efforts willbe dedicated to the intensive investigation of off-target effects atthe whole-genome level, in order to understand the safe use of thisgenome-editing modality.

Materials and MethodsExperimental details and methods can be found in SI Appendix, including thesynthesis and characterizations of the APC, heat-inducible Cas9/dCas9 plasmidconstruction, target design and sgRNA plasmid construction, in vitro photo-thermal studies, in vitro and in vivo APC-mediated optogenetic control ofprogrammable genome editing, T7E1 assay and deep-sequencing analysis, off-target analysis, temperature control by irradiation, in vivo optogenetic genomeediting in varying animal models, hematological and histological analysis, andstatistical analysis. All animal treatments and procedures were approved by theLaboratory Animal Welfare and Ethics Committee of Zhejiang University.

2404 | www.pnas.org/cgi/doi/10.1073/pnas.1912220117 Chen et al.

Dow

nloa

ded

by g

uest

on

May

14,

202

1

Data Availability Statement. The authors declare that all data supporting thefindings of this study are available within the paper and SI Appendix. Thedeep-sequencing data generated in this paper are available in the NCBISequence Read Archive (Bioproject ID PRJNA599254).

ACKNOWLEDGMENTS. We thank National Key Research and Develop-ment Program of China (2018YFA0901800), National Natural Science Foun-dation of China (81872807), Fundamental Research Funds for the Central

Universities (2018XZZX001-14), and Thousand Talents Plan (Y.P.) for thefinancial support of this work. We also acknowledge Dr. Di Wu for helpfuldiscussions on the synthesis of gold nanorods, and Shuaishuai Zhang forhelp with the construction of plasmids. We also appreciate Prof. Xing ChangfromWestlake University for the helpful discussions of deep-sequencing analysis,Prof. Xue Gao from Rice University for the useful discussions of off-targetanalysis, and Prof. Yunxian Yu from Zhejiang University for the helpful adviceon statistical analysis.

1. M. Jinek et al., A programmable dual-RNA-guided DNA endonuclease in adaptivebacterial immunity. Science 337, 816–821 (2012).

2. D. B. Cox, R. J. Platt, F. Zhang, Therapeutic genome editing: Prospects and challenges.Nat. Med. 21, 121–131 (2015).

3. T. Wan et al., Material solutions for delivery of CRISPR/Cas-based genome editingtools: Current status and future outlook. Mater. Today 26, 40–66 (2019).

4. H. Nishimasu et al., Crystal structure of Cas9 in complex with guide RNA and targetDNA. Cell 156, 935–949 (2014).

5. H. X. Wang et al., CRISPR/Cas9-based genome editing for disease modeling andtherapy: Challenges and opportunities for nonviral delivery. Chem. Rev. 117, 9874–9906 (2017).

6. L. Yang et al., Genome-wide inactivation of porcine endogenous retroviruses (PERVs).Science 350, 1101–1104 (2015).

7. Y. L. Min, R. Bassel-Duby, E. N. Olson, CRISPR correction of Duchenne muscular dys-trophy. Annu. Rev. Med. 70, 239–255 (2019).

8. O. Shalem, N. E. Sanjana, F. Zhang, High-throughput functional genomics usingCRISPR-Cas9. Nat. Rev. Genet. 16, 299–311 (2015).

9. R. T. Manguso et al., In vivo CRISPR screening identifies Ptpn2 as a cancer immuno-therapy target. Nature 547, 413–418 (2017).

10. J. A. Soppe, R. J. Lebbink, Antiviral goes viral: Harnessing CRISPR/Cas9 to combat vi-ruses in humans. Trends Microbiol. 25, 833–850 (2017).

11. K. I. Liu et al., A chemical-inducible CRISPR-Cas9 system for rapid control of genomeediting. Nat. Chem. Biol. 12, 980–987 (2016).

12. B. J. Aubrey et al., An inducible lentiviral guide RNA platform enables the identifi-cation of tumor-essential genes and tumor-promoting mutations in vivo. Cell Rep. 10,1422–1432 (2015).

13. L. E. Dow et al., Inducible in vivo genome editing with CRISPR-Cas9. Nat. Biotechnol.33, 390–394 (2015).

14. D. P. Nguyen et al., Ligand-binding domains of nuclear receptors facilitate tightcontrol of split CRISPR activity. Nat. Commun. 7, 12009 (2016).

15. J. Hemphill, E. K. Borchardt, K. Brown, A. Asokan, A. Deiters, Optical control ofCRISPR/Cas9 gene editing. J. Am. Chem. Soc. 137, 5642–5645 (2015).

16. Y. Nihongaki, F. Kawano, T. Nakajima, M. Sato, Photoactivatable CRISPR-Cas9 foroptogenetic genome editing. Nat. Biotechnol. 33, 755–760 (2015).

17. B. Zetsche, S. E. Volz, F. Zhang, A split-Cas9 architecture for inducible genome editingand transcription modulation. Nat. Biotechnol. 33, 139–142 (2015).

18. B. Maji et al., Multidimensional chemical control of CRISPR-Cas9. Nat. Chem. Biol. 13,9–11 (2017).

19. K. M. Davis, V. Pattanayak, D. B. Thompson, J. A. Zuris, D. R. Liu, Small molecule-triggered Cas9 protein with improved genome-editing specificity. Nat. Chem. Biol. 11,316–318 (2015).

20. P. K. Jain et al., Development of light-activated CRISPR using guide RNAs with pho-tocleavable protectors. Angew. Chem. Int. Ed. Engl. 55, 12440–12444 (2016).

21. L. R. Polstein, C. A. Gersbach, A light-inducible CRISPR-Cas9 system for control ofendogenous gene activation. Nat. Chem. Biol. 11, 198–200 (2015).

22. F. Bubeck et al., Engineered anti-CRISPR proteins for optogenetic control of CRISPR-Cas9. Nat. Methods 15, 924–927 (2018).

23. J. Shao et al., Synthetic far-red light-mediated CRISPR-dCas9 device for inducingfunctional neuronal differentiation. Proc. Natl. Acad. Sci. U.S.A. 115, E6722–E6730(2018).

24. C. Ash, M. Dubec, K. Donne, T. Bashford, Effect of wavelength and beam width onpenetration in light-tissue interaction using computational methods. Lasers Med. Sci.32, 1909–1918 (2017).

25. Y. Pan et al., Near-infrared upconversion–activated CRISPR-Cas9 system: A remote-controlled gene editing platform. Sci. Adv. 5, eaav7199 (2019).

26. D. D. Mosser, J. Duchaine, B. Massie, The DNA-binding activity of the human heatshock transcription factor is regulated in vivo by hsp70. Mol. Cell. Biol. 13, 5427–5438(1993).

27. K. Abravaya, M. P. Myers, S. P. Murphy, R. I. Morimoto, The human heat shock proteinhsp70 interacts with HSF, the transcription factor that regulates heat shock geneexpression. Genes Dev. 6, 1153–1164 (1992).

28. X. Li et al., In vitro and in vivo photothermal cancer therapeutic effects of goldnanorods modified with mushroom β-glucan. J. Agric. Food Chem. 66, 4091–4098(2018).

29. Y. S. Chen, Y. Zhao, S. J. Yoon, S. S. Gambhir, S. Emelianov, Miniature gold nanorodsfor photoacoustic molecular imaging in the second near-infrared optical window.Nat. Nanotechnol. 14, 465–472 (2019).

30. Z. Zhang et al., Cationic polymer-mediated CRISPR/Cas9 plasmid delivery for genomeediting. Macromol. Rapid Commun. 40, e1800068 (2019).

31. H. A. Andersson, Y. S. Kim, B. E. O’Neill, Z. Z. Shi, R. E. Serda, HSP70 promoter-drivenactivation of gene expression for immunotherapy using gold nanorods and near in-frared light. Vaccines (Basel) 2, 216–227 (2014).

32. G. Xiang, X. Zhang, C. An, C. Cheng, H. Wang, Temperature effect on CRISPR-Cas9mediated genome editing. J. Genet. Genomics 44, 199–205 (2017).

33. P. Wang et al., Thermo-triggered release of CRISPR-Cas9 system by lipid-encapsulatedgold nanoparticles for tumor therapy. Angew. Chem. Int. Ed. Engl. 57, 1491–1496(2018).

34. E. Song et al., RNA interference targeting Fas protects mice from fulminant hepatitis.Nat. Med. 9, 347–351 (2003).

35. K. Lee et al., In vivo delivery of transcription factors with multifunctional oligonu-cleotides. Nat. Mater. 14, 701–706 (2015).

36. S. Bae, J. Park, J. S. Kim, Cas-OFFinder: A fast and versatile algorithm that searches forpotential off-target sites of Cas9 RNA-guided endonucleases. Bioinformatics 30,1473–1475 (2014).

37. J. Spies et al., 53BP1 nuclear bodies enforce replication timing at under-replicatedDNA to limit heritable DNA damage. Nat. Cell Biol. 21, 487–497 (2019).

38. A. Hendel, E. J. Fine, G. Bao, M. H. Porteus, Quantifying on- and off-target genomeediting. Trends Biotechnol. 33, 132–140 (2015).

39. W. Ma et al., Dual quantification of microRNAs and telomerase in living cells. J. Am.Chem. Soc. 139, 11752–11759 (2017).

40. S. Chen et al., Near-infrared deep brain stimulation via upconversion nanoparticle-mediated optogenetics. Science 359, 679–684 (2018).

41. P. Tan, L. He, G. Han, Y. Zhou, Optogenetic immunomodulation: Shedding light onantitumor immunity. Trends Biotechnol. 35, 215–226 (2017).

42. Y. Jiang et al., Metabolizable semiconducting polymer nanoparticles for second near-infrared photoacoustic imaging. Adv. Mater. 31, e1808166 (2019).

43. S. Zhu, R. Tian, A. L. Antaris, X. Chen, H. Dai, Near-infrared-II molecular dyes for cancerimaging and surgery. Adv. Mater. 31, e1900321 (2019).

44. J. Conde, N. Oliva, Y. Zhang, N. Artzi, Local triple-combination therapy results in tu-mour regression and prevents recurrence in a colon cancer model. Nat. Mater. 15,1128–1138 (2016).

45. E. Hinde et al., Pair correlation microscopy reveals the role of nanoparticle shape inintracellular transport and site of drug release. Nat. Nanotechnol. 12, 81–89 (2017).

46. R. I. Morimoto, Cells in stress: Transcriptional activation of heat shock genes. Science259, 1409–1410 (1993).

47. K. Abravaya, B. Phillips, R. I. Morimoto, Attenuation of the heat shock response inHeLa cells is mediated by the release of bound heat shock transcription factor and ismodulated by changes in growth and in heat shock temperatures. Genes Dev. 5,2117–2127 (1991).

48. Y. Lyu et al., Dendronized semiconducting polymer as photothermal nanocarrier forremote activation of gene expression. Angew. Chem. Int. Ed. Engl. 56, 9155–9159(2017).

49. E. Miyako et al., Photothermic regulation of gene expression triggered by laser-in-duced carbon nanohorns. Proc. Natl. Acad. Sci. U.S.A. 109, 7523–7528 (2012).

50. B. Raj et al., Simultaneous single-cell profiling of lineages and cell types in the ver-tebrate brain. Nat. Biotechnol. 36, 442–450 (2018).

51. S. T. Yen et al., Somatic mosaicism and allele complexity induced by CRISPR/Cas9 RNAinjections in mouse zygotes. Dev. Biol. 393, 3–9 (2014).

52. S. A. Gangopadhyay et al., Precision control of CRISPR/Cas9 using small molecules andlight. Biochemistry 58, 234–244 (2019).

53. J. Cao et al., An easy and efficient inducible CRISPR/Cas9 platform with improvedspecificity for multiple gene targeting. Nucleic Acids Res. 44, e149 (2016).

Chen et al. PNAS | February 4, 2020 | vol. 117 | no. 5 | 2405

APP

LIED

BIOLO

GICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

May

14,

202

1