Genomic Targeting of Epigenetic Probes Using a Chemically Tailored Cas9 System

Genomic targeting of epigenetic probes using a chemically tailored Cas9 system

Glen P. Liszczaka, Zachary Z. Browna, Samuel H. Kima, Rob C. Oslunda, Yael Davida, and Tom W. Muira,1

aDepartment of Chemistry, Princeton University, Princeton, NJ 08544

Edited by James A. Wells, University of California, San Francisco, CA, and approved December 13, 2016 (received for review September 20, 2016)

Recent advances in the field of programmable DNA-binding proteins Here, we report a method that combines the versatility of have led to the development of facile methods for genomic pharmacologic manipulation with the specificity of a genetically localization of genetically encodable entities. Despite the extensive programmable DNA-binding protein. Our strategy uses a utility of these tools, locus-specific delivery of synthetic molecules chemically tailored dCas9 to display a pharmacologic agent at a remains limited by a lack of adequate technologies. Here we combine genetic locus of interest in live mammalian cells (Fig. 1). This is trans the flexibility of chemical synthesis with the specificity of a pro- accomplished using split intein-mediated protein -splicing grammable DNA-binding protein by using protein trans-splicing to (PTS) to site-specifically link a recombinant dCas9:guide RNA ligate synthetic elements to a nuclease-deficient Cas9 (dCas9) (gRNA) complex to the synthetic cargo of choice (22). Indeed, in vitro and subsequently deliver the dCas9 cargo to live cells. we show that the remarkable specificity, efficiency, and speed of PTS allow the direct generation of desired dCas9 conjugates The versatility of this technology is demonstrated by delivering within the cell culture media, thereby facilitating a streamlined dCas9 fusions that include either the small-molecule bromodomain “one-pot” approach for genomic targeting of the reaction product. and extra-terminal family bromodomain inhibitor JQ1 or a pep- We used this strategy to direct the epigenetic probes JQ1 (a tide-based PRC1 chromodomain ligand, which are capable of small molecule) and UNC3866 (a modified peptide) to spe- recruiting endogenous copies of their cognate binding partners cific genomic loci, where they interact with their target his- to targeted genomic binding sites. We expect that this technology tone recognition domains, the bromodomain and extra-terminal will allow for the genomic localization of a wide array of small (BET) family bromodomains and Polycomb repressive complex molecules and modified proteinaceous materials. 1 (PRC1) chromodomains, respectively. Therefore, we offer a strategy to direct endogenous epigenetic proteins and macro- chemical biology | epigenome engineering | protein delivery | protein molecular complexes to specific loci in the absence of genetic semisynthesis | intein splicing manipulation or overexpression of exogenous protein. Further- more, the highly modular nature of this approach makes it easily ll eukaryotic genomes are exquisitely packaged into the adaptable to a wide variety of synthetic entities. Anucleus as chromatin, wherein DNA is wrapped around Results histone proteins. To achieve the transcriptional output associ- Loci-Specific Delivery of Modified dCas9. Several bioconjugation ated with a particular cell type, chromatin is dynamically deco- techniques, including maleimide chemistry and HaloTag label- rated with a variety of covalent modifications and effector ing, have been used previously to conjugate synthetic moieties, molecules at specific genetic loci (1, 2). The resulting epigenomic such as biotin and fluorophores, to Cas9 (23, 24). PTS offers landscape establishes proper DNA accessibility and genomic struc- several advantages over these methods that make it ideal for our ture (3, 4), and its misregulation can serve as an underlying cause of approach; importantly, reactions can be performed in cell media, genetic disorders, including cancer (5). Thus, deliberate perturba- are complete within minutes, and do not require downstream tion of chromatin effector function has become a cornerstone for purification, thus allowing us to directly deliver the reaction dissecting the molecular mechanisms of transcription regulation pathways in live cells. Significance The activity of chromatin regulatory proteins can be manipu- lated using both genetic and pharmacologic approaches (6, 7). Programmable DNA-binding proteins such as nuclease-deficient The former includes technologies that alter the total cellular Cas9 (dCas9) offer a simple system for genomic localization of levels of a target transcript through transgene overexpression, genetically encodable biomolecules. These tools have enabled knockdown, or knockout, whereas the latter encompasses the use characterization of numerous different chromatin effectors at of small-molecule chemical probes that can modulate the func- specific genetic elements. The delivery of fully synthetic cargos tional output of a protein of interest. Although these techniques to specific loci is currently limited by a lack of adequate tech- are capable of globally affecting protein activities, locus-specific nologies. Here we used protein trans-splicing to ligate chemical epigenomic manipulation has been demonstrated to be a more moieties to dCas9 in vitro and subsequently deliver these mol- effective method for probing downstream transcriptional outputs ecules to live cells. We demonstrate the effectiveness and ver- – (8 10). Targeted epigenome engineering is accomplished by fusing satility of this method by delivering both small-molecule and genetically encodable chromatin effector proteins to programmable proteinaceous epigenetic probes to endogenous genomic sites DNA-binding proteins, such as transcription activator-like effectors where they interact with their target proteins. This facile and (11), zinc fingers (12), and nuclease-deficient Cas9 (dCas9) modular strategy will find broad use in locus-specific targeting – (13 16). A complementary approach to the genetic fusion strategy of chemical cargos to study nuclear processes. is genomic localization of purely synthetic entities. This opens the door for subnuclear targeting of epigenetic probes, including Author contributions: G.P.L. and T.W.M. designed research; G.P.L., Z.Z.B., S.H.K., R.C.O., small molecules and modified proteinaceous material (17, 18). and Y.D. performed research; G.P.L. and T.W.M. analyzed data; and G.P.L. and T.W.M. Examples of synthetic targeting agents include pyrrole-imidazole wrote the paper. polyamides and peptide nucleic acids (19, 20). Despite the ele- The authors declare no conflict of interest. gance of these strategies, however, their widespread application This article is a PNAS Direct Submission. has been limited by the need for specialized chemical synthesis 1To whom correspondence should be addressed. Email: [email protected].

methods and by constraints associated with their genomic tar- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. BIOCHEMISTRY geting (i.e., locus specificity) (19, 21). 1073/pnas.1615723114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1615723114 PNAS | January 24, 2017 | vol. 114 | no. 4 | 681–686 Downloaded by guest on September 30, 2021 purification, i.e., a one-pot process. We found that the splicing reaction was >90% complete within 20 min, at which point we used a previously reported cationic lipid-mediated delivery strategy (25) to introduce the modified dCas9-gRNA complexes at an optimal concentration of 100 nM into 293T cells (SI Ap- pendix, Fig. S3). Delivery of the dCas9-VP64:gRNA complexes into cells led to transcriptional activation of the gRNA target gene, as measured by reverse-transcription quantitative PCR (RT-qPCR) analysis (Fig. 2C). Moreover, the level of mRNA induction was similar to activation levels observed following delivery of the recombinantly expressed dCas9-VP64 genetic fusion protein counterpart (SI Appendix,Fig.S4), and the signal was completely

Fig. 1. Strategy for localizing synthetic molecules to specific genomic loci using a modified CRISPR/Cas9 system. Purified dCas9-IntN, IntC-cargo, and gRNAs are mixed in a one-pot reaction (1) to afford a dCas9-cargo:gRNA complex along with the excised intein fragments. This reaction mixture is then directly delivered to live cultured cells using a cationic lipid-mediated transfection strategy (2). Once inside the cell, the dCas9-cargo:gRNA complex enters the nucleus and binds to its target DNA sequence, resulting in localized activity of the synthetic cargo (3).

product to cells. In addition, the intein tags are excised from the final construct and thus will not effect Cas9 delivery or downstream applications. To explore the feasibility of our proposed methodology, we used a PTS reaction to ligate dCas9 to a potent transcription- Fig. 2. Cellular delivery of PTS-modified dCas9. (A) Schematic of the splicing activating peptide, VP64 (Fig. 2A). Previous studies have shown reaction between recombinantly expressed dCas9-NLS-IntN (1) and IntC- that when targeted to the promoter region of genes such as VP64-HA yields the dCas9-NLS-VP64-HA delivery product (2) and the inert interleukin-1 receptor antagonist (IL1RN) and neurotrophin-3 split intein complex. (B) The reaction from A as monitored by an SDS/PAGE NTF3 gel shift assay. The dCas9-NLS-IntN starting material (1) undergoes conver- ( ), the dCas9-VP64 fusion protein recruits transcriptional sion to the dCas9-NLS-VP64 delivery product (2) (96% in 20 min as measured machinery that induces local histone modifications, as well as a by densitometry). This reaction was performed using 1 μMof(1),2μM buildup of corresponding mRNA levels (10, 25). This epigenetic IntC-VP64-HA, and a total of 1 μMofIL1RN gRNAs. The gel was imaged using and transcriptional plasticity make these regions suitable targets Coomassie brilliant blue stain. (C) Gene activation in the presence of dif- for this study. For our pilot experiment, we generated dCas9- ferent dCas9-cargo molecules delivered into 293T cells. ΔVP64 represents IntN (a genetic fusion of dCas9 and the N-terminal fragment of cargo in which the VP64 sequence has been replaced with a cysteine residue. a split intein) as well as IntC-VP64 (a genetic fusion of the dCas9-cargo fusions were generated via PTS. Cells were incubated for 20 h in C-terminal fragment of a split intein and VP64) by recombinant the presence of a given dCas9 complex (final concentration, 100 nM) before protein expression in Escherichia coli (SI Appendix, Fig. S1). The harvesting and subsequent mRNA quantification via RT-qPCR. The relative requisite gRNAs were generated using an in vitro transcription mRNA level represents the fold increase in signal that is observed relative to system (SI Appendix, Fig. S1 and Table S1). Once purified, the an off-target delivery. mRNA signals from all samples are normalized to GAPDH mRNA before fold differences are calculated. Error bars represent three delivery components were mixed, and the PTS reaction was SEM from three biological replicates. *P ≤ 0.05; **P ≤ 0.01. (D) 293T cells monitored by an SDS/PAGE gel shift assay that detects con- containing a GFP reporter plasmid that have been treated with dCas9-VP64 version of dCas9-IntN to dCas9-VP64 (Fig. 2B and SI Appendix, protein in complex with either reporter-targeted or off-target gRNAs. Cells Fig. S2). Notably, the PTS reaction was carried out directly in were visualized by microscopy at 36 h after dCas9-VP64 transfection. (Scale cell media to allow delivery of the complex without further bar: 100 μm.)

682 | www.pnas.org/cgi/doi/10.1073/pnas.1615723114 Liszczak et al. Downloaded by guest on September 30, 2021 abolished when VP64 was deleted from the cargo sequence (Fig. 2C). Taken together, these results validate our ability to deliver PTS-modified dCas9 to specific genomic loci. Because the signal from the RNA activation assay represents a population average, we sought to implement a secondary assay to assess the efficiency of dCas9 delivery at the single-cell level. We reasoned that activation of a GFP reporter plasmid would allow us to quantify the percentage of cells with sustained localization of functionally active dCas9, which will be a requirement for many applications of this technology. Delivery of the dCas9- VP64 splice product to 293T cells containing a GFP reporter plasmid led to robust activation of GFP when targeting the upstream activation sequence in the reporter (Fig. 2D). We also used flow cytometry to quantify the efficiency of delivery, and determined that 30% of cells were GFP-positive (SI Appendix, Fig. S5). This is in line with previously reported Cas9 nuclease delivery studies, which found genome editing efficiencies ranging from 30% to 50% in 293T cells when a lipid-based protein transfection approach was used (25, 26). As noted above, these efficiencies are compatible with common cell-based assays, such as imaging and qPCR-based techniques.

Small-Molecule Conjugation to dCas9. We next asked whether this strategy could be extended to allow ligation of a fully synthetic entity to dCas9-gRNA complexes for cellular delivery. The use of small-molecule epigenetic probes is an attractive option for dCas9 conjugation because it allows for genomically localized modulation of target chromatin effector function, thereby expanding the utility of these pharmacologic tools. A prominent example of an epigenetic effector ligand is the bromodomain inhibitor JQ1 (27). This molecule interacts with the BET family of proteins and competitively blocks their association with acetyl- lysine residues on chromatin. In this sense, JQ1 can be consid- ered a stable mimic of the acetyl-lysine mark, which, importantly, cannot be removed through cellular deacetylase activity. We wondered whether a dCas9-JQ1 conjugate would be sufficient for the recruitment of BET proteins to target DNA sequences. Synthetic access to the dCas9-JQ1 conjugate was accomplished using PTS. In this case, we relied on previous structure- activity data to guide the design of an IntC-JQ1 construct in which the probe was predicted to retain activity (28). Specifically, the free-acid form of JQ1 was attached via a flexible hydrophilic linker to a cargo peptide bearing a reactive cysteine (SI Appen- dix, Fig. S6 and Schemes S1 and S2). This JQ1 linker was then ligated via expressed protein ligation to a recombinantly generated polypeptide containing the IntC sequence (Fig. 3A and SI Appendix, Fig. S7). Notably, our synthetic route was designed to be highly modular, allowing other synthetic entities to be easily substituted. This is illustrated by the preparation of a control molecule containing the functionally inactive enantiomer of JQ1, (−)-JQ1 (SI Appendix, Figs. S8 and S9). Importantly, the reaction of these IntC-JQ1 constructs with dCas9-IntN was found to be both rapid and efficient (>90% complete in 20 min), indicating that the PTS reaction is not affected by the chemical nature of the IntC cargo (SI Appendix, Fig. S10). We next asked whether dCas9-tethered JQ1 retained the ability to interact with its cognate binding partners. To this end, we generated a GST-tagged Brd4 construct containing the tandem bromodomain motif (Brd4T; amino acids 49–460) for use in in vitro Fig. 3. dCas9-JQ1 interacts with Brd4 in vitro and in live cells. (A)Thestruc- binding assays with dCas9-JQ1. Consistent with our design, the ture of IntC-fused JQ1 (JQ1 moiety shown in red). For simplicity, the 36-aa IntC sequence is represented as a blue box. Not depicted is a short spacer peptide (wavy bond) between the IntC sequence and the cargo peptide shown in black. (The complete sequence is provided in SI Appendix, Materials and replaced with a cysteine residue. 293T cells were incubated for 24 h in the Methods.) (B) An in vitro GST pull-down assay showing the interaction be- presence of 100 nM dCas9-complexes before harvesting. Prey protein over- tween GST-tagged Brd4T (1 μM) and either dCas9-JQ1 (1 μM) or dCas9-(-)-JQ1 expression plasmids and the luciferase reporter were transfected during (1 μM) in the presence of increasing concentrations of free JQ1 (0–100 μM). dCas9 delivery. The relative luciferase activation for each experiment rep- The SDS/PAGE gels were imaged using Coomassie brilliant blue stain. (C) resents the fold increase in signal observed for the Gal4-localized sample Schematic of the luciferase assay used to detect the interaction between relative to the signal observed from an off-target delivery. Notations for dCas9-JQ1 conjugates and Brd4T in cells. (D) Luciferase activation levels in exogenous prey proteins are as follows: WT, VP64-Brd4T (WT)-VP64; Mut,

the presence of different prey proteins and dCas9-bait complexes. The ΔJQ1 VP64-Brd4T (Y97A/Y390A)-VP64; −, no exogenous prey protein. Error bars BIOCHEMISTRY bait is identical to the JQ1 conjugate, except that the JQ1 cargo peptide is represent SEM from three biological replicates. **P ≤ 0.01.

Liszczak et al. PNAS | January 24, 2017 | vol. 114 | no. 4 | 683 Downloaded by guest on September 30, 2021 dCas9-JQ1 conjugate, but not the dCas9-(−)-JQ1 control conjugate, was able interact with GST-Brd4T in a manner compet- itive with free JQ1 (Fig. 3B and SI Appendix, Fig. S11).

dCas9-JQ1 Recruits Brd4 to Genomic Loci in Live Cells. Encouraged by the in vitro binding assays, we sought to develop a sensitive and facile assay for detection of the dCas9-JQ1–Brd4 interaction in live cells. Accordingly, we established a modified two-hybrid luciferase reporter assay that detects dCas9-JQ1 (“bait”) recruitment of a transiently transfected VP64-Brd4T- VP64 fusion protein (“prey”)(Fig.3C). Cellular delivery of dCas9- JQ1-gRNA complexes that target Gal4 binding sites within the upstream activation sequence of the luciferase reporter plasmid (SI Appendix,TableS1) led to an eightfold increase in luciferase signal relative to signal from a control delivery using off-target gRNA sequences (Fig. 3D). In addition, no stimulation of luciferase signal was observed on delivery of control dCas9 conjugates that either lacked JQ1 (ΔJQ1; SI Appendix,Fig.S12) or contained the inactive enantiomer of the drug, (−)-JQ1 (Fig. 3D). We also performed this assay with a modified prey protein containing mutations in each Brd4T bromodomain (Y97A/Y390A), which are known to ablate JQ1 binding (29). Unexpectedly, we still observed a threefold increase in luciferase signal when this mutated Brd4T was used in our two- hybrid assay (Fig. 3D). Conceivably, this stimulation might be due to residual binding of dCas9-JQ1 to the mutated Brd4T protein; however, omission of the prey construct altogether led to a similar result (Fig. 3D). This finding suggests that endogenous BET proteins, which are known transcriptional coactivators (30, 31), are recruited by dCas9-JQ1 and can activate luciferase expression, albeit to a lesser degree than the VP64-Brd4T-VP64 fusion. Gratifyingly, this activation also could be inhibited by the addition of free JQ1 to these deliveries, whereas dCas9-VP64 activation levels remained unaffected, further validating the role of endogenous BET proteins in gene activation by dCas9-JQ1 SI Appendix Fig. 4. Multimeric display of JQ1 enhances activity and recruits endogenous ( , Fig. S13). Brd4 to targeted genomic loci. (A) The structure of the IntC-fused multimeric We wondered whether increasing the number of JQ1 mole- JQ1 construct (JQ1 × 4). For simplicity, the IntC sequence is represented as a cules localized to a specific genetic element would be an effective blue box and JQ1 is depicted as a red star. Not depicted are short spacer strategy for enhancing BET protein recruitment levels. In the peptides (wavy bonds) between JQ1 and the cargo peptide shown in black, construct described above, the maximum number of cargo mol- as well as the IntC sequence and the cargo peptide. More details are pro- ecules targeted to a locus by a single dCas9 conjugate is equal to vided in SI Appendix, Materials and Methods.(B) Luciferase activation levels the number of gRNA binding sites used. We imagined that in- in the presence of different dCas9-bait complexes and prey proteins (de- creasing the number of JQ1 copies on each dCas9 conjugate scribed in Fig. 3D). Assays were performed, and relative luciferase signal was would be an effective way to circumvent this limitation. Thus, we calculated as described for monomeric JQ1 luciferase activation experiments. designed and synthesized a branched cargo peptide that allows (C) Binding of delivered dCas9-JQ1 × 4atIL1RN and NTF3 promoter regions four JQ1 moieties to be displayed (SI Appendix, Figs. S14 and as determined by bioChIP-qPCR. (D) Anti-Brd4 ChIP-qPCR enrichment levels × S15 and Schemes S3 and S4). This was subsequently ligated to at IL1RN and NTF3 promoter regions when dCas9-JQ1 4 was delivered to the IntC polypeptide using our modular expressed protein liga- these sites. In C and D, 293T cells were incubated for 8 h with 100 nM of dCas9-JQ1 × 4:gRNA. Relative biotin or Brd4 enrichment represents the fold tion strategy (Fig. 4A and SI Appendix, Fig. S16). This branched × increase in percent input observed at the target locus relative to an off- construct (IntC-JQ1 4) was then reacted with dCas9-IntN to yield target delivery. Within each sample, percent input signals at each promoter dCas9-JQ1 × 4, which retained a robust interaction with Brd4T SI Appendix are normalized to the unmodified RHOB housekeeping gene promoter in vitro, as verified by our GST pull-down assay ( ,Fig. percent input signal before calculating fold differences. Error bars represent S17). Remarkably, cellular delivery of the dCas9-JQ1 × 4construct SEM from three biological replicates. *P ≤ 0.05; **P ≤ 0.01. in complex with gRNA resulted in a 27-fold increase in luciferase activity relative to the signal from an off-target delivery (Fig. 4B). This presumably reflects improved recruitment of the VP64-Brd4T- biotin tag present on the JQ1 dendrimer scaffold, showed VP64 prey construct. The multimeric display strategy also led to an gRNA-dependent localization of the dCas9 conjugate to the enhanced luciferase signal in the absence of exogenous Brd4T, targeted promoters (Fig. 4C). Encouraged by this result, we then suggesting improved recruitment of endogenous BET proteins × B explored whether genomic localization of dCas9-JQ1 4 was (Fig. 4 ). Thus, increasing the number of probe molecules linked accompanied by recruitment of endogenous Brd4. Indeed, ChIP- to dCas9 through dendritic display is an effective strategy for in- qPCR assays using a Brd4-specific antibody revealed locus-specific creasing the functional output of both exogenous and endogenous recruitment of this protein as a function of our targeting system copies of BET proteins. D Our observation that both dCas9-JQ1 and dCas9-JQ1 × 4 (Fig. 4 ). These results illustrate the ability of our strategy to cause gene activation in the absence of the VP64-Brd4T-VP64 infiltrate a native cellular environment and redirect endogenous fusion led us to further explore the ability of these conjugates to machinery to genomic loci of interest. recruit endogenous BET proteins to designated genomic loci. To first ensure that dCas9-JQ1 × 4 could be localized to endogenous Delivery of a Modified Peptide Cargo for Targeted Recruitment of the genes, we targeted the construct to the promoter region of either Endogenous PRC1 Complex. We next sought to deliver an epige- IL1RN or NTF3 and performed a biotin-based ChIP-qPCR as- netic probe that specifically interacts with an endogenous mul- say (bioChIP). This assay, which exploits the splicing-dependent tisubunit protein complex. In contrast, genetic fusion methods

684 | www.pnas.org/cgi/doi/10.1073/pnas.1615723114 Liszczak et al. Downloaded by guest on September 30, 2021 are less applicable to multisubunitproteincomplexes, because in (Ring1b and PCGF4/Bmi1), whereas the dCas9-JQ1 × 4 control most cases, only a single subunit can be fused to the targeting vector. molecule interacted specifically with Brd4, as expected (Fig. 5C). As an ideal candidate for these studies, we chose to focus on the To confirm successful delivery and localization of dCas9- PRC1, an archetypal epigenetic silencing factor involved in estab- UNC3866, we again performed a bioChIP-qPCR assay and observed lishment of cell identity during development (32). This complex is robust localization of the modified delivery molecule at targeted composed of a minimum of four protein subunits with varying promoters, with >20-fold enrichment at IL1RN and ∼10-fold numbers of homologs, including the chromobox protein family (Cbx) enrichment at NTF3 (Fig. 5D). Finally, having shown that dCas9- that interacts with modified histone H3 (H3K27me3) to recruit the UNC3866 is capable of binding PRC1 and that it localizes to des- complex to chromatin (33, 34). Importantly, a recently developed ignated genomic loci after cellular delivery, we interrogated the high-affinity peptide ligand for Cbx proteins, UNC3866 (35), has a potential for the conjugate to support targeted genomic recruitment structure and properties that seem well suited for locus-specific PRC1 of endogenous PRC1 by performing a ChIP-qPCR assay using a recruitment via our dCas9 conjugation approach (Fig. 5 A and B). Ring1b-specific antibody. Gratifyingly, we observed specific localiza- Following our established synthetic approach, we generated an tion of this PRC1 subunit at dCas9-UNC3866 target sites (Fig. 5E). IntC-UNC3866 molecule that could be directly spliced onto This result confirms the localization of multiple PRC1 complex dCas9-IntN quickly and efficiently (>90% complete in 20 min) components, because UNC3866 can only recruit Ring1b via its in- (SI Appendix, Figs. S18–S20 and Scheme S5). To confirm that the teraction with the Cbx subunits (35). dCas9-UNC3866 conjugate is able interact with the PRC1 complex, we incubated the molecule with 293T cell lysates and performed a Discussion subsequent biotin chemiprecipitation assay. We found that Genomic localization of synthetic entities holds great promise dCas9-UNC3866 was able to specifically pull down multiple for dissecting epigenetic mechanisms and other facets of chro- components of the PRC1 complex, including the target protein matin biology (17). Progress in this area has been hampered by a (Cbx4), as well as both subunits of the E3 ligase heterodimer lack of techniques for generating facile DNA-binding conjugates with a high degree of sequence specificity. By combining flexible strategies in protein chemistry with the remarkably specific and easily programmable CRISPR/Cas9 system, we have developed a method that we anticipate will be broadly useful for locus-specific targeting of chemical cargos, including epigenetic probes, chemical imaging agents, and other tools for studying nuclear processes. We have shown here that PTS can be used to chemoselectively modify dCas9 with proteinaceous material as well as small molecules. Furthermore, the reaction is complete in minutes and can be performed directly in cell media to allow for streamlined delivery to cells without an additional buffer exchange or purification step. A key feature of our strategy is its modularity. Different probes can be site-specifically attached to the generic dCas9-IntN building block by incorporating unique payloads into the corresponding IntC-cargo component. Importantly, we have developed an efficient semisynthetic route to the IntC-cargo reactant that provides flexibility in both the chemical nature of the cargo and its architecture. This is illustrated by the generation of IntC components with either small-molecule (JQ1) or synthetic peptide-based (UNC3866) cargos. We also show that incorporation of multiple copies of a synthetic entity onto dCas9 can be accomplished using this method via a dendrimeric display strategy. Ultimately, in the case of JQ1, the increased local concentration of the molecule enabled by this multimeric display proved instrumental in driving a robust functional output at designated genomic loci. Conceivably, this scaffold could be adapted to allow the conjugation of multiple different cargoes to the same dCas9 vehicle. It is also important to note that there are limitations regarding the exact composi- tion of the cargo that can be delivered into live cells using this technique. For example, during the course of this study, we attempted to deliver a dendrimeric version of the UNC3866 Fig. 5. Recruitment of the PRC1 complex to genomic loci by dCas9-UNC3866. molecule containing four copies of the probe (SI Appendix, Figs. (A) The structure of the IntC-fused UNC3866 molecule (IntC-UNC3866). The S21–S23 and Scheme S6), but although we were able to syn- UNC3866 moiety is shown in red, and the Linker-biotin region is shown in black. thesize this molecule and show splicing onto dCas9, we were For simplicity, the IntC sequence is represented as a blue box. Not depicted is a unable to detect genomic localization using the bioChIP assay short spacer peptide (wavy bond) between the IntC sequence and the cargo (SI Appendix, Fig. S24). It is possible that the size, hydropho- peptide. More details are provided in SI Appendix, Materials and Methods. bicity, and/or some other property of the molecule prevent it from (B) Schematic representing recruitment of the PRC1 complex to a gRNA- efficiently penetrating the cells, escaping endosomes, and/or binding site by dCas9-UNC3866. (C) Western blot analysis of the 293T cell × engaging the genome. We also note that other site-specific lysate chemiprecipitation by either dCas9-UNC3866 or dCas9-JQ1 4. bioconjugation strategies could be adapted to our approach (36), (D) Binding of delivered dCas9-UNC3866 at IL1RN and NTF3 promoter re- although we stress that the PTS-based system used in this study is gions as determined by bioChIP-qPCR. (E) Anti-Ring1b ChIP-qPCR enrichment levels at IL1RN and NTF3 promoter regions when dCas9-UNC3866 was delivered fast and highly efficient, and allows the reaction to be performed to these sites. For C and D, 293T cells were incubated for 8 h with 100 nM of directly in the cell culture medium. Regardless, the key concept dCas9-UNC3866:gRNA. Relative biotin or Ring1b enrichment represents the in this study is the ability to direct a synthetic molecule, such as fold increase in percent input observed at the target locus relative to an off- an epigenetic probe, to a specific region of chromatin using the target delivery. Within each sample, percent input signals at each promoter dCas9-targeting vehicle. were normalized to the RHOB promoter percent input before fold differ- The specific application of the technology presented here involves

ences were calculated. Error bars represent SEM from three biological rep- directing a pharmacologic agent to a designated part of the genome, BIOCHEMISTRY licates. *P ≤ 0.05; **P ≤ 0.01. which in the broader sense is an example of subcellular-targeted

Liszczak et al. PNAS | January 24, 2017 | vol. 114 | no. 4 | 685 Downloaded by guest on September 30, 2021 pharmacology (37). The present work also offers an alternative chromatin substrate, so-called “cross-talk” (40). One advantage to using genetic fusions to dCas9 for genomic localization of of the present method is its potential to bypass the need for the chromatin effectors. This would be an attractive option for sys- enzyme altogether by allowing direct genomic targeting of a tems in which the introduction of exogenous DNA is difficult or nominal reaction product, for example, a modified histone peptide. to be avoided, such as certain primary cell lines (38), as well as In the application described here, we use chemical probes as a when protein overexpression fails to recapitulate biologically stable surrogate for histone acetyl-lysine (JQ1) and histone trime- relevant molecules/complexes. We also note that the protein thylation (UNC3866) marks, and show that it is possible to recruit delivery strategy used here results in a pulse of cellular activity specific chromatin reader proteins to designated genes. This ap- rather than the sustained activity associated with overexpression proach should be adaptable to the delivery of other synthetically of purely genetic Cas9 fusions. This makes the delivery approach accessible “actuators” to chromatin, including histone peptides better suited to some problems than to others. We have dem- bearing preinstalled posttranslational modifications, as well as small onstrated that the method allows for temporally controlled re- molecules that target other epigenetic factors (3). As a further ex- cruitment of complex endogenous factors to specific genetic loci. tension, it also should be possible to deliver fluorophores and cross- This naturally lends itself to other pulse-chase type experiments linkers to specific genomic loci for use in live cell imaging and in which delivery leads to an immediate change in the local probing of local protein content or chromatin contact points, chromatin environment. By the same token, processes that re- respectively. Cumulatively, such chemical tailoring of dCas9 has quire more prolonged activity to induce a robust output may be the potential to unlock a new type of cellular biochemistry re- better suited to a purely genetic strategy. For example, tran- search involving precise functionalization of eukaryotic chro- scriptional responses following the delivery of dCas9-activator matin and leading to a greater understanding of associated proteins are typically weaker than those associated with expres- regulatory mechanisms. sion of similar constructs (25). Genetically encoded programmable DNA-binding proteins Materials and Methods have emerged as powerful tools for studying genomic processes The production of recombinant proteins, gRNA molecules, and splicing- (16, 39). The CRISPR/Cas9 system is especially versatile competent cargo molecules, as well as protocols for all assays and delivery and shows great promise for exploring the causal biochemical procedures used in this study are described in detail in SI Appendix, Materials relationships underlying epigenetic regulation. In particular, the and Methods. ability to genetically fuse a chromatin-modifying enzyme, such as a histone acetyltransferase, to dCas9 and target that activity to a ACKNOWLEDGMENTS. We thank the current and former members of the designated genomic locus offers the potential to locally decorate T.W.M. laboratory, in particular A. Stevens and R. Thompson, for stimulating chromatin with a histone mark and then study the functional discussions. We also thank Christina DeCoste (Princeton University Flow Cytometry Facility) and the staff of the Princeton University Microarray Fa- consequences (10). However, the specificity and substrate scope of cility. This work was supported by National Institutes of Health (NIH) Grants these writer enzymes is not always clear, and, moreover, many R37 GM086868, R01 GM107047 and P01 CA196539. G.P.L. was supported by appear to be regulated by preexisting modifications on the NIH Research Service Award 1F32GM110880.

1. Goldberg AD, Allis CD, Bernstein E (2007) Epigenetics: A landscape takes shape. Cell 21. Kaihatsu K, Janowski BA, Corey DR (2004) Recognition of chromosomal DNA by PNAs. 128(4):635–638. Chem Biol 11(6):749–758. 2. Dunham I, et al.; ENCODE Project Consortium (2012) An integrated encyclopedia of 22. Shah NH, Dann GP, Vila-Perelló M, Liu Z, Muir TW (2012) Ultrafast protein splicing is DNA elements in the human genome. Nature 489(7414):57–74. common among cyanobacterial split inteins: Implications for protein engineering. 3. Fierz B, Muir TW (2012) Chromatin as an expansive canvas for chemical biology. Nat J Am Chem Soc 134(28):11338–11341. Chem Biol 8(5):417–427. 23. O’Connell MR, et al. (2014) Programmable RNA recognition and cleavage by CRISPR/ 4. Badeaux AI, Shi Y (2013) Emerging roles for chromatin as a signal integration and Cas9. Nature 516(7530):263–266. storage platform. Nat Rev Mol Cell Biol 14(4):211–224. 24. Deng W, Shi X, Tjian R, Lionnet T, Singer RH (2015) CASFISH: CRISPR/Cas9-mediated in 5. Dawson MA, Kouzarides T (2012) Cancer epigenetics: From mechanism to therapy. situ labeling of genomic loci in fixed cells. Proc Natl Acad Sci USA 112(38): – Cell 150(1):12–27. 11870 11875. 6. Cole PA (2008) Chemical probes for histone-modifying enzymes. Nat Chem Biol 4(10): 25. Zuris JA, et al. (2015) Cationic lipid-mediated delivery of proteins enables efficient – 590–597. protein-based genome editing in vitro and in vivo. Nat Biotechnol 33(1):73 80. 7. Arrowsmith CH, Bountra C, Fish PV, Lee K, Schapira M (2012) Epigenetic protein 26. Liang X, et al. (2015) Rapid and highly efficient mammalian cell engineering via Cas9 – families: A new frontier for drug discovery. Nat Rev Drug Discov 11(5):384–400. protein transfection. J Biotechnol 208:44 53. 8. Rivenbark AG, et al. (2012) Epigenetic reprogramming of cancer cells via targeted 27. Filippakopoulos P, et al. (2010) Selective inhibition of BET bromodomains. Nature – DNA methylation. Epigenetics 7(4):350–360. 468(7327):1067 1073. 9. Mendenhall EM, et al. (2013) Locus-specific editing of histone modifications at en- 28. Anders L, et al. (2014) Genome-wide localization of small molecules. Nat Biotechnol 32(1):92–96. dogenous enhancers. Nat Biotechnol 31(12):1133–1136. 29. Floyd SR, et al. (2013) The bromodomain protein Brd4 insulates chromatin from DNA 10. Hilton IB, et al. (2015) Epigenome editing by a CRISPR-Cas9–based acetyltransferase damage signalling. Nature 498(7453):246–250. activates genes from promoters and enhancers. Nat Biotechnol 33(5):510–517. 30. LeRoy G, Rickards B, Flint SJ (2008) The double bromodomain proteins Brd2 and Brd3 11. Konermann S, et al. (2013) Optical control of mammalian endogenous transcription couple histone acetylation to transcription. Mol Cell 30(1):51–60. and epigenetic states. Nature 500(7463):472–476. 31. Rahman S, et al. (2011) The Brd4 extraterminal domain confers transcription activa- 12. Gersbach CA, Gaj T, Barbas CF, 3rd (2014) Synthetic zinc finger proteins: The advent of tion independent of pTEFb by recruiting multiple proteins, including NSD3. Mol Cell targeted gene regulation and genome modification technologies. Acc Chem Res Biol 31(13):2641–2652. 47(8):2309–2318. 32. Di Croce L, Helin K (2013) Transcriptional regulation by Polycomb group proteins. Nat 13. Cong L, et al. (2013) Multiplex genome engineering using CRISPR/Cas systems. Science Struct Mol Biol 20(10):1147–1155. – 339(6121):819 823. 33. Simon JA, Kingston RE (2009) Mechanisms of polycomb gene silencing: Knowns and 14. Mali P, et al. (2013) RNA-guided human genome engineering via Cas9. Science unknowns. Nat Rev Mol Cell Biol 10(10):697–708. – 339(6121):823 826. 34. Gao Z, et al. (2012) PCGF homologs, CBX proteins, and RYBP define functionally – 15. Perez-Pinera P, et al. (2013) RNA-guided gene activation by CRISPR-Cas9 based distinct PRC1 family complexes. Mol Cell 45(3):344–356. – transcription factors. Nat Methods 10(10):973 976. 35. Stuckey JI, et al. (2016) A cellular chemical probe targeting the chromodomains of 16. Keung AJ, Bashor CJ, Kiriakov S, Collins JJ, Khalil AS (2014) Using targeted chromatin Polycomb repressive complex 1. Nat Chem Biol 12(3):180–187. regulators to engineer combinatorial and spatial transcriptional regulation. Cell 36. Stephanopoulos N, Francis MB (2011) Choosing an effective protein bioconjugation 158(1):110–120. strategy. Nat Chem Biol 7(12):876–884. 17. Højfeldt JW, Van Dyke AR, Mapp AK (2011) Transforming ligands into transcriptional 37. Rajendran L, Knölker HJ, Simons K (2010) Subcellular targeting strategies for drug regulators: Building blocks for bifunctional molecules. Chem Soc Rev 40(8):4286–4294. design and delivery. Nat Rev Drug Discov 9(1):29–42. 18. Kang JS, Meier JL, Dervan PB (2014) Design of sequence-specific DNA binding mole- 38. Kim S, Kim D, Cho SW, Kim J, Kim JS (2014) Highly efficient RNA-guided genome editing in cules for DNA methyltransferase inhibition. J Am Chem Soc 136(9):3687–3694. human cells via delivery of purified Cas9 ribonucleoproteins. Genome Res 24(6):1012–1019. 19. Dervan PB, Edelson BS (2003) Recognition of the DNA minor groove by pyrrole- 39. Shalem O, Sanjana NE, Zhang F (2015) High-throughput functional genomics using imidazole polyamides. Curr Opin Struct Biol 13(3):284–299. CRISPR-Cas9. Nat Rev Genet 16(5):299–311. 20. Nielsen PE (2010) Peptide nucleic acids (PNA) in chemical biology and drug discovery. 40. Suganuma T, Workman JL (2011) Signals and combinatorial functions of histone Chem Biodivers 7(4):786–804. modifications. Annu Rev Biochem 80:473–499.

686 | www.pnas.org/cgi/doi/10.1073/pnas.1615723114 Liszczak et al. Downloaded by guest on September 30, 2021