CRISPR-mediated multiplexed live cell imaging of nonrepetitive genomic loci with one guide RNA per locus

Three-dimensional (3D) structures of the genome are dynamic, heterogeneous and functionally important. Live cell imaging has become the leading method for chromatin dynamics tracking. However, existing CRISPR- and TALE-based genomic labeling techniques have been hampered by laborious protocols and are ineffective in labeling non-repetitive sequences. Here, we report a versatile CRISPR/Casilio-based imaging method that allows for a nonrepetitive genomic locus to be labeled using one guide RNA. We construct Casilio dual-color probes to visualize the dynamic interactions of DNA elements in single live cells in the presence or absence of the cohesin subunit RAD21. Using a three-color palette, we track the dynamic 3D locations of multiple reference points along a chromatin loop. Casilio imaging reveals intercellular heterogeneity and interallelic asynchrony in chromatin interaction dynamics, underscoring the importance of studying genome structures in 4D.

T hree-dimensional (3D) organization of the genome has pervasive roles in the regulation of transcription, DNA replication, and DNA repair [1][2][3] . Sequencing methods 4,5 can uncover genome-wide chromatin interactions in bulk cell populations but are unable to resolve intercellular heterogeneity or temporal dynamics (the "fourth dimension") which instead require live cell imaging techniques 6 . Four-dimensional (4D) live cell imaging of genomic loci can be achieved by inserting into the target loci large tandem arrays of sequences that are bound by well-characterized DNA binding domains fused to fluorescent proteins providing important insights into the dynamics of chromatin interaction and transcription 7,8 . However, these techniques require tedious genome engineering and can be confounded by the side effects of disruptive introduction of exogenous sequences at the target loci. Live-cell imaging of the unmodified genome can be conducted with the use of fluorescent proteins fused to programmable DNA binding proteins such as zinc finger proteins (ZFPs) and transcription activator-like effectors (TALEs), but the complicated cloning protocols requiring weeks of development limit their scalability [9][10][11][12] . CRISPR/Cas systems offer versatile DNA binding proteins (e.g., Cas9) that can be programmed by a target-specific guide RNA (gRNA), revolutionizing the field of genome biology 13 . Fluorescent proteins or fluorophores recruited by dCas9 or assembled on the gRNA scaffold offer easily programmable platforms for live cell genomic imaging [14][15][16][17][18][19][20][21][22] (Supplementary Table 1). However, despite current techniques providing robust labeling of repetitive sequences using a single gRNA, imaging of nonrepetitive sequences requires simultaneous delivery of up to tens of different gRNA per target to achieve sufficient signal, precluding broad adoption.
We previously engineered a modular multitasking CRISPR/ Cas-based platform called Casilio by combining dCas9 and engineered Pumilio/FBF (PUF)-tethered effectors 23 (Fig. 1). An effector domain carrying out a specific function (e.g., transcription activation) is fused to a unique PUF RNA binding domain that can bind to one or more copies of a specific 8-mer motif inserted to the 3' end of a gRNA (Fig. 1a). PUF domains contain peptide subunits that can each be programmed to recognize an RNA-base by changing amino acids contacting the RNA, thereby allowing designed PUF domains to bind different 8-mer sequences 24,25 . Casilio capitalizes on the versatility of PUF domains to allow simultaneous and orthogonal delivery of effector functions to distinct genome loci, as well as multimerization of effector molecules, for efficient epigenetic editing 23,26 (Fig. 1b).
In this work, we develop Casilio for imaging nonrepetitive DNA sequences with one gRNA per locus. We apply Casilio to visualize dynamic chromatin interactions in the presence or absence of cohesin subunit RAD21 in live cells. Finally, we expand fluorescent palette of Casilio to allow simultaneous labeling of three loci along a chromatin loop to track its folding dynamics.

Results
One-gRNA labeling of a nonrepetitive locus by Casilio. We reasoned that the immense multimerization capability of Casilio will enable concentration of fluorescent proteins at a target locus to achieve sufficient signal using reduced number of target DNA binding sites while the multiplexing capability will allow simultaneous multicolor labeling of distinct loci in the same cell (Fig. 1c, d). The MUC4 gene on chromosome 3 was imaged in the first and subsequent CRISPR-based imaging studies by tiling gRNAs over the target region 14,18,22 . In addition to nonrepetitive sequences, the MUC4 gene harbors repetitive sequences allowing convenient DNA-FISH co-labeling validation using a single oligonucleotide probe 14 . Therefore, we first tested Casilio for imaging nonrepetitive sequences within the MUC4 gene. We designed gMUC4 which binds a nonrepetitive sequence within MUC4 and inserted 15 copies of PUF Binding Site c (15xPBSc) to recruit Clover-PUFc. To allow the assessment of the specificity of Casilio, we co-labeled the MUC4 region with a Cy5-MUC4 DNA fluorescence in situ hybridization (FISH) probe in U2OS and ARPE-19 cells labelled with Casilio/gMUC4 (Fig. 2a-c). Casilio achieves highly specific labeling of the nonrepetitive locus using only one gRNA with 96% and 100% specificity in U2OS and ARPE-19 cells, respectively. Additional support for specificity is that the formation of fluorescent spots is dependent on the presence of dCas9 (Fig. 2d), and that the number of fluorescent spots is consistent with the ploidy of the diploid ARPE-19 (Fig. 2d) and haploid HAP1 (Fig. 2e) cells. Spot formation efficiency was around~68% for ARPE-19 ( Fig. 2d) and~54% for HAP1 (Fig. 2e). These together demonstrate that robust and specific labeling of nonrepetitive DNA sequence by Casilio requires only one gRNA.

Tracking of chromatin interactions with dual-color labeling.
To test whether Casilio can be applied to study the dynamics of chromatin interactions in live cells, we selected a chromatin interaction, the MASP1-BCL6 loop, with 362 kb genomic distance between the interaction anchors from a published RAD21 ChIA-PET experiment performed in ARPE-19 cells (ENCSR110JOO, Michael Snyder lab). We then designed a pair of gRNAs to bind nonrepetitive sequences next to the anchors (Fig. 3a). We used Clover-PUFc to label the 5' anchor (locus A) and PUF9R-iRFP670 to label the 3' anchor (locus B). Live cell microscopy of ARPE-19 cells transfected with dual-color Casilio probe pairs revealed dynamic chromatin interactions at second timescales and showcased a high degree of intercellular heterogeneity and interallelic asynchrony of chromatin interaction at this loop (Fig. 3b, c; Supplementary Video 1; Supplementary Figs. 1, 2a). Such insights cannot be readily obtained via bulk cell sequencing technologies, supporting the advantages of Casilio live-cell imaging for studying the dynamics of chromatin interactions at single cell and single allele level. To provide future users a method to assess the specificity of labeling in live-cell settings, we developed a "competitor" assay. Here, a "cold" specific competitor gRNA without any PUF Binding Sites (PBS) for fluorescent protein recruitment, but overlaps the target, is co-introduced into cells to interfere with the binding of Casilio imaging complexes and the formation of target spots (Fig. 3d). A non-specific competitor gRNA that does not overlap the target (such as GAL4 27 ) is used as a control. To denote a competitor gRNA, we prefix the target with "c" to mean cold or competitor. Specific competitor gRNA cA overlapping with MASP1-BCL6 locus A greatly reduced the formation of the Clover spots compared to non-specific competitor cGAL4 control (Fig. 3e). As expected, the iRFP670 spots formed at the noninterfered location (MASP1-BCL6 locus B) served as an internal control that was unaffected by either competitor gRNAs. To delineate the surrounding chromatin structure over an expanded MASP1-BCL6 region, we tracked an additional location 3' (locus R) 312 kb from locus B (Fig. 3f, g). A-B pair distances (mean 1.36 μm, median 0.97 μm) are significantly lower (Mann-Whitney two-sided p-value = 6.01 × 10 −29 ) than B-R pair distances (mean 1.81 μm, median 1.40 μm), consistent with the ChIA-PET data which shows more extensive interactions between A-B compared to B-R (Fig. 3f, h).
To test if Casilio can be applied to study the function of cohesin complexes in chromatin interactions, we took advantage of an engineered HCT116 cell line with endogenous RAD21 fused to an auxin-inducible degron (HCT116/RAD21-mAID) 28 that allows for the conditional depletion of RAD21 by auxin treatment. A previous study performed Hi-C assays on this cell line and demonstrated that the depletion of RAD21 resulted in loss of loop domains including ã 500 kb loop domain between the IER5L promoter (IER5L-P) and its super-enhancer (IER5L-SE) 29 . We used Clover-PUFc and PUF9R-iRFP670 to label the IER5L promoter and super-enhancer, respectively (

Example diresidue Base
Gln Cys A Gln Asn U

Glu Ser G
Arg Ser C Fig. 1 Schematic of Casilio. a In Casilio, gRNA is appended with one or more binding sites for Pumilio/FBF (PUF) RNA binding domain-tethered effectors. Each PUF domain contains 8 peptide subunits each programmable to recognize one of the four RNA bases. A collection of PUF-effectors can be created by fusing different effectors to PUF programmed to recognize different 8-mer PUF binding sequence (PBS). Effectors are recruited to dCas9/gRNA-PBS complex according to the fused PUF domain. b Casilio architecture allows multiple molecules of the same or different effectors (E 1 , E 2 , …, etc.) to be recruited to a target (multimerization or complex formation) as well as different set of effectors to be recruited at different targets in the same cell (multiplexing). This system is expandable to potentially over 65,000 possible effectors. c Casilio-Imaging capitalizes on the multimerization capability to recruit 15 or more fluorescent proteins to a target as opposed to requiring target site tiling of dCas9-GFP direct fusions using 12 or more individual gRNAs in previous studies. d The potential for Casilio to achieve multimodal operations simultaneously, for example, demethylating histone (using LSD1 module), acetylating histone (CBP), methylating DNA (DNMT3A), activating (p65HSF1) different targets, and at the same time monitoring multiple locations with different fluorescent proteins in live cells. dCas9/gRNA-PBS not drawn for simplicity.  (Fig. 4d), consistent with the reduced interaction between the promoter and super enhancer of IER5L upon RAD21 depletion 29 . We also used Casilio to image a "RAD21-independent" interaction connecting two H3K27acenriched super-enhancers and displaying an opposite behavior, in which interaction frequency increased upon RAD21 depletion 29 These data demonstrate that Casilio can image interaction between cis-regulatory elements to study the contribution of protein factors, such as cohesin, on dynamic chromatin interactions underlying gene regulation. Visualization of folding dynamics with a three-color palette. Imaging specific interactions of non-coding elements such as enhancers and promoters provide new insights into gene regulation, while visualizing a continuous stretch of genomic region will improve our understanding of structural folding dynamics and illuminate the process of chromatin loop formation. Given that Casilio can image each nonrepetitive locus with a single gRNA, we next explored the possibility of simultaneously imaging multiple nonrepetitive loci, with the ultimate goal of tracking the "live" structure of a continuous genomic region. We call this technique, where sequential Casilio probes are deployed across a stretch of genomic DNA, "Programmable Imaging of Structure with Casilio Emitted sequence of Signal" or PISCES.
To increase the number of nonrepetitive loci that can be labelled, we developed a three-color palette. We used three gRNAs (IER5L P-20xPBS 48107 , M-20xPBSc, and SE-25xPBS9R) (Fig. 5a) for labeling three nonrepetitive locations at the IER5L P-SE loop, with a three-color fluorescent protein palette  Supplementary  Figs. 2f, g, 6). These experiments demonstrated the use of dual or triple-color fluorescent protein combinations to encode a sequence of PISCES probes for visualizing multiple reference points along a genomic region.

Discussion
In this study, we present a CRISPR/Cas-based method for live cell fluorescence imaging of nonrepetitive genomic loci with low gRNA requirement (one gRNA per locus) and high spatiotemporal resolutions. Although gRNA-scaffold methods based on MS2/PP7/boxB hairpins have been attempted before to reduce gRNA requirement for non-repetitive DNA labeling, the addition of 14 or more of these hairpins to gRNA greatly reduces its expression 18 , whereas in Casilio, addition of up to 47 copies of PBS did not affect dCas9 binding activity 23 . The high signal of target-bound Casilio complexes compared to the background unbound complexes can be explained by the observation that gRNA is unstable unless complexed with dCas9 at the target sites 21 . These unique properties of Casilio enable the reduction of its gRNA requirement compared with previously published CRISPR-based approaches and allow imaging experiments to be conducted with standard, widely available microscopy systems. This, in combination with the convenient oligonucleotide-based cloning protocol 30 , translates into low-cost genome imaging (in our lab, less than USD$5 per target). The "one-gRNA-per-target" requirement not only significantly reduces the technical challenge of using live cell imaging to study chromatin interactions in hardto-transfect cells, but also it simplifies the design of genome-wide gRNA libraries for imaging. We applied Casilio to visualize the dynamic interactions of DNA elements in native, unmodified chromosomes, in the presence or absence of RAD21, showing RAD21-dependent and independent loops display opposite behaviors upon RAD21 depletion. By developing a three fluorescent protein palette, we showed that chromatin loops can be imaged at multiple reference points over time. These revealed the highly heterogeneous and dynamic nature of chromatin interactions and structural changes not attainable with bulk cell sequencing approaches, underscoring the need for studying the 4D nucleome with high spatiotemporal resolution in live cells. As proof-of-principle, we visualized the dynamics of a few cis- regulatory elements in the presence or absence of one of the many proteins involved in the regulation of the 3D genome. Casilio-Imaging should be applied in future studies to track chromatin dynamics over longer time courses upon depletion and replenishment of RAD21, CTCF, WAPL, NIPBL, and PolII, etc. 29,31,32 , across a spectrum of cis-regulatory elements (e.g., enhancers, silencers, promoters) in different epigenetic contexts in order to decipher the dynamic molecular processes underlying 3D genome establishment, maintenance and plasticity. One caveat to bear in mind when conducting and interpreting live-cell genome imaging experiments is the potential of the interference of the natural genome folding processes by the binding of the CRISPR reagents, which might be partially alleviated with careful gRNA designs. It is thus an important future goal to systematically evaluate probe design and produce a design pipeline that takes into account the potential interference of the underlying genome folding processes under investigation. Future development of the Casilio imaging toolkit should include higher-density PISCES probe-set with an expanded palette of fluorescent proteins, in order to enable the folding of the genomic region to be delineated at a higher resolution. The multiplexing and expandable nature of Casilio in adapting new effector functions 23 , in combination with the advance in genomic imaging demonstrated here, will enable "onthe-fly" and "plug-and-play" perturbation of the (epi)genome (e.g., using Casilio activator, repressor and epigenetic editing modules 23,26 ) and concurrent read-out of dynamic 3D chromatin interactions, thereby providing unprecedented flexibility and power for the study of the 4D nucleome (Fig. 1d).

Methods
Design of chromatin interaction imaging experiments. Fastq files of ARPE-19 RAD21 ChIA-PET experiment (ENCSR110JOO) were downloaded from ENCODE database and processed by ChIA-PET2 33 . Interactions were ranked by PET count. Top-ranked interactions with different PET distances were selected. To avoid interference of CTCF binding, design regions were selected by shifting external to PET regions. JACKIE 34 was then used to identify unique 1-copy CRISPR sites in the hg38 genome overlapping design regions of selected loops and further filtered for specificity by Cas-OFFinder 35 . ChIA-PET loops are displayed on WashU EpiGenome Browser 36 .
All plasmids were subjected to restriction diagnostic tests and sequenced to ensure correct cloning. Plasmids are deposited on Addgene (Supplementary Table 2  Live-cell confocal microscopy. Imaging was performed at 41-51 h posttransfection. Prior to imaging, cells were stained with 0.5-1.0 μg/ml Hoechst 33342 prepared in cell culture media for 15-30 min, followed by two media washes. Images were acquired with the Dragonfly High Speed Confocal Platform 505 (Andor) using an iXon EMCCD camera and a Leica HCX PL APO ×40/1.10 W CORR objective or a Leica HC PL APO ×63/1.47NA OIL CORR TIRF objective mounted on a Leica DMi8 inverted microscope equipped with a live-cell environmental chamber (Okolab) at humidified 37°C and 5% CO 2 . Imaging mode was Confocal 25 μm. Hoechst images were acquired with a 200 mW solid state 405 nm laser and 450/ 50 nm BP emission filter. Clover images were acquired with a 150 mW solid state 488 nm laser and 525/50 nm BP emission filter. mRuby2 images were acquired with a 150 mW solid state 561 nm laser and 620/60 nm BP emission filter. iRFP670 images were acquired with a 140 mW solid state 637 nm laser and 700/75 nm BP emission filter. Z-series covering the full nucleus was acquired at 0.27 μm step size for ×40 objective or 0.13 μm step size for ×63 objective. For time-lapse imaging, the Z-series was acquired at 0.32 μm step size for ×40 objective or 0.16-0.2 μm step size for ×63 objective. Two-color images were acquired every 15 s. Three-color images were acquired every 15 or 24 s. Images are a maximum intensity projection of Z-series. Images were processed in Fusion software using ClearView-GPU deconvolution with the Robust (Iterative) algorithm (pre-sharpening 0, 5 iterations, and de-noising filter size 0.1). Linear adjustments in maximum and minimum levels were applied equally across the entire image to display only nuclear spot signals, and not background nuclear fluorescence. Example images of our image processing pipeline are shown in Supplementary Fig. 2.
Image analysis. To measure two targeted genomic loci distance in time-lapse images, Fiji image analysis software was used 39 . Z-series acquired at 0.32 μm step size for ×40 objective, and 0.16-0.2 μm step size for ×63 objective was used. If the nucleus drifted over time, the Correct 3D Drift plugin was used 40 . For segmenting and tracking spots, the TrackMate plugin was used 41 . Blob diameter was set at 2.0 μm. For each channel, threshold was set to include two spots with the maximum intensity in the 3D volume of the nucleus. Simple LAP Tracker used 2 μm linking max distance, 2 μm gap closing max distance, and 2 gap closing max frame gap. Analysis produced "Spots in track statistics" file which was used to run python script to calculate 3D distances between spots and generate 3D tracks. To determine and correct for potential chromatic aberrations, Invitrogen Tetraspeck 0.1 μm Microspheres (ThermoFisher #T7279) were imaged. On the Dragonfly High Speed Confocal Platform 505 (Andor), Z-series were acquired for both the Leica HCX PL APO ×40/1.10 W CORR objective and the Leica HC PL APO ×63/1.47NA OIL CORR TIRF objective. TrackMate was then used to localize the positions of the microspheres. iRFP670 and mRuby2 chromatic aberrations in each X,Y,Z dimension were calculated relative to Clover, and corrected by offsetting the shifts in each X,Y,Z dimension as follows:

Objective
Fluorescent Statistics and reproducibility. No statistical method was used to predetermine sample size. Sample sizes were chosen as customary in the field and were sufficient to obtain statistical significance. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment.