Simple and versatile imaging of genomic loci in live mammalian cells and early pre-implantation embryos using CAS-LiveFISH

Visualizing the 4D genome in live cells is essential for understanding its regulation. Programmable DNA-binding probes, such as fluorescent clustered regularly interspaced short palindromic repeats (CRISPR) and transcription activator-like effector (TALE) proteins have recently emerged as powerful tools for imaging specific genomic loci in live cells. However, many such systems rely on genetically-encoded components, often requiring multiple constructs that each must be separately optimized, thus limiting their use. Here we develop efficient and versatile systems, based on in vitro transcribed single-guide-RNAs (sgRNAs) and fluorescently-tagged recombinant, catalytically-inactivated Cas9 (dCas9) proteins. Controlled cell delivery of pre-assembled dCas9-sgRNA ribonucleoprotein (RNP) complexes enables robust genomic imaging in live cells and in early mouse embryos. We further demonstrate multiplex tagging of up to 3 genes, tracking detailed movements of chromatin segments and imaging spatial relationships between a distal enhancer and a target gene, with nanometer resolution in live cells. This simple and effective approach should facilitate visualizing chromatin dynamics and nuclear architecture in various living systems.

The dynamic genome organization in the nucleus directs many important processes, such as gene expression regulation, epigenetic inheritance, recombination, DNA replication and genome maintenance, in normal physiology and in disease. Fluorescence in-situ hybridization (FISH) methods have been widely used to detect specific DNA sequences, providing a wealth of information on 3D genome architecture. However FISH relies on chemical fixation, precluding tracking real-time dynamics in live cells. Moreover, harsh treatments with high heat and denaturants might compromise the integrity of genomic structures, especially on the kilobase and nanometer scales. To address some of these limitations, fluorescent programmable DNA binding proteins have emerged as a promising tool for visualizing specific genomic loci in live cells. Fluorescent TALE proteins, engineered to recognize specific DNA sequences, such as major satellite and telomere repeats, were used to visualize repetitive DNA sequences in live human and mouse cells as well as in fly and mouse embryos [1][2][3] . Although engineering a separate TALE protein is required for each genomic locus, the use of CRISPR-Cas9 proteins, whose DNA recognition can be programmed by a single-guide RNA, significantly simplified targeting 4,5 . Multi-color approaches based on orthogonal dCas9 proteins 6,7 , or by fusing various RNA protein binding motifs to the gRNA [8][9][10] , have also been demonstrated.
Most of the previous CRISPR imaging applications focused on visualizing repetitive DNA elements, where the signal amplification from multiple bound dCas9 molecules facilitates detection. Although significantly more challenging, dCas9 can also be used to tag non-repetitive DNA elements, when programmed with multiple gRNAs that target a specific locus in tandem 4,11,12 . However, robust tagging for non-repetitive locus imaging can be challenged by low efficiencies. Often the system needs to be engineered to optimize expression levels of fluorescent dCas9 as well as gRNAs. This issue is further exacerbated by the additional complexity of multi-color systems, which use multiple expression constructs, each one needing separate optimization, limiting the broad use of CRISPR for genome imaging applications.
Here we report a simple and versatile CRISPR imaging system based on in vitro transcribed gRNAs and fluorescent recombinant dCas9 proteins. Controlled cell delivery enables straightforward optimization of the system to achieve robust tagging and high signal-to-noise (SNR) imaging. Simultaneous delivery of multiple gRNAs is also simplified, allowing tandem tagging of endogenous loci. Using this approach, CRISPR/Cas9-mediated

Results
Imaging of genomic loci by delivery of in vitro transcribed gRNAs and recombinant fluorescent dCas9. We first test the ability of in vitro transcribed gRNAs to target cell-expressed dCas9-EGFP to specific genomic loci (Fig. 1a). We deliver gRNAs targeting telomeres in live U-2 OS cells using electroporation, together with a plasmid encoding for BFP-LifeAct to assist identification of transfected cells. Telomeres are visible in transfected cells, while a control gRNA (sgGal4) does not result in any nuclear puncta, demonstrating the specificity of the approach (Fig. 1b, Supplementary Fig. 1). As an alternative, in vitro transcribed gRNAs are microinjected in live U-2 OS cells, together with Alexa 647-Benzylguanine, as a nuclear "counter-stain" for SNAP-tagged RNA Pol II. Like electroporation, microinjection of gRNAs results in robust dCas9-EGFP nuclear puncta (Fig. 1c), suggesting efficient tagging of telomeres and illustrating versatility among cell delivery approaches.
Based on the results with in vitro transcribed gRNAs, we next test the ability of pre-assembled fluorescent dCas9-gRNA ribonucleoproteins (RNPs) to tag specific genomic loci (Fig. 1d). We first prepare fluorescent dCas9-SNAPf and dCas9-CLIPf proteins ( Supplementary Fig. 2) and validate their in vitro DNA binding activity and specificity when assembled with in vitro transcribed gRNAs into RNPs ( Supplementary Fig. 3). We next test the preassembled RNPs in live cells. Delivery of Alexa 647-labelled RNPs targeting telomeres into U-2 OS cells by microinjection results in numerous nuclear puncta (Fig. 1e), similar to that achieved by telomere tagging with cell-expressed dCas9-EGFP. Co-delivery of two types of RNPs into live U-2 OS cells, TMR-labelled RNPs targeting telomeres and Alexa 647-labeled RNPs targeting α-satellite repeats, results in distinct, non-overlapping nuclear puncta (Fig. 1f). To further test the specificity of tagging, we assemble Atto 488-or TMR-labelled RNPs targeting telomeres, and non-targeting (sgGal4) Alexa 647-labelled RNPs as a control. Co-delivery into live HeLa cells results in robust telomere tagging, visible in the Atto 488 or TMR channel, while no nuclear puncta are observable in the Alexa 647 channel (Fig. 1g). Together, these results are consistent with in vitro competition experiments that show minimum cross-talk ( Supplementary Fig. 4a,b) and illustrate specificity in two-color genome tagging as well as versatility among cell types.
CAS-LiveFISH imaging of genomic loci in live pre-implantation embryos. Although numerous demonstrations of CRISPR-based genomic imaging have been reported in cultured cells, applications in living organisms have been thus far limited. We reason that the versatility of the pre-assembled fluorescent RNP system could also be used to tag genomic loci in live mammalian pre-implantation embryos. We thus microinject preassembled TMR-labelled dCas9 RNPs targeting telomeres, into 2-cell stage mouse embryos (Fig. 2a). Nuclear TMR puncta can be discerned (Fig. 2b), while a co-injected Alexa 647-Gal4 RNPs show only diffuse nuclear signal ( Supplementary Fig. 5). TMR-tagged telomeres can be also observed at the 4-cell stage, ~ 1 day after microinjection (Fig. 2c), after the second cleavage division, indicating that fluorescent RNPs are stable enough in vivo to potentially track dynamics occurring over ~ 1 day time-scales in the early mouse embryo. These results show successful proof-of-principle use of the CAS-LiveFISH system for imaging genomic loci in live mammalian embryos.
Pools of pre-assembled RNPs enable visualizing the HPV-18 integration site in live HeLa cells. HeLa cells contain a genomic integration site of Human Papilloma Virus 18 (HPV-18) near the MYC oncogene (Fig. 3a). HPV integration plays a prominent role in the initiation of cervical cancers 13 and in HeLa cells the HPV integration site creates a powerful transcriptional enhancer that drives MYC activation 14 . The HeLa HPV-18 integration site contains parts of the viral genome, as well as local integration-associated rearrangements of the host genome, with host and viral sequences amplified to copy numbers of ~ 4-30 14 .
We design 36 distinct gRNAs targeting the HPV integration site and assemble them with Alexa 647 dCas9 into fluorescent RNPs. Delivery of this HPV-targeting pool of RNPs into live HeLa cells typically results in ~ 2-4 nuclear puncta per cell (Fig. 3b,c). Occasionally we also resolve doublets in close proximity, suggestive of replicated loci in S/G2 4 . To verify the specificity of dCas9 tagging, we perform FISH with a probe labelling the extended MYC locus. Two-color imaging shows that the HPV-Alexa 647 puncta are within ~ 200-300 nm from the centroid of the MYC FISH puncta ( Supplementary Fig. 6a,b). These results demonstrate successful locusspecific tagging of the low-repeat containing HPV integration site, using pools of pre-assembled RNPs.
Multiplex CAS-LiveFISH imaging of up to 3 genes. We reason that the minimal crosstalk between separately assembled two-color RNPs should also be extendable to higher multiplexing. We thus assemble Atto 488-and TMR-labeled dCas9 RNPs programmed with gRNAs targeting the MUC1 and MUC4 genes, respectively, and co-deliver them together with Alexa 647-HPV RNPs in live HeLa cells ( Supplementary Fig. 7). 3-color imaging reveals robust and distinct puncta for Atto 488, TMR and Alexa 647, indicating efficient and specific simultaneous tagging of 3 separate genes. Higher multiplexing should also be achievable by adding more spectral channels and/or by spectral unmixing techniques.

CAS-LiveFISH enables imaging spatial relations between HPV integration site and MYC in live HeLa cells. Proximity-ligation assays, such as Hi-C and ChIA-PET, suggested that activation of MYC in
HeLa cells by the transcriptional enhancer created at the HPV integration site involves long-range genomic interactions (Fig. 3a). However, the physical distances between enhancer and target gene at MYC have not been measured in live cells. To leverage CAS-LiveFISH for imaging spatial relationships between two genomic loci, we first create a pool of 288 in vitro transcribed gRNAs targeting ~ 40 kb around the MYC gene and assemble them with TMR-labelled dCas9 into fluorescent RNPs. Co-delivery of TMR-MYC and Alexa 647-HPV RNPs in live HeLa cells results in Alexa 647 puncta that often co-localize with TMR puncta (Fig. 3d). As a control, co-delivery of TMR-Gal4 and Alexa 647-HPV RNPs results in Alexa 647 puncta but no discernible TMR puncta (Supplementary Fig. 8), further indicating specific tagging of the two genomic loci with minimal crosstalk.  Fig. 9a-c). We further validate the precision of distance measurements in live cells using dual TMR-HPV and Alexa 647-HPV tagging. The two-color HPV-HPV distances have a tight distribution with an average 2D distance of ~ 60 nm, consistent with σ xy ~ 39 nm 2D localization errors in each color (σ x,y ~ 27 nm each in x, y) and 32 nm two-color 2D registration errors (Supplementary Fig. 10a,b). Using this two-color nanometer co-localization procedure, we measure the relative distances between TMR-MYC and Alexa 647-HPV nuclear puncta. We observe a range of 2D MYC-HPV distances, with the majority (~ 75%) being within 200 nm proximity, while a smaller sub-population (~ 10%) is further separated, at distances of 300 nm-1 µm (Fig. 3e). These results illustrate how two-color tagging of genomic loci can provide insights into spatial relations of cis-elements and target genes, with nanometer resolution in live cells.
Pre-assembled RNPs enable tracking the real-time motion of single gene loci. Understanding the movements of genomic loci in the nucleus is important for elucidating key processes such as gene regulation, DNA translocations and recombination. We use dCas9 RNP tagging and analyze the movement of the MUC4 gene as well as the HPV integration site in live HeLa cells (Fig. 4a, Supplementary Video 1). The meansquared-displacement (MSD) vs. time of both MUC4 and HPV follow a scaling behaviour, MSD ~ Dt α , with α < 1 (Fig. 4b), indicating sub-diffusive motion, typical for tagged genomic loci. For MUC4 α ≈ 0.5, in close agreement with previously observed α ≈ 0.5 for other endogenous genes and cis-elements 11,15 . Scaling exponent α = 0.5 characterizes the movement of individual monomers in a polymer chain (Rouse model) 16 . Interestingly, the motion of HPV appears strongly sub-difussive, with α ≈ 0.35, significantly deviating from the free polymerlike physical picture. This confined motion could reflect spatial constraints such as local compartmentalization, highly cross-linked chromatin chains or other possible mechanisms 16,17 . We also analyzed the velocity autocorrelation function (ACF) for the MUC4 and HPV traces 18 . The velocity ACFs for both loci exhibit negative dips (Fig. 4c), indicating that the movement dynamics might be dominated by restoring forces often occurring inside a viscoelastic environment 16,18,19 . Both loci exhibit self-similar motions, as the velocity ACFs collapse upon timeaxis rescaling. The HPV velocity ACF exhibits a more pronounced negative dip than MUC4, consistent with the more strongly confined nature of the HPV motion vs. MUC4. Taken together, these results demonstrate how pre-assembled fluorescent RNP tagging can provide quantitative insights into the detailed movements of specific genomic loci in the nucleus of live cells.

Discussion
Here we demonstrate CAS-LiveFISH, an approach for visualizing specific genomic loci in live mammalian cells and early embryos. CAS-LiveFISH relies on cell delivery of pre-assembled fluorescent dCas9-gRNA RNPs, bypassing the need for extensive genetic manipulations and optimization of multiple expression constructs. Previous works had begun to recognize the versatility of pre-assembled RNPs for genomic imaging. A variant of FISH (CAS-FISH), based on fluorescent dCas9 RNPs instead of DNA hybridization probes 20 , showed rapid, cost-effective and convenient multi-color labelling of genomic loci, in fixed cells. More recently, use of dCas9 RNPs with fluorescent CRISPR RNAs (crRNAs) achieved high SNR detection of highly-repetitive genomic loci in cell lines, as well as in primary human cells 21 . The latter study also illustrated the potential of RNP-based approaches for medical diagnostics and for genomic imaging in systems that are not readily amenable to genetic manipulations. Here we significantly expand these previous works by demonstrating robust and efficient imaging of specific low-repeat containing and non-repetitive genomic loci, in live cells. dCas9 RNPs with fluorescent crRNAs could potentially achieve higher detection SNR, however use of fluorescent dCas9 and non-fluorescent in vitro transcribed gRNAs is significantly more cost-effective than pools of 10-100 s of individually chemically synthesized and purified fluorescent crRNAs. Future developments of low-cost methods to prepare in vitro transcribed pools of fluorescent gRNAs, such as using dye-binding aptamers 22,23 , could enable increased detection SNR, due to the rapid degradation of background gRNAs that are not "protected" in a target DNA-bound dCas9:RNA:DNA ternary complex. Finally, combining our DNA targeting RNP system with RNA targeting CRISPR proteins, such as dCas13 21,24 , would facilitate simultaneous imaging of cis-elements, target genes and nascent transcripts, as well as various other chromatin associated RNAs. Such direct visualization approaches would transform our ability to probe the links between genome organization and gene expression regulation as well as the nuclear organization, dynamics and function of non-coding regulatory RNAs.
CAS-LiveFISH by delivery of pre-assembled dCas9 RNPs can also be extended to imaging genomic loci in live early mouse embryos. The fluorescent RNPs are stable enough to potentially image over ~ 1 day-long timescales. With emerging microscopy techniques to image in thick multi-cellular samples with increased sensitivity and resolution [25][26][27] , imaging fainter signals, down to single non-repetitive genomic loci, and possibly together with simultaneous and coincidental single locus activity readouts and detection of regulatory molecules should also be within reach. These new capabilities could facilitate visualizing several important genomic processes at specific gene loci, in the context of early mammalian development.
A long-standing "mystery" of genome biology is how distal enhancers communicate with target genes to activate transcription 28 . Mechanisms involving direct physical interactions, such as long-range chromatin looping, are thought to facilitate this process, but it has been very challenging to directly measure the spatial relationships between distal enhancers and target genes in live cells. Progress has been made by introducing artificial exogenous tags, such as arrays of target sites for bacterial DNA-binding proteins 29,30 . CAS-LiveFISH enables directly tagging and imaging the endogenous genomic sequences, without needing genomic integrations of exogenous tags and possibly better reflecting the native chromatin organization. Our observations in live HeLa cells reveal that the distal enhancer created at the HPV integration site is often in < 200 nm proximity to the target MYC gene. An emerging view of distal gene regulation mechanisms proposes that approximate enhancer-promoter proximity 28,30 , rather that direct molecular contact, might be adequate for transcription activation. High-local concentrations of Pol II regulatory factors (RFs) created around the enhancer might thus facilitate such actionat-a-distance. Interestingly, the physical distances are within the range (~ 200 nm) over which Pol II RFs cluster in the vicinity of pluripotency genes like Pou5f1 (aka Oct4) and Nanog in embryonic stem cells 15 . CAS-LiveFISH, combined with molecular imaging of Pol II RFs could further elucidate the physical nature and underlying genome topologies behind these important regulatory macromolecular assemblies.
Finally, the observations of highly sub-diffusive motion of the HPV integration site in HeLa cells are particularly striking. Previous work had also showed strongly sub-diffusive motions for Tetracycline operator arrays integrated in immunoglobulin gene segments in B-lymphocytes 17 . It was further postulated that such highly constrained motion might reflect a network of cross-linked chromatin, close to the boundary of a sol-gel phase transition. Intriguingly, such a picture might also apply to highly-active clusters of transcriptional enhancers, bound by high densities of transcription factors and various co-activator proteins 28,31 . The HPV-16 integration site in a related cervical neoplasia cell line colocalizes with a prominent nuclear focus that contains large amounts of Brd4, Mediator and H3K27 acetylated chromatin 32 . Further extensions to simultaneously visualize the HPV integration site in HeLa cells, together with RNA Polymerase II and various Pol II regulatory factors, as recently demonstrated for endogenous genes in mouse embryonic stem cells 15 , will be critical for testing various models for the physical organization of this important class of transcription regulatory units and their links to activation of important oncogenes like MYC.

Preparation of In vitro transcribed gRNAs.
A synthetic DNA fragment (IDT Gblocks) containing the full-length sequence of an optimized SP sgRNA 4 was inserted into the BamHI and EcoRI sites of a modified pLVX-shRNA2 vector (Clonetech) in which ZsGreen1 had been replaced by BFP. DNA templates for T7 transcription were prepared by PCR reactions using a common reverse primer and unique forward primers containing the T7 promoter and the seed sequence. The sgRNAs were in vitro synthesized with HiScribe T7 High Yield RNA Synthesis Kit (NEB). Transcription reactions were treated with RNase-free DNaseI (NEB) to digest the DNA template at 37 °C for 1 h. The sgRNAs were purified by phenol:chloroform: isoamyl alcohol (25:24:1) extraction and ethanol precipitation. The sgRNA pellets were dissolved in 20 mM HEPES (pH 7.5), 150 mM KCl, 10% glycerol and 1 mM TCEP. To refold purified sgRNAs, the sgRNAs were incubated at 70 °C for 5 min and slowly cooled down to room temperature. MgCl 2 was then added to 1 mM final concentration and the sgRNA samples were incubated at 50 °C for 5 min and slowly cooled down to room temperature. The sgRNAs concentration was estimated based on absorption measurements with a NanoDrop 2000 sprectrophotometer (Thermo Scientific) and the samples were stored at − 80 °C. The sequences of nucleic acids used in this study can be found in Supplementary Tables 1-3.
Binding and electrophoretic mobility shift assays. Fluorescent dCas9-gRNA RNP complexes were assembled by incubating fluorescent dCas9 protein with a ~ 30-fold molar excess of sgRNA for 10 min at 37 °C. DNA binding reactions contained 5 nM Cy3-or Cy5-labelled DNA duplex and increasing concentrations of dCas9-sgRNA RNPs. In competition assays, unlabelled DNA duplex was added at 5 nM. DNA binding reactions were incubated for 15 min at 37 °C. dCas9-sgRNA:DNA tertiary complex formation was analyzed by 4% polyacrylamide native gel electrophoresis in 0.5× TBE buffer. Fluorescent bands in the gels were visualized using a fluorescent scanner (Typhoon; GE).
Mouse strains, husbandry and embryo recovery. Animal work was approved by MSKCC Institutional Animal Care and Use Committee (IACUC) and all methods were performed in accordance with the relevant IACUC guidelines and regulations. Animals were housed in a pathogen-free facility under a 12 h light/dark cycle. Mouse strain CD1 (Charles River) was used in this study. Embryos were obtained from mated females 6-12 weeks old. Embryo sex was not determined. To recover recently fertilized embryos, the infunibulum was flushed with FHM media (Sigma-Aldrich) and recovered embryos were washed and placed in FHM media for microinjection.

Delivery of sgRNAs and dCas9-gRNA RNPs into live cells and embryos. U-2 OS cells sta-
bly expressing spdCas9-GFP were nuclofected with Amaxa Cell Line Nucleofector Kit V (Lonza). Cells were detached with 0.05% trypsin/EDTA and spun down by centrifugation at 1000 rpm for 5 min. 1 × 10 6 cells were resuspended in 100 μl room-temperature Nucleofector Solution and mixed with sgRNA and a LifeAct-BFP expressing plasmid. Each sample was transferred into a cuvette and nucleofected using Nucleofector Program X-001. 500 μl of pre-equilibrated culture medium was then added to the cuvette and the cells were gently transferred into an 8-chamber coverglass (LabTek, 155411 Live-cell and live-embryo imaging. Imaging was performed with a home-built microscope setup 15 , featuring a 60 × 1.49 NA objective lens (Nikon MRD01691), 405 nm (Thorlabs diode), 488 nm (Coherent Sapphire HP 500 mW), 532 nm (Coherent Verdi G2 2000 mW) and 640 nm (Coherent Cube 100 mW) excitation lasers in epifluorescence configuration, a piezoelectric 3D nanopositioning stage (Physik Instrumente P-517.3CD and E-710.3CD controller) a quad-view 4-color imaging device (Photometrics) and an EM-CCD camera (Ixon3 897; Andor). Final magnification is 160 nm/pxl. Simultaneous two-or three-color imaging was performed on separate quadrants of the EM-CCD using the quad-view device.The quad-view was equipped with dichroics beam splitters (T495LPXRU, T560LPXR, and T645LPXR; Chroma) and emission filters (ET450/50 m, ET525/50 m, ET605/52 m, and ET700/50 m; Chroma) for separating and imaging the different colours. Imaging was performed at 3 frames/sec. Fine focusing was achieved by translating the sample in z with a nanopositioning stage (P-517.3CD, with E-710.3CD controller; Physik Instrumente). A home-built stage incubator and a separate objective lens heater maintained the sample at 37 °C. The stage incubator atmosphere was adjusted with independent O 2 , N 2 and CO 2 mass-flow controllers (Omega). For some of the embryo imaging experiments, to achieve a larger field-of-view fitting whole embryos, we used a second epifluorescence microscope (Zeiss, Axiovert 200), equipped with an oil-immersion objective lens (Zeiss, Plan-apochromat 63 × 1.4 NA), a mercury lamp illuminator (VSG HBO100/001-26E), filter sets for DAPI (used for BFP), GFP, TMR (used for RFP) and Cy5 (used for SiR), and a scientific CCD (Hamamatsu, ORCA Flash part #C47428012AG).
Image and data analysis, and two-color nanometer distance measurements. Image analysis was performed in IDL (ITT Visual Information Solution, version 8.0.1). First, nuclear puncta were identified using as local maxima in band-pass filtered and background-subtracted images 33 . Then, we performed nanometer localization analysis for each identified punctum: to obtain xy coordinates for the image center of mass, with precision higher than the optical resolution, the processed images were fitted to a 2D elliptical Gaussian peak function using non-linear least-squares fitting 34,35 . For calibrating the registration between xy coordinates in the Alexa 647 and TMR images with nanometer accuracy 35,36 , we used the coordinates of 100 nm Tetraspek beads (ThermoFisher, T7284) scattered throughout the field of view. Two-color registration was performed using a second-order polynomial spatial warping transformation 37,38 . MSD and velocity ACFs were calculated with a MATLAB routine (Matworks, version 2010b). Observed MSDs were corrected by subtracting a constant, MSD loc_error , to account for the effect of localization errors 39 . MSD loc_error was estimated by tracking the relative motion of a single genomic locus, simultaneously tagged with dCas9-Alexa 647 and dCas9-TMR 17 . Graphing, linear least-squares fitting and statistical analysis was performed using Origin (OriginLab, version 8.5.0).