Genome Oligopaint via Local Denaturation Fluorescence in Situ Hybridization (GOLD FISH)

Summary

Cas9 in complex with a programmable guide RNA targets specific double-stranded DNA for cleavage. By harnessing Cas9 as a programmable loader of superhelicase to genomic DNA, here we report a physiological-temperature DNA FISH method termed Genome Oligopaint via Local Denaturation Fluorescence in Situ Hybridization (GOLD FISH). Instead of global denaturation as in conventional DNA FISH, loading a superhelicase at a Cas9-generated nick allows for local DNA denaturation, reducing non-specific binding of probes and avoiding harsh treatments such as heat denaturation. GOLD FISH relies on Cas9 cleaving target DNA sequences and avoids the high nuclear background associated with other genome labeling methods that rely on Cas9 binding. The excellent signal brightness and specificity enable us to image non-repetitive genomic DNA loci and analyze the conformational differences between active and inactive X chromosomes. Finally, GOLD FISH could be used for rapid identification of HER2 gene amplification in a patient tissue.

Graphical Abstract

eTOC blurb

Wang et al. develop Genome Oligopaint via Local Denaturation Fluorescence in Situ Hybridization (GOLD FISH), a DNA FISH method utilizing Cas9 and superhelicase to locally denature target DNA for probe binding. GOLD FISH avoids harsh conditions such as heat denaturation and robustly detects non-repetitive DNA loci with high signal-to-background ratio.

Introduction

The CRISPR-Cas9 system from Streptococcus pyogenes has been widely used for genome editing in cells (Cong et al., 2013; Doudna and Charpentier, 2014; Mali et al., 2013b). In the CRISPR-Cas9 system, the Cas9 endonuclease can be programed with a guide RNA to target a desired DNA sequence (Gasiunas et al., 2012; Jinek et al., 2012; Sapranauskas et al., 2011). An on-target DNA substrate of Cas9 ribonucleoprotein (RNP) contains a 20-nucleotide (nt) protospacer region complementary to the spacer sequence of guide RNA, and a protospacer adjacent motif (PAM, 5’-NGG-3’ for Streptococcus pyogenes Cas9; N representing any nucleotide) (Jinek et al., 2012). After the Cas9 RNP binds to an on-target DNA substrate, the target strand (TS) and non-target strand (NTS) of the DNA substrate are cleaved by HNH nuclease domain and RuvC nuclease domain, respectively (Gasiunas et al., 2012; Jinek et al., 2012; Sapranauskas et al., 2011). After catalysis in vitro, Cas9 remains stably bound to the cleaved DNA substrate (Singh et al., 2016; Sternberg et al., 2014). Engineering the active sites of the nuclease domains creates the Cas9_dHNH variant (Cas9 with H840A mutation) that cuts only the NTS, and the dCas9 variant (Cas9 with H840A and D10A mutations) that is inactive for DNA cleavage (Jinek et al., 2012). CRISPR-mediated transcriptional activation and repression platforms (CRISPRa and CRISPRi) were developed utilizing dCas9 (Gilbert et al., 2013; Maeder et al., 2013; Mali et al., 2013a; Perez-Pinera et al., 2013; Qi et al., 2013).

DNA fluorescence in situ hybridization (DNA FISH) allows for direct visualization of specific DNA sequences in situ, making it a powerful tool to study chromatin conformation and gene localization (Beliveau et al., 2012; Boettiger et al., 2016; Levsky and Singer, 2003; Wang et al., 2016). Conventional DNA FISH requires harsh conditions such as high temperature and concentrated formamide to globally denature genomic DNA for probe hybridization, which risk disrupting the integrity of biological structures such as heat-labile epitopes of proteins and increase the likelihood of DNA FISH probes binding to off-target genomic DNA sequences that are exposed due to global denaturing. By exploiting the high binding affinity of dCas9 RNP to specific DNA sequences, fluorescently labeled dCas9 RNP has been adopted for genomic loci imaging in live cells or in fixed cells without global genomic DNA denaturation (Chen et al., 2013; Chen et al., 2016a; Deng et al., 2015; Hong et al., 2018; Ma et al., 2018; Neguembor et al., 2018; Qin et al., 2017; Wang et al., 2019). Visualization of non-repetitive genomic sequences has been achieved with dCas9-binding-based genomic imaging methods (Chen et al., 2013; Deng et al., 2015; Hong et al., 2018; Mao et al., 2019; Qin et al., 2017; Shao et al., 2018), but the signal-to-background ratio was compromised by non-specific binding of dCas9 RNP to off-targets in genomic DNA (Knight Spencer et al., 2017; Lakadamyali and Cosma, 2020; Wu et al., 2019). Cas9 RNP can tolerate up to eleven PAM-distal mismatches on the DNA substrate for stable binding in vitro (Singh et al., 2016; Sternberg et al., 2014). However, more than three PAM-distal mismatches drastically reduce or inhibit DNA cleavage activities of Cas9 RNP (Sternberg et al., 2015), indicating that the cleavage specificity of Cas9 RNP is much higher than its specificity for stable binding. Conformational activation of Cas9 is dependent of the base-pairing between guide RNA and target DNA, but it is independent of whether the nuclease domains are engineered to be catalytically dead or not (Anders et al., 2014; Dagdas et al., 2017; Huai et al., 2017; Jiang et al., 2016; Nishimasu et al., 2014; Sternberg et al., 2015). We therefore expect that Cas9 nickase variants such as Cas9_dHNH should have similar cleavage specificity as Cas9, and a genomic imaging method that relies on Cas9 or Cas9 nickase variants cleaving target DNA would have higher labeling specificity than Cas9-binding-based genomic imaging methods.

In this study, we show that the post-cleavage Cas9_dHNH-RNA-DNA ternary complex can recruit a 3’ to 5’ DNA helicase to unwind double-stranded DNA (dsDNA) beyond the protospacer. Exploiting this observation, we demonstrate a physiological-temperature DNA FISH method that leverages the high cleavage specificity of Cas9_dHNH to label target genomic DNA.

Design

Previous cell-based studies suggested that the NTS in Cas9-RNA-DNA complex is exposed and available for annealing to an exogenous DNA strand (Richardson et al., 2016). Cleavage on the NTS reveals ~17-nt of single-stranded DNA (ssDNA) with a 3’ hydroxyl end, the NTS 3’ flap (Figure 1A) (Jinek et al., 2012). A recent single-molecule study showed that the NTS 3’ flap can be digested by a ssDNA-specific exonuclease, suggesting the 3’ hydroxyl end may be exposed to solvent (Wang et al., 2020). Rep-X is a highly processive 3’ to 5’ DNA helicase engineered from E. coli Rep helicase through conformational control (Arslan et al., 2015; Hua et al., 2018) based on mechanistic understanding of its activity regulation (Cheng et al., 2001; Comstock et al., 2015; Korolev et al., 1997; Lee et al., 2013). The ssDNA translocating and dsDNA unwinding activities of Rep-X are powered by ATP hydrolysis. We hypothesized that Rep-X can be loaded onto the NTS 3’ flap, translocate along the NTS, and unwind the dsDNA downstream of the protospacer (Figure 1A). If the hypothesis is true, Cas9 RNP can function as a programmable loader of Rep-X to genomic DNA and the loaded Rep-X can unwind the downstream genomic DNA until it encounters an insurmountable blockade (Figure 1B). If Rep-X loaded onto a cleaved NTS unwinds a long enough stretch of genomic DNA, the resulting ssDNA could be targeted by fluorescently labeled oligonucleotide probes for site specific imaging of genomic DNA in the cell (Figure 1B). In this scheme that we call Genome Oligopaint via Local Denaturation Fluorescence in Situ Hybridization (GOLD FISH), after cell fixation, we first add the non-target strand nickase (Cas9_dHNH RNP) to bind and cleave specific DNA sequences in the cells. Next, we add Rep-X and ATP to unwind the dsDNA downstream of each Cas9_dHNH cleavage site and expose the FISH target sequences (FISH-TS, Figure 1B). ATP is then removed. We surmised that the NTS that Rep-X translocates along may be removed from chromatin by Rep-X, for example if it hits the another nick generated by Cas9_dHNH nearby, or form secondary structures, or remain bound by Rep-X, preventing reannealing of the unwound genomic DNA (Figure 1B). Finally, the Cy5-labeled FISH probes are added to hybridize with complementary FISH-TS sequences.

Figure 1. Cas9 exposes the NTS 3’ flap for Rep-X loading and unwinding the downstream dsDNA.

(A) Schematic of Rep-X loaded on the NTS 3’ flap and translocating along the DNA strand. The NTS 3’ flap is outlined in a dash box.

(B) Schematic of GOLD FISH. The unwound DNA may not rezip behind the translocating helicase for three possible reasons indicated.

(C) Schematic for the DNA helicase invasion assay. A nick is indicated in the figure at 20 bp downstream of the protospacer.

(D) Representative images at different time points during the DNA helicase invasion assay. Note that the images were taken at different locations on the slide surface. Scale bar, 5 μm.

(E) Spot number per imaging view decreased with time in the DNA helicase invasion assay when using Cas9_dHNH.

(F) Spot number per imaging view did not change with time with dCas9 or in the absence of ATP. Error bar represent mean ± SD (n = 20). n represents number of different imaging views. The experiments were repeated twice independently with similar results.

Results

Cas9_dHNH exposes cleaved NTS for Rep-X superhelicase loading

To test if Rep-X could be loaded onto the Cas9_dHNH-generated NTS 3’ flap and unwind the dsDNA beyond the protospacer at the single-molecule level, we developed a DNA helicase invasion assay (Figure 1C). In this assay, the Cas9_dHNH-RNA-DNA ternary complex was assembled in Mg²⁺-containing imaging buffer and immobilized on a quartz slide through biotin-NeutrAvidin interaction (Figure 1C). We internally labeled the PAM-distal region of the NTS with Cy5 so that if the loaded Rep-X fully unwound the 20-bp DNA downstream of the protospacer, fluorescence signal of Cy5 would be lost from the surface-tethered DNA (Figure 1C). The internal Cy5-labeling on the NTS does not affect the DNA cleavage activity of Cas9 (Singh et al., 2018). When Rep-X and ATP were added together, the average number of fluorescent spots per image area decreased over time (Figures 1D and 1E), indicating that the Cy5-labeled DNA strand downstream of the NTS cleavage site was unwound by Rep-X and lost from the surface (Figure 1C). Control experiments using dCas9 or without ATP did not show Cy5 spots decrease with time (Figure 1F), indicating that both DNA nicking by Cas9_dHNH and ATP hydrolysis-driven DNA unwinding by Rep-X are necessary to remove the downstream NTS DNA. Together, our data suggest that Rep-X can invade into the Cas9_dHNH-RNA-DNA complex through the cleaved NTS and unwind the downstream dsDNA, supporting the design principle of GOLD FISH.

GOLD FISH allows efficient labeling of a repetitive region within the MUC4 gene at physiological temperature

We first tested GOLD FISH on a repetitive region within the MUC4 gene (MUC4-R) in IMR-90 cells, a human female diploid fibroblast strain (Nichols et al., 1977), so that a single guide RNA and a single Cy5-labeled FISH probe could be used to decorate the gene with multiple fluorophores. The GOLD FISH images of MUC4-R obtained using epifluorescence microscopy showed that 93% of cells contained 2 to 4 bright foci with low background (Figures 2A and 2B). Percentages of cells showing a particular number of foci are indicated above the corresponding histogram bin). We referred previously measured percentages of IMR-90 cells in G0/G1 (should have 2 foci) and S/G2/M (should have 4 foci) (Shi et al., 2007), and estimated the efficiency of detection to be 90%. Control experiments performed using dCas9 or without ATP showed weak or no detectable foci (Figures S1A and S1B), indicating that both DNA nicking by Cas9_dHNH and ATP hydrolysis-driven DNA unwinding by Rep-X are necessary to obtain bright GOLD FISH signals.

Figure 2. GOLD FISH targeting a repetitive region within the MUC4 gene (MUC4-R).

(A) A representative image of GOLD FISH against MUC4-R in IMR-90 cells. A single cell outlined in green is magnified on the upper–right corner. Probes were Cy5-labeled, and nuclei were stained using Hoechst 33342. Scale bar, 10 μm.

(B) (Left panel) Histogram of number of MUC4 foci detected in each cell (n = 78). Percentages of total cells are indicated. (Right panel) Box plot of signal-to-background ratio of detected FISH spots. Mean ± SD are represented using a black line and box. Each dot represents one FISH spot (n = 163).

(C) Schematic for GOLD FISH using ATTO550-labeled guide RNA. ATTO550 was conjugated at the 5’ end of tracrRNA.

(D) Quantification of co-localized loci from ATTO550-guide RNA and Cy5-GOLD FISH probe. The black numbers indicate spot number examined in each channel.

(E) A representative image of MUC4 fluorescent signals from ATTO550-guide RNA (green) and Cy5-GOLD FISH probes (magenta) in HEK293ft cells. Scale bar, 5 μm.

(F) Comparison of the signal-to-background ratio of ATTO550 foci and Cy5 foci from the co-localization assay, and CASFISH foci using Cas9_dHNH and dCas9. Mean ± SD are represented using a line and box in the box plot. Each dot represents one FISH spot (n = 135, 146, 197 and 154 for ‘ATTO550 foci’, ‘Cy5 foci’, ‘CASFISH foci Cas9_dHNH’ and ‘’CASFISH foci dCas9’ respectively). ***P<0.001 (Student’s t-test).

To test if GOLD FISH works in other cell types and if the FISH probes co-localize with Cas9 binding sites, we used an ATTO550-labeled guide RNA and performed GOLD FISH of MUC4-R in HEK293ft cells (co-localization assay, Figure 2C). We found the guide RNA, and by inference Cas9_dHNH, remained visibly bound to target DNA under our experimental condition, and 90% of Cy5 loci co-localized with ATTO550 loci (Figures 2D and 2E). Next, we access whether GOLD FISH reduces nuclear background arising from non-specific binding of Cas9_dHNH RNP. Of note, it is possible that some on- and off-target ATTO550-labeled Cas9_dHNH RNP dissociated from the cells during the Rep-X unwinding and FISH probe hybridization steps in GOLD FISH (Figure 2E). To fairly compare GOLD FISH signals with signals from labeled Cas9_dHNH RNP binding, we also performed Cas9-mediated fluorescence in situ hybridization (CASFISH) against MUC4-R using either Cas9_dHNH or dCas9 with the ATTO550-labeled guide RNA (Figures S1C and S1D) (Deng et al., 2015). We measured the signal-to-background ratios of the ATTO550 foci and Cy5 foci from the co-localization assay (S/B_{ATTO550-colocalization} and S/B_{Cy5-colocalization}), as well as that of the CASFISH foci (S/B_{CASFISH-Cas9dHNH} and S/B_{CASFISH-dCas9}) (Figure 2F). We found S/B_{Cy5-colocalization} (17.4 ± 6.6, mean ± S.D.) was substantially greater than S/B_{ATTO550-colocalization} (2.7 ± 1.4, mean ± S.D.), S/B_{CASFISH-Cas9dHNH} (1.3 ± 0.8, mean ± S.D.) and S/B_{CASFISH-dCas9} (0.9 ± 0.4, mean ± S.D.). Control experiments performed without Cas9 enzyme showed that the nuclear background arising from cellular autofluorescence and non-specific intracellular binding of ATTO550-labeled guide RNA was negligible (Figure S1E). Therefore, higher non-specific binding of labeled Cas9_dHNH RNP is responsible for the lower signal-to-background ratio of the ATTO550 foci in the co-localization assay and the CASFISH foci (Figure 2F). S/B_{ATTO550-colocalization} being higher than S/B_{CASFISH-Cas9dHNH} also suggests that some non-specifically bound Cas9_dHNH RNP dissociated from the cells during the Rep-X unwinding and FISH probe hybridization steps in GOLD FISH (Figure 2F). Together, these data indicate that GOLD FISH, which requires Cas9 cleavage of target DNA to achieve efficient labeling (Figures 2B and S1A), has much reduced non-specific labeling compared to the genomic imaging methods relying on labeled Cas9 RNP binding alone.

GOLD FISH enables robust labeling of non-repetitive DNA sequences and provides chromatin conformational information

Low nonspecific binding of GOLD FISH should greatly facilitate non-repetitive loci imaging, which is generally much more challenging due to the need to include guide RNAs and FISH probes of multiple sequences at the same time. For example, if m different guide-RNA sequences and n different FISH probes are used, the total concentration of guide RNAs and FISH probes would have to be m and n times higher, respectively, to achieve the same signal level for each probe, potentially increasing background arising from nonspecific probe binding. A previous CASFISH study used 73 different guide RNAs to label a non-repetitive region within the MUC4 gene and observed compromised labeling efficiency and increased background (Deng et al., 2015). In order to test the capability of GOLD FISH in targeting non-repetitive DNA sequences, we designed 9 different guide RNAs (MUC4-NR guide-RNA set 1) targeting a 2.3-kilobases (kb) non-repetitive region within the MUC4 gene (MUC4-NR), with an approximate spacing of 300 base pair (bp) between them, and 57 different Cy5-labeled FISH probes that bind regions between the guide RNAs (Figure 3A, top). Remarkably, GOLD FISH efficiently labeled the MUC4-NR region (Figure 3A). 89% of cells had 2 to 4 FISH loci and the average signal-to-background ratio was 7.8 (Figures 3B and S2, percentages of cells with specific number of foci are shown above the histogram in Figure 3B). The specificity of MUC4-NR FISH loci was verified by colocalization with MUC4-R loci (Figures 3A and 3C). The excellent labeling efficiency and signal-to-background ratio of MUC4-NR GOLD FISH confirm that it is capable of non-repetitive loci imaging without high nuclear background.

Figure 3. GOLD FISH targeting a non-repetitive region within the MUC4 gene.

(A) (Top) A schematic showing Cas9 binding sites and probe targeting region for GOLD FISH against MUC4 non-repetitive region (MUC4-NR) using MUC4-NR guide-RNA set 1. (Bottom) A representative image of GOLD FISH against MUC4-NR and MUC4 repetitive region (MUC4-R) in an IMR-90 cell. Scale bar, 5 μm.

(B) (Left panel) Histogram of number of MUC4-NR foci detected in each cell (n = 78). Percentages of total cells are indicated. (Right panel) Box plot of signal-to-background ratio of detected MUC4-NR FISH spots, mean ± SD are represented using a black line and box. Each dot represents a FISH spot. n = 167.

(C) Quantification of co-localized foci from MUC4-NR and MUC4-R GOLD FISH. The black numbers indicate spot number examined in each channel.

(D) (Top) A schematic showing 11 guide RNAs (MUC4-NR guide-RNA set 2) designed to target sites flanking the probes tiling region of MUC4-NR. (Bottom) A representative image of GOLD FISH using the MUC4-NR guide-RNA set 2 in an IMR-90 cell. Scale bar, 5μm.

(E) (Left panel) Histogram of number of MUC4-NR foci detected in each cell using the MUC4-NR guide-RNA set 2 (n = 76). Percentages of total cells are indicated. (Right panel) Box plot of signal-to-background ratio of MUC4-NR FISH spots using the MUC4-NR guide-RNA set 2. Mean ± SD are represented using a black line and box. Each dot represents a FISH spot (n = 90).

(F) (Top) A schematic showing MUC4-I1 guide RNA designed to target a repetitive region that is 30-kb away from the probes tiling region of MUC4-NR. (Bottom) A representative image of GOLD FISH using the MUC4-I1 guide RNA in an IMR-90 cell. Scale bar, 5μm.

(G) Histogram of number of MUC4-NR foci detected in each cell using the MUC4-I1 guide RNA (n = 70). Percentages of total cells are indicated.

Rep-X can unwind thousands of base pairs of dsDNA in vitro (Arslan et al., 2015). To examine whether Rep-X is similarly processive on the genomic DNA, we performed GOLD FISH using the same fluorescently labeled probes targeting the MUC4-NR region but a new set of guide RNAs (MUC4-NR guide-RNA set 2). This set contains 11 different guide RNAs targeting a 2.4-kb region next to the MUC4-NR probe tiling region (Figure 3D, top). Only 39% of cells had ≥ 2 detectable FISH loci, and the detectable loci had 30% lower signal-to-background ratio on average in comparison with using the MUC4-NR guide-RNA set 1 (Figures 3D and 3E). Next, we designed another set of guide RNAs (MUC4-I1) targeting a 2.5-kb repetitive region (~50 copies) that is 30 kb away from the MUC4-NR probe tiling region (Figure 3F, top). We found 91% cells did not have detectable focus (Figures 3F and 3G). Although it is possible that the MUC4-NR guide-RNA set 1, MUC4-NR guide-RNA set 2 and MUC4-I1 have different on-target activities which could lead to variability in labeling efficiency, the reduced labeling efficiency when unwinding initiates a few kb away (Figures 3D and 3F) is more likely because that Rep-X failed to efficiently unwind the chromatin from the Cas9 cleavage sites to the probe tiling region. We presume this is because the crowded nuclear environment and presence of nucleosomes inside cells prevent even a superhelicase from unwinding a very long stretch of chromatin. Therefore, we conclude that GOLD FISH locally denatures the targeted chromatin for FISH probe hybridization.

Next, we investigated if GOLD FISH can be used to quantitatively assess chromatin conformations. CASFISH and CRISPR/Cas9-mediated proximity ligation assay (CasPLA) are previously reported Cas9-mediated genomic imaging methods that are capable of labeling nonrepetitive loci in fixed cells (Deng et al., 2015; Zhang et al., 2018). The two methods use a solution of methanol and acetic acid (MAA) as the cell fixative. GOLD FISH experiments described above were also performed in MAA-fixed cells. However, it is known that fixation with methanol may cause a nuclear shrinkage (Boettiger et al., 2016). We found that, although the MAA fixation largely preserved the nuclear morphology, the projected nuclear area is reduced by 10% (Figures S3A and S3B). To overcome this issue, we combined a previously developed buffered ethanol (BE70) with MAA fixation (BE70-MAA) (Perry et al., 2016): the cells were fixed in BE70, and further permeabilized in MAA. We found that the BE70-MAA fixation method preserved the nucleus with a less than 1.2% reduction in projected nuclear area (Figures S3A and S3B). Next, we measured the spatial distance between two genomic regions by two-color GOLD FISH. Of the 40 topologically associated domains (TADs) previously identified in chromosome X (ChrX) of IMR-90 cells (Dixon et al., 2012), we chose two non-repetitive regions located in the 5^th and the 37^th TAD (TAD5 and TAD37, respectively). The genomic distance between TAD5 and TAD37 is 125 megabases (Mb, Figure 4A, top). Two-color GOLD FISH against TAD5 and TAD37 was performed in the BE70-MAA-fixed IMR-90 cells (Figures 4A). We were also able to distinguish inactive chromosome X (Xi) from active chromosome X (Xa) by concurrent MacroH2A.1 immunostaining (Figures 4A and S3C) (Costanzi and Pehrson, 1998). The average three-dimensional distance between TAD5 and TAD37 was 1.5 μm for Xi and 2.9 μm for Xa, close to the previously reported values (Figure 4B) (Wang et al., 2016). Control GOLD FISH experiments using TAD5 guide RNAs with TAD37 probes or TAD37 guide RNAs with TAD5 probes showed that more than 96% cells had no detectable focus (Figures S3D and S3E), further supporting that GOLD FISH denatures genomic DNA locally. These results demonstrate that GOLD FISH can robustly target multiple non-repetitive DNA loci and provide chromatin structural information.

Figure 4. GOLD FISH shows conformational differences of active ChrX and inactive ChrX by multi-color imaging and chromosomal scale paint.

(A) (Top) A schematic of TAD5 and TAD37 regions in chromosome X. (Bottom) A representative image of GOLD FISH against TAD5 (magenta) and TAD37 (green) in an IMR-90 cell. MacroH2A.1 immunostaining (cyan) was used to distinguish inactive ChrX from active ChrX. White arrow indicates the inactive ChrX. Scale bar, 5 μm.

(B) Box plots of distance between TAD5 and TAD37 of ChrX for active ChrX and inactive ChrX. Mean ± SD are represented using a line and box. Each dot represents a ChrX measured (n=83 for active ChrX and 86 for inactive ChrX). ***P<0.001 (Student’s t-test).

(C) DNA probe design for ChrX paint GOLD FISH. The primary probe has two Priming regions for PCR amplification, a Readout region complementary to fluorescently labeled Readout probe and an Encoding region for hybridization to genomic DNA. Cas9 RNP and Rep-X are omitted in this figure.

(D) A representative image of p-arm (green) and q-arm (magenta) of ChrX ‘painted’ by GOLD FISH. MacroH2A.1 immunostaining (cyan) was used to distinguish inactive ChrX from active ChrX. White arrow indicates the inactive ChrX. Scale bar, 5 μm.

(E) Box plots of center of mass distance between p-arm and q-arm of ChrX for active ChrX and inactive ChrX. Each dot represents a ChrX measured (n = 101 for active ChrX and 99 for inactive ChrX). Mean ± SD are represented using a line and box. ***P<0.001 (Student’s t-test).

Chromosomal scale GOLD FISH

We further extended GOLD FISH to the chromosomal scale and imaged the p-arm and q-arm of ChrX in the BE70-MAA-fixed IMR-90 cells. We designed 3,287 guide RNAs and 2,307 FISH probes targeting all 40 TADs of ChrX in the IMR-90 cells. The FISH probes consist of unlabeled primary probes and fluorescently labeled readout probes (Figure 4C). Each primary probe contains an encoding region complementary to genomic DNA, a readout region complementary to a specific readout probe, and two primer regions for amplification of the primary probe library (Figure 4C). The probes against the p-arm and the q-arm of ChrX were labeled with Cy3 and Cy5, respectively (Figure 4C). The GOLD FISH signals of the p-arm and the q-arm were cloud-like (Figures 4D and S4A), and MacroH2A.1 immunostaining was performed to distinguish Xi from Xa (Figures 4D). The average center-of-mass distance between the p-arm and the q-arm for Xa is 28% larger than that for Xi (Figure 4E), and the average volume of Xa is 53% greater than that of Xi (Figure S4B). This is consistent with a previous finding that Xi adopts more compact conformations than Xa (Wang et al., 2016). Our data suggest GOLD FISH is scalable ranging from a single locus as short as 2.3 kb to chromosomal scale ‘paint’.

GOLD FISH allows for rapid identification of HER2 gene amplification in tissue samples

Finally, we show that GOLD FISH is applicable to non-repetitive DNA sequences in pathologic tissue samples. DNA FISH is widely used for diagnosis of molecular pathologies like Human Epidermal Growth Factor Receptor 2 (HER2) gene amplification in breast cancer patients, where the HER2 FISH spot number is compared to an enumeration gene or region of chromosome 17 (e.g. centromere region of chromosome 17 (CEP17)) to calculate the gene amplification state (Figure 5A) (Furrer et al., 2015). Tissue samples fixed by non-crosslinking fixatives have several advantages compared to crosslinking-fixed tissue samples including higher quality and quantity of DNA, RNA and protein extraction (Oberauner-Wappis et al., 2016; Perry et al., 2016). Non-crosslinking fixation also allows faster probe hybridization to sequences of interest (Shaffer et al., 2013). However, the HER2 gene amplification testing in the non-crosslinking-fixed tissue samples requires an 18 to 24 hours crosslinking reaction prior to overnight conventional DNA FISH (Oberauner-Wappis et al., 2016), which extends the experimental procedures to days. To test whether GOLD FISH can rapidly detect non-repetitive sequences in the non-crosslinking-fixed tissue samples, we performed GOLD FISH targeting the HER2 gene and CEP17 in BE70-MAA -fixed human breast cancer tissue sections (10 μm thick), in parallel with immunostaining of HER2 protein. GOLD FISH efficiently labeled target sequences within 6 hours (including fixation time, Figure 5B). By quantifying the numbers of HER2 and CEP17 foci per cell, we found 88% of cells had more than 4 copies of HER2 gene accompanied by high expression level of HER2 protein, while no more than 4 copies of CEP17 foci was observed (Figures 5B-5E, percentages of cells with specific number of foci are indicated above the histograms). The ratio of HER2/CEP17 foci numbers in each cell was 6.1 ± 3.8 (mean ± SD), indicating the HER2 gene amplification in the sample examined (Figure 5F) (Wolff et al., 2013). Notably, although the Retinoic Acid Receptor Alpha (RARA) gene was suggested as a chromosome 17 enumeration gene (Figure 5A) (Tse et al., 2011), we found RARA was co-amplified with HER2 gene (Figures S5A and 5F) (Varga et al., 2012). Therefore, RARA did not faithfully show the copy number of chromosome 17 in this tissue sample. Together, our analysis suggests that GOLD FISH can be directly applied to non-crosslinking fixed tissue samples for rapid DNA detection.

Figure 5. HER2 gene amplification detection in human tissue samples.

(A) A schematic of HER2 gene, CEP17 and RARA gene in chromosome 17 (Chr17).

(B) A representative view of GOLD FISH against HER2 gene (yellow) with HER2 protein immunostaining (red) and DNA staining by Hoechst 33342 (blue) on a breast cancer tissue sample from a patient. Sub-regions outlined in green boxes are zoomed showing HER2 amplified cells and HER2 non-amplified cells, respectively. Scale bar, 10 μm.

(C) A representative view of GOLD FISH against HER2 gene (yellow, left) and CEP17 (green, right) with HER2 protein immunostaining (red) and DNA staining by Hoechst 33342 (blue). Scale bar, 10 μm.

(D) Histograms of number of HER2 foci in each cell of the breast cancer tissue sample (n = 161). Percentages of total cells are indicated.

(E) Histograms of number of CEP17 foci in each cell (n = 161). Percentages of total cells are indicated.

(F) Box plots of loci number ratio of HER2 to CEP17 and HER2 to RARA. Each dot represents one cell. Mean ± SD are represented using a line and box (n = 159 for HER2/CEP17 and n = 95 for HER2/RARA). ***P<0.001 (Student’s t-test).

Discussion

In this study, we repurposed the Cas9_dHNH RNP as a programmable loader of superhelicase to genomic DNA. We showed that Cas9_dHNH cleavage exposes a ssDNA region on the NTS, allowing Rep-X superhelicase to load on and unwind downstream dsDNA. Based on this, we developed GOLD FISH, a superhelicase-mediated physiological-temperature DNA FISH method. GOLD FISH leverages the high specificity of Cas9_dHNH cleavage to trigger targeted genomic DNA denaturing and shows several advantages when compared to other genomic imaging methods.

Current Cas9-mediated genomic imaging methods rely on the binding of directly or indirectly labeled dCas9 RNP to target DNA (Chen et al., 2013; Chen et al., 2016a; Deng et al., 2015; Hong et al., 2018; Ma et al., 2018; Mao et al., 2019; Qin et al., 2017; Shao et al., 2018; Wang et al., 2019). In contrast, GOLD FISH adds DNA cleavage as a prerequisite for efficient labeling (Figure S1A). Cas9’s cleavage specificity is much higher than its stable binding specificity (Singh et al., 2016; Sternberg et al., 2015; Sternberg et al., 2014). This could explain our observation that GOLD FISH shows excellent labeling specificity and avoids high nuclear background even when it targets non-repetitive loci.

Conventional DNA FISH denatures genomic DNA globally by heat and concentrated formamide treatments to enable probe hybridization. In contrast, GOLD FISH locally denatures targeted chromatin under much milder experimental conditions as we demonstrated through several examples. Targeted chromatin denaturing also reduces the likelihood of non-specific binding of FISH probes to the genome. CO-FISH and RASER FISH are DNA FISH methods that do not require heat denaturation, and RASER FISH has been used for super-resolution imaging of chromatin conformations (Brown et al., 2018; Williams and Bailey, 2009). However, CO-FISH and RASER FISH non-specifically and globally digest genomic DNA for probe hybridization, and require an overnight BrdU treatment in live cells prior to cell fixation (Brown et al., 2018; Williams and Bailey, 2009). BrdU may alter DNA stability, transcriptional/translational level, and lengthen the cell cycle (Taupin, 2007). In contrast, GOLD FISH does not require any treatment in live cells before cell fixation and therefore can also be applied to patient tissue samples as we demonstrated using human breast cancer tissue. The mild conditions also allow rapid GOLD FISH on tissue samples fixed by a non-crosslinking fixative. The HER2 GOLD FISH experiment in the 10-μm-thick non-crosslinking-fixed tissue sections took only 6 hours, while conventional HER2 DNA FISH in 2-μm-thick non-crosslinking-fixed tissue sections requires days (Oberauner-Wappis et al., 2016).

We have optimized the pipelines of guide-RNA and FISH probe synthesis to make GOLD FISH easy to implement and cost-effective. The oligonucleotide probes of GOLD FISH for targeting a few kilobases of non-repetitive genomic DNA were synthesized using an enzymatic approach (Gaspar et al., 2017). Oligonucleotides without any labeling or modification were purchased, and desired fluorophores were conjugated to the 3’ end of each oligonucleotide by using terminal deoxynucleotidyl transferase (TdT) (Gaspar et al., 2017). Each set of probes was labeled in a single TdT reaction. Ideally, a guide-RNA set for GOLD FISH targeting non-repetitive DNA sequences should have an equal amount of each guide-RNA species. However, the in vitro synthesis efficiencies of different canonical crRNAs can be dramatically different (Figure S5B), likely because T7 transcription is sensitive to the first two or more nucleotides of the template DNA. This would cause different crRNA species to be present at different concentrations if transcribed together. In this case, different guide-RNA species must be synthesized individually, then combined in equal amounts, which is labor-intensive. To overcome this challenge, we adopted 5’ extended crRNA in most of guide-RNA designs in this study (Kocak et al., 2019). We found different crRNAs with a common 10-nt 5’ extension have similar synthesis efficiencies (Figure S5B). This scheme enables the synthesis of multiple guide RNAs in a single reaction, giving similar synthesis efficiencies (therefore similar concentration) across the pool. Cas9 and Rep-X can be produced in large scales in terms of the amounts needed for GOLD FISH (Arslan et al., 2015; Jinek et al., 2012). A detailed cost estimate of GOLD FISH can be found in ‘Method Details’.

GOLD FISH has less stringent specificity requirements for designing FISH probes. Nonspecific annealing of probes to the rest of the genome is not a major concern because of targeted local denaturing of the genome. In contrast, conventional DNA FISH has stringent requirements to avoid annealing to the globally denatured genome. Therefore, GOLD FISH enables similar or higher probe density compared to the state-of-the-art DNA FISH methods such as OligoMiner and iFISH (Figure S5C) (Beliveau et al., 2018; Gelali et al., 2019). The higher probe density of GOLD FISH enabled efficient detection of a non-repetitive locus as short as 2.3 kb in human genome using epifluorescence microscopy (Figure 3A). We have demonstrated GOLD FISH can ‘paint’ the X chromosome with differently colored fluorophores using the primary probes and readout probes (Figure 4C), which is the scheme originally developed for multiplexed FISH experiments (Chen et al., 2015; Mateo et al., 2019; Wang et al., 2016). GOLD FISH differs from traditional DNA FISH only in the denaturation step, and therefore should be readily extendable to highly multiplexed FISH experiments.

It is intriguing how far Rep-X can unwind from the loading site in the complex and dense nuclear environment. The MUC4-NR data in Figure 3 show that the distance between the guides and the FISH probes impacts the signal. When the distance between each FISH probe and the first ‘upstream’ guide for the probe is less than 400 bp, 89% cells showed 2-4 foci (Figures 3A and 3B). However, when the distance is on average 1.2 kb, only 39% cells showed 2-4 foci (Figures 3D and 3E). When the distance is ~30 kb, 91% cells did not have detectable focus (Figures 3F and 3G). For other tested loci (MUC4-R, TAD5, TAD37, HER2 and RARA) which gave excellent signals, the distance between each FISH probe and the first ‘upstream’ guide of the probe was less than 400 bp. The data suggest that Rep-X should be able to efficiently unwind ~ 400 bp, and larger distance between FISH probes and guides may lead to decreased labeling efficiency. Systematic investigation in the future may reveal the extent of unwinding more precisely and whether it depends on the chromatin environment.

Limitations

GOLD FISH uses Cas9_dHNH RNP to create a 3’ flap for Rep-X loading. Therefore, the target locus should have enough sites that can be cleaved by Cas9 (e.g. nine Cas9 cleavage sites were sufficient for targeting MUC4-NR). The labeling efficiency of GOLD FISH may be compromised if crRNA has very low on-target activity (e.g., crRNA targeting a protospacer with very low or high GC content should be avoided) (Wang et al., 2014). The presence of nucleosomes and epigenetic modifications may also affect the ability of Cas9_dHNH to access and cleave target DNA, therefore influencing the labeling efficiency (Chen et al., 2016b; Horlbeck et al., 2016; Yarrington et al., 2018). crRNA designing tools with on-target activity prediction might be helpful (Cui et al., 2018). Because GOLD FISH uses oligonucleotide probes for hybridization with sequences of interest, targeting sequences that can form complexed structures such as G-quadruplex might lead to decreased labeling efficiency. Repeated sequences should not be problematic as potential target loci as long as there are PAM sequences for Cas9 targeting, as we have demonstrated for the MUC4-R repetitive locus (Figure 2). The ‘difficult’ sequences mentioned above, and other repeated sequences may be tested in future studies to develop a robust guideline for GOLD FISH.

We unambiguously identified the HER2 amplification in the patient tissue sample using GOLD FISH (Figure 5F), but it is possible that we underestimated the HER2 and RARA copy numbers in some cells. For example, two copies of HER2 loci are spatially within the diffraction limit resolution of epifluorescence microscopy cannot be distinguished. Combination of GOLD FISH with super-resolution microscopy would allow more accurate measurements of the gene copy numbers.

GOLD FISH does not require global heat denaturation of genomic DNA, which potentially improves the preservation of chromatin structures. However, crosslinking fixatives are not compatible with GOLD FISH. GOLD FISH of MUC4-NR did not show detectable signals in paraformaldehyde (PFA)-fixed cells (Figure S5D), likely because the PFA crosslinking interfered with Cas9 finding its target DNA and/or because Rep-X cannot translocate/unwind long enough along the genomic DNA in PFA-fixed cells. Therefore, we used two non-crosslinking fixation methods in this work. The first method was MAA fixation (Figures 2 and 3). The second method was BE70-MAA fixation (Figures 4 and 5). We showed that the cells fixed using BE70-MAA had minimal reduction in projected nuclear area (Figures S3A and S3B), and the GOLD FISH-measured spatial distances between TAD5 and TAD37 were close to previously reported values measured using conventional DNA FISH (Figure 4B) (Wang et al., 2016). But it is possible that some ultra-fine chromatin structures may be altered by BE70-MAA fixation. Further electron microscopy- or super-resolution microscopy-based study is required to assess the ultra-fine structure preservation of chromatin in GOLD FISH.

STAR Methods Text

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Taekjip Ha (tjha@jhu.edu).

Materials Availability

This study did not generate new unique reagents.

Data and Code Availability

Raw imaging data can be accessed on Mendeley Data (http://dx.doi.org/10.17632/twj8ft95j5.2). The raw imaging data are single- or multi-frame TIFF files (except for exception is Figure S5B), and the brightness/contrast may need to be adjusted to achieve optimal visualization. The code generated during this study is available at GitHub (https://github.com/ashleefeng/cas9).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human Cell Lines

IMR-90 human female diploid fibroblast cells were purchased from American Type Culture Collection (ATCC, CCL-186) and cultured at 37 °C in 5% CO₂ in EMEM (ATCC, 30-2003) with 1 mM sodium pyruvate and 10% fetal bovine serum (FBS, ThermoFisher). IMR-90 cell line authentication was performed by the vendor. HEK293ft human female cells were a generous gift from the Regot lab (Johns Hopkins University School of Medicine). HEK293ft cell line authentication was not performed. HEK293ft cells were cultured at 37 °C in 5% CO₂ in DMEM (Corning) with 4.5 g/L glucose, L-glutamate, 1 mM sodium pyruvate, 1X antibiotic antimycotic solution (Sigma-Aldrich), and 10% FBS. Imaging dishes were coated with 1 μg/cm² fibronectin for 60 min, then washed with PBS before plating.

Human Tissue Samples

Human breast cancer primary patient tissue was procured from ProteoGenex, which collected the samples with informed consent from the donor and approved by the Institutional Review Board/Independent Ethics Committee (IBR/IEC). The donor was 57 years old, female, with a breast cancer grade of G3. Samples were positive for estrogen receptor, progesterone receptor, and HER2 expression by immunofluorescence. We embedded the tissue in OCT media, froze it, sectioned it to 10 μm (OTF5000 cryostat – Bright Instruments), and adhered it collagen coated 21 mm² glass coverslips for imaging.

METHOD DETAILS

Cas9 expression and purification

The expression plasmid of Cas9_dHNH (i.e. Cas9 (H840A)) was a gift from Jennifer Doudna (Addgene plasmid # 39316; http://n2t.net/addgene:39316; RRID:Addgene 39316). To express and purify the Cas9, the plasmid was transformed into E.coli strain BL21 Rosetta 2 (DE3) (EMD Biosciences). The cells were grown in Terrific Broth (TB) at 37 °C to an optical density at 600nm of 0.6. At this point IPTG was added to a final concentration of 0.5 mM to induce expression. Cells were left at 18°C overnight (12-16 hrs) and harvested the next day. Cells were resuspended in lysis buffer (50 mM Tris, pH 7.5, 500 mM NaCl, 5% (v/v) glycerol and 1 mM TCEP) supplemented with protease inhibitor cocktail (Roche) and with Lysozyme (Sigma Aldrich). Cells were then lysed using an Emulsiflex-C5 homogenizer (Avestin). Insoluble material was pelleted at 18,000 rpm for 30 minutes at 4 °C and soluble lysate was collected and incubated with IMAC nickel affinity resin (Bio-Rad) for 30 minutes at 4 °C with gentle agitation. Resin was collected and washed in a column with 500 ml of lysis buffer. Resin was then incubated with lysis buffer supplemented with 250 mM Imidazole and first 10 ml of elute was collected. Removal of the 6His-MBP tag was performed by addition of 1mg of TEV enzyme to the elute, incubated at 4 °C for 1 hour with no agitation. Sample was then introduced to IMAC resin again and the flow-through was collected and run over a HiLoad 26/600 S200 Superdex column (GE Healthcare) equilibrated with a buffer containing 100 mM potassium chloride, 20 mM tris pH 7.5 (at 25 °C), 5 mM magnesium chloride, and 5% (v/v) glycerol. Sample was then collected, concentrated with centrifugation columns, and then flash frozen in liquid nitrogen to be stored at −80 °C until further use. dCas9 was a generous gift from the laboratory of Jennifer Doudna (University of California, Berkeley).

Rep-X expression and purification

Rep-X was prepared as previously described (Arslan et al., 2015). pET28a(+) vector containing rep (C18L/C43S/C167V/C612A/S400C) was transformed into E. coli B21(DE3) (Sigma-Aldrich, CMC0014) and plated out on LB agar containing 50 μg/ml kanamycin at 37°C overnight. From the plate, a single colony was grown in 5 ml TB medium containing 50 μg/ml kanamycin at 30 °C overnight. The cells were transformed into 500 ml of TB medium containing 50 μg/ml kanamycin and grown at 37 °C. When OD reached the range between 0.3 and 0.4, the cells were moved to an 18 °C incubator. When OD reaches 0.6 to 0.8, the cells were induced expression with 0.5 mM IPTG and continue growth overnight. The cells were harvested by centrifugation for 15 min at 5000 rpm and 4 °C. The pellet was resuspended in 40 ml of the lysis buffer (50 mM Tris-HCl pH 7.5, 5 mM Imidazole, 200 mM NaCl, 20% (w/v) sucrose, 15% (v/v) glycerol, 17.5 ug/ml PMSF, and 0.2 mg/ml Lysozyme) and sonicate to lyse the cells. The lysed cell mix was centrifuged at 14,000 rpm at 4 °C for 30-60 min and collect the supernatant. The supernatant was stir-mixed with pre-equilibrated Ni-NTA resin for 1.5 hours at 4 °C. Ni-NTA purification was performed by washing the protein-bound resin with buffer A (50 mM Tris-HCl pH 7.5, 5 mM Imidazole, 150 mM NaCl, 25% (v/v) glycerol), followed by buffer A1M (50 mM Tris-HCl pH 7.5, 5 mM Imidazole, 1 M NaCl, 25% (v/v) glycerol) to remove any DNA residue, and final washed the protein-bound resin with buffer A, then eluted the Rep variant with imidazole buffer (50 mM Tris-HCl pH 7.5, 205 mM Imidazole, 150 mM NaCl, 25% (v/v) glycerol). 20 μM eluted Rep variant was mixed with 100 μM BMOE crosslinker to self-crosslink into Rep-X. The reaction was stir-mixed at room temperature for 1 hour. The excess crosslinker and Imidazole was removed by an overnight dialysis and stored in Rep-X storage buffer (50% glycerol, 600 mM NaCl, 50 mM Tris-HCl pH 7.5) at −80 °C.

Preparation of nucleic acids for single-molecule assays

DNA oligonucleotides were purchased from Integrated DNA Technologies (IDT). Cy5 N-hydroxysuccinimido (NHS) dyes were conjugated to DNA through a thymine modified with an amine group through a C6 linker (/iAmMC6T/). dsDNA targets were assembled by mixing the target strand (TS), non-target strand (NTS) and a 22-nt biotinylated adaptor strand at 1:1.25:1 ratio in T50 buffer (10 mM Tris-HCl pH 8, 50 mM NaCl) and incubating at 95 °C for 1 min, then cooling down to room temperature over 1 hour. The polyethylene glycol (PEG)-passivated flow chamber surface was purchased from Johns Hopkins University Slide Production Core for Microscopy. crRNA and tracrRNA were synthesized in vitro using HiScribe™ T7 Quick High Yield RNA Synthesis Kit (NEB, E2050S) according to the manufacturer's instructions. The guide RNA was annealed by mixing crRNA and tracrRNA at 1:1.25 ratio in Nuclease Free Duplex Buffer (IDT), and incubating at 95 °C for 30 seconds, then slowly cooling down to room temperature over 1 hour. The DNA and RNA sequences are listed in ‘Key Resources Table’.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-mH2A1 antibody	Abcam	Cat#ab183041
Goat anti-Rabbit Alexa Flour 750	Invitrogen	Cat#A21039
Anti-Her2/erbb2 antibody	Cell Signaling Technology	Cat#2165S
Bacterial and Virus Strains
BL21(DE3) Chemically Competent Cells	Sigma-Aldrich	Cat#CMC0014
Biological Samples
Human breast cancer primary patient tissue	ProteoGenex	N/A
Chemicals, Peptides, and Recombinant Proteins
Alt-R® S.p. Cas9 H840A Nickase V3	IDT	Cat#1081064
Rep-X	Arslan et al., 2015	N/A
Cas9dHNH	Jinek et al., 2012	N/A
dCas9	Laboratory of Jennifer Doudna	N/A
RNase Cocktail™ Enzyme Mix	Invitrogen	Cat#AM2286
Hoechst 33342 Ready Flow™ Reagent	Invitrogen	Cat#R37165
Terminal Deoxynucleotidyl Transferase	Thermo Scientific	Cat#EP0162
Maxima H-reverse transcriptase	Thermo Scientific	Cat#EP0753
Cy3 NHS Ester	GE Healthcare	Cat#PA13101
Cy5 NHS Ester	GE Healthcare	Cat#PA15100
Amino-11-ddUTP	Lumiprobe	Cat#A5040
Critical Commercial Assays
HiScribe™ T7 Quick High Yield RNA Synthesis Kit	NEB	Cat#E2050S
Phusion® Hot Start Flex 2X Master Mix	NEB	Cat#M0536S
RNA Clean & Concentrator-25	Zymo Research	Cat#R1017
DNA Clean & Concentrator-100	Zymo Research	Cat# D4029
Deposited Data
Raw images (z-stack TIFF)	This paper; Mendeley Data	http://dx.doi.org/10.17632/twj8ft95j5.2
Experimental Models: Cell Lines
IMR-90	ATCC	Cat#CCL-186
HEK293ft	Laboratory of Sergi Regot	N/A
Oligonucleotides
TracrRNA in DNA helicase invasion assay: GGACAGCAUAGCAAGUUAAAAUAAGGCUAGUCCGUUAUCAACUUGAAAAAGUGGCACCGAGUCGGUGCUUUUU	This paper	N/A
crRNA in DNA helicase invasion assay: GAUGUAUAAAGAUGAGACGCGUUUUAGAGCUAUGCUGUUUUG	This paper	N/A
22-nt biotinylated adaptor annealing to target strand in DNA helicase invasion assay: /5Biosg/AACGCAACGTCGTCAGCTGTCT	This paper	N/A
Target strand in DNA helicase invasion assay: AGCAAGCTGACGTTTGTACTCCAGCGTCTCATCTTTATACATCAGCAGAGATTTCTGCTGTGCAGACAGCTGACGACGTTGCG	This paper	N/A
Non-target strand for Cy5-labeling in DNA helicase invasion assay: GCACAGCAGAAATCTCTGCTGATG/iAmMC6T/ATAAAGATGAGACGCTGGAGTACAAACGTCAGCTTGCT	This paper	N/A
Alt-R® CRISPR-Cas9 tracrRNA	IDT	Cat#1072533
Alt-R® CRISPR-Cas9 tracrRNA, ATTO™ 550	IDT	Cat#1075927
Alt-R® CRISPR-Cas9 crRNA against MUC4-I1: /AltR1/GGUAUGGGUGUGGAAGGUAUGUUUUAGAGCUAUGCU/AltR2/	IDT	N/A
See Table S1 and Table S2 for the sequences of DNA oligonucleotides used in GOLD FISH experiments	This paper	N/A
Recombinant DNA
pMJ826	Jinek et al., 2012	RRID: Addgene 39316
rep (C18L/C43S/C167V/C612A/S400C)	Arslan et al., 2015	N/A
Software and Algorithms
Fiji/ImageJ	Schneider et al., 2012	https://imagej.net/Fiji
MATLAB R2019b	MathWorks	https://www.mathworks.com/
OriginPro 2020	OriginLab	https://www.originlab.com/
Python 3.7	Python	https://www.python.org/
Oligoarray 2.1	Rouillard, 2003	http://berry.engin.umich.edu/oligoarray2_1/
iFISH Probe Designer	Gelali et al., 2019	http://ifish4u.org/probe-design/
NIS-Elements AR	Nikon	https://www.microscope.healthcare.nikon.com/products/software/nis-elements/
FISH-quant	Mueller et al., 2013	https://code.google.com/archive/p/fish-quant/
Other
Detailed bench protocols of GOLD FISH experiments	This paper	Methods S1
The code used for searching Cas9 binding sites for chromosomal scale ‘paint’	This paper	https://github.com/ashleefeng/cas9

Microscopy and data acquisition for single-molecule assays

Microscopy was performed on Nikon Eclipse Ti microscope and custom prism type TIRFM module. The system was driven by home-built software (smCamera 2.0). Nikon 60X/1.27 NA objective (CFI Plan Apo IR 60XC WI) was used. Illumination was provided by solid-state lasers (Coherent, 641 nm) combined and coupled to an optical fiber. Emission was collected using long-pass filters (T540LPXR UF3, T635LPXR UF3, T760LPXR UF3) and a custom laser-blocking notch filter (ZET488/543/638/750M) from Chroma. Images were recorded using an electron-multiplying charge-coupled device (EMCCD; Andor iXon 897).

Single-molecule fluorescence imaging and quantification

For the DNA helicase invasion assay, Cy5-labeled dsDNA target (with 22-nt biotinylated adaptor strand) was immobilized on the PEG-passivated flow chamber surface using NeutrAvidin-biotin interaction. 100 nM Cas9 RNP was assembled by mixing 100 nM Cas9 and 100 nM wild-type gRNA and incubating for 10min at room temperature in Mg²⁺-containing imaging buffer (20 mM Tris-HCl pH 8, 100 mM KCl, 5 mM MgCl₂, 5% (vol/vol) glycerol, 0.2 mg/ml BSA and saturated Trolox (> 5 mM), 0.8% (w/v) dextrose) supplied with GLOXY (1 mg/ml glucose oxidase, 0.04 mg/ml catalase). The Cas9 RNP was flowed into the DNA-immobilized chamber and incubated for 20 min at room temperature. Short movies of 10 frames at 10 Hz with 641 nm laser excitation were taken at 20 different imaging views. The first 5 frame of each movie were averaged and Cy5 spot number per imaging view was measured as 0 min time point data. Then 100 nM Rep-X with 1 mM ATP in Mg²⁺-containing imaging buffer supplied with GLOXY were flowed into the chamber. The Cy5 spot number per imaging view was measured from 20 different imaging areas each again at different time points after flowing in Rep-X.

Genome sequences

The human genome assembly hg38 was used in this study and downloaded from genome.ucsc.edu. The coordinates of non-repetitive loci are listed below:

MUC4-NR (Chr3:195808789-195811123)

TAD5 (ChrX: 18579431-18584379)

TAD37 (ChrX:143999562-144006499)

HER2 (Chr17:39706827-39710552)

RARA (Chr17:40348168-40355149)

The coordinates of target sequences for p-arm/q-arm of ChrX ‘paint’ are listed in Table S2.

Cas9 binding site and probe design for GOLD FISH

For GOLD FISH against a short target region (<10 kb), all potential Cas9 binding sites (i.e. all PAM sequences) within the target region were found using Benchling (https://benchling.com/). Cas9 binding sites were chosen manually with the following constraints: adjacent Cas9 binding sites were generally spaced by 50 to 300 bp; all guide RNAs hybridized to the same strand (i.e. FISH-TS, Figure 1B) so that Rep-X would translocating in the same direction along the other strand (i.e. Rep-X translocating strand, Figure 1B). The average spacing between consecutive Cas9 binding sites for MUC4-NR, TAD5, TAD37, HER2 and RARA are 266 bp, 166 bp, 163 bp, 93 bp and 188 bp, respectively. We arranged Cas9 binding sites relatively close to each other to increase the likelihood that Rep-X could peel off the Rep-X translocating strand between the two adjacent Cas9 binding sites (Figure 1B). Next, desired oligonucleotide probes against the FISH-TS were designed using Oligoarray 2.1 (Figure 1B) (Rouillard, 2003). The DNA sequences between adjacent Cas9 binding sites were loaded into Oligoarray 2.1 with the following constraints: Length: 18- to 30-nt; Tm: 72 °C to 90 °C; %GC: 30-70; Max. Tm for structure: 54 °C; Min. Tm to consider X-hybrid: 54 °C; and there was no consecutive repeat of 5 or more identical nucleotides. For MUC4-R and MUC4-NR probes, no specificity filtering was performed. For TAD5, TAD37, HER2 and RARA, two specificity filters were applied: Probes with more than 30 non-specific bindings on human genome were removed; Probes that can non-specifically bind to human noncoding RNA and E.coli tRNA were also removed. We applied the probe filtering for the following reasons. First, if Cas9 and Rep-X non-specifically unwound a stretch of repetitive genomic DNA, and a probe that could non-specifically bind to the repetitive genomic DNA might give a detectable false positive signal. Second, RNA molecules in the cells might not be digested completely by RNAse. Probes annealing to abundant RNA (e.g. rRNA) or RNA molecules containing repetitive sequences might also give false positive signals. Third, E.coli tRNA was used as a blocking reagent. Forth, probe density remained high although the specificity filtering was applied (Figure S5C). The excellent signal-to-background ratio with MUC4-NR GOLD FISH indicates the probe specificity filtering was not necessary in terms of keeping nuclear background low (Figures 3A and 3B). The colocalization of MUC4-R and MUC4-NR signals suggests false positive signal was rare even without the probe specificity filtering (Figure 3C). The sequences of probes and template DNA for crRNA synthesis are listed in Table S1.

For p-arm/q-arm of ChrX ‘paint’, Cas9 binding sites were found using custom-written scripts. The Cas9 binding sites were restricted within the central 300-kb regions of TADs in ChrX in IMR-90 cells (Dixon et al., 2012). All guide RNAs hybridized to the same strand (i.e. FISH-TS, Figure 1B) so that Rep-X would translocating in the same direction along the other strand (i.e. Rep-X translocating strand, Figure 1B). To increase the likelihood that Rep-X could peel off the Rep-X translocating strand (Figure 1B), most of adjacent Cas9 binding sites were spaced by 50 to 200 bp. 50 to 100 Cas9 binding sites were designed for each TAD. Next, desired oligonucleotide probes against ChrX were designed (Figure 4C). The probes were designed as previously described (Wang et al., 2016), with the following modifications: each primary probe contains 4 regions: a 20-nt forward priming region, a 20-nt readout region, a 20-nt encoding region for hybridization to genomic DNA and a 20-nt reverse priming region (Figure 4C). To generate the encoding region sequences, the sequences between adjacent Cas9 binding sites (which spaced less than 200 bp) were loaded into Oligoarray 2.1 with the following constraints: Length: 20 nt; Tm: 72 °C to 90 °C; %GC: 30-70; Max. Tm for structure: 54 °C; Min. Tm to consider X-hybrid: 54 °C; and there was no consecutive repeat of 5 or more identical nucleotides; The generated encoding region sequences with more than 10 non-specific bindings on human genome or can bind to human non-coding RNA were removed. To generate the priming regions and readout region, computationally designed 25mer sequences (Xu et al., 2009) were loaded into Oligoarray 2.1 with the following constraints: Length: 20 nt; Tm: 75 °C to 90 °C; %GC: 40-60; Max. Tm for structure: 54 °C; Min. Tm to consider X-hybrid: 54 °C; and Sequence to avoid in the oligo: 'GGGG;CCCC;TTTT;AAAA;ATATAT;TATATA;ACACAC;CACACA;CGCGCG;GCGCGC'; The generated priming region sequences and readout region sequences with at least 1 non-specific binding on human genome were removed. Amount thousands of candidate sequences satisfying all the constraints, four sequences were chosen as the priming regions and readout regions in this study (2 sequences for the priming regions and 2 sequences for the readout regions). Finally, the primary probes were assembled using the encoding region sequences, priming region sequences and readout region sequences as indicated in Figure 4C. The primary probes with at least 9 non-specific bindings to human genome or at least one non-specific binding to E.coli tRNA or human non-coding RNA were excluded using BLAST+ (here a ‘non-specific binding’ refers to the primary probe contains > 16 nt homology sequence to an off-target sequence). The sequences of primers and template DNA for synthesizing the primary probes and the crRNAs are listed in Table S2. The sequences of readout probes are also listed in Table S2.

Preparation of guide RNAs for GOLD FISH

For GOLD FISH against a short target region (<10 kb), template DNA for in vitro transcribing crRNAs were purchased from IDT. The template DNA of a crRNA was partially double stranded, including a double-stranded T7 promoter region and a single-stranded template region (Figure S5B). crRNAs were transcribed using HiScribe™ T7 Quick High Yield RNA Synthesis Kit (NEB). Different crRNAs have different protospacer sequences at 5’ end, and we found the transcription efficiency of crRNA heavily depends on its 5’ end sequence. Therefore, different crRNA would have different transcription efficiencies (Figure S5B). To make the transcription efficiency homogeneous for different crRNA, we adopted 5’ extended crRNA (Kocak et al., 2019). A common 10-nt extension at 5’ of each crRNA made the transcription efficiency homogeneous (Figure S5B). For MUC4-R, MUC4-NR, TAD5 and CEP17, canonical crRNAs were used, and each crRNA was transcribed separately. For TAD37, HER2 and RARA, the 5’ extended crRNAs were used, and each set of crRNAs were transcribed together in a single reaction. The transcribed crRNAs were purified by polyacrylamide gel electrophoresis.

For p-arm/q-arm of ChrX ‘paint’, an oligopool of template DNA (Twist Bioscience) was amplified to a dsDNA pool using Phusion^® Hot Start Flex 2X Master Mix (NEB, M0536S) by limit-cycle PCR (no more than 10 cycles). The crRNAs for ChrX ‘paint’ was in vitro transcribed using the dsDNA pool and HiScribe™ T7 Quick High Yield RNA Synthesis Kit (NEB). The synthesized crRNA was purified using RNA Clean & Concentrator Kits (Zymo, R1017).

To assemble guide RNAs, the Alt-R^® CRISPR-Cas9 tracrRNA (IDT) or Alt-R^® CRISPR-Cas9 tracrRNA, ATTO™ 550 (IDT) and desired crRNAs were mixed at 1:1 ratio in Nuclease-Free Duplex Buffer (IDT) and incubated at 95 °C for 30 s, then slowly cooled down to room temperature over 1 hour.

Synthesis of DNA probe for GOLD FISH

For GOLD FISH against a short target region (<10 kb), designed DNA oligonucleotides (without any labeling/modification) were purchased from IDT, and fluorescently labeled as previously described (Gaspar et al., 2017). Briefly, to conjugate an amino-ddUTP at the 3’ end of each oligonucleotide, 66.7 μM DNA oligonucleotides, 200 μM Amino-11-ddUTP (Lumiprobe) and 0.4U/μl Terminal Deoxynucleotidyl Transferase (TdT, Thermo Scientific, EP0162) were mixed in 1X TdT Reaction buffer (Thermo Scientific) and incubated overnight at 37 °C. The reaction was cleaned up by ethanol precipitations and P4 beads (Bio-Rad, #1504124) spin columns. Next, the DNA oligonucleotides conjugated with amino-ddUTP were mixed with 100 μg of Cy3-NHS or Cy5-NHS (Lumiprobe or GE Healthcare) in 0.1 M sodium bicarbonate and incubated overnight at room temperature, and cleaned up by ethanol precipitations and P4 beads (Bio-Rad, #1504124) spin columns. We generally achieved ~90% labeling efficiency. In some cases, unlabeled oligonucleotides were removed by high-performance liquid chromatography (HPLC).

For p-arm/q-arm of ChrX ‘paint’, an oligopool of template DNA for synthesizing primary probes were purchased from Twist Bioscience, and the primary probes were synthesized as previously described (Moffitt and Zhuang, 2016). In short, the oligopool of template DNA was amplified to a dsDNA pool using Phusion^® Hot Start Flex 2X Master Mix (NEB) by limit-cycle PCR (no more than 10 cycles). One of the primers we used for the limit-cycle PCR contained a T7 promoter sequence. The dsDNA pool was cleaned up by using DNA Clean & Concentrator-100 (Zymo, D4029).

Next, the dsDNA pool was further amplified and converted into single-stranded RNA (ssRNA) pool using HiScribe™ T7 Quick High Yield RNA Synthesis Kit (NEB). The ssRNA pool was converted back to desired primary probe pool by using Maxima H-reverse transcriptase (Thermo Scientific, EP0753) followed by alkaline hydrolysis. The primary probe pool was further purified to remove enzyme and excess dNTPs, and we found RNA Clean & Concentrator Kits (Zymo, R1017) worked excellently for this purpose. The primers for synthesizing the primary probes and Cy3- or Cy5-labeled secondary readout probes were purchase from IDT. The sequences of primers and template DNA for synthesizing the primary probes and the crRNAs are listed in Table S2. The sequences of readout probes are also listed in Table S2.

Cost estimate of GOLD FISH

Here we estimate the cost of GOLD FISH targeting a non-repetitive genomic locus (a few kb long). Assume GOLD FISH will be performed in an imaging dish with 12-millimeter-diameter glass bottom surface.

Guide RNAs.

Alt-R® CRISPR-Cas9 tracrRNA can be purchased from IDT, each GOLD FISH experiment consumes ~ 100 pmol of tracrRNA ($0.6 to $1.9). Template DNA for in vitro transcribing crRNAs can be purchased from IDT (oPools Oligo Pools). A set of template DNA strands (which can transcribe up to 47 different crRNAs) costs $99. The crRNAs can be in a single reaction using HiScribe™ T7 Quick High Yield RNA Synthesis Kit ($5.24 per reaction).

Oligonucleotide probes.

DNA oligonucleotides without any labeling or modification can be purchase from IDT in a 500 picomole DNA Plate Oligo. The plate requires at least 96 oligonucleotides to be ordered. We found ~ 60 probe oligos (on average 21-nt for each probe) would be enough for GOLD FISH to achieve excellent signals. Therefore, a plate containing 60 oligonucleotide probes and 36 random oligonucleotides (15-nt each) can be purchased from IDT ($180). To label the 60 oligonucleotide probes, terminal deoxynucleotidyl transferase (ThermoFisher, EP0162), amino-11-ddUTP (Lumiprobe) and NHS-form of fluorophores were used ($6 to $32).

Cas9 and Rep-X.

Each GOLD FISH experiment consumes ~ 100 pmol of Cas9_dHNH. 100 pmol of Alt-R® S.p. Cas9 H840A Nickase V3 (IDT) costs $23 to $32. If Cas9_dHNH is produced in the lab, the cost on Cas9 per experiment can be substantially lower. We produced 200 nmol of Rep-X using reagents of less than $300, while each GOLD FISH experiment consumes only 45 pmol of Rep-X. Therefore, once Rep-X is produced in the lab, the cost on Rep-X per experiment can be very low.

Fixation of cultured cell and tissue section

We adopted two different fixation methods for GOLD FISH: methanol-acetic acid (MAA) fixation and Buffered Ethanol (BE70)-based fixation (BE70-MAA). Both methods permeabilized cells during the fixation steps.

For GOLD FISH against MUC4-R and MUC4-NR, MAA fixation was used (except Figure S5D). Cells were briefly washed once with PBS and fixed at −20 °C for 20 min in pre-chilled MAA solution (methanol and acetic acid mixed at 1:1 ratio), then washed three times (5 min each wash at room temperature unless indicated) with PBS.

For GOLD FISH against other genomic regions, BE70-MAA fixation was used. This fixation method has two steps: BE70 fixation and MAA treatment. The BE70 buffer were prepared as previously described (Perry et al., 2016). To make 50 ml of BE70 buffer, 2.5 ml of 10X PBS (pH 7.4) was mixed with 1 ml of 50% glycerol and 0.25 ml of glacial acetic acid. The mixture was adjusted to pH 4.3 by adding NaOH. The solution was then filled to 15 ml with distilled water and mixed with 35 ml of absolute (200 proof) EtOH. Cells were briefly washed once with PBS and fixed at room temperature for 25 min in BE70 buffer, then washed twice with PBS. Cells were then incubated at −20 °C for 20 min in pre-chilled MAA solution and washed three times with PBS. We found the incubation in MAA solution was necessary for efficient GOLD FISH labeling in the BE70-fixed cells, presumably because MAA solution further permeabilized the cells.

For the PFA-fixed cells (Figure S5D), cells were fixed with 4% paraformaldehyde (PFA) in PBS at room temperature for 10 min. The cells were washed three times with PBS and incubated in freshly made 1 mg/ml sodium borohydride for 10 min at room temperature. The cells were washed twice with PBS, and further permeabilized with 0.5% (v/v) Triton X-100 in PBS for 10 min at room temperature. The cells were washed twice with PBS and incubated with 0.1 M HCl for 5 min at room temperature. Finally, the cells were washed three times with PBS.

For CASFISH, cells were fixed as previously described (Deng et al., 2015). Cells were fixed at −20 °C for 20 min in pre-chilled MAA solution, then washed three times with PBS.

GOLD FISH

In this work, GOLD FISH was performed against different genomic sequences (e.g. repetitive, non-repetitive, and chromosome ‘paint’), and the GOLD FISH protocol has evolved with the development of the method. Therefore, individual GOLD FISH experiments were performed with different parameters (e.g. Cas9 RNP and oligo probe concentrations). To avoid confusions, here we describe a standard GOLD FISH protocol. Detailed protocols of each GOLD FISH experiment presented in this work can be found in Methods S1.

Step 1: Targeted chromatin denaturation.

After the cell fixation, Cas9 RNP (20 nM to 40 nM per guide RNA species, e.g. the MUC4-NR guide-RNA set 1 contains 9 different guide RNAs, then the total concentration of guide RNA in this step would be 180 to 360 nM) was assembled by mixing equal amount of Cas9_dHNH and guide RNA in Binding-Blocking buffer (20 mM Hepes pH 7.5, 100 mM KCl, 5 mM MgCl₂, 5% (v/v) glycerol and 0.1% (v/v) TWEEN-20, 1% (w/v) BSA, freshly added 1 mM DTT, freshly added 0.1 mg/ml E.coli tRNA) and incubated for 10 min at room temperature. The cells were incubated in Binding-Blocking buffer for 10 min at 37 °C, and the Cas9 RNP was added to the cells and incubated for 30 to 60 min at 37 °C. After the incubation, free Cas9 RNP were removed. Rep-X (100 to 400 nM) in Binding-Blocking buffer supplied with 2 mM ATP were added to the cells and incubated at 37 °C for 30 min. The cells were washed three times (5 min each wash at room temperature) with PBS.

Step 2: RNAse digestion (optional).

RNase Cocktail™ Enzyme Mix (Invitrogen, AM2286) was diluted 100 times in PBS and incubated with the cells for 1 hour at 37 °C. The cells were washed three times (5 min each wash at room temperature) with PBS.

Step 3: FISH probe hybridization.

The cells were incubated in freshly made hybridization buffer (10% to 20% (v/v) formamide, 2X saline-sodium citrate (SSC), 0.1 mg/ml E.coli tRNA, 10% (w/v) dextran sulfate, 2 mg/ml BSA) for 10 min at 37 °C. Next, fluorescently labeled oligonucleotide probes (2.5 nM per probe, e.g. the MUC4-NR probe set contains 57 different oligonucleotide probes, then the total concentration of probes in this step should be 142.5 nM) in the hybridization buffer were applied to the cells and incubated for 1 hour at room temperature (repetitive targets) or 37 °C (non-repetitive targets). The cells were washed twice (15 min each wash) with wash buffer (20% formamide, 2X SSC) at 37 °C and once with PBS at room temperature for 5 min.

Step 4: Preparation for imaging.

(Optional) one drop of Hoechst 33342 Ready Flow™ Reagent (Invitrogen, R37165) was mixed with 2 ml of PBS and incubated with the cells for 2 min at room temperature. Finally, FISH-imaging buffer (20 mM Tris-HCl pH 7.5, 100 mM KCl, 5 mM MgCl₂, 5% (vol/vol) glycerol, 0.2 mg/ml BSA and saturated Trolox (> 5 mM), 0.8% (w/v) dextrose) supplied with GLOXY (1 mg/ml glucose oxidase, 0.04 mg/ml catalase) was added to the cells for imaging.

CASFISH

CASFISH experiments were performed as previously described (Deng et al., 2015). The fixed cells were incubated with CASFISH-blocking/reaction buffer (20 mM Hepes pH 7.5, 100 mM KCl, 5 mM MgCl₂, 5% (v/v) glycerol and 0.1% (v/v) TWEEN-20, 1% (w/v) BSA, freshly added 1 mM DTT) at 37 °C for 15 min. Five nM Cas9_dHNH or dCas9 was mixed with 5 nM ATTO550-labeled guide RNA and incubated in the CASFISH-blocking/reaction buffer for 10 min at room temperature, and stored on ice before next step. The assembled Cas9 RNP was applied to the cells and incubated at 37 °C for 30 min. After the incubation, the cells were washed three times with CASFISH-blocking/reaction buffer (5 min each at room temperature).

Microscopy and data acquisition for GOLD FISH

Epifluorescence microscopy was performed on Nikon Eclipse Ti2 microscope with Nikon Plan Apo λ 60x Oil objective and Intermediate Magnification switching of 1.0x/1.5x. The system was driven by NIS-Elements AR software. Illumination was provided by high power LED. Emission was collected using filter sets: ET - Sedat Quad (Chroma, 89100) for Hoechst 33342 channel, ET - Gold FISH (Chroma, 49304) for Cy3 or ATTO550 channel, ET - Cy5 Narrow Excitation (Chroma, 49009) for Cy5 channel, and ET - Cy7 (Chroma, 49007) for Alexa750 channel. Images were recorded as z-stacks (21 to 35 steps), with 200 nm or 300 nm step size using a digital CMOS camera (ORCA-Flash 4.0 C11440, Hamamatsu), except for several images which were recorded using an EMCCD camera (Andor iXon 888) (Figures 3F, S3D and S3E). TetraSpeck™ Microspheres (T7279, Invitrogen) were also imaged in the same way for correction of chromatic aberration between Cy3/ATTO550 channel and Cy5 channel.

Comparison of live and after-GOLD FISH cells

DNA in live IMR-90 cells was stained with Hoechst 33342 Ready Flow™ Reagent (Invitrogen, R37165) and imaged at the focus plane where the nuclear edges were the sharpest. The coordinates of imaged cells were recorded so that the same cells could be found again after GOLD FISH protocol. Next, some cells (Figure S3A, top) were fixed using the BE70-based fixation method (i.e. BE70 fixation followed by MAA treatment). Other cells (Figure S3A, bottom) were fixed using the MAA fixation method. Next, the protocol of GOLD FISH against TAD5 and TAD37 was performed on all cells. After the GOLD FISH, the previously imaged cells were found, and their nuclei were imaged again at the focus plane where the nucleus edges were sharpest. The images of nuclei before and after GOLD FISH were split into sub-images, and each sub-image contained only one nucleus. The area of nucleus in each sub-image were automatically measured using the ‘Threshold’ function with ‘IsoDATA’ parameter in Fiji/ImageJ (Schneider et al., 2012). For each cell, the ratio of nuclear area after GOLD FISH (Area_GOLDFISH) to nuclear area when the cell was alive (Area_Live) was calculated (Figure S3B).

Data analysis for GOLD FISH

Generate representative images shown in figures.

Images were processed using Fiji/ImageJ. Z-stack images were projected to a single plane using the ‘Max Intensity’ Z-Projection function. The contrasts of images were linearly adjusted by changing the minimum and maximum values using the ‘brightness/contrast’ function in Image J for optimal visualization purpose only. The correction of chromatic aberration between Cy3/ATTO550 channel and Cy5 channel was performed using the TetraSpeck™ Microspheres images with custom-written MATLAB scripts.

Foci fitting and signal-to-background measurement.

FISH-quant was used to find foci in each cell and fitted with three-dimensional (3D) Gaussian function (Mueller et al., 2013). Spatial coordinates (x, y and z), amplitude (A_signal) and background (BGD_FISH-quant) were extracted from the 3D Gaussian fitting. The average background (BGD_coverslip) was calculated from multiple areas where there was no cell. To calculate signal-to-background ratio (S/B), we used

$$S∕B=\frac{A_{signal}}{BG{D}_{FISH-quant}-BG{D}_{coverslip}}$$

TAD5 and TAD37 distance measurement.

After the chromatic aberration correction, the distance between TAD5 and TAD37 was measured:

$$Distance=\sqrt{(x_{\text{TAD5}}-x_{\text{TAD37}}{)}^2+(y_{\text{TAD5}}-y_{\text{TAD37}}{)}^2+(z_{\text{TAD5}}-z_{\text{TAD37}}{)}^2}$$

Center of Mass distance and volume measurement.

The Z-stack images of ChrX ‘paint’ were background-subtracted using the ‘Subtract background’ function in Fiji/ImageJ with rolling ball radius of 15 pixels. After the chromatic aberration correction between Cy3 channel and Cy5 channel, each ChrX was cropped into a small region manually. The mean and standard deviation of residual nuclear background (BGD_mean and BGD_STDEV) were measured. A threshold (T) was set at:

$$T=BG{D}_{mean}+1.5\ast BG{D}_{STDEV}$$

The pixels within the cropped region with intensities higher than T were selected. The center of mass coordinate ($\stackrel{\rightharpoonup }{CoM}$) of p-arm or q-arm of ChrX was calculated using the coordinates $\stackrel{\rightharpoonup }{x}$ of each selected pixel and intensity I of each selected pixel as weighting factors:

$$\stackrel{\rightharpoonup }{CoM}=\frac{{\sum }_{I_i>T}(\stackrel{\rightharpoonup }{x_i}\ast I_i)}{{\sum }_{I_i>T}\stackrel{\rightharpoonup }{x_i}}$$

The CoM distances between p-arm and q-arm of each ChrX were calculated:

$$Distance=\sqrt{(\stackrel{\rightharpoonup }{Co{M}_{p-arm}}-\stackrel{\rightharpoonup }{Co{M}_{q-arm}}{)}^2}$$

The volumes of each ChrX were calculated:

$$Volume=Numberofselectedpixels\ast pixelsize$$

QUANTIFICATION AND STATISTICAL ANALYSIS

FISH foci were fitted with three-dimensional Gaussian functions using FISH-quant to obtain foci number per cell, foci intensity and background (Mueller et al., 2013). The nuclear area of each cell was automatically measured using the ‘Threshold’ function with ‘IsoDATA’ parameter in Fiji/ImageJ (Schneider et al., 2012). Two or more cells with overlapping nuclei were excluded from quantifications. Statistical analyses were conducted using Student’s t-test. n represents number of cells (except for Figure 1E, where n represents number of imaging view measured). Standard deviation (SD) are shown in this work. OriginPro 2020 was used for the statistical analysis. Statistical details of experiments such as values of n can be found in the figure legends.

ADDITIONAL RESOURCES

Detailed Protocol

A detailed bench protocol of GOLD FISH experiments can be found in Methods S1.

Supplementary Material

Table S1

Table S1. Sequences of DNA used for GOLD FISH against MUC4-R, MUC4-NR, TAD5, TAD37, HER2, RARA and CEP17. Related to STAR Methods.

Table S2

Table S2. Sequences of DNA used for ChrX ‘paint’. Related to STAR Methods.

Supplementary material

Methods S1. Detailed protocols of GOLD FISH experiments. Related to STAR Methods.

Highlights.

• Cas9 can function as a programmable loader of superhelicase to genomic DNA.

• GOLD FISH locally unwinds target genomic DNA for FISH probe hybridization.

• GOLD FISH allows imaging of non-repetitive genomic DNA with low background.

• GOLD FISH is scalable from targeting a 2.3-kb locus to chromosomal-scale ‘paint’.