Simultaneous DNA/RNA HiFISH
We sought to develop a protocol for the sensitive visualization of cellular DNA and RNA through single-step hybridization of a mixture of DNA and RNA probes (Fig. 1). To do so, we optimized all steps of the FISH protocol, including permeabilization, hybridization, and washes in a 384-well plate format (Fig. 1; see Materials and Methods).
For optimal results, cells were grown to 80% density, and standard 4% paraformaldehyde fixation was used. For cell permeabilization, a saponin/triton combination supplemented with an RNase inhibitor was used to preserve RNA integrity. After deproteination with HCl, the cells were equilibrated overnight in 50% formamide in preparation for subsequent hybridization.
The hybridization step employed a customized buffer designed for optimal DNA and RNA probe binding, containing dextran sulfate, formamide, SSC, Tween-20, sodium citrate, and an RNase inhibitor (see Materials and Methods for details). The concentration of the components in the hybridization buffer was systematically adjusted in pilot experiments to an intermediate level between standard DNA and RNA hybridization buffers (Shaffer et al. 2013; Finn and Misteli 2021), creating an environment conducive to both DNA and RNA probe hybridization. The hybridization buffer was explicitly formulated for simultaneous DNA/RNA HiFISH, with the formamide concentration adjusted to be 20% lower than that typically used in DNA FISH protocols, yet 20% higher than that in RNA FISH, to balance the conditions for both types of probes. Moreover, we halved the concentration of Tween-20 to approximate the viscosity characteristic of RNA FISH buffers. Dextran sulfate and SSC concentrations were kept at standard levels compatible with both DNA and RNA FISH. This tailored approach ensures the stability and hybridization efficacy of both DNA and RNA probes. Additionally, the buffer was supplemented with an RNase inhibitor, and a molecular grade 1 mM sodium citrate buffer was used, aimed at stabilizing and minimizing inherent DNA and RNA base hydrolysis by inhibiting the activity of both DNase and RNase during the hybridization process. Detection was further optimized using a pre-hybridization wash buffer containing 10% formamide and 2XSSC. We either used commercially available pre-labeled DNA probes or in-house generated fluorescently labeled BAC probes (see Materials and Methods). For RNA probes, we used multiple 20-base single-stranded DNA oligonucleotides, individually labeled and designed to bind distinct segments of the target RNA via Watson-Crick base pairing (Supplementary Fig. 1).
During the process of optimizing the simultaneous FISH protocol to a 384-well format, we noted heterogeneity in heat distribution among wells during the heat denaturation step, especially when using various commercial heat blocks. Because uneven heating across the plate could introduce inconsistencies in technical replicates, we systematically assessed multiple heat blocks. We identified the ThermoMixer® C-PCR 384 (Eppendorf) as the optimal choice, providing a reliable and uniform heat distribution across the 384-well plate.
Hybridization efficiency was further optimized by a prolonged incubation period of up to 48 hours. All steps used molecular-grade RNase-free reagents, and treatment of reagents with RNAsecure™ ensured RNase inactivation and thorough RNase decontamination measures were implemented on equipment and benchtops using RNaseZap™ before experiments. Commercially available or in-house labeled BAC probes worked equally well, and Stellaris® RNA FISH probes were routinely used. Following a sequential series of standard rinsing steps, cells are stained with DAPI and mounted for imaging using a high-throughput confocal microscope per standard FISH protocol (see Materials and Methods).
To evaluate our approach, we performed simultaneous DNA/RNA HiFISH utilizing BAC FISH probes targeting the downstream regions of the MYC and EGFR genes on human chromosomes 8 and 7, respectively (Supplementary Fig. 1). We detected both DNA and RNA for MYC and EGFR in HBEC and HFF cells with high efficiency (Fig. 2, 3; Supplementary Fig. 2, 3). MYC DNA signals were detected in 98 ± 1.1% of HBECs and 99 ± 0.9% of HFF cells, and both copies of MYC were detected in 72 ± 5.7% of HBEC cells (Fig. 2c) and 77 ± 2.9% of HFF cells (Fig. 2i). Likewise, EGFR was detected in 89 ± 4.4% of HBECs and 90 ± 9.2% of HFFs, 57 ± 7.6% of HBEC cells (Fig. 3c), and 58 ± 18% of HFF cells showing two signals (Fig. 3i). These values are well within detection efficiencies previously reported in hiFISH approaches (Shachar et al. 2015; Finn et al. 2019). The few DNA signals that were missed are likely due to suboptimal FISH hybridization, weak FISH signal, or high background signal, which can reduce signal detection.
For RNA detection, in line with the demonstrated variable expression of MYC in individual cells (Liu et al. 2023), 6.4 ± 2.9% of HBECs showed biallelic expression, monoallelic expression in 26 ± 3.4% of cells and no expression in 66 ± 6.3% of HBEC cells (Fig. 2e). Similarly, in HFF, 2.0 ± 0.6% showed biallelic expression, monoallelic expression in 16 ± 4.8% and no expression in 82 ± 5.1% of cells (Fig. 2k). In contrast, EGFR RNA is more highly expressed in both HBEC and HFF and accordingly more active alleles were detected in HBEC; 44 ± 5.0% of cells showed biallelic expression, 30 ± 2.7% monoallelic expressed, and only 20 ± 4.5% no expression. In HFF, 34 ± 1.7% of cells expressed EGFR biallelically, 38 ± 3.1% monoallelically expressed, and 22 ± 5.0% did not express the gene. More than the expected two transcription sites per nucleus were detected in only 3.2 ± 3.1% of cells for both genes and cell lines due to false positive detection of FISH signals. We conclude that simultaneous detection of DNA and RNA using a single-hybridization step is feasible.
Sequential DNA/RNA HiFISH
To complement the simultaneous FISH detection, we developed an alternative in which RNA and DNA are detected in a sequential fashion based on previously published protocols for each nucleic acid species(Raj and Tyagi 2010; Shaffer et al. 2013; Finn and Misteli 2021; Finn et al. 2022) with optimizations implemented for high-throughput image processing (Fig. 1). This method involves first RNA detection, imaging of the RNA signals, followed by DNA FISH and a second round of imaging to detect DNA signals. The two sets of images are then accurately superimposed using an image registration algorithm. In our hands, detection of DNA prior to RNA yielded sub-optimal results, and we focused on optimizing protocols that detect RNA before DNA.
After cell culture and fixation, the RNA FISH protocol is initiated with overnight permeabilization at 4°C using 70% ethanol. The subsequent hybridization involves washing with 10% formamide buffer, followed by the introduction of RNA probes in a standard hybridization buffer (Shaffer et al. 2013). After overnight hybridization, a series of rinsing steps is performed, and cells are stained with DAPI for imaging of RNA signals.
After imaging of the RNA signals, plates are returned for DNA FISH, which involves re-permeabilization with saponin/triton, deproteination with HCl, and equilibration in 50% formamide, followed by hybridization with DNA BAC probes in standard hybridization buffer and washes as per established protocols (Shachar et al. 2015; Finn et al. 2019) and finally imaging of the DNA signal by high-throughput microscopy.
After imaging of the DNA signals, an image alignment algorithm is employed for DAPI-stained DNA and RNA image registration, ensuring precise spatial alignment of the RNA and DNA signals for combined analysis of DNA and RNA signals (see Materials and Methods for details). The algorithm utilized a technique akin to aligning two complex patterns by identifying reference points within each image. In this case, it scrutinizes the distinct DAPI staining patterns of individual nuclei present in both the DNA and RNA images, relying on specific high-intensity areas—corresponding to the DAPI-stained regions—as reference points for alignment. By evaluating these identifiable features in RNA and DNA FISH images and leveraging cross-correlation techniques, the algorithm calculated the optimal translation vector, essentially a set of instructions determining the precise shift needed to match the RNA image to the DNA image. This alignment process enabled us to generate an accurate combined image of the RNA and DNA signals, allowing us to examine and compare the DNA and RNA signals (Fig. 2, 3; Supplementary Fig. 2, 3; see Materials and Methods for details).
In validation experiments, sequential RNA/DNA HiFISH using BAC FISH probes targeted MYC and EGFR. High detection efficiency was observed for both DNA and RNA signals in HBEC and HFF cells (Fig. 2, 3; Supplementary Fig. 2, 3). MYC DNA signals were detected in 98 ± 1.3% of HBEC and 95 ± 3.6% of HFF cells, with 75 ± 8.6% of HBEC cells (Fig. 2d) and 65 ± 13.5% of HFF cells (Fig. 2j) showing two copies of MYC. EGFR was detected in 83 ± 15.6% and 99 ± 0.6% of HBECs and HFFs, respectively, with 56 ± 8.1% of HBEC cells (Fig. 3d) and 69 ± 1.0% of HFF cells (Fig. 3j) showing two signals.
For MYC RNA detection, 27 ± 3.2% of HBECs exhibited monoallelic expression, 4.6 ± 1.5% showed biallelic expression, and 68 ± 4.9% showed no expression (Fig. 2f), similar to the simultaneous detection method. For HFF, 81 ± 4.8% of cells did not express MYC, 16 ± 4.8% showed monoallelic expression and 2.0 ± 0.2% showed biallelic expression (Fig. 2i). Conversely, EGFR RNA was more highly expressed in both HBEC and HFF. In HBEC, 27 ± 21% of cells showed biallelic expression, 33 ± 0.6% monoallelic expressed, while 38 ± 23% of cells were silent (Fig. 3f). In HFF, 10 ± 3.0% of cells were biallelic expressed, 33 ± 2.2% were monoallelic expressed, and 56 ± 1.0% of cells were silent (Fig. 3i).
When directly compared, simultaneous and sequential DNA/RNA HiFISH gave similar results but exhibited some nuanced differences. Simultaneous HiFISH was slightly more robust in detecting MYC DNA signals, with detection rates of 98–99% in HBECs in HFF cells vs. 95–985%, in sequential FISH. The detection efficiency of EGFR DNA signals was similar for both methods. In terms of RNA detection, MYC RNA displayed comparable expression in simultaneous and sequential detection of MYC active alleles (33 ± 6.2% vs 31 ± 4.8% in HBECs, 18 ± 5.3% vs 18 ± 5.0% in HFFs, respectively). In contrast, detection of the more highly expressed EGFR RNA by simultaneous FISH was slightly more efficient compared to sequential detection with detection efficiencies in HBEC of 74 ± 4.5% vs 61 ± 22%, and in HFFs 72 ± 1.4% vs 42 ± 0.7%. As expected, differences between the FISH methods were more pronounced for the more highly expressed EGFR gene, whereas the two methods were more similar in detection efficiency for the more lowly expressed MYC gene. Considering the detection sensitivities, we conclude that both methods are suitable for the high-sensitivity detection of DNA and RNA at the single-allele level.
Application of allele-level DNA/RNA HiFISH to compare the radial position of active and inactive gene alleles
The RNA/DNA detection pipelines developed here can be used to probe the behavior of active and inactive alleles in the same cell nucleus. As proof-of-principle for their utility, we applied DNA/RNA HiFISH to ask whether the nuclear position of the active and inactive alleles differ (Fig. 4). To do so, we visualized active and inactive alleles of MYC or EGFR in HBEC or HFF using both simultaneous and sequential DNA/RNA HiFISH. We defined active alleles as DNA FISH signals associated with an RNA signal within 1.0 micron. DNA FISH signals without an RNA signal were defined as inactive. For each active or inactive allele, we determine its radial position relative to the center of the cell nucleus using a previously described distance transform method, which assigns each allele a value between 0 (center of the nucleus) and 1 (periphery) (see Materials and Methods for details). We analyzed between 1,280 and 5,118 alleles per sample. To ensure the reliability of our analyses, we excluded from analysis nuclei that contained an incorrect number of detected signals for both DNA and RNA, typically under 10–30% nuclei in the sample.
In line with prior observations (Shachar et al. 2015), the overall distribution of the radial position for the two genes was distinct (Fig. 4). MYC showed preferential localization to the periphery of the nucleus with both simultaneous and sequential methods in both HBEC (mean radial distance = 0.7 ± 0.1) and HFF (mean radial distance = 0.7 ± 0.0). This observation aligns with previous findings indicating peripheral localization of MYC in normal HCECs and cancer HCT-116 cells (Scholz et al. 2019). In contrast, EGFR exhibited a more uniform distribution within the nucleus in both HBEC (mean radial distance = 0.6 ± 0.0) and HFF (mean radial distance = 0.6 ± 0.0) (Fig. 4e - h).
A comparison of the radial position of active vs. inactive alleles for MYC and EGFR showed no difference in localization. For MYC active and inactive alleles in HBEC showed similar peripheral positioning using both simultaneous (mean radial distance = 0.7 ± 0.2 for both) or sequential DNA/RNA HiFISH (KS test p-values: 0.37, 0.22, respectively) (Fig. 4; Supplementary Table 1). Similarly, the radial positioning of MYC in HFF showed similar peripheral positioning for both active (mean radial distance = 0.7 ± 0.3) and inactive (mean radial distance = 0.7 ± 0.2) alleles using either simultaneous FISH or sequential detection (mean radial distance = 0.7 ± 0.2) (KS test p-values: 0.86, 0.13, respectively) (Fig. 4; Supplementary Table 1).
Like MYC, the radial positioning of EGFR did not show a difference between active and inactive alleles in both HBEC and HFF (mean radial distance = 0.6 ± 0.2; KS test p-value: 0.51 ± 0.2) with the exception of inactive alleles in HFF when using the sequential method (mean radial distance = 0.7 ± 0.2) which may be attributed to smaller sample size (KS test p-value: 0.05) (Fig. 4; Supplementary Table 1).
Comparing the radial distance distributions obtained by simultaneous versus sequential DNA/RNA HiFISH, we observed a slight but statistically significant difference, with an average KS test p-value of 0.04 ± 0.06 (Supplementary Table 2). Sequential FISH exhibited a mean radial distance skew of 0.05 ± 0.02 units towards the nuclear periphery compared to simultaneous FISH (Fig. 4; Supplementary Table 2). We suspect this discrepancy may arise from cell shrinkage during sequential FISH due to the use of a 70% ethanol overnight permeabilization step.
Together, these results demonstrate sensitive detection of active and inactive gene alleles and suggest that the radial position of MYC and EGFR alleles is independent of their expression status.