Research was authorized by the University of Washington Institutional Animal Care and Use Committee. Wild-type AB Danio rerio were raised under standard conditions in the UW SLU Aquatics facility on a 14h/10 h light/dark cycle, with lights on at 9 a.m. Male and female fish were euthanized over an ice bath, followed by cervical dislocation and enucleation. For the 11 p.m. timepoint euthanasia and dissections were performed under infrared light with night vision goggles.

For histology, eyes were collected at 9 a.m. and 11 p.m. (4 animals per timepoint), pierced several times across the cornea with a needle, and the vitreous cavity was flushed with 4% paraformaldehyde fixative in phosphate buffered solution (PBS). Following overnight immersion in fixative at 4°C, eyes were rinsed in PBS, then bleached with 10% H₂O₂ in PBS overnight at 37°C to clarify the pigmented epithelium. Eyes were cryoprotected in 20% sucrose, the anterior halves dissected away, and eyecups were embedded and frozen in OCT cryomolds. Eyecups were cryosectioned at 20 μm onto slides that contained one section from each timepoint for parallel staining and analysis; slides were stored at −20°C.

For bulk genomic analysis, eyes were collected at 9 a.m. and 11 p.m. (4-5 animals per timepoint), and the retinas were isolated in PBS. Retinas were snap frozen in microfuge tubes over liquid nitrogen and stored at −80°C until DNA extraction.

Three commercially-available ACD RNAscope™ probes were used. As a positive control, Dr-polr2a-C3 reported mRNA transcripts from a nuclear-encoded zebrafish housekeeping gene. As a negative control, 3-plex DapB reported transcripts from a bacterial gene. To explore mtDNA and mtRNA we used probes against the mitochondrial gene MT-ND5, which encodes a highly-conserved subunit of respiratory complex I and was identified by the manufacturer as a good candidate reporter. The “coding” probe Dr-MT-ND5 contains sequences antisense to the MT-ND5 gene, binding to regions of the noncoding (sense) mtDNA strand, and to MT-ND5 mtRNA transcripts.

To report mtDNA, we employed a strategy similar to one reported for chromogenic RNAscope™ with paraffin-embedded samples (Chen et al., 2020). The “noncoding” probe Dr-MT-ND5-sense-C3 was custom designed to contain sense sequences of the same mitochondrial gene; it binds regions of the coding (antisense) mtDNA strand. Antisense mtRNA transcripts from the noncoding strand of mtDNA are typically degraded (Pietras et al., 2018; Jedynak-Slyvka et al., 2021), so this probe reports primarily mtDNA.

Sections were treated with the RNAscope™ Fluorescent Multiplex v2 kit. The standard protocol for fixed frozen sections was used including pretreatment, probe hybridization and fluorescent labelling, with the following modifications: 1) Target retrieval was reduced to 5 min 2) Protease III digestion was reduced to 15 min 3) Probe and dye concentrations were optimized to best visualize single molecules (see Table 2). 4) Following the final wash, sections were not mounted and instead immediately underwent IHC. Incubations were conducted in staining chambers in a standard incubator. MT-ND5 probes were conjugated to the fluorescent dye Opal™ 520 diluted in RNAscope™ TSA buffer. Polr2a (C3) and DapB (C2) control probes were conjugated to Opal™ 520 and Opal™ 570, respectively.

To verify specificity of mtDNA and mtRNA detection with the MT-ND5 probes, an additional incubation in either PBS or RNase A was performed. Following post-protease water washes and just prior to probe hybridization, sections were incubated 30 min at 37°C in either PBS or 50 μg/mL RNase A in PBS, followed by 3 water washes. This was followed by standard probe hybridization, fluorescent labelling, and IHC. Additional validation strategies are outlined in the Section 3.

Following RNAscope™, sections underwent IHC using a primary antibody against MTCO1 (a subunit of respiratory complex IV) to report mitochondrial volume. Sections were washed 3 times with PBS, blocked 1 h at room temperature (RT) with blocking buffer, and incubated overnight at RT with anti-MTCO1 diluted in blocking buffer. They were then washed 3 times with PBS, incubated 2 h at RT with Alexa Fluor™ 633-conjugated secondary antibodies and DAPI counterstain, washed 3 times with PBS, and mounted with Vectashield and a coverslip. Table 1 lists primary antibodies used in this study, along with others the authors have found to be compatible for multiplexing IHC with RNAscope™.

Imaging was conducted using a Leica SP8 confocal microscope with a ×63 oil objective to capture single Z-stacks and Z-stack montages. LAS-X software was used for acquisition, and images were deconvolved using Leica Lightning in adaptive mode. Images were processed in ImageJ, where mitochondrial clusters of individual rods and short-wavelength cones were identified based on morphology and position in the retina (cyan and yellow boxes in Figure 2A). Single clusters were cropped out for quantification using the MTCO1 mitochondrial channel as a mask. Confocal images presented are maximum intensity projections over 12 µm tissue depth.

Mitochondrial clusters of rod photoreceptors were found to have well-resolved individual RNAscope™ MT-ND5 puncta. Single cluster Z-stacks (n = 3 cells per condition) were analyzed using the 3D Objects Counter plugin in ImageJ to report the number of puncta per cluster. MT-ND5 puntca in cone clusters were poorly resolved in areas due to very dense packing of mitochondria. Single cluster Z-stacks (n = 30 cells per standard timepoint, n = 12 cells per timepoint for RNase validation) were analyzed using the 3D Objects Counter plugin in ImageJ to instead report total puncta volume relative to overall mitochondrial volume measured using the MTCO1 channel. For both rods and cones, analysis was conducted blind to sample condition and timepoint. Mann-Whitney statistical tests were performed using Microsoft Excel with the Real Statistics resource pack.

DNA was isolated from frozen retinas (n = 3-4 animals per timepoint, each with three technical replicates) using an AllPrep DNA/RNA Mini Kit, then quality checked via spectrophotometer. mtDNA:gDNA ratios were determined using the qPCR 2^−ΔΔCT method (Livak and Schmittgen, 2001) relative to the 9 a.m. timepoint. Mitochondrial-encoded NADH dehydrogenase 1 (mt-nd1) served as the mtDNA target and polg1 was the nuclear-encoded DNA target, a strategy reported previously in zebrafish (Artuso et al., 2012). Primer sequences for these targets are listed in Table 2. qPCR CT values were generated using extracted DNA on the 7500 Fast Real-Time PCR System and SYBR Green Real-Time PCR Master Mix. The number of mtDNA copies per cell was determined according to the following equation (Artuso et al., 2012): m t D N A   c o p i e s = n u m b e r   o f   m t D N A   c o p i e s n u m b e r   o f   a l l e l e   o f   n D N A 2

To determine the relative contribution of RNA to probe signals, we employed an optional RNase incubation step prior to probe hybridization (Figure 1B). We used an RNAscope™ positive control probe against messenger RNA (mRNA) transcripts from the nuclear-encoded housekeeping gene Polr2a, and a negative control probe against mRNA transcripts from the bacterial gene DapB. In zebrafish retina sections, the Polr2a probe labeled puncta around nuclei that were absent with RNase pretreatment; no puncta were visible with the DapB probe (Figure 1C).

Further enzymatic validation using DNase to digest mtDNA has been reported for paraffin sections (Chen et al., 2020). We attempted this repeatedly with fixed frozen retina sections but results were difficult to interpret, as DNases removed most of the RNA signal from the positive control mRNA probe. One explanation is that structural damage to the tissue from DNase is more severe in fixed frozen sections, causing RNA to float away before probe hybridization. This is an important control that could be successful in other tissues, but likely necessitates paraffin sections.

To confirm that MT-ND5 probe signals were localized to mitochondria, we expanded the RNAscope™ procedure to include subsequent immunolabelling (Figure 1B). The pretreatment conditions required for RNAscope™ destroyed antigenicity of some primary antibody targets, while antigenicity was unaffected or improved for others. Table 1 lists primary antibodies we have used to successfully immunolabel various cell structures following ISH in zebrafish, macaque, or human tissue cryosections.

A primary antibody against MTCO1, a subunit of respiratory complex IV, displayed a robust mitochondrial staining pattern (Figure 1C, second row) following ISH that is typical of zebrafish retina stained using conventional IHC (Giarmarco et al., 2020). When ISH against MT-ND5 and IHC against MTCO1 were multiplexed, signals from coding and noncoding probes were confined to mitochondria (Figure 2A), demonstrating that the MT-ND5 probes are not binding off-target nuclear-encoded genes. Mitochondria in zebrafish photoreceptors are tightly packed and do not immunolabel uniformly (see U-shaped structures in Figure 1C, second row), but IHC is still useful for broad morphometrics like cluster size and outline.

Immunolabelling with antibodies against dsRNA or dsDNA is also useful for validation of mitochondrial probes. We did not employ this method in our samples because IHC quantification is confounded by limitations with antibody penetration. However, dual labeling with RNAscope™ and anti-dsRNA has been reported for cultured primate cells (Liu et al., 2018).

Given the difficulty using DNase as a control for RNAscope TM superscript on cryosections, bulk analysis using quantitative PCR (qPCR) can be used to confirm findings about mtDNA. By comparing the number of mitochondrial and nuclear genomes in a sample, mtDNA copy number per cell can be calculated (Artuso et al., 2012). While cell-specific information is lost, patterns in mtDNA copy numbers found using RNAscope TM superscript should be consistent with observations made using qPCR. We demonstrate validation using qPCR in a proof-of-principle experiment below (Figure 3C).