Enhanced single RNA imaging reveals dynamic gene expression in live animals

Imaging endogenous mRNAs in live animals is technically challenging. Here, we describe an MS2-based signal amplification with the Suntag system that enables live-cell RNA imaging of high temporal resolution and with 8xMS2 stem-loops, which overcomes the obstacle of inserting a 1300 nt 24xMS2 into the genome for the imaging of endogenous mRNAs. Using this tool, we were able to image the activation of gene expression and the dynamics of endogenous mRNAs in the epidermis of live C. elegans.


Introduction
RNAs relay genetic information from DNA to proteins or function by themself. Live-cell imaging of RNAs at a single-molecule level is crucial to uncovering their roles in gene expression regulation (Buxbaum et al., 2015). Various tools have been developed to visualize RNAs in live cells (Braselmann et al., 2020;Le et al., 2022), including RNA-binding protein-fluorescent protein approaches (Bertrand et al., 1998), CRISPR-based systems (Nelles et al., 2016;Yang et al., 2019), and those utilizing fluorophore-RNA aptamer pairs Paige et al., 2011;Sunbul et al., 2021). The MS2-based system is the most widely used and represents the current gold standard for singlemolecule RNA imaging in live cells (Braselmann et al., 2020;Bertrand et al., 1998). MS2 is a short RNA stem-loop bound specifically by the bacteriophage MS2 coat protein (MCP). To image RNA, 24xMS2 are placed at the 3′UTR or 5′UTR, and a fluorescent protein is fused to the MCP (MCP-FP). When coexpressed in cells, up to 48 fluorescent proteins (2 × 24, with two MCPs bound to one MS2) will be recruited to the RNA through MS2-MCP binding. This forms a fluorescent spot indicating a single RNA molecule (Braselmann et al., 2020).
The MS2 system has been successfully used to trace the whole mRNA life-cycle from transcription, to nuclear export, subcellular localization, translation, and to final degradation (Braselmann et al., 2020;Le et al., 2022). However, most RNA imaging studies in animal cells have been performed using exogenous mRNAs in cultured cell lines. 24xMS2 (about 1300 nt in length) have to be knocked into a specific genomic locus to image endogenous mRNA. The difficulty and low efficiency of the knock-in of long sequences into the genome represent a significant obstacle toward visualizing endogenous mRNA using the MS2 system. Thus, it is not surprising that less than 10 endogenous mRNAs have been imaged in live animal cells at a single-molecule level, and examples of endogenous mRNAs imaged in live animals remain extremely rare (Halstead et al., 2015;Levo et al., 2022;Dufourt et al., 2021;Park et al., 2014;Das et al., 2018;Zimyanin et al., 2008;Forrest and Gavis, 2003;Lee et al., 2022). Since overexpressed mRNAs may not faithfully recapitulate endogenous mRNA expression and dynamics, the development of more sensitive techniques for endogenous mRNA imaging is of great value.

Results
In this study, we reasoned that combining the MS2 with a signal amplifier may allow the recruitment of more fluorescent proteins to the RNA with fewer MS2 repeats (i.e., 8xMS2 -see Figure 1A). To achieve this, we combined the MS2 and Suntag systems. Suntag is a 19 amino acid protein tag that binds to its specific single-chain variable fragment (scFv) antibody (Tanenbaum et al., 2014). We fused MCP with a 24xSuntag array and linked scFv with sfGFP. When coexpressed in cells, one MS2 interacts with two MCP-24xSuntag molecules, further recruiting 2 × 24 GFP molecules ( Figure 1A). As the Suntag serves as a signal amplifier, the combined system was named as MS2-based signal Amplification with Suntag System (MASS). When an 8xMS2 is placed into the 3′UTR, up to 384 (2 × 8 × 24) GFP can then be tethered to a single mRNA through the MCP-Suntag-scFv-sfGFP interaction ( Figure 1A). This leads to the formation of an intense GFP spot associated with single mRNA, facilitating live RNA imaging.
As proof of concept, 8xMS2 V1 4 was fused to the 3′UTR of β-ACTIN mRNA and transfected into HeLa cells. When all the required elements of the MASS (MS2, MCP, 24xSuntag, and scFv antibody) were present, bright GFP foci were readily detected ( Figure 1B). As controls, no GFP foci were detected when omitting any one of these elements ( Figure 1B). With MCP-24xSuntag, an MCP molecule could be labeled with up to 24 GFPs. Under our imaging conditions (100-500 ms exposure time), MCP-24xSuntag particles were not detected ( Figure 1B), probably because MCP-24xSuntag are diffusing too fast to be imaged as a spot. Thus, the GFP foci clearly represent β-ACTIN RNA molecules.
In addition, we performed MASS combined with single-molecule RNA fluorescence in situ hybridization (smFISH) using a probe against the MS2 stem-loop and a probe against the linker region between the MS2 stem-loops ( Figure 1C, Figure 1-figure supplement 1A, B). We found that MASS detected a similar number of GFP foci compared to the spots detected by smFISH (Figure 1-figure supplement 1C, Figure 1-source data 1). Moreover, the majority of GFP foci (72%) colocalized with the smFISH spots of β-ACTIN-3′UTR-8xMS2 mRNAs ( Figure 1C, D, Figure 1-figure supplement 1B, Figure 1-source data 1). It is reported that not all MS2 stem-loop will be bound by the MCP . As only 8xMS2 was used in MASS, it is likely that some mRNAs were not fully bound by MCP and were not detected. On the other hand, in theory, only up to 16 probes will be hybridized with the 8xMS2 stem-loops and the linker regions in the smFISH experiment, and it is possible that some mRNAs were miss labeled by smFISH. Therefore, 100% colocalization of MASS foci with the smFISH spots was hard to achieve. Taken together, our data indicated that MASS is able to detect single mRNA molecules and label the majority of mRNAs from a specific gene in live cells.
It was reported that tagging mRNAs with MS2 stem-loops might affect mRNA stability, which could be counteracted by using improved versions of MS2 repeats (Li et al., 2022;Vera et al., 2019). We found GFP foci of β-ACTIN-3′UTR-8xMS2 mRNAs were detected regardless of using MS2-V1 4 or MS2-V7 21 with 2× tandem repeats of monomer MCPs (tdMCP) (Figure 1-figure supplement 2A). Therefore, MASS could be performed using different versions of MS2 stem-loops and MCPs. To directly test whether mRNA stability was affected while imaging by MASS, we examined the stability of three mRNAs: MYC, HSPA1A, and KIF18B, which were reported as medium stable mRNAs (Vera et al., 2019;Sharova et al., 2009). We found that the stability of those mRNAs was not significantly affected either by tagging of 8xMS2 V1 or by coexpression of the MASS imaging system (Figure 1-figure supplement 2B, C, Figure 1-source data 1). It is worth noting that we only examined three mRNAs in this study. The stability of specific mRNAs might be affected by MASS. If so, an improved version of MS2 should be used for the imaging experiment. It has been reported that β-ACTIN mRNAs with 3′UTR can localize to the lamellipodia (Katz et al., 2012). In support of this, we observed that the GFP foci of β-ACTIN-3′UTR-8XMS2 mRNAs were indeed localized to the lamellipodia in HeLa cells (Figure 1-figure supplement 3A and Video 1). To further test whether mRNA localization was affected by MASS, we imaged β-ACTIN-3′UTR mRNA with MASS or with the conventional 24xMS2 system in NIH/3T3 cells, which is a mouse fibroblast cell line. We found that GFP foci of β-ACTIN-3′UTR mRNAs detected by MASS or 24xMS2 system showed similar localization (Figure 1-figure supplement 3B). Thus, these data suggested that MASS did not affect RNA subcellular localization.
Haven established that MASS with MCP-24xSuntag could image single RNA molecules; we next sought to test whether MASS could be performed with shorter repeats of Suntag arrays (MCP-12xSuntag, MCP-6xSuntag) and compare MASS to the conventional 24xMS2 image system.
With the conventional 24xMS2 mRNA imaging system, a nuclear localization signal (NLS) was usually fused to MCP to localize NLS-MCP-GFP into GFP. NLS-MCP-GFP will be exported into the cytoplasm with mRNAs when binding to mRNAs. This strategy allows the detection of β-ACTIN-3′UTR mRNAs with a signal-to-noise ratio of 1.21 ( Figure 1-figure supplement 4A, B, Figure 1-source data 1). MASS with MCP-24xSuntag showed the highest signal-to-ratio of 1.79. MCP-12xSuntag and MCP-6xSuntag labeled β-ACTIN-3′UTR-8xMS2 mRNAs with a similar signal-to-noise ratio of 1.42 and 1.48, which are better than the conventional 24xMS2 system ( One critical concern about MASS is that intense tagging of mRNAs may affect the dynamics of mRNAs. To address this, we performed live-cell imaging of β-ACTIN mRNA using the conventional 24xMS2 system or MASS with different lengths of Suntag arrays (MCP-24xSuntag, MCP-12xSuntag, and MCP-6xSuntag). The velocity of mRNA movement in each imaging condition was measured. We found that compared to the conventional 24xMS2 system, mRNA labeled by MASS with MCP-24xSuntag or MCP-12xSuntag showed a stem-loops and probes against the linker region between the MS2 stem-loops in HeLa cells. Scale bar, 5 µm. See also Figure 1-figure supplement 1. (D) Quantification of the total and colocalized foci of β-ACTIN-8xMS2 mRNAs detected by smFISH or MASS with tdMCP-24xSuntag in HeLa cells. A total of 16 cells from three independent smFISH experiments were analyzed. See also Figure 1-source data 1. (E-H) Time-lapse imaging of β-ACTIN-8xMS2 mRNA dynamics in HeLa cells. sfGFP foci (β-ACTIN mRNAs) are shown. Constructs of β-ACTIN-8xMS2, MCP-24xSuntag, and scFv-sfGFP were cotransfected into HeLa cells. Images were taken 12 hr after transfection. (E) A fusion event of two sfGFP spots (white arrows). (F) A fission event: with large sfGFP foci split into three spots (white arrows). (G) Transient interactions of an sfGFP spot (yellow arrow) between two spots (white arrows). (H) An sfGFP spot showing no movement over a 10-s period (white arrow). Scale bars, 1 µm.
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. qRT-PCR (Quantitative Reverse Transcription PCR), number of foci detected by smFISH and MASS, signal-to-noise ratio, velocity, and intensity of the foci of mRNA detected in HeLa cells.       In contrast, mRNAs labeled by MASS with MCP-6xSuntag showed a similar velocity to that labeled with the conventional 24xMS2 system (Figure 1-figure supplement 5A, B, Figure 1-source data 1). Those data pointed out that when MASS is used to measure the speed of mRNA movement, a short Suntag array (MCP-6xSuntag) should be used. We next measured the average intensity of each GFP foci of β-ACTIN mRNA labeled using the conventional 24xMS2 system or MASS with different lengths of Suntag arrays (MCP-24xSuntag, MCP-12xSuntag, and MCP-6xSuntag). We found GFP foci detected by MASS showed higher intensity than those detected by the 24xMS2 system ( Figure 1figure supplement 6, Figure 1-source data 1). These data support that with MASS, a single mRNA molecule could be tethered with more GFP molecules, forming a GFP spot of higher fluorescence intensity. Such an application, retains the ability to image mRNA using low-power lasers, thus lowering any unwanted phototoxicity and photobleaching and still allowing the tracking of mRNA dynamics in high temporal resolution.
We then performed time-lapse imaging of the β-ACTIN-3′UTR-8XMS2 mRNA with a time interval of 1 s in HeLa cells (Video 2). We found that foci of mRNAs showed various dynamics: (1) Fusion. GFP spots fused into a more prominent spot ( Figure 1E and Video 3); (2) Fission. Large GFP foci split into smaller spots ( Figure 1F and Video 4); (3) Transient interaction. GFP foci touched each other briefly, then moved away ( Figure 1G and Video 5), suggesting there are dynamic RNA-RNA interactions in cells; (4) Dynamic movement or anchoring. Despite most foci of β-ACTIN-3′UTR-8XMS2 mRNAs showing dynamic movement in cells, other foci were far more static, showing little movement ( Figure 1H and Video 6), suggesting that these latter mRNAs may be anchored to subcellular structures.
Our ultimate goal was to develop tools for endogenous mRNA imaging in live animals. It has been previously reported that the knock-in of short sequences into the genome is far more efficient than those of longer sequences (Wang et al., 2022;Paix et al., 2017). The MASS exploits this advantage as only 8xMS2 (350 nt) needs to be inserted into a genomic locus, thus overcoming the previous obstacle of the requirement of inserting a long 1300 nt 24xMS2 into the genome for live-cell imaging of endogenous mRNA.
We then used the nematode C. elegans to specifically examine whether the MASS could be used for RNA imaging in live animals. An 8xMS2 was placed into the 3′UTR of cdc-42 mRNA ( Next, we set out to visualize gene expression activation and the dynamics of endogenous mRNAs in live animals. We used the skin of C. elegans as a model, which is composed of an epidermal epithelium with multiple nuclei. Upon wounding the epithelium via laser or needle, specific gene expressions and downstream signaling cascades for wound repair are then triggered and activated (Xu and Chisholm, 2014;Figure 2A). To this end, 8xMS2 was knocked into the 3′UTR region of two endogenous genes, C42D4.3 and mai-1 ( Figure 2B and Figure 2-figure supplement 2A), the expression levels of which were reported to increase significantly after wounding (Fu et al., 2020). MCP-24xSuntag and scFv-sfGFP were expressed in the epidermis with the tissue-specific promoter semo-1 and col-19. We then used a UV laser to injure an area of the epidermis. Prior to wounding, few sfGFP foci were detected in either wild-type (WT) or in 8xMS2 knock-in animals ( Figure 2B and  2A,. This result suggests that the wounding had activated C42D4.3 and mai-1 mRNA expression. In agreement with this, qRT-PCR showed that C42D4.3 and mai-1 mRNA levels were upregulated more than eightfold 15 min after injury ( Figure 2C and As the mRNA expression level of C42D4.3 and mai-1 significantly increased after wounding, we expected a boost in transcription. GFP will form extensive foci in the nuclei with active transcription. However, we failed to detect the appearance of bigger GFP foci in the nucleus. The epidermis of C. elegans is a syncytium with 139 nuclei located in different focal planes. With our microscopy, we could image only one focal plane, in which there are usually 4-10 nuclei. Therefore, it is likely that the nuclei with active transcription were out of focus, and therefore the GFP foci formed at the transcription site were not detected. Next, we tracked the dynamics of endogenous C42D4.3-8xMS2 mRNA in the C. elegans epidermis after wounding. We found that in proximity to the injury site GFP foci were detected as early as 1 min after wounding ( Figure 2D, Figure 2-figure supplement 3, and Videos 7 and 8). As a control, we pretreated the C. elegans with Actinomycin D, which potently inhibits gene transcription. In such cases, no GFP foci could be detected after wounding (Figure 2-figure supplement 4). This indicated that GFP foci are newly synthesized C42D4.3-8xMS2 mRNAs. Our data demonstrated that gene expression activation and transcription occur extremely fast (in this case, within 1 min after the stimulation). In addition, the appearance of GFP foci gradually spreads from the area around the injury site to distal regions. The total foci number steadily increased in the epidermis ( Figure 2D Figure 2-source data 1). These data suggested that C42D4.3 mRNAs undergo clustering after wounding and form RNA granules in vivo. In agreement with our observation, it has been previously reported that mRNAs formed large clusters and are co-translated in Drosophila embryos (Dufourt et al., 2021).

Discussion
It has been recently reported that the SunRISER and MoonRISER system by combination of PP7 and Suntag or Moontag enables imaging of single exogenous mRNAs in living cells. Through computational and experimental approaches, the authors further optimized the SunRISER system and showed that SunRISER provided an excellent approach for long-term imaging of overexpressed mRNA in living cells (Guo and Lee, 2022). The principle of the SunRISER system and MASS is similar, which uses a protein signal amplifier to generate a higher fluorescence signal for RNA imaging. In this study, we compared MASS to the conventional 24xMS2 mRNA imaging system and characterized the effect of MASS on mRNA stability and dynamics. In addition, we primarily explored the application of MASS to image endogenous RNA in live animals, which has not been tested in the previous study. Here, we showed such signal amplification strategy is valuable for imaging endogenous mRNAs that utilizing only 8xMS2 in comparison to the 24xMS2 used in the classic MS2-based live-cell RNA imaging. The advantage of a short MS2 is of prime benefits for imaging endogenous mRNA as it reduced difficulties involved in inserting a long 1300 nt 24xMS2 into a genomic locus. We expect this tool will help promote studies of RNA transcription, nuclear export, subcellular localization, translation, RNA sensing, and degradation, at the endogenous level in culture cells and live animals.

Cell lines
The HEK293T

Constructs for mammalian cells
All PCR reactions were performed using KOD One PCR Master Mix -Blue-(TOYOBO). Recombinational cloning was performed with the ClonExpress One Step Cloning Kit (Vazyme).
To make pcDNA-puro-BFP-β-ACTIN-3′UTR, a β-ACTIN coding sequence with a 3′UTR of 373 bp was PCR amplified from the cDNA of HEK293T/17 cells and inserted into the pcDNA-puro-BFP vector with KpnI and BamHI sites through recombinational cloning. The primers were β-ACTIN F and β-ACTIN R.
To make the pcDNA3.1-NLS-tdMCP-sfGFP, the NLS sequence was synthesized and inserted into the pcDNA3.1-tdMCP-sfGFP vector with NheI restriction sites. The sequence of NLS is as follows: ccaa aaaa gaaa agaa aagt t.

Constructs for C. elegans
Recombinational cloning was performed with the ClonExpress One-Step Cloning Kit (Vazyme).
Constructs will be deposited to Addgene.

Live-cell imaging
For images in Figure 1 and Videos 1-6: HeLa cells were plated on 35-mm glass-bottom dishes (NEST, 801001) at a density of 0.13 × 10 6 . The indicated constructs were transfected into HeLa cells. Twelve to twenty-four hours after transfection, cells were imaged using a spinning-disk confocal microscope with a ×60 objective (Nikon T2 Microscope; Apo ×60 oil; 1.4 NA) using the SoRa mode. Exposure time was 500 ms. For time-lapse imaging, the time interval was set to 1 or 2 s. Images were analyzed with FIJI (ImageJ).
For images shown in supplements: HeLa cells were plated on 35-mm glass-bottom dishes (BGI, BGX-03520-100) at a cell number of 0.8 × 10 6 to grow for 12 hr. The indicated constructs were transfected into HeLa cells. Twelve hours after transfection, cells were imaged by a spinning-disk confocal microscope with a ×60 objective (Olympus SpinSR10 Microscope; Apo ×60 oil; 1.5 NA) using the SoRa mode. Different parameters were used: Images shown in supplements: Exposure time was 500ms. Time-lapse imaging for calculating the speed of mRNA movement: Exposure time was 200 ms, and the interval was set to 217 ms. Fifteen frames are recorded. Z-stack imaging for calculating intensity and signal-to-noise: Exposure time was 500 ms. Stacks of 5 planes with a z-spacing of 0.5 μm were obtained by using range mode.

Single-molecule FISH and image acquisition in cells
Two 3′ Cy3 fluorescently labeled DNA oligos (probe-1: catg ggtg atcc tcat gt, probe-2: ttct agag tcga cctg ca) as smFISH probes against MS2 stem-loops and the linker regions were synthesized by Tsingke. HeLa cells were plated on 35-mm glass-bottom dishes (BGI, BGX-03520-100) at a density of 0.8 × 10 6 to grow for 12 hr and transfected with the indicated constructs. Twelve hours after transfection, cells were washed with phosphate-buffered saline (PBS), fixed with 2% formaldehyde at 37°C for 10 min, and washed twice each for 5 min with PBS. PBS was then discarded, and 2 ml 70% ethanol was added. The plates were kept at 4°C for 8 hr. The 70% ethanol was aspirated, 1 ml wash buffer was added (2× SSC (saline sodium citrate), 10% formamide in RNase-free water), and incubated at RT for 5 min. Hybridization mix was prepared by mixing 10% Dextran sulfate, 10% formamide, 2× SSC, 2 mM ribonucleoside vanadyl complex (NEB), 200 μg/ml yeast tRNA (Sigma, 10109495001), 10 nM probe-1, and 10 nM probe-2. To each plate, 800 μl hybridization mix was added and hybridized at 37°C for 16 hr. Fixed cells on plates were washed twice for 30 min (each time 15 min) with pre-warmed wash buffer (1 ml, 37°C) in the dark, followed by one quick wash with PBST, and kept in PBST for imaging within 3 hr. Images were captured using confocal ZEISS LSM 880 with Airyscan super-resolution mode.

Drug treatment
The Actinomycin D stock solution was dissolved in DMSO (Dimethyl sulfoxide) and diluted with M9 to a working concentration of 30 µM. Young adult stage worms were incubated in 100 µl Actinomycin D (APExBIO; Catalog No. A4448) solution (containing E. coli OP50) using a 1.5-ml microcentrifuge tube at 20°C for 3 hr. The worms were then transferred to fresh NGM plates to dry before wounding and imaging.

mRNA stability test in C. elegans
Worms (n = 200) of WT, C42D4.3-8xMS2 knock-in animals, animals expressing the MASS imaging system, and C42D4.3-8xMS2 knock-in animals expressing the MASS imaging system were incubated in 200 µl Actinomycin D solution (containing E. coil OP50) using the 1.5 ml tubes at 20°C for 0, 3, or 6 hr. The treated worms were used to extract RNA for qRT-PCR (Quantitative Reverse Transcription PCR) of C42D4.3 expression.

Quantification and statistical analysis in cells smFISH analysis
To quantify mRNA numbers detected by smFISH and MASS, also the ratio of colocalization, Fiji Plugin (https://imagej.net/imagej-wiki-static/Fiji)-ComDet was used for spot detection in two-color channel images. Data were analyzed in Excel, and the scatter diagram was generated using GraphPad Prism9.

The mRNA expression level by qRT-PCR
Statistical analyses were performed using GraphPad Prism 9. One-way ANOVA for multiple comparisons and non-paired t-tests were used for two comparisons. NS indicates not significant, *indicates p < 0.05, **indicates p < 0.01, *** indicates p < 0.001, **** indicates p < 0.0001. Unless elsewhere stated, bars represent means ± SD.

Intensity and signal-to-noise
Single-plane images were used for analysis using the Fiji plugin Trackmate (Tinevez et al., 2017). Particle size was estimated at 0.3 μm and a Differences of Gaussian (DoG) filter was applied to detect all spots. The 'Simple LAP tracker' particle-linking algorithm was used and the linking max distance was 15 μm, the gap closing distance was 15 μm, and the gap closing max frame was 2. The intensity and signal-to-noise ratio were produced by TrackMate and served as source data for the generation of histogram graphs by SPSS, and curve diagrams by Excel.

Velocity analysis
For the conventional 24xMS2 mRNA imaging system, an NLS was fused to MCP to localize NLS-MCP-GFP into the nucleus, which allows the detection of mRNAs in the cytoplasm with a high signalto-noise ratio. However, mRNAs in the nucleus cannot be detected clearly. Therefore, when the velocity analysis was performed, signals in the nucleus were excluded and only mRNA foci in the cytoplasm were included for analysis.
Single-molecule tracking was performed in 2D using the Fiji plugin Trackmate (Tinevez et al., 2017). To improve the accuracy of spot detection, we cropped the cytoplasm into several parts with no overlapping. Single particles were segmented frame-by-frame with a time interval of 217 ms which was set before starting the TrackMate. And the parameters for tracking single spots were the same as described in 'Intensity and signal-to-noise'. The velocity of each mRNA foci was produced by Track-Mate and served as source data for the generation of histogram graphs by SPSS, and curve diagrams by Excel.