Imaging of DNA and RNA in living eukaryotic cells to reveal spatio-temporal dynamics of gene expression

Abstract

This review focuses on imaging DNA and single RNA molecules in live cells in order to define the functional organization and dynamic processes in eukaryotic cells. Here, the latest advances in technologies to visualize individual DNA locus and RNAs in live eukaryotic cells in real time are discussed. Single-molecule fluorescence microscopy provides the spatial and temporal resolution to reveal mechanisms regulating fundamental cell functions. Novel insights on the regulation of the nuclear architecture, transcription, post-transcriptional RNA processing and RNA localization provided by multicolor fluorescence microscopy are reviewed. A perspective on the future use of live imaging technologies and overcoming their current limitations is provided.

1. INTRODUCTION

Eukaryotic cells are sophisticated multitasking units. They synthesize their components and control their quality, adapt to their microenvironment and can divide to create new cells. These actions demand extraordinary coordination occurring in a small volume (average of 85 μm³ in yeast and 5700 μm³ in neurons) packed with millions of molecules. Most knowledge about molecular interactions that lead to cell functionality originates from biochemical extraction of its components and performing assays in vitro to determine their functions. These approaches provide an overview of the cellular components in a population of cells but not their dynamics. In order to investigate when, where and how components of a cell interact in situ, imaging these molecular events in living cells is necessary. In this chapter, the insights obtained from individual cells about the regulation of two critical biomolecules- DNA and RNA are reviewed.

Recent advances in microscopy, and fluorescent tagging technologies to label DNA loci and single RNA molecules in live cells have opened avenues to validate the results from biochemical approaches and gain novel insights into the molecular behavior in their native environment. With the addition of time dimension, precise kinetics of these molecular interactions can be quantified. The regulation of gene expression in different subcellular compartments, like transcription in the nucleus (1–3), can now be visualized in real time in whole organisms (4, 5), and efforts are underway to study how nucleic acids mediate communication between the different compartments of the eukaryotic cell. Besides the well-defined compartments, the discovery of membraneless structures may facilitate molecular interactions in smaller microdomains in the nucleus and in the cytoplasm (6, 7). The fast dynamics of the assembly of these bimolecular condensates provide the means to regulate stochastic events, like transcription in nuclear hubs (8, 9), or localization of RNAs in stress granules (SGs) upon stress (10–12). Therefore, visualizing molecules in live cells is a unique approach to study cell biology because it provides the means to integrate the cellular organization and the kinetics of basic physiological processes in the life of a cell, like DNA replication and RNA transcription and translation.

2. DETECTION OF DNA AND RNA FOR REAL-TIME IMAGING

The ability to visualize specific DNA loci and RNA molecules in live cells relies heavily on optical advances in fluorescence microscopy and engineering proteins or oligonucleotides which are fluorescently tagged and bind the sequence of interest. One of the purposes to achieve specific locus labelling is to characterize the locus behavior in its native genomic environment and functionally correlate it with its transcriptional status with temporal resolution. Additionally, the spatial context of the locus demonstrates how the dynamic formation of transcriptional hubs in certain nuclear territories regulates transcription. To this end, imaging of single DNA locus should be accompanied by the visualization of the transcripts or the regulatory RNAs influencing the transcriptional outcome. In this section, approaches to label DNA and RNA are summarized. The detection of these labeled molecules with precision relies heavily on the latest advances in fluorescence microscopy, that have been described in great detail in other reviews (13). Depending on the specimen being imaged such as yeast, mammalian cells or developing Drosophila embryos, the microscope and the illumination conditions need to be optimized. Different imaging modalities are available. Epifluorescence microscopes with EMCCDs work effectively for mammalian cells to capture fast temporal events, although out of plane light increase the background and may limit signal to noise ratios (SNR). Strategies to limit out of plane fluorescence include total internal reflection fluorescence microscopy (TIRFM), highly inclined and laminated optical sheet (HILO) illumination and spinning disk confocal microscopy. TIRFM and HILO have been widely used to image transcription factors dynamics, however, these approaches have limited depth of illumination. To achieve high sample penetration depth particularly for tissue or whole organisms, light-sheet fluorescence microscopy (LSFM) is the method of choice, where a focused sheet of light is used to illuminate only a thin section of a sample(14). This structured light sheet increases acquisition speed and minimizes phototoxic damage to the cells, thereby enabling imaging of 3-dimensional dynamics across large volumes (15). Therefore, by a combination of microscopes and detectors, one can visualize both long-lasting events or processes that require fast image acquisition.

Detection of specific DNA loci or RNA with high SNR is based on multimerizing fluorescent molecules. Alternative strategies to improve SNR involve decreasing the fluorescence background by using fluorogenic aptamers and dyes, which are non-emitters in their non-bound state, and fluoresce only when bound to target nucleic acid sequences. As novel technologies for specific targeting emerge, combination of new and existing approaches provide the means to interrogate DNA and RNA dynamics in live cells. Some of these approaches have provided single-molecule resolution (listed in Table 1), despite the challenges of non-repetitive DNA sequences and individual RNA molecules.

Table 1.

Summary of molecules that have been labeled for live imaging, the labeling method used, and the biological process that was reported to be visualized

Category	Target	Single molecule detection	Biological phenomenon	Method(s) and Reference(s)
IncRNA	Xist	Yes	Chromatin dynamics	MS2-MCP (98), PUM-HD (99)
	TERRA	Yes	Chromatin dynamics	PUM-HD (71, 72), MS2-MCP (73, 74)
	NEAT1	Yes	Localization and foci detection	MB (157)
Small ncRNA	miRNA	Theoretically possible	Localization and foci detection	Activity-dependent modification (129)
	miRNA	No	Localization and foci detection	MB (158), Riboglow (55)
	Ribosomal RNA	No	Localization and foci detection	Aptamer (54)
	U6 RNA	No	Localization and foci detection	Aptamer (54)
	scaRNA	No	Localization and foci detection	Aptamer (54)
	mtRNA	No	Localization and foci detection	PUM-HD (159)
	snoRNA (snR30)	No	Localization and foci detection	Aptamer (160)
	U1 RNA	No	Localization in U bodies	Riboglow (55)
	U3 small nuclear RNA and 28S ribosomal RNA	No	Cellular and subcellular localization	ECHO-liveFISH in tissue (63)
mRNA	mRNA	Yes	Transcription, mRNA export, and localization	MS2-MCP (45), PUM-HD (64), MB (60), aptamer (115)
	mRNA	No	RCas9	RCas9 (65), Cas13a (66)
DNA	Xic	Yes	Chromatin dynamics	tetO/TetR system (95)
	Replication fork	Yes	Chromatin dynamics	TALE (67), tetO/TetR, and lacO/LacI systems (68, 69)
	Telomere	Yes	Chromatin dynamics	lacO/LacI system (70)
	Ribosomal DNA condensation	No	Chromatin dynamics	CRISPR (161)
	Chromosomal locus	Yes	Chromatin dynamics	dCas9-SunTag system (160a), CRISPR liveFISH (33)
	Loci interactions	Yes (single locus detection)	Chromatin compaction, decompaction, and interaction	tetO/TetR and lacO/LacI systems (4, 91), CRISPR (20), and CRISPRainbow (29)

2.1. DNA: Loci Tagging and Labeling

Chromatin architecture and dynamics have critical roles in the spatiotemporal regulation of gene expression in all eukaryotic cells. Although the spatial organization of a specific locus can be characterized using fluorescent in situ hybridization (FISH) on fixed cells, the dynamic interactions of chromatin during different stages of gene expression can only be addressed using live imaging methods. For a long time, these live imaging approaches mostly relied on the use of large arrays of the lac operator/Lac repressor (lacO/LacI) system, in which up to 256 lac operator (lacO) repeats were inserted into the genome. The fluorescently tagged lacO binding protein, Lac repressor, bound to these repeats with high affinity and specificity, providing the means to image chromatin dynamics in living cells and organisms (16–18). The analogous tet operator/Tet repressor (tetO/TetR) system employs a similar strategy (Figure 1a, i). These arrays can be inserted into any DNA region of interest and are powerful approaches to image chromosome dynamics over long timescales, such as during cell division.With the advent of the clustered regularly interspaced short palindromic repeat–CRISPR-associated protein 9 (CRISPR-Cas9) system, these repeats could be targeted to specific genomic loci (19). Although these are robust approaches with high SNRs, the insertion of these long arrays could affect chromatin structure and the function of the targeted loci (20). This problem was circumvented by programming DNA-binding proteins, such as zinc fingers or transcription activator-like effectors (TALEs), to bind to specific DNA sequences. These proteins are fused to a fluorescent protein or quantum dots and work effectively for labeling repetitive DNA (21–23) (Figure 1a, ii). However, their low SNR and inefficient multimerization limit their use for labelingDNA. In recent years, the CRISPR-Cas9 system has proved effective at directing the targeting and labeling of specific chromosomal loci with high precision. The first study used an enhanced green fluorescent protein (EGFP)-tagged catalytically inactive Cas9 (dCas9) protein and a structurally optimized small guide RNA (sgRNA) to image repetitive elements in telomeres and coding genes (24) (Figure 1a, iii). While this method allowed for visualization of telomere dynamics and Mucin 4 [a highly repetitive sequence (25)], its ability to label nonrepetitive genomic sequences was not robust. The dCas9 system led to a gamut of labeling strategies for imaging any loci of choice in various cell types (26).

Figure 1. Fluorescent imaging approaches for DNA/RNA detection in living cells.

(a) Detection of genomic DNA at single-locus resolution in living cells. (i) Fluorescent protein (FP)-labeled operator-binding proteins are tethered to multiple repeats of an operator inserted into a desired DNA locus (e.g., lac operator/Lac repressor and tet operator/Tet repressor systems). (ii) Transcription activator-like effectors (TALEs): TALEs are DNA-binding proteins that can be designed to target any DNA sequence (21). Customized FP-tagged TALEs can be tethered to specific endogenous genomic loci and imaged. (iii) Clustered regularly interspaced short palindromic repeat (CRISPR)-catalytically inactive Cas9 (dCas9): FP-tagged dCas9 is tethered to specific endogenous genomic loci that are recognized with specific guide RNA (gRNA). (b) Single-molecule RNA detection in living cells. (i) The stem-loop/coat protein system is composed of a specific RNA stem-loop structure inserted into desired mRNAs [e.g., MS2–MS2 coat protein (MCP), PP7–PP7 coat protein (PCP), λN-boxB] and an FP-labeled RNA-binding protein (coat protein) binding to the stem-loops with high affinity and specificity. (ii) Green fluorescent protein (GFP)-mimic aptamer: RNA aptamers inserted into desired mRNAs are labeled with fluorogenic ligands (fluorogens) that specifically recognize the aptamer (e.g., Spinach,Mango). (iii) CRISPR-RCas9 or Cas13a: This approach is based on CRISPR-dCas9 for DNA labeling, but RNA-targeting Cas9 (RCas9) or Cas13a is used for the specific labeling of desired endogenous mRNAs. (iv) Pumilio homology domain (PUM-HD): PUM-HD is an RNA-binding protein domain that can be designed to bind any eight-base RNA sequence. A pair of adjacent endogenous RNA sequences is labeled with biomolecular fluorescence complementation using designed PUM-HD tagged with FP fragments. (v) Oligonucleotide-based: Also known as fluorescence in vivo hybridization, complementary oligonucleotides with synthetic fluorescence dyes target desired mRNAs with RNA–DNA base paring (e.g., molecular beacons, forced intercalation probes, and quenched autoligation probes) (62).

(B) Single-molecule RNA detection in living cells. (a) Stem loop/Coat protein: This system is comprised of specific RNA stem loop structure inserted into desired mRNAs (e.g. MS2-MCP, PP7-PCP, λN-boxB), and FP-labeled RNA binding protein (Coat protein) binding to the stem loops with high affinity and specificity. (b) GFP-mimic aptamer: RNA aptamers inserted into desired mRNAs are labeled with fluorogenic ligands (Fluorogen) that specifically recognize the aptamer (e.g. Spinach, Mango) (c) CRISPR/rCas9 or Cas13a: This approach is based on CRISPR/dCas9 for DNA labeling but RNA-targeting Cas9 (RCas) or Cas13a is used for the specific labeling of desired endogenous mRNAs. (d) PUM-HD (Pumilio homology domain): PUM-HD is an RNA-binding protein domain that can be designed to bind any 8-base RNA sequence. A pair of adjacent endogenous RNA sequences will be labelled with biomolecular fluorescence complementation using designed PUM-HD tagged with fluorescent protein fragments. (e) Oligonucleotide-based: Also known as Fluorescence In Vivo Hybridization (FIVH), complementary oligonucleotides with synthetic fluorescence dyes target desired mRNA with RNA-DNA base paring (e.g. molecular beacons, FIT probes, QUAL probes) (62).

The combination of CRISPR-Cas9 with the bacteriophage-derived RNA stem-loop motifs MS2 and PP7 (see Section 2.2) made it possible to simultaneously visualize two distinct genomic loci. The design involved sgRNAs containing MS2 and PP7 aptamers that are bound by fluorescently tagged MS2 capsid protein (MCP) and PP7 coat protein (PCP), respectively (20, 27). This technology labels an individual chromosomal locus with as few as four unique sgRNAs and renders nonrepetitive regions of the genome visible for live imaging. For a more flexible labeling of multiple genomic loci in living cells, two adaptations of the CRISPR-Cas9 system have been made: (a) multicolor versions of CRISPR specifically using dCas9 from three bacterial orthologues (28) and (b) engineered sgRNAs that can bind to combinations of different fluorescent proteins, also known as the CRISPRainbow system (29) (Table 1). CRISPRainbow utilizes sgRNAs that have been genetically modified at their 3’end to be tagged with a pair of any of the three RNA aptamers (MS2, PP7, and λN-boxB). The cognate coat proteins are fused to spectrally distinct fluorescent proteins, allowing each sgRNA to be bound by a pair of the same or different coat proteins. Depending on the stem-loop pair on the RNA, up to six combinations of colors can be generated using primary fluorophores (red, green, and blue) when sgRNAs have identical aptamer pairs, or three secondary colors can be generated using pairs of different aptamer sequences that bind to two distinct proteins. This multiplexing approach enables the simultaneous imaging of up to six chromosomal loci in living cells and reveals their dynamic interactions. While CRISPRainbow represents a significant technological advancement for genomic DNA imaging, the unbound sgRNAs that contribute to the background, as well as the variable quantum yield of different fluorescent proteins, may limit its utility for long-term imaging with high sensitivity in living cells. An improved multicolor CRISPR-based imaging system increases the stability of the guide RNAs (gRNAs) by inserting octet arrays of stem-loop aptamers at the 3’end of the gRNA scaffold. This CRISPR-Sirius system confers multiple loci tracking with increased brightness, a big technological leap for imaging low-copy genomic loci or single-copy genes (30).

Approaches employing dCas9 and an engineered sgRNA harboring a unique molecular beacon (MB) target sequence (sgRNA-MTS), may provide more flexibility and lower background (31). MB is a RNA labeling method using single stranded oligonucleotide probes with a fluorophore and a quencher at the two ends (see RNA labelling below). In contrast to the sgRNA-MTS approach which detects highly repetitive elements, the most recent version of the CRISPR/MB system, CRISPR/dual-FRET MB, enables imaging of non-repetitive locus with higher specificity (32). The binding of two distinct MBs with FRET fluorophore pair to the same gRNA causes the FRET emission. Only when two MBs bind to gRNA which targets the specific genomic locus, this approach achieves imaging of non-repetitive genomic locus with higher specificity. CRISPR/MB system may provide the flexibility in choosing any combination of target sequence and MBs, however, the efficacy of hybridization and stability of the MBs need to be characterized in detail.

Instead of expressing CRISPR components in live cells, dCas9-EGFP and Cy3-labeled gRNA can be assembled in vitro for form complexes called fluorescent ribonucleoproteins (fRNPs), which were delivered into the cells using electroporation (33). Higher SNR was achieved by the stabilization of DNA bound fRNAs while the most unbound gRNAs were degraded within 4 hours. This LiveFISH method was applied to label DNA and RNA using dCas9 and dCas13 respectively, enabling simultaneous visualization of genomic DNA and RNA transcripts in living cells. The dual DNA-RNA LiveFISH approach combined with other genetic manipulation technologies will prove to be a powerful tool to interrogate the causal relationships between genome organization and transcription dynamics.

2.2. RNA: mRNA and Noncoding RNAs

mRNA was the first molecule to be viewed at single-molecule resolution in living cells through the use of a strategy to introduce a tandem array of bacteriophage-derived stem-loops (referred to as MS2) into the endogenous locus of a gene (34).MS2 is an RNA aptamer that can be multimerized and integrated within the intron, coding sequence, or untranslated region (UTR) of a transcript as an array. Most studies utilize arrays of 24 stem-loops at the 3’UTR; however, array lengths may vary from 12 to 128 repeats, depending on the gene of interest, sensitivity of detection, and timescales of imaging. Each stem-loop is tightly bound by the coat protein (MCP) and fused to fluorescent proteins. MS2–MCP–fluorescent protein binding renders a single mRNA visible because it is decorated with multiple fluorescent molecules (Figure 1b, i). Several versions of the MCP–fluorescent protein have been used, with the best so far being a tandem MCP fused to two fluorescent proteins for bright and homogenous labeling of mRNAs (35). Tagging methods analogous to the MS2-MCP system, such as PP7, U1, and λN-boxB, have been developed over the years, conferring the ability for multiplexed mRNA imaging in living cells (36–38). Owing to the central role ofmRNAin gene expression, quantification of the imaging data has provided unprecedented insights into the regulation of transcription,mRNA export and localization, and,more recently, translation and decay (1, 2, 36, 38–43). The MS2 and PP7 systems have been reengineered over the years to accommodate the demands of specific biological questions (44, 45). One such example is the latest MS2 version with reduced affinity for the MCP to overcome the challenges of degradation of the MS2-MCP cassette, as observed in yeast (45–48). Although this tagging technology requires introducing MS2 aptamer repeats into RNAs, these reporter systems have faithfully captured the fate of single mRNAs without significantly perturbing their endogenous regulation. In fact, several mouse models of genes tagged with either the MS2 or PP7 technology are available that do not show any effect on gene function; these can now be used to monitor the behavior of mRNAs in living tissue (49–51). Nonetheless, genetic modifications must insert the cassette into the genome to tag the transcribed mRNA and must express the fluorescent binding protein. These requirements preclude the tagging of small noncoding RNAs (ncRNAs) that will experience an unnatural increase in size by more than a kilobase that will probably interfere with their proper folding, processing, and biological function.

An alternative strategy to eliminate the protein component of the MS2-MCP and its equivalent systems was to make an RNA aptamer that could fluoresce on its own. Spinach and Broccoli aptamers were the first iterations (52, 53) but these RNA mimics of GFP have poor folding properties in vivo, low quantum yields, and rapid photobleaching.Two recent approaches that partially overcame these caveats are the Mango III fluorogenic aptamer and the Riboglow modular system (Figure 1b, ii). Mango III is a closed RNA stem with a space to bind with high affinity to thiazole orange 1 dye–biotin and increase its quantum yield to 1,000 times the initial yield when it binds with its ligand. It is more photostable than the aptamer Broccoli and has been successfully used to localize RNA polymerase III (RNAP III)-dependent transcripts such as 5S and U6 RNAs (54). The Riboglow system is based on the bacterial cobalamin riboswitch RNA. Cobalamin quenches the fluorescence of a synthetic fluorophore until it is bound to the RNA riboswitch (55). The increase in the quantum yield of the fluorophore probes and their stability enables it to localize U1 small nuclear RNP in Cajal bodies in the nucleus and U1 bodies in the cytoplasm.Thus, Riboglow provides an alternative system to visualize only highly expressed ncRNAs in live mammalian cells. Two new fluorogenic aptamers, both called Peppers, have been recently developed to either increase the spectral range of RNA labeling (56) or introduce a fluorogenic protein component that is stabilized by RNA binding (57). In general, fluorogenic aptamers are not as bright as fluorescent coat proteins, and, as with the MS2 system, they demand genetic insertion to tag the RNA under investigation.

Several systems attempt to circumvent the problem of genetic insertion to tag endogenous RNAs. Examples are oligonucleotide-based RNA detection (also known as fluorescence in vivo hybridization), pumilio homology domain (PUM-HD), and the RNA-targeting Cas9 (RCas9) or Cas13a system (Figure 1b (iii–v; Table 1). One oligonucleotide-based approach is theMB, which contains a fluorophore or a quencher at each end of the oligonucleotide (58). In the absence of its target sequence, MBs have very low background fluorescence due to the formation of a stable duplex system whereby the quencher significantly quenches the fluorophore. Upon hybridization of theMB to a complementary sequence, the duplex is disrupted, leading to fluorescence emission. The latest version is the ratiometric bimolecular beacons (RBMBs), which are two-component MBs (59). The first component is a stem-loop-forming oligonucleotide with one fluorophore that is quenched by the second component, an antisense oligonucleotide of the stem tagged at both ends.The end closest to the stem-loop has the quencher, and the distal one has a spectrally different fluorophore. Binding of the RBMB to the target mRNA switches the labeling of the RBMB from one to two fluorophores. This ratiometric approach decreases the false positive signals obtained from mislocalization of MBs in the nucleus (60). RBMBs have been successfully used to track the motion of single RNAs in living cells (61). Several other oligonucleotide-base approaches have been attempted, such as forced intercalation probes and quenched autoligation probes (62). A recent example is the visualization of the 28S ribosomal RNA and U3 snRNA intranuclear foci dynamics in the brain after in vivo electroporation of exciton-controlled hybridizationsensitive fluorescent oligonucleotide (ECHO)-liveFISH probes (63). General limitations of oligonucleotide-based RNA detection approaches are the low SNR, the requirement for injection of fluorescence-labeled oligonucleotide, the trapping of probes by endosomes, and the low target specificity.

An alternative approach for labeling an endogenous RNA sequence is the PUM-HD system, a programmable RNA binding protein that targets any 8-base RNA sequence (64). Since it recognizes a short RNA sequence, the specificity of labelling and the possibility of off-target effects are the limitation of this approach. Specificity has been achieved with the catalytically inactive Cas13a system and the RCas9 by using a mismatch PAM as part of the oligonucleotide that binds the RNA sequence. These systems do not elicit the activity of RNAseH and reports on mRNA localization through a fluorescently tagged Cas9 (65, 66). The caveat of MB, PUM-HD and RCas9 systems is that they cannot achieve single molecule resolution unless several of these molecules are delivered to the cell and bind to specific sequences of the same mRNA. Hence, although promising for RNA tagging, much work needs to be done in order to improve their brightness and make small RNAs visible.

Overall, there is a plethora of systems engineered to visualize RNAs during their life cycle, and their use depends on the experimental requirements that will answer a specific biological question. As of today, the most used RNA tagging systems are the MS2 and analogs because they provide long-term single-molecule tracking. This review provides several examples of live imaging technologies used to investigate posttranscriptional RNA processing, localization, and fate and to address fundamental cellular processes such as chromatin remodeling, nuclear organization, and transcription (Figure 2).

Figure 2. Cellular processes addressed using single-molecule imaging approaches.

Single-molecule imaging approaches have been applied to visualize the processes occurring in the nucleus and cytoplasm: remodeling of chromatin and RNA metabolism. The illustration summarizes the cellular processes characterizing the molecular dynamics, kinetics, and temporal order of events using live imaging approaches. The overall life cycle of the mRNA from step (1), transcription, to step (4) , decay, in the different subcellular compartments is shown. Translation may occur before or after step (3), mRNA localization, depending on the mRNA and the cell response to environmental clues. The inset for chromatin remodeling shows a magnified view of the nucleus, where DNA associated with histones (nucleosomes) is organized into topologically associated domains (TADs). These sequences within TADs self-interact with each other more frequently than do sequences outside TADs. TADs may be repressed, often when they are associated with the nuclear lamina, or they may become more flexible domains regulating transcription (active TADs). Following decompaction of the chromatin, active transcription from multiple genes may occur simultaneously (e.g., Genes 2 and 3) when they share the same transcriptional hub. The shaded green area indicates a protein-rich core, housing transcription factors (TFs), mediators, and RNA polymerase II (RNAP II).

3. DNA REPLICATION AND TELOMERE DYNAMICS

DNA replication is a fundamental process in proliferating cells. The precise duplication of DNA prior to mitosis is tightly controlled during development and cell differentiation. In Drosophila, it is known that embryonic developmental progression extends the duration of the cell cycle. Using the TALE system to tag two specific DNA satellite sequences, one in the X chromosome and another in each of the two autosomes, showed that developmental progression altered the timing of the replication of those two satellites independently (67). Live DNA imaging shed light into the heterogeneity of the replication rates as fly development progressed, i.e. although both satellite sequences showed an overall delayed timing of initiation and increased duration of replication with mitotic cycles, the replication rates of the two satellite sequences varied depending on the cycle number. During mitotic cell division, a precise program controls the initiation timing of different origins of replication. In single DT40 chicken cells, the activity of six single loci was monitored using the TetO/TetR system (68). The quantitative detection of the allelic asynchrony at six single loci revealed the stochasticity of the programed replication initiation timing. In addition, the correlation between nuclear positioning and the allelic synchrony in late replicating loci suggested the strict control of replication timing close to nuclear lamina.

The replication rates of single replicating forks were measured in yeast using the labelling of two adjacent DNA loci with the LacO/GFP-LacI and the TetO/tdTomato-TetR systems (69). The study found the impact of major leading and lagging replication factors on the replication rates of single replicating forks. Fast temporal imaging of these loci showed that the replication of a 30 kb sequence took between 12 and 15 minutes, as calculated from the difference in the duplication time between the fluorescent signals. Unexpectedly, the polymerase accessory subunits had limited contribution in determining replication rates, but maturation of the Okazaki fragment was crucial for the fork progression (69).

Each cell division is accompanied by the shortening of the telomeres. Telomeres are unique structures that cap the end of chromosomes and protect their ends from sticking to each other. The telomere length depends on the addition of the telomere repeat sequence by the ribonucleoprotein telomerase. Telomerase has a ncRNA (telomerase component 1 (TLC1)), that is used as a template with binding protein(s), that catalyze the addition of repeat sequence to the 3’end of telomeres. The telomerase recruitment to the telomeres during the cell cycle was tracked by tagging TLC1 with the MS2-MCP system (70). Single molecule tracking of TLC1 molecules showed that they freely diffuse during G1 and G2 phases of the cell cycle but they colocalize with telomere loci, labeled with the LacO/LacI system, in late S phase. Telomerase associates longer (up to 45 s) at the telomere in S phase than in G1 phase (5 s). The telomere sequence, a six nucleotide repeat, is also transcribed into a functional long ncRNA, the telomeric repeat-containing RNA (TERRA). TERRA participates in the telomere biogenesis inducing its heterochromatin formation and capping of the telomere. Imaging approaches have addressed TERRA localization and molecular dynamics in human cells (71, 72) and yeast (73, 74). Single TERRA molecules were visualized with either the MS2-MCP system or the PUM-HD system against the repetitive UUAGGG sequence. TERRA is transcribed from short telomeres (73), it freely diffused and formed foci and transiently co-localized with the nuclear ribonucleoprotein hnRNPA1 (71) and the telomere (72, 73). Although these reports identified TERRA foci and its colocalization with some regulatory factors, additional approaches are needed to understand the diverse functions of TERRA on the telomere biogenesis.

DNA replication and the complex dynamics of factors affecting chromosome integrity advocate for a regulatory role of nuclear organization as well as the roles of specific proteins and ncRNAs.

4. SPATIOTEMPORAL DYNAMICS OF GENE EXPRESSION IN THE NUCLEUS

4.1. Chromatin Architecture Regulates Gene Expression

In the nucleus, the genome is highly organized and compacted into the 3D volume to form higherorder chromatin structures (75, 76). The configuration of the chromatin is regulated in a dynamic manner and determines the spatial and temporal features of replication, transcription, and gene silencing. In this section, we review the different examples of how chromatin organization affects loci interactions and gene activation.

4.1.1. Higher-order chromatin structure and heterochromatin organization

Chromatin organization occurs at multiple levels, from nucleosomes (DNA wrapped around histones) to epigenetic modifications of histones, higher-order domains such as heterochromatin and euchromatin, and finally the physical movement of chromatin and long-range interactions.Technologies such as chromosome conformation capture (3C and Hi-C) have yielded high-resolution contact probability maps of the genome (77).This has led to the identification of self-interacting genomic regions known as topologically associated domains (TADs) in various cell types (78).TADs play instructive roles in transcriptional regulation, and direct visualization of TAD organization at gene clusters (such as the HoxD locus in embryonic stem cells) has been done in fixed cells (79, 80). How these TADs are formed and maintained by a functional interplay of cohesins and CCCTC binding factors (CTCFs) is only beginning to be visualized in real time by imaging of the factors (81). Recent work has highlighted how these CTCFs self-organize and form chromatin loops in a locus-specific mechanism, a process potentially regulated by RNA or RNAs (82).

Progress has also been made in imaging single nucleosomes by single-molecule tracking photoactivated localization microscopy (spt-PALM), exhibiting that they form compact domains, frequently in the heterochromatin region (83). These transcriptionally repressed heterochromatin domains are epigenetically defined by the methylation of histone H3 at lysine 9 and recruitment of its binding partner HP1α. Real-time imaging of HP1α in cells along with a transgene array labeled with both LacO and MS2 system allowed monitoring the dynamics of HP1α association with the chromatin and the impact on transcription (84). The study characterized how the dynamic loss of HP1α from the heterochromatin and an exchange of histones mediates transition into a transcriptionally active state. More studies on the dynamic behavior of HP1α protein have revealed that the protein can undergo liquid-liquid demixing in vitro and phase separate to form heterochromatin domains (85). The domain formation is sensitive to disruption of weak hydrophobic interactions and exhibit restricted diffusion of molecules. These features are conserved across different isoforms of human HP1 and dependent on the phosphorylation status of the protein (86). It remains to be elucidated how the nuclear environment promotes such domain formation, or whether it is an intrinsic property of these proteins. In the future, applying these technologies correlating HP1α clustering dynamics with the genomic loci will be powerful tools to resolve how higher order nuclear organization is achieved, and its effect on transcription.

4.1.2. Chromosome, decompaction, interactions, and gene expression

The physical interactions between DNA regions, particularly between enhancers and promoters, are known to impact gene expression, as shown by 3C and Hi-C studies (77, 87). These interactions can occur both in cis and in trans. Trans interactions, in which the enhancer of one chromosome interacts with the promoter of another chromosome, are an example of long-distance interactions in the nucleus. Recent work has shown that insulators play an instructive role in aligning the two alleles and facilitate this transcoactivation. This phenomenon, known as transvection, has been imaged in Drosophila by insertingMS2 and PP7 stem-loops into reporter mRNAs transcribed from homologous alleles on complementary chromosomes (88, 89). One of the alleles has the enhancer, which is shared, resulting in coactivation of PP7 and MS2 reporter genes both in cis and in trans. This sharing of resources during coactivation of genes aligns with the emerging concepts of formation of transcriptional hubs with observed clustering of RNAP II and associated activators.

To understand the causal relationship between macroscale 3D organization and gene expression, a programmable 3D genome organizer named CRISPR-GO has been designed by coupling the CRISPR-dCas9 system with nuclear compartment-specific proteins (90). This allowed an inducible system for dynamic repositioning of specific genomic loci to different nuclear compartments such as the nuclear periphery, Cajal bodies and PML bodies. By combining CRISPR-GO with live-cell CRISPR-Cas9 imaging, the real-time interactions of chromatin with the nuclear compartments was elucidated- for example, targeting loci to Cajal bodies repressed gene expression of not only the reporters, but also the distal endogenous genes. This study shed light on how genome positioning and the environment may be instrumental in determining the transcription from these loci. While the macroscopic interactions could be mapped out, the higher-order interactions enabling transcription from one compartment versus the other remains to be characterized.

Dynamic changes in chromatin architecture and chromosomal localization are related to the transcriptional status of genes. To quantify the changes in chromatin conformation in response to transcriptional activation, compaction of the Gal locus was visualized in single live S. cerevisiae (91). The Gal locus extends for 31 kb in the DNA and contains the GAL7, GAL10 and GAL1 genes, which are transcriptionally active when the carbon source in growth media is changed to galactose. Impact of transcription on chromatin decompaction was measured by inserting an array of LacO loops labelled with GFP and an array of TetO sequences labelled with mCherry in two different positions on the GAL locus. The intermolecular distance between the two fluorescent foci doubled during transcriptional activation compared to inactivate states suggesting ongoing chromatin de-compaction during this process. This de-compaction was triggered by nucleosome eviction by the SWI/SNF nucleosome remodeler complex and by transcription initiation. Although the direct role of transcription progression on chromatin structure remains to be deciphered this result indicates the absence of promoter-terminator looping in the GAL locus. Unexpectedly, histone acetylation affected neither transcription nor the level of chromatin compaction. These findings differ from a previous study showing that the acetylation of histone 3 at lysine27 (H3K27ac) primes a robust transcriptional response of the glucocorticoid receptor (GR) in mammalian cells upon hormone stimulation (92). The dynamics of histone modifications and RNAPII initiation and elongation were detected by using labelled antigen-binding fragments (Fabs) on an array of the GR receptor integrated in the genome. Single live cell imaging showed that H3K27ac favors the recruitment of the transcription factor GR and accelerates the escape of RNAPII from the promoter.

The above mentioned publications were designed to achieve different goals. While the work done on the GAL genes offers a better spatial resolution on chromatin structure, the one done on the GR offers a better time resolution of the events that lead to an efficient transcription. The discrepancies between both studies on the role of H3 acetylation on transcription could be explained by the use of different model organisms. It is possible that posttranslational histone modification may be critical to arrange the spatial compaction of the chromatin in mammalian cells because their nuclear organization is more complex than yeast. Alternatively, the dramatic burst of transcription of the GAL genes might be higher than that of the GR and therefore, more dependent on histone remodelers than modifiers. Finally, the GAL10 gene antisense strand is transcribed into a ncRNA, which prevents transcriptional leakage under repression conditions (93). Thus, gene transcriptional activity might be differentially influenced by histone modification for transcription of nuclear ncRNAs and chromatin organization. This idea is supported by recent work showing that the establishment of neuronal surface identity depends on a lncRNA that leads to DNA methylation and promotes the choice for clusters of protocadherin alpha genes (94).

4.1.3. X chromosome inactivation

X chromosome inactivation (XCI) provides an important example of the influence of lncRNAs, chromatin remodeling and nuclear positioning on gene activity (Figure 3). In female mammalian cells, only one X chromosome remains active to compensate for chromosome dosage between female (XX) and male (XY) cells. XCI is established during early development and maintained through mitotic cell divisions. Silencing of the X chromosome is achieved by the transcription regulation of Xist (X-inactive specific transcript) lncRNA and its antisense transcript Tsix from the X-inactivation center (Xic). Single-locus labeling of X chromosome centers (Xic) using the TetO system revealed that homologous pairing events determine which X chromosome will be silenced during differentiation (95). The mobility of both X chromosomes increased prior to XCI in undifferentiated cells, then X-chromosomes paired for 45 minutes. After pairing, the monoallelic downregulation of Tsix induces the upregulation of the Xist, which will coat whole X chromosome, silencing its transcription. It has been suggested that direct association of the X chromosome with nuclear lamina is required for Xist RNA spreading (96), however, the pairing/lamina localization models might be still under debate since a recent report demonstrated that reducing Xic pairing as well as relocation of X chromosome to the nuclear lamina does not influence monoallelic Xist transcription (97). Xist transcription and its spread over the whole inactivating X chromosome territory were visualized in living cells using the MS2-MCP system (98). Xist spread on the X chromosome led to the binding of epigenetic and heterochromatin regulators, culminating in transcriptional silencing of the chromosome. Interestingly, transcription of Xist remains active and it is necessary for the turnover of bound Xist transcripts by newly transcribed Xist RNA. Xist RNA has a half-life of 4–6 hrs, and a stable association with the chromosome that depends on the functional tethering of a cofactor enhancer of zeste homolog 2 (Ezh2) (99). XCI is one of the best-characterized processes of chromatin regulation; understanding the entire process of XCI may reveal common molecular mechanisms underlying chromatin dynamics and transcriptional regulation.

Figure 3. Molecular mechanisms in X chromosome inactivation (XCI) characterized by live imaging approaches.

(a) A single-locus DNA labeling approach using the tetO/TetR system identified X-inactivation center (Xic) pairing prior to initiation of XCI, suggesting it plays a critical role in the choice of which X chromosome is silenced. (b) Using simultaneous detection of X-inactive specific transcript (Xist) and Tsix using RNA fluorescent in situ hybridization (FISH) in fixed cells, the temporal order of chromatin remodeling and transcriptional regulation was determined (95). (c) Single-molecule detection of Xist RNA using the MS2–MS2 capsid protein (MCP) (98) or pumilio homology domain (PUM-HD) (99) system also revealed the dynamic nature of Xist long noncoding RNA in coating and inactivating the X chromosome.

4.1.4. Formation of transcriptional hubs

Transcription of any gene results from a highly coordinated assembly of multiple components of the transcriptional machinery on the DNA. Advances in both biomolecular labeling strategies with the self-labeling HaloTag dyes (100) and imaging modalities such as LSFM have led to the ability to track the different components of the transcriptional machinery with high precision (101). Quantitative measurements of the residence times of TFs, cofactors or mediators on the DNA, and their binding kinetics (102) have yielded insights into the temporal regulation of how individual components interact with each other and with chromatin. While these tagging systems work effectively in mammalian cells, attempts to use fluorescent tags or dyes in vivo need further optimization. A recently developed protein tag referred to as the LlamaTag, which directly binds to mature fluorescent proteins, has enabled visualization of rapid changes in TF concentration in live Drosophila embryos (103). How dynamic changes in TF concentrations affect the assembly of the transcriptional machinery has been explored both biochemically and by imaging. One hypothesis is that TFs, cofactors or mediators, and RNAP II all contain intrinsically disordered, low-complexity domains. The self-assembling nature of these domains allows them to engage in stable interactions that persist over time to facilitate the formation of transcriptional factories (104–106). Biochemical characterization of these factories has highlighted that many of their components could result in phase-separated condensates in the nucleus (107). In living cells, a temporally coordinated cluster formation would facilitate cooperative interactions between the different components of the transcription factory in a timely manner, thereby leading to a controlled onset of transcription. Simultaneous imaging of RNAP II and a transcriptional mediator showed in real time that both these proteins are incorporated in the same phase-separated condensate, thus indicating its heterotypic nature (104). Although these studies provided high temporal dynamics of the transcriptional hubs, it was not evident how hub formation occurred on certain chromatin domains (see the inset in Figure 2). Combined DNA FISH and HaloTag TF imaging revealed the formation of hubs on microsatellite repeat regions (108). However, how these self-assembling hubs or condensates affect transcription from endogenous genes has yet to be shown. Simultaneous imaging of these clusters with tagged loci in the future may provide insights into whether such assembly occurs at active gene loci (i.e., whether actively transcribing genes induce the clustering) or the clustering induces the transcription. Further biological relevance of these hubs can be extended toward understanding whether the formation of clusters or condensates changes, especially on inducible genes, which are triggered by external stimuli.

One such example is genes encoding for different Heat Shock Proteins (HSPs), which coalesce into discrete spots in the yeast nucleus upon transcription stimulation. Interallelic clustering in haploid yeast cells was measured as co-localization of different loci labelled with different repeats of the LacO-GFP system (109). The interaction between the HSP104 and HSP12 locus was dependent on transcriptional activation because mutations on the TATA box abolished it. This result and the failure of constitutive genes to coalesce, suggested that specific transcriptional factories are formed in response to environmental stimulus and co-regulated by unique transcription factors, like the heat shock factor 1 (HSF1 (109)). It remains to be determined if co-localization of HSP genes upon thermal stress is conserved in other organisms with different nuclear organization. In cultured mammalian fibroblasts different alleles of HSP70 did not coalesce upon transcriptional activation under thermal stress (110), rather the locus moved from the nuclear periphery to speckles (111). The movement is dependent on nuclear actin polymerization and its inhibition prevents transcriptional activation as measured with the MS2-MCP system. Therefore, transcription depended on the contact of the gene with nuclear speckles that might serve as transcriptional hubs to tether HSP70 and probably other genes (19). Further evidence that the location of the genes matters was provided by the study where double stranded breaks were introduced at two distinct chromatin locations: a promoter-proximal region downstream of the transcription start site and a region within an internal exon (intergenic) (112). Transcription from these locations after induction of double stranded DNA (dsDNA) breaks was monitored using two reporter constructs, using MS2 and/or PP7 tagging. While suppression of preexisting transcription occurred after such breaks, damage at the intergenic regions resulted in transcriptional recovery. This site-specific transcriptional recovery was possibly due to local nucleosome depletion, thereby highlighting the importance of chromatin regions influencing DNA-damage induced transcriptional response.

The recent advances in labeling of DNA loci with CRISPR-Cas9 and tagging endogenous genes in various organisms have provided insights into the dynamic changes in the nuclear architecture and the assembly of transcriptional factories, and how these changes regulate transcription.

4.2. Transcription Dynamics: Initiation and Elongation

Transcription kinetics with high temporal sensitivity has been determined with technologies to tag endogenous genes using the CRISRP-Cas9 system and the development of fluorescent tags for labeling mRNAs. Transcription initiation occurs in an irregular stochastic fashion, where a period of transcriptional activity is followed by an inactive period. This leads to the well-established “random telegraph” model (113). It considers stochastic switching between transcriptionally inactive and active states, referred to as “bursts”. Transcriptional bursting is one of the underlying components of gene expression noise and the heterogeneity observed between genetically identical cells. Based on the nature of the bursting behavior, one can begin to infer the molecular regulation of bursting, which is often unique to each gene and even each allele. Some of the key parameters analyzed are- amplitude (the intensity of transcription), frequency (how often bursts occur), and duration (how long bursts persist). While these analyses provide insights into transcriptional regulation, more detailed understanding requires mathematical models incorporating different promoter states. Several studies have shown that transcription initiation is a multi-step process, and often promoters fluctuate in different time scales in different organisms. This has been described in other reviews (36).

Besides tracking transcription from single genes, use of orthogonal tags like MS2 and PP7 systems have furthered our understanding of how multiple genes behave with respect to each other in their native state. This is best exemplified by measuring transcription from two distinct MS2 and PP7-labeled genes in Drosophila embryos which are controlled by the same snail shadow enhancer. When the enhancer is placed equidistant from MS2 and PP7 reporter genes in a symmetric orientation, similar bursting frequencies of both reporters were observed. However, when the enhancer was placed closer to the PP7 reporter, sustained transcription was observed from the PP7 locus, with significant increase in burst frequency than the MS2 reporter (88). These results showed that competition between genes for the same enhancer can modulate bursting frequencies. Recent evidence using reporter constructs and endogenous genes have highlighted how histone modifications can also play a role in modulating bursting frequencies. For example, transcriptional bursting of a luciferase reporter controlled by a circadian gene promoter revealed that the burst frequency was modulated by the circadian time, whereas the burst size was determined by the site of integration. The circadian changes are correlated with the histone acetylation levels, and CRISPR-Cas9–mediated acetylation of the promoter was sufficient to change the burst frequency (114). Such regulation of burst frequency, but not burst size by acetylating histones has been shown for inducible genes like c-fos from fixed cell studies (115). Based on both fixed cell and live imaging studies, the burst size appears to be modulated primarily by the kinetics of the transcription factor binding (49, 116), whereas the histone modifications impact the burst frequency.

Besides transcription initiation, direct measurements of the dynamics of transcription elongation in single cells has been made possible by positioning the MS2 and PP7 stem loops in the 5’ and 3’ ends of the same gene. By ensuring that the nucleotide sequence between the two labels is sufficiently long, the time between the appearance of 5’ and 3’ labels was used to directly measure RNAP II elongation rates (1). Using a reporter gene in yeast where MDN1 is driven by the GAL promoter, the average RNAP II elongation rate was measured at 25 ± 2 bases/sec upon galactose induction. Significant cell-to-cell heterogeneity was observed in the measured elongation rates, which varied widely across organisms. Similar experimental design in Drosophila embryos yielded an average elongation rate of 40 bases/sec, and this was independent of the transcription factor gradients along the anterior-posterior axis of the embryos. Interestingly using a Ponasterone A inducible promoter in cancer U2OS cells and by placing the stem-loop cassettes 3 kb apart, RNAP II elongation rates were measured to be much faster, at 100 bases/sec. The different elongation rates may arise due to the differences in the reporter designs and the genomic environment where the integration of the reporter occurs, or alternatively stochastic pausing of polymerases. Utilizing strategies such as reporter genes with both MS2 and PP7 integrated at a specific locus with varying promoters, as performed in yeast (117), would help to decouple the contribution of the promoter and the genomic environment from the trans-regulatory elements in determining elongation rates. In the future, insertion of the stem loops in the 5’ and 3’ of the endogenous genes will further validate whether these measurements are affected by cis or trans regulatory elements of the genome.

Transcription dynamics mark the beginning of the extensive regulation of gene expression. The post-transcriptional processing of the nascent transcripts, cytoplasmic export and localization, and their translational status determines the fate of these transcripts.

5. RNA PROCESSING AND LOCALIZATION

One property common to many cellular RNAs is that they are post-transcriptionally modified. This processing affects the subcellular localization and fate of RNAs, and can be rapidly modified to regulate gene expression in response to environmental changes. Imaging post-transcriptional events occurring both in the nucleus and in the cytoplasm have characterized the life cycle of mRNAs with high temporal resolution and several reviews have been written describing each step in the life-cycle of single mRNAs in great detail (36–38, 118, 119). To summarize, co-transcriptional pre-mRNA splicing in living cells was visualized by labeling introns using either the MS2 system (120–122) or the MBs (123). Frequencies of splicing monitored at single transcriptional sites revealed stochasticity of co- and post-transcriptional splicing (122) and differential splicing kinetics between two introns in same transcript (120). The dynamics of single mRNA nucleocytoplasmic export through nuclear pores has been investigated using several approaches. With the MS2-MCP system either rapid export of mRNPs (0.5 seconds) (124) or a three-step process including docking, transport, and release (durations of ~80, 5–20, ~80 milliseconds) was characterized (40). Three-dimensional export routes and selectivity mechanism were addressed with Single-Point Edge-Excitation sub-Diffraction (SPEED) microscopy (125). Interestingly, ~36% of mRNP molecules that encountered the nuclear pore successfully completed export. Using LSFM, ~25% of mRNP molecules exported rapidly after encountering with nuclear envelope (65 ms up to several seconds) in Chironomus tentans (126). The functional association of mRNA export factor NXF1 with cytoplasmic side of nuclear pore was identified using real-time single-molecule mRNA tracking as well as stimulated emission depletion (STED) super resolution microscopy and fluorescence lifetime imaging microscopy (FLIM)-FRET in fixed cells (127). Live imaging of PP7 labelled mRNAs in yeast indicated that once the RNA reached the nuclear periphery, it is not immediately exported. Instead, mRNAs scan the nuclear periphery till they find a nuclear pore that enables their export. The scanning behavior is influenced by the nuclear basket protein MLP1/2 and the mRNA binding protein Nab2 (128). Once exported into the cytoplasm, these mRNPs may either undergo a dynamic exchange of RNA binding proteins, or translate immediately, or localize to subcellular compartments to be translated in response to specific cues. In this section we discuss the recent advances in imaging that allow us to follow the fate of the mRNAs from RNA localization to translation to decay, as well as the less explored journey of ncRNAs.

5.1. Localization of Noncoding RNA

ncRNAs are functional RNA transcripts that play fundamental roles in a broad range of cellular processes. They are categorized into two types, small (<30nts) and long (>200nts). Live imaging of some nuclear lncRNAs, such as Xist and TERRA, and their function in nuclear organization and chromatin dynamics have been reviewed in the previous section (1C). Small ncRNAs include transfer RNAs (tRNA), microRNAs (miRNAs), small interfering RNAs (siRNAs), Piwi-interacting RNAs (piRNAs), small nucleolar RNAs (snoRNAs) and the spliceosomal small ncRNPs. Single-molecule imaging of small ncRNAs is precluded by the difficulty of multimerizing fluorophores in such a small molecule without perturbing its function and maturation.

Two biological processes that would benefit greatly from high-resolution single molecule approaches are the complex life-cycle of splicing factors and how miRNA binding kinetics affect the fate of mRNAs. Unperturbed tagging of the splicing factor U1 snRNA has been possible with the riboglow system (Table 1) (55). This system enabled visualization of U1 snRNP localization in Cajal bodies in the nucleus and in U-bodies in the cytoplasm upon stress induction. Nonetheless, it was not powerful enough to track single U1 molecules and investigate the impact of stress on its maturation process. Regarding the regulation of mRNAs by the binding of miRNAs, a recent imaging approach attempted to answer how the low copy number of miRNA could regulate a specific mRNA target (129). It was hypothesized that specificity depended on the spatial proximity between miRNA and mRNA. Using a probe with a fluorophore and quencher, a fluorogenic system was designed that reported on miRNA maturation. The binding of the probe to the pre-miRNA allowed the quencher to be removed after dicer processing and the fluorophore to remain bound and reveal the mature miRNA. This conferred high specificity and low background enabling visualization of miRNA maturation after local stimulation of neuronal dendrites (129). The functional impact of miRNA was measured by inhibition of translation of the mRNA targets that localized in the same dendrite. It remains to be determined if the target specificity observed in neurons is cell-type or microRNA-dependent.

5.2. Link Between Localization and mRNA Translation

The sophisticated interplay of cis- and trans- regulatory elements in the mRNA life-cycle determines where and when an mRNA will be translated. Local mRNA translation plays a crucial role in highly regulated processes, like embryo development in Drosophila (44), synapse formation in stimulated dendrites in neurons (extensively reviewed by (118, 130–132) and cell cycle progression in the budding yeast S. serevisiae (133). The general accepted model is that mRNAs travel in granules in a translationally repressed state to the specific location in the cell where protein is needed. This has been well exemplified during mRNA transport along microtubules in neurons, where β-actin mRNAs bound by ZBP1 (IGF2BP1) are packed in granules, until local activity unmasks these mRNAs for translation (39). The cycle of unmasking and re-masking is around fifteen minutes. Active transport of mRNAs is regulated by cis- and trans-factors, which influences how they associate with the motor proteins to travel along microtubule tracks (reviewed in (118)). A recent study has identified a new mode of transportation of the RNA granules by hitchhiking on lysosomes in mammalian cells, including axonal transport in primary neurons (134). Annexin A11 (ANXA11) tethers β-actin mRNA containing granules to lysosomes by means of phase separation and membrane binding properties of its N- and C-terminal regions respectively (134). In neurons, transport of mRNAs to distal regions by docking on lysosomes may have implications in cellular homeostasis. In the case of Rgs4 mRNA, its 3’UTR fine tunes the direct transport of the mRNA to dendrites and its association with synapses (135).

Besides mRNAs and the RNA binding proteins, the translation factors are often packaged in complexes with the mRNAs, which are transported to sites of action. In migrating mouse embryonic fibroblasts, simultaneous imaging and co-tracking of single β-actin mRNAs using the MS2 system and ribosomes labelled with photoactivatable fluorescence proteins revealed active translation at focal adhesions, resulting in reduced mRNA diffusion speeds (136). However, not all mRNAs are locally translated to influence migration in fibroblasts. For example, APC-dependent mRNAs are globally translated in the cell, and they undergo translational silencing by granule formation at sites of protrusion retraction (137). In yeast, specific granules harboring the mRNAs of translation factors, also known as translation factories, are transported to the bud tip (138). Localization of the translation factories in the bud-tip is myosin dependent and relies on the RNA binding protein She2 and the scaffold protein She3. To visualize how mRNA transport and local translation are coupled in all these cell types, dual labeling of mRNA with MS2/PP7 aptamers and the nascent peptides with SunTag system will be extremely informative. SunTag is a series of the GCN4 epitope that are genetically integrated in the 5’ end of the coding sequence of the mRNA of interest (139). On-going translation kinetics are assessed by the rapid binding of fluorescently-tagged single chain antibody fragments that recognize the GCN4 epitope as soon as it is translated in the cell. (43, 140–143). Since the GCN4 epitope is derived from yeast, a modified SunTag system will be required for translation kinetics there. New versions of epitopes which can report nascent protein synthesis has been developed (144–146).

The unifying theme obtained from mRNA transport in various organisms is that functionality is achieved by confining the regulator and its molecular targets in the same cellular location. In one specific situation, this consensus is challenged; the localization of mRNAs with ncRNAs, translation factors and decay enzymes in SGs and processing bodies (PB) during the stress response. The recruitment of mRNAs to SGs and PBs in the cytoplasm is a dynamic process. Different mRNAs show different residency time in these biomolecular condensates and competence to traffic back and forth between these condensates and the cytoplasm (42, 147). The translation status of these mRNAs were assessed using the SunTag system, which showed that mRNAs cease translation prior to entry into SGs and resumes translation rapidly once SGs have been dissolved (11, 12). Hence, the recruitment of mRNAs to SGs and PBs might provide a protection mechanism for the mRNAs during stress and the means for a rapid re-start of translation once the stress ceases (42, 147).

Localization of RNAs in specific sites of the cell is a conserved and dynamic process regulated by cis- and trans-elements. It provides the cell with the plasticity to express proteins in response to internal requirements and external cues. Efficiency of mRNA translation can also influence the decay of the mRNA.

5.3. mRNA Turnover

Capturing the dynamics of mRNA degradation at single-molecule resolution has been technically challenging. The initial MS2-MCP system generated fluorescent degradation intermediates due to the high-affinity binding of MCP to MS2, which could not be displaced by the mRNA degradation machinery in yeast (45–48). This problem was resolved by generating a version of the MS2 system that degrades simultaneously with the coding sequence (45). Imaging challenges have yet to be overcome to quantifymRNA degradation accurately. One technical limitation is the measurement of mRNA degradation as the result of the disappearance of the fluorescent signal emitted by single mRNA molecules. The fast cytosolic diffusion of mRNAs makes it difficult to distinguish degradation from the loss of signal caused by freemolecular diffusion into out-of-focus z-planes. Multifocal imaging could compensate for this molecular diffusion, but in turn, it causes signal loss due to rapid photobleaching. Two independent solutions have been found to circumvent these problems and answer two unrelated questions on decay. The first solution, known as the 3’RNA end accumulation during turnover (TREAT) system, was used to study the fate of the mRNAs in PBs (42) (Figure 4b, i). The system arose from the two-color labeling method using PP7 and MS2 stem-loops, which was originally reported as a translation detection tool known as translating RNA imaging by coat protein knockoff (TRICK) (44) (Figure 4a, i). In the TRICK system, translated mRNAs can be distinguished from untranslated mRNAs by the displacement of fluorescence-PP7 labeling on the open reading frame by an elongating ribosome. For TREAT, instead of inserting the PP7 stem-loops in an open reading frame, each mRNA is tagged in its 3’UTR with the MS2 and PP7 systems separated by a viral pseudoknot that inhibits degradation by the enzyme 5’-3’ exoribonuclease 1. Intact mRNAs are detected by bothMS2 and PP7 signals, and upon degradation, the decay fragments carrying the pseudoknots are easily identified by the exclusive detection of the signal from the PP7-PCP system. The conversion of an mRNA from two-color to a single color provides a highly quantitative detection system of mRNA degradation.

Figure 4. Fluorescent imaging approaches to detect translation and mRNA decay in living cells.

(a) Live-cell imaging of translation on single-molecule mRNA. (i) Translating mRNA imaging by coat protein knockoff (TRICK). By measuring the displacement of fluorescence-PP7 labeling from open reading frames on reporter mRNA by the first elongating ribosome, this imaging method distinguishes translated from untranslated mRNA (44). (ii) Nascent peptide detection of translating ribosomes using the SunTag system. The nascent peptides of SunTag are labeled with superfolder GFP (sfGFP)-tagged single-chain variable fragment (scFv-sfGFP) on translating mRNAs (43, 143–146). (b) Live-cell imaging of single mRNA decay. (i) 3’-RNA end accumulation during turnover (TREAT): Similar to the two-color labeling scheme of TRICK, but both PP7 and MS2 stem-loops are located in the 3’ untranslated region. Pseudoknots are inserted between PP7 and MS2 stem-loops to distinguish the decay fragment (one color) from intact mRNA (two color) (42). (ii) Nonsense-mediated decay (NMD) reporter with the SunTag system. The combination of the SunTag translational reporter system and CAAX-tethered mRNA allows one to monitor the translation kinetics of NMD (41). Normal termination codons (STOPs), premature termination codons (PTCs), and release factors (RFs) are shown.

The second solution is an mRNA tethering system. It has been successfully used to investigate the dynamics of nonsense-mediated decay (NMD). NMD is an mRNA surveillance mechanism that cleaves mRNAs containing premature termination codons (PTCs) in a translation-dependent manner and therefore limits the production of truncated proteins. NMD on single mRNAs was visualized in real time by tagging mRNAs in their 3’UTR with PP7 and tethering them to the plasma membrane with the CAAX-PCP system to prevent their diffusion (41) (Figure 4b, ii). The reporter mRNA containing the PTC also encoded the SunTag translation detection system in the N terminus (43, 139, 141–143). The SunTag signal provided the means to assess the probability of a ribosome triggering NMD as measured by the separation of the SunTag and PP7 fluorescent signals caused by endonucleolytic cleavage during NMD (148). Since the separation events of the SunTag and PP7 fluorescent signals can also represent the peptide release from translation termination at PTC, the authors accounted for the moment of the 3’ end decay by measuring the vanishing of the PP7-PCP signal (Figure 4b, ii). Simultaneous detection of the SunTag and PP7 signals suggests that not only the first ribosome but each ribosome that terminates translation at the PTC has an equal probability of triggering NMD. Also, the efficiency of NMD was influenced by the exon sequence downstream of the PTC and the PTC-to-intron distance. Given that most PTC-containing mRNAs are expected to be degraded immediately after mRNA export before reaching the plasma membrane (149), CAAX-tethered mRNA molecules used to study NMD might be selectively capturing mRNA molecules that have escaped from or were subjected to slowNMD. However, the tethering approach offers an attractive benefit to track themoment of mRNA decay in living cells.Using these imaging tools, it is now possible to determine the kinetics of mRNA decay and the sequence specificity of this process (whether the efficiency of the decay of 3’end regions is sequence dependent or whether some sequences can escape degradation). Recent publications suggested that sequences of PTC-containing mRNA could participate in a genetic compensation response (150, 151). Identifying and imaging the behavior of these RNAs will provide insights into the feedback loop described for mRNA decay on transcriptional regulation and even nuclear organization.

6. CONCLUSIONS AND PERSPECTIVE

Simultaneous imaging of specific DNA locus and RNA are providing a fresh perspective on how temporal coordination between different events, from DNA replication to transcription to mRNA processing, translation and decay is attained despite stochasticity at every step. Imaging living cells in action at the single molecule level is opening new venues to explore chromatin dynamics and RNA functions that influence gene expression and go beyond the synthesis of proteins. It provides the means to track these molecules and answer when and where their actions influence the functional events in a cell. These observations pave the way to design multicolor imaging experiments and determine how several factors efficiently interact. For example, the simultaneous imaging of specific nuclear components, like mediator, enhancers and nascent RNAs, affords a way to distinguish productive intermolecular interactions from spurious ones as a function of time, and how this contributes to restructuring the nuclear architecture for genetic decisions. Visualizing the temporal order of their interactions will also define the factors/events that facilitate or fine tune cellular processes. Therefore, it is necessary to choose the factors that need to be tagged and quantify their behavior in live cells in order to characterize the events that ultimately define the cell fate and functionality.

Two goals for the future of the single-molecule live imaging field would be multiplexed imaging of several factors to understand their temporal dynamics relative to each other and study the cellular processes in whole organisms. Several efforts are being made to take on the challenge to visualize simultaneously more than three components involved in a cellular action at a high imaging speed without photo bleaching or photo damaging the sample. A pallet of new generation of fluorescent dyes has been introduced that are brighter and more photostable. The photoactivatable and photoconvertible versions, like the new Janelia Fluorophore Halo-dyes provide greater localization precision (152, 153). Faster and more sensitive microscopes like light-sheet microscopy for 3D imaging allows to localize single particles with high sensitivity with minimal phototoxic damage to cells (154, 155). New algorithms to analyze and provide statistics to the data generated by imaging approaches are constantly being developed (156). Although advanced imaging technologies have expanded our approaches for visualizing the various steps of gene expression with high temporal and spatial resolution, the wide-spread use of these technologies have been technically challenging in intact tissue and whole organisms. Most labeling schemes for single-molecule approaches rely on insertion of multiple repeat sequences such as stem-loops or binding motifs, and multimerizing the fluorescent proteins, which raises concerns about the genetic manipulations of the native DNA sequences or loading the RNAs with bulky fluorescent molecules. This also poses significant challenges for labeling small nucleic acids like ncRNAs, miRNAs, and imaging them as single molecules over long periods of time. To address these concerns, we foresee that DNA and RNA imaging systems based on non-genetic tagging need to be optimized for brightness and specificity to allow detection of endogenous molecules in the physiological context of tissue and live animals. These advances are needed to investigate the structure and organization of the nucleus, the communication between the nucleus and the cytoplasm, and the role of bimolecular condensates in cells when functioning in their tissue microenvironment. Other fields that will certainly benefit from these technologies are cytogenetic diagnosis (33) in primary cells and RNA therapeutics. Localizing the therapeutic RNA molecules and following them over time in their physiological relevant environment will enable fine-tuning of therapeutically administered dosage.

A dynamic cellular environment provides the plasticity for meaningful interactions, some of which can influence the steps of chromatin organization and gene expression. To decipher the functional cellular network, further development of novel tagging systems, not just limited to fluorescent proteins, are required for multiplex detection of the different components at various steps of DNA and RNA’s life-cycle. This future era of technological revolution requires integration of the findings of single-molecule fluorescence microscopy with biochemical approaches that offer a complementary perspective on the interactions and localization of a molecule during its life.