mRNAs are localized to subcellular domains in neurons as a mechanism to facilitate localized gene expression and functional targeting of proteins [1, 2]. Recent methods to detect mRNAs have expanded greatly, yet the use of live-cell imaging to study mRNA dynamics remains limited [3]. Previous work using the bacteriophage-derived MS2-MS2-Coat Protein (MCP) system to detect dynamic fluorescent protein-labeled mRNAs in living cells has demonstrated diverse aspects of RNA biology [4, 5]. Yet many reports indicated that the signal-to-noise ratio (SNR) is too low to reliably detect RNA movements, given the instantaneous velocities of RNA granules at greater than one micrometer per second [6]. The local translation of mRNA is critical for many forms of synaptic plasticity [7], and a better understanding of mRNA dynamics in dendrites is required. Both the MS2 tagging and microinjection methods have been used to a limited degree to detect neuronal mRNA dynamics, with the MS2 tagging being a more readily available approach [8]. However, there remain several obstacles to imaging mRNAs in living neurons [5]. Therefore, to gain a deeper understanding of mRNAs trafficking in cells, improved detection methods are required.

Several reports have applied microinjected fluorescently-labeled mRNAs to track their movements in dendrites [9, 10]. However, successful use of this method requires specialized equipment and training, and yet still, mRNAs synthesized in vitro may not recapitulate dynamics similar to their endogenous counterparts. Notwithstanding these caveats, microinjected RNAs have shown the capability to form particles and undergo nuclear-cytoplasmic shuttling in live cells [11]. Alternatively, the MS2 system, which utilizes genetic encoding of MS2 mRNAs, allows for single particle tracking [8, 12] and has been used to study mRNAs such as Calcium/Calmodulin-dependent protein kinase type II subunit alpha (CaMKII$\alpha$) mRNA and Arc, Kv4.2, and $\beta$-actin mRNAs [6, 13, 14, 15]. Analysis of these time-lapse images showed that single granules can be observed under optimized conditions. However, while the MCP is an obligate dimer, recent biophysical data reveal that a significant fraction of MCP exists in a monomeric state, contributing to increased background fluorescence [16].

We sought to improve the SNR of the MS2 system in neurons to better observe mRNA dynamics, given how few RNAs have been imaged in dendrites, and compared this to an improved microinjection method. We showed that the choice of promoter and ratio of RNA to fluorescent protein construct are critical to use MS2 successfully. Unexpectedly, we also observed that the majority of significantly displaced mRNAs are the smallest and dimmest ones detected. By optimization of both MS2 and microinjection methods, we observed mRNAs moving actively into and out of synapses, as well as splitting off from larger granules, and that they co-transport with their cognate mRNA-binding proteins (RBPs) in dendrites. We characterized the mRNAs and demonstrated that they fell into three distinct sizes, the smallest of which actively transports in dendrites with multiple copies of diverse mRNAs.

To accomplish this study, we needed to significantly improve the transfection efficiency since this was the rate-limiting step in our efforts to image RNAs with MS2 in primary hippocampal neurons. We co-transfected CFP (driven by cytomegalovirus (CMV) promoter) with MS2-RNA as a marker for transfected cells and tested several transfection protocols. To our surprise, the greatest factor improving transfection of neurons was the presence of dividing glia, as the omission of the routine primary neuronal additive cytosine arabinoside (AraC, to prevent glial proliferation) yielded ~3-fold higher transfection rates (Fig. 1A). This was in the presence of a similar density of neurons as demonstrated by Tau staining (Fig. 1B). Given the abundance of transfected glial cells, we developed plasmids that carry the Synapsin promoter to drive the expression of fluorescent proteins exclusively in neurons but not glial cells (Fig. 1C).

We found that a critical aspect of MS2-RNA labeling requires optimizing the expression of the MS2 Coat Protein-Fluorescent Protein (MCP-FP) relative to the MS2-RNA (Fig. 2A). This aspect has been highlighted previously [4]. Without the optimal ratio of expression, too few or no visible granules are formed, with diffuse MCP-FP background obstructing visualization. To optimize the MS2-RNA labeling system, we made some improvements on three parameters that affect MS2-RNA granule formation: the MCP-FP:MS2-RNA ratio; MS2-RNA promoter; and MCP-FP promoter. To assess the effects of a given parameter on RNA granule visualization, we quantified the number and brightness of the motile population, which were small RNA granules (1 µm diameter; ‘signal’), as well as the dendritic background caused by free/unbound MCP-FP (‘noise’) and obtained an SNR for the tested conditions (Fig. 2B).

Firstly, we compared the different ratios of MCP-RFP (red fluorescent protein):MS2-RNA using an established dendritic mRNA 3^′ UTR reporter, CaMKII$\alpha$ [13]. Here a range of ratios was tested, with the most striking qualitative difference between 1:3 and 1:9, as RNA granules with the 1:9 ratio showed a 20% increase in SNR (Fig. 2C, 1:3 Rous sarcoma virus (RSV) vs. 1:9 RSV), resulting in increased number of small RNA granules (Fig. 2D, 1:3 RSV vs. 1:9 RSV) and reduced dendritic background (Fig. 2E, 1:3 RSV vs. 1:9 RSV). Below the 1:3 ratio, very few granules were observed, and no cells could be imaged for RNA dynamics. Next, we assessed the importance of the promoter driving the MS2-RNA expression and found that substituting the RSV promoter for a stronger CMV promoter resulted in a further 12% gain in SNR (Fig. 2C, 1:9 RSV vs. 1:9 CMV), with no further decrease in average dendritic background (Fig. 2E, 1:9 RSV vs. 1:9 CMV).

We next assessed the effects of the promoter driving the MCP-RFP on RNA granule visualization and found it to be of critical importance. While most reports using the MS2-RNA labeling system use the CMV promoter to drive MCP-FP [1, 18], we found that this promoter leads to a vast excess of MCP-RFP in dendrites which results in a very high dendritic background that obstructs RNA granule visualization. MCP-RFP driven by weaker promoters such as Ubiquitin (Ubc) or Synapsin (Syn) exhibited a ~50% decrease in the dendritic background when compared to MCP-RFP driven by the CMV promoter (Fig. 3A,B, Ubc vs. CMV and Syn vs. CMV). Furthermore, the MCP-RFP brightness in the nucleus was quantified as a means to assess the amount of excess MCP-RFP within a neuron under the different promoters since the MCP-RFP contains a nuclear localization signal (NLS) that sequesters free MCP-RFP back to the nucleus. This analysis showed that neurons expressing MCP-RFP under the CMV promoter had a 3-fold and 10-fold higher nuclear MCP-RFP fluorescence than those using the Ubc or Syn promoters, respectively (Fig. 3C,D, Ubc vs. CMV and Syn vs. CMV). These weaker promoters resulted in a 50% reduction in the number of large RNA granules ($>$1 µm diameter, data not shown), while keeping the number of small RNA granules comparable (Fig. 3E, Ubc vs. CMV and Syn vs. CMV) and with a 2-fold increase in the SNR of small RNA granules (Fig. 3F, Ubc vs. CMV and Syn vs. CMV). These data suggested that it is critical to maximize the amount of RNA (by increasing the ratio to 1:9 and using a strong CMV promoter) while minimizing the expression of MCP-RFP (by using less MCP-RFP and using a weak promoter), thus ensuring that the MS2-RNA, not the MCP-FP, is in excess.

Other reports have successfully used microinjected mRNAs in neurons to track granules [19]. Therefore, we compared the optimized MS2 system to microinjected fluorescently labeled mRNAs. The mRNAs were microinjected into the nucleus, where they were initially restricted, and then were accumulated at the cell body of hippocampal neurons (Fig. 4A,B). Microinjected mRNAs in dendrites exhibited ~100% higher SNR over MS2-RNAs with Syn-MCP-RFP and ~300% higher than MS2-RNAs with CMV-MCP-RFP (Fig. 4C). In addition, microinjected neurons exhibited 2-fold more small moving granules in dendrites, although there were also 2-fold large ($>$1 µm diameter), non-motile granules than in MS2-RNA cells (not shown). We attribute the increased SNR to an indirect labeling technique we employed, which used amine-modified UTP during in vitro RNA synthesis followed by conjugation with N-hydroxysuccinimide (NHS) ester-modified Alexa Fluor 555 (see “Materials and Methods”).

Given the increased number of motile mRNA granules that microinjected neurons exhibited, as well as improved SNR, we aimed to use both microinjected and MS2 techniques to observe novel RNA dynamics in a time-lapse manner, notwithstanding the caveat that microinjected mRNAs may not undergo the same nuclear processing as endogenous counterparts and therefore, may not function as similarly as those genetically encoded like MS2-RNAs. To test the fidelity of microinjected mRNAs, several established localized mRNAs were microinjected, which exhibited granular dendritic motility (Fig. 4A,B). These included the 3^′ untranslated regions (UTRs) of CaMKII$\alpha$ [13] and Fmr1 [17], two mRNAs critical for learning and memory function and shown to be involved in Fragile X Syndrome pathophysiology. The mRNA granules moved in both anterograde and retrograde directions and reached speeds of up to 4 µm/sec (Fig. 4D). GAPDH mRNAs, which are not actively localized to dendrites, were mostly stationary within the cell bodies, with occasional granules restricted to the somatodendritic domains (Fig. 4A,B). These data demonstrated the fidelity of microinjected mRNAs in neurons under these controlled conditions, and that those specific mRNAs are targeted to dendrites through 3^′ UTR elements.

Using RNA granule-tracking software (Imaris, Abingdon, United Kingdom), we observed that the number of smaller granules, i.e., ones that had diameters of 0.250–0.625 µm and showed processive movements greater than 3 µm, represented the majority of motile RNA granules. Larger granules of 0.625–1.0 µm were more sparsely observed and were less motile, with oscillatory movements, typically less than 3 µm net displacements (Fig. 4E). Highly motile small granules often merged with these larger granules briefly before moving further in subsequent frames. Many small granules paused within these larger complexes on their trajectory. Most microinjected RNA granules exhibited speeds of more than 1 µm/s, in agreement with previous reports (Fig. 4D) [6, 20].

To visualize novel aspects of RNA dynamics in hippocampal dendrites, we first used the optimized MS2 system (1:9 ratio of Syn-MCP:CMV-MS2). We co-transfected the MS2-CamKII$\alpha$ 3^′ UTR/MCP-RFP with PSD-95-GFP to visualize post-synaptic terminals and confirmed that mRNAs were within synapses by analyzing colocalization of the mRNA with PSD-95 (Fig. 5A). We also observed the preferential localization of RNAs at the base of synapses within dendrites, where MS2-CaMKII$\alpha$ 3^′ UTR/MCP-RFP was co-localized with Fragile X mental retardation protein (FMRP-GFP) (Fig. 5B). Interestingly, the splitting of an RNA granule was also observed within the dendritic spine head itself (Fig. 5D). MS2-CaMKII$\alpha$ 3^′ UTR/MCP-RFP was visualized within a spine as a single granule that then split into two distinct particles. Given the rapid time course imaging of this event (~0.4 FPS), it was clear that these particles originated from the single precursor granule, as shown in the surface reconstruction and 3D rendering movie generated using Imaris (Supplementary Material 2). Using a distinct mRNA and RNA-binding protein (RBP) combination, MS2-$\beta$-actin 3^′ UTR/MCP-RFP and zipcode-binding protein 1 (ZBP1-GFP) also formed motile granules that were detected in dendritic spines that were co-labeled with Cerulean, and this messenger ribonucleoprotein (mRNP) was also observed to go into spines (Fig. 5C). ZBP1 is known to bind to the ‘zipcode’ sequence of the$\beta$-actin mRNA 3^′ UTR with high specificity [21]. We observed MS2-$\beta$-actin 3^′ UTR/MCP-RFP moving together with ZBP1-GFP in a fast bi-directional manner in dendritic shafts, consistent with reports for each component individually [6, 21]. Two distinct motile MS2-$\beta$-actin 3^′ UTR/MCP-RFP/ZBP1-GFP granules were observed within a synapse and moved to the base of the spine within seconds, where a large preexisting granule was docked. Subsequently, two medium-sized granules in the dendritic shaft coalesced with the large granule at the base. A small granule then moved back into the synapse and exited seconds later (Fig. 5E). We observed many mRNPs consisting of MS2-$\beta$-actin 3^′ UTR/MCP-RFP and ZBP1-GFP within dendritic spines. To our knowledge, this is the first time mRNAs have been visualized moving into synapses in living cells.

To further visualize and characterize the dynamic of mRNAs and cognate RBPs using the improved methods, dual-colored mRNPs were followed. MS2-CaMKII$\alpha$ 3^′ UTR/MCP-RFP/FMRP-GFP mRNP granules were observed to move together within dendrites in both the anterograde and retrograde directions. Close examination of dual-labeled mRNPs within the dendritic shaft showed single granules splitting into two distinct granules that then moved in opposite trajectories (Fig. 6A), with both resulting particles retaining both the mRNA and protein components. Quantification of fluorescence intensities of each component showed that the sum of the resulting granules was roughly equivalent to the original granule before splitting (Fig. 6B,C). This observation of a motile granule splitting was unexpected due to recent reports suggesting that mRNAs transport mostly in the form of granules containing single RNAs in neuronal dendrites [22]. Together these data demonstrated the co-transport of mRNAs and cognate RBPs within dendrites and synapses and that one motile RNA granule may contain multiple copies of a single mRNA species within their mRNA/RBP structure. Furthermore, another mRNA/RBP pair (MS2-$\beta$-actin 3^′ UTR/MCP-RFP/ZBP1-GFP) also showed co-transport of the mRNA and protein components, with the mRNP traveling bi-directionally and with speeds of ~1 µm/s (Fig. 6D, E) [6].

To more directly test whether motile mRNPs contain more than one mRNA, both CaMKII$\alpha$ and Fmr1 mRNAs were labeled with distinct colors and microinjected into the nucleus of hippocampal neurons. After 30 minutes, these Alexa Fluor-labeled mRNAs started to enter dendrites rapidly. Several granules showed co-labeling for both mRNAs and rapid imaging revealed granules moving together within dendrites in both anterograde and retrograde directions (Fig. 7A,B). These granules were less than 0.6 µm in diameter, as observed previously for granules that move significant distances ($>$3 µm displacements). The velocities averaged ~1 µm/s, consistent with those of single mRNAs alone or mRBPs such as FMRP. Another distinct combination of two mRNAs was CaMKII$\alpha$ mRNA and a novel dendritically localized transcript for a synaptic protein, the Neuroligin1 (NLGN1) mRNA. Larger particles (0.625–1 µm) containing the two transcripts only displayed oscillatory movements consistent with those observed for larger single mRNA granules (by MS2 or Alexa-labeling) or GFP-labeled mRBPs (such as FMRP). These granules moved back and forth with total displacements of less than 3 µm in each direction. This was in contrast to small (0.625 µm) dual-labeled granules, which moved at speeds over 2 µm/s, similar to dual-labeled CaMKII$\alpha$ and Fmr1 mRNA granules (Fig. 7C,D). Furthermore, co-injection of three different labeled mRNAs (CamKII$\alpha$, Fmr1, and Arc 3^′ UTRs) also showed certain granules containing all three mRNAs moving actively in dendrites over four minutes (Fig. 7E,F). These data showed that individual motile granules in dendrites can contain more than one copy of distinct mRNAs that display motilities similar to those observed for single mRNA granules.

Dendritic mRNA localization is a key aspect of synaptic plasticity, as it allows the rapid and site-specific delivery of proteins required for synaptic modulation. In this study, we have optimized existing methods to visualize dendritic mRNAs in live neurons, providing an important tool for the study of synaptic activity-dependent mRNA transport. We focused on MS2 and microinjection as they are the most commonly used methods to study mRNA transport in live cells. Our first trials with the MS2 system yielded very few transfected cells with very high fluorescent backgrounds and very few particles undergoing directed transport. To resolve these issues, we first optimized transfection conditions to yield more transfectants and then optimized the plasmids to increase MS2-RNA expression and decrease MCP-FP expression. We found that the presence of glial cells in neuronal cultures increased the transfection efficiency and that the Lipofectamine LTX transfection reagent was far superior to the commonly used Lipofectamine 2000 (data not shown). Further, increasing the ratio of MCP-FP:MS2-RNA from 1:3 to 1:9 yielded brighter RNA granules. Additionally, placing the MS2-RNA under a stronger CMV promoter and the MCP-FP under weaker promoters (Ubiquitin and the neuron-specific Synapsin promoter) also yielded granules with highly increased SNRs.

We also studied microinjection as an alternative method for observing dynamic mRNA granules. Although microinjected RNAs don’t undergo nuclear processing to the extent of endogenous or overexpressed MS2 mRNAs, they do have a degree of nuclear processing when injected into the nucleus, as RNAs injected directly into the cytoplasm form mostly small granules but not larger granules that likely represent mature mRNPs (data not shown). To ensure bright mRNA granules, the in vitro transcribed RNA was labeled indirectly with fluorescent dyes, which yielded 3$\times$ brighter RNA when compared to directly-labeled RNA.

Using these optimized approaches, we detected many novel behaviors for mRNA particles. We observed RNA particles undergoing fast bidirectional transport, as expected, and also found that these particles not only stop momentarily at the base of dendritic spines, but some also travel into spines. We confirmed this result using the MS2 system, where we observed an mRNP composed of both $\beta$-actin mRNA and ZBP1 protein traveling in and out of synapses in a time frame of minutes. This raises the possibility that dendritic mRNAs may not only be translated at the base of spines following synaptic stimulation, but some proteins may be translated in the actual synapse.

While mRNPs have been thoroughly described in the past through biochemical and fixed cell methods, the co-transport of an mRNA and its cognate RBPs has not been demonstrated in live cells. This may be due to the difficulty in co-transfecting three plasmids into primary neurons, coupled with the fact that until now, the MS2 system yielded relatively few actively transporting granules. Optimized transfection conditions and MS2 plasmid construction allowed us to obtain a large number of transfected cells with sufficiently bright transporting granules. We were thus able to visualize the rapid transport of CamkII$\alpha$ 3^′ UTR/FMRP-GFP and $\beta$-actin 3^′ UTR/ZBP1-GFP mRNPs in live neurons, confirming what has until now only been hypothesized.

The scope of this study was to optimize live-cell mRNA visualization techniques to further understand the role of dendritic mRNA transport in local translation during synaptic plasticity. Microinjection provides brighter and more motile RNA granules than the MS2 system, making it ideal for studying the fast dynamics of mRNA transport. However, microinjection requires expensive equipment. The MS2 system, on the other hand, is widely available, and substantial improvements were observed by just changing the ratio of plasmids. Optimized transfections resulting in brighter RNA granules make MS2 ideal for studying mRNP dynamics.

Several studies have suggested that mRNAs travel singly in dendrites (Batish et al. [22]). Approximately ~800 mRNAs have been identified in dendrites, suggesting that colocalization of dendritic mRNAs is higher than can be currently measured; our microinjection data showed up to three different synaptic plasticity-related mRNA species transporting together, showing that the dendritic machinery is capable of transporting at least up to three different mRNAs together. Further, new methods to study the transcriptome in-situ promise to provide genome-wide localization of RNAs in neurons [23, 24, 25]. These techniques will confirm whether different mRNAs travel alone or are co-transported together in dendrites.

All data analyzed during this study are included in this published article. Raw data sets for current study are available from the author (IC, HW and JD) upon request.

IJC, HW and JD designed the research study. IJC performed the research. TZ and MN provided help and advice on the design. IJC and JD analyzed the data. IJC, HW and JD wrote the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Mice were housed in the AALAC-approved animal facility at Hunter College. Animal facility protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of Hunter College (approval number: HC49032).

IJC received funding from the Howard Hughes Medical Institute Undergraduate Research Scholar Program and the Research Initiative for Scientific Enhancement (RISE) Program at CUNY Hunter College. This work was supported by National Institutes of Health Grants GM084805 and National Science Foundation (NSF) Grants 0819022 and 0960367 to JD.

The authors declare no conflict of interest.

Supplementary material associated with this article can be found, in the online version, at https://doi.org/10.31083/j.fbl2912430.