Differential regulation of local mRNA dynamics and translation following long-term potentiation and depression

Decades of work have demonstrated that mRNAs are localized and translated within neuronal dendrites and axons to provide proteins for remodeling and maintaining growth cones or synapses. It remains unknown, however, whether specific forms of plasticity differentially regulate the dynamics and translation of individual mRNA species. To address these issues, we targeted three individual synaptically-localized mRNAs, CamkIIa, Beta actin, Psd95, and used molecular beacons to track endogenous mRNA movements and reporters and Crispr-Cas9 gene editing to track their translation. We found widespread alterations in mRNA behavior during two forms of synaptic plasticity, long-term potentiation (LTP) and depression (LTD). Changes in mRNA dynamics following plasticity resulted in an enrichment of mRNA in the vicinity of dendritic spines. Both the reporters and tagging of endogenous proteins revealed the transcript-specific stimulation of protein synthesis following LTP or LTD. The plasticity-induced enrichment of mRNA near synapses could be uncoupled from its translational status. The enrichment of mRNA in the proximity of spines allows for localized signaling pathways to decode plasticity milieus and stimulate a specific translational profile, resulting in a customized remodeling of the synaptic proteome.


Abstract.
Decades of work have demonstrated that mRNAs are localized and translated within neuronal dendrites and axons to provide proteins for remodeling and maintaining growth cones or synapses. It remains unknown, however, whether specific forms of plasticity differentially regulate the dynamics and translation of individual mRNA species. To address these issues, we targeted three individual synaptically-localized mRNAs, CamkIIa, Beta actin, Psd95, and used molecular beacons to track endogenous mRNA movements and reporters and Crispr-Cas9 gene editing to track their translation. We found widespread alterations in mRNA behavior during two forms of synaptic plasticity, long-term potentiation (LTP) and depression (LTD). Changes in mRNA dynamics following plasticity resulted in an enrichment of mRNA in the vicinity of dendritic spines. Both the reporters and tagging of endogenous proteins revealed the transcript-specific stimulation of protein synthesis following LTP or LTD. The plasticity-induced enrichment of mRNA near synapses could be uncoupled from its translational status. The enrichment of mRNA in the proximity of spines allows for localized signaling pathways to decode plasticity milieus and stimulate a specific translational profile, resulting in a customized remodeling of the synaptic proteome.

Introduction.
Synaptic plasticity requires the rapid and robust remodeling of the proteome1. Both the strengthening (long term potentiation-LTP) and the weakening of synaptic connections To address this, we used molecular beacons to track and quantify the dynamics of 3 endogenous mRNAs under basal conditions and after plasticity. We found that induction of either long-term potentiation (LTP) or metabotropic glutamate receptor long-term depression (mGluR-LTD) resulted in widespread attenuation of mRNA motility and led to an enrichment of mRNA near dendritic spines. These enhanced mRNA dynamics and availability near synapses was accompanied for some, but not all, mRNAs by enhanced translation of either a reporter or a CRISPR/Cas9 tagged endogenous protein. This dissociation allows for the enrichment of mRNAs near spines where localized signaling pathways can control which specific sets of transcripts are translated.

Results.
Tracking endogenous mRNA dynamics in live neurons.
To assess endogenous mRNA dynamics, we focused on Camk2a, Beta actin and Psd95 as they are abundant in neuronal dendrites28 and are translationally regulated by plasticity18, 22,29,30. To track these mRNAs, we employed molecular beacons ( Figure   1A Interestingly, we observed a heterogenous size distribution for the mRNA puncta -with larger pronounced particles seen near the soma (Supplemental videos [1][2][3], similar to what has been reported previously for Beta actin25. In addition, we detected a number of apparent dendritic mRNA-mRNA fusion events (Supplemental Figure 1A, Supplemental Video 4), suggesting that these mRNAs can exist in a heterogenous copy number state-contrary to proposed models of single mRNA transport in axons and dendrites29, 32,33. To assess this quantitatively, we analyzed the intensity of the individual beacon puncta (i.e. mRNA granules) and compared it to a commercially synthesized oligonucleotide containing a single ATTO647N fluorophore (see methods). Using this standard, we determined that a sizeable fraction of each mRNA exhibited an intensity consistent with a single mRNA species ( Figure 1B&C, Supplemental Figure 1B&C). Beta actin mRNAs, in particular, were often detected in a range consistent with a single copy number state ( Figure 1C, Supplemental Figure   1B&C), in line with previous reports29,32. Interestingly, while both Camk2a and Psd95 mRNAs also existed as a single copies, a large fraction of the population existed in a multimeric state ( Figure 1C, Supplemental Figure 1B&C), indicating that higher-order (containing more than a single mRNA) transporting mRNA granules exist within the dendrite. Furthermore, the relative abundance of either a single or multimeric state appears to be a transcript-specific feature-likely determined through specific sets of RNA-binding proteins bound to particular mRNAs.
To capture the dynamic behavior of each mRNA species within dendrites, we employed a semi-automated tracking approach. Using a custom-written analysis pipeline (see methods) we quantified the beacon mRNA dynamics (Figure 2A-C), distance traveled ( Figure 2D) and transport velocity ( Figure 2E) for all 3 mRNA targets.
To assess mRNA dynamics, we measured the percentage of time (during the entire imaging epoch) a detected granule spent actively moving within the dendrite (either anterogradely: away from the cell body, or retrogradely: towards the cell body) or exhibited a confined behavior. For all experiments we imaged at 1 frame per second for up to 20 minutes (see methods). Confined behavior was assigned to periods of time when an mRNA granule exhibited restricted (< 0.5um) movement within the dendrite (see methods). Similar to previous reports22,32, we found that all 3 mRNAs spent most velocities (~1um/s) for both anterograde and retrograde transport, consistent with the mixed polarity of microtubules within the dendrite34. We noted that while the majority of mRNA molecules we measured (orange and magenta highlighted puncta, Figure   2A&B) alternated between periods of confined vs. active transport, a small fraction of particles (green) showed little to no active transport during the entire imaging session.
We therefore further distinguished confined vs truly stationary events (Supplemental Figure 2A, see methods) and found ~6% of the confined population were better characterized as stationary events. For all downstream analyses, we removed these stationary events from the analysis.

Translational inhibition alters mRNA dynamics within the dendrite.
With the above measurements of basal mRNAs dynamics, we next assessed if we could alter their dynamic properties. We first assessed if perturbing the translational status of an mRNA could affect its motility. To alter the translation status of an mRNA we used two mechanistically distinct translational elongation inhibitors ( Figure 3A): puromycin, which causes release of the nascent peptide chain and ribosomal dissociation from the mRNA35, and anisomycin, which freezes elongating ribosomes on mRNAs36. As such, puromycin promote the transition to a ribosome-free mRNA state, whereas anisomycin causes ribosome accumulation on the mRNA. Using our analysis pipeline, we quantified the effects of these treatments on mRNA dynamics with Psd95 mRNA being slightly more dynamic relative to the other two mRNAs ( Figure   2). Since displacing ribosomes for Camk2α and Beta actin enhanced mRNA motilitywe predicted that freezing the ribosomes on the mRNA should lead to the opposite effect. Indeed, anisomycin treatment led to decreased mRNA motility ( Figure 3D Figure 2B&C). Furthermore, neither puromycin nor anisomycin affected the stationary population (Supplemental figure 2D&E). Given that transport was unaffected by either perturbation of translation, (Supplemental Figure 2B&C), our data is consistent with the idea that mRNAs are transported in a quiescent non-translating state23,37. Taken together, these data illustrate that the translational status of a given mRNA will affect its dynamics within the dendrite.

Plasticity stalls mRNA transport and accumulates mRNAs near dendritic spines.
We next assessed if we could modulate mRNA dynamics with physiologically relevant manipulations-specifically synaptic plasticity. We examined how mRNA dynamics are altered during two forms of protein synthesis-dependent plasticity, chemically-induced long-term potentiation (cLTP)38 and metabotropic glutamate receptor-mediated longterm depression (mGluR-LTD)39 (Supplemental Figure 3A&B). We induced cLTP To assess more precisely the location of mRNA deposition during these enhanced periods of mRNA confinement, we performed high resolution single molecule fluorescent in situ hybridization in dendrites immediately after induction of cLTP and mGluR-LTD (smFISH, see methods) ( Figure 4E&F). We measured the mean distance of an mRNA granule to its nearest dendritic spine and found that this distance decreased significantly with both cLTP and mGluR-LTD induction for all three mRNAs ( Figure 4F). Taken together with the altered dynamics observed with the molecular beacons ( Figure 4A&C) our data support increased spine association of these mRNAs during plasticity. This enhanced association may fuel local translation of these mRNAs to induce and maintain both forms of structural plasticity.

Exploring the dynamics of protein synthesis in real-time.
To assess directly whether translation of these three mRNAs is enhanced during cLTP and mGluR-LTD, we used translational reporters18 ( Figure 5) comprising an optimized super-folder GFP40 (sfGFP, see methods) flanked by the corresponding dendritically To assess if the same pattern of translation following plasticity was also observed with endogenous transcripts, we used CRISPR/Cas9 gene editing in neurons42 to tag endogenous CAMK2 or BETA ACTIN (n-terminal) or PSD-95 (c-terminal) protein with the fast-folding Venus fluorescent protein43 ( Figure 6A, Supplemental Figure 4). Similar to sfGFP, the fast-folding nature of Venus (t1/2 maturation = 2-5 minutes) allowed us to rapidly assess the translational regulation of these three proteins. All three proteins were successfully tagged and exhibited their characteristic localization patterns (Supplemental figure 5). Venus-tagged CAMK2 and BETA ACTIN were To obtain better temporal and spatial resolution on the translational responses to plasticity induction, we validated the above data with a method that couples general metabolic labeling of nascent proteins with a specific label for a protein-of-interest using the proximity ligation assay (Puro-PLA44, see methods, Supplemental Figure   7A). With this method we again examined changes in local dendritic protein synthesis during plasticity ( Figure  also support that this translation can occur, at least in part, locally within the dendrite where we detect changes in mRNA dynamics. Taken together, these data indicate that alterations in mRNA dynamics and protein synthesis underlie the manifestation of specific forms of synaptic plasticity.

Discussion
Here we investigated the interplay between mRNA dynamics and translation within neuronal dendrites during two different forms of synaptic plasticity. We characterized the dynamics and translation of three individual endogenous mRNAs: Camk2a, Beta actin and Psd95, during basal neuronal activity and plasticity. In live hippocampal neurons we provide evidence that mRNAs exist in heterogenous copy number organizational states ( Figure 1). The preference for single mRNA copy vs higher order state appears to be a transcript-specific feature, as Camk2a and Psd95 exhibit an enhanced preference for higher order multimeric states compared to the previously described single mRNA state of Beta actin29. We found that ribosome association directly influences mRNA dynamics-suggesting that mRNA translation is likely restricted to non-transporting mRNAs exclusively within the dendrite. Consistent with our data, previous work in fibroblasts45 and axons32 has shown that diffusion of Beta actin increases when ribosomes are displaced. These data fit with recent proteomic analysis on isolated mRNA transport granules46 which detected only a subset of ribosomal proteins associated with granules.
In our experiments, mRNAs were persistently sequestered, exhibiting reduced mobility, following plasticity (e.g. for entire duration of imaging, ~20 min). The mechanisms underlying the initial capture and enduring sequestration of mRNAs are not well understood. Previous work with Beta actin capture during glutamate uncaging suggest that it involves actin remodeling22. Indeed, the structural spine plasticity characteristic of both cLTP and mGluR-LTD involve modulation of actin cytoskeleton47.
Whether actin remodeling broadly promotes mRNA sequestration at or near dendritic spines remains assessed. The translation of all three mRNAs was enhanced by LTP, but LTD only enhanced the translation of Camk2a and Beta actin. Psd95 mRNA has previously been characterized as an mGluR-LTD regulated transcript16,48,49. However, following washout of the mGluR agonist (similar to the conditions used here) PSD-95 protein is rapidly degraded50. The above literature is consistent with our data, where Psd95 mRNA is retained at spines but its protein synthesis is unchanged relative to the control state. Taken together, our data dissociates the accumulation of mRNAs near synapses from their translational status22. Given specific signaling cascades are turned on by distinct forms of plasticity2, these cascades likely influence changes in post-translational modifications of RNA binding proteins on particular transcriptsregulating their translatability. Consistent with this, activation of PKA signaling alone is sufficient to enhance dendritic protein synthesis11. LTP, known classically for its dependence on CamkII signaling51 also triggers a number of other classical signaling cascades including PKA52, PKC53,54, MAPK/ERK55, PI3K56, mTOR57 and Src58. mGluR-LTD on the other hand, is less clear in its signaling requirements -it involves activation of PKC59 and PI3K/AKT/mTOR60 however the role of CamkII61,62is debated. Once sequestered by an active synapse, mRNAs could be "synaptically decoded" by the activation of these signaling pathways to determine whether and when a given mRNA species will be translated or not. E.M.S. designed experiments, supervised the project, and wrote the paper. All authors edited the paper.

Competing interests:
The authors declare no competing financial interests.
Data and materials availability: All data is available in the manuscript or the supplementary materials.

Solid-phase synthesis
Milli-Q water was treated with DEPC (0.1%) overnight and autoclaved.
The following oligonucleotides were synthesized on an ABI392 instrument: were used with a gradient from 5% to 100% MeOH in 39 minutes. ESI-MS spectra were recorded on a Bruker micrOTOF-Q device in negative ionization mode.

Hippocampal neurons
Dissociated rat hippocampal neuron cultures were prepared and maintained as to normal media or E4 with calcium and magnesium.

Imaging of molecular beacons
Investigation of the mRNA dynamics was carried out using a Leica DMi8 TIRF microscope. Differential interference contrast (DIC) microscopy was used to identify neurons with well-isolated dendrites. mRNA dynamics were recorded for 20 minutes at a rate of 1 Hz in epi-fluorescence mode. ATTO565 fluorophores were excited using a 561 nm diode laser which provided 1.8 kW/cm2 of intensity at the sample plane.
ATTO647n was imaged with a 638 nm laser which produced 2.0 kW/cm2 at the sample plane. The fluorescence was recorded with a scientific-CMOS camera (Leica-DFC9000GT). The exposure time was fixed to 200 ms and 2x2 camera binning and set the digitalization to 12 bit (low noise) was used to limit the data volume.

Quantifying mRNA dynamics
To extract information on the mRNA dynamics a custom MATLAB script was used. For each neuron a single dendrite was segmented by manually drawing its profile. The script was divided into filtering the images and rendering the puncta, tracking the mRNAs and extracting information regarding their dynamics. To render the puncta the background was subtracted by applying a mean filter. The pixel that represents the local maximum around a region of approximately 400 nm x 400 nm was then identified and selected. Puncta were rendered in a binary array. This pipeline was repeated each frame of the time series and exported as a movie. mRNA tracking was performed using the Motion-Based Multiple Object Tracking function of MATLAB taking the first 100 frames as training for the model. Any particle that did not appear in consecutive frames was discarded. After tracking was complete puncta that appear for longer than 20 frames (20 s) were retained and information such as the puncta coordinates, their distance travelled, their velocities and directionalities was extracted. Puncta that moved less than 500 nm throughout the imaging session were classified as fully stationary and were not included in directionality calculations. From the velocity data sets per puncta, the percent of time spent in the confined state was calculated by assessing the total number of frames a puncta was detected and how many frames this puncta exhibited a velocity from -500nm/s to 500nm/s. Similarly, the percent time spent anterograde or retrograde was calculated by the fraction of time >500nm/s or <-500nm/s over the total number of frames detected. For all events detected in the cell, the average time spent in confined, anterograde or retrograde was calculated.

Translational inhibitors and mRNA dynamics experiments
Beacons were transfected and imaged as described above, and puromycin and anisomycin were used at the concentrations indicated above. For assessing translational inhibitor effect on mRNA dynamics, two similarly looking neurons (containing a similar number of beacons and a similar morphology) were selected per Mattek dish. The first neuron was imaged as a control reference cell. Following the 20 minute imaging window for the control neuron, puromycin was added for 5 minutes or anisomycin was added for 20 minutes, prior to the start of imaging for the second, treated, cell. Pairwise assessment ( Figure 3B&D) between the control and treated cell per dish was used to assess the effect of the drugs on mRNA dynamic properties.

Spine size experiments
Neurons DIV17+ were transfected with myrVenus 12 hours prior to imaging and then imaged using a Leica DMi8 TIRF microscope. A 100x oil objective (HC PL APO 100x/1.47 OIL) was used to record a field-of-view of 133 µm x 133 µm using a 488 laser line. Samples were imaged for 10 minutes at baseline 1 frame every minute.
Mock treatment, mGluR-LTD or cLTP was induced for 5 minutes, and samples were imaged 1 frame/minute during the induction phase. Following washout of drugs, neurons were imaged for 90 minutes post induction. For anisomycin treatments, anisomycin was added 20 minutes prior to the start of the experiment and kept in the media continuously throughout the experiment. Drift was corrected using the built-in correct 3D drift plugin in ImageJ/FIJI. An area of 5-10 spines per dendrite were then quantified over the imaging window, using the mean size of the first 10 baseline frames for normalization.

Plasticity and mRNA dynamics experiments
Beacons were transfected and imaged and plasticity was induced as described above.
For assessing the effect of plasticity on mRNA dynamics, two similar neurons (beacon number and morphology) were selected per mattek dish. The first neuron was imaged as a control reference cell. Following the 20 minute imaging window for the control cell, plasticity was induced and the stimulation washed out, prior to commencing imaging of the second, treated, cell. Pairwise assessment ( Figure 4A&C) between the control and treated cell per dish was used to assess the effect of plasticity on mRNA dynamic properties.

Fluorescence In situ hybridization
All steps were performed at room temperature, unless stated otherwise. Hippocampal

FRAP translational reporters
A codon optimized superfolder GFP was custom synthesized (Eurofins) and cloned into a plasmid backbone driven by a CMV promoter. 3 Fluorescence intensity was measured in a 50um dendritic segment from the raw image. FRAP was calculated from background-corrected fluorescence intensity by normalizing the change in fluorescence (F-F0) to pre-photobleaching fluorescence (Fi).
Mobile fraction and t1/2 values were extracted from data fitted to a one phase exponential association.

Puro-PLA
Detection of newly synthesized proteins by proximity ligation was performed as