A genetically encoded probe for imaging HA-tagged protein translation, localization, and dynamics in living cells and animals

To expand the toolbox of imaging in living cells, we have engineered a new single chain variable fragment (scFv) that binds the classic linear HA epitope with high affinity and specificity in vivo. The resulting probe, which we call the HA frankenbody, is capable of lighting up in multiple colors HA-tagged nuclear, cytoplasmic, and membrane proteins in diverse living cell types. The HA frankenbody also enables state-of-the-art single-molecule experiments, which we demonstrate by tracking single mRNA translation dynamics in living U2OS cells and neurons. In combination with the SunTag, we track two mRNA species simultaneously to demonstrate comparative single-molecule studies of translation can now be done with genetically encoded tools alone. Finally, we use the HA frankenbody to precisely quantify the expression of HA tagged proteins in developing zebrafish embryos. The versatility of the HA frankenbody makes it a powerful new tool for imaging protein dynamics in vivo. One-sentence summary A genetically encodable intracellular single-chain variable fragment that selectively binds the HA epitope (YPYDVPDYA) with high affinity in living cells and organisms can be used to quantify HA-tagged protein translation, localization, and dynamics.


INTRODUCTION
Live-cell imaging is critical for tracking the dynamics of cell signaling. The discovery and development of the green fluorescent protein (GFP), for example, has literally revolutionized the field of cell biology 1,2 . GFP can be genetically fused to a protein of interest (POI) to light it up selectively and track its expression and localization in living cells and organisms. While incredibly powerful, GFP-tagging has several limitations that have made it difficult to image the full lifecycles of proteins. First, long fluorophore maturation times prevent co-translational imaging of GFP-tagged nascent peptide chains 3,4 . By the time the GFP tag folds, matures and lights up, translation of the nascent peptide chain is long over. Similarly, slow GFP maturation times have made it difficult to image short-lived transcription factors that are critical for development and embryogenesis 5 . Again, before the GFP tag has time to light up, the transcription factor may already be degraded. Second, GFP fusion tags cannot discriminate posttranslational modifications (PTM) to proteins 6 , such as acetylation, methylation and phosphorylation, nor can they discriminate protein conformational changes 7 . Without the ability to directly image these important protein subpopulations, their unique functionality is difficult to quantify and assess. Third, GFP fusion tags are relatively large, permanently attached, and dim. It is therefore difficult to detect and/or amplify fluorescence signal. This severely limits the length of time a single tagged protein can be tracked in a living cell before the protein is either photobleached or the cell is photodamaged.
To address these limitations of GFP, an alternative live-cell imaging methodology has recently emerged that uses antibody-based probes 8 . In this methodology, probes built from antibodies, such as antigen binding fragments (Fabs) 9 , single-chain variable fragments (scFvs) [10][11][12] and camelid nanobodies [13][14][15][16] , are conjugated or genetically fused with mature fluorophores. When expressed or loaded into cells, the probes dynamically bind and light up epitopes within POIs as soon as the epitopes are accessible. With this emerging methodology, it is possible to visualize and quantify the co-translational dynamics of nascent peptide chains [17][18][19][20][21] , capture the dynamics and localization of shortlived transcription factor dynamics in living embryos 5 , track single molecules for extended periods of time 22,23 , and selectively track the spatiotemporal dynamics of PTMs 24 and specific protein conformational changes 7 .
While there is great potential for antibody-based probes in live-cell imaging, so far only a small handful have been developed and tested. Arguably the most straight-forward probes to develop and test are Fab, since they can be digested from commercially available antibodies and conjugated with dyes using standard kits. For example, we have generated a variety of complementary Fab to track both endogenous histone modifications and single mRNA translation dynamics 6,17,24,25 . Unfortunately, Fab have not been widely adopted, in large part because of the difficulties associated with loading them into living systems. While robust adherent cells, such as Hela or U2OS, can be easily bead loaded in mass 26 , more sensitive cell types, including neurons and embryonic stem cells, have proven refractive to most loading procedures. Additionally, Fab are expensive to work with, typically requiring milligrams of purified full-length antibody as starting material. Fab may also change considerably from batch to batch, which can lead to unwanted variability between experiments that complicates downstream analyses.
Given the drawbacks of Fab, genetically encoded probes are an attractive alternative. Since these probes can be integrated into plasmids, they can easily be distributed and cell lines and/or transgenic organisms can be generated that stably express the probes, all without the worry of batch-to-batch variability. The only downside of genetically encoded antibody-based probes is they are not straightforward to develop.
Both scFv and camelid nanobodies require a large initial investment, as either existing hybridomas or immunized animals are necessary to get the sequences of individual antibody chains. Worse, even after specific sequences are determined, there is a good chance that antibody-based probe derived from the sequences will not fold and function properly in vivo. The problem is that antibodies have evolved to be secreted from cells, so their folding and maturation is more often than not disrupted when expressed within the reduced intracellular environment. This leads to low-expression and aggregation inside living cells. For example, in our experience developing scFv against endogenous histone modifications (i.e. mintbodies), only around 5% of antibody heavy and light-chain sequences properly fold when integrated into an scFv framework and expressed within living mammalian cells. Camelid nanobodies also suffer from misfolding, instability and aggregation when expressed inside living cells. Thus, when it comes to antibody-based probes for imaging, sequence information is usually not enough. Instead, extensive protein engineering, directed evolution, and mutagenesis are typically needed to generate an ideal probe that functions well both inside and outside living cells.
A case in point is the SunTag scFv, the only genetically encoded antibody-based probe capable of binding a small epitope co-translationally in living cells. The SunTag scFv binds a 19 aa epitope (EELLSKNYHLENEVARLKK) that is repeated 24 times within a single SunTag. As multiple scFvs tightly bind the SunTag co-translationally, fluorescence signal from individual tagged POIs can be greatly amplified, enabling both single mRNA translation imaging and long-term single molecule tracking in vivo [18][19][20][21]27,28 .
The SunTag imaging technology was developed over many years, starting with work in 1998 by the Plϋckthun lab. Briefly, after generating antibodies from immunized mice, ribosome display was used to affinity select high-performance scFv variants 29 . With the aid of crystal structures and computational design, the complementarity determining regions (CDRs or loops) of these scFv were grafted onto an in-vitro optimized hyperstable scFv scaffold mutagenized at about 20 individual residues to minimize grafting mismatch 30 . This led to a stable version 1 probe that was later tested in 2014 by Tanenbaum et al to stain mitochondria in living cells 23 . The probe was further optimized via the addition of stabilizing sfGFP and GB1 domain to eliminate aggregation at higher expression levels and the original epitope was optimized to version 4 via directed mutagenesis.
The large amount of work required to develop the SunTag imaging technology highlights the difficulty of generating new antibody-based probes that are suitable for demanding live-cell imaging applications. To confront this problem, we here develop an alternative strategy for developing new scFvs for live-cell imaging. To bypass many of the difficulties associated with probe development, we begin with a diverse set of scFv scaffolds that have already been proven to fold properly and function within the reduced cytoplasm of living cells. Onto these scaffolds, we loop graft all six CDRs from an epitopespecific antibody 31 . Depending on the compatibility of the scaffold and CDRs, this produces a hybrid scFv that retains the folding stability of the scaffold, while acquiring the binding specificity of the grafted CDRs.
To demonstrate the efficiency of our approach, we use it to generate two new hybrid scFvs (with a 40% success rate) that bind to the classic linear HA epitope (YPYDVPDYA) 32 in living cells and organisms. We extensively tested one of these hybrid scFvs, which we call the "HA frankenbody," by using it to label in multiple colors a variety of proteins in diverse live-cell environments. In all cases we tested, the HA frankenbody binds its target HA-epitope with high affinity and selectivity, making it suitable for a wide range of imaging applications. This makes the HA frankenbody a powerful new imaging tool to study complex protein dynamics in living systems with high spatiotemporal resolution. We anticipate that as additional antibody sequences become available, our general strategy to develop frankenbodies will become easier and more efficient, leading to a complementary set of scFvs for highly multiplexed imaging of the full lifecycle of proteins in vitro and in vivo.

Design strategy and initial screening of frankenbodies
We engineered the HA-frankenbody from six complementarity determining regions (CDRs, or loops) within the heavy and light chains of a published anti-HA scFv (parental full-length antibody: 12CA5) 33,34 (Fig. 1A). On its own, this wildtype anti-HA scFv (wtHA-scFv) does not fold properly in the reduced intracellular environment, and therefore displays little to no affinity for HA epitopes in living human U2OS cells 33 . We figured we could address this folding issue by grafting the CDRs onto more stable and sequence similar scFv scaffolds (Fig. 1A). To test this, we selected five scFv scaffolds that have already been successfully used for live-cell imaging purposes and that have a wide range of sequence identity compared to wtHA-scFv. In particular, the sequences of the heavy chain variable regions (VH) were 47-89% identical, while the sequences of the light chain variable regions (VL) were 50-67% identical. The five scaffolds we chose included (1) an scFv that specifically binds histone H4 mono-methylated at Lysine 20 (H4K20me; 15F11) 35 ; (2) an H3K9ac specific scFv (13C7) 36 ; (3) an H420me2-specific scFv (2E2, unpublished); (4) a SunTag-specific scFv 23 ; and (5) a bone Gla protein (BGP) specific scFv (KTM219) 33 . Among these scaffolds, 15F11 and 2E2 have the greatest sequence identity compared to the wtscFv-HA (Fig. 1B). We therefore hypothesized there would be a higher chance of grafting success with either of these scaffolds.
To verify our hypothesis, we grafted the anti-HA scFv CDR loops onto our five chosen scFv scaffolds. We refer to the resulting five chimeric scFvs as . For example, 15 11 specifies the chimeric scFv that was generated by loop grafting the anti-HA CDRs onto the 15F11 scFv scaffold. To screen our chimeras, we fused each with the green fluorescent protein mEGFP and co-transfected each of the resulting plasmids into U2OS cells, together with a plasmid encoding 4×HA-tagged histone H2B fused to the red fluorescent protein mCherry (HA-mCh-H2B). If a chimeric scFv binds to the HA epitope in living cells, it should co-localize with the HA-tagged H2B in the nucleus, as shown in Fig. 1C. Live-cell imaging revealed 15 11 and 2 2 were superior, displaying little to no misfolding and/or aggregation, strong expression, and excellent co-localization with H2B in the nucleus. In contrast, the other three scFvs did not show any co-localization signal ( Fig. 1D). Moreover, in control cells lacking HA tags, both 15 11 and 2 2 displayed uniform expression (Fig. 1E), indicative of free diffusion without non-specific binding.
According to our screen, both 15 11 and 2 2 worked equivalently well in living cells. We chose the 15 11 variant for additional screening, which we herein refer to as the "HA frankenbody" due to its construction via grafting.

Multicolor labeling of HA-tagged proteins in diverse intracellular environments
We tested the HA frankenbody in a variety of different settings. First, as we already demonstrated in the initial screen, the HA frankenbody colocalizes with HA-tagged histone H2B in the nucleus of living U2OS cells ( Fig. 2A). This demonstrates frankenbodies can pass through the nuclear pore and bind target nuclear proteins. We next wanted to test if the HA frankenbody can work equally well in the cell cytoplasm, another reducing environment that can interfere with intradomain disulfide bond formation 30 . We tested this by creating a new target plasmid encoding the cytoplasmic protein β-actin fused with a 4×HA-tag and mCherry (HA-mCh-β-actin). When this plasmid was expressed in cells, co-expressed frankenbodies again took on the distinct localization pattern of their targets, in this case colocalizing with HA-mCh-β-actin along filamentous actin fibers (Fig. 2B). We therefore conclude that both nuclear and cytoplasmic HA-tagged proteins can be selectively labeled with high efficiency by the HA frankenbody in living human cells.
To test if frankenbodies could also work in more sensitive cell types, we co- To ensure the HA frankenbody is as broadly applicable as possible, we wanted to test if it could tolerate different fluorescent protein fusion partners that might be needed in multicolor imaging applications. GFP and its derivatives are generally superior fusion partners because their high stability actually helps stabilize and solubilize the tagged protein. This was observed, for example, during the development of the SunTag scFv 23 .
To test how well the HA frankenbody tolerates different tags, we fused it to mCherry, HaloTag 38 and SNAP-tag 39 . Encouragingly, all three frankenbody constructs colocalized with HA-tagged H2B in the nucleus of living U2OS cells, similar to the original GFP-tagged frankenbody (Fig. 2D, upper). Furthermore, all three constructs displayed relatively diffuse localization patterns in cells lacking the HA-tag (Fig. 2D, lower). These data indicate the HA frankenbody can indeed tolerate different fusion partners, including green (GFP), red (mCherry), and far-red (SNAP-tag/HaloTag with far-red ligands) fluorescent proteins. Thus, the HA frankenbody can label HA-tagged proteins in a rainbow of colors in living cells.

Immunostaining and Western blotting with purified recombinant frankenbody
We wondered if the HA frankenbody has the potential to replace costly anti-HA antibodies in traditional assays such as immunostaining and Western blots. To test this, we cloned the frankenbody gene fused with mEGFP and a hexahistidine tag into an E.
coli expression vector, pET23b. We expressed the recombinant frankenbody and purified the soluble portion from E. coli. Using this purified fraction, we immunostained fixed cells expressing HA-tagged H2B and HA-tagged β-actin. Similar to our observations in living cells, the purified HA frankenbody beautifully stained both the HA-tagged nuclear and cytoplasmic proteins, but now with almost no observable background signal (Fig. 3A, B).
We next tested the suitability of the HA frankenbody for Western blotting. For this, we harvested U2OS cells 24h after transiently transfecting either HA-tagged H2B or HAtagged β-actin. In contrast to the parental 12CA5 full-length anti-HA antibody, which was stained using a secondary antibody and visualized via a sensitive chemiluminescent substrate, the frankenbody Western blot used the GFP signal alone for detection.
Nevertheless, similar dark and sharp bands were seen on the frankenbody membrane as the 12CA5 membrane (Fig. 3C). Although several of the bands were dimmer than those seen using 12C15, we attributed the difference to signal amplification from the secondary antibody and the chemiluminescent substrate. In principle, a similar signal-to-noise could be attained using the GFP-tagged HA frankenbody with secondary antibodies against GFP. Together, our Western blot and immunostaining results strongly suggest the HA frankenbody can serve as a cost-effective replacement for full-length HA antibody in widely used in vitro applications.

HA frankenbody specifically binds the HA epitope for minutes at a time in live cells
An ideal imaging probe binds its target with high affinity to maximize the fraction of target epitopes bound and thereby increase signal-to-noise. In general, a high bound fraction is established by a large ratio of probe:target binding off to on times; in other words, the time a probe remains bound to a target is ideally much longer than the time it takes a probe to bind a target. Although the latter depends sensitively on the concentrations of both target and probe, the former is a fixed biophysical parameter that is useful for planning and interpreting experiments. With this in mind, we set out to measure the length of time the HA frankenbody remains bound to the HA epitope in living cells.
To accurately measure the binding kinetics of HA frankenbody, we performed

Tracking single mRNA translation in living U2OS cells with the HA frankenbody
A major advantage of the HA frankenbody over other intrabodies is the small size and linearity of its epitope, just 9 aa in length. This means the epitope is quickly translated by the ribosome and becomes available for binding almost immediately. The HA frankenbody therefore has the potential to bind HA-tagged nascent peptides cotranslationally, much like purified anti-HA antibody fragments are capable of 17 . By simply repeating the HA epitope multiple times within a tag, fluorescence can furthermore be amplified for sensitive single molecule tracking 22 .
To test the potential of HA frankenbody for imaging translation dynamics, we cotransfected a GFP-tagged version (FB-GFP) into U2OS cells together with our standard translation reporter. The reporter encodes a 10x HA spaghetti monster tag N-terminally fused to the nuclear protein KDM5B. In addition, the reporter contains a 24x MS2 stem loop repeat in the 3' untranslated region to label and track single mRNA 17  To ensure the frankenbody can also light up translation sites in multiple colors, we repeated experiments, but now using our other frankenbody constructs. For this, we cotransfected cells with either our mCherry or HaloTag frankenbody plasmids (FB-mCh or FB-Halo), together with our KDM5B translation reporter (Fig. 5D). In both cases, we could easily detect bright translation sites that responded to puromycin treatment ( Fig. 5E and 5F, Movie S3 and S4). Collectively, these data demonstrate the HA frankenbody can be used to image translation in three colors spanning the imaging spectrum, demonstrating its potential and flexibility in multicolor experiments.

Multiplexed imaging of single mRNA translation dynamics in living U2OS cells
With the ability to image translation in more than one color, we combined the HA frankenbody with the SunTag imaging system to simultaneously quantify the translation kinetics of two distinct mRNA species co-expressed in single living cells. Previously, the SunTag scFv has been fused to GFP (Sun-GFP) to monitor translation [18][19][20][21] . We therefore coupled this probe 23 (after removing its HA epitope) with our complementary mCherrytagged frankenbody probe (FB-mCh) (Fig. 6A). Co-transfecting these into living U2OS cells together with plasmids encoding SunTag-kif18b and smHA-KDM5B, we observed two distinct types of translation sites, those labeled entirely green by Sun-GFP and those labeled entirely magenta by frankenbody (Fig. 6B, Movie S5). After co-tracking hundreds of these translation sites, we quantified their mobilities. This revealed they both move with a similar kinetic, having a diffusion coefficient of 0.024±0.003 µm 2 /sec for Sun-GFP and 0.019±0.001 µm 2 /sec for FB-mCh. The similarity of their movement despite their different sequences shows that different mRNA types can nonetheless be translated in similar micro-environments.
Since both translation sites were labeled in different colors, we next wondered if they ever co-localized. The observation of colocalized translation sites would provide further evidence for multi-RNA "translation factories," which our group 17 and another 21 have recently observed. In particular, we showed that approximately 5% of our smHA-KDM5B mRNA reporter is within multi-RNA factories 17 . Although we looked, we were unable to detect co-localized magenta and green translation sites. Since the smHA-KDM5B mRNA has different 3' and 5' untranslated regions (UTRs) as well as a different open reading frame (ORF) than SunTag-kif18b mRNA, while the translation reporters, used for the colocalization observation in our previous study, share identical 3' and 5' UTRs, but slightly different ORFs, these suggest that the composition of factories may be dictated in part through mRNA sequence elements.

Monitoring local translation in living neurons with the HA frankenbody
Local translation is implicated in neuronal plasticity, memory formation, and disease 42 . The ability to image local translation at the single molecule level in living neurons would therefore be a valuable research tool to better understand these processes.
To facilitate this, we tested if the HA frankenbody could be used to monitor single mRNA translation in living primary rat cortical neurons. When we co-transfected these cells with our KDM5B reporter 17 and the GFP-tagged HA frankenbody (Fig. 7A), we were able to see distinct bright spots that diffused throughout the cell cytoplasm (Fig. 7B, Movie S6), reminiscent of the translation sites we had observed in U2OS cells. Again, we confirmed these were indeed translation sites by adding the translational inhibitor puromycin (Fig. 7C, Movie S7). Just seconds after the drug was added, the bright spots disappeared, just as they had in U2OS cells.
Unlike the mobility of mRNA we observed in U2OS cells, mRNA within neuronal dendrites displayed obvious directed motion events. For example, we regularly saw mRNA zip along linear paths within dendrites, achieving rapid translocations with rapid retrograde and anterograde transport over large distances up to 8 microns (Fig. 7D, Movie S6). The strong frankenbody signal within these fast-moving sites suggest translation is still active, despite the motored movement. These data therefore provide further support for a model in which translation is not necessarily repressed during trafficking 20 .

Monitoring zebrafish development with the HA frankenbody
Arguably the most demanding types of imaging applications are in whole living animals. To verify that the HA frankenbody can also be applied in this way, we used it to monitor development in zebrafish embryos. The environment within embryos is complex and contains many potential non-specific binding targets. To express HA frankenbody in this complex environment, we microinjected mRNA encoding GFP-tagged frankenbody (FB-GFP) and HA-mCh-H2B into the yolk of one-cell stage zebrafish eggs. With this setup, HA frankenbody is expressed (i.e. translated) immediately without having to wait for the onset of transcription after the maternal-zygotic transition 43 . Following the initial injection, we co-loaded a positive control Fab (Cy5 conjugated) which specifically binds and lights up endogenous histone acetylation (H3K9ac) in the cell nucleus 25 (Fig. S3A).
We began imaging embryo development with the HA frankenbody around the fouror eight-cell stage. At all timepoints we could see colocalization of the frankenbody with the HA-mCh-H2B target (Fig. 8A, Movie S8), although at earlier timepoints the concentrations of both were lower and therefore marked the nuclei only dimly compared to the positive control Fab. Nevertheless, we could detect all three signals in the nuclei of single mother and daughter cells throughout the entire 80-minute imaging time course (Fig. 8B, upper). Moreover, the signal from the frankenbody in the nucleus tightly correlated with the target HA-mCh-H2B signal, increasing steadily from a nuclear-tocytoplasmic ratio of one to nearly 2.5 (Fig. 8B, lower). This single-cell trend was also observed in the population of cells (Fig. 8C). As a negative control, we repeated experiments in zebrafish embryos lacking target HA-mCh-H2B. In this case, the frankenbody was evenly distributed throughout the embryo (Fig. S3B, C, Movie S9), displaying a nuclear-to-cytoplasmic ratio close to one at all times. This confirms the HA frankenbody binds the HA epitope selectively and tightly in vivo, so it can be used to accurately monitor the concentration of target HA-tagged proteins in living organisms. including primary rat cortical neurons and zebrafish embryos.

DISCUSSION
A major advantage of the HA frankenbody is it can be used to image single mRNA translation dynamics in living cells. This is because the target HA epitope (YPYDVPDYA, 9 aa) is small and linear. It therefore emerges quickly from the ribosome, so it can be cotranslationally labeled by frankenbody almost immediately. It is also short enough to repeat many times in a single tag for signal amplification, as in the HA spaghetti monster tag 22 . In principle, epitopes could even be made conditionally accessible within a protein to monitor conformational changes. In contrast, almost all other antibody-based probes bind to 3D epitopes that span a large length of linear sequence space. In general, 3D epitopes take a relatively long time to be translated and fold before they become accessible for probe binding. Furthermore, they are too big to repeat more than a few times, so fluorescence is difficult to amplify to the degree necessary for single molecule tracking 22,23 .
Here we used the HA frankenbody to image single mRNA translation in both living U2OS cells and primary neurons. Unlike Fab, which cause neurons to peel during the loading procedure, HA frankenbody can be expressed in neurons without issue via transfection. We exploited this to demonstrate the mobility of translating mRNA is cell-type dependent. While our KDM5B translation reporter mRNA displayed largely nondirectional, diffusive movement in U2OS cells, in neurons they were often motored.
Neurons can be notoriously long, so motored mRNA movement provides a solution to the unique challenge of local protein production in distal neuronal dendrites and axons 44,45 .
An open question is if translation is repressed during transport. On the one hand, certain mRNA are known to be actively repressed during trafficking 3  Given our success with CDR grafting in this study, in principle it should be relatively straightforward to generate more scFvs that bind additional targets besides the HA epitope tag. However, the functionality of CDR grafted scFvs is still difficult to predict 30 , so it remains unclear how generalizable the method is. According to our initial screen of scaffolds for frankenbodies, scFvs that have similar sequences will generally have a higher chance of being compatible grafting partners. We are therefore optimistic that as the price of determining antibody sequences continues to decrease, more scFvs will be constructed, tested and verified to fold and function in living cells. The availability of compatible grafting partners will therefore only increase, meaning less effort will be required to generate additional frankenbodies in the future. Ultimately, we envision a panel of complementary frankenbodies will be available to image in multiple colors the full lifecycles of proteins in vivo with high spatiotemporal resolution.

ACKNOWLEDGMENTS:
We thank all members of the Stasevich lab for input and helpful

Plasmids Construction
Each chimeric anti-HA scFv tested in this study was constructed by grafting six CDR loops of an anti-HA antibody 12CA5 onto each selected scFv scaffold. The 15

Nascent Chain Tracking
Nascent chain tracking was performed as described previously 17  For multiplexed imaging, 2 reporter plasmids, smHA-KDM5B and SunTag-kif18b, as well as 2 probes, FB-mCh and Sun-GFP (with the HA epitope removed), were transiently transfected into U2OS cells plated on a MatTek chamber 4~6 hours before imaging. 3 hours later, the medium of the cells was changed to phenol-red-free complete DMEM medium. The cells were then ready for imaging.

Imaging condition for translation and colocalization assays
To image single mRNAs and their translation status with FB, a custom-built widefield fluorescence microscope based on an inclined illumination (HILO) scheme was used 17,57 .
Briefly z-stacks with a step size of 500 nm (6 µm in total) were imaged using the piezoelectric stage. This resulted in our total cellular imaging rate of 1 Hz for imaging either red or green signals, and 0.5 Hz for imaging both red and green signals regardless of far-red imaging.
For Fig. 1D, 1E, 2A, and 2D, a single plane of the cells was imaged continuously at 6.5 Hz for 100 time points and averaged throughout the time. For Fig. 2

Particle tracking
Single particle detection and tracking was performed on maximum intensity projection images with custom Mathematica (Wolfram Research) code, as previously described 17 .
Briefly, the images were processed with a bandpass filter to highlight particles, and then binarized to detect their intensity-centroids as positions using the built-in Mathematica routine ComponentMeasurements. Detected particles were tracked and linked through time via a nearest neighbor search. The precise coordinates (super-resolved locations) of mRNAs and translation sites were determined by fitting (using the built-in Mathematica routine NonlinearModelFit) the original images to 2D Gaussians of the following form: where is the background fluorescence, the particle intensity, ( 0 , 0 ) the particle location, and ( , ) spreads of the particle. The offset between the two cameras was registered using the built-in Mathematica routine FindGeometricTransform to find the transform function that best aligned the fitted positions of 100 nm diameter Tetraspeck beads evenly spread out across the image field-of-view.
For Fig. 6C, average mean squared displacements were calculated from the Gaussian-fitted coordinates (from 2D maximum intensity projection images). The diffusion constant was obtained by fitting the first 10 time points to a line with slope = 4 , where is the diffusion coefficient.

Puromycin Treatment
U2OS cells transiently transfected with smHA-KDM5B and FB or bead loaded with smHA-KDM5B, FB and MCP-HaloTag were imaged as above with 10 s intervals between frames. After acquiring 5 or 10 timepoints as pre-treatment images, cells were treated with a final concentration of 0.2 mg/mL puromycin right before acquiring the 6 th or 11 th timepoint. After puromycin was added, the cells were imaged under the same conditions used for the pre-treatment imaging until the translation spots disappeared.

Neuron Culture and Transfection
Rat cortical neurons were obtained from the discarded cortices of embryonic day (E)18 fetuses which were previously dissected to obtain the hippocampus, and frozen in ComponentMeasurements requires binary masks of the objects to be measured. Binary masks of the nuclei were made using the built-in Mathematica function Binarize with an appropriate intensity threshold to highlight just nuclei in images from Cy5-labeled Fab (specific to endogenous histone H3 Lys9 acetylation). Masks of the cytoplasm around each nuclei were made by dilating the nuclear masks by 4 pixels (using the built-in command Dilation) and then subtracting from the dilated mask the original nuclear masks dilated by 1 pixel. This creates ring-like masks around each nuclei, from which the average cytoplasmic intensity was measured.

Surface plasma resonance
Binding kinetics of purified FB to the HA epitope tag was measured by surface plasma resonance (OpenSPR, Nicoyalife). After biotin-labeled HA peptide was captured by a Streptavidin sensor chip (Nicoyalife), diluted purified FB-GFP in PBS running buffer, pH 7.4, was slowly flowed over the sensor chip for 5 min to allow interaction. The running buffer was then allowed to flow for 10 min to collect the dissociation data. The non-specific