Assembling an intermediate filament network by dynamic cotranslation

Abstract

We have been able to observe the dynamic interactions between a specific messenger RNA (mRNA) and its protein product in vivo by studying the synthesis and assembly of peripherin intermediate filaments (IFs). The results show that peripherin mRNA-containing particles (messenger ribonucleoproteins [mRNPs]) move mainly along microtubules (MT). These mRNPs are translationally silent, initiating translation when they cease moving. Many peripherin mRNPs contain multiple mRNAs, possibly amplifying the total amount of protein synthesized within these "translation factories." This mRNA clustering is dependent on MT, regulatory sequences within the RNA and the nascent protein. Peripherin is cotranslationally assembled into insoluble, nonfilamentous particles that are precursors to the long IF that form extensive cytoskeletal networks. The results show that the motility and targeting of peripherin mRNPs, their translational control, and the assembly of an IF cytoskeletal system are linked together in a process we have termed dynamic cotranslation.

Figure 1.

FRAP analysis of a PC12 cell expressing GFP-peripherin. (A) A series of images of a single PC12 cell transfected with GFP-peripherin for 24 h, followed by NGF treatment for 4 h. This cell was photobleached in its entirety, and recovery of fluorescence was monitored by time-lapse imaging for a period of 10 min at 10-s intervals (Video 1). The cell, which is in the early stage of neurite extension, is depicted before photobleaching in i (phase and fluorescence) and ii (fluorescence). Particles were not seen immediately after photobleaching (iii), but were seen within 1 min after bleaching (iv). This rapid appearance of fluorescent particles after whole-cell photobleaching is completely inhibited by cycloheximide, as demonstrated by identical time-lapse observations over a 10-min period (Video 2 and not depicted). (B) The majority of peripherin particles move rapidly through the cytoplasm along microtubule tracks (Video 1; Helfand et al., 2004) making it extremely difficult to determine whether an individual fluorescent particle could recover its fluorescence. Therefore, PC12 cells were transfected with GFP-peripherin for 24 h, treated with NGF for 2 h, and exposed to nocodazole to disassemble microtubules and inhibit peripherin particle motility. In this experiment, one half of the cell (prebleach image shown in v) was left unbleached, as shown in the 10-s postbleach image (vi), to ensure that particles were immobilized. Time-lapse observations demonstrated that fluorescent particles appeared within 1 min (vii) and that after 10 min (viii) ∼30% of the particles (n = 200) seen before photobleaching had recovered their fluorescence (Video 3). There was no significant increase in the number of particles that recovered after a longer recovery period of 30 min (not depicted). The 10-min postbleach image (viii) and the higher magnification image (ix) are displayed as an overlay between the prebleach (green) and the 10-min after bleach (red) images. Overlap between the green and the red appear yellow and represent particles that recovered their fluorescence. Not all the particles that recovered are yellow, but are red because of slight changes in particle position and shape over the 10-min observation period, leading to an incomplete overlap. Videos 1–3 are available at http://www.jcb.org/cgi/content/full/jcb.200511033/DC1. Bars, 10 μm.

Figure 2.

Peripherin mRNA association with peripherin particles. (A and D) PC12 cells were transfected with GFP-peripherin for 24 h, followed by exposure to NGF for 4 h, and processed for FISH to determine whether peripherin particles were associated with peripherin mRNA. Biotinylated antisense RNA probes were used for detection of peripherin mRNA (red). Peripherin protein is shown in green (Fig. S1, A and B, for split images). Approximately 29% of the peripherin particles (n = 1,776 particles in 20 cells) are associated with peripherin mRNA (D, arrows). An association was defined as a minimum of 1/4 overlap between the green and red signals, and the quantitation was performed in thin regions of the cytoplasm, not in the thick, juxtanuclear regions where potential for random overlap was higher because of the large amount of both peripherin mRNA and protein. (B and E) GFP-peripherin–transfected PC12 cells were treated with puromycin before processing for FISH with probes specific for peripherin mRNA. Peripherin protein particles are shown in green and peripherin mRNA is in red (Fig. S1, C and D, for split images). This treatment reduced the association between peripherin particles and peripherin mRNA to statistically insignificant levels (n = 750 peripherin particles in 15 cells). (C and F) As a control for the specificity of the association between peripherin mRNA and peripherin particles, we also determined whether actin mRNA was associated with peripherin particles. To this end, PC12 cells that were transfected as in A were processed for FISH using biotinylated antisense-RNA probes for actin mRNA. Unlike the overlapping distributions of peripherin particles and peripherin mRNA throughout the cytoplasm (A), the actin mRNA (C, red) appears to localize in cytoplasmic regions separate from the majority of peripherin protein particles (green; Fig. S1, E and F, for split images). There is no statistically significant association between actin mRNA and peripherin particles (F; n = 500 peripherin particles in 10 cells). (G) Because ∼70% of the peripherin particles move at any given time, it was of interest to determine whether particles associated with mRNA are motile. Time-lapse images of moving GFP-peripherin particles (n = 40) were captured (i–iii) and the cells were fixed with formaldehyde directly on the microscope stage. These cells (n = 15) were then processed for FISH to detect peripherin mRNA and the motile particles were relocated in the confocal microscope (iv). None of the moving particles (green; arrowheads) was associated with peripherin mRNA (red). In contrast, some stationary particles (asterisks) within the same microscope field of view were associated with peripherin mRNA. Fig. S1 and Videos 4 and 5 are available at http://www.jcb.org/cgi/content/full/jcb.200511033/DC1. Bars, 5 μm.

Figure 3.

Simultaneous imaging of peripherin mRNA and its protein product in vivo. (A) A schematic representation of the system permitting the direct observation of both peripherin mRNA and its protein product. A construct containing the ECFP-peripherin CDS, followed by R24 and the peripherin 3′UTR, is introduced into PC12 cells along with a plasmid encoding for YFP-MS2 (a bacteriophage coat protein). When the ECFP-peripherin-R24-3′UTR construct is transcribed, the R24 sequence forms 24 stem loops, each consisting of an identical sequence of 19 nucleotides (1). The YFP-MS2 protein binds to these stem-loop structures, thereby providing a visible reporter for localizing and tracking the movements of peripherin mRNAs (2). The translation product of this YFP-MS2–tagged peripherin mRNA is CFP-peripherin (3). (B) A region of the cytoplasm of a living PC12 cell cotransfected with the pECFP-peripherin-R24-3′UTR and YFP-MS2 constructs for 24 h and subsequently treated with NGF for 4 h (i, fluorescence image; ii, phase image). Time-lapse imaging of this cell expressing both constructs showed rapid movements of YFP-MS2–tagged peripherin mRNA (red) and its protein product, CFP-peripherin (green), throughout the cell body as well as in early developing neurites (Video 6). (C) Higher magnification time-lapse observations (iii–v) of a PC12 cell cotransfected and treated with NGF as described in B reveal that peripherin mRNPs (red; arrowheads) move rapidly away from stationary peripherin protein particles (green; blue asterisks; Video 7). Movement of peripherin protein particles away from stationary mRNPs was not observed. Videos 6 and 7 are available at http://www.jcb.org/cgi/content/full/jcb.200511033/DC1. Bars: (B) 5 μm; (C) 1 μm.

Figure 4.

Quantification of peripherin mRNA. Cells transfected with ECFP-peripherin-R24-3′UTR construct for 24 h, followed by 4 h in NGF, were processed for double FISH with a Cy3-labeled probe to peripherin mRNA and a Cy5-labeled probe to the MS2-binding repeat region. This made it possible to quantitatively analyze peripherin mRNAs that could be detected in live cell observations using the MS2 tagging method. 3D Z-stacks of the Cy3 and Cy5 probe were obtained. These images were deconvolved, and the cells were analyzed by a single RNA molecule quantification method (Femino et al., 1998) to determine the number of RNA transcripts in each mRNP. (A) A color-coded map of the mRNPs containing 1–30 mRNA molecules in a single cell. The map is a compressed image of all the Z-slices. Low-level fluorescent signals caused by background noise and with values lower than that for a single probe (as in B) have been excluded. (B) A deconvolved image of a single Z-slice in the middle of the cell used for the color-coded map in A. (C) The distribution of the number of peripherin mRNA molecules per mRNP. The data show that ∼50% of the mRNPs contain one mRNA (red) and the other half contain two or more peripherin mRNAs displayed according to the other colors indicated. (D) Analyses of the distribution of total cellular peripherin mRNA show that the single mRNAs account for only 20% of the total peripherin mRNA; therefore, 80% of the total peripherin mRNA exists in clusters of two or more. Bars, 2 μm.

Figure 5.

Insights into peripherin mRNA clustering. The relative roles of the nascent peripherin protein chain, the 3′UTR of peripherin mRNA, and MT in the formation of peripherin mRNA clusters were studied. (A) To determine the role of the nascent peripherin protein in mRNA clustering, a peripherin translation-null construct (ECFP-TAA-peripherin-R24-3′UTR) was created by inserting a stop codon (TAA) between the CFP and peripherin CDSs in the ECFP-peripherin-R24-3′UTR construct. To confirm the effectiveness of the inserted stop codon in preventing the translation of peripherin, Rat2 cells that are null for peripherin were transfected with either the ECFP-peripherin-R24-3′UTR construct as a control or the ECFP-TAA-peripherin-R24-3′UTR construct. After 24 h, whole cell lysates of these transfected Rat2 cells were separated by SDS-PAGE, transferred to nitrocellulose, and immuno-blotted with antibodies against CFP (GFP antibody) and peripherin. Vimentin, the endogenous Rat2 IF protein was used as a loading control. In cells expressing the control plasmid (CFP-Periph), both the peripherin and GFP antibodies recognized a product with a molecular weight of ∼81 kD, the predicted size of the CFP-peripherin fusion protein. In cells expressing the construct with the TAA insertion, there was no detectable peripherin by immunoblotting, and the GFP antibody recognized an ∼27-kD product, the predicted size of CFP. (B) A histogram comparing the distributions of peripherin mRNP clusters in control cells (blue bars, ECFP-peripherin-R24-3′UTR), ECFP-TAA-peripherin-R24-3′UTR–transfected cells (red bars), cells expressing the 3′UTR-null construct (yellow bars, ECFP-peripherin-R24), and control cells (ECFP-peripherin-R24-3′UTR) treated with nocodazole during NGF-induced differentiation (green bars). PC12 cells were transfected with the various constructs for 24 h, followed by NGF for 4 h. The cells were prepared for FISH and quantitatively analyzed as described in Fig. 4. Error bars depict the SEM. (C) A histogram comparing the distributions of total peripherin mRNA in control cells (blue bars, ECFP-peripherin-R24-3′UTR); ECFP-TAA-peripherin-R24-3′UTR transfected cells (red bars); cells expressing the 3′UTR-null ECFP-peripherin-R24 construct (yellow bars); and control cells (ECFP-peripherin-R24-3′UTR) treated with nocodazole.

Figure 6.

Observing peripherin particle formation in vivo. (A) A live cell cotransfected with both ECFP-peripherin-R24-3′UTR and YFP-MS2 for 24 h, followed by NGF treatment for 4 h, was treated with nocodazole for 30 min to depolymerize microtubules. This treatment inhibited the motility of both peripherin particles and mRNPs, thereby facilitating the observation of peripherin particle formation in association with mRNPs. Time-lapse imaging shows an mRNP (red) that appears to be engaged in protein synthesis (green; arrowhead) in a thin peripheral region of the cytoplasm (Video 8). The green signal appears first as a dim haze (i and ii) that becomes brighter and more distinct with time (iii), resulting in the formation of a typical IF particle by 180 s (iv). (B) The association of ribosomes with peripherin mRNPs. Cells expressing both ECFP-peripherin-R24-3′UTR and YFP-MS2 were processed for indirect immunofluorescence using an antibody against the S6 subunit of ribosomes. Approximately 70% of peripherin mRNPs (vi, light blue) associated with peripherin protein particles (v; green) were also associated with ribosomes (vii, red) as shown in the triple overlay in viii. (C) Distribution of peripherin particles. 29% of peripherin particles are associated with peripherin mRNA. Of this fraction, ∼70% are also associated with ribosomes. The remaining 30% may represent complexes in which the translation process has ceased and the ribosomes have dissociated. Video 8 is available at http://www.jcb.org/cgi/content/full/jcb.200511033/DC1. Bars, 5 μm.

Figure 7.

Keratin 18 and 8 mRNA colocalize in a subpopulation of mRNPs. HeLa cells were trypsinized and replated for 2 h; then they were processed for double FISH to detect keratin 18 and 8 mRNA. (A) For keratin mRNA detection a biotinylated probe specific for keratin 18 mRNA (red) and a digoxigenin-labeled probe specific for keratin 8 mRNA (green) were used. (B) A ventral view of an early spreading HeLa cell hybridized with both K18 (red) and K8 probes (green). Approximately 15% of both types of mRNA signals appeared to colocalize or overlap completely (arrowheads; n = 8 cells; n ≅ 900 mRNPs; P ≅ 0.0004). Bar, 5 μm.

Figure 8.

A model depicting dynamic cotranslation mRNA particles. mRNPs (red) containing one or more peripherin mRNA are moved along MT (blue) by molecular motors such as kinesin and dynein. These motile mRNPs are translationally silent. When these complexes stop moving they associate with ribosomes (yellow) and engage in translational activity. The regulation of mRNP motility could involve the inactivation of their associated motors, the uncoupling of these motors, or the dissociation from their MT tracks. Many mRNPs appear to contain two or more mRNAs. As the IF protein is synthesized from multiple IF mRNAs, the protein chains are cotranslationally assembled into higher order structures that appear as nonfilamentous particles (green). Once a critical concentration of protein is achieved, the ribosomes disengage, followed by the dissociation of the mRNPs, which can either reasssociate with molecular motors or activate preexisting motors and move rapidly away along MT. The newly synthesized IF particles could assemble into IF at their site of synthesis or recruit motors to begin their journey as IF precursors to different regions of the cytoplasm. Once they reach their targets, they assemble into short IF (squiggles) that link up in tandem to form longer filaments (Video 9, available at http://www.jcb.org/cgi/content/full/jcb.200511033/DC1; Helfand et al., 2003a).