mRNP components are differentially expressed in quiescent muscle stem cells versus myofibers
To investigate mRNP protein distribution in muscle, we used isolated murine myofibers ex vivo, complete with resident MuSCs in their niche. At a sub-cellular level, mRNP granule components are known to partition between a diffuse cytoplasmic distribution and punctate granules, where puncta represent functional complexes of RNA and proteins [32, 33]. Immunofluorescence analysis revealed that mRNP granule proteins are organized in puncta that were enriched in quiescent Pax7+ MuSC (Fig. 1). Myofibers also showed puncta, organized in striated patterns that suggest association with underlying cytoskeletal elements. Specifically, the translational repressor Fmrp showed punctate immunolabelling that was highly enriched in the cytoplasm of Pax7+ MuSC (Fig. 1B-D), but also associated with sarcomeres in myofibers (Fig. 1E-G), with distinct non-sarcomeric enrichment in the cytoplasm adjacent to Pax7- myonuclei (Fig. 1C, C’). Interestingly, Fmrp was also located in MuSC nuclei (Fig. 1B, B’, D, D’), but not in myonuclei. Another translational repressor GW182, which is involved in the Ago-miRNA pathway, showed a similar distribution to Fmrp: discrete cytoplasmic puncta in Pax7+ MuSC and in zones near myonuclei, with smaller puncta in myofibers, arranged in a distinct pattern reflecting sarcomeric organisation (Fig. 1K). Thus, proteins implicated in translational repression are located in mRNP granules clearly evident in MuSC.
To determine the distribution of proteins involved in mRNA turnover, we examined expression of key regulators of mRNA, the decapping enhancer Dcp1a and the 3’5’ exoribonuclease Xrn1 [14]. Dcp1a protein was not detected in MuSC (marked by MuSC-enriched membrane protein Caveolin 1 (Cav) (Fig. 1L), but formed a fine striated pattern in the myofiber cytoplasm, largely perpendicular to expected sarcomeric organization. Similarly, Xrn1 was not present in MuSC, but exhibited a clear striated pattern in myofibers (Fig. 1M). These observations reveal that components of the mRNA storage/stabilization complex (Fmrp, GW182) are highly expressed in the nuclei of quiescent MuSC, while the nonsense-mediated decay complex components (Dcp1a, Xrn1) are not.
Fmrp knockout mice exhibit altered muscle stem cell proliferation
To examine whether Fmrp observed in mRNP granules in quiescent MuSC is important for stem cell function in vivo, we analyzed skeletal muscle from the Fmr1 knockout (Fmr1-/-) mouse. First, quantification of cross-sectional area of muscle fibers in cryo-sections of adult tibialis anterior muscle revealed that muscle fibers in Fmr1-/- muscle showed drastically reduced caliber [mean±SD of 619 μm2 ± 200] compared to age-matched wild-type (WT) mice [mean±SD of 1518 μm2 ± 438, n=250 fibers; p value <0.0001] (Fig 2A, B). Second, FACS-isolated Fmr1-/- CD45-VCAM-1+ MuSCs proliferate less compared to WT controls (Fig 2C). While the proportions of sorted cells that were MuSCs were similar in Fmr1-/- and WT (Fig. 2C), when equal numbers were plated, there were fewer Fmr1-/- cells over the course of 6 days in culture compared to WT (Fig. 2D). Together, these preliminary results indicate that Fmrp expression is required for achieving normal fiber caliber in post-natal adult skeletal muscle, and that this phenotype may be linked to a defect of knockout MuSC in proliferation, reactivation from quiescence or survival in culture.
Differential expression of mRNP proteins in quiescent, proliferating and differentiated muscle cells in culture
To explore the muscle cell-intrinsic functions of Fmrp, we examined the expression of a series of mRNP proteins (schematized in Fig. 3A), in a tractable adult MuSC-derived murine C2C12 culture model that permits the generation of pure populations of proliferating myoblasts (MB), quiescent myoblasts (G0) or differentiated myotubes (MT) [2, 25]. In particular, this model allows the entry to mitotic quiescence to be examined, a limitation in other similar techniques. The three cellular states were distinguished using expression of Myogenin (Myog), a master regulator of myogenic differentiation, and Cyclin D1, a canonical marker of proliferation: MB are Cyclin D1+ Myog-, MT are Cyclin D1- Myog+ and G0 are Cyclin D1- Myog- (Fig. 3B). We investigated the abundance of Fmrp and Dcp1a, together with other categories of mRNP components, namely: (i) proteins involved in translation repression and formation of SGs (Fmrp, Fxr1, Tia1) [10, 34], and translation initiation (eIF-4e), (ii) proteins involved in the PB nonsense mediated mRNA decay pathway (Dcp1a, Pat1 and Edc4), and (iii) proteins known to shuttle between these two complexes, (LSm4, Xrn1, Gw182 and Ago2) [10, 33, 35] (Fig. 3C, D). Briefly, the translation repressors Fmrp and Tia1 continued to be expressed in G0 at levels equivalent to MB or greater, but in MT, Fmrp was down-regulated. Consistent with reversible suppression of translation in G0, eIF-4e, the cap-binding component of the rate limiting translation initiator eIF-4F complex, is strongly down-regulated in G0, but up-regulated in MT. Dcp1a (and another mRNA decay factor Edc4) were less abundant in both G0 and MT compared to MB. Overall, the quantitative analysis of mRNP granule protein expression (Fig 3C,D) revealed that in G0, proteins involved in mRNA turnover such as Dcp1a, were under-represented by comparison to both MB and MT, while proteins involved in mRNA storage/stabilization/translational stalling (Fmrp, Tia1) were sustained.
To assess whether changes in expression of mRNP proteins resulted from changes in expression of their mRNAs, we used bio-informatic analysis of available transcriptome data [36], which revealed that genes encoding translational stalling complex proteins, such as Fmrp and Tia1 are up-regulated in G0 (Table 1). Taken together, this analysis suggests that compared to proliferating MB, the nonsense-mediated mRNA degrading machinery is suppressed in both non-dividing states (G0 and MT), but that translational repression capacity is maintained in G0.
Distinct organization and dynamics of mRNP granules in two mitotically inactive states
To compare the distribution and dynamics of mRNP complexes in different cellular states in culture, we examined the staining pattern of Fmrp and Dcp1a using immunofluorescence confocal microscopy (Fig. 4A). As active mRNPs self-assemble into observable puncta and disassemble upon releasing bound mRNA [37], sub-cellular staining patterns are a reflection of the activity state of these complexes. In asynchronously proliferating MB, Dcp1a and Fmrp were present in small, numerous, non-overlapping cytoplasmic puncta, consistent with their participation in distinct complexes with distinct functions. In G0, whereas Dcp1a immunolabelling was low and diffuse (not punctate), the size and intensity of cytoplasmic Fmrp granules dramatically increased, and nuclear-localized Fmrp granules were also prominent, while total Fmrp protein level was maintained (Fig. 3D), suggesting enhanced granule assembly, and greater translational repression. Notably, mRNP immuno-detection patterns in cultured G0 cells (Fig. 4A) reflected the patterns observed in vivo in MuSC (Fig. 1) with respect to (i) increased Fmrp puncta and reduced Dcp1a puncta and (ii) the appearance of Fmrp puncta in the G0 nucleus.
We next tested the effect of cell-cycle reactivation on mRNP granules. Three hr after reactivation from quiescence (R3) Fmrp puncta disappeared and Dcp1a puncta re-appeared, consistent with the abundance of Dcp1a protein in cycling MB (Fig. 3D). In MT, Fmrp was organized as small, dispersed cytoplasmic granules, while Dcp1a puncta were reduced compared to MB (Fig. 4A). There was a general similarity in abundance of puncta in culture and in vivo: i.e. similar patterns in MT and myofiber versus G0 and MuSC, suggesting an association of particular mRNP granule dynamics with these distinct cellular states. The degree to which three additional mRNP proteins (Edc4, Pat1, Ago2) were organized into puncta also varied between cellular states (Fig. S1). Taken together, these immuno-localization studies indicate that translational repression complexes (Fmrp, Pat1) are more prominent in G0 than in MB and MT, and that nonsense-mediated mRNA decay complexes (Dcp1a, Edc4, Ago2) are more prominent in MB and MT than in G0, both of which are consistent with earlier reports of transcript stabilization in quiescent cells [37, 38, 39].
Global translation rates and expression of translation initiation factors are suppressed in G0
To compare global translation rates between the proliferating, differentiated and quiescent states, we used incorporation of O-propargyl-puromycin (OPP) to biosynthetically label nascent proteins (Fig. 4B,C). Rapidly growing MB pulsed with OPP for 1 hr showed strongly labeled cytoplasm and nucleoli, possibly reflecting the nucleolar location of newly-synthesized ribosomal proteins during ribosome biogenesis. In fused MT, cytoplasmic OPP labeling predominated, likely reflecting greater synthesis of sarcomeric and other non-ribosomal proteins. By contrast, G0 cells showed low and variable OPP labeling of the cytoplasm. Many G0 cells were essentially unlabeled above background levels (Fig. 4B,C). Moreover, nucleoli could not be distinguished (Fig. 4B). These findings indicate lower rates of protein synthesis and ribosome assembly in G0 cells.
To investigate translation by an independent method, we analysed expression of two translation initiation factors, eIF-4E (the rate limiting factor in cap-dependent translation that also regulates mRNA export) and eIF4G (a scaffold for assembly of the eIF-4F complex comprising eIF-4E, eIF-4G and eIF-4A on the 5’ cap). Both proteins were present in MB and MT, but both were much reduced in G0 (Fig. 4D), and strongly re-induced in a punctate pattern after re-activation from quiescence (R3), when protein synthesis begins to recover (Fig. 4C,D). The altered abundance of eIF-4E was consistent with our western analysis (Fig. 3C,D). These initiation factors showed some nuclear localization in G0, which was greatly enhanced during synchronous reactivation, but not detected in either cycling MB or MT, possibly reflecting involvement in upstream functions such as mRNA export, that are important for cell cycle re-entry (Fig. 4D). Taken together, these findings are consistent with the notion of G0 as a state where global translational repression is coupled to mRNA stabilization in granules, keeping cells primed for cell cycle re-entry [39, 40].
Quiescent myoblasts exhibit puromycin-resistant mRNP complexes in G0
Quiescent cells show low transcriptional activity compared to proliferative or differentiated states. Although many transcripts are specifically induced in G0 [1, 41], and some must be translated into proteins required for the maintenance of quiescence(44) , the data above suggest that a number of G0-induced transcripts may also be sequestered in non-polysomal compartments, to be mobilized for protein synthesis required for the return to the cell cycle [17]. To visualize ongoing translation activity directly, we analysed steady state polysome profiles in each cellular state. To ensure polysome integrity during isolation and display, cells were treated briefly with cycloheximide (CHX) prior to lysis, to arrest translating ribosomes on mRNAs, followed by separation on sucrose density gradients (Fig. 5A-C). The profile of RNA-protein complexes was quantified in density-separated fractions and analysed by immuno-blotting. A second profile was run from cells in each state that were treated briefly with puromycin (Puro), that successfully disengaged mRNA from translating ribosomes, removing the polysome profile (Fig. 5A-C). With respect to mRNP granule dynamics, CHX rapidly dissociates mRNP granules, whereas Puro promotes their assembly [10].
In MB, in addition to the strong ribosomal subunit peaks in fractions 3, 4, 5 (containing free 40S, 60S subunits and 80S monosomes, respectively), a range of polysome peaks was visible (fractions 6-9), which was sensitive to Puro treatment, showing that the cells were engaged in active translation (Fig. 5A). Western blot profiles confirmed that the ribosome-containing heavier fractions (3-9, marked by the presence of the ribosomal protein P0) were largely devoid of mRNP proteins Xrn1 and Fmrp, which were enriched in the non-polysomal fractions 1-2. On treatment with Puro, disruption of polysomes was evident and accompanied by loss of P0 protein from fractions 6-9 (Fig. 5A). MT also displayed very active translation, showing monosome and polysome peaks similar to the profile in MB (Fig 5B). Similarly, Xrn1 and Fmrp were detected at low levels in fraction 6, but otherwise these mRNP proteins were largely absent from the polysome fractions 7-9. As in proliferating MB, Puro treatment led to the loss of polysome peaks in MT.
In G0 cells, by contrast, polysomes were nearly undetectable and fewer monosomes were seen, consistent with the accumulation of P0 in the lighter complexes (Fractions 1-4 (Fig. 5C). Nevertheless, in G0 cells, P0 persisted in high molecular weight complexes (Fractions 6-9 in Puro vs CHX), that were insensitive to Puro (Fig. 5C). Together, these observations suggest the presence of heavy mRNP complexes in G0 cells that are not engaged in active translation. These heavy mRNPs could be stalled polysomes, or mRNA captured in other heterogeneous paused complexes along with ribosomes, but not undergoing active translation. Treatment of G0 cells with Puro increased mRNPs in the heavy fraction 7-9, the opposite of the effect of Puro in MB (Fig. 5A,C). The sustained enrichment in the heavier fractions (7-9) in puromycin-treated G0 cells suggests that ribosomal proteins are present in non-canonical high molecular weight complexes in G0 cells, which are absent in MB and MT. Taken together with OPP incorporation and eIF-4E expression levels, these results demonstrate that proliferating and differentiated cells are actively engaged in translation, while quiescent cells show markedly suppressed protein synthesis, potentially associated with sequestered and stalled ribosomes.
Transcripts accumulate in a non-polysomal mRNP compartment specifically in G0
To probe the distribution of specific transcripts between actively translating and inactive sequestered compartments, we used qRT-PCR analysis on RNA isolated from the mRNP, monosome- and polysome-containing fractions (Fig. 5D). We selected mRNAs whose levels are (i) unchanged (Gapdh) (ii) suppressed in G0 (Cyclin D1, MyoD) or (iii) maintained/induced in G0 (Myf5, Cdkn1b/p27). Consistent with the bulk polysome profile that shows low polysome assembly in G0, all transcripts tested show substantial enrichment in mRNP and monosome compartments and <10% in the polysome fraction in G0 cells (Fig. 5D). By contrast, in both MB and MT, all five mRNAs were enriched on polysomes, with barely detectable presence in the mRNP fraction, consistent with the high rates of protein synthesis typical of these states. Taken together with the repressed global rates of protein synthesis and increased accumulation of mRNP proteins in visible puncta, we conclude that mRNA sequestration in a non-translated compartment is a broad regulatory process that is enhanced in reversible G0, but not in post-mitotic MT.
Dcp1a and Fmrp reciprocally regulate their protein abundance and granule assembly
Since Fmrp and Dcp1a are known to regulate distinct aspects of mRNA function (translation vs. turnover) and were found in different complexes, we considered the possibility that these proteins might also cross-regulate. We used siRNA-mediated knockdown to perturb the levels of each protein and evaluated the effect of knockdown of one protein on abundance of the other protein using western blotting (Fig. 6A). Proliferating myoblasts were transfected with siRNA smart pools (comprising four independent siRNAs) designed to target either Dcp1a or Fmr1 mRNAs. A non-targeting siRNA pool was used as a control. Knockdown efficiency was confirmed to be 70-85% for Fmrp and 40-50% for Dcp1a by western blotting (Fig. 6A). Indeed, knockdown of Fmrp led to an induction of Dcp1a protein levels and vice-versa, knockdown of Dcp1a was accompanied by higher levels of Fmrp (Fig. 6A). This reciprocal regulation at the level of protein abundance was accompanied by increased detection of the respective protein in cytoplasmic puncta (Fig. 6B). Quantification of the fluorescent intensity of cytoplasmic staining (Fig. 6C) revealed that knockdown of Fmrp was readily observed as reduced immunofluorescence, and accompanied by an enhanced intensity of Dcp1a, and reciprocally, knockdown of Dcp1a, led to loss of Dcp1a detection and enhanced intensity of Fmrp. Taken together, these experiments reveal cross-regulation of Fmrp and Dcp1a not only at the level of protein abundance, but also at the level of protein assembly into puncta.
Fmrp and Dcp1a play opposing roles in control of MB proliferation
The results so far indicate differential mRNP granule protein abundance and assembly in distinct cellular states: the quiescent state is enriched in translational silencing/repressive complexes, whereas proliferating and differentiated cells are enriched in the classical nonsense-mediated mRNA decay complex. Further, reducing the abundance of translational repressor Fmrp by knockdown led to increased abundance and assembly of mRNA decay regulator Dcp1a and vice versa. To determine whether the differential enrichment of the decay and repressive complexes plays a role in the maintenance of a particular cellular state, we examined the phenotypes of the knockdown cells. Knockdown of Dcp1a in proliferating MB caused cells to proliferate more rapidly than control siRNA-treated cells, as evidenced by a significant increase in cell number and 5-Ethynyl-2´-deoxyuridine (EdU) incorporation (Fig. 7A). By contrast, knockdown of Fmrp led to reduced EdU incorporation (Fig.7A), mimicking the reduced proliferative capacity seen in MuSC from Fmr1-/- mice (Fig. 2D). Together, these results indicate that compromising the expression of key decapping and repressive/silencing mRNP proteins differentially affects proliferation in myoblasts. Dcp1a and Fmrp thus exert opposing effects on cell proliferation, possibly by targeting different transcripts for degradation, translational repression and/or sequestration. To identify the regulatory nodes at which Fmrp and Dcp1a might exert their effects, we evaluated the expression of cell cycle regulatory proteins by western blotting. In Dcp1a knockdown cells, Cyclin A2 protein expression increased, consistent with enhanced proliferation (Fig 7E). In Fmrp knockdown cells, by contrast, Cyclin A2 and Cyclin E protein levels were decreased, consistent with reduced proliferative capacity (Fig 7E).
Knockdown of Fmrp and Dcp1a on cell cycle regulators during quiescence and re-activation
As knockdown of Fmrp and Dcp1a had opposing effects on MB proliferation we evaluated the consequences of the knockdowns on expression of cell cycle regulators in MT, G0 and R3. In G0 conditions, we found nearly 5-fold increase in Cyclin A2 protein expression (Fig 7E) consistent with the increased proliferation observed in the MB condition. Fmrp knockdown, however did not affect protein expression of Cyclin A2 or Cyclin E (Fig 7E), consistent with the unchanged EdU incorporation in either knockdown in G0 (Fig 7C). Moreover, Fmrp knockdown G0 cells displayed reduced levels of mRNAs encoding Cyclin A2, B1, E and ki67, but negligible change in the levels of transcripts encoding either cell cycle inhibitors (Cdkn1a/p21, Cdkn1b/p27) or myogenic regulatory factors (MyoD, Pax7 and Myf5) (Fig. S2A). By contrast, Dcp1a knockdown increased levels of pro-proliferative transcripts including ki67 and Cyclin A2, B1, D1 and E, and reduced levels of the anti-proliferative Cdk inhibitor, Cdkn1a/p21, consistent with increased proliferation (Fig S2). Interestingly, both MyoD and Pax7 transcript levels were strongly reduced (Fig S2). Taken together, the reciprocal molecular phenotypes of Fmrp and Dcp1a knockdowns in cells in G0 were consistent with the observed reciprocal effect on proliferation. Notably, the results suggest that when either Fmrp or Dcp1a expression was compromised, cells entered an aberrant G0, since they arrested when cultured in suspension but expressed atypical levels of cell cycle regulators.
Knockdown of Dcp1a and Fmrp had a marked impact during reactivation of quiescent cells. At 3 hr of reactivation, Dcp1a knockdown cells already displayed increased EdU incorporation compared to control cells (Fig.7D). Supporting the premature entry into S phase, Dcp1a knockdown cells showed increased expression of both Cyclin A2 and Cyclin E proteins (Fig 7E). By contrast, Fmrp knockdown did not affect S phase re-entry and was accompanied by reduction of both Cyclin A2 and Cyclin E protein levels (Fig 7E). Together, these data are consistent with opposing effects of Fmrp and Dcp1 on proliferation, and suggest that Fmrp and Dcp1a modulate quiescence entry/exit potentially by targeting stability/utilization of cyclin transcripts.
Knockdown of either Fmrp or Dcp1a compromises myogenic differentiation
To assess the effects of depletion of Fmrp and Dcp1a on myogenesis, knockdown myoblasts were induced to differentiate for 2 days. As in proliferative conditions, Dcp1a knockdown in low serum conditions also led to sustained EdU incorporation with a corresponding increase in Cyclin A2 protein (Figs. 7B and 7E). By contrast, Fmrp knockdown lead to negligible EdU incorporation accompanied by drastic reduction in Cyclin A2 protein compared to control. Notably, the cross-regulation of Fmrp by Dcp1a knockdown (see previous section) was most pronounced in myotubes, and correlated with a pronounced suppression of Myogenin protein in the same sample, consistent with translation suppressive function of Fmrp. However, maintenance of the knockdown cells in differentiation conditions showed that loss of either Fmrp and Dcp1a negatively affected differentiation as evidenced by reduced Myogenin protein abundance, decreased frequency of Myogenin+ nuclei, reduced Myosin Heavy Chain protein expression and significantly reduced fusion index (Fig. 8 A-D). Taken together, these results indicate that despite their opposing effects on the cell cycle, optimal levels of both Dcp1a and Fmrp are required for myogenesis.
In summary, our data support a model (Fig. 9), where Fmrp and Dcp1a reciprocally regulate each other at the level of protein abundance and granule assembly, differentially regulate the expression of cell cycle and myogenic proteins and thereby play critical and opposing roles in the transitions between proliferation and reversible quiescence, ultimately leading to compromised differentiation.