Time-resolved quantitative proteomics implicates the core snRNP protein SmB together with SMN in neural trafficking

The five uridine-rich spliceosomal snRNPs, U1, U2, U4, U5 and U6 are essential components of the spliceosome required for removal by splicing of introns from pre-mRNA. With the exception of U6 snRNP, splicing U snRNPs have a complex pathway of biogenesis involving cytoplasmic and nuclear stages (reviewed by Fischer et al., 2011). Following their transcription by RNA polymerase II in the nucleus, the small nuclear RNAs (snRNAs) are exported into the cytoplasm where early events in snRNP maturation occur including hypermethylation of the 5′ cap of the snRNA and trimming of the 3′ end. Additionally, a heptameric ring of core Sm proteins is assembled onto the Sm-binding site of the snRNA. Assembly of the Sm core onto the snRNA requires the SMN (survival motor neurons) complex (reviewed by Battle et al., 2006). This process has been the subject of considerable attention because lowered levels of SMN result in the inherited neurodegenerative condition spinal muscular atrophy (SMA) (Lefebvre et al., 1995; Lefebvre et al., 1997). It is currently unclear whether the specific pathological defects in SMA, which affect predominantly spinal motor neurons, are caused by a decreased capacity for splicing snRNP maturation or by defects in other, as yet uncharacterized, roles for SMN in neural cells. Although numerous reports have been made of accumulations of Sm proteins and/or SMN in the cytoplasm, primarily in cells overexpressing SMN or mutant SMN (Liu and Gall, 2007; Pellizzoni et al., 1998; Sleeman et al., 2003), the cellular location and organization of the addition of Sm proteins to the snRNA during assembly of the snRNP is unknown. Furthermore, the identity and role of SMN-rich cytoplasmic accumulations has been a subject of debate. In Drosophila, U bodies, which contain snRNAs and SMN (Liu and Gall, 2007), have been proposed to play a role in snRNP assembly, whereas SMN-rich structures in mammalian neural cells, reported in some cases to lack Sm proteins, have been implicated in the localization of mRNAs (Fallini et al., 2010; Fallini et al., 2011; Fan and Simard, 2002; Hubers et al., 2011; Jablonka et al., 2001; Peter et al., 2011; Sharma et al., 2005; Todd et al., 2010a; Todd et al., 2010b; Zhang et al., 2006), with a proposed role in the local translation of mRNAs required for motor neuron function. Following the assembly of the Sm protein core, the partially assembled snRNP is imported into the nucleus. Once imported, snRNPs accumulate transiently in the Cajal body (CB) where the core snRNA is modified. Further maturation of snRNPs involves the addition of a number of additional proteins specific to each class of snRNP (for example, U1 snRNP contains the proteins U1A, U170K and U1C). Again, it is not known exactly where or how these stages are carried out, but it is assumed that these proteins are added after the snRNP has been re-imported into the nucleus. Mature snRNPs are believed to be stored in splicing speckles from which they are recruited to active spliceosomes when required (reviewed by Spector and Lamond, 2011).

In order to dissect the molecular mechanisms and subcellular organization of different stages of splicing snRNP biogenesis, in particular cytoplasmic stages of potential relevance to SMA, we used a time-resolved quantitative proteomic approach. Differential labeling of cells containing FP–tagged newly assembled snRNPs (transiently transfected with FP–SmB) versus FP-tagged mature snRNPs (lines stably expressing FP–SmB) was achieved by culturing cells in tissue culture media containing different isotopes of the essential amino acids arginine and lysine. This technique, known as stable isotope labeling by amino acids in culture (SILAC) (Ong et al., 2002), was then combined with affinity purification and mass spectrometry to facilitate comparison of the interactomes of snRNPs at these two distinct stages. Investigations informed by these data revealed a family of novel cytoplasmic vesicles, enriched in SmB and SMN. Unexpectedly, these vesicles appear to be heterogeneous in their constituents other than SMN and SmB, suggesting that they are involved in several aspects of cellular RNA metabolism. The SmB- and SMN-rich vesicles are upregulated during neural differentiation, disrupted by overexpression of SMN or reduced expression of SmB and are highly mobile in neurites, implicating them in the pathology of SMA.

Our previous analysis of the splicing snRNP maturation pathway using FP-tagged Sm proteins revealed that newly expressed Sm proteins are first detected diffusely in the cytoplasm, before accumulating transiently in Cajal bodies prior to forming the steady-state localization of mature splicing snRNPs in splicing speckles (Sleeman and Lamond, 1999). Because the Sm proteins are assembled into snRNPs before reaching speckles (Sleeman et al., 2001; Sleeman 2007) (supplementary material Fig. S1), this defines a pathway of maturation for splicing snRNPs. This pathway does not influence the localization of pre-existing snRNPs within the cell. HeLa cells constitutively expressing YFP–SmB can be fused with cells constitutively expressing CFP–SmB, and the import into the nucleus of newly assembled snRNPs monitored. This results in differently localized mature and newly assembled snRNPs (Fig. 1A). In cells stably expressing FP–SmB, the majority of FP-tagged snRNPs are mature, with only a few newly assembled snRNPs present. Importantly for our experimental approach, the use of fluorescently tagged Sm proteins enables the maturity of tagged snRNPs to be assessed using live cell microscopy immediately before processing the cells for proteomic analysis.

To gain information about complexes involved in different stages of snRNP biogenesis, we adopted an affinity purification and mass spectrometry (AP/MS) strategy involving differential stable isotope labeling, which allows a quantitative comparison of the relative distribution of interacting proteins between mature and newly assembled snRNPs and the ready elimination of non-specific contaminants. Mature snRNPs were isolated from a cell line constitutively expressing a YFP-tagged version of the core snRNP protein SmB (Sleeman et al., 2003), whereas newly assembled snRNPs were isolated from cells transiently transfected with YFP–SmB and harvested after 24 hours, before the snRNPs had accumulated in nuclear speckles. The subcellular distribution of the YFP–SmB protein was assessed by fluorescence microscopy immediately before harvesting the cells (supplementary material Fig. S1). The isolation was carried out using the high-affinity GFP-Trap® reagent conjugated to agarose beads (Chromotek) (Fig. 1B).

This time-resolved screen identified a total of 344 proteins, 152 of which were deemed likely contaminants because of their presence in the ‘bead proteome’ previously established using the same reagents in HeLa cells (Trinkle-Mulcahy et al., 2008). In addition to identifying the remaining six members of the core Sm complex, we identified six members of the Lsm (like Sm) protein family, five of which form the core of splicing snRNP U6 and one of which, Lsm11, is present with SmB in the core of the low-abundance U7 snRNP, which plays a role in histone RNA processing. Detection of Lsm11 confirms the sensitivity of the screen. Other key groups of proteins identified include non-Sm snRNP proteins and other splicing factors, SMN and several of the gemin proteins (members of the SMN complex), and members of the PRMT5 cytoplasmic methylation complex (Table 1).

Direct comparison of the degree of enrichment of detected proteins above background with mature (log H∶L) versus newly assembled (log M∶L) snRNPs clearly identifies interaction partners, both known and novel, and highlights key differences between the two pools (Fig. 2A). For example, although Sm proteins (yellow) and several other snRNP proteins (orange) are clearly enriched in both samples, non-snRNP splicing factors (green) are more enriched with mature snRNPs, whereas a larger number of transport factors (purple) co-purify with newly assembled snRNPs. The log H∶M ratios directly measure relative enrichment with mature (H) versus newly assembled (M) snRNPs, and information about relative abundance can be added by plotting these ratios versus the summed intensities of all peptides identified for that particular protein (normalized by molecular weight) (Fig. 2B). As expected, the most highly enriched protein was the YFP–SmB bait protein. Although slightly more was pulled down from transiently transfected cells (newly assembled snRNPs), it still fell below the chosen threshold (>1 log above or below the median log H∶M) and was therefore not significant. The most highly enriched and abundant mature snRNP interactors were the Sm and other snRNP proteins. Importantly, a large number of non-snRNP splicing factors are also present in the Sm protein interactome, demonstrating conclusively for the first time that FP-tagged Sm proteins are not only assembled into splicing snRNPs, but are also recruited into the spliceosome. Most of the proteins known to be involved in snRNP maturation including the SMN complex and the PRMT5 complex involved in methylation of Sm proteins are plotted near or slightly below the axis, indicating that they show a slight enrichment in the newly assembled sample. This is consistent with the suggestion that snRNP maturation is highly active in HeLa cells, as evidenced by their prominent Cajal bodies (Kaiser et al., 2008; Matera and Shpargel, 2006; Sleeman et al., 2001). Proteins enriched more highly in the newly assembled snRNP sample also include Crm1, which we have previously demonstrated to have a role in the targeting of new snRNPs to Cajal bodies (Sleeman, 2007) and a number of proteins potentially involved in the trafficking of new snRNPs in the cytoplasm. Because little is known about the physical organization and dynamics of the early stages of snRNP assembly in the cytoplasm, the identification of dynein heavy chain: a microtubule associated motor protein, and coatomer proteins beta and gamma, which are usually associated with endoplasmic reticulum trafficking vesicles was of particular interest. Early stages of snRNP assembly are of potential importance in understanding the molecular pathology of SMA. Dynein light intermediate chain (DYNCILI2) was also identified in the screen, falling just outside the threshold defining proteins preferentially interacting with newly assembled snRNPs.

In HeLa cells, cytoplasmic accumulations containing Sm proteins and SMN have been reported in cells overexpressing SMN (Sleeman et al., 2003), whereas U-bodies containing snRNPs, SMN and members of the SMN complex have been implicated in splicing snRNP maturation in Drosophila egg chambers (Cauchi et al., 2010; Liu and Gall, 2007). SMN-rich granules have also been detected in the cytoplasm of neural cells (Fallini et al., 2010; Jablonka et al., 2001; Peter et al., 2011; Sharma et al., 2005; Todd et al., 2010a; Todd et al., 2010b; Zhang et al., 2006). These have been widely reported to lack Sm proteins, leading to the suggestion that they are involved in alternative neural-specific roles for SMN. The potential importance of cytoplasmic stages of snRNP maturation in neural cells and its significance for SMA led us to investigate the cytoplasmic distribution of mCherry-tagged SmB in the human neuroblastoma cell line, SH-SY5Y. Time-lapse fluorescence microscopy of the SHY5mCherrySmB cell line, stably expressing mCherry–SmB in the SH-SY5Y human neuroblastoma background (Clelland et al., 2012), revealed small, rapidly moving, punctate structures within the cytoplasm (Fig. 3). Cytoplasmic structures were not observed in cells stably expressing the U1-snRNP-specific protein U170K or the non-snRNP splicing factor ASF/SF2 [also known as serine/arginine-rich splicing factor 1 (SFRS1)] fused to mCherry (Clelland et al., 2012) (our unpublished observations). Differentiation of SH-SY5Y cells using retinoic acid (RA) followed by brain-derived neurotrophic factor (BDNF) leads to the production of extensive neurites (Clelland et al., 2009; Encinas et al., 2000). Differentiated cells from line SHY5mCherrySmB showed prominent SmB-containing structures moving rapidly in a bi-directional manner within the neurites (Fig. 3 and supplementary material Movie 1). The increased prominence and mobility of cytoplasmic structures containing SmB in differentiated cells suggests that they play an important role in mature neural cells.

To gain detailed structural information about the cytoplasmic distribution of mCherry–SmB in the neurites of SH-SY5Y cells, we carried out correlative confocal and electron microscopy. Cells from line SHY5mCherry–SmB were imaged by confocal microscopy, with selected cells processed for subsequent transmission electron microscopy. Careful spatial alignment of the resulting confocal and electron microscope images revealed that the mCherry–SmB signal is found in vesicles of around 50 nm diameter with a slightly irregular outline (Fig. 4). In some cases, these vesicles were found adjacent to granular structures of similar size.

Our initial SILAC screen identified dynein heavy chain (DYNHCI), part of a molecular motor responsible for transport of cargo along microtubules, and two subunits of the coatomer complex, γCOP and βCOP, associated with COPI-trafficking vesicles, as proteins interacting preferentially with newly expressed SmB (Fig. 2 and Table 1). To confirm the interactions between SmB and these two protein complexes, immunoprecipitations were carried out. Whole-cell lysates from line SHY5mCherrySmB were incubated with RFP-Trap®-conjugated agarose beads (Chromotek) or unconjugated agarose beads and the isolated proteins used for western blot analysis with antibodies against DYNHCI and γCOP. Both of these endogenous proteins co-immunoprecipitated with mCherry–SmB (Fig. 5A). Because the coatomer complex is associated with COPI vesicles, normally linked to retrograde transport within the Golgi network and endoplasmic reticulum (reviewed by Szul and Sztul, 2011), we next addressed the localization of mCherry–SmB with respect to COPI vesicles within the cytoplasm. We first established an SH-SY5Y cell line, SHY1GFPεCOP, stably expressing GFP-tagged εCOP (a gift from Presley and Lippencott-Schwartz) (Presley et al., 2002). Immunoprecipitation using GFP-Trap_A followed by mass spectrometric analysis confirmed that the GFP–εCOP associated with other members of the COPI complex (unpublished observations). mCherrySmB was then transiently expressed in this cell line and the two fluorescent proteins imaged in living cells. The resulting time-lapse sequences indicate that the mCherry–SmB structures represent a subset of GFP–εCOP-tagged vesicles (Fig. 5B and supplementary material Movie 2). Further evidence that the mCherry–SmB structures are a type of membrane vesicle was obtained by incubating differentiated cells from line SHY5mCherry–SmB with the fluorescent lipid dye, BODIPY-488 (4 µg/ml, 16 hours). Time-lapse imaging showed colocalization between the mCherry–SmB vesicles and small lipid-rich structures stained with BODIPY-488 (Fig. 5C,D; supplementary material Movies 3 and 4).

The identification of dynein, a microtubule motor, in our time-resolved proteomic screen suggested that newly expressed Sm proteins interact with microtubules. To investigate the trafficking of the newly identified Sm protein vesicles, differentiated cells from line SHY5mCherrySmB were imaged using time-lapse microscopy before and after treatment with inhibitors of cytoskeletal dynamics. In control cells, the vesicles showed saltatory movement characteristic of microtubule trafficking with maximum velocities of up to 2.5 µm/second (Fig. 6A). Disruption of microtubules with nocodazole (30 µM for 3 hours) resulted in the formation of larger, irregular SmB vesicles (Fig. 6B) and dramatically reduced their mobility (Fig. 6A). Inhibition of actin microfilaments with cytochalasin D (2 µg/ml for 2.5 hours) did not alter the appearance or mobility of the SmB vesicles (Fig. 6A). To confirm that the cytochalasin D treatment had been effective, cells were stained with phalloidin (Fig. 6C).

A number of recent studies have identified mobile granules in neural cells that contain the survival motor neuron (SMN) protein. These granules showed partial association with other members of the SMN complex (Fallini et al., 2010; Todd et al., 2010a; Todd et al., 2010b; Zhang et al., 2006) and with the COPI complex (Peter et al., 2011) but were not demonstrated to contain Sm proteins. To investigate the contents of the SmB vesicles, differentiated cells from line SHY5mCherrySmB were used for immunodetection of SMN, gemin2 and Sm proteins. However, although the SmB vesicles were clearly visible in fixed cells, they were disrupted on permeabilization of the cells using detergents or solvents (supplementary material Fig. S2). This reinforces the conclusion that they are lipid-based structures. Furthermore, we also observed that different primary antibodies used to detect SMN appear to stain different cytoplasmic granules and show only partial colocalization when used together in the same cells (supplementary material Fig. S3).

To circumvent these technical difficulties, cells from line SHY5mCherrySmB were transfected with plasmid constructs to express GFP-tagged versions of representative proteins from the classes identified in the SILAC screen. SMN, gemin2 and SmD1 are proteins associated with snRNP maturation events; U5snRNP 100K represents non-Sm snRNP proteins; ASF/SF2 represents other splicing factors and CRM1 represents transport factors. We also included GFP–DCP1A (decapping mRNA 1) to determine whether the SmB-rich structures colocalized with P-bodies. Transfected cells were imaged either live, or fixed with paraformaldehyde but not permeabilized (Fig. 7; supplementary material Movies 5 and 6). GFP–SMN colocalized with mCherry–SmB in vesicles near the cell body and in neurites Fig. 7A,B). GFP–Gemin2 (Fig. 7C and supplementary material Movie 5) and GFP–SmD1 (Fig. 7D; supplementary material Movie 6) showed colocalization with a subset of the mCherry–SmB vesicles, but the colocalization was less complete than that seen between SmB and SMN. By contrast, GFP–U5 snRNP 100K (Fig. 7E) and GFP–ASF/SF2 (Fig. 7F) did not colocalize with the mCherry–SmB cytoplasmic vesicles (Fig. 7E). Both GFP–CRM1 (Fig. 7G) and GFP–DCP1A (Fig. 7H) showed cytoplasmic accumulations, but these did not colocalize with the SmB vesicles. This suggests that SmB and SMN form the core of the mobile vesicles, with Gemin2 and SmD1 present in a proportion of the vesicles, which are, therefore, implicated as the cytoplasmic sites of snRNP assembly. The roles for the rest of the SmB vesicles are not clear. The absence of U5snRNP 100K and ASF/SF2 is in agreement with current models of later stages of snRNP maturation and spliceosome assembly taking place within the nucleus. CRM1 does not appear to be important in the SmB vesicles, whereas the lack of colocalization with DCP1A, together with the absence of P-body-associated factors from the SILAC screen, indicates that these structures are not P-bodies so are not likely to be involved in mRNA degradation.

In the experiments described above, we observed that cells expressing high levels of GFP–SMN showed a significant disruption of the mCherry–SmB cytoplasmic vesicles. The vesicles in these cells were larger and less mobile than vesicles in control cells (Fig. 8). Furthermore, they also contained higher levels of mCherry–SmB protein than seen in untransfected cells [0.23±0.05% (mean ± s.e.m.) of total cellular mCherrySmB signal is found in vesicles in control cells; 3.79±1.9% in cells overexpressing GFP–SMN; Student's t-test, P<0.05, n = 11], suggesting that the SMN protein is instrumental in the formation and dynamics of the vesicles and in the regulation of their composition. Reduction of the amount of SMN present in cells from line SHY5mCherrySmB, using siRNA sequences that result in ∼70% depletion of SMN (Clelland et al., 2012) did not, however, result in the loss of SmB cytoplasmic vesicles (Fig. 8D). Furthermore, depletion of PRMT5 by ∼75% (Clelland et al., 2009) did not abolish the vesicles. Methylation of Sm proteins by PRMT5 is required for their addition to snRNAs during early snRNP maturation. To further investigate the requirements for SmB vesicle formation, we used siRNA to reduce the amount of SmB in an SH-SY5Y cell line constitutively expressing GFP–SMN (Clelland et al., 2009) (Fig. 8E). This cell line expresses relatively high levels of GFP–SMN protein, resulting in the accumulation of GFP–SMN in a small number of cytoplasmic structures in addition to nuclear CB/gems. Reduction of SMN results in the loss of both nuclear and cytoplasmic GFP–SMN. Reduction of SmB by ∼60% in these cells resulted in the complete loss of GFP–SMN accumulations from the cytoplasm together with a marked increase in nuclear bodies containing GFP–SMN. Owing to the striking nature of this result, we investigated whether it was a direct result of SmB depletion or an indirect consequence of reduced snRNP assembly by inhibiting snRNP assembly with leptomycin B in the same cell line. Treatment with LMB, in contrast to depletion of SmB, resulted in the appearance of a large number of GFP–SMN-rich structures in the cytoplasm, but no increase in nuclear accumulations of GFP–SMN. This unexpected result strongly suggests that SmB has a key role in the formation and maintenance of SMN-rich cytoplasmic structures.

Using the core snRNP protein SmB tagged with YFP, we have established a technique to identify proteins involved in different stages of the complex biogenesis pathway of splicing snRNPs. Our SILAC screen showed a clear enrichment of non-snRNP splicing factors and snRNP-specific proteins in the interactome of mature snRNPs when compared with that of newly expressed SmB. By contrast, an enrichment of the majority of snRNP processing factors, including SMN and a number of the gemin proteins, was seen in the newly expressed sample. Importantly, the nuclear export factor CRM1, which has a role in targeting newly imported snRNPs to the Cajal body, was enriched in the newly expressed interactome, confirming that differences observed in the interactomes reflect different stages of snRNP biogenesis and are not simply a consequence of higher amounts of cytoplasmic YFP–SmB in the newly expressed sample. Confirmation of the interaction of SmB with both γCOP and DYNHCI in cell lines stably expressing SmB demonstrates that interactions identified by transiently expressing SmB are biologically relevant and not a result of a transient overexpression. The identification of a number of non-snRNP splicing factors in the YFP–SmB interactomes demonstrates that FP-tagged Sm proteins are incorporated into active spliceosomes, whereas the grouping of different classes of protein seen by plotting log H∶L against log M∶L has the potential to reveal information about the relationships between proteins identified in this and similar screens. Of particular interest in this respect is the position of the minor U11-U12 snRNP associated protein U1SNRNPBP (SNRNP35), which is clearly separate from the major spliceosomal snRNP proteins in the top left section of the graph.

Among the proteins found to associate preferentially with newly expressed SmB was the molecular motor protein dynein heavy chain 1. Examination of the cytoplasmic distribution and dynamics of mCherry–SmB in the neuroblastoma cell line, SH-SY5Y, revealed highly mobile structures. Importantly, these structures were more prominent in cells differentiated using retinoic acid followed by BDNF. This suggests they might be of particular importance in neural cells. Analysis of the movement of these structures showed the saltatory motion associated with microtubule trafficking. The cessation of their movement in cells treated with nocodazole confirmed that they are transported along microtubules. The identification of γCOP, a member of the COPI complex usually associated with vesicle trafficking in the endoplasmic reticulum, in the interactome of newly expressed SmB was unexpected and led us to characterize the cytoplasmic SmB structures further. Expression of mCherry–SmB in an SH-SY5Y cell line constitutively expressing εCOP, another member of the coatomer complex, demonstrated that these two proteins colocalize in the mobile structures, whereas staining of cells expressing mCherry–SmB with the lipophilic dye BODIPY-488 demonstrates the SmB-containing structures to be rich in lipids. Together with their extreme sensitivity to extraction with detergents or alcohol when fixed with paraformaldehyde for immunofluorescence, this strongly suggests that the SmB-containing structures are a novel class of lipid-rich vesicle associated with the coatomer complex. Correlative confocal fluorescence and transmission electron microscopy reveals that the SmB-containing structures have morphology that is consistent with lipid-rich vesicles. Interestingly, some of the vesicles appear to be immediately adjacent to granular structures of a similar size. This is consistent with recent electron microscopic analysis of SMN-containing structures associated with the Golgi complex in mouse motor-neuron-like NSC34 cells (Ting et al., 2012).

A number of other cytoplasmic structures containing snRNP components and processing factors, including SMN, have been documented. In Drosophila, ‘U’ bodies, containing snRNAs, Sm proteins and the SMN protein, have been identified as a potential location for cytoplasmic stages of snRNP maturation (Liu and Gall, 2007). By contrast, several reports have identified structures in various neural cells containing the SMN protein and other members of the SMN complex in the apparent absence of any snRNP components (Fallini et al., 2010; Jablonka et al., 2001; Peter et al., 2011; Sharma et al., 2005; Todd et al., 2010a; Todd et al., 2010b; Zhang et al., 2006). One of these reports (Peter et al., 2011) also demonstrated an association of SMN with the coatomer complex and with mRNA encoding β-actin, suggesting that the SMN-containing structures are involved in trafficking mRNAs to growth cones and represent a nerve-cell-specific role for SMN beyond its role in splicing snRNP assembly. The SmB vesicles that we have detected contain both SMN and the coatomer complex. They also show some colocalization with Gemin2 and SmD1, although Gemin2 and SmD1 appear to localize to only a subset of the SmB- and SMN-rich vesicles. Previous work from several groups has demonstrated that Gemin2 and other members of the snRNP assembly machinery colocalize with a subset of SMN-positive structures in a number of neural cell types (reviewed by Fallini et al., 2012). Combined with our new data, this is suggestive of a family of related vesicles, with SMN and SmB as core components. The vesicles containing Gemin2, which is known to be responsible for the addition of Sm proteins to the snRNA core (Zhang et al., 2011), and other Sm proteins are likely to be involved in splicing snRNP assembly. Others might have different roles, for example in mRNA trafficking or signal recognition particle (SRP) assembly (Piazzon et al., 2013). We propose that SmB and SMN are core components of a family of cytoplasmic vesicles involved in ribonucleoprotein assembly, RNA trafficking and potentially other areas of RNA metabolism.

SmB, together with SmD3, has been implicated in the developmentally important localization of oskar mRNA in Drosophila (Gonsalvez et al., 2010) and Sm proteins are also involved in the maintenance of germ granule integrity in C. elegans (Barbee et al., 2002). A subset of Sm proteins including SmB have been demonstrated to interact with telomerase RNA and scaRNAs in mammalian cells (Fu and Collins, 2006; Fu and Collins, 2007), and in yeast, Sm proteins have recently been implicated in telomerase RNA processing (Tang et al., 2012). Taken together, these data demonstrate an expanding number of roles for SmB in the processing and regulation of RNAs beyond its role in pre-mRNA splicing. The bacterial Sm-like protein Hfq, the likely evolutionary origin of both Sm and LSm proteins, has been implicated as a general cofactor for chromosomally encoded regulatory RNAs (reviewed by Valentin-Hansen et al., 2004). We have now demonstrated that SmB is vital for the formation of SMN-rich cytoplasmic structures in neural cells, implicating it in the pathology of SMA whether or not that pathology is linked to snRNP assembly.

The importance of dynein-based microtubule trafficking in neurons is well established (reviewed by Hirokawa et al., 2010), with dynein implicated in the pathologies of a number of neurodegenerative conditions (reviewed by Banks and Fisher, 2008; Eschbach and Dupuis, 2011). We have demonstrated that SmB interacts with DYNCH1 and forms vesicles that are trafficked on microtubules. Clearly, further analysis of these vesicles is required to fully understand their function or functions within neurons. It is, however, clear that both SMN and SmB are key components of these vesicles. There is increasing evidence for a role for SMN in the transport and local translation of axonal mRNAs (reviewed by Fallini et al., 2012). The effect of overexpression of GFP–SMN in cell lines constitutively expressing mCherry–SmB is striking, with large, almost immobile structures forming. These abnormal structures also recruit a substantial amount of SmB. This is highly reminiscent of the effect of overexpression of SMN on the structure of the nucleus in SH-SY5Y cells (Clelland et al., 2009), where it results in the formation of a Cajal body complex containing SMN, coilin and Sm proteins in place of the separate CBs and gems usually seen in this cell line. This further implicates SMN in the formation of structures, both nuclear and cytoplasmic, involved in splicing snRNP metabolism. Surprisingly, however, many of the SmB- and SMN-rich vesicles did not colocalize with other members of the Sm protein family or of the SMN complex, making it unlikely that all of these vesicles are involved in snRNP maturation. Importantly, our data suggest that the relative levels of expression of SmB and SMN determine the formation and mobility of a family of cytoplasmic vesicles implicated both in snRNP maturation, analogous to Drosophila U bodies, and in the trafficking of mRNAs, analogous to SMN-containing structures identified in neural cells. Interestingly, previous work has shown that GFP-tagged human SMN protein lacking exon 7, the most common mutation seen in SMA, is unable to localize to cytoplasmic structures in chick neurons (Zhang et al., 2003). We have shown that SMN and SmB are key to the integrity of a family of cytoplasmic trafficking vesicles, providing a mechanistic link between the SMN and SmB proteins and defects in RNA localization and trafficking within axons proposed in models of SMA (Fallini et al., 2012).

The SMN complex has a well-established role in assembling the Sm protein ring onto snRNAs during splicing snRNP maturation, with alterations both in the cellular snRNP repertoire and in pre-mRNA splicing patterns documented in several models of SMN (Bäumer et al., 2009; Boulisfane et al., 2011; Campion et al., 2010; Gabanella et al., 2007; Zhang et al., 2008). Despite this, it is difficult to reconcile a defect in a process as ubiquitous as snRNP assembly with the motor-neuron-specific symptoms seen in SMA, making a role for SMN in mRNA trafficking in neural cells an attractive alternative hypothesis. The involvement of SmB in SMN trafficking vesicles in neurites highlights the closely interlinked nature of mRNA processing and trafficking pathways. Taken together with the recent documentation of mis-splicing of Stasimon, which is proposed to have a role in vesicular cargo transport, in a Drosophila model of SMA (Imlach et al., 2012; Lotti et al., 2012), our data suggest that the defects caused in cells by lowered levels of SMN converge on pathways of RNP transport that are of particular importance in cells, such as motor neurons, where local translation of mRNAs, at some distance from their site of transcription within the nucleus, is of key importance.

HeLa and SH-SY5Y cells were from ATCC. Cells were cultured in DMEM with 10% FBS at 37°C, 5% CO₂. Transfections were carried out using Effectene (Qiagen) according to the manufacturer's instructions. pYFP–SmB, pCFP–SmB, pGFP–ASF, stable HeLa cell lines expressing YFP– and CFP–SmB and stable SH-SY5Y cell lines expressing mCherry–SmB and GFP–SMN have been described previously (Clelland et al., 2009; Sleeman et al., 1998; Sleeman et al., 2003). pGFP–εCOP (Presley et al., 2002) was a gift from J. Presley and J. Lippencott-Schwartz. pGFP–DCP1A (Aizer et al., 2008) was a gift from Y. Shav-Tal. pGFP–U5snRNP 100K was a gift from A. I. Lamond. SH-SY5Y cell lines stably expressing GFP–εCOP were derived by selection with 200 µg/ml G418 (Roche) following transfection. Differentiation of SH-SY5Y cells using retinoic acid and BDNF was carried out as described previously (Clelland et al., 2009), as were heterokaryon assays (Sleeman et al., 2001). BODIPY-488 (Life Technologies) was added to culture medium at 2 µg/ml overnight and leptomycin B (LMB) at 100 ng/ml for 16 hours.

For live-cell microscopy, cells were grown on 40-mm-diameter glass coverslips and transfected as appropriate. Coverslips were transferred to an open chamber (Zeiss) within an environmental incubator (Solent Scientific, Segensworth, UK) on an Olympus DeltaVision RT microscope (Applied Precision), maintained at 37°C with 5% CO₂. Time-lapse images were taken at ∼1 second intervals using single z-sections or three z-sections at 0.5 µm spacing for deconvolution, 100× NA 1.35 objective. Exposure times were between 0.2 and 1.0 second, giving maximum pixel intensities of between 500 and 750. For fixed cell microscopy, z-sections of 0.2 µm spacing were taken through the entire sample with exposures giving maximum pixel intensities of between 3000 and 4500. Deconvolution was carried out and movies generated using Volocity software (Perkin Elmer). Vesicles were tracked manually in Volocity. mCherry–SmB signals in whole cells and cytoplasmic vesicles were quantified using Volocity by identifying objects by their intensity and size. Line profiles were generated using SoftWorx (Applied Precision). Montages and overlays of images were made using Adobe Photoshop CS4. Bar charts and statistical analyses were generated using Prism4 (GraphPad).

Cells on CELLocate coverslips (Eppendorf) were fixed in 4% paraformaldehyde in PBS and imaged on an LSM700 confocal microscope using the 40× Plan-Neofluar NA 1.3 and 100× alpha-Apochromat NA 1.46 objectives. Z stacks of fluorescent cells were collected and located on the CELLocate grid using DIC optics. Cells were re-fixed in 2.5% glutaraldehyde in 0.1M cacodylate buffer pH 7.2 and post-fixed in 1% OsO₄ with 1.5% sodium ferrocyanide in cacodylate buffer. Coverslips were dehydrated in alcohol and embedded in Durcupan resin (Sigma). Coverslips were removed from the resin stubs leaving the CELLocate grid impression on the resin, and the region containing the required cells was trimmed and serially sectioned, collecting every section through the cells. Sections were stained with lead citrate only. Fluorescent cells were identified from their relationship to other cells in the DIC image and TEM images were taken from all the sections collected until the whole of the cell had been imaged. Stacks of images were imported into Volocity (Perkin-Elmer) and cross-correlated with the TEM images.

Lysates were prepared as described previously (Sleeman et al., 2003). To isolate complexes containing mCherry–SmB, total cell lysates were cleared overnight with agarose beads then incubated for 2 hours with RFP–Trap_A (Chromotek) or agarose beads. Following five washes with RIPA buffer (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 1% v/v NP40, 0.5% w/v NaDOC and protease inhibitors) samples were separated bySDS-PAGE (Novex, Invitrogen) and transferred to nitrocellulose membranes (Hybond-C+, GE Healthcare) for immunoblotting. Antibodies used were rat mAb anti-RFP (Chromotek, 1∶500); goat polyclonal anti-γCOP (Santa Cruz, 1∶250) and polyclonal rabbit anti-DYNHCI (Santa Cruz, 1∶200). Secondary antibodies were conjugated to horseradish peroxidase (HRP) (Pierce, 1∶20,000). Detection was carried out with ECL Plus (GE Healthcare) and imaged using a Fujifilm LAS-3000 imaging system.

Reduction of protein expression was achieved by transfecting the appropriate cell lines with siRNAs (Thermo Scientific) using viromer green (Lipocalyx GmbH) according to the manufacturer's instructions. Cells were lyzed for assay by western blotting or fixed with paraformaldehyde for fluorescence microscopy 48 hours after transfection. Sequences used were SMN: CAGUGGAAAGUUGGGGACA, negative control targeting luciferase, UAAGGCUAUGAAGAGAUAC; positive control targeting cyclophilin B, GGAAAGACUGUUCCAAAAA; PRMT5, a mixture of CAACAGAGAUCCUAUGAUU, CCAAGUGACCGUAGUCUCA, UGGCACAACUUCCGGACUU, CGAAAUAGCUGACACACUA; SmB, a mixture of CCCACAAGGAAGAGGUACU, GCAUAUUGAUUACAGGAUG, CCGUAAGGCUGUACAUAGU, CAAUGACAGUAGAGGGACC.

Quantitative SILAC-based AP/MS experiments were carried out as previously described (Trinkle-Mulcahy et al., 2008). HeLa cells stably expressing free GFP were encoded by passaging in medium containing ‘light’ L-arginine and L-lysine (R0K0; Sigma-Aldrich). HeLa cells stably expressing YFP–SmB were encoded by passaging in medium containing ‘heavy’ [¹³C/¹⁵N]L-arginine and [¹³C/¹⁵N]L-lysine (R10K8; Cambridge Isotope Laboratories, Andover, MA). Parental cells were encoded by passaging for 7 days in medium containing ‘medium’ [¹³C ]L-arginine and [4,4,5,5-D4]L-lysine (R6K4) and then transiently transfected with YFP–SmB for 24 hours. The subcellular distribution of YFP–SmB in these samples was determined, by fluorescence microscopy, to be diffuse with some accumulation in CBs immediately before harvesting. Cells were harvested and extracts prepared by sonicating in RIPA buffer and centrifuging at 2800 g for 10 minutes at 4°C. Total protein concentrations were measured using a Biophotometer (Eppendorf). For affinity purification of free GFP and YFP-tagged SmB, equivalent total protein amounts of each extract were incubated with GFP-Trap_A™ (Chromotek, Martinsried, Germany) for 1 hour at 4°C. Extracts were then removed and the beads washed four times with ice-cold RIPA buffer. Proteins were eluted, separated by 1D SDS-PAGE and trypsin-digested for MS analysis, as described previously (Trinkle-Mulcahy et al., 2008).

An aliquot of the tryptic digest (prepared in 5% acetonitrile, 0.1% trifluoroacetic acid in water) was analyzed by LC-MS on an LTQ-Orbitrap mass spectrometer system (ThermoElectron, Rockford, IL) as previously described (Trinkle-Mulcahy et al., 2008). The Orbitrap was set to analyze the survey scans at 60,000 resolution and the top five ions in each duty cycle selected for MS/MS in the LTQ linear ion trap. Database searching (against the human IPI database v3.68; 87,061 entries) and quantification were performed using MaxQuant software v1.1.1.14 and the Andromeda search engine (Cox et al., 2009; Cox et al., 2011). The following criteria were used: peptide tolerance  =  10 p.p.m., trypsin as the enzyme (two missed cleavages allowed), carboxyamidomethylation of cysteine as a fixed modification. Variable modifications were oxidation of methionine and N-terminal acetylation. Medium SILAC labels were Arg6 and Lys4, heavy SILAC labels were Arg10 and Lys8. Minimum ratio count was 2 and quantification was based on razor and unique peptides. Peptide and protein FDR was 0.01. The full dataset (minus common environmental contaminants as per http://maxquant.org and proteins identified via the decoy database) is provided as supplementary material Table S1.

For the 24 hour sample, HeLa cells were transfected with pYFP–SmB and harvested 24 hours after transfection. For the stable cell line sample, cells from line YFP–SmB E1 (Sleeman et al., 2003) were harvested in parallel with the 24 hour sample. Immunoprecipitation was carried out as previously published (Sleeman et al., 2003) using anti-2,2,7-trimethylguanosine conjugated to agarose beads (Millipore NA02A), with agarose beads as control. 2 mg total cell lysate was used as input for each sample. 10 ng of pre-cleared lysate and unbound protein were separated by SDS-PAGE together with the agarose control beads and TMG antibody beads. Subsequent detection was carried out using Y12 anti-Sm (Abcam, dilution 1∶500) and rabbit polyclonal anti-GFP (Abcam, dilution 1∶500). Secondary antibodies were goat anti-mouse IRDye 680RD and goat anti-rabbit IRDye 800CW (LI-COR, dilution 1∶20,000). Blots were imaged using a LI-COR Odyssey CLx. Both sets of samples were analyzed on the same gel.

HeLa cells were transfected with pYFP–SmB and lyzed 24 hours after transfection using ice-cold 50 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 1% (v/v) Nonidet P-40, 1% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 2 mM EDTA plus complete protease inhibitor cocktail (Roche). Pre-cleared protein lysates were incubated with GFP-Trap_A (Chromotek) for 2 hours at 4°C or with agarose beads, washed four times with lysis buffer and the isolated fraction used for total RNA preparation using RNAeasy (Qiagen) according to the manufacturer's instructions. The samples were treated with DNase (Invitrogen) according to the manufacturer's instructions then used for cDNA synthesis using random hexamer primers and RevertAid reverse transcriptase (Thermo Scientific) according to the manufacturer's instructions. The cDNAs were used as templates for PCR using primers to U2 (ATCGCTTCTCGGCCTTTTGGCT and TCGATGCGTGGAGTGGACGGA) and U6 (CTCGCTTCGGCAGCACA and AACGCTTCACGAATTTGCGT) snRNAs (IDT). The resultant PCR products were electrophoresed on a 1% agarose gel containing SafeView (NBS).

The authors thank Lawrence Puente (Ottawa Hospital Research Institute Proteomics Core Facility) for technical support; J. Lippencott-Schwartz (NIH, Bethesda) and J. Presley (McGill University, Quebec) for pεCOP-GFP; Y. Shav-Tal (Bar-Ilan University, Israel) for pGFP-DCP1A; A. I. Lamond (University of Dundee, UK) for pU5snRNP 100K; M. Fuszard and C. Botting (BSRC MS and Proteomics Facility, University of St Andrews) for MS analysis of εCOP-GFP cell lines; G. Morris (RJAH Orthopaedic Hospital, Oswestry, UK) for MANSMA1 antibody.