|Home | About | Journals | Submit | Contact Us | Français|
The survival of motor neurons protein (SMN) is essential for the biogenesis of small nuclear RNA (snRNA)-ribonucleoproteins (snRNPs), the major components of the pre-mRNA splicing machinery. Though it is ubiquitously expressed, SMN deficiency causes the most common motor neuron degenerative disease, spinal muscular atrophy (SMA). We show here that SMN deficiency, similar to that which occurs in severe SMA, has unexpected cell type-specific effects on the repertoire of snRNAs and mRNAs. It alters the stoichiometry of snRNAs and causes widespread pre-mRNA splicing defects in numerous transcripts of diverse genes, preferentially those containing a large number of introns, in SMN-deficient mouse tissues. These findings reveal a key role for the SMN complex in RNA metabolism and in splicing regulation, and indicate that SMA is a general splicing disease that is not restricted to motor neurons.
The survival of motor neurons (SMN) protein is part of a large multi-protein complex (the SMN complex) that also contains at least seven additional proteins, termed Gemins 2–8 (Baccon et al., 2002; Carissimi et al., 2006; Charroux et al., 1999; Charroux et al., 2000; Gubitz et al., 2002; Liu et al., 1997; Pellizzoni et al., 2002a). The SMN complex is an assemblyosome that is essential for the biogenesis of small nuclear ribonucleoprotein particles (snRNPs) (Fischer et al., 1997; Meister et al., 2001; Pellizzoni et al., 2002b; Yong et al., 2004). snRNPs are major components of the spliceosome, the machinery that carries out pre-mRNA splicing (Nilsen, 2003; Will and Luhrmann, 2001). Each of the spliceosomal snRNPs (except for U6 and U6atac) is comprised of one snRNA (U1, U2, U4, U5, U11, U12, and U4atac) and a set of seven common proteins called the Sm proteins, as well as one or more proteins that are specific to each snRNA (Kambach et al., 1999; Nilsen, 2003; Will and Luhrmann, 2001). The assembly of heptameric Sm protein rings on snRNAs (Sm cores) is strictly dependent on the SMN complex (Meister et al., 2001; Pellizzoni et al., 2002b; Wan et al., 2005).
Genetic lesions in the SMN gene (SMN1) that result in functional SMN protein deficiency are the cause of spinal muscular atrophy (SMA) (Lefebvre et al., 1995), the most common motor neuron degenerative disease and the leading genetic cause of infant mortality (Cifuentes-Diaz et al., 2002; Iannaccone et al., 2004; Talbot and Davies, 2001). A second copy of the gene, SMN2, contains a single nucleotide mutation in exon 7 which results in frequent skipping of this exon and, consequently, low levels of functional SMN protein (Lorson et al., 1999). Although SMN is ubiquitously expressed in all tissues and is essential for viability of all cells in diverse eukaryotic organisms (Paushkin et al., 2002; Schrank et al., 1997; Wang and Dreyfuss, 2001), the question of how reduction in SMN levels results in an apparently cell-selective motor neuron phenotype remains unexplained (Monani, 2005). The SMN-dependent Sm core assembly reaction can be studied in cell extracts, and the amount of Sm core assembly on the snRNAs corresponds directly to the amount of SMN protein and can be modulated by the other components of the SMN complex (Battle et al., 2006b; Feng et al., 2005; Wan et al., 2005). Indeed, extracts of SMA patient cells have a lower snRNP assembly capacity, corresponding to the lower amount of SMN protein they contain (Wan et al., 2005). These and other findings indicate that the deficiency in the activity of the SMN complex in snRNP assembly is relevant to the pathogenesis of SMA (Winkler et al., 2005). Several other functions for the SMN complex in RNA metabolism have been suggested, including functions in transcription (Pellizzoni et al., 2001b), pre-mRNA splicing (Fischer et al., 1997; Pellizzoni et al., 1998), the biogenesis of small nucleolar RNPs (snoRNPs) (Charroux et al., 2000; Pellizzoni et al., 2001a; Whitehead et al., 2002), and in axonal RNA transport (Jablonka et al., 2004; Pagliardini et al., 2000; Rossoll et al., 2003; Zhang et al., 2003), but these remain uncharacterized.
The studies described here were initiated to determine what effect a deficiency in SMN, and hence in the capacity of cells to produce snRNPs, has on the steady-state levels of spliceosomal snRNPs. We quantitated snRNAs in cell lines with reductions in SMN levels and an SMA mouse model, an SMN-deficient organism. Unexpectedly, we found that SMN deficiency does not cause a uniform reduction in snRNPs but rather has cell type- and snRNA-specific effects, resulting in profoundly different repertoires of snRNPs compared to those that contain normal SMN levels. In light of the alterations in the normal repertoire of the major constituents of the splicing machinery, we used exon microarrays to examine in detail if any changes occurred in the splicing patterns in several tissues of SMA mice compared to their unaffected littermates. We show that splicing defects occur in numerous mRNAs in multiple tissues of SMA mice, and conclude that SMN plays a key general role in RNA processing. The observation that deficiency of the ubiquitously expressed SMN results in cell type-specific splicing defects provides a new perspective on SMA and a plausible explanation for the selectivity of SMA pathogenesis.
To study the effect of reduced levels of SMN on the amount of snRNAs, we established an inducible RNAi system for regulated knockdown of SMN. Short hairpin RNAs (shRNAs) specifically targeting either the human or mouse SMN mRNAs under the transcriptional control of a tetracycline-inducible H1 promoter were constructed in a lentiviral vector (Figure 1A). The two vectors were packaged into viral particles and used first to transduce a HeLa cell line (T-REx) constitutively expressing the tetracycline repressor (TR) protein. In this case, the non-targeting mouse SMN shRNA was used as the control. Binding of doxycycline to the TR activates transcription of the shRNA, resulting in RNAi, and the extent of SMN reduction was depended on the multiplicity of infection and on doxycycline concentration, and was determined by quantitative Western blots. Reduction of SMN level to ~20% (Figure S1A) of control had no effect on cell growth rate or morphology (data not shown). SMN levels could be further reduced to ~5% of normal, and the cells survived under these conditions for an additional 2–3 days before cell death was evident (data not shown). The capacity of assembly of Sm cores on snRNAs was measured using a quantitative assay in cell extracts (Wan et al., 2005), and it corresponded directly to the amount of SMN protein (Figure S1A).
To determine the effect of the reduction in snRNP assembly capacity on the steady-state level of snRNAs, we carried out real-time RT-PCR measurements for each of the spliceosomal snRNAs. Despite the strong reduction (almost 80%) in snRNP assembly capacity, there was no significant change in snRNA levels (Figure S1B). A further reduction in SMN levels, to ~15% or less of normal levels, was necessary to observe significant changes in snRNA levels. Using Western blots, we measured the levels of all the major components of the SMN complex, including SMN, the Gemins, and Unrip, as well as α-tubulin and Magoh which served as loading controls (Figure 1B). In addition to the strong reduction in SMN, a severe reduction was observed also for Gemin8 (87% before and 94% after induction), Gemin2 (58% before and 86% after induction) and Gemin3 (67% before and 75% after induction). In contrast, Gemin5, Gemin4 and Unrip remained unchanged. The snRNP assembly activity in these cells was also strongly reduced in close correlation with the amount of SMN protein remaining in these cells after RNAi induction (data not shown).
As shown in Figure 1C, in the absence of doxycycline and with only ~15% of SMN remaining, a moderate reduction in U4, U5, U12 and U4atac snRNAs was observed. However, a more severe decrease in SMN, to ~5% of normal, resulted in a pronounced reduction, albeit to different extents, of most of the snRNAs except U2 and U6. A very similar profile was observed measuring the snRNAs in mature snRNPs obtained by immunoprecipitation with Y12, a monoclonal antibody against Sm proteins (Figure S2). This is consistent with previous reports that the pool of snRNAs outside of snRNPs is relatively very small, as snRNAs that are not assembled with Sm cores are unstable (Sauterer et al., 1988; Zieve et al., 1988). These observations suggest that HeLa cells have a large excess capacity to produce snRNPs over what they require for viability. Importantly, once the level of SMN drops below a critical threshold (~15–20% of normal in these cells), the cellular contents of snRNAs is altered, but the change is not uniform as different snRNAs are affected to different extents.
To determine if a similar change in snRNAs occurs upon SMN deficiency in other cell types and species, we carried out similar measurements in a mouse motor neuron-derived cell line (MN1). MN1 cells were transduced with lentivirus expressing either the shRNA targeting mSMN or the non-targeting hSMN as control. Measurements of the amount of each of the snRNAs in MN1 cells where SMN levels were reduced to less than 10% of normal (Figure S3A), showed a small increase in U1 snRNA (16%) and a large decrease in U11 snRNA (40%) compared to control (Figure S3B). Importantly, the change in the profile of snRNAs in MN1 and in HeLa cells with a similar degree of SMN deficiency were markedly different, indicating that SMN deficiency causes cell type-specific alterations in the repertoire of snRNAs.
To assess whether the observations described above can be extended to the level of an organism and their potential relevance to SMA, we performed similar studies in tissues of SMN-deficient mice. For this, we used a moderate SMA mouse model corresponding to a type II SMA phenotype (Le et al., 2005), as the affected animals survive for up to 14 days after birth, which allows the collection of various tissues at different stages as the disease progresses. Here, we used both 6-day old mice, at which time the phenotype begins to become evident, and 11-day old mice representing an advanced stage of the disease. Tissues harvested from homozygous mSmn−/− SMA mouse littermates and their unaffected heterozygous littermates at 6 (P6) and 11 days (P11) after birth were used for preparation of whole tissue extracts for Western blot, measurement of snRNP assembly activity, and isolation of total RNA for quantitative RT-PCR.
As expected, the level of SMN was severely reduced to ~15% of control in both brain and kidney (Figure 2A). Consistent with observations in other SMN deficiency systems (Wang and Dreyfuss, 2001), Gemin2 also showed significant reduction in SMA mice (~55% in brain and ~48% in kidney), and Gemin8 was severely reduced to an extent similar to that of SMN in both brain and kidney of SMA mice, indicating a key role for SMN in maintaining the stability of the entire SMN complex. The extent of reduction in the level of SMN and other SMN complex components was very similar between P6 and P11 mice (data not shown). snRNP assembly activity was measured in total tissue extracts from brain and kidney in P6 and from brain, kidney, spinal cord and heart in P11 control and SMA mice. Considerable levels of snRNP assembly activity were detected in tissue extracts from P6 mice, whereas extracts of tissues of P11 mice had extremely low levels of activity, in accordance with the previously reported down-regulation of SMN complex activity during mouse development (Gabanella et al., 2005). In close parallel to the reduction in SMN levels, snRNP assembly activity in SMA mice was reduced by 75–80% in brain and kidney of P6 mice compared to the unaffected controls (Figure 2B).
Total RNA isolated from brain of P6 mice and from brain, spinal cord, kidney, heart and skeletal muscle of P11 mice was used for real-time RT-PCR measurements of snRNAs. Significant changes in the stoichiometry of snRNAs were already apparent in P6 SMA mice (Figure 3). Marked reductions in some of the minor spliceosomal snRNAs in P11 SMA mice were detected, including U11, U12 and U4atac, which were reduced by ~30% to ~60% in brain, spinal cord and heart compared to the corresponding unaffected controls (Figure 3). This reduction in the minor (U12-dependent) spliceosomal pathway snRNAs is tissue-specific, as the same snRNAs in kidney and skeletal muscle are relatively unaffected in the SMA mice (Figure 3). In addition, we note that the levels of some major spliceosomal snRNAs, particularly U6 snRNA in heart, U2 and U6 snRNAs in kidney, and U1, U2, U5 and U6 snRNAs in skeletal muscle of SMA mice, were significantly increased compared to those of control mice (Figure 3), while the levels of 5S and 5.8S rRNAs, which were used as input normalization controls, were not significantly changed (data not shown). Recently, after submission of this manuscript, Gabanella et al. have also reported a decrease in some of the minor snRNAs in SMA mouse tissues (Gabanella et al., 2007).
Similar results were obtained using limited primer extension, an alternative method for measuring the amounts of RNA, which also illustrated the very large difference in abundance of the major and minor spliceosomal snRNAs. Shown in Figure S4, the measurement of U1, U11, U12 and U4atac snRNAs from brains of P11 control and SMA mice indicated a 25–35% reduction in the levels of U11, U12 and U4atac snRNAs in SMA mice, while U1 snRNA remained unchanged, consistent with the real-time RT-PCR data described above.
Although the alterations in the normal repertoire of snRNAs in SMN-deficient cells and in SMA mouse tissues were remarkable, they did not appear to cause a catastrophic collapse of splicing because, except for motor neurons, there was no evidence of cell death. We therefore used the high resolution afforded by exon microarrays to ask if there are any changes in splicing in SMA compared to control mice. To do so, total RNA from three tissues, spinal cord, brain, and kidney of three SMA mice and three unaffected littermates, each prepared and processed separately, were analyzed using the Affymetrix mouse exon 1.0 ST microarray. Among the 266,200 probe sets supported by putative full-length mRNA, around 200,000 probe sets with significant signals above background and representing exons of ~20,000 out of ~30,000 genes in the mouse genome (Waterston et al., 2002) were included in the analysis. With a False Discovery Rate (FDR) set at less than 0.1, 259, 73, and 633 genes from spinal cord, brain, and kidney were identified having potential splicing pattern changes, respectively (Table S1). The top 30 affected genes from each tissue, ranked by absolute fold change of the most affected exon of each transcript, are listed in Table 1. A relatively small number of genes showed ≥ 1.5 fold change across all exons (42, 0, and 264 genes from spinal cord, brain, and kidney, respectively), indicating a much higher prevalence of splicing changes compared to transcription changes (Table S2).
To assess the validity of the exon array data we performed real-time RT-PCR, on 31 genes (indicated by Q under Validation & Characterization in Table 1), using exon junction specific primers and total RNA from the same control and SMA mice used in the exon array. Except for one, where the signal with the selected primers was too low for definitive measurement, all confirmed the changes observed in the exon array, indicating a validation rate of 97%.
In addition to measuring the amount of the exon in the tissue from which the signal was originally selected for the purpose of validation of the exon array data, we also determined the expression levels of the same exon in several other tissues, including brain, spinal cord, muscle, heart, and kidney, in both the control and SMA mice (Figure 4). These comprehensive measurements were performed on eight of the genes listed in Table 1 (indicated by Q*), including three types of transporter proteins (Abca8a, Kcng4 and Slc38a5), an extracellular matrix protein (Col5a1), Uspl1, Agxt2l1, Hif3a, and Chodl. Importantly, these measurements revealed a wide variation in the levels of these exons in different tissues, indicating the alterations in splicing are tissue-specific.
Figure 5 presents a summary of the exon array data for seven of these genes and indicates several additional validation experiments performed in the control and SMA mice. The height of the bars indicates the fold change between SMA and control for the probe sets representing each of the exons. The signal intensity for each of these is shown in Figure S5. The small error bars reflect the low variability among the three control and the three SMA mice used in the exon array experiments. Importantly, these measurements indicated that both transcript- and tissue-specific alterations in splicing occurred upon SMN deficiency.
To further validate and characterize the specific changes in the transcripts, we performed high-throughput (HT) RT-PCR and analyzed the products by high resolution capillary electrophoresis (Klinck et al., 2008). This large-scale analysis, consisting of more than 4,000 RT-PCR reactions in pairs of control and SMA, surveyed 22 of the genes in Table 1 (indicated as R). All the data of these experiments, including the sequences and locations of primers used for the HT RT-PCR, the electrophoregrams, the sizes and molarities of the products, and their positions on the gene maps, are accessible online (see Experimental Procedures). Because of space limitations, only a few representative examples are presented here. The locations of specific changes observed in some of these genes are indicated in Figure 5 (brackets labeled by R), and electrophoregrams of four pairs of RT-PCR reactions are shown in Figure 6. These data revealed numerous perturbations in the normal transcriptome. At least 17 of the examined genes (77%) contained changes predicted by the exon array data, but a considerable number of additional changes were documented. In some cases, the changes could be explained as alternative splicing where the ratio of two spliced isoforms is altered, however many others suggest aberrant splicing with isoforms present in SMA mice that are not found in the unaffected control littermates, and often do not correspond to a previously reported spliced isoform, or are outright defective as they produce a transcript that does not conform to normal splicing sites.
Examples of some of the many striking cases of specific alternative splicing changes and aberrant splicing events detected in the SMA mice are presented in Figure 6. The electrophoregrams illustrate several types of transcript changes in different tissues of the SMA mice, including changes in relative amounts of two likely alternatively spliced exons (Agxt2l1 & Lrdd), changes in the amount of several exons in the same transcript (Abca8a), and an example of aberrant splicing (Hif3a). Another two cases of aberrant splicing events were shown in Figures 5 and S6, where aberrant splicing products were detected by RT-PCR, then sequenced, and their amounts were measured by real-time RT-PCR in control and SMA mice. For Col5a1, the aberrant transcript, which was only detectable in the SMA mice, skipped the last 223 nt of exon 62 and first 21 nt of exon 63, creating premature termination codons (PTCs) in the new open reading frame (ORF) (Figures 5 and S6A). Similarly, an aberrant transcript from Uaca in which exon 3 is skipped and PTCs were generated in the ORF was detected in SMA mice (Figures 5 and S6B). The generation of PTCs explains why the relative amount of the aberrant transcripts is, in many cases, relatively low and difficult to detect. The reduction in the level of the normal transcript, for example, ~50% for Col5a1, is consistent with a scenario in which about half of the pre-mRNA is aberrantly spliced but only a fraction of that is detectable because it is eliminated by nonsense-mediated decay (NMD).
Functional annotations of the most affected genes in the three tissues using Database for Annotation, Visualization and Integrated Discovery (DAVID), indicated that a wide diversity of genes were affected and revealed an over-representation of genes expressing transporter and extracellular matrix proteins. The functional annotations of the 30 most highly affected genes for each of the tissues are listed in Table 1. Although the number of affected genes common to all three tissues is small (only 2 genes), transcripts of transporter and extracellular matrix genes are particularly affected in all three tissues. These classes of genes have particularly large numbers of exons. In fact, among the 259, 73, and 633 affected genes in SMA mice spinal cord, brain, and kidney, the median number of exons in the affected transcripts is 14, 10, and 12, respectively. In the mouse genome, the median number of exons per transcript is only 6 (Waterston et al., 2002). This suggests that pre-mRNAs with more introns (exons) to splice are more vulnerable to splicing changes in SMA mice.
Our findings here demonstrate that SMN deficiency causes profound changes in cellular RNA metabolism. It alters the repertoire of snRNAs and perturbs pre-mRNA splicing, leading to numerous splicing defects. The defects are widespread and cell-type specific, affecting mRNAs of functionally diverse genes. These surprising discoveries place the SMN complex, the essential machinery for the biogenesis of snRNPs, as a major factor in splicing regulation. Furthermore, they provide an example of how a deficiency in a ubiquitous protein can result in tissue-specific changes, and cast a new light on the pathophysiology of SMA.
A large degree of SMN decrease (> 80%) is required to cause a significant change in the levels of snRNAs, or cause cell death in cultured cells, suggesting that cells normally contain a large excess capacity of SMN complex to maintain their normal inventory of snRNAs. While a reduction in the amount of snRNAs upon decrease in SMN was not unexpected, its non-uniformity across different snRNAs and its cell type-specificity were unanticipated. Immunoprecipitation of snRNPs from total extract, which could be reliably performed from tissue cultured cells but not from mouse tissues, showed a similar stoichiometry for snRNAs in total RNA preparations and in snRNPs, consistent with previous observations that the pool of unassembled snRNAs is very small, and suggest that the cells have a similarly altered repertoire of snRNPs. The biogenesis of snRNPs is a highly regulated and intricate stepwise process, and the SMN complex is essential for the key step of the Sm core assembly on which all subsequent steps depend (Battle et al., 2006a; Massenet et al., 2002; Narayanan et al., 2002; Will and Luhrmann, 2001). The steady state level of each snRNA is determined by the gene copy number of that snRNA and by many other factors, including its rate of transcription, processing, transport, assembly, modifications and turnover, as well as availability of both the common and snRNA-specific proteins. Many of these factors could be regulated differently in different cell types and at various stages of development. It is also possible that the SMN complex, in addition to its direct function in Sm core assembly, also plays a role in other steps of snRNP biogenesis and snRNA metabolism. Little is known about the regulation of the biogenesis and turnover of snRNPs in different tissues, and it will be of interest to study these processes under both normal and SMN deficient conditions.
The splicing changes we observe are widespread, affecting hundreds of genes in different tissues of SMN-deficient mice. Different splicing changes occur in different tissues, but the same splicing changes are observed independently in the same tissue in each of the individual SMN-deficient mice. The very high confirmation rate of changes identified by the exon array and the validation of numerous specific cases by RT-PCR, attest to the high reliability of the data set we obtained. Using the stringent criteria we applied to selection of the affected transcripts, only a fraction of genes for which robust signals were obtained showed splicing perturbations (< 2% of genes). However, the number of affected transcripts is likely to be much higher, because the signals for any given exon represent the average for all transcripts in the tissue. To be detected, a significant change in the signal of an exon would have to occur in a large fraction of the transcripts that contain it. Thus, an aberrant splicing of an exon could be severe in some or even all of the transcripts that contain it in a specific cell population and may not be detectable in the total tissue sample. It is possible that a specific splicing defect that is detrimental to motor neurons occurs but was not detected in the total RNA sample from spinal cord. The lack of an effect on the level of the majority of the transcripts nevertheless indicates that general Pol II transcription and mRNA turnover have not been significantly affected, consistent with cDNA expression microarray experiments using different SMA mouse models with milder phenotypes (Balabanian et al., 2007; Olaso et al., 2006). In many cases, the aberrant splicing generates an mRNA that contains a premature termination codon. These transcripts are present only at very low levels probably because they were subject to nonsense-mediated decay.
Importantly, although some of the changes we observe in specific exons may correspond to shifts in alternative splicing patterns, most of them are not; they are abnormal RNA processing events that give rise to aberrant splicing products, which normally are not produced. We conclude this because many of the exons that are significantly changed in the SMA mice, particularly exon skipping events, are not involved in known alternative splicing events, such as Hif3a (Figure 6), Col5a1 and Uaca (Figure S6). In addition, it is not possible to distinguish based on the available data whether all copies of an affected transcript where more than one exon is aberrantly spliced have the same defect or whether the defects are distributed among many different transcripts.
The mechanistic basis for the widespread splicing defects caused by changes in the levels of SMN, hitherto known to function in the biogenesis of snRNPs, remains to be determined. Splicing is dependent both on snRNPs and RNA-binding proteins (hnRNPs and SR proteins). Knockdowns of mRNA processing factors, including spliceosomal protein components, have been shown to have differential effects on the splicing of a subset of pre-mRNAs in yeast (Clark et al., 2002; Pleiss et al., 2007) and on alternative splicing of a few pre-mRNAs in Drosophila (Park et al., 2004). Rather than depletion of an individual splicing factor, deficiency in SMN, which is not itself a component of the splicing machinery, causes changes in multiple snRNAs, leading to an altered snRNP repertoire. Despite this, there is no reason to believe that the composition of snRNPs within each spliceosome is altered in SMN-deficient cells. Importantly, while changes in the stoichiometry of snRNPs have not been previously associated with global regulation or defects in splicing, our observations merit consideration of such a link. Like the major spliceosomal snRNPs, many of the hnRNP and SR proteins that play a role in splicing are extremely abundant and present in vast excess over their high-affinity binding sites on pre-mRNAs (Dreyfuss et al., 2002; Dreyfuss et al., 1993), yet even moderate changes in their relative stoichiometry have significant effects on alternative splicing patterns (Black, 2003; Hou et al., 2002; Kashima and Manley, 2003; Licatalosi and Darnell, 2006; Martinez-Contreras et al., 2006; Paradis et al., 2007; Wang and Manley, 1995; Zhu et al., 2001). The complexity of the network of interactions among the snRNPs and between snRNPs and the enormous assortment of splicing factors precludes meaningful prediction of specific splicing outcomes based on a particular snRNA repertoire, at this time. However, we suggest that a change in the stoichiometry of snRNPs perturbs this network and affects the efficiency, rate and fidelity of spliceosome assembly on different introns. Because each cell type has a unique assortment of splicing factors and SMN deficiency causes its snRNP repertoire to change in a unique way, the resulting perturbations are distinct and give rise to cell-type specific effects on splicing. We note, however, that alternative mechanisms cannot be ruled out. For example, the SMN complex could have an snRNA-independent effect on splicing, as experiments with an SMN mutant in vitro have suggested (Pellizzoni et al., 1998), or in other aspects of RNA metabolism. Another possible snRNA-dependent scenario for the splicing perturbations is that some unassembled snRNAs accumulate in SMN-deficient cells, which might then sequester snRNP proteins or other proteins required for splicing. The important conclusion nevertheless remains, that the stoichiometry of the major constituents of the splicing machinery (the “snRNPertoire”) is altered and that transcripts reflecting a faulty splicing process accumulate in SMN-deficient cells.
Further studies will be required to uncover the specific mechanism at fault for particular transcripts out of the hundreds or thousands that are affected. The splicing changes we observe in SMN deficiency appear to reflect a general defect in splicing and do not resemble the splicing changes that result from modulation of specific splicing factors (Black, 2003; Caceres et al., 1994; Licatalosi and Darnell, 2006; Ule et al., 2005; Wang and Manley, 1995). However, examination of the affected transcripts reveals several striking features. Pre-mRNAs containing a large number of introns are much more likely to suffer aberrant splicing, suggesting that the likelihood of splicing defects increases with the number of splicing events the transcript undergoes. We note, in addition, that a considerable number of the aberrantly spliced transcripts have more than one splicing abnormality (see Table 1). This appears to exceed the probability of random events and suggests that the affected transcripts have a specific susceptibility to the suboptimal splicing machinery in SMN-deficient cells and that a splicing defect in one intron may be propagated to other introns in the same transcript. We suggest that having a large number of introns makes a transcript more vulnerable to splicing defects upon decreased SMN complex activity, presenting a challenge to evolutionarily more advanced organisms possessing many multi-intron genes. Classification of the aberrantly spliced transcripts by gene ontology indicates that proteins of diverse functions are affected, and reveals a very high preponderance of membrane transporters and extracellular matrix proteins (Table 1). Further analysis of the affected transcripts will be required to determine if there are sequence motifs common to these RNAs that cause their splicing defects. Nevertheless, it is apparent that many of the affected transcripts contain a large number of introns which likely explain their susceptibility.
A new view of SMA emerges from these findings. The selective degeneration of motor neurons is a characteristic feature of SMA, and several potential explanations for this pathogenesis have been suggested, including the possibility of a specific function for SMN in these cells (Monani, 2005). However, our findings that SMN deficiency causes cell type-specific perturbations in the composition of the major components of the spliceosome and a defective transcriptome throughout all cells and tissues, not only motor neurons, indicate that SMA is a general splicing disease. An unanticipated alternative explanation for the motor-neuron selective phenotype in SMA may therefore be that cell-specific factors that influence the relative abundance of individual snRNPs, rather than a cell-specific function of SMN itself, account for the cell-specific phenotype, but their effect is manifested only when SMN, the common factor in the general pathway, becomes limiting. The attrition of motor neurons may be caused by one or more aberrant transcripts or by the cumulative effect of many splicing defects. However, it now seems possible, that splicing defects in surrounding cells, such as those that make up the matrix on which motor neurons depend, may be the cause or a contributing factor to the demise of the motor neurons. As the severity of SMA is directly correlated with the degree of SMN deficiency, it is likely that even moderate reductions in SMN levels alter the snRNP repertoire and perturb splicing, albeit more subtly than those we have studied here. Together, the observations we describe here establish a key role for the SMN complex in gene regulation, particularly in the maintenance of splicing machinery.
All experiments with mice were carried out in accordance with the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee (IACUC Protocol # 800714). The mice were anesthetized with isoflurane and then sacrificed at the indicated age by cervical dislocation. Tissues were collected by manual dissection and either used immediately or frozen in liquid nitrogen and stored at −80°C until further processing.
Total RNA was extracted from culture cells and tissues using mirVana miRNA isolation kit (Ambion). 5S, 5.8S rRNA, and snRNA-specific primers were used to generate cDNA using Advantage RT-for-PCR kit (Clontech). 100 ng (or 375 ng for minor snRNAs) of total RNA was utilized as template in a 20 μl reaction. 1% of the cDNA was used for each real-time PCR reaction.
Real-time PCR experiments for snRNA quantification were carried out on an Applied Biosystems 7500 fast real-time PCR system using SYBR Green I dye chemistry. The same reverse primers were used in both RT and real-time PCR. Each snRNA from mouse tissue or cell samples was measured in triplicates. Absolute quantification was performed using ABI 7500 Fast System Detection Software. Details of all primers and probes are in Table S3.
Biotinylated sense-strand DNA targets were prepared using the Affymetrix GeneChip Whole Transcript (WT) Sense Target Labeling Assay according to the manufacturer’s directions. 1μg of total RNA from normal (n=3) and SMA (n=3) mouse spinal cord, brain and kidney was used as input for the first round of amplification. 10 μg of the resulting cRNA was used to proceed to the second round of amplification. A hybridization cocktail including 5.5 μg of fragmented, end-labeled ssDNA was applied to GeneChip Mouse Exon 1.0 ST arrays. Hybridization was performed using F450–001 fluidics wash and stain script on the Affymetrix GeneChip Fluidics Station 450. Arrays were scanned using the Affymetrix GCS 3000 7G and GeneChip Operating Software (GCOS) to produce .CEL intensity files.
Probeset intensities were calculated from the CEL files of the 18 samples using the RMA algorithm with default settings at both the gene level and the probeset level in Partek Genomic Suite 6.3, using the core probe sets as defined by Affymetrix. Probesets with maximum RMA intensity of 3 across all samples were excluded to eliminate probesets with low expression levels. Alternative splicing multi-way ANOVA was applied using Partek defaults with terms (probeset ID and group) reflecting not only experimental conditions but also terms that allow detection of alternative splicing events that differ between the sample groups. FDR was calculated and genes with FDR < 0.1 for differential expression or alternative splicing were considered. ANOVA was also applied at the exon level to determine differential expression of exons not grouped by transcript. Functional annotation was performed using DAVID 2007 (http://david.abcc.ncifcrf.gov/home.jsp). The data were analyzed comparing the proportions of genes with altered splicing against the total genes expressed above background on the microarray.
The exon array data can be accessed at NCBI Gene Expression Omnibus repository (accession number GSE10599).
HT RT-PCR, capillary electrophoresis, and data analysis were performed as previously described (Klinck et al., 2008). PCR primer locations, sequences and electrophoregrams of all reactions are available at http://palace.lgfus.ca/login, using dreyfuss_public and public for username and password, respectively, or upon request from the authors at http://www.med.upenn.edu/dreyfusslab.
We are grateful to Drs. Roscoe Klinck and Philippe Thibault for expert and generous help with the HT RT-PCR, Dr. Gordon Lutz for providing the SMA mice, Dr. John Tobias for technical support on exon array data analysis and Drs. Luigi Naldini and Giuliana Ferrari for the pRRL.SIN vector. We thank Isabela Oliva and Quynh Nguyen for technical assistance and the members of our laboratory for stimulating discussions and comments on this manuscript. This work was supported by the Association Française Contre les Myopathies (AFM). G.D. is an Investigator of the Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.