The data presented here shed interesting light upon crossregulation between PTB, nPTB, and ROD1, as well as indicating a high degree of functional redundancy between PTB and nPTB. PTB was already known to autoregulate its own expression by repressing its own exon 11 (Wollerton et al., 2004
). The feedback loop may serve to prevent overexpression of PTB, to upregulate expression upon PTB export to the cytoplasm, and to reduce stochastic noise in PTB expression (Becskei and Serrano, 2000
). Although splice-sensitive microarray analysis of cells knocked down for UPF1 has been used to argue against the widespread harnessing of regulated alternative splicing to NMD (Pan et al., 2006
), more recent computational and array analyses have indicated that AS-NMD is widespread among both SR proteins and hnRNP proteins, where it is associated with ultraconserved genomic regions (Lareau et al., 2007; Ni et al., 2007
). Similar findings had previously been reported with plant SR protein genes (Kalyna et al., 2006
). These reports speculate that the AS-NMD events might generally be involved in autoregulatory feedback loops, as previously demonstrated for PTB (Wollerton et al., 2004
) and SC35 (Sureau et al., 2001
). Our findings demonstrate that such nonproductive splicing events can also be harnessed to crossregulate the expression of families of regulators. In HeLa cells, PTB switches nPTB off by promoting skipping of nPTB exon 10 (A–2C), while PTB and nPTB both promote the nonproductive splicing of ROD1 (F). Although nPTB exon 10 is associated with an ultraconserved genomic region, PTB exon 11 and ROD1 exon 2 also have extended conserved upstream regions that coincide with extended AG exclusion zones, indicative of branchpoint sequences remote from the exon (Gooding et al., 2006
) (Figure S3
). The reason for coincidence of ultraconserved regions with AS-NMD events is a mystery. However, for the PTB, nPTB, and ROD1 exons, the extended regions of upstream conservation can be explained in part by the distant branchpoint arrangement and the likely location of regulatory elements.
nPTB protein is usually expressed with neuronal specificity in the retina and brain (Kikuchi et al., 2000; Markovtsov et al., 2000; Polydorides et al., 2000; Rahman et al., 2002
), although nPTB transcripts can be detected in cells in which nPTB protein is not present (Boutz et al., 2007a
). The regulated skipping of nPTB, PTB, and ROD1 exons provides a mechanism for the cell to respond to excessive levels of any or all of the PTB paralog proteins. The relative sensitivities of the exon-skipping events appear to be tuned to facilitate PTB expression in preference to nPTB or ROD1. Thus the normal levels of PTB protein in HeLa cells result in 80% inclusion of PTB exon 11 (Wollerton et al., 2004
), 16% inclusion of nPTB exon 10, and 2% inclusion of ROD1 exon 2 (). Reduced PTB protein levels may therefore be a prerequisite for nPTB or ROD1 protein expression. In our experiments, this was achieved by PTB knockdown in HeLa and PAC-1 cells, but in vivo, this might result from developmentally programmed reduction in PTB transcription (Boutz et al., 2007b
). We also showed that nPTB protein is equally effective as a repressor of PTB exon 11 () and so potentially could modulate PTB levels. Whether this control is actually exerted in vivo is an open question. The crossregulation and partial functional redundancy of PTB and its paralogs will have implications for the analysis of transgenic knockout mice. On the one hand, many of the normal functions of the knocked-out factor may be masked by compensatory upregulation of the paralog(s). However, this may have the advantage of restricting phenotypes to the events that are critically dependent upon the individual paralog that has been knocked out.
Expression of nPTB appears to be intricately regulated at multiple levels, in addition to the PTB-induced exon 10 skipping. For example, upregulation of the miR133 family of miRNAs during striated muscle differentiation translationally silences nPTB expression (Boutz et al., 2007a
). The reduction in PTB+nPTB activity thereby allows various muscle-specific splicing events to occur, such as Tpm2
exon 7 and 10 selection (B). Another unusual feature of nPTB expression, shared by ROD1, is that it has an extremely suboptimal codon content, with a high proportion of codons ending in A or U, which are relatively infrequent in the human genome (Lander et al., 2001
). This limits its ability to be translated in vitro or overexpressed in transfected cells and may also modulate its expression in vivo. Indeed we were only able to overexpress nPTB () by using a codon-optimized construct with 211 silent mutations (F. Robinson and C.W.J.S., unpublished data).
The fact that very few changes were observed in 2D DiGE analysis of PTB knockdown samples unless nPTB was also knocked down () suggests a high extent of functional redundancy between PTB and nPTB in HeLa cells. The functional redundancy is not surprising in view of the >70% amino acid identity between PTB and its paralogs. Nevertheless, it contrasts with the example of the SRC
N1 exon, where nPTB is much less repressive (Markovtsov et al., 2000
). PTB and nPTB also have a differential effect upon internal ribosome entry segment (IRES) driven translation (Guest et al., 2004; Mitchell et al., 2003; Pilipenko et al., 2001
). It may be that N1 is representative of a subset of exons that are differentially sensitive to PTB nPTB, whereas many other splicing events are equally repressed by either paralog. Among the AS events that we analyzed (), the sensitivity to knockdown of PTB alone varied, with substantial effects upon PTB exon 11 () but negligible effects upon actinin splicing (). It may be that there is a continuum of AS events with differential sensitivities to repression by PTB or nPTB. Presumably, this differential sensitivity to PTB and nPTB is written into an “RNA code” and might be decipherable upon analysis of a sufficient number of splicing events that are coregulated (Blencowe, 2006; Matlin et al., 2005; Ule et al., 2006
). A similar situation probably pertains with events regulated by PTB and ROD1 in immune cells (Lynch, 2004
The role of PTB in regulating splicing of FGFR2 exon IIIb (Wagner and Garcia-Blanco, 2002
) and PTB exon 11 (Wollerton et al., 2004
) has been demonstrated previously by RNAi of PTB. However, before uncovering the upregulation of nPTB upon PTB knockdown, we were unable to obtain convincing RNAi evidence for the role of PTB in regulating Tpm1
() or α-actinin () splicing. Likewise the previous evidence for PTB regulation of mammalian TPM2
exon 7 was based upon PTB binding to an intronic silencer rather than functional evidence. Splicing of a chicken β-TM construct in HeLa cells was shown to respond partially to RNAi of PTB (Sauliere et al., 2006
). The lack of complete derepression was taken to indicate the involvement of other repressor factors. Our data suggest that the additional repressor is nPTB.
Our data suggest that quantitative proteomics might provide a useful complementary approach to methods for global analyses of alternative splicing (Blencowe, 2006; Matlin et al., 2005
). Here we identified eight splicing events that are targets of PTB in the PTBP2
, and GANAB
genes (). The TPM2
-regulated exons are usually muscle specific, whereas ANXA7
exon 6 is used in muscle and brain. This is consistent with previous indications that PTB commonly represses muscle and brain-specific exons. It will be interesting to see whether a strong enrichment of muscle and neuron-specific exons is maintained as we increase the number of validated PTB-regulated events. Likewise, although the number of validated events currently prevents a robust analysis for enrichment of particular regulatory sequence motifs, it is clear that the newly discovered PTB-regulated events share some common features with the previously characterized examples (Figure S3
). These include a high frequency of known PTB binding motifs and potential branchpoint sequences located beyond the conventional ~40 nt distance from the exon. We aim to compare analysis at the proteome level with the use of splice-sensitive microarrays (Blanchette et al., 2005; Johnson et al., 2003; Pan et al., 2006; Relogio et al., 2005; Sugnet et al., 2006; Ule et al., 2005
) in order to rigorously determine the extent to which proteomic analysis can contribute to understanding the global roles of splicing regulators. In view of the fact that PTB and other splicing regulators also influence gene expression at other levels, including translation (Mitchell et al., 2005; Sanford et al., 2004
), it seems likely that proteomic analysis will usefully complement splice-sensitive transcriptome analysis. Nonetheless, the data presented here already demonstrate that unanticipated insights can be gained from this approach.