Proteins that bind to individual sequence elements within the DCS RNA.
The DCS acts in vivo as part of an intronic splicing enhancer needed for inclusion of the N1 exon (56
). The DCS also contains a CUCUCU element that mediates repression of N1 splicing in nonneural cells and that cross-links to PTB in a full-length spliceable src
). A short DCS RNA probe was previously shown to assemble into two RNA-protein complexes, “nonspecific” and “specific” (53
). The nonspecific complex can be competed away by any RNA and is hence non-sequence specific. It is also present in both HeLa and WERI extracts. This complex contains a prominent 50- to 55-kDa protein observable by shortwave UV cross-linking. The specific complex is enriched by removing the 55-kDa protein by ammonium sulfate fractionation of the extract. This specific complex contains hnRNP H, hnRNP F, and KSRP as well as other factors. Although present in HeLa extract, the specific complex is more prominent in WERI extract (16
). This complex is also sequence specific, since it is competed efficiently only with a DCS RNA. Although PTB binds to the CUCUCU element of the DCS in a full length pre-mRNA, we had not previously seen PTB binding to this element in the complexes formed on a short DCS RNA probe. To characterize proteins interacting specifically with the CUCUCU element of the DCS, we generated DCS RNAs where the U's in this element were replaced with the photoaffinity label 4-thiouridine (4-thioU). The 5′ phosphate of each cytosine in the CUCUCU element was further replaced with 32
P. This modified DCS RNA was incubated with HeLa or WERI nuclear extract and irradiated with 366-nm UV light. The extract was treated with RNase A, and the cross-linked proteins were resolved by SDS-PAGE.
As with shortwave UV cross-linking, 4-thioU cross-linking to the DCS RNA in the unfractionated extract gave a prominent band of 50 to 55 kDa as well as a band of approximately 100 kDa (Fig. A lanes 3 and 4). These proteins are thought to be the La and nucleolin proteins nonspecifically bound to the RNA (see below). 4-ThioU cross-linking also gave a 60-kDa band (marked nPTB) that was not previously observed with shortwave UV cross-linking. This 60-kDa protein was more prominent in WERI extract. We showed previously that the RNP complex that is specific for the DCS sequence is enriched in the 40% ammonium sulfate pellet fraction (ASP40) of the extract (53
). 4-ThioU cross-linking in this fraction again gave protein bands similar to those seen by shortwave UV (53
) (Fig. A lanes 1 and 2). These proteins were identified by immunoprecipitation (Fig. B). The top band, migrating at 80-kDa, was identified as KSRP, and the 70-kDa band was identified as FBP, which is highly related to KSRP (Fig. B, lanes 3 and 4) (18
). The other prominent band common to both extracts runs at 55 kDa and was identified as hnRNP H, as indicated by immunoprecipitation with Y12 anti-SM antibodies (lane 2). We previously found that hnRNP H is recognized by this antibody (16
). KSRP and hnRNP H were identified earlier as components of the specific DCS complex. The coimmunoprecipitation of KSRP and FBP with hnRNP H is presumably due to their presence in the same complex, since anti-SM serum has not been observed to react with KSRP and FBP. This coimmunoprecipitation varies with the reaction conditions and is not seen when precipitating with anti-KSRP or anti-FBP antibodies (lanes 3 and 4).
FIG. 1 Structure and composition of the DCS complex. (A) nPTB cross-links to the CUCUCU element of the DCS RNA. The DCS RNA sequence is shown at the top. Modified bases are shown as hollow letters. Site-specific labeling of proteins binding to the DCS RNA is (more ...)
Interestingly, the 60-kDa protein seen binding in the whole nuclear extract is also enriched in the ASP40 fraction and is immunoprecipitable with anti-PTB antibodies (Fig. B, lane 5). We reported earlier on a WERI-specific form of PTB that bound to the N1 3′ splice site (13
). This may be the same protein seen here binding to the DCS sequence in WERI extract. As explained below, we call this protein nPTB (for neurally enriched homolog of PTB). However, since WERI extract also contains the normal HeLa form of PTB, these bands could be mixtures of the two proteins. Nevertheless, it is clear that a PTB-related protein binds the CUCUCU element of the DCS.
An equivalent DCS-cross-linked PTB band is present in HeLa extract but is less pronounced. It is not clear why PTB cross-linking in HeLa extract is inefficient. It could be that PTB actually binds less well to the DCS than nPTB does. Alternatively, the precise interaction of nPTB and PTB with the RNA may differ, leading to inefficient cross-linking of PTB despite its binding to the RNA. Given that the DCS CUCUCU element is needed for splicing repression in HeLa extract, we thought that the latter possibility was more likely (see below).
We next extended this cross-linking approach to more precisely localize the binding sites for each protein along the DCS RNA. By incorporating the 4-thioU and 32P at different positions, we could assess which nucleotides were in close contact with each protein (Fig. C). When the photoaffinity label was placed just 5′ or just 3′ of the G5 element in the DCS, the predominant cross-linked bands were hnRNP H and F (lanes 1 and 2 and lanes 5 and 6, respectively). This was also seen when the RNA was 32P labeled in the G tract without the 4-thioU and the cross-linking was by shortwave UV (lanes 3 and 4). Interestingly, hnRNP F, although present in both extracts, cross-linked well to the RNA only in the WERI extract. When the 4-thioU was moved downstream to the CU tract, cross-linking to the F protein was lost but cross-linking to hnRNP H, nPTB, KSRP, and FBP was seen (lanes 7 to 12). This contact with the H protein was primarily with the 5′-most U in the CU tract; when the label was only in the central U, the intensity of the H band was decreased (lanes 9 and 10). As before, the PTB band was enriched in the WERI extract over the HeLa extract. When the 4-thioU was placed in the element UGCAUG, just downstream of the CU tract, cross-linking was weaker but was predominantly to KSRP, FBP, and an unidentified protein of about 50 kDa. With the label in this position, cross-linking to PTB or nPTB and to hnRNP H and F was lost. The proteins that bind to this position are particularly interesting because this UGCAUG element is crucial to the activity of the DCS as an enhancer. Overall, these results give a clearly ordered arrangement of the proteins along the DCS sequence, going from 5′ to 3′, hnRNP H and F then nPTB, then KSRP and FBP.
The importance of each element in the binding of the individual proteins was assessed by repeating the cross-linking on RNAs where the small sequence elements were mutated (Fig. D and E). These experiments used labels in two positions along the RNA sequence, just 5′ of the G tract (modification U8 [Fig. D and E, bottom]) or within the CU tract (modification U18 [Fig. D and E, top]). Cross-linking at U8 adjacent to the G tract gave predominantly the hnRNP H and F proteins as before, unless the G tract was mutated, which caused a loss of hnRNP H and F cross-linking (Fig. E, bottom). This was also seen with the label at U18 and confirms the need for the G tract in hnRNP H and F protein binding (Fig. E, lanes 2 and 3). The effects of the other mutations were most easily observed with the label at U18 (Fig. E, top). Mutation of the CU tract reduced nPTB cross-linking and increased cross-linking to the hnRNP H and F proteins (lanes 4 and 5). It is not clear whether the increased signal for hnRNP H and F is due to increased binding of these proteins to the mutant RNA, perhaps due to changes in secondary structure, or to increased interaction with the label at this position because of the loss of PTB binding (see below). Mutation of the GCA sequence within the UGCAUG element does not affect nPTB binding but reduces KSRP and FBP binding (compare lanes 6 and 7 with lanes 8 and 9). In sum, the cross-linking to the mutant RNAs confirms the position of contact for each protein along the RNA.
The cross-linking experiments delineated the relative position of each protein along the RNA. However, these experiments did not give any information on the stoichiometry of the proteins binding to the DCS and did not distinguish how many complexes are formed by the various cross-linked proteins. For example, when we see both KSRP and nPTB cross-linking, we do not know whether these are binding simultaneously in one complex or independently in separate complexes. Previous gel shift and antibody supershift experiments indicated that hnRNP H and KSRP are in the same complex (16
). As a first step in testing whether PTB is also in this complex with hnRNP H and KSRP, we performed coimmunoprecipitation experiments (data not shown). Using an anti-PTB monoclonal antibody, we immunoprecipitated PTB from HeLa extract in the presence or absence of the DCS RNA. KSRP was strongly coprecipitated with PTB only in the presence of the DCS RNA, indicating that these proteins do indeed interact with the DCS RNA in the same complex (data not shown). Since the majority of the KSRP is present in the DCS complex under these conditions, these results place PTB in a complex with both hnRNP H and KSRP rather than in a complex that is independent of the previously characterized DCS complex. Moreover, since these experiments were done with HeLa extract containing PTB rather than nPTB, they demonstrate that PTB is indeed binding to the DCS in spite of its weak cross-linking. The common binding of these proteins into a single complex is addressed further below.
Characterization of DCS RNA binding proteins by RNA affinity chromatography.
To further examine the proteins binding to the DCS RNA in the absence of cross-linking, we performed RNA affinity chromatography. After periodate treatment, the DCS RNA was chemically attached through its 3′ end to adipoyl hydrazido-Sepharose. We find that this resin shows lower nonspecific protein binding than do RNA affinity resins based on the avidin-biotin interaction. To further minimize nonspecific protein binding, nuclear extract was supplemented with 0.1 mg of heparin per ml before being loaded on the column. Based on gel shift data, this amount of heparin disrupts nonspecific RNA-protein complexes while leaving the specific DCS complex intact.
The RNA resin was incubated with either HeLa or WERI nuclear extract, packed into a column, and washed extensively in 100 mM salt buffer. DCS RNA binding proteins were then eluted in a step gradient of KCl. In a control experiment, a column carrying RNA unrelated to the DCS sequence was used. The protein composition of each fraction was analyzed after gel separation by Coomassie blue staining and Western blot analysis (Fig. ). The elution profiles for the HeLa and WERI nuclear extracts on the DCS RNA column were very similar. The most prominent WERI protein, eluting between 0.3 and 0.5 M salt, was nPTB, as confirmed by Western blot using an anti-PTB serum that reacts with both proteins. In HeLa extract, PTB eluted in the same fractions but ran as a doublet on the gel. Note that due to differences in the running of the two gels, they cannot be perfectly aligned. The PTB in the HeLa panel is slightly below that in the WERI panel. The binding of the nPTB and PTB proteins was very efficient; these proteins were almost completely depleted from the WERI flowthrough fraction (Fig. B).
FIG. 2 Purification of proteins bound to the DCS RNA or an unrelated RNA in the WERI-1 or HeLa extracts using RNA affinity chromatography. (A) Coomassie blue-stained gels displaying fractions from the DCS RNA affinity column (lanes 1 to 6 and lanes 8 to 13) (more ...)
Western blot analysis of the column fractions identified other proteins binding to the DCS RNA. Known components of the DCS gel shift complex, KSRP, FBP, and hnRNP H, all bound the DCS RNA column and eluted in 0.3 to 0.5 M salt (Fig. B). By Coomassie staining, hnRNP H appeared to bind to the column less efficiently than nPTB or KSRP did. The control RNA binding proteins hnRNP A1 and U2AF65 both failed to bind to the column and eluted in the flowthrough fraction. None of these proteins bound well to the nonspecific RNA column, although some KSRP and hnRNP A1 were present in the 0.3 M fraction.
Another protein eluting from the DCS RNA column was identified as SC35. This 35-kDa protein was reactive with MAb104 anti-SR and anti-SC35 monoclonal antibodies but not with ASF/SF2 antibodies. This was the only SR protein seen binding to the DCS column. However, it also bound to the unrelated RNA column. Given the known effects of SR proteins on splicing, we are investigating whether SC35 plays a role as a DCS binding protein, even though its binding specificity is somewhat in doubt.
Several other proteins also bound to the DCS RNA column. These were identified by MS analysis of their tryptic peptides as nucleolin, La antigen, hnRNP C, and actin. The nucleolin, SC35, and La antigen proteins bound tightly to the DCS RNA, eluting mostly at 0.5 to 1 M salt. Both La and nucleolin recognize short stem-loop RNA structures containing pyrimidine-rich sequences on their termini, and the DCS RNA can potentially fold into such a structure (22
). Thus, the binding of these proteins to the column is probably due to similarities between the DCS RNA and natural RNA substrates for these proteins. La is presumably the ~50-kDa protein observed in the nonspecific gel shift complex with the DCS RNA (53
). hnRNP C eluted from the column at 0.3 M salt. hnRNP C has affinity for pyrimidine-rich RNAs, and such elements are present in the DCS RNA (69
). Some actin protein was also observed to stick very tightly to the column, eluting only in 6 M urea. This may result from polymerized actin filaments, trapped at the top of the column, eluting from the column when denatured into monomers.
Most of the proteins binding to the DCS RNA, including nPTB and PTB, were not retained on the unrelated RNA column, which showed a different spectrum of bands. The exceptions to this were SC35 and nucleolin, which bound tightly to both RNA columns. The nucleolin elution profile varied from batch to batch of the extract. Note that there is a prominent nucleolin breakdown product migrating just below KSRP in some of the WERI and HeLa extract fractions.
The RNA affinity chromatography was repeated with RNAs carrying mutations in each of the various subelements of the DCS RNA (data not shown). This confirmed that PTB and nPTB required the CU tract for stable binding, hnRNP H binding needed the G tract and the UGCAUG element, and KSRP binding needed the CU tract and the UGCAUG element.
Finally, the affinity chromatography results indicate that HeLa PTB binds to the DCS RNA even though it does not cross-link efficiently. Moreover, as seen previously, the eluted HeLa PTB migrates as a doublet with slightly different mobility from the WERI nPTB-PTB mixture.
Purification and cloning of nPTB.
The different cross-linking patterns of PTB in the two extracts indicated a difference in the protein from these two sources. To allow a detailed comparison of the PTBs in the two extracts, we purified the two proteins using a modification of the published PTB purification protocol (61
). The HeLa and WERI PTBs copurified over the first three columns: DEAE, heparin, and poly(U)-Sepharose. When the WERI extract fractions from the poly(U) column were loaded onto a HiTrap Cibacron Blue Sepharose column and eluted with a shallow gradient of KCl, the two forms of PTB were resolved. Under these conditions, nPTB came off the column earlier than the PTB isoforms seen in HeLa extract (Fig. A). As a result, a nearly homogeneous sample of the nPTB protein, free of PTB, was obtained (Fig. B).
FIG. 3 Purification of human nPTB. (A) Purification scheme. Coomassie blue-stained gels of PTB- and nPTB-containing fractions from the HiTrap Blue column are shown below the scheme. (B) Coomassie blue staining of an SDS–10% polyacrylamide gel (more ...)
nPTB was subjected to tryptic digestion followed by microsequencing of seven peptides. Two of the analyzed peptides matched the PTB sequence exactly, whereas the others contained one or more changes in amino acid sequence from the corresponding PTB peptides (Fig. C). The unique nPTB peptides were distributed along the PTB sequence, indicating that nPTB is a new protein rather than a new PTB splice variant.
Searching GenBank databases for the nPTB peptide sequences resulted in only one positive match. There were no matches to ESTs. However, the longest identified peptide, NNQFQALLQYGDPVNAQQAK, matched the sequence from the end of the human genomic BAC clone AQ006967. This clone contained 56 nt of an apparent nPTB exon encoding most of the peptide. Reverse transcription-PCR (RT-PCR) of WERI cell total RNA, using the 56-nt exon oligonucleotide as a sense-oriented primer and an 18-nt degenerate antisense primer corresponding to the FFQDHK peptide, produced a partial nPTB cDNA, 821 nt in length. Both the 56-nt exon sequence and the RT-PCR product were used to screen a WERI cell λZap cDNA library. A clone containing the apparent full-length cDNA of nPTB was identified and sequenced (Fig. ). This 3,060-nt cDNA contains a 1,593-nt open reading frame encoding a 531-amino-acid 57,454-Da protein. All seven nPTB peptides are present in the amino acid sequence of the encoded protein (Fig. ).
Human nPTB cDNA sequence. The NLS sequences are shaded. The four RRM domains are boxed. Sequenced peptides are in bold. The asterisk indicates the stop codon.
As expected, nPTB is very similar to PTB (~74% identical, depending on how the gaps are weighted), containing four unusual RNA recognition motif (RRM) domains and a putative bipartite nuclear localization signal (NLS) near the N terminus. This NLS domain consists of two short sequences, GVKRG and KKFK, separated by a 29-amino-acid gap, and matches the consensus nucleoplasmin NLS better than the corresponding PTB domain does (63
) (Fig. ). The four 80-amino-acid RRM domains of nPTB can be aligned with each other and are similar to the PTB RRM domains. A typical RRM domain contains two consensus elements, RNP1 and RNP2, that comprise β-strands 3 and 1 of the domain (32
). Like PTB, the RNP elements of nPTB diverge from those in the consensus RRM (Fig. A). Alignment of the four nPTB RRM domains also shows common residues within α-helix A, within β-strand 2 and following β-strand 4. In the known structures of RRMs complexed with RNA, the β-sheets both form the hydrophobic core of the domain and make extensive contact with the RNA on the interaction surface (1
). The conservation of the nPTB sequence in these regions may reflect folding constraints or could result from similarities in RNA recognition.
FIG. 5 Comparative analysis of the nPTB amino acid sequence. (A) Structural organization of the nPTB RRMs. RNP1 and RNP2 consensus sequences for the typical RRM and for the nPTB RRMs are shown at the top. The domain secondary structure is schematically shown (more ...)
The variation in sequence between human nPTB and PTB far exceeds that between PTBs from different mammals. Human nPTB is presumably the homolog of the mouse brain PTB protein recently isolated by the Darnell laboratory, since they are nearly identical (accession no. AF095718
). An EST database search with the nPTB sequence also identified the partial cDNA sequence of an apparent zebrafish nPTB (accession no. AA566427
). Phylogenetic comparison of the known PTB sequences places nPTB on a distinct branch, roughly equally related to all the mammalian PTBs (Fig. C). nPTB is also 67% identical to ROD1, another mammalian PTB homolog identified as a regulator of differentiation in Schizosaccharomyces pombe
). Searches of the complete Caenorhabditis elegans
and Drosophila melanogaster
genomes each uncovered a single gene with similarity to PTB. These proteins were divergent from both mammalian proteins, but at several positions they were more similar to PTB than nPTB. These results are consistent with the idea that nPTB is relatively recently evolved and is perhaps specific to the vertebrates.
Tissue distribution of nPTB.
The biochemical assays for nPTB indicated its presence in WERI but not HeLa extract. We next examined the expression of nPTB and PTB on a multiple-tissue Northern blot (Fig. A). The full-length nPTB cDNA and a 1-kb fragment of the PTB coding region were used as probes. A human β-actin cDNA was used as a control. nPTB mRNA appeared as a single band approximately 3.4 kb in length, while PTB gave rise to a doublet of 3.6- and 4.7-kb bands. The upper band may be a cross-hybridizing mRNA or possibly an immature form of PTB mRNA. PTB mRNA was present in a wide range of tissues, with the lowest levels observed in the brain. In contrast, nPTB mRNA was most abundantly expressed in the brain, although there was detectable expression in other tissues, notably the heart and skeletal muscle.
FIG. 6 Tissue and cell line distribution of nPTB. (A) Northern blot analysis of poly(A)+ RNA from the indicated human tissues using an nPTB oligonucleotide probe (top) or the PTB cDNA probe (middle) probe. Loading was verified by hybridization to a β-actin (more ...)
The level of nPTB mRNA was also examined in tissue culture cell lines (Fig. B). Abundant expression was observed in the WERI-1 neural cell line, as expected. Only minimal mRNA was observed in the nonneural cell lines, HeLa and HEK293. Significantly, only low levels of nPTB mRNA were seen in the human neuroblastoma cell line, LA-N-5. This was confirmed by RT-PCR experiments, where LA-N-5 cells showed higher expression of nPTB mRNA than did HeLa or HEK cells but significantly lower expression than did WERI-1 cells (data not shown). Since LA-N-5 cells show strong inclusion of the src N1 exon, the level of nPTB is not the only factor that determines N1 splicing. These cell lines all expressed abundant PTB mRNA, confirming that WERI cells express both PTB and nPTB (Fig. B). This was further confirmed by Western blot analysis (Fig. C). nPTB could be observed in WERI cells as a reactive band running between the two major isoforms of PTB.
nPTB is a weaker repressor of N1 splicing than is PTB.
To assess the functional differences between nPTB and PTB, we analyzed the behavior of each protein in the in vitro splicing assay by using two src
minigene transcripts. BS7 contains both the upstream and downstream regulatory sequences, each containing a pair of CUCUCU elements (Fig. A). BS27, obtained by precise deletion of the intron upstream of the N1 exon, contains only the two downstream CUCUCU elements. We previously showed that BS27 is spliced in the HeLa nuclear extract, where BS7 is repressed. Both transcripts splice equally well in WERI extract (12
FIG. 7 Effect of PTB and nPTB on N1 exon splicing in vitro. (A) Maps of the src splicing substrates. Black boxes indicate splicing-regulatory elements. (B) Splicing of adenovirus major late (lanes 1 to 8) and β-globin (lanes 9 to 16) transcripts in WERI-1 (more ...)
Purified HeLa PTB efficiently repressed BS7 splicing when titrated into the WERI nuclear extract (Fig. C, left, lanes 20 to 22). The levels of both the splicing intermediates and products were decreased 3- to 10-fold by the added PTB. Similarly, the 0.3 M fraction of the HeLa extract from the DCS RNA column (Fig. ) also repressed splicing (Fig. C, left, lanes 14 to 16). The splicing repression required the same amount of PTB whether it was introduced as 0.3 M fraction or as a purified protein. At the highest point of the titration, the total PTB concentration was about twofold higher than that of the endogenous protein in HeLa extract. Bacterially expressed recombinant PTB also repressed splicing but was about 10 times less active than the purified protein (data not shown). This may be due to aberrant folding of the recombinant protein, to a lack of posttranscriptional modification, or to other effects.
PTB was a much stronger repressor of splicing than was nPTB. At the highest concentration of nPTB, the levels of the second-step products of splicing were somewhat reduced but the levels of the intermediate products of the first step were unaffected (Fig. C, left, lanes 17 to 19). Most interestingly, when the 0.3 M RNA affinity column fraction of the WERI extract was used, virtually no repression of splicing was observed (lanes 11 to 13). Thus, the presence of other factors in the 0.3 M fraction may further reduce the repressor activity of nPTB.
Control introns from adenovirus and β-globin were unaffected by either nPTB or PTB or by the column fractions, indicating the specificity of nPTB and PTB for the src substrate (Fig. B and data not shown). Moreover, the inhibitory effect of PTB and nPTB required the CU element containing sequences upstream of the N1 exon, since the BS27 RNA was only weakly repressed by the PTB fraction and not at all repressed by nPTB (Fig. C, left, lanes 1 to 8).
The amounts of PTB and nPTB used in these experiments were carefully measured to ensure that equal amounts of protein were added in each of the titrations. These fractions also have equal binding activity for the N1 3′ splice site, as shown below. To further ensure that the observed difference in repression activity was not due to variations in our preparation of the proteins, we repeated these experiments with several independently isolated samples of PTB and nPTB and with two preparations of WERI extract. The results were always the same; PTB strongly repressed splicing, while nPTB only partially reduced the second-step products. Another titration of PTB and nPTB into a different sample of WERI extract is shown in Fig. C (right).
Assembly of the DCS complex from purified components.
Three sequence elements were defined previously in functional assays and are defined here as binding sites for particular proteins: the G tract, the CU tract, and the UGCAUG element. These three elements all reside within a 20-nt portion of the DCS sequence. The RNA contacts of an RRM or a KH domain can be between 4 and 7 nt, making these elements reasonable binding sites for individual RNA binding proteins (43
). However, one imagines that there must be extensive protein-protein contacts between the DCS complex proteins if they are simultaneously bound to this short sequence. To examine how the binding of each DCS protein affected the binding of the others, we performed gel shift experiments using recombinant or purified fractions of the individual proteins.
First, we tested whether nPTB and PTB were distinguishable in their binding properties. Purified nPTB and PTB were used in electrophoretic mobility shift assays with both the DCS RNA and the polypyrimidine tract of the N1 3′ splice site as probes (Fig. A, right and left panels, respectively). Both proteins bound the 3′ splice site RNA, with saturation reached at the same concentration of protein, although the mobilities of the PTB and nPTB complexes were slightly different. In contrast, with the DCS RNA probe, nPTB formed a larger and more abundant complex than PTB did. In the PTB binding reaction, the predominant RNA-protein complex was much faster migrating and only weakly present in the nPTB reaction. The overall binding of the two proteins to the DCS RNA sequence was much weaker than that to the 3′ splice site. Nevertheless, the affinity of nPTB for the DCS RNA probe was clearly higher than that of PTB.
FIG. 8 Binding of purified nPTB and PTB to the src N1 exon splicing-regulatory elements in the presence and absence of other protein components of the DCS complex. (A) The left panel shows a gel mobility shift analysis of nPTB (lanes 2 to 6) or PTB (lanes 7 (more ...)
Next, we examined the effect of recombinant hnRNP H on the binding of nPTB and PTB. hnRNP H interacted very weakly with the DCS RNA by itself (Fig. B, right, lane 2). However, the effect of combining the proteins was striking; binding of both nPTB and PTB was improved by hnRNP H (Fig. B, left). Complex formation in the presence of 300 ng of nPTB and 400 ng of hnRNP H was 4- to 10-fold higher than for either protein alone (Fig. B, right, lane 4). Oddly, the gel mobility of the complex formed with the combination of hnRNP H and nPTB was similar to that formed with nPTB alone (see below). hnRNP H also stimulated PTB binding, although equivalent reactions with PTB and hnRNP H exhibited at least fourfold less complex formation than with nPTB at all points in the titration (Fig. B, left). From these results, it is clear that hnRNP H strongly affects both nPTB and PTB binding to the DCS RNA.
A purified fraction containing almost exclusively KSRP and FBP did not give an observable complex with the DCS RNA in the gel shift assay (Fig. B, right, lane 3). Similarly, combining the KSRP fraction with hnRNP H gave only a weak new complex (lane 5). This KSRP fraction, when added to a complex containing both hnRNP H and nPTB, bound strongly to the DCS RNA (lane 6). The mobility of this hnRNP H-nPTB-KSRP complex was lower than that of the hnRNP H-nPTB complex and about equivalent to that of the complex formed in the crude ASP40 fraction (lane 7).
These gel shift experiments with purified proteins indicate that nPTB and/or PTB are essential for the cooperative assembly of hnRNP H and KSRP into the RNP complex. However, the comigration of the hnRNP H-nPTB complex with the complex formed with nPTB alone was confusing. A combination of cross-linking and antibody supershift experiments indicated that both proteins were indeed in the hnRNP H-nPTB complex (data not shown). The comigration was thus apparently due to a change in the stoichiometry or conformation of the nPTB complex upon addition of H. In addition to the CU tract, the DCS RNA used for these experiments has a second potential PTB binding site. This is a group of pyrimidine residues at the extreme 3′ end of the DCS RNA that were known to affect PTB binding in the RNA affinity chromatography assay (data not shown). To resolve the nPTB and hnRNP H-nPTB complexes, we shortened the DCS probe to remove these residues (Fig. C, top, probe WTs). With this WTs probe, nPTB gives a weak complex on its own (Fig. C, lane 3). hnRNP H alone gave a faster-migrating complex that was barely detectable (lane 2). The combination of these two proteins allowed the formation of a strong RNP complex band that was larger than that of either protein alone, as well as a weak band still higher in the gel (lane 4). Finally, the addition of KSRP and FBP shifted the nPTB-hnRNP H complex higher in the gel to make a three-protein DCS complex (lane 5).
The resolution of the nPTB and hnRNP H-nPTB complexes allowed us to examine the importance of the different sequence elements in the binding of each protein, using the gel shift assay. These results agreed well with the cross-linking and affinity chromatography results. Mutation of the G tract eliminated the weak binding of hnRNP H, as expected (Fig. C, lane 7). Interestingly, this mutation stimulated nPTB binding (compare lane 8 with lane 3). The DCS sequence can form a secondary structure that pairs the G tract with the CU tract. Mutation of the G tract presumably allows easier access of nPTB to the CU element by disrupting this RNA structure. This may be part of the stimulatory activity of hnRNP H for nPTB binding; by binding to the G tract, hnRNP H may make the nPTB site more accessible. The addition of hnRNP H to the DCS with a mutant G tract did not further stimulate nPTB binding, although a faint higher-molecular-weight complex was seen (lane 9). The addition of KSRP and FBP shifted the PTB complex further, although this complex was smeared and was apparently less stable than the hnRNP H-nPTB-KSRP complex seen on wild-type RNA (lane 10).
Mutation of the CU tract also disrupted the potential secondary structure between this element and the G tract. However, the binding of hnRNP H was only weakly stimulated by this change (lane 12). Presumably, hnRNP H needs nPTB to bind well, whether the secondary structure is there or not. As expected, the CU tract mutation nearly eliminated nPTB binding (lane 13). Interestingly, the combination of hnRNP H and nPTB gave some complexes with this RNA (lane 14). These ranged in mobility from the size of hnRNP H alone to the size of the full DCS complex. It is possible that the interaction between nPTB and hnRNP H stimulates binding even if not all the normal RNA contacts can be made to the mutant RNA. These hnRNP H-nPTB complexes coalesced into a strong DCS complex in the presence of KSRP (lane 15).
Mutation of the UGCAUG element also behaved as predicted. PTB formed a complex on its own (lane 18). This was shifted to an hnRNP H-nPTB complex when hnRNP H was added, although the stimulation of binding was not as strong as that seen with the wild-type RNA (lane 19). Most significantly, the UGCAUG is essential to KSRP binding, since the hnRNP H-nPTB complex did not show a substantial further shift in the presence of KSRP (lane 20).
Taken together, the binding data provide a consistent picture of the DCS complex (Fig. ). This complex is held together by both protein-RNA contacts and protein-protein contacts. The G tract is the binding site for hnRNP H and F, although it is not clear yet whether hnRNP H and F are in separate complexes or bind as a heterodimer (16
). nPTB binds to the CU tract, and this interaction is critical in stimulating the subsequent assembly of the other proteins. Finally, KSRP and FBP bind to the UGCAUG element, although, again, these two proteins may bind separately or together.
FIG. 9 Diagram of the DCS RNP complex. The position of each protein along the DCS RNA is indicated. These contacts are supported by cross-linking, affinity chromatography, and gel shift assays using wild-type and mutant DCS RNA sequences. Homologous pairs of (more ...)