Splicing factor 1 (SF1) functions at early stages of pre‐mRNA splicing and contributes to splice site recognition by interacting with the essential splicing factor U2AF65 and binding to the intron branch site. We have identified an 80 kDa substrate of cGMP‐dependent protein kinase‐I (PKG‐I) isolated from rat brain, which is identical to SF1. PKG phosphorylates SF1 at Ser20, which inhibits the SF1–U2AF65 interaction leading to a block of pre‐spliceosome assembly. Mutation of Ser20 to Ala or Thr also inhibits the interaction with U2AF65, indicating that Ser20 is essential for binding. SF1 is phosphorylated in vitro by PKG, but not by cAMP‐dependent protein kinase A (PKA). Phosphorylation of SF1 also occurs in cultured neuronal cells and is increased on Ser20 in response to a cGMP analogue. These results suggest a new role for PKG in mammalian pre‐mRNA splicing by regulating in a phosphorylation‐dependent manner the association of SF1 with U2AF65 and spliceosome assembly.
The splicing of introns from nuclear pre‐mRNA is catalysed by the spliceosome. This is a large ribonucleoprotein complex that assembles by a network of interactions between the pre‐mRNA, small nuclear ribonucleoprotein particles (snRNPs) and many non‐snRNP proteins (for a review see Krämer, 1996). These interactions contribute to the accuracy and specificity of splice site recognition that is necessary for the generation of mRNAs encoding functional proteins. Elements essential for the definition of splice sites are confined to the intron extremities and these are recognized several times during the assembly of the catalytically active spliceosome (Krämer, 1996; Reed, 1996). The 5′ splice site is initially defined by base pairing with U1 snRNA in the early complex E, an interaction that is later replaced by U6 snRNA. In the same complex, two conserved elements that are located upstream of the 3′ terminal AG dinucleotide of the intron, the branch site and a polypyrimidine tract, are recognized by two protein factors. The large subunit of splicing factor U2AF (U2AF65) binds to the polypyrimidine tract, whereas SF1 (also termed branch point bridging protein, BBP) interacts with the branch site (Berglund et al., 1997). U2AF65 and SF1 contact one another (Abovich and Rosbash, 1997; Rain et al., 1998), resulting in cooperative binding of both proteins to the pre‐mRNA (Berglund et al., 1998a). Direct contacts between U2AF65 and the branch site have also been reported (Valcárcel et al., 1996); however, it is not known whether binding of SF1 and U2AF65 to this region is mutually exclusive or occurs at the same time. SF1 is implicated in the bridging of splice sites by associating, directly or indirectly, with the U1 snRNP bound to the 5′ splice site (Abovich and Rosbash, 1997; Bedford et al., 1998). Members of the SR family of splicing factors, which contain Ser/Arg‐rich domains, also participate in complex E formation and function in the juxtaposition of splice sites. These proteins can bind directly to the pre‐mRNA and engage in interactions between a U1 snRNP‐specific protein (U1 70K) and the 35 kDa subunit of U2AF (Fu, 1995). Binding of U2AF65 to the pre‐mRNA and the presence of SF1 are essential for the following step in the assembly pathway, the formation of a short helix between U2 snRNA and the branch site, which results in assembly of pre‐splicing complex A (Krämer, 1992; Zamore et al., 1992). Completion of spliceosome assembly requires many additional proteins and involves the incorporation of the U4·U6/U5 snRNP into complex A, resulting in the formation of complex B. After a conformational change, complex C is formed and intron removal is catalysed within this complex in two consecutive transesterification reactions.
Splicing is regulated by reversible protein phosphorylation (see Manley and Tacke, 1996; Xiao and Manley, 1998). Among the 50 or more non‐snRNP proteins that participate in splicing, the best studied targets for phosphorylation are the SR proteins. Their phosphorylation is required at the onset of spliceosome assembly and for the incorporation of the U4·U6/U5 snRNP into pre‐splicing complex A, whereas dephosphorylation events are essential for catalysis. These modifications affect the interaction of SR proteins with each other, other splicing components and the pre‐mRNA, thus contributing to the fine‐tuning of the splicing reaction. Several kinases that phosphorylate SR proteins have been identified. Among these are SRPK1 and SRPK2, which specifically phosphorylate SR proteins (Gui et al., 1994; Wang,H.Y. et al., 1998) and the dual specificity kinase Clk/Sty (Colwill et al., 1996). To date, the only non‐SR protein shown to be phosphorylated is SF3b155/SAP155, a component of the U2 snRNP. SF3b155 is phosphorylated in spliceosomal complexes concomitant with or just after the first catalytic step (Wang,C. et al., 1998) and is a target for cyclin E/cyclin‐dependent kinase 2 (Seghezzi et al., 1998). In addition to regulating spliceosome assembly, phosphoryl‐ation affects the intranuclear distribution of splicing proteins (Misteli et al., 1997 and references therein) and alternative splice site selection (Cardinali et al., 1994; Duncan et al., 1997; Kanopka et al., 1998). Despite the advances in unravelling how the activities of splicing factors are modulated by phosphorylation, the upstream signals that trigger this regulation are largely unknown. Cellular signalling mechanisms that are, for example, mediated by insulin action (Chalfant et al., 1995) or involve signalling by calcium and protein kinase C (PKC) (König et al., 1998), can regulate alternative splicing events; however, in these cases the downstream targets remain to be explored.
A major signalling pathway common to many cells is mediated by the second messenger cGMP. Intracellular levels of cGMP are responsive to hormones of the atrial natriuretic peptide family and to a variety of other hormones which activate or induce the synthesis of nitric oxide synthase leading to the production of nitric oxide (Wang and Robinson, 1997). cGMP plays a role in a variety of physiological responses, from learning and memory to apoptosis. Major targets for cGMP signalling are the two Ser/Thr protein kinases PKG‐I and PKG‐II (Wang and Robinson, 1997 and references therein). Knockout mice lacking PKG‐I have severe vascular and intestinal dysfunctions (Pfeifer et al., 1998), while those lacking PKG‐II have defects in intestinal secretion and severe dwarfism (Pfeifer et al., 1996). PKG has a relatively limited substrate specificity compared with the highly related protein kinase A (PKA) and only a small number of substrates have been identified (Wang and Robinson, 1997). One target for PKG‐I is c‐raf1, phosphorylation of which blocks growth factor signalling to MAP kinase leading to reduced cellular proliferation (Suhasini et al., 1998). New approaches have revealed novel putative PKG substrates in brain and other tissues by in vitro phosphorylation of proteins in tissue extracts (Wang and Robinson, 1995). The identification of these substrates is crucial for the understanding of PKG signalling. Here, we identify one of these substrates, an 80 kDa rat brain protein, as splicing factor SF1. PKG‐I phosphorylates SF1 at Ser20, which interferes with the SF1–U2AF65 interaction and results in an inhibition of splicing complex formation. Our data suggest a new role for PKG and the cGMP signalling pathway in pre‐mRNA processing.
Identification of an 80 kDa brain PKG substrate as SF1
We have previously identified 48 proteins in rat brain that show a higher specificity for phosphorylation by PKG than by PKA (Wang and Robinson, 1995). An 80 kDa protein (P80) that was specifically phosphorylated by PKG was purified and characterized. A rat brain extract enriched in Ca2+‐sensitive lipid‐binding proteins was subjected to Q‐Sepharose chromatography and proteins were eluted with a 0–400 mM NaCl gradient (Figure 1). Aliquots of each column fraction were phosphorylated in the presence of exogenous PKG or the catalytic subunit of PKA. Equal specific activities of PKG and PKA were used, as determined by phosphorylation of a synthetic peptide containing amino acids 8–21 of phospholamban (PL8–21). As previously reported, virtually no specific PKG substrates could be detected in the crude cellular extract (Figure 1A, lanes 1–3) nor in the unbound material (lanes 4–6). P80 was eluted from the column at 100 mM NaCl (in fractions QF10–14) and was phosphorylated by PKG, but not PKA (Figure 1A, lanes 7–9). Fractions containing P80 were pooled and subjected to chromatography on S‐Sepharose. Bound proteins were eluted with a 0–400 mM NaCl gradient and aliquots of each fraction were phosphorylated. P80 eluted at 200–250 mM NaCl (in fractions SF22–26; Figure 1B, lanes 10–12). At this point ∼10 μg of P80 were obtained that was sufficiently pure for amino acid sequencing (Figure 1C).
P80 was in‐gel digested with Endo‐Lys‐C and peptides were separated by high‐performance liquid chromatography (HPLC). Amino acid sequencing of four of the major peptides revealed 100% identity of P80 with human SF1 (SF119–28 RSRWNQDTME, SF194–103 LNTREFRTRK, SF1228–239 QGIETPEDQNDL and SF1298–308 PGDPQSAQDKxR). The peptides originated from four distinct regions of the protein and single amino acid sequences were obtained in all cases, suggesting that P80 fully accounted for the Coomassie Blue‐stained protein. To complete the identification of P80 as SF1, antibodies against two synthetic peptides of SF1 were raised in sheep and used for immunoblotting. The antibodies strongly reacted with recombinant SF1. P80 was detected in the crude brain extract and in S‐Sepharose column fractions (data not shown). These results demonstrate that P80 is identical to SF1.
To study the relationship between P80 and SF1 further, recombinant His6‐tagged versions of SF1 (Figure 2A) were phosphorylated by PKG, PKA and PKC. SF1 proteins containing the N‐terminal 320 (C4) or 137 (C2) amino acids were efficiently phosphorylated by PKG (Figure 2B, lanes 1 and 2). No or only weak phosphorylation was observed in the presence of PKC or the catalytic subunit of PKA, respectively (Figure 2B, lanes 5–12). The activity of PKC and PKA was confirmed by their ability to phosphorylate the SR protein ASF/SF2 or histone H1 (Figure 2B, lanes 13–15), which are known substrates for these enzymes (Colwill et al., 1996). When SF1 proteins with N‐terminal truncations of 28 (N1a) or 93 (N1) amino acids were used, phosphorylation by PKG was abolished. Thus, recombinant SF1 serves as a specific substrate for PKG in vitro as observed with SF1 purified from rat brain. Moreover, the N‐terminal 28 amino acids are essential for this reaction, suggesting that the phosphorylation site is localized within this region.
SF1 is specifically phosphorylated by PKG but not by other protein kinases
Next the specificity of SF1 phosphorylation by various protein kinases was addressed. As shown in Figure 3A, rat brain SF1 was an excellent substrate for PKG, but was not phosphorylated by six other protein kinases: the catalytic subunit of PKA, cdc2 kinase, c‐src, casein kinase 2 (CK2), PKC and calcium and calmodulin‐dependent protein kinase II (CaMPK‐II). The activity of each kinase was confirmed in separate experiments by their ability to phosphorylate purified rat brain dynamin I (PKC, cdc2 and CK2) or to phosphorylate proteins present in other S‐Sepharose column fractions (data not shown). This result demonstrates that SF1 is a specific substrate for PKG, but not for six other protein kinases tested. As shown in Figure 2B, recombinant SF1‐C4 is also preferentially phosphorylated by PKG compared with PKA, and is not phosphorylated by PKC.
PKG phosphorylates SF1 at Ser20 in vitro
To determine whether recombinant and purified SF1 are phosphorylated by PKG at the same site(s), native SF1 purified from rat brain (nSF1) and recombinant SF1‐C4 (rSF1) were phosphorylated with PKG and subjected to phosphopeptide mapping with V8 protease (Figure 3B). A major phosphorylated peptide of 16 kDa was detected with both proteins, suggesting that PKG phosphorylates the same site or region in native and recombinant SF1. The minor, slower migrating phosphopeptide detected with nSF1 represents a partial proteolytic product that in some experiments is further digested to the 16 kDa peptide. PKG is a Ser/Thr‐specific protein kinase and phosphoamino acid analysis revealed that PKG phosphorylated nSF1 exclusively on Ser (Figure 3C).
Studies with synthetic peptide libraries (Tegge et al., 1995) and comparison of the sequences of the few known PKG substrates (Wang and Robinson, 1997) have revealed the motif RKxS/T (x being any amino acid) as the optimal sequence context for PKG phosphorylation. Amino acids 17–20 of SF1 showed an excellent match to this motif (Figure 2A). Taken together with the observation that the N‐terminal 28 amino acids of SF1 were required for phosphorylation, Ser20 would be expected to represent the phosphate acceptor. To test this, PKG‐phosphorylated SF1‐C4 was subjected to digestion by Lysyl‐endopeptidase C and the peptides were separated by HPLC. Two peaks of radioactivity were obtained. Since SF1 was exclusively phosphorylated on Ser, both fractions were derivatized to convert phosphoserine to S‐propylcysteine, which is detectable by HPLC because of its specific elution after leucine (data not shown). The sequences of both peptides contained Ser20 (RSRWNQDTME and RKRSRWNQDT) and the derived moiety revealed that the phosphorylation site was Ser20.
To confirm this result, Ser20 of SF1‐C4 was mutated to Thr (SF1‐C4‐S20T) or Ala (SF1‐C4‐S20A). Compared with SF1‐C4, SF1‐C4‐S20T was phosphorylated by PKG at a greatly reduced rate (Figure 4). Thus, Ser serves as a much better target for PKG than Thr in the context of this sequence. Mutation of Ser20 to Ala completely abolished phosphorylation, providing further evidence that Ser20 represents the only PKG phosphorylation site in SF1.
The kinetics of phosphorylation were determined with a synthetic peptide, SF115–25, that included Ser20. The kinetics (Km and Vmax) of phosphorylation of SF115–25 by PKG and PKA were compared with those of one of the best PKG peptide substrates, PL8–21, which exhibits similar kinetics for PKG and PKA (Kemp and Pearson, 1991; Table I). SF115–25 served as a high affinity PKG substrate and was phosphorylated at similar rates as PL8–21 (Table I). In contrast to PL8–21, however, SF115–25 was phosphorylated 13.6‐fold less by PKA. Therefore the specificity for phosphorylation by PKG as compared with PKA was the same for SF115–25 and the native and recombinant proteins.
Phosphorylation of SF1 on Ser20 in intact cells
To determine whether SF1 phosphorylation could be of physiological relevance, the phosphorylation of SF1 in intact cells was investigated using the rat brain‐derived neuronal‐like cell line B35 (Schubert et al., 1974). These cells differentiate in response to cAMP stimulation and extend axonal‐like processes. Differentiated cells were labelled with 32Pi, washed and stimulated with membrane‐permeable cGMP or cAMP analogues (8‐pCPT‐cGMP or 8‐pCPT‐cAMP). SF1 was immunoprecipitated with the anti‐SF1 peptide antibody and immunoprecipitates were separated by SDS–PAGE in parallel with in vitro PKG‐phosphorylated SF1 as a marker. A protein co‐migrating with SF1 was weakly phosphorylated in B35 cells and increased in phosphorylation in response to stimulation with the cGMP analogue (Figure 5, lanes 4 and 5). The response was specific, since the cAMP analogue did not induce phosphorylation of the protein (Figure 5, lane 6). Control experiments indicated that SF1 was not immunoprecipitated with pre‐immune serum (Figure 5, lanes 7–9). A faster migrating 76 kDa 32P‐labelled band immunoprecipitated with the SF1 serum was non‐specific as it was also observed with pre‐immune serum. This result suggests that SF1 is phosphorylated in vivo in response to extracellular stimuli that activate the cGMP signalling pathway.
The next aim was to determine whether SF1 was phosphorylated on Ser20 in vivo. Polyclonal antibodies were raised in sheep against a synthetic peptide encompassing the phosphorylation site and which included a phosphorylated serine [SF117–24 RKRS(P)RWNQ]. The specificity of the antibodies was confirmed by Western blotting. Anti‐SF1 antibodies recognized recombinant SF1, regardless of its phosphorylation by PKG (Figure 6A), while the anti‐phospho‐Ser20 antibodies preferentially detected PKG‐phosphorylated SF1 (Figure 6B). When SF1 was immunoprecipitated from duplicate plates of B35 cells with protein G–Sepharose‐purified anti‐SF1 antibodies, a constant amount of SF1 was recovered from control and cGMP‐stimulated cells. This was determined by probing the blots with the same anti‐SF1 antibodies as those used for immunoprecipitation (Figure 6C). Notably, the non‐specific 76 kDa band (Figure 5) was not detected. The blot was then stripped and re‐probed with protein G–Sepharose‐purified anti‐phospho‐Ser20 antibodies, which primarily detected SF1 from the cells that had been stimulated with the cGMP analogue (Figure 6D). This demonstrates that SF1 is phosphorylated in vivo on Ser20 in response to a cGMP stimulus.
Phosphorylation of SF1 regulates its interaction with U2AF65 and inhibits pre‐spliceosome assembly
SF1 functions during early stages of spliceosome assembly by binding to the intron branch site and interacting with splicing factor U2AF65 (see Introduction). To determine whether these activities are modulated by phosphorylation we first examined the interaction of phosphorylated SF1 with U2AF65. His6‐tagged SF1‐C4 was phosphorylated or mock‐phosphorylated (no ATP) by PKG, purified by metal‐affinity chromatography and incubated with glutathione S‐transferase (GST)‐tagged U2AF65 bound to glutathione–agarose beads. After removal of unbound material, bound proteins were eluted with SDS sample buffer, separated by SDS–PAGE and blotted to a nylon membrane. SF1‐C4 was detected by antibodies directed against the N‐terminal His6‐tag. As previously shown for non‐phosphorylated SF1 (Rain et al., 1998), mock‐phosphorylated SF1‐C4 bound to GST–U2AF65 (Figure 7A). After phosphorylation, the SF1–U2AF65 interaction was severely reduced at all concentrations of SF1‐C4 tested. No binding was observed with GST bound to the beads or with glutathione beads alone. As a control, a similar experiment was performed after incubation of SF1‐C4 with PKA, which results in only weak phosphorylation of the protein. In this case the mock‐phosphorylated and phosphorylated proteins interacted with U2AF65 with no apparent differences (data not shown). We conclude that the specific phosphorylation of SF1 at Ser20 by PKG inhibits its interaction with U2AF65.
Analysis of functional domains in SF1 has revealed that sequences encompassing the KH domain (amino acids 136–228) are essential for RNA binding (Berglund et al., 1998b; Rain et al., 1998). Consistent with these results, phosphorylation of SF1‐C4 did not affect the ability of the protein to bind RNA, as tested by UV crosslinking of the protein to a pre‐mRNA derived from the adenovirus major late transcription unit (data not shown).
The N‐terminal 28 amino acids of SF1 are essential for both binding to U2AF65 and pre‐spliceosome formation (Rain et al., 1998). Given the result that SF1 phosphorylated at Ser20 cannot bind to U2AF65 we anticipated that PKG phosphorylation would also inhibit the activity of SF1 in spliceosome assembly. SF1‐C4 was PKG‐phosphorylated, purified and added to reactions containing partially purified splicing components that are essential for the formation of pre‐splicing complex A (Figure 7B; Krämer and Utans, 1991). Splicing complexes A and B/C are assembled in HeLa cell nuclear extracts (Figure 7B, lane 1). Complex A was efficiently assembled in the reconstituted system when SF1 purified from HeLa cells was added (Figure 7B, lane 3), as compared with a control performed in the absence of SF1 (lane 2; the background activity observed in this reaction is due to a small contamination of the SF3a/b fraction with SF1). Complex A was also formed upon addition of increasing amounts of mock‐phosphorylated SF1‐C4 (Figure 7B, lanes 4–6). In contrast, PKG phosphorylation of SF1‐C4 completely inhibited spliceosome assembly (Figure 7B, lanes 7–9). Incubations were performed in the presence of okadaic acid to inhibit protein phosphatases present in fractions containing the partially purified splicing factors. In the absence of okadaic acid the phosphorylated SF1‐C4 was dephosphorylated within 5 min of incubation at 30°C and was fully active in complex assembly (data not shown). These results demonstrate that phosphorylation of SF1 at Ser20 inhibits pre‐spliceosome formation, presumably due to a failure to interact with U2AF65.
Ser20 is required for SF1 function
The role of Ser20 for SF1 activity was further analysed with SF1 proteins carrying Ser20 mutations. SF1‐C4, SF1‐C4‐S20T and SF1‐C4‐S20A were incubated with PKG in the presence or absence of ATP and reacted with GST–U2AF65. Mutation of Ser20 to Thr severely reduced the interaction, and binding of SF1 to U2AF65 was virtually abolished with the Ala20 mutant protein (Figure 8A). Again, phosphorylation of SF1‐C4 and SF‐C4‐S20T inhibited the interaction with U2AF65. Thus Ser20 is essential for the binding of SF1 to U2AF65.
The mutant proteins were tested for their activity in pre‐spliceosome assembly. When 2 or 20 ng of SF1‐C4‐S20A were added to the reaction, only background levels of complex A were observed (Figure 8B, lanes 7 and 8). Addition of 100 ng of SF1‐C4‐S20A resulted in the assembly of complex A, but the efficiency was reduced (to 57%) when compared with SF1‐C4 (Figure 8B, lanes 6 and 9). Mutant protein SF1‐C4‐S20T was active at all concentrations tested (Figure 8B, lanes 10–12); however, the efficiency of complex formation was lower than that observed with the SF1‐C4 control (ranging from 15 to 88%). We conclude that the activity in pre‐spliceosome assembly of the mutant proteins reflects the efficiency of their interactions with U2AF65. It might be surprising that at high concentrations (100 ng) both mutant proteins allow for the assembly of complex A, while U2AF65 binding has been severely compromised. However, interactions of SF1 with proteins other than U2AF65 most likely contribute to the association of SF1 with the spliceosome (Abovich and Rosbash, 1997; Bedford et al., 1998). Our results demonstrate that both PKG phosphorylation of SF1 at Ser20 and mutation of this residue interfere with the interaction between SF1 and U2AF65 and result in reduced levels of pre‐spliceosome assembly. This explains the requirement of the N‐terminal 28 amino acids of SF1 for contacts with U2AF65 and splicing complex formation (Rain et al., 1998).
Splicing takes place in the active spliceosome, a dynamic complex of proteins and snRNPs that assembles in a stepwise manner on the pre‐mRNA. The initial stage of spliceosome assembly involves an interaction between the two splicing factors SF1 and U2AF65. Here we show that SF1 is phosphorylated by PKG on Ser20, which blocks its interaction with U2AF65 and inhibits assembly of spliceosomal complex A. The ability of SF1 to interact with U2AF65, but not with RNA, was found to be regulated by PKG phosphorylation, but not by PKA. The phosphorylation of SF1 by PKG is the first indication that pre‐mRNA splicing may be regulated by cGMP‐PKG signalling.
Several criteria were used to identify rat brain P80 as SF1. Amino acid sequencing of four Lys‐C‐derived peptides revealed 100% identity with SF1 and antibodies raised against synthetic peptides of SF1 specifically recognized P80. Comparison of SF1 and P80 phosphorylation in vitro revealed that both were phosphorylated in a similar site by phosphopeptide mapping. The PKG phosphorylation site in SF1 was determined to be Ser20 by direct sequencing of proteolytic products combined with the development of an improved method of chemical conversion of phosphoserine to S‐propylcysteine. The site was confirmed by kinetic analysis of phosphorylation of the synthetic peptide SF115–25, which contained Ser20 as the only Ser residue. The sequence context of the phosphate acceptor site in SF1 is KKRKRSRWN, which agrees well with the motif necessary for efficient PKG phosphorylation determined in studies with peptide libraries on cellulose papers (Tegge et al., 1995). Those studies revealed the minimal motif RKxS/T and highlighted a key role for additional basic amino acids on either side of the RK for optimal Vmax. The PKG phosphorylation site was determined with recombinant SF1 that lacked the C‐terminal proline‐rich region and sequences specific to different SF1 isoforms (Krämer et al., 1998 and references therein), raising the possibility that Ser20 might not be the only phosphate acceptor in SF1. Two observations argue against this. First, when the phosphorylation of native and recombinant SF1 was compared by phosphopeptide mapping with V8 protease, a major 16 kDa product was observed with both proteins, suggesting phosphorylation in the same region. PKG failed to phosphorylate SF1‐N1a (amino acids 29–370) or a Ser20 to Ala mutation in SF1‐C4 (amino acids 2–320), demonstrating the presence of a single PKG phosphorylation site in the N‐terminal half of the protein. Secondly, no consensus PKG‐I phosphoryl‐ation motifs are found in the regions of SF1 that are absent from the recombinant proteins used in this study. Therefore Ser20 is the only phosphorylation site for PKG‐I in SF1.
SF1 phosphorylation revealed a high degree of selectivity for PKG. Only four other relatively specific PKG substrates have been previously identified and had their phosphorylation sites sequenced (Wang and Robinson, 1997). SF1 did not serve as an in vitro substrate for PKC, cdc2, c‐src or CaMPK‐II, although potential phosphorylation sites for some of these enzymes are found in SF1. Moreover, SF1 was a poor substrate for PKA, the protein kinase that is most closely related to PKG (Wang and Robinson, 1997). This kinase selectivity was emphasized when the kinetics of phosphorylation of SF115–25 were compared with those of PL8–21. Whereas PKG and PKA phosphorylated PL8–21 with very similar kinetics (Kemp and Pearson, 1991; Table I), SF115–25 was a much better substrate for PKG than for PKA. PKG and PKA are known to show overlapping substrate specificities and the canonical PKA phosphorylation site (RRxS/T) is also recognized by PKG (Tegge et al., 1995). Therefore, the minimal motif determined for optimal PKG phosphorylation, RKxS/T, does not account for kinase selectivity and little information is available regarding sequences that may contribute to this. The presence of a Phe C‐terminal to the phosphorylation site has been suggested as a negative determinant for PKA phosphorylation (Colbran et al., 1992). SF1 also contains an aromatic amino acid (Trp) in a similar position, but whether this residue accounts for kinase selectivity remains to be tested. In contrast, other substrates for PKG that exhibit this selectivity, VASP and G‐substrate, do not have aromatic amino acids C‐terminal to the phosphate acceptor site (Wang and Robinson, 1997 and references therein).
SF1 is phosphorylated in vivo after appropriate stimulation of the cGMP‐PKG signalling pathway. Immunoprecipitates from lysates of 32P‐labelled intact neuronal cells contained very low levels of phosphorylated SF1, suggesting that the protein normally predominates in its dephospho‐form in these cells. Stimulation of cells with a cGMP analogue increased phosphorylation, while a cAMP analogue had no effect. Phosphorylation site‐specific antibodies also revealed that elevated phosphorylation occurred on Ser20, although in vivo phosphorylation at additional sites cannot be ruled out at this stage. Therefore, SF1 is most likely a target for PKG signalling in vivo. SF1 exists in several isoforms which share the N‐terminal 448 amino acids, but vary in the length of the proline‐rich region and their C‐termini (Krämer et al., 1998 and references therein). Therefore all known isoforms contain Ser20 and we predict that they may all be targets for PKG‐I. Notably, only the 80 kDa form of SF1 and a 76 kDa non‐specific protein were immunoprecipitated from 32P‐labelled cells, but since the latter band was not immunoreactive for SF1 it was not considered further as a potential splice variant. The 80 kDa band observed in the immunoprecipitates may contain more than one SF1 isoform or smaller forms of SF1 may not be expressed in the B35 cell line.
PKG phosphorylation of SF1 blocked its ability to bind to U2AF65 and inhibited pre‐spliceosome assembly. Another known property of SF1, RNA binding, was unaffected. SF1 and U2AF65 bind independently of each other to the branch site and the polypyrimidine tract, respectively (Zamore et al., 1992; Abovich and Rosbash, 1997; Berglund et al., 1997). Upon interaction the proteins cooperate to form a ternary complex with the pre‐mRNA, and the affinity of SF1 and U2AF65 for the RNA is increased several fold (Berglund et al., 1998a). The results that the assembly of complex A is severely decreased in the presence of phosphorylated SF1 or SF1‐C4‐S20A demonstrate that the contacts between SF1 and U2AF65 are crucial for the incorporation of the U2 snRNP into the spliceosome and thus for the initial recognition of the branch site by SF1.
Phosphorylation of Ser20 introduces two negative charges into a highly positively charged region of SF1 which may explain the block in protein–protein inter‐ action. However, mutation of Ser to Thr or Ala also compromised the interaction between SF1 and U2AF65, suggesting that the nature of Ser20 itself is an essential feature of the binding site. Ser20 and neighbouring sequences are well conserved between human, Drosophila and Caenorhabditis elegans SF1 (R.Mazroui and A.Krämer, manuscript in preparation), but not in Saccharomyces cerevisiae or Schizosaccharomyces pombe. Thus, interactions between SF1 and U2AF65 may rely on the same principle in flies and worms. In addition, the conservation of these sequences suggests the possibility of PKG regulation of SF1 activity in these species. Although Ser20 is located at or near the U2AF65 interaction site in SF1, it may not be the only region responsible for binding. A two‐hybrid screen between yeast homologues of SF1 and U2AF65 defined an additional interaction region of ∼100 amino acids immediately N‐terminal to the KH domain, which are highly conserved during evolution (Rain et al., 1998; R.Mazroui and A.Krämer, manuscript in preparation). It is possible that this region also contributes to the interaction between human SF1 and U2AF65. Whether sequences corresponding to Ser20 in human SF1 that are not well conserved in the yeast SF1 homologues are required for interactions with U2AF65 homologues remains to be tested.
How could cGMP‐PKG regulate splicing? Protein kinases that phosphorylate the SR or SR‐like proteins, such as SRPK1, SRPK2 and Clk/Sty, are normally found in the nucleus and are implicated in regulating protein–protein and protein–RNA interactions during the course of the splicing reaction (Xiao and Manley, 1998 and references therein). In contrast, PKG‐I is a cytoplasmic protein, which makes it unlikely that it constitutively modulates SF1 activity during the spliceosome cycle. Upon stimulation of cells with cGMP, a cryptic nuclear localization signal (NLS) in the ATP binding domain of the protein is exposed, and PKG translocates to the nucleus (Gudi et al., 1997). Immunofluoresence studies confirmed that PKG is present in B35 neuronal cells and is detected in the nucleus when the cells are stimulated with a cGMP, but not cAMP analogue (X.Wang and P.J.Robinson, unpublished results). In agreement with this observation, SF1 phosphorylation was highly increased on Ser20 after stimulation with cGMP. These results have three implications. First, phosphorylation is likely to be mediated by nuclear PKG‐I, because SF1 shows an intranuclear distribution typical of many splicing components (D.Nesic and A.Krämer, unpublished results). Phosphorylation by cytoplasmic PKG would be expected if SF1 shuttled between the nucleus and the cytoplasm, as observed for various hnRNP proteins and a subset of SR proteins (Nigg, 1997; Cáceres et al., 1998). In this respect we note that Ser20 is immediately adjacent to a putative NLS in SF1, which raises the possibility that PKG phosphorylation also regulates the intracellular localization of SF1. Secondly, in unstimulated cells we do not expect a regulation of SF1 activity by PKG. However, upon stimulation of cGMP signalling, splicing would be negatively regulated by PKG through interference with the SF1–U2AF65 interaction and spliceosome assembly. Thirdly, we do not expect that PKG will be present in all cells, which limits this regulatory pathway to those cells which express PKG. PKG is found in a wide variety of cell types in low levels, and in high levels in a limited number of cells like Purkinje neurons or smooth muscle cells. However, this does not preclude another unidentified cellular protein kinase from regulating SF1 in cells which do not express PKG.
PKG or cGMP signalling pathways have been linked to gene expression at the transcriptional level (Gudi et al., 1997), but effects on pre‐mRNA splicing have not been previously reported. It is known that alternative splice‐site selection can be modulated by signal transduction pathways. For example, levels of mRNA for PKCβI in BC3H‐1 myocytes quickly decline with insulin treatment, and concomitantly mRNA and protein levels increase for its alternatively spliced variant PKCβII (Chalfant et al., 1995). High levels of phorbol ester mimic insulin action in switching alternative splicing between these transcripts. In another study, it has been demonstrated that CD44 alternative splicing is regulated in vivo by signalling through synergistic action of PKC and calcium (König et al., 1998). These studies demonstrate that cell signalling pathways can influence alternative splicing, but the protein kinases involved or their final targets are not known. Our results highlight a new role for PKG in splicing by regulating the association of SF1 with U2AF65 in a phosphorylation‐dependent manner and thereby controlling early steps in spliceosome assembly. The increase in SF1 phosphorylation by cGMP analogues in intact cells provides a first hint that splicing may also be modulated by PKG in vivo when cells are appropriately activated.
Materials and methods
[γ‐32P]ATP (3000 Ci/mmol) was from Amersham. l‐phosphatidylserine (PS), 1,2‐diolein, cAMP, cGMP, anti‐polyHis antibodies and Staphylococcus aureus V8 protease were from Sigma. Leupeptin, calmodulin, histone H1, protein kinase CK2 (human, recombinant), p34cdc2/cyclin B (human, recombinant) and c‐src were from Calbiochem. 8‐pCPT‐cGMP and 8‐pCPT‐cAMP were from Biolog Life Science Institute (Bremen, Germany). Okadaic acid was from Gibco‐BRL. Rat brain CaMPK‐II was a generous gift from Dr John Rostas, University of Newcastle, NSW, Australia. Synthetic peptide substrates were SF115–25 (KKRKRSRWNQD) from Auspep, Melbourne; PL8–21 (TRSAIRRASTIEMP) from Macromolecular Resources, CO; and the Walsh inhibitor of PKA (PKI5–24 TTYADFIASGRTGRRNAIHD) from the Peptide Synthesis Group of Louisiana State University. All peptides were purified by HPLC to >90% purity.
P80 was purified from a fraction enriched in Ca2+‐sensitive lipid‐binding proteins (Wang and Robinson, 1995) prepared from 40 rat brains. Proteins in this fraction were precipitated by (NH4)2SO4 to 80% saturation, dialysed against buffer A [20 mM Tris–HCl, 1 mM dithiothreitol (DTT), 1 mM EDTA, 0.1% Tween 80 pH 7.7] and applied to a Q‐Sepharose high‐performance anion exchange column (10 × 1.6 cm, Amersham Pharmacia Biotech). After washing in buffer A, bound proteins were eluted with a 0–400 mM NaCl gradient at 2 ml/min and sixty 2 ml fractions were collected. Aliquots of every second fraction were phosphorylated (see below) and fractions containing P80 were pooled. The pool was diluted to adjust the NaCl concentration to <40 mM and the pH to 7.0, and applied to an S‐Sepharose high‐performance cation exchange column (10 × 1.6 cm, Amersham Pharmacia Biotech) which was pre‐equilibrated with buffer A, pH 7.0. After extensive washing, bound proteins were eluted with 0–400 mM NaCl gradient at 1 ml/min and sixty 1 ml fractions were collected. Aliquots of the column fractions were phosphorylated and fractions containing P80 were pooled, concentrated and desalted with an ultra free‐4 centrifugal filter device (Millipore, MA) and stored at −20°C in buffer A, pH 7.0.
The catalytic subunit of PKA and protein kinases C (a mix of PKCα, β and γ) were purified from bovine lung and rat brain as previously described (Robinson et al., 1993). Bovine lung PKG‐I was purified to homogeneity using DEAE cellulose and cAMP‐affinity columns with additional steps as described (Robinson et al., 1993) and was stored at −70°C in 10% glycerol with 0.05% Tween 80.
Phosphorylation of brain extracts or column fractions was performed in a final volume of 80 μl. The reaction contained 30 mM Tris–HCl pH 7.4, 1 mM EGTA, 1 mM MgSO4, 40 μM ATP, 0.125 μCi/μl [γ‐32P]ATP and 0.05% Tween 80. Where appropriate, the following activators of protein kinases were included: PKC—1.2 mM CaCl2 (generating 200 μM free Ca2+), 40 μg/ml PS and 4 μg/ml diolein; CaMPK‐II—10 μg/ml of calmodulin and 1.2 mM CaCl2; or PKG—10 μM cGMP. A peptide derived from the Walsh inhibitor of PKA, PKI (0.01 mg/ml), was routinely included in the assays for PKG and PKC to prevent cross‐activation by cGMP of any PKA potentially present in column fractions. After prewarming the tubes to 30°C for 5 min, reactions were initiated by addition of 10 μl of ice‐cold brain samples, followed immediately by the purified protein kinase (20–40 ng/reaction). Reactions were terminated after 5 min by addition of 50 μl of SDS sample buffer and rapid freezing of samples on dry ice, as previously described (Robinson and Dunkley, 1983). Phosphoproteins were detected by gel electrophoresis and autoradiography (Robinson and Dunkley, 1983) using 7.5–15% acrylamide linear gradients and 20 cm gels (Protean II system, Bio‐Rad).
Phosphopeptide maps using V8 protease were performed on native SF1 and recombinant SF1 excised from dried gels as described (Dunkley et al., 1986). Phosphoamino acid analysis of 32P‐labelled proteins excised from polyacrylamide gels was performed under the conditions described previously (Robinson, 1991).
Protein kinase activity was determined in the presence of 30 mM Tris–HCl pH 7.4, 1 mM EGTA, 200 μM ATP, 2 μCi [γ‐32P]ATP, 10 mM MgSO4 in 40 μl final reaction volumes. Incubations were for 5 min at 30°C using the synthetic peptide substrates PL8–21 (Kemp and Pearson, 1991) or SF115–25 at 0.1 mg/ml. The peptides were used over a broad concentration range (0.003–0.3 mg/ml) for determination of substrate kinetics. Reactions were initiated by addition of 40 ng of PKG or 20 ng of the catalytic subunit of PKA. These amounts of PKG and PKA were determined in prior experiments to phosphorylate PL8–21 to the same level, since this substrate has the same Vmax for both protein kinases (Kemp and Pearson, 1991). Reactions were terminated by the addition of 75 mM phosphoric acid and aliquots were spotted onto Whatman P81 paper, washed 3 times for 10 min each in 75 mM phosphoric acid, dried and counted by liquid scintillation techniques (Robinson, 1992). Kinetic constants Km and Vmax were determined with the PC program Enzfit (written by Robin Leatherbarrow, Sigma).
Amino acid sequencing and phosphorylation site determination
For amino acid sequencing P80 was excised from a dried Coomassie Blue‐stained polyacrylamide gel, digested with Endo‐Lys‐C (from Achromobacter lyticus) and the resultant peptides were separated by HPLC. Four major peaks were sequenced with an Applied Biosystems Procise sequencer model 494‐HT.
To determine the phosphorylation site, recombinant SF1‐C4 (50 μg) was phosphorylated with PKG and [γ‐32P]ATP as above (except that the incubation volume was increased to 500 μl and the incubation time was 30 min) prior to digestion with Endo‐Lys‐C and separation of resultant peptides by HPLC. The labelled fraction was detected by Cerenkov counting. Two peaks of radioactivity were detected, a major peak (peptide 1, SF119–28 RSRWNQDTME) and a minor peak. Since we had determined that the only phosphorylated amino acid in SF1 was phosphoserine (see Results), to confirm which amino acid in each fraction was phosphorylated, a new method to detect phosphoserine was developed. Phosphoserine was converted to its thioether derivative, S‐propylcysteine, using a method which combines the addition of an alkanethiol to the α‐β unsaturated dehydroalanine (Reynolds et al., 1994) generated by the Ba2+‐catalysed β‐elimination of phosphoserine (Byford, 1991). The column fractions containing the 32P label were lyophilized and incubated with 50 μl of freshly prepared ethanolic solution saturated with barium hydroxide containing 20% 1‐propanethiol (350 μl ethanol, 110 μl 1‐propanethiol, 65 μl 5 M NaOH, 60 μl 0.1 M BaCl2; excess Ba(OH)2 precipitates were removed by centrifugation) for 1 h under argon at 50°C. The reaction was terminated by the addition of 10 μl of 10% trifluoroacetic acid (TFA) and the mixture was dried in a centrifugal evaporator. The residue was redissolved into 10% aqueous TFA/50% acetonitrile (50 μl) and re‐evaporated to remove residual thiol. The peptides were dissolved into the same TFA‐acetonitrile mixture (15 μl) and dried onto 8‐mm TFA‐activated glass fibre micro‐filters (Perkin Elmer) which had been pre‐treated with 1.5 mg of BioBrene Plus (Perkin Elmer). The samples were subjected to N‐terminal sequence analysis as above and PTH‐S‐propylcysteine was observed to elute, fully resolved, after PTH‐leucine.
SF1 deletion mutants containing N‐terminal His6‐tags have been described previously (Rain et al., 1998) and are shown schematically in Figure 2A. Point mutations at Ser20 were introduced by PCR (Rain et al., 1998) with complementary primers corresponding to the cDNA sequence of amino acids 15–25 of SF1. The Ser codon AGC was changed to ACC to generate SF1‐C4‐S20T and to GCC for SF1‐C4‐S20A. Amplified DNA was transformed into XL1‐Blue cells and mutations were confirmed by sequencing (T7 DNA sequencing kit, Pharmacia). Mutant SF1 plasmids were transformed into the Escherichia coli strain TOP10 (Invitrogen). Recombinant SF1 proteins were expressed and purified on Talon affinity resin as described (Rain et al., 1998). Purified proteins were dialysed against buffer D (Dignam et al., 1983). A plasmid encoding GST–U2AF65 (Zamore et al., 1992) was kindly provided by M.Green and J.Valcárcel. Expression of GST and GST–U2AF65 was performed as described (Rain et al., 1998). A plasmid encoding ASF/SF2 was a gift from J.Manley and the protein was expressed and purified as described (Ge et al., 1991).
GST binding assays
Protein phosphorylation was performed as described above, but in the presence of 1 mM ATP and omitting the [γ‐32P]ATP. Mock‐phosphorylation was performed in the absence of ATP. SF1 proteins were separated from PKG by purification on Talon affinity resin prior to interaction with U2AF65 (Rain et al., 1998). Proteins were separated in 12% SDS–polyacrylamide minigels and blotted onto Protran membranes (Schleicher and Schüll; Kyhse Andersen, 1984). SF1 was detected by incubation with anti‐poly‐His antibodies (1:3000 dilution) and anti‐rabbit secondary antibodies (1:2000 dilution, DAKO) followed by incubation with enhanced chemiluminescence (ECL) reagents (Amersham).
Splicing complexes were assembled on a synthetic 32P‐labelled pre‐mRNA substrate derived from the adenovirus major late transcription unit, as described previously (Krämer and Utans, 1991). Reactions contained partially purified SF3a, SF3b, U2AF, U1 and U2 snRNPs and SF1 isolated from HeLa cells (Krämer, 1992) or recombinant SF1‐C4 and its mutated variants. The SR proteins that are essential for the reaction are present in the snRNP fractions. Okadaic acid (1 μM) was added to inhibit endogenous protein phosphatases. Phosphorylation of SF1‐C4 and Talon purification were done as above except that PKG was omitted from the mock phosphorylation reaction. A reaction performed in HeLa cell nuclear extract (Dignam et al., 1983) served as a control for spliceosome assembly. Splicing complexes were separated in a non‐denaturing 4% polyacrylamide gel.
SF1 antibodies were raised in sheep against two different synthetic peptides: SF11–18 MATGANATPLDFPSKKRK and SF187–105 YNSEGKRLNTREFRTRKKL. The peptides were conjugated to diphtheria toxoid as a carrier and were used to immunize the same sheep. Phosphorylation site‐specific polyclonal antibodies to phospho‐Ser20 were also raised in sheep against a synthetic peptide encompassing the phosphorylation site, and including a phosphorylated serine residue at the time of peptide synthesis [SF117–24 RKRS(P)RWNQ, with a cysteine added at the C‐terminus for conjugation to diphtheria toxoid]. Peptides were synthesized by Chiron (Melbourne, Australia). Immunoblotting was performed by chemiluminescent detection (Pierce SuperSignal).
Tissue culture, intact cell labelling and immunoprecipitation
B35 cells were grown in six‐well plates in DME with 10% fetal calf serum and were differentiated with 100 μM dibutyryl‐cAMP for 2 days. The cells were washed twice with phosphate‐free medium and labelled with 500 μCi/ml 32Pi for 1.5 h. After washing with phosphate‐free medium to remove free 32Pi the cells were divided into three aliquots: (i) control, without additions; (ii) stimulated by the cell membrane permeable PKG agonist 8‐p‐chloro‐phenylthio‐cGMP (8‐pCPT‐cGMP, 200 μM); and (iii) stimulated by the PKA agonist 8‐pCPT‐cAMP (200 μM). The cells were stimulated for 30 min, washed and lysed with 20 mM Tris, 1 mM DTT, 1% Triton X‐100, 7 μg/ml leupeptin, 1 mM EGTA and 1 mM EDTA pH 7.4. SF1 was immunoprecipitated from the 32P‐labelled cell lysate as described (Robinson et al., 1993) and detected in gels by autoradiography.
For the detection of the in vivo phosphorylation site, cells were treated the same way but were not radiolabelled. Prior to immunoprecipitation the cell lysate was applied to a Q‐Sepharose column (1 ml of resin), washed with buffer A and proteins were batch‐eluted with buffer A plus 300 mM NaCl. SF1 was immunoprecipitated from the eluates and proteins were run on SDS–PAGE minigels and transferred to nitrocellulose. In these experiments the SF1 antibodies were first purified on protein G–Sepharose before immunoprecipitation. The blots were probed with the same anti‐SF1 antibodies, stripped with 0.2 M NaOH for 30 min at room temperature and re‐probed with anti‐phospho‐Ser20 antibodies. Protein G–Sepharose‐purified antibodies were used for both immunoprecipitation and blotting. The Q‐Sepharose step eliminated most of the background bands detected with the antibodies.
The authors wish to thank Eric Reynolds for the suggestion for the phosphoserine conversion to S‐propylcysteine, John Rostas for the purified CaMPK‐II, John Haycock for advice on the phosphorylation site antibodies, Michael Green and Juan Valcárcel for the gift of GST–U2AF65, Jim Manley for the plasmid encoding ASF/SF2 and Peter Rowe for comments on the manuscript. This work was supported by grants from the Australian National Health and Medical Research Council (NHMRC to P.J.R. and X.W.), the Swiss National Science Foundation and the State of Geneva (A.K.) and the Australian Research Council (P.J.M.).
- Copyright © 1999 European Molecular Biology Organization