Introduction
Trans-acting factors in the regulation of CD45 alternative splicing
Apart from the homology between cis-sequences discussed in chapter 5, trans-acting splicing factors are most likely to be conserved between different species. Mouse cells are most likely to regulate the alternative splicing of human CD45 in a similar tissue specific manner to human cells. This appears to be the case in our somatic cell hybrids and CD45 minigene transfectants (Streuli and Saito, 1989) and has previously been shown by interspecies transient heterokaryon experiments (Rothstein et al., 1992). In these, mouse T cell and thymocyte nuclei provided factors that were able to influence alternative splicing of human CD45. B cells, however, were not able to provide the postulated positive regulatory factors that would change alternative splicing of CD45 in T cells.
A trans-acting factor that affects CD45 splicing is likely to contain the following features: Firstly, it is likely to be conserved between different species since CD45 is regulated similarly in human/rodent hybrids and interspecies transient heterokaryons. Secondly, it can directly or indirectly interact with regions within exon A of CD45, either through interaction with other basal splicing factors or possibly directly with cis-acting regions within alternative exons of CD45, such as the purine-rich elements that have been mentioned earlier. Thirdly, it should be regulated itself either by differential expression, alternative splicing or posttranslational modification (i.e. phosphorylation) in response to cell signalling, differentiation or cell cycle. From an evolutionary viewpoint it would be logical that this factor existed much earlier (as the process of splicing became important for evolution of eukaryotes) than the CD45 gene (for which expression is restricted to immunological cells in higher mammals). Therefore a member of a family of similar splicing factors that control the splicing of a variety of different pre-mRNA products in different tissues and developmental processes may be involved.
The existence of positively or negatively transacting factors that may modulate alternative splicing of CD45 pre-mRNA has been postulated previously (Rothstein et al., 1992; Streuli and Saito, 1989; Tsai et al., 1989). However, these studies were mainly centred on the involvement of regulatory cis-sequences in the modulation of splicing events within the CD45
gene itself and at that time none of the postulated sequence specific splicing factors had been isolated. The first considerable advances to our understanding of the control of alternative splicing have been made since a family of splicing factors were discovered, commonly referred to as SR proteins. One of their most remarkable characteristics was their involvement in both constitutive and alternative splicing (Fu, 1995; Manley and Tacke, 1996).
Trans-acting factors implicated in alternative splicing pathwavs
The general idea that trans-acting factors are involved in the selection of alternative splice sites was proposed a while ago (Breitbart et al., 1987; Leff et al., 1986). In the light of this theory a number of indications accumulated that alternatively spliced exons contain either weaker splicing signals or are o f suboptimal length when com pared with constitutively spliced exons (Lamond et al., 1987; Stamm et al., 1994). Consequently, it was suggested that additional factors are required that recognize these exons, as individual splicing units via unique sequence elements (Cooper and Ordahl, 1989; Dobinson et al., 1989; Hedley and Maniatis, 1991; Maddon et al., 1987; Ryner and Baker, 1991; Xu et al., 1993; Yeakley et al.,
1993).
Logically, alternatively spliced CD45 isoforms may be generated by a similar mechanism requiring the presence of either positively or negatively transacting splicing factors (Streuli and Saito, 1989; Tsai et al., 1989). This was further supported by cell fusion experiments between human and mouse T and B cell lines, in which specific changes in the splicing pattern of CD45 were most satisfactory explained by the existence o f the postulated factors (Rothstein et al., 1992).
The question if SR proteins interact with the variable exons of CD45 pre-m RNA was addressed in these studies. In order to investigate the role of individual SR proteins in alternative CD45 pre-m RN A splicing the previously described CD45 m inigenes were analysed together w ith the cloned versions of a num ber of SR proteins in the COS- transfection model. Furthermore, the contribution of each domain of two individual SR proteins to their specific splicing activity was examined by domain-swap chimera and domain deletion studies.
Results
In order to examine whether distinct SR proteins can act as regulators of alternative CD45 splicing, we used a COS-co-transfection system in which the genes for various cloned SR proteins (figure 6.1) together with the LCA-2 minigene (figure 5.1) were used. If splice site choice of the CD45 pre-mRNA was affected by interaction of individual SR proteins directly or indirectly with CD45 pre-mRNA one should be able to detect shifts in the ratio of the two spliced products by semi-quantitative RT-PCR. A similar method was applied to examine the splicing activity of the human homologue of the Drosophila SWAP protein (Sarkissian et al.,
1996) The splicing factors initially tested were SRp20, SRpBOa, SRpSOb, SRpSOc, SRp40, SRp55, and SRp75, which had been cloned into the mammalian expression vector pCG (W ilson et al., 1995) in which transcription is driven by the CMV promoter. Constructs containing the wild type SR splicing factors were cloned and provided by G. Screaton (Nuffield Dept of Clin. Med., Radcliffe Hosp., Headinton, Oxford).
Alternative splicing of LCA-2 was analysed as previously described (chapter 5), using the LCA-2 minigene construct which was cotransfected with PCG-SRp vector (molar ratio of 1:5) by lipofection into COS-7 cells (figure 6.2). The LCA-2 minigene alone, or together with an empty pCG vector showed two mRNA species representing the two transcripts o f CD45 including or excluding exon A, with almost equal intensities. The majority of the splicing factors which were cotransfected with the CD45 minigene caused a weak shift in the pattern of splicing slightly favouring either the RO or the RA form; no gross differences in alternative minigene splicing were detected, with one exception (figure 6.3). Co-expression of SRp30c caused a dramatic increase in splicing to the CD45RO form two days after transfection by almost 10 fold compared to the control (empty pCG-vector), when quantified by densitometry (figure 6.4). This activity was consistently observed by three different constructs used (pCG, pCG-GlO, and pCI) (data not shown).
Dose dependent promotion o f CD45RO splicing by SRp30c
To test whether the effect of SRp30c was dose dependent, different amounts of plasmid were titrated with the empty vector (PCG), to give a constant amount of 5 pg DNA and transfected into COS cells together with a constant amount of 1 pg LCA-2 (titration: 0.1 pg, 1.0 pg, 2.5 pg, 4.0 pg, 4.9 pg, 5.0 pg SRp30c). The more SRp 30c was transfected the greater was the shift from CD45RA to CD45RO spliced forms, establishing that this effect is concentration dependent as quantitated by densitometry (figure 6.4).
RT-PCR using primers specific for the cloned SRp30c transcripts but not the endogenuous SRp30c mRNA was performed. Figure 6.5 shows that the band intensity of SRp30c relative to
the intensity of the HPRT control correlates well with the amount o f SRp 30c vector transfected into COS cells, indicating that the shift in CD45 minigene splicing is dose dependent on the level of SRp30c.
Effects on the alternative splicing pattern of pSEC-LCAl-7
As mentioned before, the LCA-2 minigene construct contains exon A only and flanking intron sequences, but the variable exons B and C are missing. If the solo exon model discussed in chapter 6 does not apply to the alternative exons of CD45 (as it was suggested in chapter 5), it seemed likely that factors controlling other variable exons may influence the splicing pattern of exon A by communication through SR domains. The effect of SRp30c on exon A splicing in LCA-2 may therefore be limited to this minigene and would not show any activity in the control of alternative splicing of the natural CD45 pre-mRNA.
To examine this possibility the pSEC -LC A l-7 construct (figure 5.1) was employed. To quantitate/relate the splicing phenomena to the level of the synthesized splice factors, their cDNA sequences were subcloned into a modified pCG vector containing a T7 tag (referred to as pCG-GlO). cDNA inserts were SRp30a (ASF/SF2), SRp30b (SC35), SRp30c, SRp30d (9G8), SRp40, SRp55, SRp75, hnRNP A1 (G. Screaton, Nuffield Dept, of Clinical Medicine, Radcliffe Hospital, Headington, Oxford). Transient transfections with these plasmids will produce the modified splicing factors as fusion proteins containing the first 11 residues o f the T7 gene 10 capsid protein in the N-terminus. The recombinant proteins can be detected using an anti-T7-tag mAb (Caceres et al., 1997).
The pSEC-LCA l-7 was cotransfected with splicing factor constructs as shown in experiments with LCA-2. SRp30c induced a strong shift towards low molecular weight isoforms of CD45. In addition, most of the two RRM domain containing splicing factors had reproducible effects in changing the CD45 splicing pattern, including SRp30a, SRp40, SRp55 and hnRNP A1 (data not shown, except for SRp30a, SRp30b and SRp30c in figure 6.6). However none of these effects were as prominent as observed with SRp30c, which shifted the splicing almost completely to the CD45RO isoforms. This shift in splicing which appears to proceed from ABC (and BC) through the intermediate exon combinations BC and B to O, in which all alternative exons have been skipped, was dependent on the concentration of SRp30c. In contrast, splicing to the 0-fo rm was inhibited when antisense RNA was expressed in COS cells after transfection of a CDM8-vector (Invitrogen) containing a reversely inserted cDNA of SRp30c (figure 6.6).
This data is in agreement with our data obtained with LCA-2, showing that the effect is dose dependent. In contrast to the results with LCA-2 though, splicing factors SRp30b (SC35), SRp20 and SRp30d (9G8), which contained only one RRM domain, shifted CD45 minigene
splicing into the other direction towards CD45RABC isoforms, although w ith different efficiency. SRpSOb has no significant effect, whereas the SRpBOd activity is strong (not shown), but not as dramatic as for SRp20, which induced striking shifts at very low concentrations (2-3 )Lig) (figure 6.6). As for SRpBOc which promoted exon exclusion, SRp20
promoted exon inclusion in a dose dependent manner as similar titration experiments have shown (G. ten Dam, Netherlands, unpublished). In conclusion there appears to be a trend towards exon exclusion for SR proteins containing two RRMs (although with different activities, the most prominent being SRpBOc) and towards exon inclusion by SRps containing one RRM (with the strongest activity given by SRp20).
Protein expression o f SRps in COS cells
In order to demonstrate that the plasmids were able to produce protein, cell lysates of transfected COS cells were analysed by SDS-PAGE and W estern blotting using anti T7 tag -antibody. All constructs tested produced proteins, but most SR proteins showed a doublett in w hich the slightly increased m olecular w eight is probably due to the extensive phosphorylation of serine residues within the RS region (Fu and Maniatis, 1992; Ge et al., 1991; Gui et al., 1994; Krainer et al., 1991; Zahler et al., 1992). Although different intensities of protein expression were detected, the changes in splicing of CD45 were unlikely to be due to differences in expression levels as all proteins were expressed at comparable levels and experim ents testing dose effects were carried out for each of the splicing factors (in collaboration with G. ten Dam, data not shown). In particular, the expression levels of SRpBOc and SRpBOa are shown in figure 6.10, and are of almost identical magnitude.
Theoretical considerations for SRp30c/30a chimeric domain swap- and SRp30c deletion- mutants
Since individual members o f the family of SR proteins of splicing factors display different effects on CD45 minigene splicing. We then wanted to investigate the activity contributed by individual domains of SR proteins in alternative CD45 splicing using SR proteins, SRpBOa (which has no gross effects on CD45 minigene splicing) and SRpBOc (which shows a dramatic effect in CD45 minigene splicing), as models. Splicing factor SRpBOc is most closely related to SRpBOa in primary structure with an overall amino acid identity of 74% (Screaton et al., 1995). Between these two factors, RNP-2 of RRM-1 differ in 6 amino acids o f 32, RNP-1 in 9 amino acids of 39, the central RRM-2 differs in 12 amino acids o f 71 amino acids (see also figure 1.4). This atypical RRM-2 however shares little homology with other splicing factors generally, whereas RRM-1 contains the invariant signature SWQDLKD which is shared by all members of the SR protein family (Birney et al., 1993). The SR domain is the least homologous region between SRpBOa and SRpBOc, which contains an unusually
short RS domain. Despite the extensive homology between SRpBOa and SRpBOc, only SRpBOc appears to regulate CD45 exon A splicing as presented earlier.
The cloning strategy to create all possible combinations of domain swap-mutants (figure 6.7) was to PCR amplify individual domains of SRpBOc using specific primers containing unique restriction sites at their 5' ends. In order to clone the digested PCR fragments into the appropriate positions of SRpBOa to replace individual domains. Two unique restrictions sites within SRpBOa separate the three individual domains, R RM l and RRM2 by S a d , and RRM2 and RS by Apal, were used for the cloning strategy. Xbal (5') and BamHl(B') sites flanking the cDNA were previously created in order to clone the SRpBOa fragment into an pCG-GlO expression vector. All cloning steps were perform ed in the TAg vector into which the constructs were subcloned and subsequently transferred into the expression vector pCG-GlO. In SRpBOa-TAg (see figure 6.1 and 6.7) R RM l can be released by restriction digestion with X bal/SacI, RRM2 by digestion with Sacl/A pal and the RS dom ain by digestion with Apal/BamHI.
Pfu-PCR of single and multiple domains of SRpBOc was performed with primers containing the appropriate restriction sites for cloning. Different swap mutants are designated by a three letter code for the sake of simplification: The first letter representing R R M l, the second letter representing RRM2 and the third letter representing the RS domain of SRpBOa (= FiF2Fr s) and SRpBOc (= CiC 2Cr s). All the theoretically possible com binations are as follows:
C1C2CRS, C1F2FRS, C1C2FRS, C1F2CRS, F1C2FRS, F1C2CRS, F1F2CRS, F1F2FRS.
We failed to clone the C1F2CRS combination. The deletion mutants cloned were C1C2ARS
and C1A2CRS, the A1C2CRS version was not generated (figure 6.8).
C1F2FRS was cloned by amplification of C1C2C RS (SRpBOc) using FBOcXba/RBOcSac
prim ers, C1C2FRS by using FBOcXba/RBOcApa p rim ers, F1C2C R S by u sin g
FBOcSac/RBOcBam primers, F1C2FRS by using FBOcSac/RBOcApa primers and F1F2CRS by
using FBOcApa/RBOcBam prim ers, C1A2CRS by using FB0cXba/RB0cA2Apa prim ers,
C1C2ARS by using FBOcXba/RBOcARSBam primers (for oligonucleotide primer sequences,
see chapter 2, illustration of the cloning strategy in figure 6.7). These PCR fragments and the TA g-FiF2FR S vector (SRpBOa) were digested with the appropriate restriction enzymes separated on a Agarose gel, excised and purified using the Qiaex PCR purification system. The fragments were ligated and miniprep plasmid DNA of individual bacterial Ampicillin resistant clones was checked for inserts by restriction digestion and PCR, using appropriate restriction enzymes and oligonucleotides. Positive clones were subsequently digested with Xbal/BamHI to release the complete splicing factor swap mutant and this was ligated into the Xbal/BamHI sites of mammalian expression vector pCG-GlO. Positive bacterial clones were identified by restriction digestion of miniprep plasmid DNA and PCR with appropriate
primers. Inserts were sequenced in both orientations using the chain termination technique with Sequenase (USB) and internal sequencing primers. In addition, all clones were tested for protein expression in COS cells by lipofection and W estern blotting o f COS lysates using T7 tag mAb (figure 6.10). Transfections into COS-7 and RT-PCR to detect variably spliced forms of the CD45 minigenes were as previously described.
The atypical RRM o f SRp30c is responsible for the specific interaction with exon A
To examine whether the specific splicing activity o f SRpBOc with CD45 minigenes was defined by one of its two RRMs or by its RS domain, we first exchanged the unusually short RS domain of SRpBOc with the corresponding RS domain of SRpBOa, to produce C1C2FRS and F1F2CRS. After transfection into COS cells the expression o f the T7 tag chimeric proteins as tested by W estern blotting to ensure that the proteins with the correct size and comparable high levels were expressed. Only the C1C2FRS protein was expressed at lower
levels than all the other fusion proteins tested.
Unfortunately, fusion protein F1F2CRS was found to be of too large size, despite the correct
sequence upstream of the 5' stop codon, but this has not been checked. Therefore the unexpected size is likely to be due to a defect stop codon, which would result in the use of the next downstream stop codon. Although most of the sequence is correct, and the fusion protein is expressed, the results obtained have to be interpreted with caution. W hen compared to wild type SRpBOa (F1F2FRS), no change in the alternative splicing activity was observed for F1F2CRS, but this could be due to defective function as already mentioned.
Similarly, C1C2FRS produced the CD45RO isoforms as the wild type SRpBOc (C1C2CRS)
construct, suggesting that the RS domains of these SR proteins do not contribute to CD45 alternative splicing activity. This result conforms with earlier data showing that RS swapping between SRpBOa and SRpBOb had no functional effect (Chandler et al., 1997), as was first reported for Drosophila factors Tra and SWAP (Li and Bingham, 1991).
It has been shown that individual RRMs of splicing factors represent functional domains that can either act independently or cooperatively in binding to RNA (Bentley and Keene, 1991; Jamison et al., 1995; Lutz and Alwine, 1994). To investigate which RRM domain of SRpBOc was mediating the interaction (directly or indirectly) with exon A o f CD45 and whether the activity was confined to one individual domain, or both domains acting together, each RRM in SRpBOc was substituted by the corresponding domain of SRpBOa and vice versa.
Interestingly, all domain swap chimeras containing the central atypical RRM of SRpBOc (F1C2CRS, F1C2FRS, C1C2FR S) prom ote splicing to C D 45RO w ith com parable
( F iF 2 C r s , C1F2FRS, no C1F2CRS was cloned) showed no strong ability to change the
alternative splicing pattern of CD45 (figure 6.9). These results are summarised in figure 6.8. To test whether the CD45 splicing activity was abrogated when single domains were absent, the deletion plasmids C1C2ARS, lacking the RS domain, and C1A2CRS lacking the central
RRM, were analysed. As expected, the mutant C1A2CRS protein showed no strong splicing
activity to exclude variable exons from CD45 pre-mRNA. The capability to exclude variable