ABSTRACT
The fibroblast growth factor receptor-2 gene contains a pair of alternative exons, K-SAM and BEK, which are spliced in a cell type specific manner. We
have shown previously that a 10 nucleotide sequence within the K-SAM exon exerts a negative effect on K-SAM exon splicing independent of cell type. We demonstrate here that this sequence works autonomously, as it can repress splicing of
a heterologous exon, the EIIIb alternative exon of the rat fibronectin gene. By introducing point mutations into the 10 nucleotide sequence, we have shown that the functional portion is limited to 4 nucleotides, TAGG,
the dinucleotide AG of which is particularly important. This short sequence may
participate in the control of splicing of exons carrying it, provided that they
carry weak splice sites.
Different mRNAs can be obtained from a single pre-mRNA by alternative splicing (
1
,
2
). Several protein isoforms can thus be obtained from a single gene. Their
production can be modulated in a tissue-specific manner by controlling splicing. This can be an important
mechanism for controlling gene expression during development. For example, the fibroblast growth factor receptor-2 (FGFR-2) gene contains a pair of mutually exclusive exons, K-SAM (or IIIb) and BEK (or IIIc), whose splicing is controlled
in this way (
3
-
6
). Epithelial cells splice the K-SAM exon with no detectable splicing of the BEK exon, while some other
cell types make the opposite choice (
7
). The subset of fibroblast growth factor (FGF) family members (there are nine
known to date) recognized by the resulting receptor depends on this splicing
choice (
8
,
9
). Making the right choice is thus important to ensure an appropriate response
to the growth factors available in a cell's environment. Indeed, an
unprogrammed change in the splicing choice accompanies (and may participate in)
malignant transformation in some cases (
10
).
Splicing is a multi-step process involving many proteins and small nuclear ribonucleoproteins
(snRNPs), which assemble on the pre-mRNA to form spliceosomes (
11
). The exon definition model (
12
) proposes that an individual exon is recognized by the coordinated binding of
splicing factors to its 3' splice site (for example U2AF and U2 snRNP) and 5' splice site (for example ASF/SF2 and U1 snRNP). Clearly any
molecule which helps or hinders these binding events could control the
efficiency of exon splicing. Thus in drosophila sex determination, the female-specific protein sxl binds to the polypyrimidine tract of a male-specific 3' splice site on the tra pre-mRNA. This prevents binding of U2AF, and use of the
male-specific splice site (
13
). Subsequently, in females, a complex composed of tra, tra-2 and several proteins of the SR family binds to a repeated exonic
sequence on the dsx pre-mRNA to activate use of a female-specific 3' splice site (
14
).
In mammalian systems, purine-rich exon splicing enhancers have been described (
15
-
18
) which function by binding SR proteins. These proteins presumably then help to
stabilize splicing factor binding to the exon's splice sites. Intron sequences
(often repeats of a short sequence motif) have also been reported to control
splicing of nearby exons, although their mechanism of action remains to be determined (
19
-
23
). A few exon sequences which repress splicing have been described (
24
-
31
), although the mechanism involved is only clear for a very few cases. An
inhibitory sequence close to the 3' splice site of a beta-tropomyosin alternative exon participates in a secondary structure
which represses use of the splice site (
27
,
28
). Another exon sequence represses splicing of a drosophila P element intron in
somatic cells. Here, a protein complex binds together with U1 snRNP to pseudo-5' splice sites within the exon and stops U1 snRNP binding to the
bona fide
5' splice site (
29
,
30
).
We have shown previously that the K-SAM exon of the FGFR-2 gene contains a sequence 5'-TAGGGCAGGC-3' which represses splicing of the exon (
31
). Deleting this sequence allows efficient splicing of the K-SAM exon in HeLa cells, which normally splice the BEK exon. The sequence
also has a negative effect on K-SAM exon splicing in SVK14 cells, which normally splice this exon. In this
case the negative effect appears to be out-weighed by positive effects transmitted by downstream intron sequences.
Given the dearth of information available concerning negatively acting exon
sequences, we decided to undertake further analysis of the K-SAM sequence.
Minigenes RK3 and RK12 have been described elsewhere (
31
). Minigenes S10+8nt and S10+50nt were made by ligation of double-stranded oligonucleotides carrying the S10 sequence (5'-TAGGGCAGGC-3') with appropriate overhangs into the
Ava
I and
Eco
RI sites respectively of RK3CAT. Minigene S10+97nt was made from RK12 by
replacing the
Eco
RV-
Sal
I CAT sequence by the sequence 5'-GATATCGGATAGTG
TAGGGCAGGC
ATCGAT-3' in double-strand form in which the S10 sequence (bold type) is flanked
by
Eco
RV and
Cla
I sites. Minigenes containing mutated versions of the S10 sequence were made
from minigene S10+97nt by replacing the
Eco
RV-
Cla
I fragment with mutated versions using synthetic oligonucleotides. Fibronectin
minigenes containing the S10 sequence were made by ligating oligonucleotides
containing the S10 sequence with
Bam
HI overhangs into the
Bam
HI site of the EIIIb alternative exon of plasmid [Delta]GATP3 (
23
).
Cells and transfection of cells, harvesting of RNA and RT-PCR analysis were as described previously (
31
,
32
). Care was taken to remain within the exponential range of amplification (20 cycles or less of
amplification were routinely used; however, the results obtained remained
essentially unchanged if 30 cycles of amplification were carried out). In
consequence, Southern blotting (
31
,
32
) was needed to visualize the PCR products. As demonstrated previously (
31
), RT-PCR results in our system correlate well with results obtained by
Northern blotting. For fibronectin minigenes, PCR primers were as follows: 5'-ATGCCGATCAGAGTTCCTGC-3' and 5'-GGCGGTGACATCAGAAGAATCAAAA-3'.
We have described previously the RK3 minigene in which the alternative exons K-SAM and BEK together with flanking exons C1 and C2 lie between the Rous sarcoma virus (RSV) long terminal repeat
and a bovine growth hormone (BGH) polyadenylation signal (
31
). To study splicing, this minigene is transfected into cells and an RT-PCR protocol with primers P1 in the C1 exon and P2 in the BGH sequence is
used to analyse RNA transcribed from the minigene. In principle, three
amplified fragments could be obtained: a 0.46 kb fragment with a
Hpa
I site (BEK spliced to C1 and C2); a 0.46 kb fragment with an
Ava
I site (SAM spliced to C1 and C2); and a 0.61 kb fragment with both
Ava
I and
Hpa
I sites (SAM spliced to BEK and C1, BEK spliced to SAM and C2). These fragments
can be distinguished by restriction enzyme mapping, and this allows
determination of which exons have been spliced. In particular, generation of a
0.20 kb
Ava
I fragment is associated with splicing of the K-SAM exon.
As described previously (
31
,
32
), the BEK exon is spliced when the RK3 minigene is transfected into HeLa cells,
but when the bulk of the K-SAM exon internal sequences are replaced by bacterial CAT sequences of the
same length (Fig.
1
A), splicing of the K-SAM/CAT exon (which we shall henceforth refer to as the CAT exon) is
activated. The RT-PCR protocol (Fig.
1
B) now yields a mixture of 0.46 kb fragments (of which about half have a
Hpa
I site and half an
Ava
I site, and which correspond to BEK and CAT transcripts respectively, Fig.
1
C) and 0.61 kb fragments with both
Ava
I and
Hpa
I sites. The latter correspond to CATBEK transcripts (Fig.
1
C). Reintroduction of the K-SAM exon S10 sequence (5'-TAGGGCAGGC-3') into the CAT exon so that its location
relative to the 3' and 5' splice sites is the same as in the K-SAM exon [97 nucleotides (nt) downstream from the 3' splice site, S10+97 nt, Fig.
1
B] is sufficient to repress markedly splicing of the CAT exon. The RT-PCR protocol detects a 0.46 kb fragment with a
Hpa
I site (Fig.
1
B), reflecting splicing of the BEK exon (Fig.
1
C).
To determine which part of the sequence 5'-TAGGGCAGGC-3' is important in repressing splicing, short blocks of
the sequence (TAG, GGC or AGGC) were mutated in the S10+97 nt minigene as
indicated in Figure
3
, before transfection of mutated minigenes into HeLa cells. Mutation of the TAG
or GGC blocks lifts repression of CAT exon splicing (presence of 0.61 kb CATBEK
fragments and 0.46 kb CAT fragments with
Ava
I sites). The experiment where the AGGC block is mutated identifies 5'-TAGGGC-3' as sufficient to repress splicing (absence of the 0.61 kb CATBEK fragment, presence of 0.46 kb BEK
fragments with
Hpa
I but not
Ava
I sites). Individual nucleotides of the 5'-TAGGGC-3' sequence were next mutated (we used the minigene
where the AGGC block was already mutated to CTTG), and the effect of these
mutations on splicing tested. As shown in Figure
3
, only mutations which change the nucleotides 5'-TAGG-3' of the sequence 5'-TAGGGC-3' abolish repression of CAT
exon splicing.
The sequence TAGG contains a stop codon, and also represents part of the 3' splice site consensus YAGG. Several cases have been described where in-frame stop codons reduce the efficiency of exon splicing (
33
and references therein). The TAG of interest here is not an in-frame stop codon for the K-SAM exon. However, as nothing is known about the mechanism which
allows stop codons to influence splicing, we decided to test whether the
sequence TAGG represses splicing because it contains a stop codon. Minigenes
based on S10+97 nt were made in which the TAGG sequence was mutated to TAAG or
TGAG, and they were transfected into HeLa cells (Fig.
4
). In neither case was CAT exon splicing repressed (we detected 0.61 kb CATBEK
fragments and 0.46 kb CAT fragments with
Ava
I sites in both cases). We conclude that the TAG in our sequence does not
function exclusively as a stop codon.
Does the exonic TAGG sequence work as a decoy 3' splice site, turning the attention of the splicing apparatus away from
the true 3' splice site to a non-functional one? It is a better fit to the consensus than the K-SAM exon's 3' splice site, uagC (intron sequence in lower case
letters, exon sequence in upper case letters), and could represent a good
competitor. Indeed, when the K-SAM exon's 3' splice site is mutated to TAGG, the exon sequence TAGG is no
longer capable of repressing splicing of the K-SAM exon in HeLa cells (Fig.
4
; the RT-PCR analysis detects 0.61 kb CATBEK fragments and 0.46 kb CAT fragments
with
Ava
I sites). However, if the TAGG sequence functions simply as a decoy, one would
imagine that the sequence CAGG would work at least as well. As shown in Figure
4
, this is not the case. If the TAGG sequence competes with the K-SAM exon's 3' splice site, it must be for a factor which prefers the sequence
TAGG.
We showed previously that the K-SAM exon S10 sequence TAGGGCAGGC represses splicing of this exon, and also
of an artificial CAT exon joined to K-SAM splice sites (
31
). The question remained open as to whether this sequence works exclusively in
the context of a K-SAM exon (perhaps reflecting particular steric constraints unique to this
exon), or whether it functions autonomously, and represents a phenomenon of
more general interest. We show here that the S10 sequence can function to
repress splicing of a rat fibronectin gene alternative exon, and thus appears
to act autonomously. We also show that the functional part of the S10 sequence
is limited to the first four residues, TAGG, of which the AG dinucleotide is of
particular importance.
That such a short sequence can repress splicing may seem surprising
a priori
, in as far as it is present in many constitutively spliced exons. However,
mutations which bring the K-SAM exon's 3' or 5' splice sites closer to the corresponding consensus
sequences render the TAGG sequence incapable of repressing splicing of the
exon, and introducing the TAGG sequence into a constitutively spliced globin
gene exon has no effect on splicing of the exon (our unpublished results). Thus
the TAGG sequence is only capable of repressing splicing of exons associated
with weak splice sites. This is the case for the rat fibronectin gene EIIIb
exon (
23
), which is inefficiently spliced in many cell types. Indeed, exons undergoing
alternative splicing are frequently associated with weak splice sites, and so
the sequence TAGG may participate in the control of splicing of many
alternative exons which carry it.
Some other alternatively spliced exons are associated with exon inhibitory
sequences which contain motifs related to the TAGG sequence. The human
fibronectin EDA exon contains the sequence CAAGG (note the AG dinucleotide),
deletion of which leads to constitutive splicing of the otherwise partially spliced exon (
25
). Exon splicing silencers within human immunodeficiency virus type 1 tat exon 2 and tat-rev exon 3 have been narrowed down to a 20 nt stretch which for the tat
exon 2 has the sequence
Could the binding of mammalian hnRNP A1 (or a similar protein with overlapping
binding specificity) to the TAGG motif be involved in repressing K-SAM exon splicing? If mammalian hnRNP A1 binds with high affinity to the
sequence UAGGGA/U, it also binds to related sequences such as the globin gene 5' and 3' splice sites with lower affinity (
34
). Several experiments have suggested a link between high levels of hnRNP A1 and
exon skipping (
35
-
37
). As with the TAGG motif, hnRNP A1 provokes exon skipping most effectively with
short exons or exons which have weak splice sites (
36
). Perhaps binding of hnRNP A1 to the UAGG motif recruits other proteins or RNPs
which hinder splicing by impeding binding of splicing factors to splice sites,
or by blocking the dialogue between 5' and 3' splice sites needed to define an exon. This could suffice to
repress splicing of an exon with weak splice sites.
The TAGG sequence contains a stop codon (out of frame for the K-SAM exon), but our results show that it is the sequence TAG rather than
the sequence of a stop codon which is functionally important. Nevertheless, it
is interesting to note that mutations which create in-frame stop codons can lead to exon skipping and be involved in human
disease. For example, a nonsense mutation (TAT G to TAG G) leads to skipping of
exon 51 of the fibrillin gene in cases of Marfan syndrome (
38
). Mutating
in vitro
TAT G to TAA G or TGA G also leads to exon skipping. However, while moving the
TAA and TGA stop codons out of frame by adding two base pairs to exon 50
restores completely splicing of exon 51, mutations which place the TAG stop
codon out of frame, but still in the context TAGG, only partially restore exon
51 splicing. This is consistent with the notion that the sequence TAGG exerts a
negative effect on splicing. It also suggests that mutations which create in-frame stop codons of the type TAG G will be particularly effective in
provoking exon skipping, cumulating the effects of a stop codon with those
linked to the creation of a splicing repression sequence.
We thank Dr R. Hynes for a kind gift of fibronectin minigenes. This work was
supported by grants from the Association pour la Recherche sur le Cancer, and
the Ligue Nationale contre le Cancer, Comité Departemental de Loire-Atlantique.
REFERENCES
Return