ABSTRACT
Circular splicing has already been described on nuclear pre-mRNA for certain splice sites far apart in the multi exonic
ETS-1
gene
and in the single 1.2 kb exon of the Sry locus. To date, it is unclear how
splice site juxtaposition occurs in normal and circular splicing. The splice site selection of an internal exon is likely to involve pairing
between splice sites across that exon. Based on this, we predict that, albeit
at low frequency, internal exons yield circular RNA by splicing as an error-prone mechanism of exon juxtaposition or, perhaps more interestingly, as a
regulated mechanism on alternative exons. To address this question, the circular exon formation was analysed at three
ETS-1
internal exons (one alternative spliced exon and two constitutive), in human
cell line and blood cell samples. Here, we show by RT-PCR and sequencing that
exon circular splicing occurs at the three individual exons that we examined. RNase protection experiments suggest that there is no
correlation between exon circle expression and exon skipping.
The c-ets-1 proto-oncogene belongs to the Ets gene family which encodes DNA-binding transcription factor proteins. Most Ets family
members act either alone or in synergy with various cofactors to regulate
numerous viral and cellular promoters (
1
). The human
c-ets
-1 locus encodes a major 6.4 kb transcript and alternatively spliced
transcripts (
2
,
3
). In addition, we previously detected by PCR a peculiar splicing reaction which
leads to additional transcripts. Indeed, we showed that the donor splice site
of two exons of the human
ets-1
pre-mRNA can splice to the acceptor splice site of an upstream exon (
4
,
5
). This reaction, originally called mis-splicing (now called circular splicing) yields circular RNA molecules
containing only exon sequences which are localised in the cytoplasm of cells.
Circular transcript formation has also been described, for the Sry transcript
in mouse testis, due to splice sites used in a 3' -> 5' direction
(
6
). In addition, it is likely that the scrambled exons described in the DCC gene
(deleted in colorectal carcinomas) (
7
), also correspond to circular transcript formation but this has not been
proved. Due to loss of ORF or the absence of ribosome attachment, these
circular RNAs are likely to be non-coding. However, eukaryotic ribosomes can initiate translation on circular
RNAs but only if the RNAs contain internal ribosome entry site elements (
8
). The biological function of these natural circular RNAs is to date unknown but
they may act as functional RNA as suggested for non-coding RNAs like XIST and H19 (
9
,
10
) or the 3' untranslated region of some mRNA (
11
).
To date, it is unclear how splice site juxtaposition occurs in normal and
circular splicing and studies of circular splicing may help to elucidate the
mechanism of normal exon juxtaposition.
ETS-1
circular splicing occurs at splice sites close to large introns (
5
). This suggest a leaky mechanism for the orderliness in splicing which occurs
more frequently for certain splice sites. It is also noteworthy that, in
addition to the proximity of large introns, the
ETS-1
circular splicing occurs efficiently with an alternative exon. Similarly, most
of the introns of the DCC gene which also yields scrambled exons are large and
some exons are also alternatively spliced (
12
). Concerning Sry, the mono-exonic structure of the gene and the inverted repeats in both ends of the
transcript are probably involved in the circular transcript formation (
6
). The suggested mechanism is similar to the intramolecular base pairing
described in the intron of the gene that encodes the 40S ribosomal subunit,
RP51. The intron complementarity has been shown to influence positively the
splice site selection and the splicing efficiency (
13
). Recently,
in vitro
spliceosome formation on a circular yeast pre-mRNA has been described and the products of the resulting splicing
reaction are the lariat-shaped intron and a mature circular mRNA (
14
). On the other hand, autocatalytic introns may also yield
in vitro
exon circle RNA with an appropriate substrate in which
the splice sites flanking a group I or group II intron are inserted at both ends
of an exon (
15
,
16
). In this case, the RNA structure of the autocatalytic intron is obviously
involved in the splice site juxtaposition for circular splicing.
Except for the intronless Sry gene, circular splicing involves an accurate
selection of both the splice sites and the exon, as in normal mRNA maturation. The splice site selection, in particular of an internal exon,
is likely to involve pairing between splices sites across the exon. Indeed, the
internal exon which is rather small (
17
,
18
), seems to be the unit for splice site recognition (
19
), but the mechanism that ensures exon juxtaposition across the intron is
unclear and this problem is even more complex in the case of transcripts
containing alternative exons. Based on this, we predict that internal exons
probably yield exon circles at a low frequency as a result of an error-prone mechanism of exon juxtaposition.
Here, we test the prediction that splicing may occur with the donor and acceptor
splice sites from an exon, producing excised exon circles and ask whether this
reaction may participate in an alternative splicing mechanism by deleting exons
from nuclear pre-mRNA or, alternatively, by splicing of an alternative exon in a lariat-shaped intron. The
ETS-1
locus is a good model to answer these questions because two exons skip during
mRNA maturation at different frequencies (
3
,
20
) and the circular splicing involving multiple exons is already known (
4
). In this present work, we show that sets of oligoprimers in opposite
directions in an exon can amplify, products from poly(A)
-
RNA by RT-PCR and that the sequence of the PCR products confirms exon circular
splicing. RNase protection experiments have been performed to quantify the
level of exon circle expression. Although the conventional
ETS-1
RNA has the potential to fully protect the antisense probe from the circular
RNA, probably via a loop-out artefact, the expression data from the exon which skips at high
frequency in this cell line shows that the signal corresponding to fully
protected probe is too low, relative to the alternative spliced transcript, to
correspond to a secondary product of the alternative splicing of this exon or
splicing event occurring in the excised lariat containing the alternative exon.
The existence of the exon circle for constitutive and alternative exons and
RNase protection experiments suggest that the exon circle splicing is
independent of alternative splicing processes.
Total RNA was prepared from cell lines using the guanidinium thiocyanate-caesium chloride method as previously described (
28
). Poly(A)
+
and poly(A)
-
RNA was isolated from total RNA using double passage over oligo(dT)-cellulose chromatography. Poly(A)
-
RNA was separated by sedimentation through a 10-20% sucrose linear gradient in 5 mM Tris-HCl pH 7.4, 2.5 mM EDTA, 0.5% SDS to collect the top fractions containing RNA
smaller in size than 18S as determined by agarose gel electrophoresis. This RNA
(1.5 [mu]g) was reverse transcribed in 10 [mu]l containing 200 ng sense primer, 1 [mu]g BSA, 1.5 mM each dNTP, 50 mM Tris-HCl pH 8.3, 75 mM KCl, 10 mM DTT, 3 mM MgCl
2
, 100 U MuMLV reverse transcriptase (Gibco-BRL) for 1 h at 37oC. Reverse transcription mixture was included in a final volume (100 [mu]l) of PCR reaction containing 100 mM Tris-HCl pH 8.3, 50 mM KCl, 1.5 mM MgCl
2
, 1 U
Taq
polymerase (Stratagene) and 200 ng of antisense primer. Amplification was
performed by 30 cycles; 92oC (1 min), 60oC (1 min) and 72oC (1 min) in a techne PHC-2 thermocycler. Aliquots of the PCR reaction (10 [mu]l) were electrophoresed on a 2% agarose-TBE gel. The
sequences of the primers are indicated in Figure
1
by < and > signs. The amplification products were cloned into the
Eco
RV site of pBluescript II SK- vector (Stratagene) and sequenced by the dideoxynucleotide method.
PCR products, as shown in Figure
1
c, were inserted into the
Eco
RV site of pBluescript SK- vector. The clones which by sequencing did not show any point mutations
were chosen to perform the RNase protection experiments which were performed,
as previously described (
5
), using total RNA from human CEM cell line. The plasmids were linearized with a
restriction enzyme and used as a template for T3 or T7 RNA polymerase
(Stratagene). Total RNA (50 [mu]g) was hybridized to 10
5
c.p.m. of
32
P-labelled uridine probe overnight at 45oC in 20 [mu]l of hybridization buffer (80% formamide, 400 mM NaCl, 40 mM
PIPES pH 6.4, 1 mM EDTA). After RNase A and T1 digestion (10 [mu]g/ml and 130 U/ml, respectively) in 300 [mu]l of digestion buffer (10 mM Tris pH 7.6, 5 mM EDTA, 0.3 M NaAc pH 7.0) at 37oC for 30 min, purified and precipitated samples were analysed by
electrophoresis on a 6% acrylamide-8 M urea gel and exposed to Cronex X-ray film in the presence of intensifying screens for 2 days.
To investigate circular exon formation, we designed oligonucleotide primers (see
Fig.
1
a and b) for three internal
ETS-1
exons which have particular properties in terms of circular or alternative splicing: the acceptor splice site of a1 exon and donor splice sites of c and d
exons are involved in multi-exonic circular splicing (
4
), and the d exon skips during mRNA maturation at high frequency (
2
,
3
). To circumvent problems associated with amplification of minor transcripts, in
particular non-specific amplifications from the highly expressed
ETS-1
mRNA, we used the top fraction from an RNA sucrose gradient, loaded with poly(A)
-
RNA from CEM cells. This fraction contains RNA molecules smaller than the 18S
as determined by agarose gel electrophoresis (data not shown). From this RNA,
ethidium bromide stained agarose gels reveal RT-PCR amplified products of 62,
63 and 104 bp during the experiments to detect circular splicing of a1, c and d
exons, respectively (Fig.
1
c). Control experiments minus reverse transcriptase do not reveal these
amplified products (data not shown). The sizes of the PCR products were as
expected for circular splicing of these exons according to the primer position
and the exon size. To confirm this observation, the PCR products were cloned
and sequenced (Fig.
1
d and e). The sequences confirm the correct amplification of the primers on the
ETS-1
transcripts and reveal the joining of exon ends at the splice site used in the
normal maturation for each of the exons examined. For example, the sequence of
the 62 bp amplified product shows that the 3' nucleotides of the a1 exon, i.e. nAAAG, are followed by the 5' nucleotides of the same exon, i.e
.
ATATn (Fig.
1
d and e). The splice site sequences of each exon are already described (
3
) and are indicated in Figure
1
b. The expression of this RNA was not restricted to the CEM cells. Indeed, these
RT-PCR products were also observed with poly(A)
-
RNA from a range of cells which expressed high levels of
ETS-1
transcript like Jurkat, HSB2 and MDA cell lines but also from human blood cell
samples (data not shown).
RNase protection experiments with
32
P-labelled uridine antisense riboprobes covering the circular splicing
junctions were performed in order to determine the expression level of circular
exons. Unfortunately, full protected probe was observed after hybridization to
an
in vitro
synthesised conventional sense RNA (data not shown) as already described for S1
nuclease mapping (
21
). However, the RNase protection experiments from the alternative d exon is informative. From total RNA of CEM cells, d exon circle expression could be
estimated as a maximum limit of expression from the RNase protection
experiments. Indeed, using the conventional 6.4 kb
ETS-1
transcript as an internal control in the RNase protection experiments, the
signal corresponding to full protected probe will correspond to the exon circle
expression and potential loop-out artefact with the linear transcript. Figure
2
shows the RNase protection experiments of d exon from total RNA of CEM cells
and, as expected, both a 105 nt protected fragment corresponding to the fully
protected probe and the 67 nt protected fragment from conventional
ETS1
mRNA are observed. Similarly, the RNase protection experiments from the a1 and c exons revealed the expected
protection products through the circular spliced junction and protection
products corresponding to the linear transcript (Fig.
3
). The 62 and 63 nt protected fragments corresponding to a1 and c exons,
respectively, as circular RNA, are observed as a faint band (Fig.
2
). Both experiments reveal smaller protected fragments around 40 nt, which
correspond to protection of one end of the exons probably by the
ETS1
mRNA. To quantify the signals, the acrylamide gels were transferred to Whatman
paper and the circular and the linear transcript protected products were
32
P counted. The length, the uridine percentage of the protection band the
surrounding background were taken into account when estimating the expression
level. These data show that the protected bands which correspond to the exon
circles and probable protections from the conventional RNA via a loop-out artefact are ~1% of the total
ETS1
expression.
In this paper we have shown that sets of oligoprimers oriented in opposite
directions in an exon are able to amplify products from poly(A)
-
RNA by RT-PCR. These products are likely to have a circular exon structure,
indeed in addition to the sequence data showing the joining of exon ends, these
RNAs are contained in the top fractions from an RNA sucrose gradient loaded
with poly(A)
-
RNA. Spliceosome and splicing reaction are involved in the formation of these
molecules as shown by the accurate selection and use of the splice sites. The
splice sites of these exons do not reveal any difference (Fig.
1
) with the splice site consensus sequence (
22
) that could explain the exon circle formation. Furthermore, the absence of 3' and 5' ends of the circular molecule may participate directly in the stability of these RNAs
and, consequently, to their presence in the cell. The circular transcripts,
involving multiple exons that we have previously identified (
4
), have been detected in the same poly(A)
-
RNA fraction. Direct evidence of the circular structure of the molecule,
produced by splicing of the acceptor sites of the d exon and the a1 exon (
4
), has been obtained by Northern blot experiments on acrylamide gels
(unpublished data). Because of the small size of these exon circles and the low
expression level, they are difficult to visualise by Northern blot. From the
present work, it is not possible to exclude intermolecular splicing, however,
to explain the small size and the absence of poly(A) tail of a linear molecule
containing two similar exons in tandem, additional events like multiple
aberrant splicing has to be imagined. To our knowledge, these data are the
first evidence of exon circles in cells from nuclear pre-mRNA. Circular splicing of Sry in adult mouse testis, where no function
has been determined, differs from the
ETS-1
exon circle formation. In this case the exon is 1231 bp in length and the
splice sites of this intronless gene are not selected in the genital ridge, its
critical site of action (
6
). Recently,
in vitro
circular exon formation has been obtained from circular yeast pre-mRNA, suggesting that free 5' and/or 3' ends of a pre-mRNA are not obligatory for a splicing reaction to
occur (
14
). The group I and group II exon circle has been obtained
in vitro
using modified transcripts which contain elements of the catalytic intron in
both sides of an exon (
15
,
16
). Another stable circular molecule has been described in T cells but this
molecule corresponds to an intron as a lariat structure (
23
).
We have shown, by RNase protection experiments, that the fully protected probe
for the three circular exons correspond to ~1% of the total protected signal in CEM cells. The expression of the circle
exons are probably smaller than this value, because of a possible protection by
the conventional
ETS-1
mRNA, via a loop-out artefact as already described (
21
).
ETS-1
transcripts resulting from alternative splicing of the d exon have been
described (
20
) and it has been noted that the cell lines in culture contain abundant d exon
skipping, far above 1% (
2
). The CEM cell line is one of the cell lines studied (Collyn-d'Hooghe, personal communication). No alternative splicing of the c and a1
exons has been observed during this study (
2
). So, there seems to be no correlation between the exon circle formation and
the alternative splicing of exons, because we observed exon circle formation
for constitutive exons and we observed similar exon circle expression for both
types of exons. Then, the circular exon formation does not occur concomitant to
an exon skipping during the mRNA maturation and this also excludes the
possibility that exon circle formation corresponds to splicing occurring on
circular substrates as, for example, a lariat shaped intron containing an
alternative exon. Indeed, the d exon skipping during the mRNA maturation forms a lariat-shaped intron which contains an intact d exon. We might think that this
secondary product could be the target of circular splicing as observed
in vitro
with a circular yeast pre-mRNA (
14
). In addition, the immature pre-mRNA that undergoes circular exon formation from a constitutive exon is
likely to be degraded because there is no conventional transcript in which the
exon is excluded in the cells.
Circular exon formation is interesting in regard to the two steps in spliceosome
assembly; the splice site selection and the exon juxtaposition. The splice site
selection for exon circle formation is correct as shown by the sequence data
and we may think that the selection probably occurs in the pre-mRNA across an exon as described in the exon definition model (
19
). The components involved in the exon definition model are not yet fully
characterised but the spliceosome particles containing U1 snRNP, U2AF and a
number of distinct protein factors which bind with some degree of selectivity
to the polypyrimidine tract and/or the AG dinucleotide are probably involved in
the exon-bridging interactions (
19
,
24
). The exon definition model does not offer a solution for the following step of
exon juxtaposition across introns. It has been proposed that interactions
between SR proteins bound to one exon with the SR proteins bound to an
adjoining exon could participate in exon juxtaposition. The hnRNPs proteins are
also potential components of this step (
19
). The exon circle formation
in vivo
shows that a correctly recognised exon may at low frequency undergo incorrect
juxtaposition by joining and splicing both splice sites of the exon.
It is noteworthy that the exon circle from
ETS-1
exons are observed by RT-PCR from different cell lines and normal human blood sample. It can be concluded that splice site recognition through the exon
(
19
) must include a mechanism for avoiding or reducing splice site juxtaposition
yielding exon circle formation. It is at present unclear to what extent the
concentration of
trans
-acting factors, like SR proteins and hnRNP proteins, may influence the level of
circular exon formation as described for alternative splicing (
25
-
27
). To date, despite the absence of this observation for other genes and other
exons, it is likely that during intron excision, many, if not all, internal
exons may be circularly spliced at a relatively low frequency by the incorrect
juxtaposition of splice sites. Further studies of the process of circular exon
production may yield insights into the process of exon juxtaposition during
maturation of pre-mRNA in eukaryotic cells.
In conclusion, these results show that, albeit at low level, all examined exons
undergo circular splicing. The nuclear spliceosomal process is error-prone and joins exon splice sites out of their functional order for
protein synthesis. A consequence of circular exon formation is the loss of
mRNA. However, given the accumulating evidence of the existence of a variety of
functional RNA molecules (
9
-
11
), a functional role for circular exonic RNA molecules cannot be excluded.
We thank M. H. Loucheux-Lefebvre for encouragement and support, D. Hétuin for experimental contribution, T. Moore for help on
manuscript preparation and M.-C. Duvieuxbourg for photographic work. This work was supported by the ARC
and the Comité du Nord and du Pas-de-Calais de la Ligue Contre le Cancer and by EC fellowship
(HMC).
REFERENCES