Oriented scanning is the leading mechanism underlying 5' splice site selection in mammals.

Abstract

Splice site selection is a key element of pre-mRNA splicing. Although it is known to involve specific recognition of short consensus sequences by the splicing machinery, the mechanisms by which 5' splice sites are accurately identified remain controversial and incompletely resolved. The human F7 gene contains in its seventh intron (IVS7) a 37-bp VNTR minisatellite whose first element spans the exon7-IVS7 boundary. As a consequence, the IVS7 authentic donor splice site is followed by several cryptic splice sites identical in sequence, referred to as 5' pseudo-sites, which normally remain silent. This region, therefore, provides a remarkable model to decipher the mechanism underlying 5' splice site selection in mammals. We previously suggested a model for splice site selection that, in the presence of consecutive splice consensus sequences, would stimulate exclusively the selection of the most upstream 5' splice site, rather than repressing the 3' following pseudo-sites. In the present study, we provide experimental support to this hypothesis by using a mutational approach involving a panel of 50 mutant and wild-type F7 constructs expressed in various cell types. We demonstrate that the F7 IVS7 5' pseudo-sites are functional, but do not compete with the authentic donor splice site. Moreover, we show that the selection of the 5' splice site follows a scanning-type mechanism, precluding competition with other functional 5' pseudo-sites available on immediate sequence context downstream of the activated one. In addition, 5' pseudo-sites with an increased complementarity to U1snRNA up to 91% do not compete with the identified scanning mechanism. Altogether, these findings, which unveil a cell type-independent 5'-3'-oriented scanning process for accurate recognition of the authentic 5' splice site, reconciliate apparently contradictory observations by establishing a hierarchy of competitiveness among the determinants involved in 5' splice site selection.

Full text

Page 1

Oriented Scanning Is the Leading Mechanism
Underlying 59 Splice Site Selection in Mammals
Keren Borensztajn1*, Marie-Laure Sobrier2, Philippe Duquesnoy2, Anne-Marie Fischer1, Jacqueline Tapon-Bretaudie `re1,
Serge Amselem2*
1 Faculte ´ de Me ´decine, Universite ´ Paris-Descartes, INSERM U428, Paris, France, 2 Ho ˆpital Henri-Mondor, INSERM U654, Cre ´teil, France
Splice site selection is a key element of pre-mRNA splicing. Although it is known to involve specific recognition of short
consensus sequences by the splicing machinery, the mechanisms by which 59 splice sites are accurately identified
remain controversial and incompletely resolved. The human F7 gene contains in its seventh intron (IVS7) a 37-bp VNTR
minisatellite whose first element spans the exon7–IVS7 boundary. As a consequence, the IVS7 authentic donor splice
site is followed by several cryptic splice sites identical in sequence, referred to as 59 pseudo-sites, which normally
remain silent. This region, therefore, provides a remarkable model to decipher the mechanism underlying 59 splice site
selection in mammals. We previously suggested a model for splice site selection that, in the presence of consecutive
splice consensus sequences, would stimulate exclusively the selection of the most upstream 59 splice site, rather than
repressing the 39 following pseudo-sites. In the present study, we provide experimental support to this hypothesis by
using a mutational approach involving a panel of 50 mutant and wild-type F7 constructs expressed in various cell
types. We demonstrate that the F7 IVS7 59 pseudo-sites are functional, but do not compete with the authentic donor
splice site. Moreover, we show that the selection of the 59 splice site follows a scanning-type mechanism, precluding
competition with other functional 59 pseudo-sites available on immediate sequence context downstream of the
activated one. In addition, 59 pseudo-sites with an increased complementarity to U1snRNA up to 91% do not compete
with the identified scanning mechanism. Altogether, these findings, which unveil a cell type–independent 59?39-
oriented scanning process for accurate recognition of the authentic 59 splice site, reconciliate apparently contradictory
observations by establishing a hierarchy of competitiveness among the determinants involved in 59 splice site
selection.
Citation: Borensztajn K, Sobrier ML, Duquesnoy P, Fischer AM, Tapon-Bretaudie `re J, et al. (2006) Oriented scanning is the leading mechanism underlying 59 splice site
selection in mammals. PLoS Genet 2(9): e138. DOI: 10.1371/journal.pgen.0020138
Introduction
A problem in mammalian pre-mRNA splicing is decipher-
ing the mechanisms underlying the recognition of authentic
signals for proper splicing. The accurate recognition of exons
is a process that requires assembly of the major spliceosome,
a macromolecular machinery that involves the coordinated
action of small nuclear RNAs (snRNAs) and more than 100
polypeptides [1,2]. Any abnormality in that process will
generate aberrant mRNAs that are either unstable or code for
defective and/or deleterious protein isoforms [3]. As a
puzzling paradox, in higher eukaryotes, introns are essentially
defined by three short and poorly conserved sequences: the 59
splice site, the branch point, and the 39 splice site [4,5].
Moreover, as splice site consensus motifs are degenerated,
within a typical mammalian transcript, several sequences, in
addition to the authentic splicing elements, referred to as
pseudo-sites, may match the consensus splice site signals, and
sometimes even better than the real splice sites. These
elements define a set of pseudo-exons that greatly outnumber
genuine exons, but that are normally not included in mature
mRNAs [6].
A number of structural features has been shown to play a
key role in 59 splice site selection [7]. One of them is the splice
site sequence itself. The 59 splice site consensus sequence
comprises nine partially conserved nucleotides at the exon–
intron boundary: MAG/guragu (with M and r standing for A
or C and a or g, respectively, and / denoting the cleavage site).
This consensus actually reflects the base pairing between the
donor splice sequence and the 59 terminus of the U1snRNA,
which is involved in the early steps of splicing [8]. For a given
splice site, the strength of this interaction is usually assessed
by its consensus value (CV) [4]. Besides this interaction,
different sets of auxiliary cis-regulatory elements, known as
splicing enhancers or silencers, contribute to the identifica-
tion of authentic 59 splice sites [1].
The situation may be complicated by the presence, within
several genes, of minisatellites, with their first monomer
element spanning an exon–intron boundary. As a conse-
quence, the exact sequence of the splice donor site is
reiterated in the following intron. Such minisatellites provide
an outstanding physiological model to study the mechanisms
underlying splice site selection. However, the resulting
Editor: Yoshihide Hayashizaki, RIKEN Genomic Sciences Center, Japan
Received October 25, 2005; Accepted July 20, 2006; Published September 1, 2006
A previous version of this article appeared as an Early Online Release on July 20,
2006 (DOI: 10.1371/journal.pgen.0020138.eor).
DOI: 10.1371/journal.pgen.0020138
Copyright: ? 2006 Borensztajn et al. This is an open-access article distributed
under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the
original author and source are credited.
Abbreviations: CHO, Chinese hamster ovarian; CV, consensus value; IVS7, intron 7
in human F7 gene; snRNA, small nuclear RNA
* To whom correspondence should be addressed. E-mail: K.S.Borensztajn@amc.uva.
nl (KB); amselem@im3.inserm.fr (SA)
PLoS Genetics | www.plosgenetics.org September 2006 | Volume 2 | Issue 9 | e138 1297

Page 2

splicing pattern has been investigated in very few cases only,
and demonstrated the activation of only the most and
second-most upstream pseudo-sites. Such is the case for the
second intronic sequence of the human interferon-inducible
6–16 gene, in which the authentic splice donor site is followed
by 25 pseudo-sites, whereas transcript analysis demonstrated
the constitutive utilisation of the two most proximal pseudo-
sites in addition to the authentic one [9]. Two pseudo-sites
are also utilised in addition to the authentic splice donor site
in the human CBS gene, encoding the cystathionine beta-
synthase, that contains a minisatellite in intron (IVS) 13 with
15?20 monomer repeats[10,11]. In the human LAMIN B2
gene, which contains a minisatellite consisting of 10?15
repeats of a 100-bp monomer, only the authentic splice donor
site is used [12]; a similar situation has been observed in the
PDGF gene that contains 16?18 repeats of an 81-bp monomer
element in intron 4 [13]. In all these cases, the mechanism
precluding the use of pseudo-sites has not been investigated.
Like the above-mentioned genes, the F7 gene (GENBANK
access: NM_000131), which encodes FVII (zymogen of a
blood coagulation serine protease), displays a peculiar
organization with a minisatellite located within IVS7 [14].
In IVS7, the first 37-bp monomer element consists of the last
four bp of exon 7 and the first 33 bp of IVS7 [15]. This
minisatellite is polymorphic, with at least five different alleles
containing five to nine repeats of this 37-bp monomer ([16–
19]). The IVS7 59 authentic donor splice site UGG/gugggu
(where / represents the cleavage site) is, therefore, followed by
four to eight copies of 59 pseudo-splice sites that are strictly
identical in sequence; however, strikingly, in physiological
conditions, only the most upstream 59 splice site is used [15].
The study of F7 pre-mRNA splicing thus provides an
opportunity to assess the mechanism underlying 59 splice
site selection for transcripts of particular clinical importance,
as judged by the potentially severe phenotype of patients with
inherited FVII deficiency [20].
In previous work, we investigated the functional conse-
quences of a transversion located within the authentic IVS7
splice donor site of the F7 gene from a patient with severe
FVII deficiency [21]. We showed that the mutation resulted in
the activation of a single cryptic site corresponding to the 59
pseudo-site located in the second IVS7 minisatellite mono-
mer repeat. This observation led us to propose the existence
of a physiological mechanism that, in the presence of
consecutive splice consensus sequences, would exclusively
stimulate and select the most upstream 59 splice site, rather
than repress the 39 following pseudo-site(s) [21]. This model
prompted us to investigate the mechanism underlying the
accurate selection and activation of one particular 59 splice
donor site among five identical sites. In the present study, we
address this issue using a mutational approach. Our findings
unveil a nuclear 59?39-oriented scanning mechanism as the
leading part of the splicing machinery, which is not cell type–
dependent.
Results
F7 IVS7 59 Pseudo-Sites Are Potentially Functional, but Do
Not Compete with the Authentic Donor Splice Site:
Evidence for a Scanning Process in 59 Splice Site Selection
To test whether each of the five IVS7 59 pseudo-sites are
potentially functional, in addition to the mutant minigene
carrying the mutation identified in our patient with severe
FVII deficiency and the wild-type construct (pF7m and pF7wt,
Figures 1 and 2A), we generated five constructs (designated
pF7m1 to pF7m5, see Materials and Methods and Figure 2A),
with sequential inactivation of one to five pseudo-sites. RT-
PCR amplification of F7 transcripts isolated from Chinese
hamster ovarian (CHO) cells transfected with the wild-type
minigene yielded a 842-bp amplicon resulting from normal
splicing of the entire IVS7 (Figure 2B, lane 1). This result,
Figure 1. Schematic Representation of the Human F7 Gene and Wild-
Type and Mutant F7 Minigenes Cloned into pTracer-CMV Vector
Sizes of exons (E, open boxes) and introns (I, horizontal lines) are
indicated. Primers Pa, Pb, Pe, Ph, and Pm are represented by horizontal
arrows indicating their respective positions in the minigene sequence.
Inset indicates the last three nucleotides of exon 7 (capital letters) and
the first ten nucleotides of IVS7 (in lowercase characters) of the wild-type
F7 gene. The invariant gt dinucleotide is indicated in underlined bold
letters. A vertical arrow indicates the nucleotide change in the mutant
sequence.
DOI: 10.1371/journal.pgen.0020138.g001
PLoS Genetics | www.plosgenetics.org September 2006 | Volume 2 | Issue 9 | e138 1298
Oriented Scanning for 59 Splice Site Selection
Synopsis
Typically, mammalian genes contain coding sequences (exons)
separated by non-coding sequences (introns). Introns are removed
during pre-mRNA splicing. The accurate recognition of introns
during splicing is essential, as any abnormality in that process will
generate abnormal mRNAs that can cause diseases. Understanding
the mechanisms of accurate splice site selection is of prime interest
to life scientists. Exon–intron borders (splice sites) are defined by
short sequences that are poorly conserved. The strength of any
splice sequence can be assessed by its degree of homology with a
splice site consensus sequence. Within exons and introns, several
sequences can match with this consensus as well as or better than
the splice sites. Using a system in which a splice site sequence is
repeated several times in the intron, the authors showed that linear
59?39 search is a leading mechanism underlying splice site selection.
This scanning mechanism is cell type–independent, and only the
most upstream splice site of all the series is selected, even if splice
sites with a better match to the consensus are in the vicinity. These
findings reconciliate contradictory observations and establish a
hierarchy among the determinants involved in splice site selection.

Page 3

which was obtained with the use of primers Pe and Pm, was
reproduced with other primer sets that flank IVS7 (unpub-
lished data).
By contrast, in similar assays performed with the mutant
constructs pF7m and pF7m1 to pF7m5 (Figure 2B, lanes 2–7),
no normal splicing could be detected. More precisely,
transient transfection of constructs pF7m1 to pF7m4 resulted
in the generation of two kinds of aberrantly spliced products
(Figure 2B, lanes 3–6), one of size that was always slightly
larger than expected. Sequencing of the corresponding RT-
PCR products demonstrated that sequential inactivation of
the 59 pseudo-sites led to the activation of the next, most
upstream functional IVS7 59 pseudo-site available. Such
activation of a cryptic site led to the retention of short
portions of IVS7 in the corresponding mutant cDNA between
exons 7 and 8, accounting for the slight size differences
between the resulting RT-PCR products (Figure 2B); sequenc-
ing of the second aberrantly spliced product, which was
smaller in size than expected, showed that it corresponds to a
unique molecular species lacking the entire exon 7 (exon
skipping), thereby demonstrating that, for those constructs,
the IVS6 donor splice site was utilized in combination with
the IVS7 acceptor splice site. Strikingly, the relative amount
of the two observed splicing products seemed to differ
according to the construct, as shown after ethidium bromide
staining (Figure 2B, lanes 3 to 6). Indeed, although our
approach was not quantitative, the intensity of the larger
products corresponding to the activation of each of the
cryptic sites clearly decreased according to the number of
inactivated pseudo-sites, while concurrently, the intensity of
the smaller product lacking exon 7 increased. In keeping with
these observations, in cells transfected with the pF7m5
construct, in which all pseudo-sites were inactivated, only
the product resulting from exon 7 skipping could be detected
(Figure 2B, lane 7), whereas in cells transfected with the pF7m
construct, in which only the first 59 splice site was inactivated,
a single product was identified, resulting from the activation
of the 59 pseudo-site located in the second 37-bp monomer
element (Figure 2B, lane 2). For all transfection experiments,
examination of 50 additional clones obtained after subclon-
ing of non-purified RT-PCR products generated with primers
Pe and Pm failed to detect any other splice variants, thereby
indicating that such species, if they indeed exist, are
expressed at extremely low levels.
Altogether, these data show that the five IVS7 59 pseudo-
sites are potentially functional but do not compete for
splicing with the IVS7 authentic donor splice site. They also
reveal that when several adjacent pseudo-sites are inactivated,
only the most upstream functional IVS7 59 pseudo-site
available is used, suggesting the existence of an accurate
mechanism underlying the selection of the active 59 splice
site.
Selection of the 59 Splice Site Follows a Scanning-Type
Mechanism Precluding Competition with Other Most
Upstream Functional 59 Pseudo-Site Available on
Immediate Sequence Context
To gain insights into the mechanism underlying selection
of the donor splice site, we next generated F7 minigenes
containing functional 59 pseudo-sites, separated by one or
several inactivated 59 splice sites (Figure 3A). As shown in
Figure 3B, RT-PCR amplification of the transcripts expressed
by CHO cells transfected with the mutant constructs pF7m6
to pF7m11—in which the first pseudo-site was always
mutated—yielded a single band slightly larger in size than
the one obtained with the wild-type construct (lanes 2 to 7); as
shown after sequencing, this larger molecular species results
from activation of the next functional 59 pseudo-site.
To confirm the hypothesis that the most upstream 59 splice
site is selected for splicing, we studied the F7 transcripts
generated from CHO cells transfected with three additional
constructs (pF7m12 to pF7m14) in which the first, second,
and fourth pseudo-sites were inactivated in combination with
either the third or the sixth pseudo-site (Figure 3A). As shown
in Figure 3B, PCR amplification of the corresponding cDNAs
resulted in the generation of two kinds of aberrantly spliced
products (lanes 8 to 10): one of size always slightly larger than
expected that resulted from the activation of the most
Figure 2. Assessment of the Ability of the IVS7 Pseudo-Sites to Be
Activated
(A) Schematic representation of the constructs used in this experiment.
Top: Organization of the wild-type F7 IVS7 proximal region (pF7wt);
exons and introns are represented by open boxes and thin lines,
respectively. The first 37-bp monomer, which spans the exon7–IVS7
boundary, is repeated within IVS7. The monomers are represented by a
thick grey line. Each pseudo-splice site is represented by a grey box, and
is separated from the next one by 28 bp. Bottom: Schematic
representation of the various F7 minigenes carrying a T-to-A transition
located at the main dinucleotide of the consensus donor splice site and
involving several consecutive 37-bp monomer elements found in IVS7.
Mutations are represented by a black cross.
(B) RT-PCR amplification of F7 transcripts isolated from CHO cells
transfected with the above-mentioned constructs. The products
corresponding to the different PCR fragments, as identified by
sequencing, are schematically represented on both sides of the figure.
DOI: 10.1371/journal.pgen.0020138.g002
PLoS Genetics | www.plosgenetics.orgSeptember 2006 | Volume 2 | Issue 9 | e138 1299
Oriented Scanning for 59 Splice Site Selection

Page 4

upstream functional IVS7 59 pseudo-site, and an additional
band, smaller in size, that corresponded to a species in which
exon 7 was skipped. Again, when two splice products were
observed, their relative proportions seemed to differ accord-
ing to the position of the available functional 59 pseudo-site.
Taken together, these results are consistent with a
scanning-type process of selection of the 59 splice site, which
precludes competition with other more downstream func-
tional 59 pseudo-sites. Indeed, other splicing mechanisms
would predict the use of the first functional 59 pseudo-sites
downstream of the inactivated splice site, whereas such a
scanning model predicts predominant use of the first
available 59 splice site.
The Identified Scanning Mechanism for 59 Splice Site
Selection Is Not Cell Type–Dependent
The above-described experiments were performed in CHO
cells. To test whether the identified scanning process is cell
type–dependent, we performed similar studies in COS-7 and
HeLa cells with F7 minigenes that, when transfected in CHO
cells, gave representative RT-PCR patterns: pF7m and
pF7m10 that were associated with the use of the most
upstream functional 59 pseudo-site available (Figure 2B, lane
2, and Figure 3B, lane 6); pF7m12, associated with the same
splicing event and exon skipping (Figure 3B, lane 8); and
pF7m5 that was associated with exon skipping only (Figure
2B, lane 7). As shown in Figure 4, transfection of these
minigenes in COS-7 and HeLa cells led to splicing patterns
similar to those observed in CHO cells. These results provide
evidence for a scanning mechanism that is not cell type–
dependent, and, therefore, strongly suggest that the machi-
nery involved in 59 splice site selection is ubiquitous.
59 Pseudo-Sites with a CV as High as 91 Do Not Compete
with the Identified Scanning Process
The wild-type 59 splice donor site sequence of IVS7 of the
human F7 gene (TGG/gtgggt), like each of the five 59 pseudo-
sites, offers a reasonable agreement with the consensus
sequence of the splice donor sites, with a CV of 74 (Figure
5A), as assessed according to Shapiro and Senapathy [4]. To
test whether 59 pseudo-sites with higher CVs could compete
with the identified scanning process, we transfected CHO
cells with a series of F7 minigenes carrying different
Figure 3. Splice Site Selection in the Context of Functional 59 Pseudo-
Sites, Separated by One or Several Inactivated 59 Splice Sites
(A) Schematic representation of the constructs used in this experiment
(see legend to Figure 2A for the meaning of each symbol).
(B) RT-PCR amplification of F7 transcripts isolated from CHO cells
transfected with the above-mentioned constructs. The products
corresponding to the different PCR fragments, as identified by
sequencing, are schematically represented on both sides of the figure.
DOI: 10.1371/journal.pgen.0020138.g003
Figure 4. The Identified Scanning Mechanism for Splice Site Selection Is
Not Cell Type–Dependent
(A) Schematic representation of the constructs used in this experiment
(see legend to Figure 2A for the meaning of each symbol).
(B) RT-PCR amplification of F7 transcripts isolated from COS-7 and HeLa
cells transfected with the above-mentioned constructs. The products
corresponding to the different PCR fragments, as identified by
sequencing, are schematically represented on both sides of the figure.
DOI: 10.1371/journal.pgen.0020138.g004
PLoS Genetics | www.plosgenetics.org September 2006 | Volume 2 | Issue 9 | e138 1300
Oriented Scanning for 59 Splice Site Selection

Page 5

combinations of 59 pseudo-sites with increased CV on a
pF7wt or a pF7m background (Figures 5 and 6). The following
mutations were introduced in order to increase the com-
plementarity of the resulting transcripts with the 59 end of
the U1snRNA sequence: a G-to-A transition at position?2 or
þ3 of the TGG/gtgggt splice site (i.e., TAG/gtgggt or TGG/
gtaggt with a CV of 82.2 or 85, respectively; see Figure 5A,
top), and a G-to-A transition at positionþ4 of the TGG/gtgggt
splice site (i.e., TGG/gtgagt with a CV of 91; see Figure 5B,
top). As shown in Figure 5A and 5B (bottom), the PCR
products generated from each construct demonstrated the
exclusive use of the most upstream functional 59 site,
regardless of its CV (i.e., 74, 82.2, 85, or 91). These results
were confirmed by subcloning and subsequent sequencing of
the non-purified RT-PCR products.
We next tested whether the most upstream functional 59
site is still used when it is followed by sequences matching a
perfect splice donor site. To this end, we designed mutant F7
minigenes carrying the following mutations that were
introduced in the splice site sequences in order to match
perfectly with the 59 end of the U1snRNA sequence, in the
context of pF7wt: a T-to-C transition at position?3, a G-to-A
transition at position ?2, and a G-to-A transition at position
þ4 of the TGG/gtgggt splice site (i.e., CAG/gtgagt with a CV of
100; see Figure 6, top). As shown in Figure 6 (bottom), two
molecular species were generated from these latest con-
structs. Although noteworthy, with such constructs one would
expect the exclusive use of the perfect sites; this experiment
and the subsequent sequencing of those products showed that
the optimal splice sequences were used with an efficiency of
approximately 70% (pF7wt_100–1) and 90% (pF7wt_100–
2), thereby demonstrating that the most upstream 59 site is
always used, even in the presence of a perfect splice site.
Taken together, these results show firstly that, even when
the CV is as high as 91%, the 59 pseudo-sites do not compete
with the identified scanning process. Secondly, competition
occurs only when those sequences match a perfect consensus
(CV of 100), and even in that case, the scanning process is still
active, as demonstrated by the use of the most upstream
functional 59 site.
No Autonomous Regulating Element Lies in the
Neighbouring Intronic Sequence
Intronic splice enhancer sequences, when placed between
competing 59 splice sites, have been shown to favour the use
of the upstream, most distal, 59 splice site [22,23]; we,
therefore, hypothesized that an intronic regulatory element,
which could overlap one or several 37-bp monomer repeats,
may either prevent the use of pseudo-sites or stimulate the
use of the most upstream one. To test this hypothesis, we
performed sequential deletions of the 37-bp monomer
repeats of pF7wt and/or pF7m; the resulting constructs
carrying one to six monomer repeats are shown in Figure
7A (top). As shown in Figure 7A (bottom), such truncations
did not change the splicing pattern observed with the pF7wt-
and the pF7m-derived constructs, suggesting that the deleted
regions did not contain any regulatory element.
Similarly, we designed F7 expression plasmids with the
insertion, in pF7wt and/or pF7m, of several 37-bp monomer
elements; the resulting constructs carrying eight to 12
monomer repeats are shown in Figure 7B (top). As shown
in Figure 7B (bottom), the choice of the 59 site used for
Figure 5. Splicing of a Series of Mutants with Increased Complementarity
of the IVS7 59 Pseudo-Sites to U1snRNA
(A) Splicing of F7 transcripts generated from F7 minigenes with a CV of
85 or 82.2.
(B) Splicing of F7 transcripts generated from F7 minigenes with a CV of
91.
(A and B) Top: Wild-type (middle line in [A], lower line in [B]) and mutated
(lower and upper lines in [A], upper line in [B]) 59 pseudo-sites in the F7
minigene. Dots show the location of the introduced mutations. Suffixes
of the constructs’ names and the resulting CVs are indicated. Middle:
Schematic representation of the constructs transfected in CHO cells.
Improved splice sites are represented by a hatched square (see legend to
Figure 2A for the meaning of each symbol). Bottom: RT-PCR amplification
of F7 transcripts isolated from CHO cells transfected with the above-
mentioned constructs. The products corresponding to the different PCR
fragments, as identified by sequencing, are schematically represented on
both sides of the figure.
DOI: 10.1371/journal.pgen.0020138.g005
PLoS Genetics | www.plosgenetics.org September 2006 | Volume 2 | Issue 9 | e1381301
Oriented Scanning for 59 Splice Site Selection