ABSTRACT
We recently compared the efficiency of six picornaviral internal ribosome entry
segments (IRESes) and the hepatitis C virus (HCV) IRES for their ability to
drive internal initiation of translation
in vitro
. Here we present the results of a similar comparison performed in six different
cultured cell lines infected with a recombinant vaccinia virus expressing the
T7 polymerase and transfected with dicistronic plasmids. The IRESes could be
divided into three groups: (i) the cardiovirus and aphthovirus IRESes (and the
HCV element) direct internal initiation efficiently in all cell lines tested;
(ii) the enterovirus and rhinovirus IRESes are at least equally efficient in
several cell lines, but are extremely inefficient in certain cell types; and
(iii) the hepatitis A virus IRES is incapable of directing efficient internal
initiation in any of the cell lines used (including human hepatocytes). These
are the same three groups found when IRESes were classified according to their
activities
in vitro
, or according to sequence homologies. In a mouse neuronal cell line, the
poliovirus and other type I IRESes were not functional in an artificial
bicistronic context. However, infectious poliovirions were produced efficiently
after transfection of these cells with a genomic length RNA. Furthermore,
activity of the type I IRESes was dramatically increased upon co-expression of the poliovirus 2A proteinase, demonstrating that while IRES
efficiency may vary considerably from one cell type to another, at least in
some cases viral proteins are capable of overcoming cell-specific translational defects.
For capped eukaryotic mRNAs, translation initiation requires scanning by the
ribosome from the 5' end of the message to the initiator AUG codon (
1
). In contrast, all picornavirus RNAs examined to date have been found to
possess an internal ribosome entry segment (IRES) (
2
-
5
). These elements, which comprise ~450-500 nt of highly structured RNA situated in the long viral 5'-untranslated region (UTR), permit the initiation of
picornavirus RNA translation in a manner which is both cap and 5'-end independent (
6
,
7
). Recently, IRESes have also been demonstrated to exist in other viral messages
(
8
,
9
) and in some cellular mRNAs (
10
-
12
). It has previously been shown that the picornaviral IRESes can be classified
into three distinct groups on the basis of primary sequence and secondary
structure conservation (for a review see
6
), and also on the basis of their requirements for efficient internal initiation
of translation
in vitro
(
13
).
Type I IRESes (those of the enteroviruses and rhinoviruses) are inefficient in
driving translation initiation in reticulocyte lysates in the absence of
specific cellular proteins, and are sensitive to even slight modifications of
KCl and MgCl
2
concentrations (
13
). Furthermore, their efficiency is dramatically enhanced
in vitro
in the presence of either of two picornaviral proteinases, the 2A proteinase of
enteroviruses and rhinoviruses or the Lb proteinase of aphthoviruses (
13
-
16
). Conversely, type II IRESes (those of the cardioviruses and aphthoviruses)
initiate translation efficiently in reticulocyte lysates in the absence of
other cell proteins, and are relatively insensitive to fluctuations in salt
concentration. Additionally, they are not dramatically affected by the presence
of the 2A or Lb proteinases (
13
,
15
,
17
). In these respects, the IRES of hepatitis C virus behaved as a type II IRES (
13
). Finally, the type III IRES (that of hepatitis A virus) is relatively
inefficient in reticulocyte lysates, but to date its activity has not been
found to be markedly stimulated by supplementation with additional cell
proteins (
18
). This IRES tolerates a wide range of salt concentrations, and is inhibited by
the presence of the viral 2A or Lb proteinases (
13
,
19
). The various picornavirus IRESes can be similarly separated on the basis of
their organization and location within the viral genome; the essential
sequences required for type II and type III IRES activity extend up to,
include, or even continue beyond the authentic start site for translation (
20
-
23
), whereas the 3' end of type I IRESes lies between 30 and 150 nt upstream of the
translation initiation site (
5
,
24
-
26
).
One of the major determinants for picornaviral species and tissue tropism is the
presence or absence of the viral receptor on the cell surface (
27
). However, several results suggest that the IRES may also represent a
determinant of viral tropism.
In vitro
studies have identified different cell factors that bind to the different
IRESes and that may be required for translation initiation (
28
-
30
). In addition, the analysis of poliovirus IRES mutants showed that translation defects could be cell type-specific, the decreased translation capacity of mutant templates being evidenced in
cells or extracts of neuronal origin, rather than in HeLa cells or their
extracts (
31
-
32
). Cell-specific determinants were recently demonstrated to exist in the
poliovirus 5'-UTR using viruses with chimeric genomes. Indeed, when the
poliovirus IRES was replaced by that of human rhinovirus, neuropathogenicity in
a mouse model was abrogated (
33
).
In the present study we have examined the translation efficiencies of six
different picornaviral IRESes in a variety of human and non-human cell lines. Our results indicate that IRES activity differs
dramatically according to the cell line used, and that the different IRESes can
once again be classified into three distinct groups on the basis of cell-specific efficiencies.
Escherichia coli
strain TG1 was used for the propagation of plasmids. All of the dicistronic
plasmids used for the present study are based on the parental construct pXLJO (
5
) which contains the
Xenopus laevis
cyclin B2 gene followed by a slightly truncated influenza virus NS gene, under
the control of the T7 Ø10 promoter. The insertion into this vector of the sequences
corresponding to the complete IRESes of poliovirus type 1 Mahoney strain (PV),
ECHO virus type 25 JV4 strain (ECHO), human rhinovirus type 2 (HRV), hepatitis
A virus p16 HM175 (HAV), encephalomyocarditis virus (EMCV), foot-and-mouth disease virus (FMDV) and hepatitis C virus (HCV) has been
described previously (
13
). The amino acid sequences of the N-terminus of the NS protein as synthesised from the seven different IRES
constructions and the control pXLJO are summarised in Figure
1
.
Neuro-2A (partially differentiated mouse neuroblastoma) and SKNBE cells
(minimally differentiated human neuroblastoma) were maintained in Dulbecco's
modified Eagle's medium (DMEM) supplemented with 10% by volume of foetal calf
serum. HeLa (human cervix epitheloid carcinoma), HepG2 (human hepatocyte),
FRhK4 (monkey foetal kidney) and L929 cells (mouse C34/An connective tissue)
were maintained in DMEM supplemented with 5% foetal calf serum. BHK21 cells
(baby hamster kidney) were maintained in Glasgow MEM, supplemented with 20 mM
HEPES, 2 mM glutamine, 10% tryptose phosphate broth and 5% foetal calf serum.
HeLa S3 cells were maintained in MEM supplemented with 10% foetal calf serum.
After transfection, all cell lines were maintained in DMEM supplemented with 5%
foetal calf serum, except after transfection with full-length Mengo virus and PV transcripts, when DMEM was supplemented with 2%
foetal calf serum. Similarly, after infection with Mengo virus and PV, cell
lines were maintained in DMEM supplemented with 2% foetal calf serum.
HeLa S3 cells (5 * 10
7
) were infected with 10
8
p.f.u. of recombinant vaccinia virus vTF7-3 which expresses the bacteriophage T7 RNA polymerase (
36
). After 3 days of incubation, the infected cells were detached from the flask
by shaking, and centrifuged for 5 min at 1800
g
. The supernatant was discarded, and cells were resuspended in 2 ml MEM
supplemented with 5% foetal calf serum. Virus was released by three successive
freeze-thaw cycles, and the virus stock (~2 * 10
9
p.f.u./ml) was stored at -70oC.
Cells were seeded 24 h before each assay in 35 mm diameter plates (6 * 10
5
cells). For all assays, cells were infected with vTF7-3 recombinant vaccinia virus at a multiplicity of infection of >10, in
DMEM without serum. After 90 min at 37oC, cells were washed with DMEM without serum and were transfected with DNA
of the pXLJ series of plasmids. These plasmids encode a dicistronic mRNA in
which the first cistron (
X.laevis
cyclin B2 gene) is under the control of its own 5' UTR and the second cistron (influenza virus NS gene) is under the
control of a viral IRES (Fig.
1
). For assays using Neuro-2A, HeLa, FRhK4, BHK21 and HepG2 cells, transfections were performed using
the DOTAP reagent (Boehringer Mannheim) with either 3 or 4 [mu]g plasmid DNA and 30 [mu]g DOTAP in DMEM with 5% foetal calf serum (final volume of 2 ml per
plate). For assays using SKNBE cells, no detectable signal was obtained using
DOTAP. Therefore, transfections were performed with the maxifectin kit (gift of
Dr Andrej Sourovoi, MBCP, Rottenburg) using 2 [mu]g DNA mixed with 2 [mu]l Enhancer and 7.5 [mu]l Unifectin, in DMEM with 5% foetal calf serum (final volume of 2 ml
per plate). For L929 cells, no detectable signal was obtained using either
method. In all assays, the transfection reagent was removed by multiple washing
with PBS at 18 h post-transfection, and cells were starved of methionine for 90 min with
methionine-free DMEM. Labelling of newly synthesised proteins was performed by
incubating cells in 1 ml methionine-free DMEM containing 15 [mu]Ci [
35
S]methionine (Amersham, 1000 [mu]Ci/mmol) for 2 h. Following labelling, cells were washed twice in ice-cold PBS, lysed with 100 [mu]l lysis buffer (10 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM EDTA and 0.5% NP-40) and centrifuged at 12 000 r.p.m. for 5 min to pellet the nuclei.
Proteins in the cytoplasmic extract supernatant were analysed by 20% SDS-PAGE, and the dried gels were exposed to [beta]-max film (Amersham).
Quantification of translation efficiencies was performed densitometrically with
a Sharp JX-330 densitometer using the NIH Image program and Macintosh software. Three
different exposures of each gel were scanned to ensure that exposures were
within the linear response range of the films. For each assay, the values
obtained for IRES-driven translation were measured as the amount of NS product relative to
that of cyclin B2, to correct for variations in transfection efficiencies. For
each cell-line tested, IRES efficiency was expressed as a percentage of the most
efficient IRES in that particular cell type.
Purified pM16 and pT7-PV1-52 plasmid DNAs were linearized with
Bam
HI and
Eco
RI, respectively, and transcribed
in vitro
using T7 RNA polymerase (New England Biolabs) as described (
37
). Confluent HeLa, L929 and Neuro-2A cell monolayers (corresponding to 10
6
cells) were transfected with 1 [mu]g RNA transcripts, or 10-fold serial dilutions thereof, as described (
37
). After incubation for 24 or 40 h at 37oC, the cell sheets were scraped into the medium, and intracellular virus
liberated by three cycles of freeze-thawing. Virus stocks were then titered on HeLa cell monolayers.
Alternatively, to determine the transfection efficiency, transfected cells were
incubated under semi-solid medium for 2 days, and plaques quantified after colouring with
crystal violet.
Artificial dicistronic mRNAs provide an excellent tool for the analysis of IRES
function both
in vitro
and
in vivo
. In the absence of an IRES between the two cistrons, translation of the
downstream cistron is extremely inefficient, since it relies on initiation by
ribosomes which recommence scanning after completion of upstream cistron
translation. However, when an IRES is used to separate the two cistrons,
downstream cistron translation is dramatically increased (reviewed in
6
). For the study described here, we have used a previously well-characterised dicistronic vector system into which the IRESes of six
picornaviruses (PV, ECHOvirus, HRV, HAV, EMCV, FMDV) and a pestivirus (HCV)
have been inserted into the intercistronic spacer to drive influenza virus NS
protein synthesis (
13
; Fig.
1
). The sequences required for internal initiation of translation extend at least
up to the authentic viral initiation codon in the case of the EMCV, FMDV and
HAV IRESes, and beyond it in the case of HCV (
23
). Thus, internal initiation of translation results in the synthesis of NS
proteins with N-terminal extensions of 4, 11, 0 and 11 amino acids, respectively (Fig.
1
). Similarly, although the 3' boundary of the ECHO IRES lies upstream of the authentic initiation
codon (
26
), for ease of construction the 5'-UTR fragment used extends beyond the authentic viral AUG codon.
Thus ECHO IRES activity results in the synthesis of an NS protein which has an
N-terminal extension of eight amino acids (Fig.
1
).
The dicistronic cDNAs used are incapable of being replicated in cultured cells.
Thus, for the analysis of translation efficiencies
in vivo
, the plasmids were introduced into cells which had first been infected with
vTF7-3, a vaccinia virus which expresses the bacteriophage T7 RNA polymerase.
RNAs were thus transcribed from the dicistronic plasmids directly in the cell
cytoplasm, and subsequently translated. This results in the synthesis of
quantities of cyclin B2 and NS proteins which are directly detectable in [
35
S]methionine labelled cytoplasmic extracts upon migration through polyacrylamide
gels (Fig.
2
). Transfection efficiency can be standardised on the basis of the level of
upstream cistron (cyclin B2) translation. Initial experiments demonstrated that
in all cases, IRES-driven translation efficiency in HeLa cells increased linearly over
plasmid DNA concentrations in the range of 1-6 [mu]g per 6 * 10
5
cells (data not shown). Thus for all subsequent experiments, between 2 and 4 [mu]g (depending on the cell line and the transfection procedure used) of each
plasmid was introduced into the different cell lines.
In order to investigate the possibility that viral tissue and/or species tropism
is to some extent mediated by the nature of the IRES, we extended the
comparison to include five different cell lines. SKNBE and Neuro-2A cells (neuronal cells of human and mouse origin respectively) were
included in the analysis since PV and ECHOvirus are neurotropic. The HepG2
human hepatocyte cell line and the FRhK4 monkey kidney cell line were chosen as
further examples of primate cells, particularly because these cells are
permissive for multiplication of HAV (
38
). Finally, baby hamster kidney cells were used as an example of non-primate, non-neuronal cells. The results of densitometric analyses of the
efficiencies of the different IRESes in driving translation in these cell lines
are summarised in Figure
3
.
The analysis of picornavirus IRES efficiency using artificial dicistronic
vectors is not comparable to the situation seen upon infection of cells with
the different viruses, since the relevant viral structural or non-structural proteins are not expressed. It is possible that translational
restrictions imposed upon certain IRESes in the different cell lines could be
circumvented if viral proteins were also expressed in the cells. To examine
this possibility, we assessed the multiplication of PV in Neuro-2A cells in a single cycle (Table
1
). While mouse cells do not possess the PV receptor and thus cannot be infected
with PV, they can be transfected with genomic length PV RNA transcripts, and
the production of infectious virus can be measured by titration on permissive
cells, such as HeLa cells. Thus, L929 cells, which are mouse cells and so lack
the PV receptor but which can otherwise support PV multiplication (
39
), were used as a positive control. Similarly, Mengo virus which multiplies well
in mouse cells was used as a positive control. The production of PV in Neuro-2A cells was at least as good as that observed in L929 cells (Table
1
), and any difference between the production of infectious poliovirions and that
of Mengo virus could be attributed to cell suffering during transfection
(compare the titer of PV in infected HeLa cells to that in transfected L929
cells). Thus, even though PV IRES activity is undetectable in Neuro-2A cells using the dicistronic assay, these cells are capable of
supporting efficient PV replication, suggesting that viral protein(s) can
overcome some block to translation. We and others have recently demonstrated
that the entero- and rhinoviral 2A proteinases are capable of stimulating translation
driven by type I IRESes (
13
-
16
,
40
). Thus, we examined the possibility that the discrepancy between the results of
the dicistronic translation assays and those of transfections with infectious
PV RNA might be due to the absence of active 2A proteinase in the former assay.
To this end, IRES activity was reassessed in Neuro-2A cells cotransfected with the dicistronic plasmids and with a plasmid
which encodes the PV 2A proteinase. Similarly to the dicistronic plasmids, RNA
corresponding to the PV 2A gene is transcribed directly in the cytoplasm of
cells which are infected with vTF7-3 vaccinia virus. As can be seen from Figure
4
, co-expression of the 2A proteinase dramatically stimulated type I IRES-driven translation in Neuro-2A cells, such that the type I IRESes were then almost as
efficient as their type II counterparts. In the case of the ECHO IRES, the only
type I IRES that gave a detectable signal in Neuro-2A cells in the absence of the 2A proteinase, the stimulation of
translation efficiency observed was ~20-fold. Translation mediated by the type II and type III IRESes was
largely unaffected by the presence of 2A, although EMCV IRES activity increased
by ~1.5-fold. Such a marginal stimulation of EMCV IRES-driven translation with either the Lb or 2A proteinases has
also been demonstrated
in vitro
(
13
,
15
).
The aims of this study were to compare the capacities of different picornaviral
IRESes to mediate internal translation initiation in a variety of cultured
cells. First, we wished to assess the relative utilities of these elements in
the design of IRES-based vectors destined for foreign gene expression in specific cell types.
Secondly, this analysis would allow the possible role of these IRESes in viral
tissue and species tropism to be examined.
IRES-mediated translation was assayed after transfection of T7 promoter-dependent dicistronic DNAs into cells which had been pre-infected with a recombinant vaccinia virus expressing the T7
polymerase. Thus, dicistronic RNAs were transcribed directly in the cell. When
the different IRESes were compared in HeLa cells, similar results were obtained
to those found in an earlier
in vitro
study using physiological concentrations of added salt and supplementation of
reticulocyte lysate reactions with saturating concentrations of HeLa cell
cytoplasmic extracts. Essentially, the type I and type II IRESes were found to
be rather similar in their global efficiency in mediating internal initiation
of translation (Fig.
2
;
13
). This confirms that cell-free translation reactions can be a relevant means of examining IRES
function. Interestingly, in pilot experiments using HeLa cells, entero- and rhinovirus IRES-driven translation, but not upstream cistron translation or
translation driven by cardio- and aphthovirus IRESes, was found to be saturated when >6 [mu]g DNA was used for transfections, and the efficiency of translation
was much reduced at even higher concentrations of DNA (data not shown). This is
similar to the situation reported for translation driven by entero- and rhinovirus IRESes
in vitro
(
5
,
41
). In that case it was concluded that the IRES-containing RNAs were in excess with respect to essential non-canonical translation factors. Thus, it seems that the limiting
nature of these specific translation factors is not an artifact of
in vitro
translation systems, but can be reproduced in cells transfected with high
concentrations of DNA. For all assays of IRES efficiency reported here, care
was taken to use non-saturating DNA concentrations, ensuring that the activities of the entero- and rhinoviral IRESes on the one hand, and the cardio- and aphthoviral IRESes on the other, could be compared.
When the
in vivo
comparison was extended to incorporate five other cell lines, the picornaviral
IRESes could be classified into three distinct groups. These same three groups
were previously found on the basis of sequence homologies (
6
,
42
,
43
), and on the characteristics of IRES-driven translation
in vitro
(
13
). HAV (type III) IRES activity was virtually undetectable in all cell lines
tested; even in human hepatocytes (HepG2 cells) this IRES was <10% as active as the most efficient one (that of FMDV). The general
inefficiency of the HAV IRES observed here might be one of the reasons for the
poor replication of HAV seen in cell culture (
44
). In this respect, it should be noted that cell culture-adapted variants of HAV have mutations in the IRES which enhance viral
replication in certain cell types (
38
). Thus, our results show that the type III IRES is an extremely poor candidate
for expression vector development. Conversely, the type II IRESes (EMCV, FMDV
and HCV) seem at first view excellent candidates for such applications, since
they functioned efficiently in all cell lines tested. In this respect, the HCV
IRES was particularly noteworthy; this element drove internal translation
initiation efficiently in all cells lines tested, including those of non-primate origin. Thus, the HCV IRES, as standardised to that of the other
type II IRESes, appears to be dramatically more efficient
in vivo
than
in vitro
, where it was consistently one of the weakest elements tested (
13
). However, the genetic organisation of the type II IRESes makes their use
fastidious, as foreign genes have to be fused extremely precisely at the 3' end of the IRES. In contrast, the type I IRESes allow easier and more
flexible foreign gene insertion. Indeed, these elements, particularly that of
ECHO virus, are good candidates for expression vectors. Depending on the cell
line, the ECHO IRES can direct translation initiation even more efficiently
than the type II elements. Nevertheless, care should be taken to ensure that
the desired target cell or tissue for foreign gene expression is not
restrictive for the chosen type I IRES, since the type I IRESes as a group
exhibited the greatest cell type-specific variations in activity, being extremely efficient in several cell
lines, and almost inactive in others.
The use of a variety of cell lines, including neuronal cells of mouse and human
origin, human hepatocytes and kidney cells from monkeys and hamsters, allowed
us to address the question of whether viral tissue or species tropism is
mediated by the IRES. The best activity for the HAV IRES was obtained in human
hepatocytes (HepG2 cells). This correlates well with both the species and
tissue tropism of this virus, which multiplies in the livers of primates (
45
). However, even in HepG2 cells HAV IRES activity was relatively poor. Thus, it
is difficult to ascertain whether the marginally stronger translation obtained
compared to other cell lines is a true reflection of tissue tropism. Given the
wide tropism of EMCV, and to a lesser extent of FMDV, it was perhaps to be
expected that these IRESes function in most cell lines. It is more surprising
that the HCV IRES also functioned efficiently in most cell lines, since this
virus shows the same restricted tropism as HAV. It should be noted that the
EMCV IRES showed significantly reduced activity when compared with the FMDV and
HCV elements both in SKNBE and BHK21 cells, suggesting that a certain degree of
host-cell translational repression/enhancement may exist amongst the different
type II IRESes. The dramatic variations of type I IRES activities in the
different cell lines tested suggests that entero- and rhinovirus translational efficiency could be a major contributing
factor to viral tissue or species tropism. However, we found no obvious
correlation between IRES activity and known virus tropism. For example,
although PV is specific for humans and is neurotropic, the PV IRES barely
functioned in SKNBE cells.
It seems rather that viral proteins are required for type I IRESes to function
efficiently in certain types of cells. The translation-restrictive Neuro-2A cell line is permissive for production of infectious PV after
transfection, i.e. translation of the full-length genome must be occurring in these cells. Furthermore, PV production
in a one step growth curve after infection of Neuro-2A cells engineered to express the PV receptor was as efficient as virus
production in HeLa cells (V.R.Racaniello, personal communication). Indeed, we
have shown here that co-expression of the PV 2A proteinase was sufficient to alleviate
translational repression of type I IRES activity in Neuro-2A cells. Since the PV 2A proteinase activated translation not only from
its own IRES, but also from those of heterologous viruses, some general
phenomenon is implicated. The entero- and rhinoviral 2A proteinases have previously been shown to modulate the
levels of IRES-driven translation by cleavage of the cellular translation initiation
factor eIF4G, via at least two mechanisms. First, translation of capped host
cell mRNAs is inhibited upon cleavage of eIF4G, reducing competition for the
translation machinery during viral infection (
14
,
46
,
47
). Secondly, the cleavage products of eIF4G specifically stimulate type I IRES-driven translation, at least
in vitro
(
48
). It is interesting to speculate that these cleavage products can also
stimulate type I IRES activity
in vivo
. Nevertheless, it cannot be excluded at this time that the increased type I
IRES activity in restrictive cells in the presence of the 2A proteinase is due
to decreased competition with cellular mRNAs upon inhibition of host cell
translation. Alternatively, it could be that the 2A proteinase acts by the
removal of some specific unidentified inhibitor present in restrictive but not
permissive cells. However, given the dramatic improvement in type I IRES
activity in the presence of the 2A proteinase, it seems unlikely that marked
differences in viral tissue/species tropism between these viruses can be
assigned to IRES activity
per se
. This contrasts with the conclusions that could be drawn from studies comparing
wild-type and attenuated PV. Lowered neurovirulence due to a single mutation in
the IRES correlated with reduced translation efficiency in neuronal cells (
31
). This was interpreted as indicative of mimicry of the host restriction
encountered by attenuated PV in the central nervous system of the infected
animal. However, the phenotypic effects of such mutations can be suppressed by
second-site mutations in the 2A proteinase gene (
49
), again suggesting interplay between the IRES and viral proteins, rather than
between the IRES and cell type-specific factors.
In conclusion, our work should facilitate the choice of the IRES to be used in
the construction of vectors for foreign gene expression in different target
cells. Additionally, our results highlight the need to examine further the
possible roles of the various viral proteins in translation and the
mechanism(s) of modulation of IRES activity by such proteins.
We thank Ewald Beck and S. M. Feinstone for the pSP449 FMDV IRES clone and the
p16 HM175 infectious HAV cDNA respectively, Ann Palmenberg for the pM16 cDNA
and Richard J. Jackson for the pXLJ-HCV and pA[Delta]802 constructs. We also thank Yves Jacob, Thérèse Couderc, Bruno Blondel and Philippe Marianneau
for the various cell lines used here, and A. Sourovoi for the
Maxifectin/Unifectin reagents. Finally, we are grateful to Yves Jacob for
stimulating discussions. A.M.B was supported by a fellowship from the ROUX
foundation (Institut Pasteur); P.L.M. was supported by a Pasteur-Weizmann fellowship.
REFERENCES
Return