ABSTRACT
The domains of the PB1 subunit of the influenza virus polymerase involved in the
interaction with the PB2 and PA subunits have been defined by mutational
analysis of PB1 protein. The experimental approach included
in vivo
competition of the PB1 activity, two-hybrid interaction assays and
in vitro
binding to PB1-specific matrices. Mutants of the PB1 gene including N-terminal, C-terminal and internal deletions and single amino acid
insertions were constructed. They were unable to support polymerase activity in
a reconstituted transcription-replication system and were tested for their competition activity when
expressed in excess over wild-type PB1 protein. The pattern of competition obtained suggested that the N-terminal 78 amino acids and the sequences between positions 506 and
659 in the PB1 protein are involved in the interaction with the other
components of the polymerase. We identified the N-terminal region of PB1 protein as responsible for the interaction with the
PA subunit by two-hybrid assays in mammalian cells. N- and C-terminal fragments of the PB1 protein were expressed as His-tagged proteins and purified on Ni
2+
-NTA resin. Such PB1-specific matrices were used in binding assays
in vitro
with metabolically labelled PB2 and PA proteins and mutants thereof. The
results obtained indicated that the N-terminal and the C-terminal regions of PB1 are responsible for binding to PA and PB2 subunits,
respectively. With this information and previously published results we propose
a preliminary model for the architecture of the influenza virus RNA polymerase.
Influenza A viruses are members of the Orthomyxoviridae family whose genome
consists of eight single-stranded RNA segments of negative polarity. These RNA molecules associate
with the nucleoprotein (NP) and the three P proteins (PB2, PB1 and PA) to form
ribonucleoprotein (RNP) complexes [reviewed in (
1
,
2
)]. In a productive infection, the parental RNAs (vRNAs) are first transcribed
into viral mRNAs by the incoming transcriptase. Cellular hnRNAs are used as cap
donors to initiate transcription (
3
) and termination occurs at an oligo-U signal that is present adjacent to the RNA panhandle structure at the 5' terminus of the vRNA templates (
4
,
5
). For viral RNA replication, a full-length RNA copy of positive polarity (cRNA) is used as an intermediate (
6
).
The influenza virus RNA polymerase is a heterocomplex composed of the three
polymerase (P) proteins-PB1, PB2 and PA-present in a 1:1:1 ratio (
7
-
9
). Some information has become available in regard to the roles of each subunit
in the complex. The PB1 protein is probably responsible for polymerase
activity, since it shares common amino acid motifs with other RNA-dependent RNA polymerases (
10
). Mutation of these conserved residues abolished the transcriptional activity (
11
). The PB2 subunit binds to cap structures (
12
,
13
) and antibodies specific for PB2 protein inhibit the cap-dependent endonuclease activity (
14
). These results suggest that PB2 protein is responsible for the transcription
initiation step (
15
). The role of PA protein is poorly understood, although the phenotypes of ts
mutants [reviewed in (
16
)] suggest its involvement in vRNA synthesis. The only biochemical activity so
far described for the PA subunit is the induction of proteolysis (
17
). The regions of the PA subunit responsible for this activity map to the N-terminal third of the protein (
18
), close to the nuclear localization signal (
19
).
In spite of the functions tentatively assigned to its subunits, the
functionality of the reconstituted polymerase in either the CAT assay [reviewed
in (
20
)],
in vitro
transcription and replication (B. Perales, unpublished results) and cap-snatching (
21
) requires the participation of every polymerase subunit and the template.
Therefore, it is essential to unravel the molecular architecture of the
complex(es) formed by the polymerase subunits. Polymerase complexes have been
described in influenza virus infected cells (
8
) or from subunit expression in baculovirus vectors (
22
). Co-expression in frog oocytes demonstrated that both PB2 and PA subunits can
form complexes with the PB1 protein but can not interact directly (
23
). Previous studies in our laboratory had determined the regions of the
influenza virus polymerase subunits PB2 and PA involved in the interaction with
the PB1 protein (
24
,
25
). A small portion of PB2 protein (amino acids 1-124) was sufficient for the interaction (
24
), but the PA sequences responsible for binding to PB1 could only be mapped to
the C-terminal three-quarters of the protein, probably due to conformational constraints
(
25
). In this report we complete the mapping of the interaction domains in the
influenza virus polymerase by analyzing the competition of a collection of PB1
protein mutants with wild-type PB1 (wtPB1)
in vivo
, as well as by
in vitro
binding assays and in a two-hybrid system in mammalian cells. While this manuscript was in
preparation, the report by Prez and Donis (
26
) was published. We confirm their conclusions indicating that the N-terminal end of the PB1 protein interacts with the PA subunit, and extend
them by proposing that the region between amino acids 506 and 659 of the PB1
protein is involved in the interaction with the PB2 subunit. Taken together the
results presented here and those recently published (
24
,
25
), we propose a model for the architecture of the influenza virus polymerase
complex.
The COS-1 cell line (
27
) was provided by Y. Gluzman and was cultivated as described (
28
). The vaccinia recombinant virus vTF7-3 (
29
) was a gift from B. Moss. The
Escherichia coli
strains MH1638 and HB101 were kindly provided by S.G. Sedgwick. The origin of
plasmids pGPB1, pGPB2, pGPA and pGNP have been described previously (
30
). The pIVACAT1-S plasmid (
31
) and pGL-G5 plasmid, containing the luciferase gene under control of the GAL4
promoter, were provided by P. Palese and P. Stäheli, respectively. The origin of plasmids for the two-hybrid assay has been described previously (
25
). The pRSETA plasmid and the Ni
2+
-NTA resin were purchased from Invitrogen. The anti-T7 tag antibody was purchased from Novagen. Cationic liposomes were
prepared as described (
32
).
The generation of single amino acid insertion mutants was carried out exactly as
described (
24
). The insertion at position 416 generated an ochre codon, leading to deletion
mutant PB2[Delta]416-757. A series of mutants (mutants PB2[Delta]83-757, PB2[Delta]592-757, PB1[Delta]659-757 and PB2[Delta]711-757) were
produced by random insertion of a tagged Tn1000 transposon derivative (
33
), which due to termination codons present close to its ends and in all frames,
resulted in C-terminal deletions of the PB1 protein. Cells harbouring transposon TnXR
(present in R388 in the MH1638
E.coli
strain) were transformed with plasmid pGPB1 and selected in medium containing
ampicillin and methicillin. Such a donor strain was mated with strain HB101
(strep
R
-recA
-
) by co-culture for 2 h at 37oC on LB-agar plates in the absence of antibiotics. Transferred cointegrates were
selected in LB-agar plates containing ampicillin, methicillin and streptomycin. After characterization by restriction analysis, the plasmid was sequenced
using primers specific for the TnXR termini to determine the site of the
integration.
The construction of N-terminal deletion mutant PB1[Delta]1-179 was carried out by digestion with
Sma
I in the plasmid polylinker and
Eag
I (position 438 in the PB1 gene), filling in with Klenow enzyme and self
ligation. The C-terminal deletion mutant PB1[Delta]393-757 was constructed by digestion of pGPB1 plasmid with
Stu
I (position 1204) and
Bgl
II (position 2259), filling in with Klenow enzyme and self ligation. Similar
strategy was used to generate a number of internal deletion mutants: plasmid
pGPB1 was digested with
Mun
I (position 228) and
Stu
I (mutant PB1[Delta]69-394),
Bsu
36I (position 246) and
Eag
I (mutant PB1[Delta]78-139),
Eag
I and
Stu
I (mutant PB1[Delta]139-394) or
Bsu
36I and
Stu
I (mutant PB1[Delta]78-394), the digested DNA was filled in with Klenow enzyme and self
ligated. In addition, mutants PB1[Delta]394-506, PB1[Delta]722-752 and PB1[Delta]722-749 were generated by digestion with
Stu
I and
Xcm
I (position 1540) or
Apa
I and
Bgl
II, respectively, trimming the ends with T4 DNA polymerase and self ligation.
Double-deletion mutant PB1-69-416 was constructed by digestion of pGPB2[Delta]416-757 plasmid with
Sma
I and
Mun
I, filling in with Klenow enzyme and self ligation.
The expression of polymerase subunits (or mutants thereof) in mammalian cells
was carried out as follows: cultures of COS-1 cells in M24 wells were infected with vTF7-3 virus at a multiplicity of infection (moi) of 10 p.f.u./cell.
After 1 h, the cells were transfected with 4 [mu]g pGPB1 (or pGPB2 or pGPA) plasmid, the corresponding mutant plasmid or
pGEM3 as a control, using cationic liposomes. Transfection and all steps thereafter were carried out in the presence of 50 [mu]g/ml citosine arabinoside (AraC). After 6 h of incubation, the cultures were
washed and incubated for 1 h in methionine and cysteine free medium. Finally, the cells were labelled for 15 h with [
35
S]met-cys (0.5 mCi/ml) in a DMEM medium containing 1/10 of the normal concentrations of these amino acids.
For expression of the C-terminal portion of PB1 protein in
E.coli
, a
Hin
dIII DNA fragment (positions 1507-2341 in the PB1 gene sequence) was subcloned into the
Hin
dIII site of pRSETA plasmid. To express the N-terminal region, the full-length PB1 open reading frame, cloned into pRSETA plasmid, was digested
with
Hin
dIII and self-ligated. In this way, the sequences between position 831 and 2341 in the
PB1 gene were eliminated. These constructs were checked by restriction analysis
and transferred to
E.coli
BL21 pLysS. The expression of the His-PB1 protein N- or C-terminal fragments was accomplished by induction with 1 mM
IPTG for 2 h at 37oC. The bacterial cells were opened by sonication in a buffer containing 500 mM NaCl, 5 mM MgCl
2
, 10% glycerol, 0.1% NP-40, 100 mM imidazol and 50 mM Tris-HCl, pH 7.5 and bound to Ni
2+
-NTA resin by incubation overnight at 4oC. The resin was washed with 50 vol of a buffer containing 1 M KCl,
5 mM MgCl
2
,
10% glycerol, 0.1% Brij36T, 20 mM Tris-HCl, pH 7.5.
The competition assays were carried out as described previously (
24
). In brief, cultures of COS-1 cells were infected with vTF7-3 virus at a moi of 10 p.f.u./cell. After virus adsorption, the
cultures were co-transfected with 500 ng pGPB2, 25 ng pGPB1, 50 ng pGPA and 2 [mu]g pGNP plasmids (
20
,
30
). A 40-fold excess of competing pGPB1 mutant plasmid (or pGEM3 as a control) was
also included in the transfecting mixture. After 5 h adsorption of DNA-liposomes to the cells at 37oC, the cultures were washed with DMEM and further transfected with
100 ng NS-CAT vRNA as described (
24
) and incubated at 37oC. Twenty to 24 hours post-infection the cells were collected in DMEM, washed in TNE and opened
by freezing and thawing three times with 0.25 M Tris-HCl pH 7.5. The extracts were used for CAT assays as described previously
(
20
).
The association of PA protein with PB1 protein or their mutants was tested in a
two-hybrid system developed for animal cells (
34
and references therein). COS-1 cells were co-transfected with 1 [mu]g each of pGL-G5 plasmid, pVP-PA plasmid and pGALPB1 plasmid (or mutants thereof),
using cationic liposomes. The mutations present in plasmids pGPB1[Delta]394-506, pGPB1[Delta]722-749, pGPB1[Delta]711-757, pGPB1[Delta]659-757, pGPB1[Delta]592-757, pGPB1[Delta]416-757 and pGPB1R641 were transferred to plasmid pGALPB1 by swapping the
Bsu
36I-
Xba
I fragment including most of the PB1 gene (positions 246-2341). The cultures were harvested 40-48 h post transfection and extracts were prepared in a buffer
containing 15 mM MgSO
4
, 4 mM EGTA, 1 mM DTT, 1% Triton X-100, 25 mM glycyl-glycine, pH 7.8. The extracts were assayed for luciferase activity
as described (
35
).
The preparation of labelled extracts containing the polymerase subunits PB1, PB2
or PA or mutants thereof was carried out as described above for the expression
of PB1 mutant proteins, except that 35 mm dishes were used and the labelling
was for 1 h. The labelled cultures were washed with PBS, scraped off the plates
and lysed in 100 [mu]l buffer containing 7.5 mM ammonium sulphate, 1 mM EDTA, 1 mM DTT, 0.025% NP-40 and 10 mM Tris-HCl, pH 7.9. After vortexing, the extract was centrifuged for 3
min at 3000 r.p.m. and 4oC and the supernatant was further centrifuged for 15 min at 10 000 r.p.m.
and 4oC. For the binding reaction, 20 [mu]l of extracts were diluted to 500 [mu]l of 100 mM NaCl, 1 mM MgCl
2
,
1% NP-40, 10 mM Tris-HCl, pH 7.5 and incubated with 20 [mu]l of the PB1-specific or control matrices for 2 h at 4oC. The resin was washed 10 times with 1 ml of the
same buffer and the retained protein was eluted by boiling in gel-loading buffer.
For western blot assays, the protein extracts were separated in SDS-polyacrylamide gels, transferred to nitrocellulose membranes (Inmobilon) and saturated with 3% bovine serum albumin for 1 h at room temperature. The membranes were incubated with either anti-PB2 serum (at 1:100 dilution), anti-PA monoclonal antibodies (at 1:10 dilution) or with the anti T7-G10 tag monoclonal antibody (at a 1:3000 dilution) for 1 h at
room temperature. After washing two times for 30 min with PBS containing 0.05%
Tween 20, the filters were incubated with a 1/3000 dilution of protein A-peroxidase or with a dilution of 1/3000 of goat anti-mouse IgG conjugated to horseradish peroxidase, respectively.
Finally, the filters were washed two times for 30 min as above and developed by
enhanced chemiluminescence (ECL, Amersham).
To study the regions of the PB1 protein involved in polymerase complex
formation, a collection of PB1 mutants were constructed whose description is
summarized in Figure
1
. It included C-terminal, N-terminal and internal deletion mutants as well as single-amino acid insertion mutants (see Materials and Methods for
details). All mutants were generated in the background of the pGEM vector,
under the control of the T7 promoter, and hence they could be expressed by
transfection into mammalian cells previously infected with a vaccinia
recombinant virus capable of expressing the T7 RNA polymerase (vTF7-3). When COS-1 cell cultures were infected-transfected by such an expression strategy and labelled
continuously with [
35
S]met-cys to ascertain the synthesis and stability of the mutant proteins, the
results presented in Figure
2
were obtained. The levels of label obtained for the insertion mutants (Fig.
2
A), the mutants containing small deletions (Fig.
2
A), those containing intermediate deletions (Fig.
2
B and C) as well as the smallest mutant protein PB1[Delta]83-757 (Fig.
2
D) were somewhat variable, but within the values obtained for wtPB1 protein (see
the two separate samples of wtPB1 shown in Fig.
2
A). In any case, the expression levels obtained exceeded those required for
competition of activity. Thus, even mutant PB1[Delta]78-130, the one expressed to the lowest level, could effectively
compete for wtPB1 activity (see Fig.
3
). The mutant proteins were characterized in regard to their ability to support
CAT activity in the transcription-replication system reconstituted
in vivo
(
24
,
30
). All mutants were completely negative (data not shown). This result is not
surprising, since relatively large portions of the protein were absent in the
various deletion mutants, and allowed us to carry out competition experiments
in vivo
to determine the functional interactions of these mutants with the rest of the
polymerase components.
As described previously for the competition experiments regarding PB2 mutants (
24
), the first step carried out was to determine the minimal amount of PB1-expressing plasmid to be transfected in order to obtain maximal CAT
activity in the reconstituted system. Such previous experiments indicated that
25 ng of pGPB1 plasmid were sufficient to yield 100% of the CAT activity in the
assay (data not shown) and such a dose was therefore used in the competition
experiments carried out subsequently. They involved the transfection of a 40-fold excess of mutant, competing plasmid, in addition to pGPB1 plasmid and
the rest of the components required for the reconstitution
in vivo
(
24
,
30
). The results of several such competition experiments are shown in Figure
3
. Since PB1 protein interacts with both the PB2 and PA subunits of the
polymerase (
23
), the most informative mutants were those able to compete fully for the PB1 wt
activity, i.e. those not affected in the interaction domains. All insertion
mutants tested, used as positive control for the competition experiments, were
able to compete almost completely for the activity of the PB1 protein (Fig.
3
; mutants PB1R72, PB1D75, PB1N309 and PB1R541). Using the results obtained with
these mutants, a value of CAT activity of 20% was established as cut-off limit to identify fully competing mutants (Fig.
3
). Two regions of the PB1 protein appeared as non-essential for the interaction with the other polymerase components: the C-terminal part of the protein (amino acids 659-757) (Fig.
3
; see mutants PB1[Delta]659-757, PB1[Delta]722-749 and PB1[Delta]722-752) and an internal portion
comprising amino acids 78-506 (Fig.
3
; see mutants PB1[Delta]78-139, PB1[Delta]139-394 and PB1[Delta]394-506). In agreement with these results,
it was not possible to compete the activity of wtPB1 by overexpression of
mutant PB1-69-416, that only encodes the internal protein region (Fig.
3
). The lack of competition observed with other deletion mutants might reflect
alterations in the interacting sequences that reduce the binding affinity or
might simply represent global misfolding of the resulting proteins. In summary,
the results of competition experiments suggest that interacting regions exist
at the very N-terminal end of PB1 protein and at its C-terminal third, excluding the C-terminus. However, these experiments do not allow the
identification of the protein partner of the interaction at each of these
separate ends of the PB1 subunit.
To identify which of the PB1 interacting domains is the responsible for the
binding of the PA subunit we made use of a two-hybrid assay based on the expression of PB1 and PA proteins fused to the
GAL4 DNA binding domain (GAL) or the VP16 activation domain (VP) in mammalian
cells. This assay was used previously to investigate the regions of PA protein
important for its interaction with PB1 (
25
). Co-transfection of wt alleles of GAL-PB1 and VP-PA led to the expression of cotransfected luciferase marker
(up to 10% of the levels obtained by transfection of intact GAL4-VP16 transactivator; data not shown). In contrast, the luciferase
activity obtained in the assay when only GAL-PB1 was transfected or when GAL-PB1 was co-transfected with a mutant allele of VP-PA (
25
) was greatly reduced (Fig.
4
) allowing an experimental window to measure the interaction of mutant versions of GAL-PB1 with wt VP-PA. The results of a series of co-transfection experiments are shown in Fig.
4
. All mutants tested, that covered serial deletions from the PB1 protein C-terminus, led to luciferase values similar to that of wt GAL-PB1, indicating that the interaction with the PA subunit is
mediated by the N-terminal domain of PB1 protein. Since deletion of the C-terminal end of PB1 did not affect binding to PA protein, it is
tempting to speculate that this region could be involved in PB1 binding to PB2.
Unfortunately, the attempts to detect PB1-PB2 interaction in the two-hybrid system were unsuccessful and hence this prediction could not
be tested using such an experimental approach.
To establish whether the PB2 binding region of the PB1 subunit maps to the C-terminal third of the protein and to confirm that the PA binding region is
located at its N-terminal end, fragments of the PB1 gene, encoding amino acids 1-266 or 494-757, were fused to a histidine tag and an antigenic T7 tag
in the pRSET vector and expressed in
E.coli
. The tagged PB1 fragments were bound to and purified on a Ni
2+
-NTA resin as indicated in Materials and Methods. The characterization of
such affinity matrices is shown in Figure
5
. The his-PB1 fragments were essentially the only proteins retained in the resin
after extensive washing, as detected by Coomassie blue staining (Fig.
5
A). Their identification as the recombinant protein fragments was carried out by
western blotting using anti-T7 tag antibodies (Fig.
5
B). As a control matrix, Ni
2+
-NTA resin was incubated with extracts of
E.coli
transformed by empty pRSETA vector and washed in parallel to the his-PB1 resins. Such a control matrix was essentially devoid of protein, as
detected by Coomassie blue staining and shown in Figure
5
A. Although detection by Coomassie blue staining does not allow us to reach any
strong conclusion on protein purity, the use of the control resin in binding
experiments (see below) indicated that the possible low level of contaminants
present did not contribute to the binding activities detected by the PB1-specific resins.
In the present report, the regions of the PB1 protein involved in the
interaction with the other polymerase subunits have been studied by means of a
series of PB1 mutants, most of which contained deletions and some had single
amino acid insertions (Fig.
1
). When these mutant genes were expressed in the vaccinia-T7 infection-transfection system (
30
) they showed accumulation levels comparable with that of wtPB1 (Fig.
2
). The interaction with the PB2 and PA subunits was studied by a set of
experimental approaches including
in vitro
binding to PB1-specific matrices, two-hybrid assays in mammalian cells and competition for the activity of
wtPB1 in a reconstituted transcription-replication system
in vivo
. Taken together, the results obtained by the three experimental approaches used
indicate that the PB1 protein contains two separate domains involved in the
interaction with the other polymerase subunits. These functional domains can be
unambiguously located at the N- and C-terminal regions of the protein and assigned to the interaction with
PA and PB2, respectively, by the pull-down experiments
in vitro
(Figs
5
-
7
). These results confirm
in vitro
the binding phenotypes of PB2 and PA mutant proteins whose interaction with PB1
had been studied
in vivo
(
24
,
25
). Furthermore, the interaction of PA protein to the N-terminal domain of PB1 was confirmed by two-hybrid assays (Fig.
4
). These results are in agreement with those presented by Prez and Donis who
used a similar two-hybrid system to map the PA-binding site of PB1 protein to the N-terminal 48 amino acids (
26
).
The fact that all mutants used were inactive in the transcription-replication system reconstituted
in vivo
allowed us to check the relevance of the interactions detected for the
biological activity of the viral polymerase by competition assays. The results
obtained are fully compatible with those obtained by the other approaches and
further define the interacting domains to the N-terminal 78 amino acids and to the region included within positions 506
and 659 in the protein sequence (Fig.
3
).
The body of evidence presented here, together with the results reported earlier
(
24
-
26
), allow us to propose a preliminary model for the architecture of the influenza
virus polymerase, as shown in Figure
8
. The N-terminal 78 amino acids of PB1 would be responsible of the binding to PA
protein through the C-terminal three quarters of the latter (
25
), while the PB1 sequences comprised within positions 506-659 would mediate interaction with the first 124 amino acids of PB2
protein (
24
). A sequence comparison among the PB1 genes of the influenza types A, B and C,
as well as the related Dhori and Thogoto viruses, show a number of conserved
amino acids within the N-terminal region responsible for the binding of PA protein. The
conservation in the 506-659 region of the PB1 protein is less apparent, although some conserved
amino acids can be identified scattered throughout. Some of the conserved amino
acids have a high propensity to be at the protein surface. Whether they
represent contact sites of PB1 and the other polymerase subunits must await
further experimentation. It is interesting to note that no overlap is apparent
between the different functional regions so far defined in the PB1 subunit, as
the nuclear localization signal (NLS) (
37
), the polymerase motifs (
10
,
11
) and the interaction domains responsible for PB2 and PA binding (
26
and this report) (Fig.
8
). Likewise, there is no overlap between the NLS, the PB1 interacting domain and
a presumptive cap-binding region (
24
) in the PB2 protein. The situation is less clear for the PA subunit of the
polymerase, because the majority of the protein seems to be required for
interaction with PB1 (
25
) (Fig.
8
).
We are indebted to J.A. Melero, A. Nieto and A. Portela for their critical
comments on the manuscript. We thank B. Moss, S.G. Sedgwick, P. Palese, J.J. Sanz and P. Stäheli for providing biological materials. The technical assistance of J.
Fernández and M. Pastor is gratefully acknowledged. S.G. and T.Z were fellows
from Programa Nacional de Formación de Personal Investigador and Human Frontiers Science Program,
respectively. This work was supported by Programa Sectorial de Promoción General del Conocimiento (grant PB94-1542), Comunidad Autónoma de Madrid (grant A0063) and by the EU HCM Program (grant
ERBCHRXCT949453).
REFERENCES
Return