Identification of Hsp90 as a stimulatory host factor involved in influenza virus RNA synthesis

Efficient transcription and replication of the influenza virus genome are dependent upon host-derived factors. Using an in vitro RNA synthesis system, we have purified and identified Hsp90 as one of the host factors that stimulate viral RNA polymerase activity. Hsp90 interacted with the PB2 subunit of the viral RNA polymerase through the amino terminal chaperone domain and the middle region containing a highly acidic domain. The acidic middle region was also responsible for its stimulatory activity. We found that a portion of Hsp90 is re-localized to the cell nucleus after viral infection. A PB2 fragment containing a Hsp90 binding domain inhibited viral gene expression in a dominant-negative manner. These results suggest that Hsp90 is a host factor for the influenza virus RNA polymerase.


In vitro Viral RNA Synthesis
In this report, all viral resources were derived from influenza A/Puerto Rico/8/34 virus.
vRNP was purified from virions as previously described (19) and was used as the enzyme source in the in vitro RNA synthesis assays. The 53 base-long model vRNA designated as 53-merVwt was synthesized by transcription with MEGAscript T7 Kits (Ambion) and synthetic DNA templates, as previously described (13). In vitro RNA synthesis was carried out at 30°C for 60 minutes in 25 µl of a reaction mixture containing 50 mM HEPES-NaOH (pH 7.9); 3 mM MgCl 2 ; 50 mM KCl; 1.5 mM dithiothreitol (DTT); 500 µM each of ATP, GTP, and CTP; 25 µM UTP; 5 µCi of [α-32 P] UTP (400 Ci/mmol); 10 U of RNase inhibitor; 25 µg/ml of actinomycin D; 250 µM ApG; 5 ng of a 53-merVwt; and vRNP (10 ng NP equivalents) in the presence or absence of the host factor fractions. A limited elongation assay was carried out in the same reaction mixture, with the exception that UTP and [α-32 P] UTP were omitted, and 500 µM each of ATP and CTP were included, as well as 25 µM GTP, and 5 µCi of [ α-32 P] GTP (400 Ci/mmol). Runoff RNA synthesis was carried out with a 35-mer adenine (A) residue-less template (35-merV-a; 5'-GUUCUUCUUCUUCUUUCUUCUGGCCUGCUUUUGCU-3'), which contained the 12 base-long conserved promoter sequence for the viral RNA polymerase at the 3' terminal region and the subsequent 23 base-long sequence lacking the A-residues. The other materials for the run-off RNA synthesis were the same as those used for the standard or limited elongation assays.

Preparation of Vectors and Recombinant Proteins
The nucleotide sequences of the plasmids used in this study were confirmed by DNA sequencing. The plasmid containing full-length Hsp90α and β cDNAs were the kind gift of Dr.

Identification of RAF-1 as Hsp90 and -
Using an in vitro influenza virus RNA synthesis system with an exogenously added 53 base-long model virus genome (53-merVwt), we found stimulatory host factors for the viral RNA polymerase, designated as RAF (RNA polymerase activating factor) -1 and RAF-2, in nuclear extracts prepared from uninfected HeLa cells (14). Previously, we demonstrated by biochemical fractionation that RAF-2 consists of 48 kDa and 36 kDa polypeptides; we found that the former peptide is identical with BAT1/UAP56, a putative splicing factor (15). Here, we purified RAF-1 as a host factor that stimulated viral RNA synthesis; purification proceeded to terminus. We then examined their stimulatory activity as regards the in vitro viral RNA synthesis. The recombinant Hsp90α (Fig. 1C, lanes 6-10) demonstrated approximately the same level of stimulatory activity as that of Hsp90β (data not shown), and recombinant Hsp90α demonstrated slightly higher stimulatory activity than did purified RAF-1/Hsp90 (lanes 2-5).
We assume that a portion of purified RAF-1/Hsp90 might have been inactivated during the process of column chromatography. These results confirmed that Hsp90α and Hsp90β are the active components in RAF-1. Hsp90α and -β have quite similar domain structure organization, and the similarity of amino acid sequences between Hsp90α and -β is approximately 93%. In the limited elongation condition lacking UTP, the 35-mer RNA was synthesized, along with short RNAs of 12-19 bases derived from endogenous viral RNA segments (lane 3). This finding suggests that the RNA synthesis from exogenously added model RNA templates was catalyzed by not only an RNA polymerase that proceeded toward and fell off of the end of the endogenous RNA template, but also by an RNA polymerase that paused and fell off at the first A residue.
RAF-1/Hsp90 also stimulated the limited elongation reaction and synthesis of the 35-mer RNA product (lane 4). These results suggest that RAF-1/Hsp90 exerted its activity at the steps prior to the early elongation stage (see Discussion). 13 suggesting that these proteins do not interact via RNA. Hsp90β also bound to PB2 as Hsp90α, indicating that PB2 would interact with a homologous region between Hsp90α and β (Fig. 3C).
Although RAF-1/Hsp90 was purified by monitoring the stimulatory activity for the viral RNA polymerase, other heat shock proteins may show similar activity. Thus, we tried to examine the effect of Hsp70 (heat shock 70 KDa protein 1A), one of the major Hsp70 proteins, in our system. In the GST pull-down assays, Hsp70 was unable to bind to PB2 as strongly as did Thus, the RAF stimulatory activity of RAF-1/Hsp90, possibly through PB2, could not be replaced by Hsp70.

Analysis of the Domains Involved in Hsp90 -PB2 Interaction
Next, we tried to determine the domains involved in the interaction between Hsp90α and PB2. Based on the reports (20,21,22,23,24) concerning the domain organization of Hsp90, we divided Hsp90α into three regions, a hydrophobic amino terminal region (N), a highly acidic middle region (M), and a carboxyl terminal homo-dimerization region (C), as shown in Fig.   5A. GST pull-down assays were carried out with three Hsp90α deletion mutants and wild-type Hsp90α containing GST at their carboxyl termini in binding buffers containing 50 mM (low-salt) or 1.0 M (high-salt) KCl. After the binding reaction was carried out at 37°C for 60 minutes, the affinity beads were washed twice with NETN buffer containing 100 mM (low salt) or 1.0 M NaCl (high salt). Hsp90α mutants and wild-type Hsp90α interacted with the PB2 subunit ( Fig.   5B), but not with the PB1 and PA subunits (data not shown). Wild-type Hsp90α was able to interact with PB2, irrespective of low-or high-salt binding and low-or high-salt washing conditions (lane 9), although the high-salt wash slightly decreased the PB2-Hsp90α interaction.
All of the Hsp90α mutants were capable of interacting with PB2 under low-salt washing conditions (1st and 3rd panels from the top), but with less efficiency than that observed in the case of wild-type Hsp90. Under the high-salt washing condition, the Hsp90α carboxyl terminal mutant (C) released PB2 (2nd and 4th panels, lane 7). Thus, the N and M regions were capable of binding to PB2 more strongly than was the C region. These results indicate that each Hsp90 region was involved with a different affinity in the Hsp90α-PB2 interaction.
The stimulatory activities of these mutants were examined in an in vitro RNA synthesis system using equal moles of Hsp90α deletion mutants (Fig. 5C). Although these three mutants were able to interact with PB2 under the low-salt condition, which was also used in the  . 5B), it is suggested that Hsp90 interacts with PB2 through N, M, or NM; moreover, the Hsp90 middle region, which contains a highly acidic region, is responsible for the RAF-1 stimulatory activity of Hsp90α.
Next, we tried to roughly estimate a Hsp90 binding site in PB2. GST pull-down assays were carried out with PB2 proteins and either an Hsp90α NM region or wild-type Hsp90α tagged with GST at their carboxyl termini (Fig. 6). The labeled wild-type PB2 and PB2 fragments, N515 and N383, (Fig. 6A)  Therefore, it is possible that the NM region, which is fully active as regards RAF-1 activity (Fig.   5C), interacts with PB2 through the region between its amino-terminus and the amino acid position 515. N383 was able to bind to wild-type Hsp90α more efficiently than to the NM region. Although we do not know the exact reason for this at present, we assume that wild-type Hsp90 binds to the N383 fragment as a chaperone as well as RAF-1. The structure of the PB2 fragment would be much less stable than that of the wild-type protein, and thus would be targeted by the Hsp90 chaperone. Since dimerization and/or oligomerization of Hsp90 is required for its chaperone activity (20,22), Hsp90α NM lacking the carboxyl terminal region that is involved in dimerization and/or oligomerization was unable to behave as a molecular chaperone.

Re-localization of Hsp90 in Infected Cells and Involvement of Hsp90 in Viral Multiplication
There appears to be a contradiction regarding the observation that the majority of Hsp90 molecules is localized in the cytoplasm of uninfected cells (Fig. 7B) (24), while viral RNA synthesis takes place in the nuclei of infected cells (Fig. 7C). We hypothesized that the intracellular localization of Hsp90 could be changed during viral infection. This was indeed the case (Fig. 7D). In infected cells, Hsp90 was localized in both the nucleus and the cytoplasm, although the total amounts of Hsp90 protein determined by Western blot analysis with rabbit anti Nuclear localization of a fraction of Hsp90 was also detected ( Fig. 7J and N) in cells co-transfected with pCAGGS-PB2-Myc, encoding a PB2-containing carboxyl-terminal myc-tag ( Fig. 7F), and pCAGGS-FLAG-Hsp90α, encoding an Hsp90α-containing amino-terminal FLAG-tag (Fig. 7J). On the other hand, nuclear localization of FLAG-Hsp90α was less often observed in cells co-expressing myc-258-401 ( Fig. 7I and M). Myc-258-401 was present around the outside periphery of the nucleus and/or in spots in the cytoplasm (Fig. 7E and G) and it co-localized with FLAG-Hsp90α at the outside periphery of the nucleus (Fig. 7M).

258-401-negative cells, the cell membrane was reactive to anti-influenza virus A/Puerto
Rico/8/34 HA antiserum (Fig. 7K). In sharp contrast, myc-258-401-positive cells were markedly reduced in their expression of HA (Fig. 7K, arrowheads). Expression of NP-myc ( Fig.   7H) did not have such an effect ( Fig. 7L and P, arrowheads). This was also the case in cells expressing GFP (data not shown). These observations suggest that the inhibitory effect of myc-

Discussion
In this manuscript, we demonstrated the possibility that influenza virus requires the host cellular stress protein, Hsp90, as a stimulatory host factor for its RNA synthesis. Hsp90 is a typical molecular chaperone highly conserved among higher eukaryotes, and mutations in this protein cause serious disorganization of cellular metabolism (26). The ubiquity of this protein would be advantageous to viral infections among animals, such as the influenza virus. In this report, we demonstrated that Hsp70 could not function as a stimulatory host factor, in spite of its functional similarity as a molecular chaperone (Fig. 4B). Hsp70 is highly conserved among species and also functions as a molecular chaperone, as does Hsp90. Therefore, it is speculated that the protein chaperone activity of Hsp90 may not be responsible for the RAF-1 stimulatory activity. In fact, we found that the stimulatory activity of Hsp90 resides in its middle region but not in its amino terminal chaperone domain (Fig. 5C), although both regions would be required for stable interaction with PB2 (Fig. 5B). Furthermore, Hsp70 is known as a general molecular chaperone, whereas interactors with Hsp90 are more specific to that protein (27). Thus, the interaction between Hsp90 and PB2 might be specific (Figs. 3 and 4), even if Hsp90 functions as a chaperone for PB2.
In this report, we used a newly developed 35-mer model viral RNA, 35-merV-a, to examine the mechanism of stimulation of viral RNA synthesis. The efficiency of RNA synthesis with the 35-merV-a seemed to be greater than that with the 53-merVwt template (Fig.   2B). This observation could be due to the assumption that the recycling of the RNA polymerase on the run-off template was more effective than that on the 53-merVwt template. In the limited elongation assay condition lacking UTP, RAF-1/Hsp90 was capable of stimulating RNA synthesis not only from 35-merV-a, but also from endogenous vRNA (Fig. 2B) (14). Instead, it is thought that RAF-1/Hsp90 facilitates the association of RNA-free RNA polymerases to template RNA and/or stabilizes the RNA polymerase during its translocation between templates. This assumption is supported by our preliminary result that pre-incubation of vRNP resulted in inactivation of the RNA polymerase activity, whereas this inactivation was suppressed to some extent in the presence of RAF-1/Hsp90 (data not shown).
The acidic middle domain of Hsp90 is involved in RAF-1 activity. We have proposed the term "acidic molecular chaperone" to describe proteins that contain highly acidic regions and function as chaperones for basic proteins, thanks to their acidic properties, which allow them to imitate the nature of a nucleic acid (12). Transcription and replication from basic protein-nucleic acid complexes, such as ribonucleoprotein and chromatin templates, require the dissociation and re-association of nucleic acid-binding proteins from and to nucleic acids. The acidic molecular chaperone prevents aggregation and inactivation of basic proteins, and facilitates association and dissociation of basic proteins to/from nucleic acids. A variety of host factors involved in viral genome functions could be categorized as acidic molecular chaperones. Acidic cytoskeletal proteins such as tubulin (9, 10) and actin (28,29)  containing an acidic region has been found to function as a chaperone for NP (15). This has also been demonstrated in the case of DNA virus. Template activating factors-I, -II, and -III contain functional highly acidic regions and stimulate adenovirus transcription and replication from viral DNA complexed with basic core proteins or histones (37,38,39). Along these lines, we hypothesized that Hsp90 may function as the acidic molecular chaperone for PB2, a basic polymerase subunit.
An alternative aspect of the Hsp90 function in influenza virus RNA synthesis is that Hsp90 may be involved in the modulation of the activity and structure of viral RNA polymerase.
It is reported that Hsp90 is involved in hepadnavirus reverse-transcription (40). Hepatitis B virus genome replication includes a reverse-transcription step of the pre-genomic RNA that is catalyzed by viral reverse transcriptase-Hsp90 complexes (41). The influenza virus RNA polymerase complex, comprised of PB2, PB1, and PA, is stably formed by strong binding between PB2 and PB1 and between PB1 and PA (25). However, it has been reported using cells expressing these subunits that a combination of PB2 and PB1 can catalyze transcription, whereas the combination of PB1 and PA supports replication (42,43). Dissociation of the complex into each subunit has not been observed under physiological conditions. It remains of interest that the PB1 and PA subunits were found to be easily released from the polymerase complex that was bound to Hsp90 in GST pull-down assays ( Fig. 3B and Fig. 4A). It seems likely that binding of Hsp90 to PB2 in the polymerase complex loosens PB2-PB1 interactions. It would be worthwhile to examine whether or not Hsp90 is one of the triggers involved in a possible conversion of the RNA polymerase complexes as regards their structure and function.
Immunoprecipitation assays have revealed the interaction between Hsp90 and the RNA polymerase in vivo (unpublished observation). However, no Hsp90 molecule was detected in 1 µg NP-equivalent purified vRNP or in virions when tested by Western blot analysis with anti-Hsp90 polyclonal antiserum and an enhanced chemiluminescence system (data not shown).
Therefore, we suggest here that there may be a mechanism(s) for the dissociation of PB2-Hsp90 interactions when nascent vRNPs are packaged in progeny virions and/or there may be a mechanism for the elimination of the Hsp90-vRNP complex from the packaging process.
It was unexpected that Hsp90/RAF-1 could be purified from uninfected HeLa cell nuclear extracts, since a major fraction of Hsp90 is localized in the cytoplasm, and only a low level of Hsp90 is present in nuclei (Fig. 7B) (24). We might have purified Hsp90 present in the nuclei, giving this minor fraction of Hsp90, and/or Hsp90 might have contaminated in the nuclear fractions during the course of cell fractionation. In infected cells (Fig. 7D), Hsp90 is relocalized to nuclei, where the transcription and replication of influenza virus take place. At present, it remains unclear whether the re-localization of Hsp90 is caused by cellular stress during infection or by co-migration with viral nuclear protein(s). Recently, it was reported that in vaccinia virus-infected cells, Hsp90 was transiently associated with virosomes that were localized in cytoplasm (44). Along these lines, it is possible that the nuclear re-localization of Hsp90 in influenza virus-infected cells takes place as a co-migration with viral RNA polymerase, possibly with the PB2 subunit. The possibility that newly synthesized PB2 is a carrier candidate (Fig 7N) is being examined further by our group.   mM HEPES-NaOH pH7.9, 50 mM KCl, 6 mM MgCl 2 , and 3 mM DTT. Subsequently, the mixtures were loaded onto 1.2 ml of a 30% to 60% glycerol density gradient buffer (50 mM HEPES-NaOH pH7.9, 50 mM KCl, 6 mM MgCl 2 , 3 mM DTT and 30% to 60% glycerol) and centrifuged at 4°C in a SW50.1 rotor at 45,000 rpm for 5 hours. Fractionation (200 µl) was carried out from the top of the gradient. One hundred microliters of each fraction were precipitated with 10% trichloroacetic acid and were loaded onto 7.5% SDS-PAGE. Proteins