Caveolin-1 influences human influenza A virus (H1N1) multiplication in cell culture

Background The threat of recurring influenza pandemics caused by new viral strains and the occurrence of escape mutants necessitate the search for potent therapeutic targets. The dependence of viruses on cellular factors provides a weak-spot in the viral multiplication strategy and a means to interfere with viral multiplication. Results Using a motif-based search strategy for antiviral targets we identified caveolin-1 (Cav-1) as a putative cellular interaction partner of human influenza A viruses, including the pandemic influenza A virus (H1N1) strains of swine origin circulating from spring 2009 on. The influence of Cav-1 on human influenza A/PR/8/34 (H1N1) virus replication was determined in inhibition and competition experiments. RNAi-mediated Cav-1 knock-down as well as transfection of a dominant-negative Cav-1 mutant results in a decrease in virus titre in infected Madin-Darby canine kidney cells (MDCK), a cell line commonly used in basic influenza research as well as in virus vaccine production. To understand the molecular basis of the phenomenon we focussed on the putative caveolin-1 binding domain (CBD) located in the lumenal, juxtamembranal portion of the M2 matrix protein which has been identified in the motif-based search. Pull-down assays and co-immunoprecipitation experiments showed that caveolin-1 binds to M2. The data suggest, that Cav-1 modulates influenza virus A replication presumably based on M2/Cav-1 interaction. Conclusion As Cav-1 is involved in the human influenza A virus life cycle, the multifunctional protein and its interaction with M2 protein of human influenza A viruses represent a promising starting point for the search for antiviral agents.


Background
In the last few years the interaction of viral matrix proteins or precursors with cellular proteins has attracted much attention in the field of medical virology due to the increase in the understanding of their interplay in late viral processes like protein transport, virus assembly and budding. Viral matrix proteins establish the link between outer shell and capsid core of enveloped viruses and bring together these parts in the virus assembly step. Moreover, matrix proteins frequently determine the place where the assembly step occurs. In influenza A viruses two M proteins are located on RNA7 of the negative-stranded, segmented RNA virus. The M1 protein functions as a typical matrix protein, while M2 exerts multiple tasks in the early and late phase of virus infection. M2 tetramers form an ion channel and in the early phase of virus infection M2 serves for the release of viral nucleocapsid by acidification of endosomes. In the late phases, M2 prevents premature activation of newly synthesized HA [1] and -in concert with M1-contributes to virus budding and morphology. The involvement in virus exit has been assigned to the cytoplasmic tail of the protein [2][3][4]. Influenza viruses bud from lipid rafts and for this event the components of the viral envelope (haemagglutin HA, neuraminidase NA, M2) and the RNA containing protein complex (vRNP) must come together to form infectious virus [5][6][7]. Interestingly, the endosomal sorting machinery (ESCRT), which has been involved in late steps of other viruses, does not contribute to influenza virus budding [6,8]. Accordingly, other routes and gates have been suggested for the transport of influenza proteins and virus assembly/budding [5].
In several previous investigations caveolin-1 (Cav-1), a multifunctional, raft-resident membrane protein has been linked to the virus replication of retroviruses HIV-1 and amphotropic mouse leukemia virus, rotavirus and respiratory syncytial virus [9][10][11][12][13]. Interestingly, a contribution of Cav-1 to HA transport has been reported for influenza virus infected MDCK cells [14]. In a recent investigation of the enveloped γ-retroviruses budding from lipid rafts we showed that caveolin-1 (Cav-1) interacts specifically with the MLV retroviral matrix protein in the Gag precursor, suggesting that Cav-1 serves in positioning the Gag precursor at lipid rafts [13]. Not surprisingly, Cav-1 is incorporated into MLV virions released from mouse NIH3T3 [13,15]. Subsequently, competition and inhibition experiments provided evidence that Cav-1 modulates MLV retrovirus production [13]. Taken together, these findings pointed to a general contribution of Cav-1 in virus replication strategy and opened the possibility that other virus families budding from lipid rafts may co-opt the functions of Cav-1. In our search for cellular/viral targets a database screen for Cav-1 binding sites notably revealed that structural proteins like matrix proteins of other viral families, e.g. Orthomyxoviridae with influenza A virus as a representative, exhibit regions of homology with a consensus motif for Cav-1 binding (Cav-1 binding domain, CBD) (Wirth, M, unpublished).
To address the biological relevance of the interplay of Cav-1 with influenza proteins we performed inhibition experiments with a dominant-negative Cav-1 mutant, knock-down by Cav-1 RNAi as well as competition experiments with M2 fusion proteins. We found, that the yield of human influenza virus progeny is affected by the presence/absence of Cav-1. The data suggest that Cav-1 can support the human influenza virus A life cycle. Pulldown and co-immunoprecipitation experiments were performed which showed binding of M2 and Cav-1.

Influenza A virus titres are affected in MDCK Cav-1 knockdown cells
We used MDCK (ATCC CCL-34), a canine kidney cell line commonly used in basic influenza virus research and vaccine production [16][17][18][19]. To elucidate the biological importance of Cav-1 in the influenza life cycle, MDCK cells were infected with a selectable retroviral Cav-1 RNAi vector carrying a puromycin-resistance gene (RVH1-Puro-Cav-1) as well as control RVH1-Puro alone [20]. We found that the Cav-1 content decreased gradually to 25% of the value in wild-type MDCK at 14 1, 1, 10). Notably, the experiments revealed similar results with an average decrease of influenza titres to 57.3% of wild-type levels ( Fig. 1). The statistical analysis of nine independent experiments revealed that the 1.5-3 fold reduction in titres observed is highly significant (paired t-test, >99% confidence, p > 0,01) When cells stably transduced with control virus vector devoid of Cav-1 interfering sequences (RVH1puro) were infected with influenza A virus (m.o.i. = 10) titres of released virus was affected only marginally. Thus, we conclude that Cav-1 reduction in MDCK is correlated with a decrease in influenza virus progeny. This suggests, that Cav-1 directly or indirectly affects the human influenza virus life cycle in MDCK cells.

A dominant-negative Cav-1 mutant decreases influenza A virus titres in MDCK cells
A dominant-negative Cav-1 mutant has been described which functionally inactivates caveolin-1 upon binding [21]. The mutant carries a F92A/V94A double mutation in the scaffolding domain (SD) of canine caveolin-1. Expression in rat adipose and COS-1 cells has been shown to interfere with the interaction of Cav-1 with the insulin receptor and impairs receptor function.
To confirm our results from knock-down experiments, we investigated the effect of expression of the dominantnegative SD mutant and over-expression of wild-type caveolin-1 on virus production. Expression efficiency Analysis using a paired t-test (n = 9) revealed that the 1,5 to 3 fold reduction in titres compared to MDCK control cells is statistically highly significant (p > 0.001).  could be monitored easily, as endogenous and transfected, recombinant caveolins differ in their mobility in SDS-PAGE due to a C-terminal myc-tag (Fig. 2 bottom). Cav-1 appeared in two isoforms, with molecular weights of 21 and 24 kD, respectively [22,23]. Expression efficiencies ranged from 7-29% (SD) and 20-50% (wt-Cav-1) with respect to endogenous Cav-1 level. Provided that the myc-tag does not interfere with Cav-1 antibody detection and assuming a 1:1 interaction of SD mutant and endogenous Cav-1, sufficient competitor amounts should be available in successfully transfected cells. Next, transiently transfected MDCK and mock-transfected MDCK cells were infected with influenza A/PR/8/34 virus one day after transfection. 24 h later supernatants were used for titre determination (Fig. 2 top). To account for between-session-variations in cell culture, values were normalized to virus production from infected wt MDCK (100%). To exclude sensitivity of influenza infection to the actual transfection process, pEGFP-N1 transfected control cells were infected with PR8 virus in a control experiment. Strikingly, SD mutant expression in MDCK cells interfered with human influenza A virus replication and decreased the viral titres on average 1.6 fold to 62% of titres from wild-type MDCK (average of three independent experiments, standard deviation = ± 15,95). Compared to processed EGFP control cells, virus yield from Cav-1 wt-or SD-transfected MDCK cells was reduced considerably, which excludes that effects observed on virus production are derived from the transfection process (data not shown). This strongly suggests that interference with Cav-1 function in MDCK cells interferes with human influenza A virus replication. Interestingly, over-expression of wild-type Cav-1 also diminished influenza virus production, since viral titres reached only 56% ± 10.53 compared to non-treated MDCK (n = 3). Thus, surplus exogenous Cav-1 interferes with endogenous Cav-1 function, too. However, compared to the SD mutant twice the amount of Cav-1 wt is necessary to account for a comparable level of inhibition, as judged by Western Blot analysis.

Search for putative Cav-1 interaction partners
In order to elucidate the molecular basis of the interaction we scanned influenza A virus proteins for putative binding motifs. Cav-1 binds to various cellular proteins like membrane receptors, soluble or membrane-associated molecules [24] as well as several viral proteins and exerts functions in localisation, transport and cellular signalling (Table 1). Signalling is preceded by phosphorylation of Cav-1, which initiates events leading either to activation of specific signalling pathways [21] or maintenance of signalling-competent, yet inactive complexes [24]. A specific, lumenal domain termed caveolin scaffolding domain (CSD, aa 82-101) which resides adjacent to the region of membrane insertion, is responsible for specific protein binding in the vast majority of cases [24,25]. Two consensus sequences have been identified in phage-display experiments and in the primary structure of Cav-1 binding partners which were termed caveolin binding domains (CBD) [26]. CBDs have been recognized in cellular [24] and viral proteins ( Table 1). The consensus sequence comprises a run of 3 aromatic residues (W, F, Y) separated by a characteristic spacing (ΦxxxxΦxxΦ; ΦxΦxxxxΦ; where x stands for any amino acid and Φ for W, F, Y). Our screening for CBDs identified putative binding regions in HA, PB2 and M2 of influenza A virus. Especially, a region in the M2 channel protein turned out to be highly conserved among human influenza A viruses (Fig. 3B). The putative CBD overlaps with a loop/helical domain immediately following the M2 transmembrane region at the lumenal site of M2 (Fig. 3A). The CBD surrounds Cys 50, which is palmitoylated and faces the membrane allowing for insertion of the palmitoyl residue into the lipid bilayer. Thus, the CBD would be located favourably for interaction with the Cav-1 scaffolding domain [27,28]. Strikingly, compared to M2 of A/PR8/34 as a reference the CDB core motif (positions F47, Y52, F55) and immediately adjacent amino acid residues are completely conserved in 8 M2 sequences available for the pandemic influenza virus of swine origin A/2009 (H1N1) (Fig. 3C). Furthermore, motif conservation is observed in a former H1N1 strain appearing in 1977, but homology is restricted to the aromatic core and to a lesser extent to adjacent residues. Surprisingly, in M2 of influenza A/ 1918 the CBD motif is not conserved, as its third aromatic residue phenylalanine is changed to leucine, a residue commonly found in the M2 of avian influenza A viruses at that position (G.-V. Hemgård and M. Wirth, unpublished observation).

Competition of Cav-1 binding with M2 affects production of influenza A/PR/8/34
These hints prompted us to investigate the effect of M2 over-expression on the influenza A virus life cycle in MDCK cells. We hypothesized, that surplus M2 fusion protein may reduce the concentration of available, functional Cav-1 by complexing. To monitor M2 protein levels and localization we generated mammalian expression vectors containing cDNAs for fusion proteins of M2 (A/ PR/8/34) with desRedExpress, a red fluorescent, tetrameric protein (pM2PR8DsRed) or EGFP (enhanced green fluorescent protein) (pM2PR8_EGFP) and transfected purified DNA into MDCK cells. Expression levels and localization of the fluorescent proteins were followed 1 and 2d after transfection. The transfection efficiency (ratio of fluorescent/nonfluorescent cells) ranged between 10 and 15%. M2 fluorescent fusion proteins initially were found in the cytoplasma and started to localize at the plasma membrane at day 1 post transfection. As expected M2DsRed and M2EGFP localization did not differ from localization of M2 after infection with A/PR8/   (Fig. 4). Interestingly, M2 expression in infected MDCK decreased viral titres to 40% (M2DsRed) and 85% of (M2EGFP) of the level of non-transfected cells. Thus, M2 over-expression interferes with human influenza virus propagation, presumably by competing with endogeneous M2 for Cav-1 interaction.

The M2 matrix protein of human influenza A interacts with Cav-1
To verify the predicted M2/Cav-1 binding, pull-down as well as co-immunoprecipitation experiments were carried out. For pull-down experiments, biotinylated peptides carrying the putative CBD of M2 or a mutant CBD with alanines instead of the motif's core aromatic residues were incubated with cell lysates and complexes were processed as specified in Material and Methods (Fig. 5A).
Results from two independent experiments show that the M2 CBD-peptide indeed pulls down caveolin-1, while the alanine-CBD mutant exhibits a strongly reduced tendency to interact with Cav-1. These results indicate that M2 of influenza A/PR/8/34 indeed exhibits at least one caveolin-binding domain.
To confirm data of the pull-down experiments, coimmunoprecipitation experiments were performed using NIH3T3 or MDCK cells after transfection of pEP24c, an expression vector containing M2 PR/8 [29] or a vector harbouring fusion protein of M2 with fluorescent marker EGFP. 24 h later cell lysates were prepared in the presence of octylglucoside, a detergent that disrupt lipid rafts, as described previously [13]. In the first series of experiments, polyclonal anti-Cav-1 antibodies were used to pull-down Cav-1 complexes from lysates. Precipitated complexes were probed for the presence of M2 after Western Blot and immunostaining. In these experiments, the Cav-1 antibodies clearly pulled down a complex that contained a M2 from pEP24c transfected MDCK cells (Fig. 5B, a) or the M2 fusion protein from pM2PR8-EGFP transfected MDCK or NIH3T3 cells (Fig. 5B, b) as well as infected MDCK cells (Fig. 5B, c left panel). In the second experimental setting, vice versa, monoclonal anti-EGFP antibodies were used to precipitate M2 binding partners and a rabbit anti-Cav-1 antibody was used to probe for the presence of caveolin (Fig 5B, c right panel). These types of experimental settings identified M2 complexed with Cav-1 and vice versa in both cell lines, NIH3T3 and MDCK. Thus, the results suggest that M2 has the capability to interact directly or indirectly with caveolin-1 in different cell lines. With respect to the type of interaction, it is notable, that caveolin-1 as well as M2 have been reported to bind cholesterol via cholesterol specific recognition domains [30,31]. This prompted us to investigate, whether cholesterol is involved in the M2/caveolin-1 interaction. For that purpose methyl-β-cyclodextrin (MβCD) was used to deplete cell lysates from cholesterol before co-immunoprecipatation (Fig. 5B b and 5c). Interestingly, in pM2PR8-EGFP transfected NIH3T3 cells as well as in PR8 virus-infected MDCK cells, signals from co-immunoprecipated proteins decreased to a certain extent, if cholesterol was removed from the lysate before pull-down. These findings imply, that cholesterol seems to support the interaction of M2 with caveolin-1.

Discussion
Viruses recruit the cellular machinery to support their own multiplication and elicit an early host response to overcome the unwanted viral invaders. In our contribution we investigated the ability of caveolin-1, a multifunctional protein, to interact with components in the influenza A virus life cycle and to interfere with influenza A virus production. Cav-1 represents an organizing element at the plasma membrane and serves on localization and accumulation of proteins in lipid rafts and transmission of signalling events [24]. Furthermore, the protein contributes to intracellular cholesterol transport and has been identified as the main determinant of caveolae, invaginations of the plasma membrane used for entry of molecules and particles into the cell.
Based on previous findings of Cav-1 involvement in the late retroviral life cycle [13] we investigated the influence    [20]. This question may be answered in a Cav-1 (-/-) MDCK cell line, which yet has to be established. Data from knock-down experiments in MDCK were supplemented by transfection of a dominant-negative Cav-1 mutant as well as Cav-1 over-expression, which decreased viral yields by 38-44%. The results are reminiscent of experiments of Nystrom et al. who observed impairment of the insulin signalling pathway upon expression of both, the dominant-negative Cav-1 mutant and the over-expressed Cav-1 wt cDNA as well [21]. Finally, competition with M2 fusion proteins impaired virus replication, too.
Taken together Cav-1 supports virus multiplication in MDCK, but the cellular pathway directing this Cav-1 property is not known. It is conceivable, that the cellular protein level of Cav-1 is important for the outcome, as it has been suggested for Cav-1 involvement in the insulin pathway [21].
Hints for the molecular basis of influenza virus/Cav-1 interaction may come from other viruses which co-opt Cav-1. It is evident that individual stages in the various viral life times are affected and different roles are allocated to Cav-1 as well (Table 1). For example, the CBD region in the HIV-1 gp41 transmembrane protein can permeate membranes and is supposed to augment the fusion step upon virus entry. Remarkably, respiratory syncytial virus (RSV), induces Cav-1 phosphorylation, which results in intracellular relocation of proteins during the paramyxovirus life-cycle. In several cases, Cav-1 functions in positioning of viral proteins to intracellular membranes (Rotavirus, SARS) or specialised regions of the plasma membrane like lipid rafts (retrovirus MLV).
To understand the molecular basis of the Cav-1 contribution to influenza A virus propagation we focussed on Cav-1 interactions mediated by the caveolin-scaffolding domain (CSD, aa 81-102) [25]. Database searches and subsequent peptide pull-down assays in combination with co-immunoprecipitation experiments suggested binding of caveolin-1 to M2 presumably to a motif in the M2 protein fitting the CBD consensus [26]. Strikingly, the motif is shared in M2 of nearly all human influenza A viruses. M2 functions within the viral life cycle as a viroporin with proton channel activity that is crucial in the entry phase [1] and as a maturation cofactor in virus budding. The cytoplasmic tail is implicated in M1 binding and facilitates virus assembly and production [2][3][4]35]. Furthermore, Schroeder et al. showed that avian M2 is a cholesterol-binding protein [31]. Most avian influenza A viruses contain two cholesterol recognition motifs (CRAC I, CRAC II) in close vicinity to the transmembrane domain in the cytoplasmic region of M2 [31,36]. Thus, cholesterol-binding and palmitoylation in combination with a short transmembrane region may direct M2 to the raft periphery in membranes and may promote clustering and merging of rafts which is then followed by the pinching-off of avian viruses [31].  [30]. Notably, to some degree Cav-1 binding to M2 is sensitive to the cholesterol depletion (this investigation). Preliminary results of mutagenesis as well as localization experiments indicate a certain role of the M2-CBD in M2 transport and localization (unpublished observations). Taken together our results demonstrate, that Cav-1 exerts an influence on influenza A virus replication and data imply that the binding of Cav-1 to the matrix protein M2 is involved. However, which function or pathway in MDCK cells actually is triggered via Cav-1 interaction with M2, remains to be determined.

Conclusion
The appearance of the aggressive bird influenza (H5N1), the 2009 outbreak of a pandemic influenza (H1N1) of swine influenza origin, and the recent occurrence and rapid dissemination of oseltamivir-resistant human influenza strains are motors that have accelerated the search for new antiviral targets and agents within the last time [37][38][39]. The investigation of cellular mechanisms involved in 'early' and 'late' viral processes and the identification of cellular actors provides a means to interfere with viral strategies. With this respect, the observed Cav-1/M2 interplay may represent a new, conserved target for e.g. therapeutic intervention with circulating and newly emerging strains of human influenza A virus. Thus, appli-cation of high-throughput screening of compound libraries will follow target identification and may result in a new antiviral agent, as exemplified for a cellular target involved in the late retroviral life cycle [40]. When this manuscript was in preparation Zhou et al. reported binding of a cytoplasmic fragment of M2 from human influenza to Cav-1 in an in vitro assay based on a Cav-1 protein fragment expressed in E. coli and CBDdependent perinuclear co-localization upon expression in CHO cells [41]. However, no experiments on the functional importance of M2/Cav-1 were performed in this investigation.

Lysis of cells
Lysates were prepared as described previously [13].

Plaque Assay
Influenza A/PR/8/34 titre was determined by plaque assay on MDCK cells. PBS-washed MDCK were inoculated with 500 μl of virus dilution for 1-2h at 37°C. Cells were covered with 2 ml of MEM medium containing 1% purified agar (Oxoid, England) and 1-2 μg/ml trypsin (Sigma). After three days incubation at 37°C, plates were stained with 0.03% neutral red staining to facilitate plaque counting.

SDS-PAGE and Western Blot
Protein concentrations were determined using the BCA kit (Pierce). 5 μg total protein was separated on a vertical 12% separating gel. Subsequently, proteins were transferred to PVDF membranes using a Transblot™ Semi-dry transfer cell (Bio-Rad). After blocking for 1 h (0.2% CA blocking reagent, Applichem) immunostaining was performed with primary antibody followed by 4 washing steps (TBS 0,02% Tween 20) and addition of the secondary antibody at appropriate dilution. The blots were developed with chemoluminescent substrate (Supersignal Femto West, Supersignal Pico West, Pierce). The band intensities were quantified using QuantityOne software (Bio-Rad) and ImageJ.

Inhibition and competition experiments Generation of Cav-1 knock-down MDCK using retrovirally mediated RNAi
The recombinant retroviral vectors were produced from 293T triple transfection of pCMV1MLVGP1, encoding MLV gagpol, pVSV-G, pRVH1-Puro-Cav-1 encoding a shRNA for Cav-1 inhibition and a puromycin resistance gene, as described [20]. For knock-down MDCK (60%-80% confluency) were infected with the respective shRNA retroviral vectors in the presence of 4 mg/ml polybrene for 48 hours. Puromycin-resistant clones were pooled and further analysed 10-27 days after infection.

Competion with M2 fusion proteins
Plasmids pM2PR8_EGFP or pM2PR8DsRed were transiently transfected into MDCK cells by lipofection. The cells were infected with influenza A/PR/8 virus 1 day after transfection. Virus titres were determined from supernatants after additional 24 h of incubation at 37°C.