Phospholipase D Facilitates Efficient Entry of Influenza Virus, Allowing Escape from Innate Immune Inhibition^*

Chemical Synthesis and Purification

PLD isoenzyme-preferring inhibitors were synthesized and characterized as described in detail previously (12, 13).

Mass Spectrometry-based Lipid Analysis

For analysis of cellular lipids, A549 cells were serum-starved for 1 h, infected at 1 m.o.i. with influenza A/California/04/2009 or other indicated strains, and then maintained in 6-well plates for the indicated treatment times. Mock-infected and influenza A-infected cells were harvested at the indicated times after infection and administration of PLD inhibitors. To extract phospholipids, ∼1 × 10⁶ cells were washed with ice-cold PBS and scraped into 1 ml of ice-cold PBS, and aliquots were taken for protein analysis. After centrifugation and PBS aspiration, the pellet was extracted by a modified Bligh and Dyer (15) method using 800 μl of ice-cold 0.1 n HCl, CH₃OH (1:1) and 400 μl of ice-cold CHCl₃. Following vortexing for 1 min at 4 °C, phases were separated by centrifugation (18,000 × g for 5 min at 4 °C). The lower organic layer was isolated, synthetic odd-carbon phospholipids (four per phospholipid class) were added as standards, and the solvent was evaporated. The resulting lipid film was dissolved in 100 μl of 58:40:2 isopropanol, hexane, 100 mm NH₄COOH (aqueous) (mobile phase A) (16). Samples containing n-butyl alcohol were extracted similarly, and 5 μl of 10 μg/ml 32:0 phosphatidylmethanol was added as an internal standard. Mass spectrometric analysis and quantitation were performed essentially as previously described (17–19). An MDS SCIEX 4000QTRAP hybrid triple quadrupole/linear ion trap mass spectrometer (Applied Biosystems, Foster City, CA) coupled to a Shimadzu HPLC system (Shimadzu Scientific Instruments, Inc., Columbia, MD) consisting of an SCL 10 APV controller, two LC 10 ADVP pumps, and a CTC HTC PAL autosampler (Leap Technologies, Carrboro, NC) was utilized for the analyses. Phospholipids were separated on a Phenomenex Luna Silica column (Phenomenex, Torrance, CA) (2 × 250 mm, 5-μm particle size) using a 20-μl sample injection. A binary gradient consisting of 58:40:2 isopropanol, hexane, 100 mm NH₄COOH (aqueous) (mobile phase A) and 50:40:10 isopropanol, hexane, 100 mm NH₄COOH (aqueous) (mobile phase B) was used for the separation. Instrumentation parameters and solvent gradient were described previously (18). Time courses were performed in two independent experiments, each in triplicate. Statistical analysis was performed by two-way ANOVA with Bonferroni's post-test.

Spatial Infection Model

A spatial infection model for testing compound efficacy and influenza replication was adapted from Lam et al. (20). Monolayers of A549 cells were grown on chamber slides. A semisolid overlay made of agar and growth medium was allowed to cure on top of the monolayer. A Pasteur pipette was used to create a consistent reservoir in the overlay, and influenza virus was introduced through this reservoir. Cultures were then incubated at 37 °C for the duration of the infection.

RNA Interference

A549 cells were transfected with 100 nm siRNA (Ambion) specific to PLD isoforms using NeoFX (Ambion) and were subsequently infected with 1 m.o.i. influenza A/Brisbane/59/2007 (H1N1) for 24 h. Innate immune proteins were knocked down using 100 nm siRNA and the Neon Transfection System (Invitrogen). Knockdown was confirmed with gene-specific TaqMan assays and the 2^ΔΔCt method using GAPDH for normalization on Western blots to confirm loss of protein.

Immunofluorescence and Live Cell Imaging

Samples were fixed in 4% formaldehyde, permeabilized with 0.3% Triton X-100, and then exposed to antisera targeting proteins of interest and corresponding fluorescent secondary antibodies alongside DAPI to visualize nuclei. Spatial infections were imaged and processed using Nikon C1Si and NIS Elements software. Confocal images were captured with a Zeiss LSM 510 NLO Meta and analyzed with Zeiss Zen 2011 and NIH ImageJ software. To determine colocalization, the PSC (Pearson-Spearman Correlation) colocalization plug-in was used to calculate Pearson's and Spearman's coefficients. Briefly, nucleoprotein (NP) and PLD2 signal channels were merged. Each Z-slice was projected to maximum intensity. NP-positive cells were masked, and the mask was used to determine whether positive signal for each channel colocalized. Each mask was in excess of 20,000 pixels to ensure a robust data set. Live cell imaging was performed using spinning disc laser-scanning confocal microscopy carried out on a Marianas spinning disk confocal imaging system (Intelligent Imaging Innovations/3i, Denver, CO) comprising a CSU22 confocal scan head (Yokogowa Electric Corp.) and solid state lasers with wavelengths of 488 and 658 nm configured on a motorized Axio Observer Z1 inverted microscope (Carl Zeiss MicroImaging) equipped with a Definite Focus system (Carl Zeiss) and spherical aberration correction optics. Time-lapsed three-dimensional imaging was performed at 37 °C in 5% (v/v) CO₂ in a humidified atmosphere using an environmental control chamber (Intelligent Imaging Innovations/3i), and images were acquired using a Plan-Apochromat 63× 1.4 numerical aperture oil objective on an Evolve 512 EMCCD camera (Photometrics, Tucson, AZ) using Slidebook 5.5 software (Intelligent Imaging Innovations/3i).

Animal Infection

All animal studies were approved by the St. Jude Children's Research Hospital Animal Care and Use Committee (Protocol number 098) following the guidelines established by the Institute of Laboratory Animal Resources approved by the Governing Board of the United States National Research Council. Female C57BL/6 mice (8–10 weeks old) were anesthetized and infected with the indicated doses and strains of influenza A virus. Mice were weighed and monitored daily; tissues were collected at the specified times and kept at −80 °C until analysis. For drug treatment, mice were administered 13 mg/kg VU0364739 or vehicle (10% DMSO, 90% PEG) every 8 h from day −1 to day 3 after infection or every 12 h during the H7N9 study. All procedures were performed according to an institutionally approved Institutional Animal Care and Use Committee protocol, which includes a requirement for daily observation and euthanasia upon detection of severe moribundity.

Titering

Infected animal lungs were titered using a standard plaque assay. Supernatant from infected cultures was titered using a traditional TCID₅₀ assay, and immunofluorescence was used to enumerate the number of infected cells in cultured samples.

Host Gene Expression

RNA was isolated from lungs and used in reverse transcription-PCR. The cDNA was then used in gene-specific TaqMan assays to determine host gene expression, and the differences in expression were quantified using the 2^ΔΔCt method. The same amount of RNA was used in each reaction, and samples were run in triplicate.

Statistical Analysis

Quantitative data are presented as mean ± S.E. of at least three independent experiments.

Influenza Infection Potentiates PLD Catalytic Activity, and That Activity Is Attenuated by VU0364739 Treatment

To determine the effect of influenza infection on PLD activity, human adenocarcinomic alveolar basal epithelial cells (A549) were infected with 1 m.o.i. human influenza strain A/California/04/2009 (H1N1) for 6 h in the presence of 0.6% n-butyl alcohol. Instead of the typical biological nucleophile water, in the presence of a primary alcohol, PLD will utilize the alcohol to produce a metabolically stable transphosphatidylation product, phosphatidylalcohol, as an alternative to hydrolysis producing PA. Fig. 1A illustrates that influenza infection markedly stimulated PLD activity as measured by the increase in phosphatidylbutanol production compared with cells that were not exposed to influenza virus (mock-infected). Treatment with a PLD2-preferring inhibitor, VU0364739, was sufficient to block the influenza-mediated PLD activity increase. As shown in Fig. 1B, both PLD1 and PLD2 were active during the viral entry phase, and complete ablation of catalytic activity required knockdown of both PLD1 and -2. This result suggests independent roles for both isoenzymes in entry.

Confocal microscopy-based experiments were conducted to assess changes in the accumulation and localization of PLD during an influenza infection. A549 cells were grown on chamber slides and infected with 1 m.o.i. influenza A/California/04/2009 (H1N1). Samples were fixed at 0, 30, 90, and 360 min postinfection and probed for influenza NP and PLD2. PLD2 began to accumulate at the periphery of the cell as early as 30 min postinfection (Fig. 1, C, D, and E). At the same time point, very low levels of NP, also occurring at the outer reaches of the cell (Fig. 1E), were detected. PLD2 staining continued to intensify 90 and 360 min postinfection (Fig. 1D), and the observed PLD2 signal was increasing as well as moving from the cytoplasm to a perinuclear region (Fig. 1E). Influenza NP signal also intensified 360 min postinfection, and NP was moving from the cytoplasm (30 and 90 min postinfection) toward the nucleus (Fig. 1, D and E). During the infection, PLD2 and NP were trafficking to similar subcellular locations. The extent of this colocalization was quantified. PLD2 and NP were increasingly colocalized at 30, 90, and 360 min postinfection (Fig. 1C) as measured by both Pearson's coefficient and rank correlation (Spearman). To further confirm that siRNA treatment was knocking down levels of PLD2 and that the PLD2 antibody was specific, we probed for PLD2 expression after infection of cells electroporated with control or PLD2 siRNA. We imaged more than 75 cells for each condition and found a consistent loss of 50% or more of stainable PLD2 in the siRNA-treated cells (1F). Together these data suggest that PLD activity is stimulated by influenza infection, endogenous PLD is redistributed during the infection, and the accumulation of PLD is occurring in the same subcellular location as influenza NP.

PLD Is a Targetable Host Factor That Facilitates Efficient Influenza Infection

To determine the role of PLD-mediated PA production in influenza infection, we used a spatial infection model in the presence of 0.6% n-butyl alcohol with tert-butyl alcohol used as a negative control. This assay utilizes the use of primary alcohols as preferred nucleophiles in a PLD transphosphatidylation reaction as an assessment of exclusively PLD-produced PA.

Twenty-four hours postinfection, infected cells were counted by anti-influenza NP staining (Fig. 2A). Infection with influenza strain A/Brisbane/59/2007 (H1N1), A/California/04/2009 (H1N1), or A/Brisbane/10/2007 (H3N2) resulted in fewer infected cells in the presence of n-butyl alcohol, indicating that blocking PLD production of PA substantially reduces the rate of cell to cell transmission of infection. Notably, averting the production of PLD-generated PA to PLD-generated phosphatidylbutanol by use of primary alcohol did not entirely prevent influenza infection but did significantly decrease the rate of infectious spread within the cultures.

To determine whether PLD1 or PLD2 was preferentially required for efficient influenza infection, RNAi was used to selectively knock down individual PLD isoforms. A549 cells were transiently transfected with siRNA that targeted PLD1 or PLD2 24 h prior to a 5 m.o.i. influenza infection of A/Brisbane/59/2007 (H1N1). Twenty-four hours postinfection, a TCID₅₀ assay was used to measure influenza virus titers in culture supernatants (Fig. 2B). RNAi of either PLD1 or PLD2 inhibited influenza replication. However, the magnitude of the effect was greater with PLD2 knockdown. Although each isoform appears to contribute to efficient influenza replication, the stronger inhibitory effects seen after PLD2 knockdown led us to focus on PLD2 inhibition and its effect on the host and viral replication for subsequent studies.

The development and optimization of the PLD inhibitors have been described (12) with the PLD2-preferring inhibitor VU0364739 having a 75-fold selectivity for PLD2 over PLD1 (13). Based on previous studies, VU0364739 was used at 10 μm in all cell-based assays unless otherwise indicated (21). To demonstrate the importance of PLD2 during influenza replication, another RNAi experiment was conducted while treating the PLD2-deficient cells with PLD2 inhibitor VU0364739. Viral titer was measured by a TCID₅₀ assay after a 24-h, 5 m.o.i. influenza A/Brisbane/59/2007 (H1N1) infection. Pretreatment with VU0364739 for 1 h caused a dramatic decrease in viral replication (Fig. 2C). Similar reductions in titer were also observed when PLD2 was knocked down by RNAi in vehicle-treated samples. Finally, the specificity of this compound was confirmed by combining the apparent marginal, non-statistically significant additive nature of RNAi with VU0364739 treatment.

These observations were extended to establish the importance of PLD2 as a host factor required for infection by other clinically relevant strains of influenza. Pretreating cells for 1 h with VU0364739 resulted in a significant decrease in the number of infected cells measured in the spatial infection assay with multiple viral strains at both 6 and 24 h postinfection (Fig. 2D). Representative images of the spatial infection model assay (Fig. 2E) illustrate the reduction in influenza spread observed after VU0364739 treatment. PLD2 inhibition by VU0364739 effectively protected A549 cells from cell to cell spread of influenza, further demonstrating that PLD2 is required for efficient influenza infection and spread. A time-of-addition study was conducted where inhibitor was added over a time course spanning 2 h before infection to 4 h postinfection. Protection from infection was observed at all times of PLD inhibitor addition (data not shown).

Inhibition of PLD Activity Dramatically Reduces Influenza Replication

The spatial infection model assay data were validated using the more traditional TCID₅₀ assay to assess viral reproduction in vitro. A549 cells were treated with 10 μm VU0364739 for 1 h before infection, and the cells were then infected with the indicated doses and strains of influenza. At the indicated time points, infectious supernatant was removed from the A549 cells and titrated on Madin-Darby canine kidney cells to measure viral reproduction. Using either a low m.o.i. (Fig. 3A) or a high m.o.i. (Fig. 3B) of influenza A/Brisbane/10/2007 (H3N2), a lower viral reproduction was observed when PLD2 was inhibited. In the case of the low m.o.i. infection, viral titers were first noted to be lower at 16 h postinfection, and the effect was sustained 24 h postinfection. Using the high m.o.i. model, when A549 cells were treated with VU0364739, a significant reduction in viral reproduction was noted 12 h postinfection and lasted through at least 24 h.

The H3N2 strain used in our previous experiments is considered a low pathogenicity strain of influenza. To determine whether host PLD is required for high pathogenicity and quickly replicating strains of influenza, VU0364739-treated A549 cells were infected with 0.01 m.o.i. influenza rg-A/Vietnam/1203/2004 (H5N1), and viral reproduction was assessed during a more severe infection. Twenty-four hours postinfection, a massive decrease in viral titer was observed when PLD2 was inhibited during H5N1 infection (Fig. 3C). Subsequent investigation to determine whether host PLD2 activity is required for infection was conducted using a recently emergent virus with pandemic potential, influenza strain A/Anhui/01/2013 (H7N9). After 1 h of 10 μm VU0364739 pretreatment, viral reproduction was effectively blocked 24 h postinfection (Fig. 3D). Viral titer was near the limit of detection when PLD catalytic activity was inhibited, consistent with PLD2 being a host factor required for low and high pathogenicity influenza infections.

Using the same model, in vitro dose-response experiments were performed using VU0364739 to determine the efficacy of the PLD2 inhibitor during H5N1 or H7N9 infection. A549 cells were treated for 1 h with varying concentrations of VU0364739 before a 0.01 m.o.i. infection of either influenza rg-A/Vietnam/1204/2004 or A/Anhui/1/2013. Supernatant was then used to measure viral reproduction in a TCID₅₀ assay 24 h postinfection. In the case of the H5N1 infection, the IC₅₀ of VU0364739 was calculated by non-linear regression analysis to be 2.1 μm (Fig. 3E). Similarly, the IC₅₀ of VU0364739 was found to be 3.4 μm when cells were infected with an H7N9 influenza strain (Fig. 3F). These IC₅₀ values are consistent with in vitro dose-response experiments using influenza A/Brisbane/59/2007 (H1N1), A/Brisbane/10/2007 (H3N2), and A/California/04/2009 (H1N1) as well (data not shown). Based on these results, it was determined that inhibition of PLD2 can significantly lower influenza reproduction in vitro and that the decrease in viral titer occurs in a dose-dependent fashion.

PLD2-preferring Inhibitor VU0364739 Reduced Viral Titer in Mouse Lungs and Delayed Mortality during Lethal H7N9 Influenza Infection

Having shown that abrogation of PLD2 activity leads to a decrease in viral spread and reproduction in vitro, we wanted to determine whether the loss of PLD2 could reduce viral titer in vivo. Female C57BL/6 (B6) mice were treated intraperitoneally with dilutions of PLD2-preferring inhibitor VU0364739 every 8 h from day −1 to 8 h after infection with 1 LD₅₀ (4000 EID₅₀) influenza A/Puerto Rico/8/1934 (H1N1) (PR8) administered intranasally on day 0. Because of solubility issues and observed acute vehicle toxicity, a long term survival study with optimal dosing to achieve the therapeutic in vivo concentration continuously was not feasible at this time, and viral titer was used as a readout to determine the role of PLD2 in a mouse model of influenza infection. Viral titer in lungs decreased significantly with PLD2 inhibitor VU0364739 treatment (Fig. 4A), and these protective effects were dose-dependent. Additionally, to determine whether PLD2 inhibition could lead to lower viral titers in a more chronic situation, mice were given 13 mg/kg VU0364739 every 8 h intraperitoneally from day −1 to day 3 after infection. Animals treated with VU0364739 had significantly less viral replication in their lungs (Fig. 4B) on day 3. Concomitant with the decrease in viral titers, 8 h after PR8 influenza infection, significant up-regulation of the innate immune proteins Mx1, OASL (2′-5′-oligoadenylate synthetase-like protein), and IFITM3 was observed when PLD2 was inhibited (Fig. 4C), indicating that the early immune response may be an important part of the mechanism of protection when mice are treated with VU0364739. IFITM3 has recently been described as a human restriction factor for influenza infection (22).

To identify survival benefits conferred by the observed reduction in viral titers, a study was conducted dosing mice with 13 mg/kg VU0364739 every 12 h from day −1 to day 3 of a lethal influenza A/Anhui/1/2013 (H7N9) infection (Fig. 4D). Relative concentrations of the PLD inhibitor in various tissues of interest over time are presented in Fig. 4E. Administration of the PLD2-preferring inhibitor resulted in a modest yet significant increase in survival and in addition delayed mortality (Fig. 4D). Although 80% of the mice administered VU0364739 eventually succumbed, death occurred considerably later after infection (compared with influenza-infected mice administered vehicle), clearly providing an extended window for further supportive therapy in a clinical setting. This demonstrates that by inhibiting PLD activity in the mouse viral replication is decreased and that antiviral responses are up-regulated, leading to some protective benefits during a lethal influenza infection. No significant toxicity or neurological impairment was noted in rats receiving identical treatment (Tables 1 and 2). This inhibitor is a preclinical compound that has not been optimized for pharmacokinetic or pharmacodynamic properties, and yet it demonstrates robust effects on viral spread and reproduction.

PLD2 Inhibition Alters Endocytosis Kinetics and Aggregation of Endocytic Proteins

Inhibition of PLD function has been shown to decrease uptake of ligands in various systems (9, 23, 24). Given that perturbation of influenza virus trafficking during the early infection process can lead to degradation of the entering virus (11), it was hypothesized that PLD inhibition during influenza infection led to alterations of entry events that occur in the first minutes of infection. To support this hypothesis, the effect of PLD2 inhibition on normal endocytosis rates was assessed using the established transferrin uptake model, a classic demonstration of a clathrin-dependent trafficking process.

A549 cells pretreated with VU0364739 were labeled with fluorescent transferrin at 4 °C for 1 h and then placed into a heated microscope for live cell imaging. Images were recorded for 1 h, and the frame and time of the transferrin fluorescence disappearance was noted (Fig. 5, A and B). Cells treated with vehicle control were able to take up, traffic, and degrade transferrin by the 14th frame, corresponding to an average time of 26 min after warming. In contrast, cells treated with VU0364739 took ∼56 min to process the fluorescent ligand. These data indicate that inhibiting PLD2 activity with VU0364739 alters trafficking kinetics by extending the endocytosis process rather than acting as a strict blockade.

Influenza entry is not entirely dependent on clathrin-mediated endocytosis, but it is widely accepted that the majority of incoming viruses enter via clathrin-dependent events primarily by the de novo formation of clathrin-coated pits on the cell surface after exposure to influenza virus (25). After entry, influenza virus needs to be properly trafficked from the membrane to the nucleus. Along the way, the endosome is acidified (26), and the hemagglutinin protein undergoes a conformational change that creates a fusion pore between the viral and vesicle membranes through which viral ribonucleoproteins gain access to the cytoplasm, eventually entering the nucleus to initiate new virus production. The accumulation of trafficking-associated proteins was visualized by confocal microscopy in A549 cells during a 0.05 m.o.i. influenza A/California/04/2009 (H1N1) infection. Signal intensity and particle size were gated to determine protein accumulation. As proteins accumulate, the brightness and size of the fluorescence increases on a per cell basis. Clathrin recruitment was inhibited 50 min postinfection and remained significantly lower (up to 80 min postinfection) with VU0364739 administration (Fig. 5C). During the first phase of viral entry, Rab5 accumulated on the early endosomes; however, in the presence of VU0364739, much less Rab5 accumulated 10 min postinfection (Fig. 5D). Similarly, VU0364739 treatment led to a significant reduction in recruitment of the late endosome marker CD63 90 min postinfection (Fig. 5E). Vesicular trafficking is also important for key elements of the late stages of viral replication, and defects in Rab11 accumulation associated with viral protein trafficking to the membrane were consistently observed (data not shown). These data indicate that when PLD2 is inhibited during an infection the normal cascade of protein accumulation required for proper endosomal maturation is disrupted, leading to inefficient trafficking of incoming virus particles, indicating that PLD2 is a host factor required for the efficient trafficking of influenza virus once within the cell.

Protection by PLD2 Inhibition Requires Intact Innate Antiviral Signaling

To assess whether the innate antiviral response was required for protection when cells were treated with the PLD2 inhibitor, an RNAi screen of essential innate immunity proteins in vitro was conducted. A549 cells were electroporated with 100 nm gene-specific siRNA and infected with 5 m.o.i. influenza A/Brisbane/10/2007 (H3N2) for 24 h; a decrease in protein expression was confirmed by Western blot analysis (data not shown). Viral titer was then measured using a traditional TCID₅₀ assay. A549 cells treated with control siRNA were protected by VU0364739 administration (Fig. 6A), showing a decrease in titer of 1.5 log units. However, VU0364739 treatment was not sufficient to protect cells transfected with siRNA targeting IRF3, Rig-I, or MxA. These three proteins are known to be vital to the innate defense against influenza. Interestingly, cells transfected with siRNA targeting the transcription factor IRF7 were still protected by VU0364739 treatment but to a lower degree compared with cells transfected with control siRNA (0.9-log unit decrease in viral titer). These experiments demonstrate that altering trafficking kinetics with a PLD2 inhibitor alone was insufficient for complete protection from influenza infection. Cells need an intact antiviral response to attack and neutralize the invading virus, suggesting that PLD2 plays a critical role in the temporal-spatial regulation of endocytosis and innate antiviral sensing/defense during influenza infection. Disruption of PLD2 regulation of these events affords the host cell more time to detect the virus and mount an immune response.

One potential benefit of the delayed trafficking kinetics is the greater window allowed for innate immune interference during influenza infection. A key component of the innate response in the respiratory epithelium is MxA, a cytoplasmic dynamin-like GTPase, which is transcriptionally induced to high levels as part of the type I IFN response but is also found at basal levels in resting cells. MxA can interfere with early virus replication events by binding to viral NP (27). MxA recruitment was measured in A549 cells postinfection by the same manner used to measure the endosomal markers. When PLD2 was inhibited during infection, MxA was observed to accumulate in foci at significant levels as early as 30 min postinfection (Fig. 6, B and C). These data support a model whereby the slower rate of endocytosis following PLD2 inhibition allows cells more time to detect and respond to incoming influenza virus. Afforded this additional time, PLD inhibitor-treated host cells were able to mount an immune response of greater magnitude and efficacy to neutralize the incoming virus, thus promoting survival of the cell and protection from degradation of the invading virus.