Artemisinins target the intermediate filament protein vimentin for human cytomegalovirus inhibition

The antimalarial agents artemisinins inhibit Cytomegalovirus (CMV) in vitro and in vivo , but their target(s) have been elusive. Using a biotin-labeled artemisinin we identified the intermediate filament protein vimentin as an artemisinin target, validated by detailed biochemical and biological assays. We provide insights into the dynamic and unique modulation of vimentin, depending on the stage of human CMV (HCMV) replication. In vitro , HCMV entry and viral progeny are reduced in vimentin-deficient fibroblasts, compared to control cells. Similarly, mouse CMV (MCMV) replication in vimentin-knockout mice is significantly reduced compared to controls in vivo , confirming the requirement of vimentin for establishment of infection. Early after HCMV infection of human foreskin fibroblasts vimentin level is stable, but as infection proceeds, vimentin is destabilized, concurrent with its phosphorylation and virus-induced calpain activity. Intriguingly, in vimentin overexpressing cells, HCMV infection is reduced compared to control cells. Binding of artesunate, an artemisinin monomer, to vimentin prevents virus-induced vimentin degradation, decreasing vimentin phosphorylation at Ser55 and Ser83 and resisting calpain digestion. In vimentin-deficient fibroblasts, the anti-HCMV activity of artesunate is reduced compared to controls . In summary, an intact and stable vimentin network is important for the initiation of HCMV replication but hinders its completion. Artesunate binding to vimentin early during infection stabilizes it and antagonizes subsequent HCMV-mediated vimentin destabilization, thus suppressing HCMV replication. Our target discovery should enable the identification of vimentin binding sites and compound moieties for binding. binding to vimentin at a series of increasing concentrations. The binding affinity ( K D ) of AS at equilibrium was determined to be 12.3 ± 1.7 µM. Data were fit to a single rectangular hyperbolic curve to calculate the K D value (see Materials and methods for details). The standard deviation was calculated from three independent measurements. The chemical structure of AS and deoxy artemisinin that lacks the endoperoxide bridge is provided. The endoperoxide bridge is circled in red.


Introduction
Repurposing of the antimalarial agents artemisinins for treatment of human cytomegalovirus (HCMV) attracted interest, fueled by clinical experience and safety data from malaria therapy (1-5). The in vitro activity of artemisinins against HCMV, but not against herpes simplex virus 1 (HSV1), indicates unique and selective antiviral activities (6). We and others reported that artemisinin-derived monomers (artemisinin, artesunate, artemether, artemisone) inhibit HCMV at µM concentrations while the dimeric versions are inhibitory at nM concentrations (1, 2,7,8). The endoperoxide bridge within the artemisinin trioxane pharmacophore is critical for compound activity (9,10), and its chemical disruption ("deoxy artemisinin") abolishes the anti-HCMV activity (2,11). Until now, an artemisinin-resistant HCMV has not been selected, suggesting virus inhibition primarily involves host-directed functions critical for virus replication (12,13). The mechanisms of HCMV inhibition by artemisinins are different from the DNA polymerase inhibitor, ganciclovir (GCV), as they inhibit GCV-resistant HCMV. The combination of artemisinins and GCV is synergistic against HCMV (7,14). The in vitro anti-HCMV activity of artesunate correlated with cell cycle stage (12); efficacious in contact-inhibited human foreskin fibroblasts (HFFs), but reduced in subconfluent HFFs. In contact-inhibited cells, HCMV induced cell cycle progression to G1/S at 24 hours post-infection (hpi), but artesunate reverted it to early G0/G1 and decreased virusinduced expression of cyclin-dependent kinases (CDKs 1, 2, 4). The cellular/microbial targets of artemisinins have been of major interest to several disciplines, including infectious diseases and cancer. Studies highlighted the complexity and promiscuity of these drugs towards multiple proteins in Plasmodium falciparum, but specific targets underlying the anti-infective or anticancer properties remain inconclusive (15,16). To better define the anti-HCMV activities of artemisinins and identify cellular functions targeted by these drugs, we synthesized a biotinlabeled trioxane semi-synthetically derived from artemisinin at the C10 position. Immunoprecipitation (IP) -mass spectrometry (MS) followed by molecular, biological, and biochemical studies confirmed the type III-intermediate filament protein vimentin as an artemisinin target, resulting in post-translational modification and stabilization of vimentin. Since artemisinins inhibit CMV in vitro and in vivo, we determined the requirement of vimentin for virus replication. Reportedly, vimentin enables HCMV trafficking into the nucleus early after infection (17). However, its functions during later stages of infection have not been studied. Our data show distinct roles of vimentin at different stages of HCMV replication. In the early stage of infection, vimentin level is stable, likely providing support for virus transport into the nucleus, but subsequently, HCMV strategy is to destabilize and degrade vimentin. This latter process requires several mechanisms, including vimentin phosphorylation and induction of calpain activity. Binding of artesunate to vimentin counteracts and reverses virus-induced vimentin degradation, primarily through decreasing its phosphorylation and inducing calpain activity, effects culminating in virus inhibition.

Identification of vimentin as artemisinin target
Trioxane C10 primary alcohol (1), derived from dihydroartemisinin (DHA), the active metabolite of all monomeric artemisinins, was coupled with carboxylic acid (2) to produce biotin-labeled trioxane, MW 552 (Fig. 1A). The biotin-labeled product will be referred to as "552". Lysates from non-infected human foreskin fibroblasts (HFFs) were treated overnight at 4 °C with DMSO, 552 (20 µM), or 552 with DHA as a competitor (200 µM), which was added 1 h prior to 552. 552-protein complexes were captured with streptavidin agarose beads followed by protein separation on SDS-PAGE and silver staining (Fig. 1B). Specific bands enriched in the 552 lane were cut and analyzed by mass spectrometry (MS). Vimentin was identified as one of the main 552-binding proteins (Table 1). To confirm the MS findings, immunoprecipitation (IP) with streptavidin agarose beads was performed in non-infected HFF cell lysates treated with DMSO, 552 (20 µM), or 552 with DHA as a competitor (200 µM). In the drug competition condition, DHA was added 1 h before 552. Lysates were incubated overnight, followed by Western blot with anti-vimentin antibody. Vimentin was detected in the 552-treated lysates, but not in the DMSO or DHA (competitor) treated samples (Fig. 1C). Next, equal quantities of purified Hisvimentin protein (100 ng) were incubated for 1 h with increasing concentrations of DHA, followed by incubation with 552 for 1 h. Proteins were blotted and probed with streptavidin-HRP. Competition with increasing concentrations of DHA resulted in reduced biotinylated adducts (Fig. 1D), while vimentin level was similar in all conditions. A surface plasmon resonance (SPR) assay was performed after immobilizing purified His-vimentin on CM5 sensor surface (Biacore T200, GE Healthcare).
Artesunate (monomeric artemisinin), and an inactive metabolite deoxy artemisinin that lacks anti-HCMV activity were tested at concentrations ranging from 0.93 µM to 33.3 µM along with at least two zero concentrations.
Artesunate showed fast association (ka) and dissociation (kd) rates at all tested concentrations based on each sensorgram (Fig. 1E). The binding affinity (KD) of artesunate binding to vimentin was calculated 12.3 ± 1.7 µM by steady-state affinity fitting with data at equilibrium (Fig. 1F). Deoxy artemisinin did not show any interaction with vimentin ( Fig. 1G &  H), indicating the endoperoxide bridge in the artemisinin pharmacophore is required for binding to vimentin.

Vimentin modulation during HCMV infection
To begin understanding whether binding of artemisinins to vimentin results in HCMV inhibition, we tested the effects of infection with two genotypically distinct strains of HCMV (TB40 and Towne) on vimentin expression and modulation. Vimentin level gradually and reproducibly reduced in infected HFFs. At 4 h post-infection (hpi) vimentin level was stable but starting at 24 hpi with HCMV-TB40 ( Fig.  2A), its level decreased. Vimentin is a highly phosphorylated protein, and its phosphorylation strongly correlates with its disassembly (18). We selected two sites to measure vimentin phosphorylation during infection -Ser55 and Ser83 (19). Ser55 is phosphorylated by cyclindependent kinase 1 (CDK1), which also recruits Polo-like kinase (PLK1) to phosphorylated vimentin-Ser55, resulting in PLK1 activation and further vimentin phosphorylation at Ser83 (20). The decrease in vimentin level during HCMV infection correlated with increased phosphorylation at the respective time points. Ser55 phosphorylation was increased from 24 hpi onwards, and Ser83 was phosphorylated at 48 and 72 hpi ( Fig. 2A).
Artesunate treatment during infection stabilizes vimentin and reduces its phosphorylation Artesunate treatment resulted in reduced expression of HCMV Towne-encoded IE1/2 and pp65 as well as vimentin Ser83 phosphorylation and recovered vimentin level (Fig. 2B). The changes in vimentin level and phosphorylation were similar to those observed with HCMV-TB40 ( Fig. 2A), indicating the two viral strains did not differ in their effects on vimentin. Deoxy artemisinin did not change vimentin level or Ser83 phosphorylation and did not inhibit pp65 expression (Fig. 2C). The disappearance of the Ser83 band at 50 kDa and the Ser55 band at 57 kDa by phosphatase assay confirmed that these were phosphorylated proteins (Fig. 2D). In addition, mass spectrometry analysis of the 50 kDa band identified vimentin. Artesunate-mediated changes in vimentin were HCMV-specific since in HSV1-infected HFFs artesunate did not modify vimentin level and HSV1-encoded ICP8 was not reduced (Fig. 2E). We investigated the effects of artesunate on vimentin level at different times during infection. When added from the time of infection up until 72 hpi artesunate treatment resulted in remarkable vimentin stabilization (Fig. 2F). When added from 24-42 hpi or 42-72 hpi artesunate had a minor effect on vimentin level. Vimentin stabilization required artesunate to be present in the cell before the onset of viral DNA replication up until 72 hpi.

Vimentin disassembly during infection
Vimentin is a cytoskeleton protein and its filaments are seen in attached and flattened cells (21,22). We tested the localization of vimentin during HCMV infection and artesunate treatment. An indirect immunofluorescence (IFA) was performed at 24 and 72 hpi with HCMV TB40 (Fig. 3, Fig. S1), revealing an altered staining pattern and reduced vimentin signal during infection and recovery of vimentin signal with artesunate. The inactive artemisinin metabolite could not rescue vimentin signal at 24 or 72 hpi. Altogether, the data indicate that vimentin is disassembled during infection and artesunate maintains its assembly condition.

Artesunate stabilizes vimentin
To confirm the role of artesunate in vimentin stabilization, we performed a cellular thermal shift assay (23,24). HFFs were treated with artesunate (30 µM) or DMSO, washed with PBS, trypsinized and resuspended in PBS containing protease inhibitors, and incubated in a thermal cycler at the indicated temperatures. Vimentin level was measured by Western blot. In the DMSO-treated samples, vimentin level decreased with increasing temperatures, while artesunate stabilized vimentin at all tested temperatures (Fig. 4A). Comparison of the cellular thermal shift by artesunate and deoxy artemisinin (30 µM) at 56 ° and 60 °C showed that only artesunate stabilized vimentin (Fig.  4B). Neither artesunate nor deoxy artemisinin had an effect on p53 stabilization (Fig. 4B). Vimentin undergoes homodimerization, followed by antiparallel association between homodimers to form soluble tetramers. Further assembly of vimentin filaments involves lateral association between eight tetramers to form a unit length filament, which is followed by longitudinal annealing of these into short filaments and radial compaction of the annealed filaments to produce mature filaments. These structures are dynamic and reorganize during mitosis or cell stress (25). Using glutaraldehyde cross-linking, we found that high molecular forms of vimentin, representing soluble tetramers and possibly unit length filaments, were increased with artesunate treatment in noninfected and HCMV-infected HFFs (Fig. 4C). In sum, artesunate treatment results in vimentin stabilization.

Proteasomal-independent vimentin degradation during HCMV infection
We investigated whether vimentin was degraded during HCMV infection through the proteasome. Using the proteasome inhibitor MG132, vimentin level did not recover at 72 hpi (Fig.  4D, upper panel). Another experiment performed at 24 and 48 hpi revealed again that vimentin level did not recover with MG132 treatment, while p53 level was recovered as expected (Fig. 4D, lower panel), indicating MG132 activity was intact. The data suggest proteasome-independent degradation of vimentin in HCMV-infected cells.

Vimentin plays distinct roles depending on the stage of HCMV replication
We reported that artemisinins inhibit HCMV after virus entry, and their inhibitory effects persist during most stages of HCMV replication (6). We therefore investigated the role of vimentin during HCMV replication. The onset of infection and HCMV migration to the nucleus were reportedly delayed in the absence of an intact vimentin network (17), but later effects of vimentin on virus replication were not studied. Using lentivirus shRNAs we knocked-down vimentin in HFFs to a degree that would still enable virus entry and replication for inhibition studies (Fig. 5A). Virus entry into vimentindeficient cells was reduced compared to entry into control transduced cells, indicated by pp65 level (Fig. 5B). For in vivo relevance of these findings, we infected vimentin-knockout and control mice (129S) with mouse CMV (MCMV). All tested tissues from vimentin knockout mice showed significantly reduced virus titer, indicating the requirement of vimentin for initiation of MCMV replication (Fig. 5C). We next measured the effects of artesunate on HCMV replication in vimentin knockdown and control cells. Cells were infected with HCMV Towne (MOI=3) and a pp28-luciferase assay was performed at 96 hpi. Artesunate activity against HCMV was reduced in vimentin knockdown cells while GCV activity was comparable between the two cell lines, (Fig. 5D). A virus yield assay revealed reduced virus progeny from vimentin knockdown cells starting from 48 hpi (Fig. 5E). Compared to control cells, inhibition of HCMV with artesunate was less effective in vimentindeficient cells (5.2-vs 1.7-fold, respectively). Thus, vimentin level directly correlates with the anti-HCMV activity of artesunate.
Given the reported timing of HCMV inhibition by artemisinins (post-entry), reduced levels of vimentin during the progression of infection, and the correlation between artemisinins' activity and vimentin stabilization we attempted to dissociate the early role of vimentin from its later effects during HCMV replication. Vimentin was overexpressed in U373 glioma cells, followed by infection with the pp28-luciferase Towne. Luciferase activity measured at 96 hpi was significantly reduced in vimentin overexpressing cells compared to controls (  (17). Taken together, while the onset of HCMV replication requires an intact vimentin, its completion is hindered by it, indicating a dynamic role of vimentin depending on the stage of HCMV replication. These results further our understanding of the vimentin-artemisinin interaction during HCMV replication. An intact vimentin network facilitates artesunate binding and HCMV suppression following virus entry. Indeed, at 4 hpi, both HCMV and artesunate maintained the level of vimentin, but starting from 24 hpi when virus strategy was to degrade vimentin, artesunate-bound vimentin counteracted these activities and stabilized it, mimicking an overexpression status.

Withaferin A, a vimentin-binding agent, inhibits HCMV replication and antagonizes the anti-HCMV activities of artesunate
Additional support for vimentin being a target of artesunate was obtained using withaferin A (WFA), a vimentin-binding compound. The anti-HCMV activity of WFA was first tested. The effective concentration resulting in 50% inhibition of pp28-luciferase activity and plaque formation (EC50) was 0.1 ± 0.01 µM, and the 50% cytotoxicity in non-infected HFFs (CC50) was 2.0 ± 0.0 µM (Fig. S2). The combination of WFA and AS was tested against HCMV Towne using plaque reduction assay and antagonism was found (Fig. 6A). The combination of artesunate and GCV was synergistic, as reported (14) (Fig. 6B). Western blots revealed that the combination of artesunate and WFA was less effective in reducing the expression of viral IE1/2 compared to each drug alone (Fig. 6C). Finally, WFA at 10, 50, and 500 µM (but not GCV, 25 µM) displaced 552 (25 µM) bound to purified human vimentin protein in an in vitro binding assay (Fig.  6D), suggesting similar/overlapping binding sites of artesunate and WFA on vimentin.

The cell-cycle stage reflects vimentin dynamics during HCMV replication
Cell cycle regulation is critical for productive HCMV replication. Moving the cell cycle towards G1/S is a hallmark of HCMV infection (29,30). Artemisinins were reported to reverse HCMV-mediated cell cycle changes and reduce the levels of CDKs. In addition, in confluent HFFs (at the time of infection), artesunate inhibited HCMV, but in subconfluent cells, its anti-HCMV activity was lost (12). Cell synchronization at G0 with serum starvation, late G1 with mimosine, or early G1 with lovastatin resulted in effective HCMV inhibition with artesunate (12). Vimentin structure is highly dynamic and reorganizes in certain physiological situations, such as during cell cycle (25). We investigated the correlation between cell cycle and vimentin changes in infected HFFs. HCMV inhibition by artesunate was associated with reduced Ser83 phosphorylation and vimentin stabilization in contact inhibited cells. However, in subconfluent cells, artesunate did not reduce HCMV pp65 expression, and although infection reduced vimentin expression, artesunate could not rescue it or reduce Ser83 phosphorylation (Fig. 7A). A bivariate flow cytometry was performed to further support the association between vimentin and cell cycle stage during infection ( Table 2, Fig. 7B). The transition of infected cells into G1/S phase was accompanied by an overall reduction in vimentin signal (mean vimentin intensity 880 vs. 485). Artesunate reverted the cells back to G0/G1 along with an increased vimentin signal at all cell cycle stages. Next, the expression level of CDKs 1, 2, 4 and PLK1 was measured in HCMV-infected cells. Infection induced CDK 1, 2, 4 and this effect was reversed by artesunate at 24 and 72 hpi (Fig.  7C). PLK1, a kinase responsible for phosphorylating the Ser55 and Ser83 residues, was induced by HCMV and reduced by artesunate. GCV did not modify the level of CDKs or PLK1 during infection. Artesunatemediated changes in CDKs and PLK1 were observed only in infected cells, but not in noninfected cells (Fig. 7D), suggesting cell cycle in non-infected cells overcomes the effect of artesunate or that viral proteins participate in vimentin reorganization. Taken together, the anti-HCMV activity of artesunate is mediated through vimentin, and cell cycle drives changes in vimentin. HCMV harnesses cell cycle and vimentin; when vimentin is already degraded, artesunate loses its effectiveness. However, at the best cell cycle stage (early G1), artesunate will bind to vimentin, and compete with HCMV to prevent its degradation, winning the battle on HCMV replication.

Discussion
Drug repurposing may expand the antiviral armamentarium. Repurposed agents may use host-directed antiviral mechanisms (31). Therefore, the identification of molecular targets critical for HCMV replication may accelerate drug development strategies. Here we identified vimentin as a major cellular target of artemisinin monomers. We defined the dynamic changes of vimentin during HCMV replication and the role of artemisinins in HCMV inhibition through vimentin stabilization and modulation of its phosphorylation. A critical advantage of our assays is the availability of an artemisinin that lacks the endoperoxide bridge, providing a direct negative control for artemisinins' bioactivity. Vimentin was identified by MS using a biotinlabeled trioxane semi-synthetically derived from artemisinin, and target confirmation studies included competition assays using cell lysates and purified His-vimentin protein, SPR, and stability assays. WFA, an agent reported to bind to vimentin (32), inhibited HCMV replication, but when used in combination with artesunate, an antagonism was found.
Vimentin is one of the most widely expressed and highly conserved protein of the type III IF family. It is a marker of epithelial-mesenchymal transition, a process in which epithelial cells acquire a mesenchymal phenotype that causes them to alter their shape and exhibit increased motility (33,34). An intact vimentin network was reportedly required for the onset of HCMV replication, facilitating the trafficking of HCMV strains AD169 and TB40 into the nucleus (17). Disruption of vimentin with acrylamide, IF bundling in cells from a patient with giant axonal neuropathy, and absence of vimentin in fibroblasts from vimentin -/-mice reduced virus entry (17). During these initial steps of infection, vimentin was not disassembled, suggesting maintenance of its network is necessary for the establishment of infection. Our studies corroborate this report and further characterize changes in vimentin throughout the early and late stages of HCMV replication. At 4 hpi vimentin expression and phosphorylation were unmodified. However, as infection proceeded vimentin structure was altered, its phosphorylation (Ser55 and Ser83) steadily increased and its level was eventually reduced. Phosphorylation is a major post-translation modification that triggers vimentin disassembly (18). Artesunate binding to vimentin resulted in its stabilization, likely through interference with its phosphorylation. Multiple phosphorylation sites on vimentin have been identified (19,20,(35)(36)(37)(38)(39)(40)(41). Among the kinases that phosphorylate vimentin are p21-activated kinase, Aurora B kinase, CAMK2 (38). Vimentin interaction with p-ERK protects it from dephosphorylation (42). Phosphorylated vimentin interacts with 14-3-3 proteins, preventing the assembly of Raf-14-3-3 and similar complexes (43). The recognition of vimentin as a substrate for multiple kinases led to the hypothesis that it acts as a scaffolding protein involved in signal transduction (18 (49,50), and we show that vimentin expression is stable even at 24 hpi in HSV-1 infected fibroblasts (Fig. 2E), suggesting it may play a distinct role during HSV-1 infection, mirrored by differences in cell cycle modulation between HSV-1 and HCMV. While HCMV moves the cell cycle towards G1/S (29,30), HSV-1 blocks it in early-to mid-G1 (51), does not induce cyclins D or E, and reduces CDK2 activity (52). These differences in the modulation of CDKs and cyclins are reflected by changes in vimentin phosphorylation. Entry into the S phase would be accompanied by induction of calpain activity and reduced vimentin levels, while the exit from S phase is expected to restore vimentin levels (53-55). A correlation between vimentin levels and the cell cycle stage has been shown for mouse plasmacytoma cells (56). Our data reveal cooperation between vimentin and cell cycle stage (Fig 8, model). The transition of HCMV-infected cells into the G1/S phase was marked by reduction in vimentin signal ( Table 2). Artesunate binding to vimentin increased vimentin signal and reversed the cell cycle into G0/G1. Thus, vimentin levels are closely linked with cell cycle, likely in a bidirectional way, and by targeting vimentin, artesunate can not only stabilize it but also control cell cycle in a way that is deleterious for HCMV. While HCMV exploits vimentin strategically at the early and late stages of infection, the reversal of cell cycle cannot be tolerated by the virus. In fact, in subconfluent cells, artesunate was inefficient in inhibiting HCMV, and although vimentin was reduced by infection, artesunate could not restore its level or reduce its phosphorylation. In confluent cells, artesunate reduced the levels of CDKs, PLK1, vimentin phosphorylation, and HCMV was inhibited.
As the cellular target of artemisinins is now known, future studies should specify the binding sites of artemisinins, the required moieties for binding, and the role of vimentin as a cofactor for HCMV proteins. Our findings may also shed light on the anti-cancer activities of artemisinins as well as other pathogens. A more direct approach for drug development can now be accomplished. 0.052 mmol, 1.2 eq), 1hydroxybenzotriazole (HOBt, 7.1 mg, 0.052 mmol, 1.2 eq) and dimethylformamide (DMF, 2 ml) were added. Artemisinin-derived monomer alcohol 1 (21.3 mg, 0.065 mmol, 1.5 eq) was added, and the reaction mixture was stirred at RT for two days. The reaction mixture was quenched with water and diluted with EtOAc. The organic layer was washed progressively with water, 3 N LiCl, and brine, then dried over MgSO4 and filtered. The solvent was removed under reduced pressure, and the residue was purified directly on silica. Gradient elution (0-8% MeOH (containing 10% NH4OH) in CH2Cl2) afforded the desired compound as a colorless, amorphous solid: 10

Identification of 552 binding proteins by liquid chromatography tandem mass spectrometry (LCMS/MS)
Cell lysates from non-infected HFFs were incubated overnight at 4 °C with DMSO, 552 (20 µM) or 552 with DHA (200 µM) added 1 h before 552. Streptavidin agarose beads were used to capture 552-protein complexes and bound proteins were eluted using 2X laemmli sample buffer. Proteins were run on tris glycine gel and visualized using silver stain. Cut destained bands enriched in the 552 lane were submitted for mass spectrometry. Silver staining was performed using Silver Quest (TM) Staining Kit (Invitrogen, cat# LC6070), according to manufacturer's protocol. Proteins in gel bands were proteolyzed with trypsin to cleave at lysines and arginines (Promega), as previously described (58). Protein identification by LCMS/MS analysis of peptides was performed using a Velos Orbitrap MS (Thermo Fisher Scientific) interfaced with a 2D nanoLC system (Eksigent, www.eksigent.com). Peptides were fractionated by reverse-phase HPLC on a 75 um x 100 mm C18 column with a 10 um emitter using 0-60% acetonitrile/0.1% formic acid gradient over 60 min at 300 nl/min. Survey scans (full ms) were acquired from 350-1800 m/z with data-dependent monitoring of up to 8 peptide masses (precursor ions), each individually isolated in a 1.9 Da window and fragmented using HCD activation collision energy 35 and 25s dynamic exclusion. Precursor and the fragment ions were analyzed at resolutions 30,000 and 15,000, respectively, with automatic gain control (AGC) target values at 1e6 with 100 ms maximum injection time (IT) and 5e4 with 300 ms maximum IT, respectively. A lock mass of siloxane (371.1012 Da) was used for on the fly recalibration.  (59). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.

Immunoprecipitation (IP) of 552
Cell lysates from non-infected HFFs were treated with DMSO, 552, or 552 with DHA as a competitor. Targeted protein was immunoprecipitated with streptavidin agarose beads (Thermo Fisher) and samples were immunoblotted for vimentin.

Determination of the dissociation equilibrium constant (KD) by SPR
Human Vimentin was reconstituted to yield 0.5 mg/mL stock concentration. The CM5 sensor surface was activated by 1-ethyl-3-(3dimethylaminopropyl) carbodiimide hydrochloride (EDC)/N-hydroxy succinimide (NHS) mixture using a Biacore T200 instrument (GE Healthcare). Vimentin protein was then diluted with 10 mM sodium acetate (pH 4.0) to 30 µg/mL right before injection and immobilized to flow cells 2 and 4 followed by ethanolamine blocking on the unoccupied surface area. The unmodified surfaces on flow cells 1 and 3 were used as reference control cells. Test compound solutions with 2-fold increasing concentrations (0.93 -33.3 µM) were applied to all four flow cells at a 30 µL/min flow rate at 37 °C in an SPR binding buffer consisting of 20 mM HEPES, pH 7.5, 250 mM NaCl, 10 mM MgCl2, 0.05% Tween-20, and 0.1% DMSO. Sensorgrams were analyzed using Biacore T200 software V3.0, and response units (RU) at each concentration were measured during the equilibration phase for steady-state affinity fittings. All data were double referenced with reference cell and zero compound concentration RU values. Steady-state affinity fittings were calculated using both Biacore T200 evaluation software and Sigmaplot 12.0 with an embedded equation (1) as shown below, where y is the response, ymax is the maximum response and x is the compound concentration. (23,24) HFFs were treated with artesunate, deoxyartemisinin (30 µM) or DMSO for 24 h. Cells were washed with PBS, trypsinized and resuspended in PBS containing protease inhibitors, followed by incubation at temperatures ranging from 50-61 °C using a gradient heating protocol in a thermal cycler (Eppendorf Mastercycler). Soluble proteins were isolated by freeze thaw protocol, mixed with 2x SDS buffer, and resolved on a Tris glycine gel. Levels of vimentin, p53 and β-actin were determined by SDS-PAGE and Western blot using specific antibodies.

Calpain activity assay
In-vitro calpain-mediated vimentin degradation assay was modified from (60 (28) HFFs were infected with HCMV and loaded with the fluorogenic calpain substrate Boc-Leu-Met-CMAC (50 µM) for 30 min. Cells were harvested and the relative fluorescence from the protease-cleaved product (AMC) of the fluorogenic substrate was measured.

Collection of vimentin subunits
Cell extracts were centrifuged at 200,000 g and supernatants were collected. Released subunits were cross-linked with glutaraldehyde (GA, 0.005%).

Transfection and generation of vimentin knockdown cells
The following vimentin plasmids were used for transient transfection in U373 (cells that achieve both good transfection and HCMV replication): pCDNA4-vimentin and pCDNA4 control. Knockdown of vimentin in HFFs was performed using lentivirus transduction with VIM shRNA: RHS3979-201759426 -TRCN0000029119 and pLKO.1 control designed by The RNAi Consortium, or TRC (GE Healthcare). Individual shRNA constructs were packaged using lentivirus as previously reported (61).

SDS-polyacrylamide gel electrophoresis and immunoblot analysis
Cell lysates containing an equivalent amount of proteins were mixed with an equal volume of sample buffer (125 mM Tris-HCL, pH 6.8, 4% SDS, 20% glycerol and 5% β-mercaptoethanol) and boiled at 100ºC for 10 min. Denatured proteins were resolved in Tris-glycine polyacrylamide gels (10-12%) and transferred to polyvinylidine difluoride membranes (Bio-Rad Laboratories, Hercules, CA) by electroblotting. Membranes were incubated in blocking solution [5% w/v nonfat dry milk and 0.1% Tween-20 in PBS (PBST)] for 1 h, washed with PBST, and incubated with antibody at 4 ºC overnight. Membranes were washed with PBST and incubated with horseradish peroxidaseconjugated secondary antibodies in PBST for 1 h at RT. Protein bands were visualized by chemiluminescence using SuperSignal West Dura and Pico reagents (Pierce Chemical, Rockford, IL). Antibodies for HCMV proteins were used at 1:2000 and included: mouse monoclonal anti-HCMV IE1 & IE2 (MAB810, Millipore, Billerica, MA), mouse monoclonal anti-HCMV UL83 (pp65, VP-C422 Vector Laboratories Inc., Burlingame, CA), and mouse anti-HCMV UL44, 10E8 (Santa Cruz Biotechnology, Santa Cruz, CA). Mouse anti-Vimentin (V9) and rabbit anti phospho-vimentin (Ser83) were from Santa Cruz. Anti-HSV ICP8 NB100-2770 was from Novus (Novus, Centennial, CO) Mouse anti phospho-vimentin (Ser55) was from Enzo Life Sciences (Farmingdale, NY). Rabbit polyclonal antibody HSP90β was from Cell Signaling Technology (Danvers, MA). Mouse monoclonal anti-p53, anti β-Actin and β-tubulin were obtained from Santa Cruz. Quantification of Western blots was performed with NIH ImageJ software. A phosphatase assay was performed to confirm the phospho-vimentin bands. After transfer, the PVDF membrane was washed twice with deionized H2O, followed by washing with buffer (5 min). The membrane was placed in FAST AP Thermosensitive Alkaline Phosphatase buffer and 100 units of FAST AP (Thermo) were added for 30 min at 37 °C shaker, followed by two washes with washing buffer. The membrane was blocked with blocking buffer for 1 h, and Ser55 antibody (1:2000) was added overnight.

IFA staining and confocal microscopy
HFFs were plated on chamber slides followed by infection with HCMV TB40 and treatment with artesunate (30 µM), deoxy-artemisinin (30 µM), and GCV (5 µM) for 24 or 72 h. Cells were fixed with 3.7% paraformaldehyde for 20 min at RT, permeabilized with ice-cold methanol for 10 min at -20 0 C and blocked with 5% bovine serum albumin in 0.5% Tween-20 for 20 min at RT. Cells were incubated with primary antibodies at 4 °C overnight, washed and incubated with fluorescently-labeled secondary antibodies (2 h, 37 °C). Fluorescence microscopy was performed using a confocal laser scanning microscope (Nikon EZ C1). Images were captured at 60X magnification and processed under identical conditions with constant parameters (including scan speed and excitation and emission wavelengths) using Nikon EZ C1 software. Quantification of cell fluorescence was performed using NIH ImageJ software; fluorescence intensity represents corrected total cell fluorescence (CTCF)/cell. Data shown are mean ± SD (n = 15).

Drug combination studies
HFFs were seeded in 24-well plates and infected with HCMV Towne at 75 plaques/well. First, a dose-response curve was generated for each drug individually to determine its EC50. Then, drugs were combined at twice their EC50, diluted in DMEM with 4% FBS, followed by serial dilution and added together after infection. Plaques were counted 10 days after infection. The Bliss model was used to calculate the effect of each drug combination on virus replication. In this model, drug combination represents the product of two probabilistically independent events as described in the following equation: Where D is the drug concentration, m is the slope, and EC50 is the effective concentration resulting in 50% virus inhibition. The combined effect of two inhibitors (FU, fractional unaffected) is computed as the product of individual effects of the two inhibitors, FU1 and FU2. If the ratio of observed fold inhibition divided by the expected fold inhibition is greater than 1, the compounds are synergistic. If the ratio is less than 1, the combination is considered antagonistic, and if it equals to 1 the combination is additive.

MCMV infection of vimentin knockout mice
Infection with MCMV Smith strain (ATCC VR-1399) was carried out in 3-4-week-old vimentin knockout (129S-Vimtm1Cba/MesDmarkJ), and wild-type 129S1/SvImJ mice (Jackson Laboratories, Bar Harbor, ME). The Animal Care and Use Committee of Johns Hopkins University approved the experimental procedures. After 2-3 days of adaptation to the housing environment, mice were infected intraperitoneally with 10 4 PFU/mice (0.1 mL in 0.8% saline). Mice were sacrificed at day 13 after infection. Salivary glands, spleen and liver were harvested and stored at -80 °C until further use. Organs were homogenized in DMEM with 4% FBS at a final concentration of 100 mg/mL. Two million mouse embryonic fibroblasts (MEFs) were seeded into 24 well plates. From each sample, 5% of the salivary gland homogenate or 10% of the spleen or liver homogenate was used for infection of MEFs in triplicates. Plaques were counted after three days.

Flow cytometry
HFFs were infected with HCMV Towne and treated with AS (30 µM). At 72 hpi, cells were harvested for bivariate cell cycle analysis on a FACSCalibur Flow Cytometer (BD Biosciences, San Jose, CA) using PI and Alexa-Fluor 488 conjugated mouse anti-human vimentin antibody. Data analysis was performed using FlowJo software (Tree Star, USA).

Statistical analysis
Statistical significance was assessed with GraphPad Prism 5.0 software. Data are presented as mean ± standard deviation, SD (n ≥ 3). Statistical significance between two groups were analyzed by two-tailed Student's t-test, and asterisks indicate the statistical significance: *, p < 0.05; **, P <0.01; *** p < 0.001. All experiments were performed at least twice.      Cell suspension was distributed into PCR tubes and incubated at the indicated temperatures for 3 minutes in a gradient thermal cycler. Heattreated cells were lysed using freeze thaw cycle twice, followed by brief vortex and centrifuged at 13,000 rpm for 10 minutes. Supernatants were transferred into fresh tubes, mixed with 2x SDS sample buffer, and a Western blot was performed using anti-vimentin and anti-p53 antibody. β-actin was used as loading control. The experiments were performed three times and data are from a representative experiment. Quantification of protein bands was performed by ImageJ and is shown below the blots. C. Cell lysates from non-infected, infected and infected AS-treated samples were cross-linked with glutaraldehyde 0.005% and vimentin forms were detected by Western blot. D. Upper: HCMV-infected or non-infected HFFs were treated with AS in the presence or absence of proteasomal inhibitor, MG132 (10 µM), added 8 h before harvesting. Expression of vimentin was measured by Western blot at 72 hpi. Lower: HFFs were infected with HCMV (MOI=1), and MG132 was added 8 h before harvesting the cell lysates. Expression level of vimentin and p53 was measured at 24 and 48 hpi. E. HCMV-infected HFFs nontreated or AStreated were used for a calpain activity assay using the fluorogenic calpain substrate Boc-Leu-Met-CMAC. At 72 hpi cells were loaded with Boc-Leu-Met-CMAC (50 µM) for 30 min before the assay. E64D (labeled "E") was used at 100 µM for 6 h prior to harvest in some of the conditions. Cells were harvested and the relative fluorescence from the protease-cleaved product (AMC) of the fluorogenic substrate was measured and summarized. Data show mean and standard deviation of triplicate values from one representative experiment, p < 0.001. NI-non infected HFFs, I -HCMV-infected HFFs, nsnot significant. F. Purified recombinant vimentin (1 µg) and calpain were used in an in vitro vimentin cleavage assay and shown is a Coomassie stain of vimentin cleavage. Lane 1-vimentin migration with activating reaction buffer (without calpain) in a 12% gel. Lane 2-in the absence of buffer, the enzyme does not digest vimentin; the glycerol in the calpain enzyme changes the migration pattern of vimentin.   1). B. HCMV entry was assayed by infecting control transduced and shVim HFFs with HCMV Towne for 2.5 h, washing with PBS and then treating with 1 µg/mL proteinase K for 30 min at 4 °C to remove viral particles adhered to the cell surface. The level of viral pp65 was measured by Western blot. C. 3-4-week-old vimentin knockout mice (Vim -/-) and wild type (WT) controls (129S) were infected intraperitoneally with 10 4 PFU/mice (0.1 mL in 0.8% saline) MCMV Smith strain and sacrificed at day 13 post infection. Salivary glands, spleen and liver were homogenized, and virus yield from organs was measured using a plaque assay in MEFs. D.
Vimentin knock-down (shVim) and shRNA control HFFs, were seeded at 2x10 6 cells/plate in a 96-well plate, infected with pp28-luciferase Towne at MOI of 5 PFU/cell and treated with AS (left) or GCV (right) at the indicated concentrations. Luciferase activity was measured after 4 days. Asterisks indicate the statistical significance: *, p < 0.05; **, P <0.01; *** p < 0.001. E. Vimentin knock-down (shVim) and shRNA control HFFs (pLKO.1) were seeded at 1x10 6 cells/plate in a 96-well plate, infected with HCMV Towne at MOI of 1 PFU/cell and treated with AS (30 µM). Supernatants were collected each day after infection up to five days. Virus yield was assayed by plaque assay in HFFs. F. U373 cells were transfected with pcDNA4A empty vector (EV) or pcDNA4A expressing vimentin (Vim). After 24 h, cells were infected with pp28-luciferase Towne. Cell lysates were collected at 96 hpi for luciferase activity. G.
Vimentin level was measured in non-infected cells by Western blot (upper). Levels of vimentin and pp65 in HCMV-infected U373 transfected with EV or pcDNA4A-vimentin plasmid were measured by Western blot (lower panel). H. HCMV entry into transfected U373 was assayed by infecting 1x10 6 cells in a 24well plate at MOI of 5 PFU/cell for 2 hours at 37 °C, washing with citric acid buffer (pH = 3) to strip off adherent virus particles and detection of pp65 by Western blot. Quantification of protein bands is shown below the blots.   . Following HCMV entry, vimentin maintains its stability to allow efficient virus trafficking to the nucleus. As infection proceeds, HCMV induces calpain as well as vimentin phosphorylation leading to vimentin cleavage. Concurrently, HCMV induces CDKs and cell cycle progression to G1/S. Overall, these linked events provide a conducive environment for lytic HCMV replication. Artesunate binding to vimentin early after virus entry prevents calpain access to vimentin as well as its phosphorylation, resulting in vimentin stability and maintenance of cell cycle at G1.