We combined novel proteomic technologies with molecular virology, genetic, RNAi, dominant interfering, and pharmacological approaches, to validate an interaction between a heretofore unrecognized tyrosine motif within HCV core and AP2M1 that is essential for viral assembly, and to identify inhibitors that disrupt this interaction and inhibit HCV assembly.
Our results suggest that core's AP2M1 binding motif is required for HCV assembly in vitro
. Y136A and V(Φ)139A core YXXΦ mutations significantly impaired AP2M1 binding, in line with the key role played by these residues in facilitating recognition of cargo proteins by AP2M1 
. Furthermore, these core mutations dramatically impaired HCV assembly, in correlation with their effect on AP2M1 binding. These results are consistent with prior alanine scanning and deletion analysis demonstrating a significant effect of either quadruple alanine substitutions or a larger deletion harboring the YXXΦ core residues on infectious virus production 
. Emergence of primary-site revertants that coincided with phenotypic reversion of the infectivity phenotype, as we detected, further indicates the requirement of maintaining a functional YXXΦ motif for supporting HCV replication in vitro
. This motif is highly conserved across all natural HCV isolates available in databases to date, suggesting that there is a requirement for the AP2M1 binding residues for productive viral infection in vivo
. The functional interchangeability of core's YXXΦ motif with homologous motifs suggests that it exerts its function via interactions with host cell proteins. Indeed, stable and transient functional RNAi approaches as well as dominant interfering and pharmacological approaches consistently demonstrated that AP2M1 is essential for efficient HCV assembly. Core's YXXΦ motif thus represents the first signal mediating binding of clathrin APs discovered and validated in the Flaviviridae family.
Binding of core to AP2M1 provides a mechanism by which HCV hijacks endocytic functions for its assembly. Few other viruses, such as HIV 
, EIAV 
, and HBV 
have been shown to rely on early endocytic functions for their assembly and budding. Unlike the established role of late endocytic functions in facilitating membrane fission events at the site of envelopment (reviewed in 
), the role mediated by these early endocytic proteins in the late stages of viral life cycles is, however, less studied.
While most studies thus far have focused on AP2M1's role in mediating endocytosis of CCV from the plasma membrane to endosomes, there is evidence that it may be involved in mediating additional intracellular membrane traffic processes. AP-2 was shown to initiate assembly of clathrin coats on lysosomes 
. Moreover, recent proteomic studies have revealed its presence on ER and Golgi membranes 
. Contacts between endosomes and ER membranes 
may help facilitate such recruitment of AP2M1 to the ER membrane. Interestingly, several endosomal membrane trafficking proteins have been previously shown to be recruited to LD in naive cells 
, illustrating the dynamic nature of these intracellular compartments. Our study did not demonstrate significant localization of endogenous or exogenous AP2M1 to LD in either naive cells or cells overexpressing AP2M1 alone, respectively. Nevertheless, in the setting of overexpression of core or HCV infection, AP2M1 appeared to significantly localize to the surface of LD. Importantly, no such colocalization was demonstrated when AP2M1 was co-expressed with core harboring the Y136A mutation, suggesting that core's YXXΦ motif hijacks AP2M1 and facilitates its recruitment to LD for mediating HCV assembly.
Core has been previously shown to localize to various compartments including LD, ER, and TGN 
. Our quantitative confocal microscopy experiments indeed demonstrated core localization to all these compartments. Nevertheless, increased localization of core to LD and decreased localization to TGN were demonstrated in cells electroporated with HCV RNA harboring a Y136A core mutation or depleted for AP2M1. These results indicate that the observed defects in infectious HCV production following disruption of the core-AP2M1 interaction are associated with altered sub-cellular localization of core. AP2M1, like other clathrin adaptors, is known to sort cargo proteins and mediate intracellular traffic (reviewed in 
). It was previously shown that AP subunits are also involved in mediating intracellular trafficking of structural retroviral proteins. An interaction between Gag's YXXΦ motif with AP2M1 was shown to confine HIV-1 exit to distinct microdomains 
. Similarly, an interaction between Gag and the δ subunit of the AP-3 complex has been shown to mediate trafficking of Gag to CD63-positive intracellular compartments and to be essential for HIV assembly 
. Binding of AP2M1 to core's YXXΦ motif may thus offer a mechanism to facilitate intracellular trafficking of core. Disruption of core's interaction with DGAT1 is associated with accumulation of core on ER membranes and failure of core to traffic to LD, consistent with a defect in an early assembly stage 
. Unlike the core-DGAT1 interaction, our data suggest that the rate limiting stage affected by disruption of the core-AP2M1 binding is transport of core away from LD. These data, combined with the reduction of core colocalization with E2 demonstrated in the context of the Y136A core mutation or AP2M1 depletion, suggest that core-AP2M1 binding may mediate a later stage of HCV assembly following accumulation of core on LD but prior to envelopment at the ER. Interestingly, we have recently identified conserved YXXΦ and dileucine motifs within the cytoplasmic tail of the E1 protein (data not shown). It is therefore tempting to speculate that core and E1 may interact on the same adaptor protein complex and that by binding both core and E1, AP2M1 may help facilitate trafficking of viral particles from their assembly sites on LD (where core is accumulated) to the envelopment sites at the ER (where the E1 protein is retained). Interestingly, such a mechanism has previously been described in HBV, where the envelope and capsid proteins both interact with γ2 adaptin as a means to traffic viral particles from assembly to budding sites 
. Consistent with its role in mediating HCV assembly, disruption of core-AP2M1 binding reduced core localization to TGN, the compartment in which viral particles are thought to mature during egress. Since ER is not only the site of presumed viral particles envelopment but also the site of core synthesis and processing, it is not surprising that no apparent decrease in localization of core to ER was demonstrated as a result of either core mutation or AP2M1 depletion. We plan to determine whether E1 binds AP2M1 and whether this interaction is cooperatively involved in mediating assembly. Future studies using live cell imaging of core trafficking will help determine whether the core-AP2M1 interaction directly mediates core transport out of LD to the envelopment sites.
Late endocytic functions have been shown to be involved in mediating production of infectious HCV virus. Either expression of dominant negative forms of Endosomal Sorting Complexes Required for Transport (ESCRT) components 
or their depletion reduced production of infectious virus 
. Moreover, inhibitors of endosomal motility disrupt HCV assembly and/or budding 
. Nevertheless, AP-made vesicles bud into the cytoplasm, which is the topological reverse of HCV budding into the lumen of intracellular compartments. Based on this topology and our IF confocal microscopy analysis, it seems lower likely that AP2M1 itself promotes vesicle formation into the ER and the pinching-off reaction required for viral envelopment.
Functionally relevant interactions between YXXΦ motifs within structural viral proteins and AP2M1 have been previously described in retroviruses 
. To the best of our knowledge, such interactions have not been previously reported in HCV or any other viral family except Retroviridae. Furthermore, AAK1 and GAK have not been reported to mediate a viral infection to date. Importantly, AP2M1's interactions with retroviral or host proteins have not been targeted pharmacologically. Using RNAi and pharmacological approaches, we show that AAK1 and GAK are involved in regulating core-AP2M1 binding. The role of AAK1 and GAK in stimulating binding of host cargo proteins to AP2M1, suggests that the HCV core protein may function as a cargo protein. Our data also suggest that AAK1 and GAK are host factors essential in mediating HCV assembly. Moreover, we discovered kinase inhibitors known to target AAK1 or GAK, which inhibit AP2M1 phosphorylation, disrupt binding of core to AP2M1, and inhibit HCV assembly as well as infectious virus production. While the discovered compounds are not entirely selective (like the majority of compounds 
), there is evidence to support their relatively good selectivity. Eventhough AAK1 and GAK are not among the cancer targets of these compounds, they are very potently bound and inhibited by sunitinib or PKC-412 and erlotinib, respectively (with Kds of binding that are lower or comparable to the major cancer targets). For example, the Kd of binding of erlotinib to its primary cancer target, EGFR, is ~1 nM compared with a Kd of binding of 3.1 nM to GAK (SuperNova Life Science 2008, Human Kinome Heat Map, http://www.supernovalifescience.com/HM/HM%2041.pdf
). Importantly, the Kds of binding of erlotinib to all other studied host protein kinases are significantly higher (with most being >1,000–10,000 nM). Moreover, these compounds had no effect on HCV RNA replication, yet significantly inhibited assembly. Last, their effect on core-AP2M1 binding and their overall effect on HCV replication () correlated with their binding affinity to AAK1 or GAK. The IC50 values of the compounds for core-AP2M1 binding were 0.04–0.2 µM, whereas their EC50 values for production of infectious virus in cells were 0.155–1.8 µM. Such ~10–25-fold differences between IC50s and EC50s values are typical for kinase inhibitors 
. It is possible that a limited bioavailability (determined by the rates of entry to the cells and/or active pumping out of the cells) of the compounds accounted for these differences 
. We hypothesize that improved drug delivery and optimization of the compounds after structure-activity relationship analysis will improve their potency as antiviral agents.
Interestingly, erlotinib has been recently reported to inhibit HCV endocytosis via its effect on EGFR 
. While sunitinib and PKC-412 do not target EGFR, as described above the latter represents the only other known major target of erlotinib besides GAK. Inhibition of GAK-mediated assembly thus represents another mechanism of action of erlotinib in the HCV life cycle, besides its effect on EGFR-mediated entry. Our data suggest that both mechanisms accounted for the effect of erlotinib on HCV infection in cells infected with culture grown HCV (). Nevertheless, its dramatic effect on infectious virus production, measured by the infectivity assays in cells electroporated with HCV RNA (), confirms its independent effect on viral assembly. Importantly, erlotinib has already demonstrated an antiviral activity in vivo
, by effectively eradicating HCV infection in humanized mice infected with HCV, without inducing any apparent toxicity 
. Moreover, a recent case report describes rapid clearance of HCV RNA in serum of a patient undergoing treatment with erlotinib for recurrent hepatocellular carcinoma 
. While attributed to its effect on EGFR, our data suggest that the observed antiviral effect likely resulted from its dual effects on EGFR and GAK. Erlotinib, sunitinib, and PKC-412 may represent a pharmacological research tool not only for studying host cell mechanisms involving AAK1 or GAK but also for studying interactions of AP2M1 with cargo proteins. Furthermore, these compounds may serve as an attractive strategy for treatment of HCV. Such a host-centered approach may potentially be associated with a higher genetic barrier for resistance and may provide an efficient means to combat all HCV genotypes, unlike genotypic specificity associated with some drugs that target viral enzymatic functions 
. Erlotinib and sunitinib are already approved anti-cancer drugs and PKC-412 is at an advanced clinical development stage. Our studies revealed no effect of these compounds on cellular viability at the concentrations used. Similarly to drugs used in current standard of care anti-HCV regimens, side effects, including some that are significant, have been reported in patients receiving erlotinib, sunitinib or PKC-412 for the treatment of various malignancies. Nevertheless, overall they seem to be well tolerated over prolonged treatment courses 
. Importantly, these drugs are delivered orally, thus avoiding the inconvenience and complications related to parenteral administration. We hypothesize that more selective and potent inhibitors of AAK1, GAK and/or core-AP2M1 binding can be obtained. Nevertheless, because erlotinib, sunitinib, and PKC-412 have already been extensively used in humans (albeit for a different indication) there may be an opportunity for repurposing them as antivirals. As recently discussed by Jilg N and Chung RT in respect to erlotinib 
, these compounds may find immediate use as components of next generation anti-HCV strategies with particular utility among patients failing to respond to standard of care regimens.
AP2M1 represents an example for an overlapping host mechanism required by HCV and lentiviruses. Indeed, a common evolutionary origin of these seemingly unrelated viruses has been previously proposed 
. We now plan to study the role of AAK1 and GAK in mediating infectious virus production of other viruses that hijack AP2M1, such as HIV. Conservative requirement for AAK1 and GAK across viral families may suggest that the identified inhibitors could potentially be used for treating HCV-HIV co-infection and possibly a broader spectrum of viral infections.
Finally, our results demonstrate the utility and advantages of the microfluidics and PCAs approaches for studying viral-host P-PIs. These sensitive formats allow studying weak and transient interactions. Protein synthesis by mammalian lysates in the presence of microsomal membranes provides natural conditions required for protein folding, thus facilitating studies of membranous proteins by the microfluidics approach. The PCAs format allows detection of P-PIs involving membrane proteins in the relevant cell model and appropriate sub-cellular compartments 
(unlike most yeast two-hybrid systems where the interaction occurs in the nucleus). By allowing nanoliter protein consumption, the microfluidics technology eliminates the need for high level protein expression and purification, which hinders the design of other protein arrays. Their scalability and ability to study pharmacological inhibition of interactions, makes these platforms ideal for high-throughput screening not only for P-PIs but also for inhibitors of identified P-PIs (as we have previously done for protein-RNA interactions 
Taken together, these results have exciting implications with respect to mechanisms of HCV assembly and design of novel antiviral strategies.