ACBD3 Is an Essential Pan-enterovirus Host Factor That Mediates the Interaction between Viral 3A Protein and Cellular Protein PI4KB

Similar to all other positive-strand RNA viruses, enteroviruses reorganize host cellular membranes for efficient genome replication. A host lipid kinase, PI4KB, plays an important role in this membrane rearrangement. The exact mechanism of how enteroviruses recruit PI4KB was unclear. Here, we revealed a role of a Golgi-residing protein, ACBD3, as a mediator of PI4KB recruitment upon enterovirus replication. ACBD3 is responsible for proper localization of enteroviral 3A proteins in host cells, which is important for 3A to recruit PI4KB. By testing ACBD3 and PI4KB mutants that abrogate the ACBD3-PI4KB interaction, we showed that this interaction is crucial for enterovirus replication. The importance of specific domains of ACBD3 was evaluated for the first time, and the domains that are essential for enterovirus replication were identified. Our findings open up a possibility for targeting ACBD3 or its interaction with enteroviruses as a novel strategy for the development of broad-spectrum antienteroviral drugs.


ACBD3 knockout inhibits replication of enterovirus A to D and rhinovirus A and B species.
Previously, we observed no effects on CVB3 and RV replication in HeLa cells in which ACBD3 was knocked down more than 90% (18,19). Here, we set out to study enterovirus replication in ACBD3 KO HeLa cells. HeLa cells lacking ACBD3 were generated with CRISPR-Cas9 technology, and the knockout was confirmed by Western blot analysis (see Fig. S1A in the supplemental material). Next, we evaluated enterovirus replication kinetics in ACBD3 KO  . RV-C was not tested, as HeLa R19 cells are not susceptible to RV-C because they lack the receptor, cadherinrelated family member 3 (CDHR-3) (21). All of the viruses clearly showed deficient replication in ACBD3 KO HeLa cells (Fig. 1A). Replication of enteroviruses was also impaired in another human cell line, haploid human cell line HAP1, in which ACBD3 was knocked out (Fig. S2). To exclude the role of ACBD3 in virus entry, we assessed viral RNA replication of subgenomic replicons, which are widely used tools to study genome replication specifically and independently of virus entry. Replication of CVB3 and EV-A71 replicons transfected in HeLa ACBD3 KO cells was reduced, which indicates a role for ACBD3 in the genome replication step (Fig. 1B). Next, to test whether ACBD3 functions in the same pathway as PI4KB in enterovirus replication, we employed a mutant virus that is less sensitive to PI4KB inhibition, CVB3 3A-H57Y (22). While the replication of wt CVB3 (RlucCVB3) was impaired in ACBD3 KO cells, the replication of RlucCVB3 3A-H57Y was significantly increased (Fig. 1C). Encephalomyelitis virus (EMCV), which belongs to the genus Cardiovirus in the Picornaviridae family, depends on PI4KA but not on PI4KB for generating ROs (23). The replication of EMCV was not affected in ACBD3 KO cells (Fig. S3), suggesting that the inhibition of CVB3 and EV-A71 replication in ACBD3 KO cells is connected to the PI4KB pathway. Overall, these results indicate that ACBD3 is an important host factor for enterovirus replication.
ACBD3 is indispensable for PI4KB recruitment. To determine the importance of ACBD3 for PI4KB recruitment during enterovirus replication, we investigated PI4KB localization in CVB3-infected ACBD3 KO cells. Since we observed delayed virus replication in ACBD3 KO cells (Fig. 1A), different time points were chosen for HeLa wt cells and ACBD3 KO cells to mitigate possible effects of different replication levels on PI4KB recruitment. As previously shown, the CVB3 3A protein colocalized with ACBD3 ( Fig. 2A) and PI4KB (Fig. 2B) throughout infection in infected HeLa wt cells, which implies that both ACBD3 and PI4KB localize to CVB3 ROs. PI4KB was more concentrated in 3Apositive cells than in 3A-negative cells, which suggests that it is actively recruited to virus replication sites. Notwithstanding the similar level of 3A expression compared to HeLa wt cells ( Fig. 2A and B), no recruitment of PI4KB was observed in infected ACBD3 KO cells at any time point (Fig. 2D). These results indicate that ACBD3 mediates recruitment of PI4KB during enterovirus replication.
Enterovirus 3A expression alone is sufficient to recruit PI4KB to membranes (9,18,19). To further investigate whether PI4KB recruitment by 3A is mediated by ACBD3, we transiently expressed the 3A proteins from representative human enteroviruses from seven different species (EV-A/B/C/D and RV-A/B/C) with either a C-terminal myc tag or an N-terminal GFP tag and examined the localization of ACBD3 and PI4KB ( Fig. 3 and Fig. S4). In HeLa wt cells, all 3A proteins colocalized with ACBD3 (Fig. 3A). PI4KB was more concentrated in cells expressing 3A than in cells that did not express 3A, and it colocalized with 3A ( Fig. 3A), which indicates that PI4KB is actively recruited by enterovirus 3A proteins. In contrast, no PI4KB recruitment was observed in ACBD3 KO cells expressing any of the enterovirus 3A proteins (Fig. 3B). These results imply that all enteroviruses utilize a shared mechanism to recruit PI4KB to replication sites, which is via a 3A-ACBD3-PI4KB interaction. Interestingly, we noticed that the localization of 3A differs from HeLa wt cells to ACBD3 KO cells ( Fig. 3 and Fig. S5). Unlike the punctate localization in HeLa wt cells (Fig. 3A), 3A proteins were dispersed throughout the cytoplasm in ACBD3 KO cells into a more reticular pattern ( Fig. 3B and Fig. S5), suggesting that ACBD3 is important for proper localization of 3A.
ACBD3 is crucial for 3A localization to the Golgi. Because the typical punctate localization of 3A on Golgi-derived membranes is lost in ACBD3 KO cells, we assessed the overall structure of the Golgi and ER as well as the colocalization between 3A and markers for the Golgi (GM130 and TGN46) and ER (calreticulin). In mock-transfected cells, no gross differences in localization of any of the above markers were observed between HeLa wt cells and ACBD3 KO cells, although in some ACBD3 KO cells, the Golgi appeared to be slightly more scattered than in HeLa wt cells (Fig. 4). Importantly, this slight Golgi scattering clearly differs from the massive Golgi scattering that can be observed upon knockout of structural Golgi proteins GRASP55 and GRASP65 (24). Nevertheless, smaller disruptions of Golgi structure that cannot be readily visualized at the light microscopy level cannot be excluded. Indeed, others have reported fragmentation of Golgi cisternae when they studied ACBD3 knockdown cells by electron microscopy (25). As previously described, the disintegration of the Golgi in enterovirusinfected cells and in 3A-expressing cells is likely a consequence of the blockage of ER-to-Golgi transport that depends on the interaction between 3A and GBF1/Arf1 (26,27). In agreement with this, overexpression of 3A caused disassembly of the Golgi apparatus in both wt and ACBD3 KO HeLa cells ( Fig. 4B and C), pointing out that the disruption of the Golgi by 3A occurs independently of ACBD3. 3A partially colocalized with the Golgi markers but not the ER marker in HeLa wt cells, whereas 3A was localized to the ER, as labeled by calreticulin, in ACBD3 KO cells ( Fig. 4B to D). This suggests that 3A cannot localize to the Golgi without ACBD3, which may contribute to the lack of PI4KB recruitment. Collectively, our results suggest that ACBD3 is not merely a mediator between 3A and PI4KB but that it plays a central role in recruiting 3A and PI4KB to facilitate virus replication.
Exogenous expression of wt ACBD3 in ACBD3 KO cells restores 3A localization, PI4KB recruitment, and enterovirus replication. To confirm that ACBD3 recruits 3A to the Golgi and mediates the interaction between 3A and PI4KB, we tested whether reconstitution of GFP-tagged ACBD3 in ACBD3 KO cells can restore 3A localization and PI4KB recruitment. For a negative control, we used a Golgi-localized GFP (i.e., GFP coupled to amino acids 1 to 60 of galactosyltransferase [GalT]), which failed to restore 3A localization (Fig. 5A, top panel) and PI4KB recruitment (Fig. 5B, top panel). When wt ACBD3 Mediates the 3A-PI4KB Interaction ® ACBD3 was reconstituted, 3A regained its punctate localization (Fig. 5A, middle panel), and PI4KB was recruited to the same sites ( Fig. 5B, middle panel). ACBD3 expressed without 3A was found in the Golgi, where it colocalized with giantin, but no concentrated PI4KB was observed (Fig. S6). These results indicate that proper 3A localization and PI4KB recruitment by 3A depend on ACBD3.
ACBD3 forms a tight complex with PI4KB (16). Recently, it was reported that one or two amino substitution(s) in the Q domain of ACBD3 (F 258 A or F 258 A/Q 259 A) can abrogate binding between purified recombinant ACBD3 and PI4KB in pulldown experiments (16,17). We confirmed that the F 258 A/Q 259 A mutant (hereafter called the FQ mutant) lost its interaction with PI4KB by coimmunoprecipitation (Fig. S7) and employed this mutant to test whether the ACBD3-PI4KB interaction is required for PI4KB recruitment and efficient enterovirus replication. Expression of the ACBD3-FQ mutant restored the punctate localization of 3A (Fig. 5A, bottom panel) but did not support PI4KB recruitment (Fig. 5B, bottom panel). Furthermore, exogenous expression of wt ACBD3 in ACBD3 KO cells restored replication of CVB3 to a level comparable to that in HeLa wt cells, while the negative control (GalT) and ACBD3-FQ mutant could not restore virus replication in ACBD3 KO cells (Fig. 5C). Taken together, we showed that 3A localization to the Golgi-derived membranes occurs in an ACBD3-dependent manner and that the interaction between ACBD3-PI4KB is crucial for PI4KB recruitment and efficient virus replication.
Reconstituted wt PI4KB in PI4KB KO cells can be recruited to membranes through the 3A-ACBD3-PI4KB interaction, thereby restoring enterovirus replication. PI4KB is recruited by 3A to ROs during enterovirus replication (8,9), and depletion of PI4KB by RNAi (9) or pharmacologic inhibition (reviewed in reference 28) have been shown to suppress virus replication. Enterovirus mutants resistant to inhibitors of PI4KB contain single amino acid substitutions in the 3A protein (e.g., H57Y for CVB3). CVB3 replication is severely impaired in PI4KB KO cells that we generated by CRISPR/Cas9 technology (Fig. S1B), while the resistant mutant virus (3A-H57Y) replicated well in PI4KB KO cells (Fig. 6A).
Two PI4KB mutants (I 43 A and D 44 A) were previously shown to reduce binding between recombinant PI4KB and ACBD3 in vitro (16,17). In agreement with this, the I 43 A mutant failed to coimmunoprecipitate ACBD3 from cells (Fig. S7). While wt PI4KB reconstituted in PI4KB KO cells colocalized with 3A and ACBD3 ( Fig. 6B and C, top panels), the PI4KB-I 43 A mutant did not colocalize with 3A and ACBD3, and instead mostly localized to the nucleus (middle panels). Unexpectedly, the D 44 A mutant did colocalize with 3A and ACBD3 (bottom panels). In line with this, wt PI4KB and the D 44 A mutant could restore enterovirus replication in PI4KB KO cells, while the I 43 A mutant and the negative controls, EGFP and a PI4KB kinase-dead mutant that lacks catalytic activity (PI4KB-KD), could not (Fig. 6D). Why the D 44 A mutant behaves differently from the I 43 A mutant is presently unclear. Possibly, the D 44 A mutant has residual interaction with ACBD3 that could not be detected in vitro. Nevertheless, these results imply that the interaction between ACBD3 and PI4KB is important for enterovirus replication.
Enterovirus replication does not require the ACB and CAR domains of ACBD3. Four domains are recognized in ACBD3; the acyl-CoA binding (ACB) domain, the charged-amino-acid region (CAR), the glutamine-rich region (Q), and the Golgi dynamics domain (GOLD) (Fig. 7A). The ACB domain, which is relatively conserved among all ACBD3 Mediates the 3A-PI4KB Interaction ® known ACBD proteins (ACBD1 to ACBD7), has been suggested to be important for binding to long-chain acyl-CoA (29) and for binding to sterol regulatory element binding protein 1 (SREBP1), causing reduction of de novo palmitate synthesis (30). The CAR domain contains a nuclear localization signal (31), yet the function of the CAR domain is unknown. The Q domain interacts with the N-terminal helix of PI4KB (12,18). The GOLD domain interacts with giantin, and by doing so, tethers ACBD3 to the Golgi membrane (31). Enterovirus and kobuvirus 3A proteins bind to the GOLD domain, most probably at the same site as giantin (7,18). To investigate the importance of the ACB and CAR domains for enterovirus replication, we tested whether N-terminal deletion mutants of ACBD3 could restore enterovirus replication in ACBD3 KO cells. mut1 and mut2, which contain intact Q and GOLD domains, could restore virus replication to a level comparable to cells reconstituted with wt ACBD3 (Fig. 7B). This is in alignment with our observation that these mutants colocalize with 3A and PI4KB in ACBD3 KO cells (Fig. 7C). In contrast, mut3, which contains only the GOLD domain, could not rescue virus replication (Fig. 7B) or PI4KB recruitment (Fig. 7C), like the negative controls (GalT and the FQ mutant) (Fig. 7B), even though all mutants colocalized with 3A (Fig. 7C). Of note, although mut3 and 3A colocalized in punctate structures, they also partly colocalized to reticular and nuclear envelope-like structures, which may indicate that PI4KB also plays a role in firmly localizing 3A and ACBD3 to Golgi-derived membranes. These results indicate that enterovirus repli-  Values were statistically evaluated compared to the EGFP control using a one-way ANOVA. ****, P Ͻ 0.0001; N.S., not significant. (C) HeLa ACBD3 KO cells were cotransfected with plasmids encoding myc-tagged CVB3 3A and EGFP-tagged ACBD3 wt, or ACBD3 N-terminal deletion mutants (mut1 to -3). The next day, the cells were fixed and stained with antibodies against PI4KB (red) and the myc tag to detect 3A (light blue). Nuclei were stained with DAPI (blue). Bars represent 10 m. cation requires the Q and GOLD domains of ACBD3 for localization of viral protein 3A to the Golgi and for hijacking PI4KB.

DISCUSSION
Both enteroviruses and kobuviruses of the Picornaviridae family coopt PI4KB to build up ROs. Viral protein 3A is responsible for PI4KB recruitment to enterovirus replication sites, yet the underlying mechanism has remained elusive. Despite the direct interaction between ACBD3 and enterovirus 3A proteins (8,18), there has not yet been a consensus about the importance of ACBD3 for enterovirus replication and PI4KB recruitment. Previously, we observed no inhibition of CVB3 and RV replication and no effects on PI4KB recruitment, even though more than 90% of ACBD3 knockdown was achieved by siRNA (18,19). In the present study in which we use ACBD3 KO cells, we showed that ACBD3 is an important host factor for replication of four different human enterovirus species (EV-A/B/C/D) and two rhinovirus species (RV-A/B). All viruses showed impaired growth in ACBD3 KO cells ( Fig. 1 and Fig. S2). In addition, neither virus infection (Fig. 2) nor the expression of enterovirus 3A proteins alone (Fig. 3) elicited PI4KB recruitment in the absence of ACBD3. In agreement with our data, the inhibition of EV-A71 and CVB3 was recently reported in ACBD3 KO cells (32)(33)(34). The discrepancy in the role of ACBD3 from KD to KO condition could result from insufficient suppression of ACBD3 function by RNA interference. In fact, this implies that the small amounts (ϳ10%) of ACBD3 that remained after knockdown are sufficient to support enterovirus replication and PI4KB recruitment. Similar issues on the differences between knockdown and knockout have been raised by others (35). For instance, the importance of cyclophilin A (CypA) in nidovirus replication was prominent only in CypA knockout cells but not in knockdown cells, even though CypA protein was undetectable after knockdown (36).
We observed that the lack of ACBD3 has a profound effect on enterovirus 3A protein localization. 3A proteins were found almost exclusively at the ER in ACBD3 KO cells (Fig. 4D), whereas in HeLa wt cells, they showed a punctate localization mostly on Golgi-derived membranes ( Fig. 4B and C). Upon reconstitution of wt ACBD3 in ACBD3 KO cells, the localization of 3A was restored to a punctate pattern (Fig. 5A). These findings hint at a new role of ACBD3 for enterovirus replication, which is more than merely being a connector between 3A proteins and PI4KB.
Considering that ACBD3 is involved in several different protein complexes, enteroviruses may take advantage of ACBD3 in several ways, more than just for PI4KB recruitment. ACBD3 may be a scaffold responsible for positioning 3A near cellular factors, including other ACBD3-interacting proteins required for RO formation. For example, ACBD3 and PV1 3A were found in a protein complex together with the putative Rab33 GTPase-activating proteins TBC1D22A/B (37). In addition, several Golgi stacking proteins such as Golgin45 and Golgi reassembly stacking protein 2 (GORASP2) were recently identified as novel interaction partners of ACBD3, and ACBD3 was proposed as a scaffold tethering Golgin45, GRASP55, and TBC1D22 for the formation of a Golgi cisternal adhesion complex at the medial Golgi (38). It is largely unknown which domains of ACBD3 are responsible for the interaction with the above-mentioned interacting partners and whether these proteins are recruited to enterovirus ROs also remains to be investigated.
The GOLD domain of ACBD3 is responsible for the interaction with enterovirus 3A protein (18,37), while the Q domain interacts with PI4KB (16,17). By utilizing mutants of ACBD3 or PI4KB which disturb the interaction with each other (Fig. 5 and 7), we show for the first time that the interaction between ACBD3 and PI4KB is crucial for enterovirus replication. Aside from the Q and GOLD domains, other domains (i.e., ACB and CAR) of ACBD3 seem to be not involved in enterovirus replication (Fig. 7). This indicates that the functions of the ACB and CAR domains, as well as the cellular proteins and/or lipids that interact with these domains, are unlikely required for enterovirus replication. Although we cannot exclude the possibility that additional or unidentified proteins that bind to the Q or GOLD domain of ACBD3 might also be important for enterovirus replication, our findings suggest that ACBD3 mainly serves to coordinate 3A and PI4KB recruitment at RO membranes, involving the Q and GOLD domains.
How exactly enterovirus 3A protein interacts with ACBD3 needs to be further investigated. The GOLD domain of ACBD3 interacts with enterovirus 3A proteins (18). Similarly, kobuvirus 3A interacts with the ACBD3 GOLD domain, and the crystal structure of kobuvirus 3A in complex with the ACBD3 GOLD domain was revealed recently (6). According to this structure, kobuvirus 3A wraps ACBD3 and stabilizes ACBD3 on membranes through the membrane binding features at the myristoylated N-terminal and hydrophobic C-terminal ends of 3A. However, enterovirus 3A proteins can bind to membranes only through the hydrophobic C terminus, and the 3A proteins of enteroviruses differ greatly in sequence from kobuvirus 3A. Therefore, the way by which enterovirus 3A interacts with ACBD3 could be different from that of kobuvirus. Thus, structural insight into the enterovirus 3A-ACBD3 GOLD complex is urgently required to understand how enterovirus 3A interacts with ACBD3.
In conclusion, our study reveals that enteroviruses employ a conserved mechanism to recruit PI4KB, which depends on the Golgi-residing protein ACBD3. Furthermore, we suggest that ACBD3 tethers viral and host proteins to form ROs. Considering the pan-enteroviral dependency on ACBD3, targeting ACBD3 or the 3A-ACBD3 interaction presents a novel strategy for broad-spectrum antiviral drug development.

MATERIALS AND METHODS
Cells and culture conditions. HAP1 wt cells and HAP1 ACBD3 KO cells were obtained from Horizon Discovery. HeLa R19 cells were obtained from G. Belov (University of Maryland and Virginia-Maryland Regional College of Veterinary Medicine, US). HAP1 cells were cultured in IMDM (Thermo Fisher Scientific) supplemented with 10% fetal calf serum (FCS) and penicillin-streptomycin. HeLa cells and HEK 293T cells (ATCC CRL-3216) were cultured in DMEM (Lonza) supplemented with 10% FCS and penicillinstreptomycin. All cells were grown at 37°C in 5% CO 2 .
Virus infection. Virus infections were carried out by incubating subconfluent HAP1 or HeLa cells for 30 min with virus. Following virus removal, fresh medium or medium containing the control inhibitors guanidine hydrochloride (2 mM) or dipyridamole (100 M) was added to the cells. To determine one-step growth kinetics for each virus, infected cells were frozen from 2 to 16 h postinfection (p.i.). Virus titers were determined by endpoint titration analysis and expressed as 50% tissue culture infectious dose (TCID 50 ). To check for the recruitment of PI4KB upon virus replication, cells were fixed for immunofluorescence staining as described below separately. To check genome replication by measuring intracellular Renilla luciferase activity, cells were lysed at 8 h p.i. and followed the manufacturer's protocol (Renilla luciferase assay system; Promega).
RNA transfection. The subgenomic replicons of CVB3 (10) and EV-A71 (42) were described previously. HeLa cells were transfected with RNA transcripts of replicon constructs. After 7 h, cells were lysed to determine intracellular firefly luciferase activity.
Immunofluorescence microscopy. HeLa cells were grown on coverslips in 24-well plates. Subconfluent cells were transfected with 200 ng of plasmids using Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer's protocol or infected with CVB3 at an MOI of 5. At 16 h posttransfection (p.t.) or 5 to 9 h p.i., cells were fixed with 4% paraformaldehyde for 15 min at room temperature. After permeabilization with 0.1% Triton X-100 in PBS for 5 min, cells were incubated with primary and secondary antibodies diluted in 2% normal goat serum in PBS. Nuclei were stained with DAPI. Coverslips were mounted with FluorSave (Calbiochem), and confocal imaging was performed with a Leica SpeII confocal microscope.
Western blot analysis. HAP1 and HeLa cells were harvested and lysed by TEN lysis buffer (50 mM Tris HCl [pH 7.4], 150 mM NaCl, 1 mM EDTA, 1% NP-40, 0.05% SDS). After 30-min incubation on ice, lysates were centrifuged for 20 min at 10,000 ϫ g. Supernatants were boiled in Laemmli sample buffer for 5 min at 95°C. Samples were run on polyacrylamide gels and transferred to PVDF membranes (Bio-Rad). The membranes were incubated with primary antibody against ACBD3 or PI4KB at 4°C overnight and then with secondary antibodies against mouse IgG or rabbit IgG for 1 h at room temperature. Images were acquired with an Odyssey imaging system (LI-COR).