Phosphorylation of the HIV-1 capsid by MELK triggers uncoating to promote viral cDNA synthesis

Regulation of capsid disassembly is crucial for efficient HIV-1 cDNA synthesis after entry, yet host factors involved in this process remain largely unknown. Here, we employ genetic screening of human T-cells to identify maternal embryonic leucine zipper kinase (MELK) as a host factor required for optimal uncoating of the HIV-1 core to promote viral cDNA synthesis. Depletion of MELK inhibited HIV-1 cDNA synthesis with a concomitant delay of capsid disassembly. MELK phosphorylated Ser-149 of the capsid in the multimerized HIV-1 core, and a mutant virus carrying a phosphorylation-mimetic amino-acid substitution of Ser-149 underwent premature capsid disassembly and earlier HIV-1 cDNA synthesis, and eventually failed to enter the nucleus. Moreover, a small-molecule MELK inhibitor reduced the efficiency of HIV-1 replication in peripheral blood mononuclear cells in a dose-dependent manner. These results reveal a previously unrecognized mechanism of HIV-1 capsid disassembly and implicate MELK as a potential target for anti-HIV therapy.


Introduction
During the course of human immunodeficiency virus type 1 (HIV-1) infection, the virus encounters numerous bottlenecks constituted by a variety of host cell proteins essential for or inhibitory to HIV-1 replication [1]. HIV-1 particles must attach to and fuse with the plasma membrane of target cells, releasing the viral core into the cytoplasm. Shortly after entry, the HIV-1 capsid (CA), a major component of the viral core, starts dissociating from the core [reviewed by [2][3][4][5]]. It has been shown that optimal dissociation of CA from the HIV-1 core is required for efficient viral cDNA synthesis in target cells [6][7][8]. Thus, (i) Rhesus monkey TRIM5α abrogates productive reverse transcription (RT) by accelerating the disassembly of CA [9,10]; (ii) CA mutations that impair HIV-1 infection are unable to achieve proper uncoating and RT [6][7][8]; (iii) The prevention of RT with RT inhibitors causes CA disassembly delay [11,12]; and (iv) Uncoating of the HIV-1 CA core is triggered following first strand transfer of reverse transcription [13]; (v) The progression of reverse transcription causes morphological and mechanical changes in the HIV-1 cores [14]. Overall, these observations suggest that proper dissociation of CA is functionally linked to reverse transcription of HIV-1. This is also supported by studies showing that cytoplasmic accumulation of CPSF6 restricts HIV-1 infection through abnormal stabilization of the HIV-1 core [15][16][17]. HIV-1 CA is likely to interact with multiple host cell factors during uncoating and trafficking to the nucleus [3][4][5]18]. However, it remains poorly understood how the HIV-1 core dissociation process is triggered and regulated by host factors. One important consideration lies in the phosphorylation of CA because previous studies have shown that its phosphorylation plays pivotal roles in the viral life cycle [19][20][21]. For example, Ser-109 located in the amino-terminal domain, Ser-149 in the flexible linker and Ser-178 in the carboxy-terminal domain have been identified as major phosphoacceptor sites in CA which are essential for virus replication [19]. Alanine substitution at Ser-109, Ser-149 or Ser-178 reduces the phosphorylation level of CA in cell-free virions and inhibits efficient viral cDNA synthesis [19,22]. Furthermore, these mutations cause aberrant CA assembly or impaired core stability [21][22][23]. Phosphorylation of other amino acid residues in CA has also been reported to contribute to viral replication [24][25][26][27]. In terms of capsid phosphorylation by virion-associated kinases, the catalytic subunit of cAMP-dependent protein kinase (C-PKA) was reported to interact with and phosphorylate CA, and thus regulate viral infectivity, although the residues that were phosphorylated were not identified [28,29]. A recent study showed that virion-associated extracellular signal-regulated kinase 2 (ERK2) phosphorylates Ser-16 in CA [30], while an earlier study showed that HIV-1 CA is not a direct substrate of MAPK/ERK2 [19]. Thus, the significance of the phosphorylation of each of the amino acid residues in CA and the contribution of host cell kinases to HIV-1 replication remains to be fully elucidated. In the current study, we performed a genome-wide RNAi screen in a human T-cell line to identify host factors that contribute to HIV-1 infection. We found that a cellular kinase, MELK, is responsible for phosphorylation of the HIV-1 CA in target cells.

Results
Genome-wide RNAi screen identifies MELK as a host factor required for HIV-1 replication To identify host cell factors involved in HIV-1 replication in human cells, we employed a genome-wide RNAi screen in the MT4C5 lymphoid cell line, a derivative of MT4 cells expressing CCR5 and susceptible to infection with CXCR4-tropic and CCR5-tropic HIV-1 strains. MT4C5 cells were transduced with ten independent pools of puromycin-marked lentivirus vectors expressing a human short hairpin RNA (shRNA) library comprising >75,000 shRNAs directed against >15,000 human genes in total. Transduced cells were then infected with the HIV-1 NL4-3 strain which normally kills infected MT4C5 cells. Surviving cells were then studied further ( Fig 1A). Using this approach, 32 individual shRNA sequences were obtained that potentially target host factors, including Cyclophilin A and Transportin-SR2 (TNPO3), which are known to be essential host factors [1,31] (S1 Table). Of these, we have characterized maternal embryonic leucine-zipper kinase (MELK) in detail. This was identified in a sub-pool resistant to HIV-1 infection. MELK is a member of the AMP-activated protein kinase-related Ser/ Thr protein kinase family [32]. Previous reports indicated that MELK is expressed mainly in the cytoplasm and is involved in different cellular processes such as cell-cycle progression, cell proliferation and pre-mRNA splicing [33][34][35][36][37]. However, involvement of MELK in HIV-1 replication has not been reported.
To determine whether endogenous MELK is involved in HIV-1 infection, we depleted this enzyme from MT4C5 cells (MELK-KD). Lentivirus-mediated stable expression of a MELKtargeting shRNA, but not that of non-targeting shRNA (Non-T), suppressed expression of MELK mRNA (Fig 1B,  In contrast, MELK depletion did not affect the infectivity of the VSV-G-pseudotyped murine leukemia virus (MLV)-based vector (Fig 1C, right panel). HIV-1 replication in MT4C5 cells with the replication-competent NL4-3 virus was markedly inhibited by MELK depletion (Fig 1D). Viral DNA synthesis by replication-competent HIV-1 proceeded more slowly than by VSVGpseudotyped HIV-1, continuing until 24 h after infection, as previously reported [38,39]. Of note, MELK depletion did not significantly affect the amount of immediate early viral cDNA quantified at the R/U5 region 2 h post-infection ( Fig 1E, upper panel), but profoundly reduced it thereafter by approximately 80% compared to Non-T (Fig 1E, upper panel). This is because amplification of the R/U5 region includes both Early and Late RT products. Similar results were obtained by quantifying viral cDNA at the pol and env regions as Late RT products (approximately 80% reduction compared to Non-T) (Fig 1E, middle and lower panels). Viral  (Fig 1E, compare Non-T, Non-T + AZT and Non-T + NVP). Collectively, these results indicate that MELK is a host factor required for efficient viral cDNA synthesis.

MELK depletion delays HIV-1 capsid disassembly
We first investigated whether MELK affects HIV-1 entry using the HIV-1 virion fusion assay with β-lactamase-Vpr chimeric protein incorporated into HIV-1 virions. This approach revealed similar efficiencies of HIV-1 entry into control Non-T and MELK-KD MT4C5 cells and a marked inhibition of HIV-1 fusion in the presence of the CXCR4 antagonist AMD3100 [40] (Fig 2A and 2B, compare Non-T and MELK-KD-2). Quantitative RT-PCR assays also showed similar amounts of incoming viral genome RNA 2 h post infection (Fig 2C, compare Non-T and MELK-KD-2). We next assessed whether it is involved in proper disassembly of the viral CA. To determine whether the viral core interacts with MELK, we purified One-STrEP-FLAG-(OSF)-tagged MELK protein expressed in HeLa cells, and viral cores from cellfree virions. The HIV-1 envelope was removed from virions and envelope-stripped cores were enriched by ultracentrifugation through a discontinuous 10% and 30% sucrose gradient with 0.1% Triton X-100 in the 10% sucrose layer, as previously reported [20] (Fig 2D). The envelope-stripped cores were characterized by transmission electron microscopy (TEM) showing recognizable~100 nm cone-shaped structures similar to authentic HIV-1 cores (Fig 2E). Purified OSF (N-terminal)-tagged Cyclophilin A (CypA) or FLAG-One-STrEP (FOS) (C-terminal)-tagged rhesus monkey Trim5α (rhT5α) proteins, known to be HIV-1 core-binding proteins [9,[41][42][43], were used as positive controls for binding to the HIV-1 core. Pull-down assays revealed that OSF-tagged CypA or FOS2-tagged rhT5α, but not OSF-tagged Green Fluorescent Protein (GFP), significantly interacted with the envelope-stripped core in a dosedependent manner (Fig 2F, lower panel CA, compare lanes 2-3, 6-7 and 8-9). Similar to CypA and rhT5α, MELK interacted with the envelope-stripped core (Fig 2F, lower panel CA, compare lanes 2-3 and 4-5). Immunoblot analyses revealed that the envelope-stripped cores, but not enveloped virions or the CA monomer, interacted in vitro with OSF-tagged MELK (S4 Fig). We next tested whether MELK affects the stability of the HIV-1 core after viral entry, using a fate-of-capsid assay, as described previously [39] and summarized in Fig 2G. Previous reports showed that uncoating was closely linked to reverse transcription, using VSVG-pseudotyped lentivirus vectors that enter target cells in large numbers [13,18,[44][45][46] and far more quickly by endocytosis than the replication-competent HIV-1 does through CD4-and CXCR4-mediated fusion with the plasma membrane. As far as we know, we used for the first time replication-competent HIV-1 in the fate of capsid assay to demonstrate how MELK acts on HIV-1 in an experimental setting more relevant to human pathology. We chose to focus on the time point 8 h post infection for this fate-of-capsid assay for the following reasons: (i) previous reports showed that reverse transcription products reached a maximum 24 h after calculated from five independent experiments. (D) Effect of MELK depletion on HIV-1 replication in MT4C5 cells. The virion-associated RT activity was monitored at the indicated time points in culture supernatants of MT4C5 (closed squares), Non-T (closed circles) and MELK-KD-2 (open circles) cells. (E) Quantitative DNA-PCR analyses of viral cDNA metabolism after HIV-1 infection of MT4C5-derived cells in the presence or absence of AZT (5 μM) or NVP (10 μM). Total DNA was extracted from a portion of the cells 2 h after infection and early viral cDNA synthesis was quantified by real-time PCR with a primer set recognizing the R/U5 region (top panel). Total DNA was extracted at the indicated time points (4, 8 and 24 h) and analyzed for the amount of late viral cDNA with a primer set recognizing the R/U5 region (top panel), env (middle panel) or pol (bottom panel) regions. The ratios of each viral cDNA level to beta-globin DNA level are given. Experiments were performed at least three times and error bars are standard deviations calculated from three independent experiments. Statistical significance was determined by one-way analysis of variance (ANOVA) with Dunnett's multiple comparison test (C). ns, not significant (P>0.05); *P<0.05, **P<0.01, ***P<0.001.  Non-T or MELK-KD-2 MT4C5 cells were mock-infected or infected with 100 or 500 ng of p24-measured amounts of NL4-3 virions containing BlaM-Vpr, based on the measured amount of p24, in the presence or absence of AMD3100 (100 nM). They were then analyzed in the fusion assay by flow cytometry using a violet laser to excite CCF2. Each experiment was performed in triplicate, repeated three times and one set of representative data is shown. (B) Relative numbers of BlaM + MELK-KD-2 MT4C5 cells are shown as percentages (%) of Non-T MT4C5 cells with standard deviations calculated from three independent experiments. (C) Virion-associated viral RNA was quantified by quantitative RT-PCR 2 h after infection of Non-T or MELK-KD-2 MT4C5 cells infection with replication-competent HIV-1 [38,39]; (ii) our requirement studies indicated that the amount of total intracellular CA 4 h after infection remained so small that it was impossible to detect pelletable CA by western blotting following ultracentrifugation (S5 Fig). Previous studies had shown that ectopic expression of the rhT5α protein accelerated uncoating and restricted HIV-1 infection, and that a reverse transcriptase inhibitor, NVP, delayed CA disassembly [9,11,12,47]. To confirm the validity of this fate-of-capsid assay, we established MT4C5 cell pools stably expressing C-terminally hemagglutinin (HA)-tagged rhT5α (rhT5α-HA) (Fig 2H, upper panel). As reported previously [47], rhT5α strongly inhibited HIV-1 infection ( Fig 2H, lower panel). Control Non-T, MELK-KD, rhT5α-HA-expressing cells and Non-T cells treated with NVP were infected with wild-type HIV-1. We found that the amount of HIV-1 cores in HIV-1-infected cells expressing rhT5α-HA 8 h post-infection was significantly lower than in control cells (Non-T) ( . This suggests that CA monomers dissociated from multimerized cores undergo degradation in living cells. Consistent with a previous report [48], degradation of incoming CA protein in our hypotonic lysis buffer was accelerated by rhT5α ( Fig 2I, panel CA of cell lysate 8h, MG132[-], compare Non-T and rhT5α-HA). The report also showed that the proteasome inhibitors MG132 and lactacystin caused markedly increased steady-state levels of incoming CA protein in the cytosol of HeLa cells expressing non-restrictive human Trim5α or rhT5α [48]. We therefore used the proteasome inhibitor MG132 to retain the CA in infected cells as far as possible in order to show how much CA was actually present in infected cells at the time of the assay. Indeed, the inhibition of CA degradation by the proteasome revealed that similar amounts of CA were present in infected cells (Fig 2I,  These results clearly indicate that MELK is required for optimal HIV-1 capsid disassembly in newly infected cells.

Catalytic activity of MELK regulates HIV-1 replication
Transduction of MELK-depleted MT4C5 expressing an shRNA targeting the 3 0 -untranslated region (3 0 -UTR) of MELK (MT4C5-MELK-KD-1) with a lentivirus vector capable of expressing wild-type MELK substantially restored HIV-1 infectivity in two independent cell pools ( Fig  3A, compare lanes 4 and 5 or 6). In contrast, a MELK mutant (T167A) that lacks catalytic activity [37]  To determine whether CA is a substrate for MELK, we prepared recombinant CA fused to GST (GST-HIV-CA) and employed an in vitro luminescent kinase assay, in which the amount of ADP produced in the kinase reaction was quantified (for details, see "Materials and methods"). MELK phosphorylated ZIPtide, a substrate for MELK [49] (Fig 3B, upper panel), which was inhibited by the small-molecule MELK inhibitor OTSSP167 [50] in a dose-dependent manner  Fig 2F implied that MELK preferentially recognized multimerized CA cores, we next determined whether MELK phosphorylates CA in a structure-dependent manner. In vitro luminescent kinase assays revealed that env-stripped HIV-1 cores were much more efficiently phosphorylated by MELK than . To explore which Ser or Thr residue(s) in CA can be phosphorylated by MELK, we generated fifteen peptides that covered all the regions containing Ser or Thr residues in CA ( Fig 3E). In vitro luminescent kinase assays revealed that peptides #8 and #9 were phosphorylated by MELK in a substrate dose-dependent manner ( Fig 3F). These results suggest that Thr-119, Ser-146, Thr-148 and Ser-149 of CA could be phosphorylation targets of MELK.

Phosphorylation of Ser-149 in CA by MELK regulates HIV-1 uncoating in target cells
We next explored whether an amino-acid substitution that mimics phosphorylation of each serine or threonine residue, Thr-119, Ser-146, Thr-148 and Ser-149, counteracts the delay of CA disassembly and reduction in viral cDNA synthesis caused by MELK depletion. We generated four mutant pNL4-3 proviral molecular clones in which each Ser or Thr residue was substituted by a glutamic acid residue so that each mutation mimics constitutive phosphorylation of the site. Mutant HIV-1 bearing T119E, S146E, T148E, or S149E mutations in CA were used to evaluate the efficiency of viral cDNA synthesis in MELK-KD MT4C5 cells. The amount of each input CA-mutated virus was normalized by its RT activity. HIV-1 bearing T119E, S146E or T148E poorly restored early ( Fig 4A) and late ( Fig 4B) cDNA synthesis in MELK-KD MT4C5 cells. In contrast, viral cDNA synthesis after infection with HIV-1 bearing the S149E mutation was robustly restored, and even at an earlier time point than wild-type HIV-1 (approximately 1.7-fold increase compared to Non-T-wt, 8 h post-infection) ( Fig 4A  and 4B, MELK-KD-2-S149E). The S149E mutation did not significantly alter the amount of incoming HIV-1 RNA in MELK-KD or control Non-T MT4C5 cells (S9G Fig, compare NL4-3wt and NL4-3 CA S149E). In control MT4C5 cells expressing non-target shRNA, only the S149E mutation caused an earlier peak and subsequent downturn in viral cDNA synthesis similar to that in MELK-KD MT4C5 cells (S9A- S9D Fig). This suggests that phosphorylation of Ser-149 is likely to play an important role in the initiation of viral cDNA synthesis. Despite maintenance of efficient cDNA synthesis by the S149E mutant, production of the 2-LTR circular form of viral cDNA, a marker for nuclear import, was undetectable ( Fig 4C, MELK-KD-2-S149E). This shows that this CA mutation promotes viral cDNA synthesis, but does not favor nuclear import. HIV-1 bearing T119E or T148E but not S146E mutations appeared to partially restore production of the 2-LTR circular form (Fig 4C), suggesting that although these mutants failed to substantially restore viral DNA synthesis, they were still competent for nuclear entry. Single-round infection assays using TZM-bl or LuSIV indicator cells revealed very low but detectable infectivity of HIV-1 bearing CA T119E or T148E. However, infection with S146E or S149E CA mutants was undetectable (S10C Fig).  To assess how the S149E mutation in CA influences the kinetics of capsid disassembly, we performed a fate-of-capsid assay in control Non-T and MELK-KD MT4C5 cells. The S149E substitution resulted in a clear decrease of the HIV-1 core not only in MELK-KD cells but also in Non-T cells, indicating that this mutation promoted the CA disassembly irrespective of the presence of MELK ( . Consequently, the ratio of pelletable S149E CA to total CA in the absence of MG132 quantified by p24 ELISA was significantly less than that of NL4-3wt in both MELK-KD and control Non-T cells (Fig 4E, compare WT and S149E). Overall, these results suggest that phosphorylation of Ser-149 by MELK is a trigger for CA disassembly in HIV-1 infection.
We also attempted to characterize HIV-1 with an S149A mutation expected to confer refractoriness to phosphorylation by MELK, but the titer of the S149A virus was too low to compare its infectivity with NL4-3wt and S149E mutant. This is consistent with a previous study that the S149A mutation affects the production of infectious virions [23]. We therefore evaluated the infectivity of VSV-G-envelope-pseudotyped NL4-3luc bearing the S149A mutation in CA (VSVG/NL4-3luc CA-S149A) in Non-T and MELK-KD MT4C5 cells. We did this because the infectivity of S149A mutant virus was reported to be rescued specifically by pseudotyping with the VSV envelope protein [22,23]. The S149A mutation greatly reduced the VSVG/NL4-3luc-derived reporter gene activity in parental and Non-T cells, and also, but less markedly, in MELK-KD cells (S11A Fig, panels MT4C5, Non-T and MELK-KD-2, compare CA-wt and CA-S149A). Additionally, depletion of MELK modestly reduced the reporter gene activity of VSVG/NL4-3luc CA-S149A compared to Non-T control cells (S11B Fig). These findings may in part reflect unidentified dysfunctionality of this mutant CA, but also raise the possibility that MELK regulates HIV-1 replication in a manner in addition to phosphorylating CA Ser-149. A GST-linked S149A CA (GST-HIV-CA S149A) was analyzed using the in vitro luminescent kinase assay to determine whether any other residues in CA can be phosphorylated by MELK (Fig 5A). Essentially no phosphorylation could be detected when activated MELK was incubated with increasing amounts of GST-HIV-CA S149A (Fig 5A, compare GST-HIV CA and GST-HIV CA S149A).
To directly test the ability of MELK to phosphorylate Ser-149 in CA, we generated rabbit polyclonal antibodies that recognize only phosphorylated Ser-149 in CA (CA-S149p). Fig 5B shows that phosphorylation of virion-associated CA was undetectable with CA-S149p (top panel CA-S149p, lanes 2 and 3). We next performed an in vitro phosphorylation assay with Fate-of-capsid assays with non-target shRNA (Non-T) or MELK-KD-2 MT4C5 cells infected with NL4-3 or its S149E CA mutant for 8 h in the presence or absence of 10 μM MG132 (MG132 [+] or MG132 [-]). HIV-1 stock inactivated by incubation at 65˚C for 30 min was used as a negative control (HI control). Cell lysates were prepared and analyzed as in Fig 2I. Aliquots of input, fraction #1 and #3 were processed for immunoblotting with anti-p24 antibody (CA). The amount of CA in each fraction in the absence of MG132 was quantified by HIV-1 p24 ELISA (MG132 [-], lower panels). Experiments were performed five times and one representative set of data is shown. (E) Percentages of pelletable CA (fraction #3) within total CA in the absence of MG132 were calculated based on the p24 ELISA data shown in Fig 4D. Total CA denotes the sum of the amount of p24 antigen which was calculated based on the p24 ELISA data of fractions #1, #2, and #3. Error bars represent the standard deviations calculated from five independent experiments. Statistical significance was determined by one-way analysis of variance (ANOVA) with Tukey's multiple comparison test (D), or unpaired two-tailed Student's t test (E). ns, not significant (P>0.05); *P<0.05, **P<0.01, ***P<0.001. https://doi.org/10.1371/journal.ppat.1006441.g004 Essential role for MELK in HIV infection Forced expression of MELK leads to premature HIV-1 capsid disassembly and viral cDNA synthesis in target cells Our observation that constitutive phosphorylation of Ser-149 caused premature capsid disassembly and early viral cDNA synthesis but failed to support nuclear import of viral DNA suggests that a well-ordered phosphorylation of Ser-149 during CA disassembly is required for optimal uncoating, viral cDNA synthesis and nuclear import. We next examined whether forced expression of MELK in MT4C5 cells affects HIV-1 infection. Expression of exogenous and endogenous MELK mRNA was monitored by RT-PCR (Fig 6A). To assess how MELK overexpression influences the kinetics of capsid disassembly, we performed a fate-of-capsid assay in control (CSII-control), MELK-expressing (CSII-MELK) and catalytically inactive MELK-expressing (CSII-MELK T167A) MT4C5 cells. Overexpression of wild-type MELK resulted in an obvious decrease in pelletable CA (Fig 6B, right panels, compare CSII-control MELK in the presence of 100 nM or 500 nM of OTSSP167 [panel: OTSSP167 (100 nM or 500 nM)]. The samples were also treated with (+) or without (-) 100 U of calf intestine alkaline phosphatase (CIAP) and immunoblotted (lower 2 panels) with CA-S149p (panel: CA-S149p) or with anti-p24 antibody [panel: CA (reprobed)]. Experiments were performed three times and one representative set of data is shown. (C) Non-T or MELK-KD-2 MT4C5 cells were infected with VSV-G-pseudotyped HIV-1 or VSV-G-pseudotyped HIV-1 CA-S149A for 8 h. The proteasome inhibitor MG132 (2 μM) was added 5 h after infection to prevent the degradation of CA proteins dissociated from the viral core in the cytoplasm [48]. Cell lysates were separated by SDS-PAGE containing Manganese(II)-Phos-tag (Mn 2+ -phos-tag) or SDS-PAGE without Mn 2+ -phos-tag (Normal), and analyzed by immunoblotting with anti-p24 antibody (CA) or anti-alpha-tubulin antibody (α-tubulin). Cell lysates were incubated for 60 min at 37˚C without (lanes 2 and 3) or with (lanes 4 and 5) calf intestine alkaline phosphatase (CIAP). "non-pCA" indicates the position of CA dephosphorylated by CIAP and "pCA" indicates phosphorylated CA. Experiments were performed at least three times and one representative set of data is shown. (D) Non-T or MELK-KD-2 MT4C5 cells were infected with VSV-G-pseudotyped HIV-1 CA-S149A for 8 h. Cell lysates were separated as in (C) and analyzed by immunoblotting with anti-p24 antibody (CA) or anti-alpha-tubulin antibody (α-tubulin). Similar results were obtained in three independent experiments and a representative result is shown. Statistical significance was determined by two-way analysis of variance (ANOVA) with Tukey's multiple comparison test (A). ns, not significant (P>0.05); *P<0.05, **P<0.01, ***P<0.001. https://doi.org/10.1371/journal.ppat.1006441.g005

MELK inhibitor suppresses HIV-1 replication
We assessed whether OTSSP167 affects HIV-1 replication in MT4C5 cells. Single-round infection assays revealed that viral infectivity in the presence of OTSSP167 was substantially reduced in a dose-dependent manner (Fig 7A). OTSSP167 compromised viral cDNA synthesis in a dose-dependent manner ( Fig 7B) and also reduced viral infectivity in PHA-stimulated peripheral blood mononuclear cells (PBMCs) derived from two healthy donors (Fig 7C).
Similar results were obtained with the macrocyclic thiazole antibiotic Siomycin A (Fig 7D-7F), initially identified as an inhibitor of the transcription factor FOXM1b and thereafter reported to reduce MELK expression in brain tumor stem-like cells in vitro [52][53][54]. MELK expression was reduced by Siomycin A in a dose-dependent manner (S13A Fig), while there was no effect on MELK mRNA levels in MT4C5 cells (S13B Fig). HIV-1 replication in MT4C5 cells with the replication-competent NL4-3 virus in the presence of Siomycin A was markedly inhibited in a dose-dependent manner (S13C Fig). These findings suggest that a small-molecule inhibitor of MELK has potential as an anti-HIV lead compound.

Discussion
The main finding of the present study is that MELK regulates CA disassembly to promote viral cDNA synthesis through the phosphorylation of Ser-149 in CA during the early stages of HIV-1 infection. MELK depletion did not significantly alter the efficiency of HIV-1 entry (Fig 2A-2C), but did impair viral cDNA synthesis in association with a significant delay of CA disassembly during the early stages of HIV-1 infection (Figs 1E, 2I and 2J). Although MELK was identified based on HIV-1 infection-resistant cells during a spreading-infection, the inhibition (GAPDH). Experiments were performed three times and one set of representative data is shown. (B) Fate-ofcapsid assays. Non-T MT4C5 cells transduced with empty lentivirus vector (CSII-control), MELK expression vector (CSII-MELK) or MELK mutant expression vector (CSII-MELK T167A) were infected with NL4-3 for 8 h. HIV-1 stock inactivated by incubation at 65˚C for 30 min was used as a negative control (HI control). Cell lysates were prepared and analyzed as in Fig 2I. Aliquots of input, soluble fraction #1 and viral core fraction #3 were processed for immunoblotting with anti-p24 antibody (CA). Input cell lysates were also analyzed by immunoblotting with anti-MELK (MELK) and anti-alpha-tubulin antibodies (α-tubulin). Experiments were performed five times and one representative set of data is shown. The amount of CA in each fraction was quantified by HIV-1 p24 ELISA (lower panels). Error bars indicate the standard deviations calculated from five independent experiments.  by MELK-depletion in single-round infection assays was not complete (Fig 1C). This seems to be due, at least in part, to the difference in MOI between single-round infection assays (MOI, 1) and the initial genome-wide RNAi screen (MOI, 0.01). The different level of inhibitory effects shown in the uncoating (~2 fold) and infectivity (~10 fold) results may be explained by the widely observed finding that changes in uncoating influence reverse transcription and nuclear translocation, thereby amplifying the effect. The marked suppression of the spread of HIV-1 infection in MELK-depleted cells further suggests a role for MELK in other stages of virus replication. This is currently under investigation. The direct interaction of MELK with envelope-stripped core, but not with the monomeric form of CA, suggests an association between the host cell MELK and the incoming viral core in the cytoplasm (Fig 2D-2F). Moreover, the results in Fig 3D indicate that the env-stripped HIV-1 core is a much better substrate of MELK than GST-HIV-CA. Host cell core-binding factors such as MELK that accelerate CA dissociation from the viral core, and those that stabilize the core, such as CPSF6 [15][16][17], are likely required to achieve optimal stability of the viral core which is necessary for efficient viral cDNA synthesis in target cells. We further document, for the first time, phosphorylation of S149 in the multimerized viral core by MELK, and provide compelling evidence of in vivo phosphorylation of S149 (Fig 5B). In vitro phosphorylation assays have shown that the HIV-1 CA is a substrate of MELK ( Fig 3C) and that Thr-119, Ser-146, Thr-148, and Ser-149 are the candidate phosphorylation targets (Fig 3F). Ser-146, Thr-148 and Ser-149 are located in the flexible linker region that may allow movement of the C-terminal domain (CTD) relative to the N-terminal domain (NTD) [55][56][57], whereas Thr-119 is located in the amino-terminal domain of CA. The restoration of viral cDNA synthesis in MELK-KD cells by the S149E mutation is consistent with a previous report that the flexible linker region has a critical role in optimal core stability and efficient HIV-1 replication [23]. The S149E mutation caused an earlier peak and subsequent reduction of viral cDNA synthesis in control cells as in MELK-KD cells (S9D Fig). Infection of cells overexpressing MELK with wild-type HIV-1 resulted in premature CA disassembly and aberrant viral cDNA synthesis (Fig 6B and 6D). These results collectively indicate that unusual phosphorylation of S149 in CA by MELK misguides capsid disassembly and viral cDNA synthesis (Figs 4 and 6 and S9D Fig). HIV-1 with T119E, S146E or T148E mutations yielded less late RT product in MELK-KD cells than in control Non-T cells probably because in these mutants S149 remains intact (S9A- S9C Fig). They produced late RT products in control Non-T cells much less efficiently than did wild-type so that the proportional reduction in the late RT product was not as marked as in the wild-type (S9A- S9C Fig). This reduction indicates that these mutants are sensitive to MELK depletion and that the essential target of MELK is not T119, S146 or T148. T119E and S146E mutants produced a little more late RT product than did the wild-type in MELK-KD cells (Fig 4B) and eventually failed to increase or retain this at 24 h post-infection, suggesting that these mutations promoted reverse transcription independently of MELK (Fig 4B). A previous report showed that the CA mutations E128A/R132A increased the stability of the viral core and impaired viral cDNA synthesis, while CA mutations Q63A/Q67A accelerated CA disassembly and viral cDNA synthesis, but severely impaired viral infectivity [6]. This appears consistent with our results that MELK depletion caused both the delay of CA disassembly and reduction in viral cDNA synthesis in target cells. We showed here that the S149E mutation, like Q63A/Q67A mutations, accelerated PHA-stimulated PBMCs were infected with VSV-G-pseudotyped NL4-3luc in the presence of increasing amounts of Siomycin A. Error bars indicate the standard deviations calculated from five independent experiments. Statistical significance was determined by two-way analysis of variance (ANOVA) with Sidak's multiple comparison test (A and D), or one-way ANOVA with Dunnett's multiple comparison test (B, C, E and F). *P<0.05, **P<0.01, ***P<0.001. https://doi.org/10.1371/journal.ppat.1006441.g007 Essential role for MELK in HIV infection CA disassembly and viral cDNA synthesis, and severely impaired nuclear import of the viral cDNA (Fig 4). This suggests that successful viral cDNA synthesis may largely depend on the S149 phosphorylation-triggered CA disassembly and that appropriate timing and perhaps location of CA disassembly is necessary for the efficient nuclear import of viral cDNA. Indeed, previous reports showed that HIV-1 CA interacts directly with the nuclear pore complex (NPC) by binding to the cyclophilin-like domain of Nup358 [58,59]. It is plausible that aberrant disassembly of S149E CA potentially at an inappropriate location renders recognition of CA by Nup358 difficult. The rapid decline in the amount of viral cDNA in S149E mutantinfected cells (Fig 4A-4C and S9D Fig) also suggests that premature uncoating promotes viral DNA degradation in the cytoplasm. Poor production of the S149E mutant virus (S10A and S10B Fig) may further explain why this residue has to be phosphorylated after entry. MELKmediated phosphorylation of Ser-149 after entry may have evolved to optimize production of infectious virions and achieve an ordered CA disassembly and efficient nuclear entry. In terms of the coupled RT and capsid disassembly, it is tempting to postulate that phosphorylation of CA leads to a conformational change that unlocks the initiation of reverse transcription. This would in turn influence the activities of cytoplasmic factors involved in the regulation of capsid disassembly, such as MELK, and thereby further promote disassembly of the core.
The poor ability of the VSVG/NL4-3luc CA-S149A virus to infect control MT4C5 cells differs from a previous observation that VSVG-pseudotyped env-deleted HIV-1 CA-S149A mutants can infect LuSIV, TZM-bl and MAGIC-5B HeLa cells. This may be due in part to the different reporter systems and cell lines used. Our observation that alanine substitution of Ser-149 (S149A) ablated CA phosphorylation in vitro by MELK suggests that the Ser-149 residue is the sole phosphorylation target of MELK in CA (Fig 5A). The lack of difference in mobility shift of S149A CA in Non-T and MELK-KD cells shown in the Phos-tag assay further reinforces this notion (Fig 5D). A previous report showed that phosphorylation of three serine residues (Ser-109, Ser-149, and Ser-178) in CA is required for efficient reverse transcription and uncoating [19]. We therefore generated S109E and S178E mutant viruses, and found that this failed to support viral cDNA synthesis in both Non-T and MELK-KD cells, suggesting that MELK is not involved in phosphorylation of Ser-109 or Ser-178 (S9E and S9F Fig). Our Phos-tag result that treatment of lysates from HIV-1-infected MELK-KD cells with CIAP further down-shifted the CA bands suggests that CA is phosphorylated by other cellular kinases in HIV-1-infected cells (Fig 5C, top panel CA, compare lanes 3 and 5). Our results from S149E mutation experiments and overexpression of MELK regarding effects on viral DNA synthesis and nuclear import strongly suggest that phosphorylation of S149 under temporally and spatially appropriate conditions is important for enabling HIV-1 to proceed through the early stages of infection. OTSSP167 inhibits the catalytic activity of MELK while Siomycin A reduces its expression, both of which interfere with the function of MELK as a kinase and similarly reduce the infectivity of HIV-1 in PBMC (Fig 7). This decreases the likelihood of off-target effects of these drugs.
In conclusion, the essential role of MELK at an early stage of HIV-1 infection exemplifies another aspect of the functional links between viral capsid disassembly, cDNA synthesis and nuclear import. These findings contribute to our understanding of early viral life-cycle events and raise the possibility of developing a new class of anti-HIV agents targeting viral capsid disassembly. Reagent Program] cells were propagated in Dulbecco's modified Eagle medium containing 10% fetal bovine serum (FBS) and penicillin/streptomycin. MT4C5 (kindly provided by Tetsuro Matano's lab in National Institute of Infectious Diseases, Japan) cells were maintained in complete RPMI 1640 medium supplemented with 10% FBS and penicillin/streptomycin. LuSIV cells (NIH AIDS Research and Reference Reagent Program) were cultured in RPMI 1640 supplemented with 10% FBS, penicillin/streptomycin, and 300 μg per ml hygromycin B. Phytohemaggulutinin (PHA)-activated PBMCs (PHA-PBMCs) were cultured in RPMI 1640 containing 10% FBS, penicillin/streptomycin, and 100 U IL-2 per ml. CD3/CD28-stimulated peripheral blood lymphocytes (PBL) were prepared using human T-Activator CD3/CD28 Dynabeads (Thermo Fisher Scientific, Waltham, MA) and cultured in RPMI 1640 containing 10% FBS, penicillin/streptomycin, and 100 U IL-2 per ml.

Pharmaceuticals
Nevirapine (NVP) and Azidothymidine (AZT) were obtained from National Institutes of Health (NIH) AIDS Research and Reference Reagent Program. AMD3100 and MG132 were obtained from Sigma-Aldrich. Siomycin A was obtained from Bioaustralis. OTSSP167 was obtained from Selleck chemicals.

Establishment of an shRNA T-cell library
A puromycin-marked lentivirus vector-based shRNA library that targets over 15,000 human genes (Sigma-Aldrich, MISSION shRNA library) was used to establish shRNA-MT4C5 cell libraries. On average, there are five shRNA sequences designed for each gene target. The library was pre-divided into ten sub-pools of approximately 8,000 shRNA constructs. MT4C5 cells were transduced with the shRNA-lentivirus library and selected with puromycin (1 μg/ ml) for 2 weeks.

shRNA-based screening
Established shRNA library-expressing MT4C5 cell pools were then infected with HIV-1 NL4-3 strain. HIV-1 NL4-3 strain normally kills infected parental or Non-T control MT4C5 cells with a slight degree of syncytia formation, indicating effective infection-induced cell death. Two weeks after infection, cells were seeded into 96-well round-bottom cell culture plates. Several sub-pools resistant to HIV-1 infection were identified. Total cellular DNA was prepared from each sub-pool and used to detect the pol region of the HIV-1 NL4-3 late reverse transcription product by quantitative PCR using TaqMan PCR (Applied Biosystems, Carlsbad, CA). Positive samples were excluded as persistently infected cells. To determine the shRNA sequences in surviving cells free from NL4-3 DNA, total cellular DNA was extracted and the DNA fragments encoding the shRNA were amplified by PCR with the primers 5'-TACAAAATACGT GACGTAGAAA-3' and 5'-TTTGTTTTTGTAATTCTTTA-3'. The PCR products were cloned into the pCR4-TOPO vector (Invitrogen Corp., Carlsbad, CA). At least 100 PCR clones were sequenced for each surviving cell pool with the primer 5'-TTTGTTTTTGTAATTCTTTA-3'.

Establishment of MT4C5 cells stably expressing MELK or a MELK mutant (T167A)
Non-T MT4C5 cells were transduced with lentivirus vectors that confer blasticidin resistance and express MELK or a MELK mutant (T167A). Transduced cell pools were established after selection with 6 μg/ml blasticidin and 2 μg/ml puromycin for 7 days.

Reconstitution of MELK in MELK-depleted MT4C5 cells
Both Non-T MT4C5 and MELK-KD-1 MT4C5 cells established with the shRNA targeting the 3 0 -UTR of MELK were transduced with lentivirus vectors encoding a blasticidin resistance gene and expressing the coding region of MELK or T167A MELK mutant. Two independent pools of reconstituted cells were established for the wild-type and mutant MELK after selection with 2 μg/ml puromycin and 6 μg/ml blasticidin.

Measurement of viral RNA levels after viral entry
For infection, 5 × 10 5 target MT4C5 cells were incubated for 2 h with HIV-1 stock containing 1 × 10 6 RT counts that were pre-treated with 100 U of DNase I (Roche Applied Science, Indianapolis, IN) in the presence of 10 mM MgCl 2 for 20 min at 37˚C. DNA-free total cellular RNA was then extracted using RNeasy Mini Kits with on-column DNase digestion (QIAGEN Inc., Valencia, CA). HIV-1 stock inactivated by incubation at 65˚C for 30 min was used as a negative control. Primers 5'-ATTCCTGAGTGGGAGTTTG-3' (nt 3780-3798) and 5'-AACTTTC TATGTAGATGGGGC-3' (nt 3863-3883) and a probe 5'-FAM-CAATACCCCTCCCTTA GTGAAGTTATGGTAC-TAMRA-3' (nt 3800-3830) were used for amplification and detection of the pol region of the HIV-1 NL4-3 virion-associated RNA by quantitative RT-PCR using TaqMan One-Step RT-PCR (Applied Biosystems, Carlsbad, CA). For standardization, a primer/probe set of the 18S ribosomal RNA was used [61]. Real-time RT-PCR was carried out in a StepOnePlus Real-Time PCR system (Applied Biosystems, Carlsbad, CA). The ratios of each viral RNA level to 18S ribosomal RNA level are given.

Fluorescence resonance energy transfer-based HIV-1 virion fusion assay
A fusion assay was performed using HIV-1 possessing β-lactamase-Vpr chimeric proteins (BlaM-Vpr) and MT4C5-derived cells loaded with CCF2 dye, a fluorescent substrate for β-lactamase, as previously described [66]. In brief, X4 HIV-1 containing BlaM-Vpr (HIV-1 NL-E-BlaM-Vpr ) [67] was obtained by cotransfecting 293T cells with pNL-E plus pMM310 [68] encoding Escherichia coli β-lactamase fused to the amino terminus of Vpr [69]. MT4C5-derived cells (1×10 6 ) were infected with 10 or 100 ng of HIV-1 NL-E-BlaM-Vpr as a measured amount of p24 by spinoculation at 1200×g for 2 h at 25˚C as previously described [70]. Thereafter, cells were washed and incubated in RPMI containing 10% heat-inactivated fetal bovine serum for 2 h at 37˚C to induce viral fusion. Cells were then washed and loaded with CCF2-AM for 1 h at RT using a GeneBLAzer In Vivo Detection Kit (Invitrogen Corp., Carlsbad, CA). The dyeloaded cells were incubated overnight at RT and assayed by flow cytometry. Cells permissive for HIV-1 fusion were detected by their fluorescence at 447 nm after excitation with a 405-nm violet laser in a FACSCanto II. Dead cells were stained with propidium iodide and were gated out during analysis. AMD3100 was used as a control for fusion inhibition. To determine the absolute copy numbers of viral DNA or 2-LTR circles in HIV-1 infected cells, we employed a calibration curve using the pNL4-3 or pGEM/NL-2LTR [71] serially diluted with a constant amount of whole cell DNA from uninfected cells. The absolute amount of beta-globin DNA determined in the same way was used to normalize the results, as described previously [39]. The ratios of each viral cDNA level to beta-globin DNA level are given. In the case of cells transduced with a lentivirus vector containing the R/U5 region, amplified viral cDNA level in HIV-1-infected cells was determined after subtraction of the level in uninfected cells.

Analysis of HIV-1 replication in human T cells
MT4C5 cells (1 × 10 5 ) were exposed to HIV-1 stock containing 10 pg of p24. Virus production was monitored for 14 days post-infection by measuring RT activity in the culture supernatants. Mean values from three independent experiments are shown.

Fate-of-capsid assay
The fate-of-capsid assay was performed with minor modifications [39] as previously described. Briefly, 5 × 10 6 of 293, 293-non-target shRNA or 293-MELK-KD-3 cells were replated in a 10 cm plastic dish one day before assay. Cells were inoculated with 5 × 10 6 RT counts of VSV-G/ NL4-3luc virus. After incubation at 4˚C for 30 min, cells were incubated at 37˚C for 4 or 8 h. MT4C5-derived cells (5 × 10 6 ) were inoculated with 2 × 10 7 RT counts of wild-type HIV-1 (HIV-1 NL4-3 strain) prepared in HeLa cells. After incubation at 4˚C for 30 min, cells were incubated at 37˚C for 2, 8 or 24 h. Cells were then washed twice with ice-cold PBS(-) containing 0.005% Trypsin/EDTA to detach virions from the cellular surface and once with ice-cold PBS (-) to remove Trypsin/EDTA. Washed cells were resuspended in 1 ml of hypotonic lysis buffer [10 mM Tris-HCl (pH 8.0), 10 mM KCl, 1 mM EDTA and protease inhibitor cocktail (NACA-LAI TESQUE, INC, Kyoto, Japan)] and incubated on ice for 15 min. Swollen cells were lysed in a 7 ml-Dounce homogenizer with a 'tight' pestle (15 gentle strokes making a half-turn of the pestle per each stroke) and cell lysates cleared by centrifugation at 2,000 × g for 3 min at 4˚C. Cleared cell extracts (0.8 ml) were layered over 20%-60% sucrose cushions prepared in PBS and centrifuged at 4˚C and 35,000 rpm for 70 min in a Beckman SW50.1 rotor; 50 μl of the cell extract was reserved as a 'cell lysate' fraction. After centrifugation, three fractions of 1.1 ml each were collected from the top of the gradient. Aliquots of each fraction of the step gradients were subsequently processed for immunoblotting. The amount of CA protein in each fraction was quantified using HIV-1 CA (p24) enzyme-linked immunosorbent assay kits (ZeptMetrix Corporation, Buffalo, NY).

Isolation of envelope-stripped cores
Envelope-stripped HIV-1 cores were prepared as described previously [20]. Briefly, HIV-1-containing culture supernatants were prepared by transiently transfecting HeLa cells with pNL4-3 using LipofectAMINE LTX PLUS (Invitrogen Corp., Carlsbad, CA). Two ml of 20% sucrose solution was placed at the bottom of model SW55 centrifuge tubes and overlaid with 3 ml of HIV-1-containing culture supernatant described above. Samples were then centrifuged for 60 min at 35,000 rpm at 4˚C. Particulate HIV-1 were resuspended in PBS(-) containing a protease inhibitor cocktail (NACALAI TESQUE, INC, Kyoto, Japan). This suspension was loaded onto the top of a discontinuous sucrose density gradient composed of 1.0 ml 30% sucrose solution at the bottom of model SW55 centrifuge tubes covered by 1.0 ml 0.1% Triton X-100 in 10% sucrose solution and then centrifuged in a model SW55Ti rotor for 120 min at 35,000 rpm at 4˚C. Particulate CA protein was used for pull-down assays with Strep-tagged MELK or processed for immunoblotting using anti-p24 antibody (CA). The amount of particulate CA protein was quantified using HIV-1 CA (p24) enzyme-linked immunosorbent assay kits (ZeptMetrix Corporation, Buffalo, NY).

Negative staining electron microscopy
Envelope-stripped HIV-1 cores isolated by ultracentrifugation were absorbed onto Formvarcoated copper grids, and stained with 2% phosphotungstic acid solution. The images were recorded with a Tecnai F20 transmission electron microscope (FEI Company, Hillsboro, OR) at 200kV.

Affinity precipitation of HIV-1 cores with Strep-tag II fusion protein
HeLa cells were transfected with pCAG-OSF-MELK, pCAG-OSF-GFP, pCAG-OSF-CypA, or pCAG-FOS2-rhT5α, harvested 48 h post-transfection and lysed in a 7 ml-Dounce homogenizer. Cell extracts were incubated with Strep-Tactin Sepharose for 2 h at 4˚C. Purified Strep-tagged protein complexes were incubated with envelope-stripped HIV-1 cores (1,000 ng p24) for 2 h at 4˚C. After extensive washing, Strep-tagged protein complexes were released by boiling in SDS-PAGE loading buffer and the proteins were analyzed by 12% SDS-PAGE and Western blotting using mouse anti-FLAG antibody (FLAG) and mouse anti-p24 antibody (CA).

Preparation of recombinant proteins and synthetic peptides
E.coli BL21 CodonPlus-RIL cells (Agilent Inc. Santa Clara, CA) transformed with pGEX-4T-3 or pGEX-HIV-CA were used for purification of GST proteins using standard methods. Fifteen independent synthetic peptides covering HIV-1 CA were designed and provided by Sigma-Aldrich (Sigma-Aldrich Co, St. Louis, MO). The amount of CA protein was quantified using HIV-1 CA (p24) enzyme-linked immunosorbent assay kits (ZeptMetrix Corporation, Buffalo, NY).

In vitro phosphorylation assay
In vitro phosphorylation assays were performed with the ADP-Glo MELK kinase assay kit, following the manufacturer's instructions (Promega Corp, Madison, WI). Briefly, 100 ng of recombinant activated MELK was incubated with either GST, GST-HIV-CA or Env-stripped HIV-1 core proteins ranging from 100 to 2,000 ng or synthetic peptides, ZIPtide, ranging from 500 to 2,000 ng for 60 min at 30˚C in the presence of ultrapure ATP. Light emission was measured using the GloMAX multidetection system (Promega Corp, Madison, WI). ZIPtide was used as a phosphorylation standard for MELK. For the detection of phosphorylated CA protein in the multimerized viral core, 200 ng of recombinant activated MELK was incubated with envelope-stripped HIV-1 core containing 100 ng of p24 at 30˚C in the presence of ultrapure ATP. The phosphorylation reaction was terminated by the addition of 2 × sample buffer. The proteins were subsequently processed for immunoblotting using rabbit polyclonal antibodies to phospho-S149-CA, mouse monoclonal antibody to HIV-1 p24 (Abcam Inc, Cambridge, MA), mouse monoclonal antibody to HIV-1 p17 (Abcam Inc, Cambridge, MA), goat polyclonal antibody to gp120 (Abcam Inc, Cambridge, MA), or rabbit monoclonal antibody to MELK (Abcam Inc, Cambridge, MA).

Phos-tag assay
Briefly, 5 × 10 6 Non-T or MELK-KD MT4C5 cells were inoculated together with 2 × 10 7 RT counts of wild-type HIV-1 (HIV-1 NL4-3 strain) or 5 × 10 6 RT counts of VSV-G/NL4-3luc or VSV-G/NL4-3 CA-S149A viruses. After incubation at 4˚C for 30 min, cells were further incubated at 37˚C for 8 h. Cells were then washed twice with ice-cold PBS(-) containing 0.005% Trypsin/EDTA to remove virions from the cell surface and once with ice-cold PBS(-) to remove Trypsin/EDTA. Washed cells were resuspended in 1 ml of hypotonic lysis buffer [10 mM Tris-HCl (pH 8.0), 10 mM KCl, 1 mM EDTA, protease inhibitor cocktail (NACALAI TESQUE, INC, Kyoto, Japan) and phosphatase inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany)] and incubated on ice for 15 min. Swollen cells were lysed in a 7 ml-Dounce homogenizer with a 'tight' pestle (15 gentle strokes making a half-turn of the pestle at each stroke) and cell lysates cleared by centrifugation at 2,000 × g for 3 min at 4˚C. Proteins were separated in 10% precast SDS-polyacrylamide gels prepared with 50 μM acrylamide-pendant Phos-tag ligand (Wako Pure Chemical Industries, Ltd., Osaka, Japan) and were analyzed by immunoblotting with mouse anti-p24 antibody.

Statistical analysis
All data are obtained from at least three independent experiments. The average values are presented with error bars indicating the standard deviation (SD) and the statistical significance was analyzed using one-way analysis of variance (ANOVA) with Dunnett's or Tukey's multiple comparison tests, two-way analysis of variance (ANOVA) with Tukey's or Sidak's multiple comparison test, or Student's t-test. All the statistical analyses were performed using Prism 6 software (GraphPad Software, Inc). P values below 0.05 (P<0.05, Ã ; P<0.01, ÃÃ ; P<0.001, ÃÃÃ ) were considered significant. Unpaired two-tailed Student's t-test was used for the data shown in Fig 2B, 2C, 2H and 2J to test whether the means of the two groups were significantly different (five biological replicates). One-way analysis of variance (ANOVA) with Dunnett's multiple comparison test was used for data in Figs 1C, 2I, 3A, 6B, 6C, 7B, 7C, 7E and 7F to determine whether the means of multiple groups were significantly different from a single group (five biological replicates). One-way ANOVA with Tukey's multiple comparison test was used in Fig 4D to determine whether the means of four groups were significantly different from each other (five biological replicates). Two-way ANOVA with Tukey's multiple comparison test was used in Figs 3C, 3D and 5A to determine significant difference by comparing of the means specified by two factors (five biological replicates). Two-way ANOVA with Sidak's multiple comparison test was used in Fig 7A and 7D to determine significant difference by pairwise comparison of the means specified by two factors (five biological replicates).