Inositol pyrophosphate mediated pyrophosphorylation of AP3B1 regulates HIV-1 Gag release

Results

The β subunit of the adaptor protein complex AP-3 (AP3B1) (Fig. 1A) was initially identified as a potential target for IP₇ phosphorylation by database searching for proteins containing serine-rich acidic regions. Adaptor protein complexes, comprising AP-1 to AP-4, mediate the sorting of transmembrane and cargo proteins to specific membrane compartments within the cell (11, 12). The AP-3 complex in particular is involved in sorting to lysosomes and related organelles (13, 14). It contains two large subunits (β and δ), a medium subunit (μ₃), and a small subunit (σ₃) (12). The AP3B1 subunit contains three main domains: the N-terminal head domain, the hinge, and the C-terminal ear domain (Fig. 1A and Fig. S1A). The putative target region of IP₇ phosphorylation lies within the hinge domain, which contains three distinct serine-rich acidic stretches that we named regions I, II, and III (Fig. 1A and Fig. S1A). To test whether human AP3B1 is a substrate of IP₇-mediated phosphorylation, we performed an IP₇ phosphorylation assay on protein extracts from yeast expressing either the full-length human AP3B1 (GST-AP3B1), the N-terminus region lacking the acidic domains [GST-AP3B1 (1–676)] or the N-terminus containing region I [GST-AP3B1 (1–706)] (Fig. 1A). Both GST-AP3B1 and GST-AP3B1 (1–706) were phosphorylated by 5β[³²P]IP₇, whereas no phosphorylation of GST-AP3B1 (1–676) could be detected (Fig. 1B). These results demonstrate that AP3B1 is a bona fide target of IP₇-mediated phosphorylation. Moreover, the lower phosphorylation levels observed in cells expressing GST-AP3B1 (1–706) indicate that the regions II and III are also likely to be targets of IP₇ pyrophosphorylation (Fig. 1B). To test this hypothesis and to further map the IP₇ pyrophosphorylation target regions of AP3B1, we cloned each of the serine-rich acidic domains and performed the same experiment described above (Fig. S1B). With the exception of clone GST-AP3B1 (576–686) and clone GST-AP3B1 (576–691) that contain only a small part of acidic region I, all of the clones were pyrophosphorylated by IP₇. These results indicate that all three AP3B1 acidic regions are potentially pyrophosphorylated by IP₇.

To investigate whether the λ-phosphate of IP₇ is transferred to a prephosphorylated serine (10), we used Escherichia coli purified GST-AP3B1 (576–902). When recombinant GST-AP3B1 (576–902) was incubated with casein kinase two (CK2) in the presence of γ[³²P]ATP robust protein phosphorylation was observed, indicating that AP3B1 is a target for CK2 phosphorylation. To test whether CK2 phosphorylation is required for IP₇-dependent pyrophosphorylation of AP3B1, recombinant GST-AP3B1 (576–902) was incubated with either native or inactivated (boiled) CK2 before performing the IP₇ phosphorylation assay (Fig. 1C). IP₇ pyrophosphorylation of AP3B1 was only observed following pretreatment with native CK2 (Fig. 1C, lane 3). We further confirmed these results by treating AP3B1 with λ-phosphatase after CK2 treatment and before the incubation with 5β[³²P]IP₇. AP3B1 pyrophosphorylation was inhibited by the λ-phosphatase pretreatment (Fig. 1C, lane 5). Taken together, these results demonstrate that CK2-dependent phosphorylation of the serine-rich acidic stretches of AP3B1 is necessary for priming IP₇-mediated pyrophosphorylation.

To demonstrate that IP₇-dependent protein pyrophosphorylation occurs in vivo, we tested whether GST-AP3B1 expressed in yeast strains containing different levels of endogenous IP₇ show different degree of in vitro 5β[³²P]IP₇-driven pyrophosphorylation. We used an IP₆-kinase yeast mutant that does not contain IP₇ (kcs1Δ) (15, 16) (Fig. S2) and an IP₇-kinase yeast mutant that accumulates large amounts of IP₇ (vip1Δ) (17) (Fig. S2). We reasoned that if AP3B1 is endogenously pyrophosphorylated, it would be a weak target for in vitro 5β[³²P]IP₇-driven pyrophosphorylation when expressed in vip1Δ, due to the high level of endogenous IP₇. Whereas AP3B1 expressed in kcs1Δ by not being naturally pyrophosphorylated due to the lack of IP₇ would be a strong target for in vitro 5β[³²P]IP₇-driven pyrophosphorylation. Protein extracts from either WT, kcs1Δ, or vip1Δ yeast strains expressing GST-AP3B1 were tested for 5β[³²P]IP₇ pyrophosphorylation in vitro (Fig. 1D). As predicted, GST-AP3B1 pyrophosphorylation was lower in vip1Δ cells than in WT yeast (Fig. 1D). Importantly, exogenous pyrophosphorylation of GST-AP3B1 was substantially higher in kcs1Δ than in WT (Fig. 1D). To demonstrate that endogenous mammalian AP3B1 can be pyrophosphorylated by IP₇, AP3B1 was immunoprecipitated from mouse embryonic fibroblast (MEF) and subjected to IP₇ pyrophosphorylation assay (Fig. 1E). A robust phosphorylation signal was observed from WT MEF (+/+), whereas phosphorylation was undetected in MEF cells from mocha mice (mh/mh) that do not contain a functional AP-3 complex and do not possess detectable AP3B1 subunit (18, 19) (Fig. 1E).

As protein phosphorylation often results in a gel mobility shift that can be detected by Western blot analysis, we asked whether IP₇-mediated protein pyrophosphorylation could also cause a mobility shift (Fig. 1F). When cell extracts of yeast expressing full-length GST-AP3B1 were rapidly processed, a dramatic gel mobility shift, which directly correlated with the endogenous IP₇ levels, was detected (Fig. 1F). A similar gel mobility shift was observed for GST-NSR1 and GST-Nucleolin, two known targets of IP₇-mediated pyrophosphorylation (9) (Fig. S3 A and B), whereas the gel migration of GST-Kif3A and HA-Ankyrin, proteins that are not IP₇ targets, were not affected (Fig. S3C). The gel mobility shift is suggestive of high stoichiometry of phosphorylation in vivo, however the inability to visualize gel shift using 5β[³²P]IP₇ pyrophosphorylation (Fig. 1 B–E) is indicative of lower stoichiometry in vitro. Various scenarios can be envisaged to explain this stoichiometry difference between the in vitro and in vivo experiments: in vivo IP₇-mediated pyrophosphorylation could stimulate additional CK2 phosphorylation or additional IP₇ pyrophosphorylation. We can also hypothesize the existence of uncharacterized cellular components regulating protein pyrophosphorylation in vivo. Alternatively, protein pyrophosphorylation might represent a step for a more complex protein modification.

To determine the physiological consequence of IP₇-mediated pyrophosphorylation of AP3B1, we investigated whether this modification could regulate protein–protein interaction. To identify interacting partners of AP3B1, we performed a yeast two-hybrid screen on a human fetal cDNA library using the target region of IP₇ phosphorylation, AP3B1 (576–902), as bait (Fig. 1A and Fig. S1). The identified binding partners were screened for the ability of the interaction to be modulated by the cellular presence of IP₇ (see below), and we choose to further analyze the interactor Kif3A, a motor protein that belongs to the kinesin superfamily (Fig. 2A and Fig. S4). Kif3A is a 70 Kda protein, and the region identified in the screen corresponded to the C-terminus that includes amino acids 601–702 (Fig. 2A and Fig. S4). GST pull down in mammalian cells of GST-AP3B1 or GST-AP3B1 (577–1094), with either Myc-Kif3A, Myc-Kif3A (1–342), or Myc-Kif3A (354–702), demonstrated that AP3B1 specifically interacts with the non-motor domain of Kif3A (Fig. 2B). The interaction between the AP3 complex and Kif3A was also confirmed by coimmunoprecipitation of endogenous AP3B1 and AP3D1 proteins in WT MEF, whereas no interaction was detected in the control mocha MEF cells (Fig. 2C).

To evaluate whether IP₇-dependent pyrophosphorylation of AP3B1 regulates AP3B1 and Kif3A interaction, we performed experiments in both yeast and HeLa cells. Homogenates of WT and kcs1Δ yeast expressing GST-AP3B1 were incubated for 2 h with homogenates of kcs1Δ yeast expressing HA-Kif3A (601–702) and subjected to GST pull down (Fig. 3A). In kcs1Δ yeast, AP3B1 is not pyrophosphorylated due to the lack of IP₇, and its interaction with Kif3A (601–702) was on average 12 times stronger when compared to AP3B1 expressed in WT yeast (Fig. 3A). To test whether this was a general effect associated with the kcs1Δ yeast strain, we performed similar experiments with ankyrin, another AP3B1-binding protein identified in the YTH screen; however no difference in binding was observed (Fig. S5). We next investigated whether a similar regulation occurs in HeLa cells by increasing the intracellular levels of IP₇ by ≈6-fold through expression of the IP₆-kinases IP₆K1 or IP₆K2 (Fig. S6). When either IP₆K1 or IP₆K2 were coexpressed with GST-AP3B1 and Myc-Kif3A, we detected on average a 4-fold decrease in AP3B1-Kif3A interaction (Fig. 3B). This effect was dependent on the levels of IP₇, as overexpression of a kinase-dead IP₆K1K/A or IP₆K2K/A had no effect on AP3B1-Kif3A interaction (Fig. 3C). We next investigated whether the reduction in the interaction between AP3B1 and Kif3A was the sole consequence of IP₇ pyrophosphorylation by performing in vitro reconstitution experiments using recombinant bacterially expressed GST-AP3B1 (576–902) and His-Kif3A (Fig. 3D). GST-AP3B1 (576–902) phosphorylated by CK2 is able to interact with His-Kif3A, however the subsequent pyrophosphorylation by IP₇ substantially weakens the binding ability (Fig. 3D). These results demonstrate that the interaction between AP3B1 and Kif3A is direct, and together with the previous results, that this interaction is negatively regulated by AP3B1 pyrophosphorylation.

We next investigated how IP₇-mediated pyrophosphorylation of AP3B1 affects AP3B1 and Kif3A function. Assembly of HIV-1 is directed by the Gag protein, the major structural protein of the virus (20). The δ subunit (AP3D1) of the AP-3 complex has been shown to bind to HIV-1 Gag and to be involved in its intracellular trafficking (21, 22). This functional role of AP3 complex led us to investigate whether IP₇-mediated pyrophosphorylation could affect HIV release through modulation of AP3B1-Kif3A interaction. We first tested whether Kif3A, like the AP-3 complex (21, 22), is involved in HIV-1 Gag release. HIV-1 Gag alone contains all of the determinants required to produce noninfectious virus-like particles (VLP) in the absence of other viral proteins (23, 24). To establish whether Kif3A plays a role in HIV-1 VLP release, we used small interfering RNA (siRNA), achieving ≈75% of Kif3A depletion 96 h after transfection as evaluated by anti-Kif3A immunoblotting (Fig. 4A). Kif3A silenced cells were transfected with Gag-GFP, and a dramatic reduction of VLP release was observed when compared with control siRNA cells transfected with Gag-GFP in HeLa cells (Fig. 4A). Importantly, the intracellular levels of Gag-GFP remained unaltered, thereby excluding the possibility that Kif3A depletion affects Gag-GFP stability (Fig. 4A). We confirmed that reduction of VLP release was due to Kif3A depletion and not an off-target effect of the siRNA by complementing Kif3A-silenced HeLa cells with mouse Kif3A (Fig. 4B). We further confirmed that AP3B1 and Kif3A interaction is involved in HIV-1 Gag release, by expressing the respective AP3B1 and Kif3A binding domains. We reasoned that if these domains behaved as dominant negative forms, the VLP release should be affected. Cotransfection of Gag-GFP with vectors encoding motorless Kif3A [Myc-Kif3A (354–702)], which binds to AP3B1 (Fig. 2B) but is unable to hydrolyze ATP or to bind to microtubules, substantially reduces VLP release (Fig. 4C). Likewise, expression of Myc-AP3B1 (576–902) that contains the Kif3A-interacting domain, with Gag-GFP, consistently reduced VLP release (Fig. 4C). These results indicate that Kif3A and the AP-3 complex are involved in HIV-1 Gag VLP release from HeLa cells.

We next investigated whether changes of IP₇ intracellular levels affect HIV-1 Gag release through modulation of AP3B1 and Kif3A interaction. The effect of increased IP₇ intracellular content was tested in HeLa cells expressing either IP₆K1 or IP₆K2 (Fig. S6) together with Gag-GFP. High levels of IP₇ consistently reduced VLP release by 30% (Fig. 5A). This reduction is noteworthy, considering that we are altering a dynamic enzymatic pathway and that the turnover of IP₇ is already high under basal conditions (1). Importantly, no reduction of VLP release was observed upon overexpression of kinase dead versions of either IP₆K1K/A or IP₆K2K/A (Fig. 5B). Most mammalian cells express three IP₆-kinases, and HeLa cells in particular express both IP₆K1 and IP₆K2. We were unable to obtain a reduction of the enzymatic activity that substantially reduced cellular IP₇ through RNA interference (RNAi), and instead we used MEF cells derived from IP₆K1 null mice that, although still possessing IP₆K2, contain reduced levels of IP₇ (7). IP₆K1 null MEF were transfected with Gag-GFP and empty GST vector. The transfection efficiency of IP₆K1 null MEF cells is substantially lower than WT MEF cells as seen by the lower ectopic expression of GST in the mutant cells, however the VLP release was substantially increased in these cells (Fig. 5C). Quantification of the ratio between intracellular and released VLP, revealed a ≈35% (±5%) increase of extracellular VLP from mutant cells when compared with WT MEF cells (Fig. 5C). Together, these findings demonstrate that changes in the cellular levels of IP₇ alter VLP release.

Discussion

In this study, we identified Kif3A as an AP3B1 interacting protein and established that this motor protein is involved in HIV-1 VLP release from HeLa cells. Evidence for the involvement of motor proteins in providing an efficient way for viruses to egress from host cell has been mounting. Kif3C, a motor protein that heterodimerizes with Kif3A, was recently identified in a whole genome RNAi screen as a late-acting protein required for viral assembly and release in HeLa cells (25). Moreover, Kif4 was shown to associate with multiple retrovirus Gag proteins and to function in Gag intracellular trafficking (26).

More importantly, we have identified AP3B1 as a target of IP₇-mediated pyrophosphorylation and used its recently established function in HIV-1 release (21, 22) as a model to address the role of this mechanism of protein posttranslational modification. Through back phosphorylation experiments, we demonstrate that AP3B1 can be pyrophosphorylated by IP₇ and that, in vivo, this modification affects the electrophoretic mobility of several target proteins. The observation that IP₇-mediated pyrophosphorylation of proteins requires conventional ATP-dependent phosphorylation as a priming event raises the possibility that cellular functions that have been previously attributed to ATP-dependent phosphorylation may also involve pyrophosphorylation. For instance, in neurons, the AP-3 complex regulates endosomal synaptic vesicle biogenesis (27, 28), and AP3B1 phosphorylation was shown to inhibit vesicular coating and to impair synaptic vesicle formation (27). It is possible that this phenomenon depends on decreased IP₇-mediated pyrophosphorylation. We have shown that IP₇-mediated pyrophosphorylation of a target protein, AP3B1, modified its interaction with a specific binding partner, Kif3A. This modification has a direct effect on one of the biological functions of AP3B1, HIV-1 Gag release. However, modification of the interaction between two interacting partners is unlikely to be the only mechanism by which IP₇-mediated pyrophosphorylation regulates the function of target proteins.

Methods

Unless otherwise stated, all reagents were from commercial sources. Please refer to SI Experimental Procedures for detailed reagent origins, cloning procedures, and routine methods.

Supplementary Material