PIKfyve-ArPIKfyve-Sac3 Core Complex

cDNA Constructs

The pCMV5- or pEGFP-based deletion and point mutants of mouse PIKfyve_S tagged with Myc or HA epitopes, including PIKfyve^WT, PIKfyve^K1831E, PIKfyve^ΔFYVE, and PIKfyve^ΔCpn+, were generated previously (16, 18, 20). The new mouse PIKfyve_S constructs are: PIKfyve^K1831EΔCpn+, with a deletion of the entire Cpn60_TCP1 domain and a portion upstream of CH homology (deleted residues 560–1231) and harboring the kinase-deficient point mutation at Lys¹⁸³¹; PIKfyve^ΔN-Cpn and PIKfyve^ΔC-Cpn, with deleted N- (deleted residues 560–749) and C-terminal halves (deleted residues 807–1032) in the Cpn60_TCP1 domain, respectively; PIKfyve^ΔFΔDΔCpn, with deleted FYVE, DEP, and Cpn60_TCP1 domains (deleted residues 1–929); PIKfyve^ΔFΔKin+, with deleted FYVE finger, kinase domain, and additional sequences upstream of the latter (deleted residues 1–199 and 1439–2052); PIKfyve^ΔSpecΔKin+, with a deleted C terminus encompassing the CH homology domain, spectrin repeats, and further downstream sequences including the kinase domain (deleted residues 1262–2052). Generation of these constructs is detailed in supplemental Material 1. The pCMV5-based cDNA constructs of human ArPIKfyve^WT tagged with Myc or HA epitopes were described previously (21). Generation of the pEGFP-based constructs of human ArPIKfyve (pEGFP-C2-HA-ArPIKfyve^WT, pEGFP-C2-ArPIKfyve-(298- 782), pEGFP-C2-HA-hArPIKfyve-(1–511), and pEGFP-C1-hArPIKfyve-(523–782)) was detailed elsewhere (14). The human Sac3 constructs, pEGFP-C3-Sac3^WT, pEGFP-C3-Sac3^D488A, pEF-BOS-Myc-Sac3, and pCMV5-HA-Sac3^WT, were described elsewhere (9). The new hSac3 point and deletion mutants are: pCMV5-HA-hSac3^D488A, pEGFP-C3-Sac3-(1–315), pEGFP-C3-Sac3-(1–574), pEGFP-C3-Sac3-(388–907), pEGFP-C3-Sac3-(478–907), and pEGFP-C3-Sac3-(610–907). Their construction is detailed in supplemental Material 2. The new constructs were confirmed by restriction mapping and immunoblotting in transfected cells.

Antibodies

Rabbit polyclonal antibodies against PIKfyve (R-7069), ArPIKfyve (WS047), and Sac3 proteins were characterized previously (9, 21, 22). They were used as protein A (R-7069) and affinity-purified forms (WS047) or as a crude antiserum (anti-Sac3). Polyclonal anti-HA (R4289, a gift by Dr. Mike Czech) was used as Protein A-purified IgG. Anti-Myc monoclonal antibody was from 9E10.2 hybridoma cells (ATCC). Anti-GFP polyclonal (Ab290) and goat anti-EEA1 (N-19) were from AbCam and Santa Cruz Biotechnology, respectively.

Cell Cultures and Transfections

COS7 cells were cultured and transiently transfected with the indicated cDNAs by Lipofectamine 2000 or Lipofectamine (Invitrogen) for immunoprecipitation and immunofluorescence microscopy analyses, respectively, under conditions described in previous studies (14, 18).

Immunofluorescence and Light Microscopy

Transfected COS7 cells grown on coverslips were subjected to immunofluorescence microscopy analysis 12 and/or 24 h post-transfection as described previously (8, 18). Permeabilized cells were stained with the primary and secondary antibodies indicated in the figure legends. Coverslips were mounted on slides using the Slow Fade Antifade kit (Invitrogen). Cells were viewed in a Nikon Eclipse TE200 inverted microscope equipped with a Plan Apo 60 × 1.4 Ph3DM oil objective and 3 standard fluorescence channels (i.e. green for GFP, red for anti-Myc/Alexa568, and blue for anti-Sac3/Alexa350 signals). Images were captured with a SPOT RT slider charge-coupled device camera (Diagnostic Instruments, Sterling Heights, MI) and processed using SPOT 3.2 and Adobe Photoshop 6.0 software. Colocalization with EEA1 was investigated by confocal microscopy (model 1X81, Olympus) by a 60× UplanApo lens. Images were captured by a cooled charge-coupled device 12-bit camera (Hamamatsu) as described previously (8, 9).

Baculoviral Vectors, Recombinant Protein Expression/Purification, and in Vitro Binding

Generations of GST-His₆-mPIKfyve in baculoviral expression vector pAcGHLT-A and that of His₆-hArPIKfyve in bacterial expression vector pRSETb were described previously (16, 21). The baculoviral expression vector pFASTBac-GST-hSac3 was a kind gift by Dr. Takenawa. His₆-ArPIKfyve and Sac3 were simultaneously expressed using a modular baculovirus-based system specifically designed for eukaryotic multiprotein expression (23) as detailed in supplemental Material 3. For the in vitro binding assays, GST-PIKfyve or His₆-PIKfyve (∼15 ng protein) purified and immobilized on GSH- or Ni-NTA-agarose as described elsewhere (16) was incubated for 2–16 h with His₆-ArPIKfyve or GST-Sac3, respectively, produced and purified separately. Immobilized GST-PIKfyve was incubated for 2 h with the His₆-ArPIKfyve-Sac3 subcomplex, purified from infected Sf21 cells on a Ni-NTA column. Beads were washed following previously published protocols (24). The reactions were analyzed by SDS-PAGE and immunoblotting.

Immunoprecipitation and Immunoblotting

Fresh cell lysates, collected in RIPA⁺ buffer (50 mm Tris/HCl buffer, pH 8.0, containing 150 mm NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate and 1× protease inhibitor mixture (1 mm phenylmethylsulfonyl fluoride, 5 μg/ml leupeptin, 5 μg/ml aprotinin, 1 μg/ml pepstatin, and 1 mm benzamidine)) were precleared (20,000 × g, 15 min, 4 °C) and subjected to immunoprecipitation 24 h post-transfection. Control immunoprecipitates with nonimmune rabbit or mouse IgG were run in parallel. Immunoprecipitations were carried out for 16 h at 4 °C with protein A-Sepharose CL-4B added in the final 1.5 h of incubation. Immunoprecipitates were washed five times with RIPA⁺ buffer and then processed by Western blotting. Immunoblotting with the antibodies indicated in the figure legends was performed subsequent to protein separation by SDS-PAGE (typically on 6% gels) and electrotransfer onto nitrocellulose membranes as described previously (14, 21). A chemiluminescence kit (Pierce) was used to detect the horseradish peroxidase-bound secondary antibodies.

Lipid and Protein Kinase Activity in Vitro Assays

PIKfyve lipid and protein kinase activities were measured in parallel using PIKfyve immunoprecipitates derived from transfected cells. For the lipid kinase assay, washed PIKfyve immunoprecipitates were subjected to 15 min of incubation at 37 °C in an assay buffer (50 mm Tris, pH 7.4, 2.5 mm MgCl₂, and 2.5 mm MnCl₂) supplemented with 50 μm ATP, [γ-³²P]ATP (12.5 μCi), and 100 μm PtdIns (from soybean, Avanti Polar Lipids Inc) as detailed elsewhere (14, 22). Lipids were extracted and analyzed by silica gel TLC using an acidic solvent system (2 m acetic acid/propanol, 1:2 v/v). Both PtdIns(5)P- and PtdIns(3,5)P₂-synthesized products were monitored simultaneously as we detailed elsewhere (21, 22). PIKfyve protein kinase was measured by the PIKfyve autophosphorylation activity in a reaction mix composed of [γ-³²P]ATP (5 μCi), 25 μm ATP, 24 mm MgCl₂, and 5 mm MnCl₂ in 50 mm Hepes, pH 7.4, and carried out for 30 min at 25 °C as described elsewhere (12, 16). Generated lipid or protein products were detected by autoradiography and quantified by radioactive counting of the scraped silica spots or cut radioactive band.

Other Methods

Protein concentration was determined by the bicinchoninic acid protein assay kit (Pierce). Protein levels were quantified from the intensity of the immunoblot bands with a laser scanner (Microteck) and UN-SCAN-IT software (Silk Scientific). Several films of different exposure times were quantified to assure the signals were within the linear range. Data are expressed as mean ± S.E. Statistical analysis was performed by either one- or two-tailed Student's t test.

ArPIKfyve and Sac3 Associate with Each Other via Their C Termini Independently of the Sac3 Phosphatase Activity

As established previously, PIKfyve interacts efficiently with ArPIKfyve and Sac3 only when the two proteins are together, whereas ArPIKfyve and Sac3 form an easily detectable heteromeric subcomplex without PIKfyve (14). That ArPIKfyve and Sac3 interact directly with each other in a manner independent of PIKfyve or other proteins was also confirmed herein in Sf21 insect cells infected with baculoviruses expressing His₆-ArPIKfyve and Sac3 simultaneously. Purification of His₆-ArPIKfyve on Ni-NTA-agarose resin reproducibly copurified Sac3 (supplemental Fig. 1). Therefore, to begin characterizing the contact sites in the PAS core complex, we first examined the molecular interaction of the ArPIKfyve-Sac3 subcomplex. We generated two overlapping GFP-based constructs of the Sac3 N- and C-terminal halves (Fig. 1A) and examined their efficiency to interact with Myc-ArPIKfyve^WT by coimmunoprecipitation analyses in double-transfected COS cells. As illustrated in Fig. 1B, whereas the N-terminal half of Sac3 (residues 1–574) coimmunoprecipitated with Myc-ArPIKfyve^WT only weakly, the C-terminal Sac3 fragment (residues 388–907) displayed the wild-type binding efficiency. With a goal to determine the minimal region that supports the interaction with ArPIKfyve, we next assessed shorter C-terminal fragments of Sac3 (Fig. 1, A and B). Sac3-(478–907) associated with an efficiency similar to that of Sac3-(388–907) or Sac3^WT, whereas the Sac3-(610–907) peptide fragment displayed 85 ± 8% of the maximal binding (Fig. 1, A and B). Sac3^D488A that harbors a point mutation in the phosphatase domain to abrogate the phosphatase activity (9) displayed the Sac3^WT association efficiency (Fig. 1, A and B). These data indicate that Sac3 interacts with ArPIKfyve^WT independently of its phosphatase activity, and a C-terminal fragment of ∼430 amino acids is required to mimic the efficiency of the Sac3^WT association with ArPIKfyve.

To determine the ArPIKfyve region interacting with Sac3, we examined two nearly overlapping GFP-based constructs of the ArPIKfyve N- and C-terminal region for their interaction with Myc-Sac3^WT by coimmunoprecipitation in double-transfected COS cells (Fig. 1C). Notably, only the fragment of the ArPIKfyve C-terminal region spanning residues 523–782 was recovered with Myc-Sac3^WT (Fig. 1D). The association of the N-terminal region was <10% of that seen with ArPIKfyve^WT binding to Sac3^WT (Fig. 1D). These data indicate that the C-terminal region of ArPIKfyve spanning residues 523–782 interacts with Sac3.

Sac3 Interacts with an ArPIKfyve Homodimer

As revealed recently, ArPIKfyve forms a homodimer (or a higher-order homooligomer) that scaffolds the PAS core (14). Intriguingly, the same C-terminal fragment that interacts with Sac3 is engaged in the ArPIKfyve homomerization (14). To better understand how the ArPIKfyve-Sac3 subcomplex is organized, we examined if ArPIKfyve could concurrently interact with Sac3 and with itself. To test this we expressed simultaneously two distinctly tagged versions of ArPIKfyve^WT together with Myc-Sac3^WT. We used Myc-ArPIKfyve^WT and GFP-HA-ArPIKfyve^WT because they exhibit markedly different electrophoretic mobilities, making their unambiguous detections possible. Importantly, both Myc-Sac3^WT and Myc-ArPIKfyve^WT were recovered in the HA-ArPIKfyve^WT immunoprecipitates (Fig. 1E). The ArPIKfyve^Nt fragment neither homodimerized nor associated with Sac3^WT to a significant extent, as evidenced by the lack of Sac3^WT or ArPIKfyve^Nt immunoreactive bands coimmunoprecipitated with anti-HA ArPIKfyve^Nt (Fig. 1E). Together, these data indicate that ArPIKfyve-ArPIKfyve and ArPIKfyve-Sac3 interactions proceed simultaneously through the same region of ArPIKfyve spanning residues 523–782. Intriguingly, this region, identified by data base searches as “Domain with unknown function,” incorporates four of the predicted five coiled-coil motifs in ArPIKfyve (Fig. 1C), conducive in forming coiled-coil helical backbone for structural stability of multiprotein interactions (25). Because Sac3 does not form homooligomeric structures (14), seen also with the yeast counterpart (26), these data are consistent with the notion that the subcomplex incorporates ArPIKfyve and Sac3 at a molar ratio of 2 (or higher) to 1.

ArPIKfyve and Sac3 N Termini Interact with PIKfyve

We next sought to determine the regions in ArPIKfyve-Sac3 interacting with PIKfyve. As found previously in ectopically transfected COS cells, PIKfyve efficiently interacts with ArPIKfyve and Sac3 only when the two are together (14). This conclusion was corroborated herein by the data from in vitro reconstituted binding of the three recombinant proteins, expressed and purified from insect or bacterial cells. Thus, GST-PIKfyve immobilized on GSH-agarose did not pull down His₆-ArPIKfyve to a significant degree (Fig. 2A). Likewise, His₆-PIKfyve immobilized on Ni-NTA-agarose did not pull down GST-Sac3 (not shown). However, if the His₆-ArPIKfyve-Sac3 subcomplex, purified from infected Sf21 on Ni-NTA-agarose, was incubated with purified GST-PIKfyve immobilized on GSH beads, both His₆-ArPIKfyve and Sac3 were readily pulled down (Fig. 2A). Therefore, to determine how PIKfyve associates with the ArPIKfyve-Sac3 subcomplex, we triple-transfected COS cells with constructs of appropriately tagged PIKfyve^WT, ArPIKfyve^WT, and C-terminal fragments of Sac3 or PIKfyve^WT, Sac3^WT, and C-terminal fragments of ArPIKfyve. Coimmunoprecipitation was carried out with antibodies against the tags in ArPIKfyve^WT or Sac3^WT. Levels of retrieved PIKfyve^WT were quantified relatively to those found with the ArPIKfyve- Sac3 wild type. As illustrated in Fig. 2B, Sac3-(610–907) and ArPIKfyve^WT poorly associated with PIKfyve^WT as judged by the low HA-PIKfyve^WT amounts coimmunoprecipitated with anti-Myc-ArPIKfyve^WT. Likewise, a longer Sac3 fragment spanning residues 388–907 did not bring about greater levels of HA-PIKfyve^WT retrieved in Myc-ArPIKfyve^WT immunoprecipitates (Fig. 2, B and C). As expected, however, complexes between ArPIKfyve^WT and the Sac3 C-terminal fragments were readily formed at the efficiency similar to that seen with the Sac3 wild type (Fig. 2, B and C). This observation indicates that even though a complex between ArPIKfyve and C-terminal Sac3 is formed, PIKfyve is not efficiently retrieved, consistent with the requirement of the Sac3 N-terminal region (1–388) in the PAS core formation. Likewise, ArPIKfyve-Sac3 complexes made of Myc-Sac3^WT and GFP-ArPIKfyve-(523–782) or GFP-ArPIKfyve-(288–782) did not efficiently retrieve HA-PIKfyve^WT (quantified in Fig. 2C), indicating the ArPIKfyve N-terminal 1–288 residues are required for an efficient PIKfyve incorporation within the PAS complex. These observations further indicate that the PAS complex could not be formed in the absence of ArPIKfyve or Sac3 N termini even though a stable subcomplex via their C termini can still be formed.

PIKfyve Cpn60_TCP1 and CHK Homology Are Essential in Interacting with the ArPIKfyve-Sac3 Subcomplex

To examine how PIKfyve associates with the ArPIKfyve-Sac3 subcomplex in the PAS core, we generated several GFP-HA or HA-based truncated mutants of PIKfyve (Fig. 3A). Their efficiency to interact with the wild-type ArPIKfyve-Sac3 subcomplex was examined relative to that of PIKfyve^WT by coimmunoprecipitation in triple-transfected COS cells. We began characterizing the binding efficiency of a truncated mutant with deleted Cpn60_TCP1 domain and N-terminal parts of the CHK homology (deleted residues 560–1231; Fig. 3A). The CHK homology (residues 1151–1461), shared by all PIKfyve orthologs, harbors conserved Cys (6 versus Fab1) or His (4 versus Fab1), positioned at the N-terminal half (residues 1155–1327), and Lys (11 versus Fab1), positioned at the C-terminal half of the homology (1295–1461). Sequence analyses by modular databases identify the region of conserved Lys as spectrin repeats (residues 1367–1464; Fig. 3A), a module that supports assembly in multiprotein complexes involved in cytoskeletal architecture and signal transduction (27). Therefore, we herein refer to the N-terminal and C-terminal part of the CHK homology as CH and spectrin repeats, respectively (Fig. 3A). Intriguingly, for reasons awaiting clarification, the Δ560–1231 mutant (herein as ΔCpn+) was known for quite some time to be devoid of lipid kinase activity (22) despite the intact lipid kinase domain (Fig. 3A). This peculiar observation points to the Cpn60_TCP1 and/or CH homology domains as a plausible candidate region for binding the ArPIKfyve-Sac3 subcomplex, thereby activating PIKfyve kinase activity. This suggestion was corroborated herein by coimmunoprecipitation analyses, shown in Fig. 3B, in which we detected only insignificant amounts of Myc-ArPIKfyve^WT and Myc-Sac3^WT (Fig. 3A) in HA-PIKfyve^ΔCpn+ immunoprecipitates. Mutants with shorter truncations in the Cpn60_TCP1 domain, namely Δ560–749 (herein as ΔN-Cpn) and Δ807–1032 (herein as ΔC-Cpn), also showed a markedly reduced ability of retrieving Myc-ArPIKfyve^WT and Myc-Sac3^WT (Fig. 3, A and C), indicative for the presence of binding determinants on both the N- and C-terminal halves of the Cpn60_TCP1 domain. In contrast, the FYVE-finger deletion mutant interacted with ArPIKfyve-Sac3 at the wild-type efficiency, as judged by the similar levels of Myc-ArPIKfyve^WT and Myc-Sac3^WT immunoreactive bands recovered in immunoprecipitates of HA-PIKfyve^ΔFYVE versus HA-PIKfyve^WT (Fig. 3, A and B).

To confirm the critical role of the Cpn60_TCP1/CH homology domains in binding the ArPIKfyve-Sac3 subcomplex, we sought to reconstitute the triple association in ectopically transfected COS cells. We assessed the efficiency at which ArPIKfyve^WT and Sac3^WT coimmunoprecipitated with a PIKfyve fragment spanning residues 1–1260. This mutant, referred to herein as PIKfyve^ΔSpecΔKin+, harbors the region of Cpn60_TCP1 and CH deleted in the ΔCpn+ mutant, but the C-terminal part downstream of the CH homology encompassing the spectrin repeats and the catalytic domain is eliminated (Fig. 3A). Surprisingly, the PIKfyve^ΔSpecΔKin+ mutant was incapable of reproducing the efficiency of the PAS wild-type associations as evidenced by only 25% potency of PIKfyve^ΔSpecΔKin+ versus PIKfyve^WT immunoprecipitates to retrieve ArPIKfyve^WT and Sac3^WT (Fig. 3, A and D). Likewise, a larger fragment that incorporates the entire CH/spectrin repeats homology along with the Cpn60_TCP1 domain, referred to herein as ΔFΔKin+, was unable to coimmunoprecipitate Myc-ArPIKfyve^WT and Myc-Sac3^WT to an extent seen with PIKfyve^WT (Fig. 3, A and E). These data indicate that the region encompassing Cpn60_TCP1 and upstream of CH is a necessary binding determinant, yet alone it is insufficient to associate with the ArPIKfyve-Sac3 subcomplex as efficiently as PIKfyve wild type.

The PIKfyve Kinase Domain, but Not Activity, Is Required for Interaction with the ArPIKfyve-Sac3 Subcomplex

The observation that a peptide fragment of the N-terminal and middle region of PIKfyve incorporating the Cpn60_TCP1 domain and upstream of the CH homology is insufficient to mimic the PIKfyve^WT association with ArPIKfyve-Sac3 suggests the C-terminal region downstream of the CH homology also plays a role in the PAS core formation and stabilization. To test this we examined the interaction of a truncated mutant harboring nearly the entire PIKfyve sequence minus the C-terminal kinase domain (Fig. 3A). As illustrated in Fig. 3, A and F, there was significantly less Myc-ArPIKfyve^WT and Myc-Sac3^WT retrieved in immunoprecipitates of PIKfyve^ΔKin versus PIKfyve^WT, suggesting that interaction sites for ArPIKfyve-Sac3 are present within the PIKfyve kinase domain. Noteworthy, although the catalytic domain is apparently important for efficient PAS core formation, the PIKfyve kinase activity itself is not. This was evidenced herein by the observed wild-type efficiency of the lipid/protein kinase-deficient HA-PIKfyve^K1831E mutant to coimmunoprecipitate Myc-ArPIKfyve^WT and Myc-Sac3^WT (Fig. 3, A and F). Consistent with the requirement for the catalytic domain in productive ternary complex formation, a PIKfyve C-terminal fragment spanning residues 930–2052 (referred to as ΔFΔDΔCpn), interacted to some extent with ArPIKfyve-Sac3 as judged by the detection of Myc-ArPIKfyve^WT and Myc-Sac3^WT in HA-PIKfyve^ΔFΔDΔCpn immunoprecipitates (Fig. 3, A and G). It should be emphasized, however, that either of the two overlapping PIKfyve halves (residues 1–1260 and 930–2052) produced only ∼¼ of the PIKfyve wild-type association with ArPIKfyve-Sac3 (Fig. 3A), suggesting ArPIKfyve/Sac3-induced alterations in the PIKfyve conformation to optimally accommodate the ArPIKfyve-Sac3 subcomplex. Together, these data are consistent with the notion that ArPIKfyve-Sac3 first docks at the Cpn60_TCP1/CH region, which induces conformational changes in PIKfyve to uncover additional binding sites stabilizing the ternary associations of the PAS core.

No Phenotypic Defects by PIKfyve^K1831E If Its ArPIKfyve-Sac3 Binding Region Is Eliminated

Having identified the Cpn60_TCP1/CH region as a major determinant in associating with ArPIKfyve-Sac3, we next assessed the outcome of the lost ArPIKfyve-Sac3 binding to the genuine PIKfyve characteristics, such as lipid or protein kinase activities, intracellular localization, and the effect on endomembrane homeostasis. Intriguingly, the PIKfyve^ΔCpn+ mutant was devoid of measurable in vitro lipid kinase activity as observed previously (22) and confirmed herein under longer exposure times (4 days) of the autoradiograms (Fig. 4A, quantified in Fig. 3A). However, PIKfyve^ΔCpn+ displayed in vitro protein kinase activity comprising 20–25% that determined for PIKfyve^WT (Fig. 4A), unlike the PIKfyve^K1831E mutant that lacked any lipid or protein kinase activities (16). These data suggest that associated ArPIKfyve-Sac3 is essential for PtdIns(3,5)P₂ and PtdIns(5)P synthesis but less critical for the protein kinase activity of PIKfyve. This conclusion is also substantiated by data demonstrating that if PAS-associated levels of endogenous ArPIKfyve are reduced by small interfering RNA-mediated knockdown or detergent stripping, the PIKfyve lipid kinase activity is markedly inhibited, whereas the autokinase activity remains unaltered (12, 21, and data not shown).

Given this lack of activity, one would assume that, like PIKfyve^K1831E and other PtdIns(3,5)P₂-deficient mutants such as PIKfyve^K1999E or PIKfyve^{K1999E/K2000E} (17, 18), expressed PIKfyve^ΔCpn+ will induce the typical endomembrane vacuolation defects. Oddly, however, defective endomembrane morphology was not seen in COS cells due to PIKfyve^ΔCpn+ (18). Considering the inability of PIKfyve^ΔCpn+ to impair endomembrane homeostasis resulted from residual PtdIns(3,5)P₂ synthesis, here we characterized the behavior of a PIKfyve^K1831E mutant with the truncated Cpn+ region (residues 560–1231). Like PIKfyve^ΔCpn+ (Fig. 3B), the PIKfyve^K1831EΔCpn+ mutant did not associate with ArPIKfyve-Sac3 (not shown but quantified in Fig. 3A). Like PIKfyve^K1831E, PIKfyve^K1831EΔCpn+ was devoid of lipid kinase activity (Fig. 4A, supplemental Fig. 2, quantified in Fig. 3A). Remarkably, however, unlike PIKfyve^K1831E that unconditionally triggers aberrant endomembrane defects (18), PIKfyve^K1831E with Cpn+ deletion failed to induce endomembrane swelling and vacuolation as apparent from the phase-contrast images of PIKfyve^K1831EΔCpn+-expressing cells (Fig. 4B). Defective cell morphology was undetectable even after prolonged PIKfyve^K1831EΔCpn+ expression (48 h, not shown), conditions under which PIKfyve^K1831E expression further exacerbates the vacuolar defects (18). Likewise, PIKfyve^K1831EΔCpn+-dependent endomembrane vacuoles were not seen upon coexpression with Sac3^WT and ArPIKfyve^WT alone or in combination (not shown).

To assess whether the inability of PIKfyve^K1831EΔCpn+ to imbalance endomembrane homeostatic mechanisms and trigger endomembrane vacuoles is related to possible mislocalization of the mutant, we treated the transfected cells with wortmannin. As with PIKfyve^K1831E (8), wortmannin rendered the PIKfyve^K1831EΔCpn+ vesicular pattern diffuse, consistent with localization of the PIKfyve^K1831EΔCpn+ mutant to PtdIns(3)P-enriched endosomes (Fig. 4B). Similarly to PIKfyve^K1831E (8), the GFP-PIKfyve^K1831EΔCpn+-positive vesicles displayed substantial colocalization (∼40%) with the EEA1 endosomal marker (Fig. 4C) and nearly completely overlapped the PIKfyve^K1831E immunofluorescence signals (not shown). Together, these data indicate that if kinase-deficient PIKfyve^K1831E is incapacitated for binding ArPIKfyve-Sac3, it does not affect endomembrane homeostasis despite the correct localization and dead kinase activity.

We also examined the lipid kinase activity of other PIKfyve-truncated mutants (supplemental Fig. 2) that displayed intact kinase domain but reduced ArPIKfyve-Sac3 binding and inspected the ability of the mutant to induce vacuolar defects. Quantitative summary of these observations from at least three independent experiments for each construct is presented in Fig. 3A. The comparative analysis allows two important conclusions. First, impeded binding to the ArPIKfyve-Sac3 subcomplex is associated with markedly reduced lipid kinase activity of the mutants (supplemental Fig. 2 and Fig. 3A). We have to point out, however, that in addition to reduced levels of bound ArPIKfyve-Sac3 subcomplex, resulting in kinase inhibition (12, 14, 21), conformational changes due to the deletions might also contribute to the drastic loss of the lipid kinase activity and could not be ruled out. Second, if the mutants fail to bind ArPIKfyve-Sac3 with efficiency as high as that of PIKfyve^WT, they are unable to trigger vacuolar defects despite drastically reduced (>30-fold) lipid kinase activity and proper localization to PtdIns(3)P, determined by the intact FYVE domain.

Sac3^WT Exacerbates the Vacuolar Defects by PtdIns(3,5)P₂-deficient PIKfyve Mutants with Intact Sac3 Binding

The above result that expression of PIKfyve^K1831EΔCpn+, in contrast to PIKfyve^K1831E, is not dominantly interfering suggests that associated ArPIKfyve-Sac3 and particularly the PtdIns(3,5)P₂-hydrolyzing activity of Sac3 contribute a great deal to the dominant-negative effect of the PtdIns(3,5)P₂-deficient point mutants on the endomembrane morphology. As we elaborated previously, the dominant-negative phenotype of kinase-deficient PIKfyve^K1831E is a complex, gradually developing process that affects different endosomal types depending on the duration of expression (8, 18). At earlier stages of COS cell transfection (9–15 h), the PIKfyve^K1831E-positive vesicles are enlarged, but translucent vacuoles are not visible. The latter emerge 15–24 h post-transfection, first at the perinuclear region and then in the whole cell. The vacuoles, with or without PIKfyve^K1831E on the limiting membrane, progressively increase in size and decrease in number as a result of fusion (8, 18). Therefore, to assess the role of Sac3 activity in induction and progression of the PIKfyve^K1831E dominant-negative effect, we examined the phenotypic changes in COS cells triple-transfected with PIKfyve^K1831E, ArPIKfyve^WT, and Sac3 in either phosphatase-active (Sac3^WT) or -deficient forms (Sac3^D488A), both forming PAS complexes with similar efficiencies (Fig. 1, A and B). Cells were monitored by immunofluorescence microscopy at two time points: 10–12 and 22–24 h post-transfection. Intriguingly, the cell phenotype under triple expression of eGFP-PIKfyve^K1831E, Myc-ArPIKfyve^WT, and HA-Sac3^WT was markedly different compared with that under expression of PIKfyve^K1831E alone. Thus, the early phase of endosome vesicle swelling seen under single eGFP-PIKfyve^K1831E expression was not manifested (Fig. 5A). Rather, the PIKfyve^K1831E/ArPIKfyve^WT/Sac3^WT-transfected cells displayed large translucent vacuoles within the whole cell as early as 10–12 h post-transfection, which persisted 24 h post-transfection (Fig. 5, A and B). By contrast, cells expressing eGFP-PIKfyve^K1831E, Myc-ArPIKfyve^WT, and HA-Sac3^D488A displayed neither dilated endosomes at the early phase nor translucent vacuoles in the later phase to a substantial degree (Fig. 5, A and B). Noteworthy, no significant Sac3-related phenotypic changes were seen if eGFP-PIKfyve^K1831E was coexpressed with Sac3^WT or Sac3^D488A in the absence of ArPIKfyve^WT (Fig. 5B).

As a further test of our hypothesis for Sac3 functioning as an active phosphatase in the PAS complex, we examined the effect of Sac3^WT versus Sac3^D488A on cell morphology under expression of the PIKfyve^K2000E point mutant. As opposed to Lys-1831 or Lys-1999, Lys in position 2000 was previously characterized to be partially engaged in PtdIns(3,5)P₂ coordination, resulting in only an ∼40% decreased in vitro PtdIns(3,5)P₂ synthesis by PIKfyve^K2000E (17). At this level of activity, PIKfyve^K2000E, unlike PIKfyve^K1831E or PIKfyve^K1999E, does not induce vacuolar defects in transfected COS cells (17). Remarkably, coexpression of Sac3^WT and ArPIKfyve with PIKfyve^K2000E triggered profound cell vacuolation, whereas coexpression of Sac3^D488A-ArPIKfyve did not substantially affect the PIKfyve^K2000E-dependent cell morphology. Quantitation of the PIKfyve^K2000E-induced vacuolar phenotype due to Sac3^WT versus Sac3^D488A from two independent transfection experiments is presented in Fig. 5C. Taken together, the results illustrated in Figs. 4 and 5 are consistent with the conclusion that Sac3 functions as a PtdIns(3,5)P₂ phosphatase in the context of the PAS complex.

ArPIKfyve^Ct That Disassembles the PAS Complex Alleviates the Phenotypic Defects by PIKfyve^K1831E

As revealed recently, ArPIKfyve homomeric interactions mediated by the ArPIKfyve C-terminal domain scaffolds the PAS complex (14). Therefore, as a final verification of our hypothesis that Sac3 functions as a phosphatase when associated with PIKfyve, we examined if disassembly of the PAS complex by ArPIKfyve^Ct (14) alleviates the phenotypic defects induced by PIKfyve^K1831E in transfected COS cells. As illustrated in Fig. 5D, expressed Myc-PIKfyve^K1831E was significantly less potent in triggering formation of aberrant vacuoles if cells coexpressed ArPIKfyve^Ct. Quantitation of 2 separate experiments counting ∼250 transfected cells revealed that coexpressed ArPIKfyve^Ct diminished the number of vacuolated cells due to PIKfyve^K1831E by 42 ± 4%. These data further indicate that the dominant phenotype of aberrant swollen vacuoles is due to two superimposed events; first, disrupted PIKfyve-catalyzed PtdIns(3,5)P₂ synthesis at endosomal microdomains determined by the PIKfyve FYVE finger and, second, Sac3-dependent hydrolysis localized at the PtdIns(3,5)P₂ synthetic sites through association of the ArPIKfyve-Sac3 subcomplex with PIKfyve.