RILP regulates vacuolar ATPase through interaction with the V1G1 subunit

Vacuolar-type H⁺ ATPases (V-ATPases) are large multi-subunit complexes that control the acidity of intracellular compartments, and they are also responsible for proton transport across the plasma membrane and thus for acidification of the extracellular environment (Forgac, 2007; Sun-Wada et al., 2004; Toei et al., 2010). They are proton pumps that are conserved in all eukaryotes, and they are organized into two domains that constitute a rotary machine (Cipriano et al., 2008; Kane, 2006; Kane and Smardon, 2003). V0 is the integral transmembrane domain responsible for proton translocation, whereas V1 is the peripheral domain located on the cytoplasmic side of the membrane, and it is responsible for ATP hydrolysis, thus providing energy to pump the protons (Cipriano et al., 2008; Kane, 2006; Kane and Smardon, 2003). Accurate control of the pH of intracellular organelles and the extracellular milieu is fundamental for many cellular processes, including the processing and degradation of macromolecules in secretory and digestive compartments, coupled transport of small molecules such as neurotransmitters, sperm maturation and osteoclast function. Defects in acidification have been recently recognized as an important cause of several severe human diseases, such as renal tubule acidosis, osteopetrosis, defects in sperm maturation and diabetes (Casey et al., 2010; Marshansky and Futai, 2008). In addition, V-ATPase activity regulates membrane fusion, early-to-late endosomal trafficking and neurosecretion (El Far and Seagar, 2011; Hurtado-Lorenzo et al., 2006; Qiu, 2012). These proton pumps are required for invasion by endothelial cells during angiogenesis. Importantly, V-ATPases also contribute to the invasive properties of tumor cells (Forgac, 2007; Sennoune et al., 2004b). In tumor cells, V-ATPases are targeted to the plasma membrane and acidify the extracellular environment, activating lysosomal enzymes that are responsible for the degradation of the extracellular matrix. Thus, V-ATPases are directly involved in invasiveness and are now considered a very promising target in metastasis (Forgac, 2007).

The reversible dissociation of the V-ATPase complex into the V1 and V0 domains represents a fundamental control mechanism of V-ATPase function that acts to vary the rate of proton pumping (Forgac, 2007). However, the reversible dissociation of single subunits within the V1 and V0 domains has also been reported to regulate V-ATPase activity (Qi et al., 2007). A complex named RAVE (regulator of ATPase of vacuoles and endosomes) regulates the assembly of the pumps (Smardon et al., 2002). In addition, extracellular pH also tightly regulates V-ATPase activity and assembly (Diakov and Kane, 2010).

Rab-interacting lysosomal protein (RILP) is an effector of the small GTPase Rab7 (Cantalupo et al., 2001; Bucci et al., 1988; Chavrier et al., 1990). Rab proteins localize to the cytosolic side of intracellular organelles and regulate membrane traffic (Aloisi and Bucci, 2013; Bucci and Chiariello, 2006; Stenmark, 2009). Rab7, in its active GTP-bound form, recruits RILP to the membrane and together they control trafficking to late endosomes and lysosomes in the endocytic route (Bucci et al., 2000; Cantalupo et al., 2001; Jordens et al., 2001). Furthermore, RILP is involved in the biogenesis of multivesicular bodies (MVBs) (Progida et al., 2007; Progida et al., 2006).

Using the yeast two-hybrid system, we have identified the V1G1 (officially known as ATP6V1G1) subunit of V-ATPase as a RILP-interacting protein. Our data indicate that RILP, through its direct interaction with V1G1, regulates the function of V-ATPase. Thus, we have discovered a new regulatory mechanism of V-ATPase function, and, given the involvement of V-ATPase in several human diseases, these data suggest that RILP could become a useful target for the development of effective therapeutic interventions.

A yeast two-hybrid screen was used to identify proteins that interact with RILP. A Gal4-binding-domain–RILP fusion construct was encoded in the pGBKT7 vector and was used in the screen to identify putative interactors. The construct was used to screen a liver cDNA library encoding proteins as C-terminal fusions with the transcriptional-activation domain of Gal4 in the pACT2 vector. Of 1.3×10⁶ primary transformants, 27 were His⁺LacZ⁺ and encoded true positives that did not activate transcription in the presence of a non-specific test bait. Seven transformants encoded the entire V1G1 subunit of the vacuolar ATPase. The interaction was revealed by the growth of yeast cells expressing RILP and V1G1 on synthetic medium lacking histidine and adenine. In the two-hybrid system, the interaction was specific as the yeasts expressing V1G1 together with Rab7 were not able to grow without histidine or without histidine and adenine (supplementary material Fig. S1A). These results were confirmed using the other reporter gene, β-galactosidase (supplementary material Fig. S1B). Using a β-galactosidase quantitative liquid assay, we confirmed the interaction between RILP and V1G1, and established, using previously described RILP deletion mutant constructs (Colucci et al., 2005a), that V1G1 binds to the N-terminal half of RILP (supplementary material Fig. S1).

To confirm the results obtained with the two-hybrid screen we investigated whether the two proteins were able to co-immunoprecipitate. HeLa cells were transfected with constructs expressing HA-tagged RILP, V1G1 or both, or were not transfected, and immunoprecipitation was performed using an anti-HA resin. The presence of immunoprecipitates of HA-tagged RILP containing the V1G1 subunit of V-ATPase confirmed the interaction (Fig. 1A). In addition, by overexpressing V1G1, RILP, RILPC33 (a truncated mutant lacking the N-terminal half of the protein) and RILPΔC2 (a truncated mutant lacking the C-terminal half of the protein), we established that V1G1 could be immunoprecipitated by RILPΔC2 but not by RILPC33 (Fig. 1B), thus confirming the data obtained with the two-hybrid system, which indicated that V1G1 binds to the N-terminal half of RILP (supplementary material Fig. S1). In order to verify the physiological significance of the interaction, we also demonstrated co-immunoprecipitation of endogenous RILP and V1G1 using a crosslink immunoprecipitation kit as described in Materials and Methods (Fig. 1C).

Subsequently, to check whether the interaction was direct, we purified bacterially expressed GST, GST-tagged RILP, GST-tagged RILPC33 and His-tagged V1G1. Glutathione resin alone or bound to GST, GST–RILP or GST–RILPC33 was incubated with His-tagged V1G1. The V1G1 subunit specifically bound only to GST-tagged RILP, demonstrating a direct interaction between V1G1 and RILP (Fig. 1D). Taken together, these results demonstrate that RILP is able to interact directly with the V1G1 subunit of the vacuolar ATPase, and this interaction is physiologically relevant.

RILP is able also to interact directly with the C-terminal domain of p150^Glued (also known as DCTN1), a component of the dynactin projecting arm domain, controlling transport of Rab7-positive organelles towards the cell center (Cantalupo et al., 2001; Jordens et al., 2001; Johansson et al., 2007; Harrison et al., 2003). Because acidification of endosomes correlates with their translocation towards the microtubule organizing center, we investigated whether the interactions of RILP with p150^Glued and V1G1 are mutually exclusive. Using purified His–V1G1 bound to Ni–NTA resin we were able to pull down FLAG–p150^Glued from lysates of transfected HeLa cells only in the presence of purified GST–RILP (Fig. 1E). These data indicate that RILP can bind to V1G1 and p150^Glued at the same time.

To investigate the role of this interaction, we decided to analyze whether modulation of RILP expression levels affected V1G1 abundance. Interestingly, upon RILP overexpression, the total amount of the V1G1 subunit in HeLa cells was reduced (Fig. 2A). By contrast, in RILP-silenced cells, the amount of V1G1 was strongly increased (Fig. 2B). Importantly, RILP overexpression or silencing did not affect V1C1 and V0D1, two other V-ATPase subunits (Fig. 2A,B), indicating that RILP regulates V1G1 specifically. Notably, expression of RILPΔC2 or RILPC33 did not affect V1G1 abundance (Fig. 2C), suggesting that only the entire protein is able to act on V1G1 protein abundance. Furthermore, overexpression of HA-tagged RILP was able to partially rescue the effect on V1G1 abundance in RILP-silenced cells, thus excluding off-target effects (Fig. 2D).

To establish whether the observed changes in the amount of V1G1 protein were a consequence of mRNA abundance, we checked V1G1 mRNA expression using real-time PCR (Fig. 2G). As expected, in cells transfected with HA–RILP, RILP mRNA was increased, whereas in RILP-silenced cells, RILP mRNA levels decreased (Fig. 2E,F). Interestingly, we found that the amount of V1G1 mRNA upon RILP overexpression was slightly increased, whereas it was decreased upon RILP silencing (Fig. 2G). Thus, mRNA expression seems to be altered in order to counteract the effects on the amount of V1G1 protein that are caused by RILP overexpression or silencing.

As we established that RILP expression levels regulate the amount of V1G1, we then decided to investigate whether RILP was responsible for the proteasomal degradation of V1G1. We overexpressed RILP and treated cells with the proteasome inhibitor MG132. Under these conditions, the endogenous V1G1 protein level did not decrease, suggesting that RILP regulates the proteasomal degradation of V1G1 (Fig. 2H). We then decided to investigate whether V1G1 was ubiquitylated. We expressed HA–V1G1 in HeLa cells in the presence or absence of the proteasomal inhibitor MG132 and in the presence or absence of RILP overexpression. Lysates of cells were then immunoprecipitated with anti-HA resin and subjected to western blot analysis using an anti-ubiquitin antibody. Ubiquitylation of V1G1 was present in all samples where HA–V1G1 was expressed and immunoprecipitated, but was increased in the presence of MG132, suggesting a strong regulation of V1G1 mediated by the proteasome. The amount of V1G1 protein decreased upon RILP overexpression, whereas treatment with MG132 partially rescued V1G1 levels and strongly increased the level of ubiquitylated forms of V1G1. Hence, taking into consideration V1G1 levels, RILP overexpression induces V1G1 ubiquitylation (Fig. 2I).

We next analyzed whether RILP is responsible for the recruitment of V1G1 to membranes. We treated cells with MG132 in order to prevent V1G1 ubiquitin-dependent proteasomal degradation induced by RILP, and we looked at the cytosolic and membrane distribution of endogenous V1G1 in control cells and in cells overexpressing RILP. As loading controls, we used GAPDH, a cytosolic protein, and V0D1, a membrane subunit of V-ATPase. Expression of HA–RILP did not change the membrane or cytosol distribution of GAPDH and V0D1. The analysis of the distribution of V1G1 showed that the amount of membrane-associated V1G1 increased upon RILP overexpression, whereas the amount of cytosolic V1G1 decreased (Fig. 3A). Indeed, RILP overexpression increased membrane localization of V1G1 by about 30%, whereas cytosolic V1G1 was reduced by 85% compared with control (Fig. 3B). These data demonstrate that V1G1 membrane localization increases upon RILP overexpression, and they suggest that RILP could recruit V1G1 to endosomal and/or lysosomal membranes.

Having shown that RILP controls the amount of V1G1 and its recruitment to membranes, we decided to investigate whether V1G1 binds to RILP when it is in complex with other V-ATPase subunits. We used an anti-HA resin to immunoprecipitate HA-tagged RILP from lysates of HeLa cells expressing HA-tagged RILP and endogenous or overexpressed V1G1. In both immunoprecipitates, we observed the presence not only of V1G1 but also of V1C1 and V0D1, two other V-ATPase subunits (Fig. 3C). Larger amounts of co-immunoprecipitated V1G1, obtained when V1G1 was overexpressed, led to increased co-immunoprecipitation of the other two subunits (Fig. 3C). These data suggest that V1G1 interacts with RILP alone or complexed with the other subunits, thus suggesting that RILP could be important for controlling the assembly of the entire pump at the endosomal membranes by recruiting V1G1.

We also used confocal immunofluorescence to analyze the intracellular localization of V1G1 upon RILP overexpression. As commercially available antibodies against the V1G1 subunit of V-ATPase do not work in immunofluorescence, we looked at the localization of the overexpressed HA-tagged V1G1 subunit. As expected, V1G1 was partially colocalized with Lamp1 on late endosomal and lysosomal structures (Fig. 4). Expression of RILP caused clustering of late endosomal and lysosomal structures, as reported previously (Cantalupo et al., 2001; Colucci et al., 2005b), and HA-tagged V1G1 was localized on the clustered structures that were generated and marked by GFP–RILP (Fig. 4A). Interestingly, upon RILP overexpression, the colocalization percentage of V1G1 with Lamp1 is increased (Fig. 4A,B). By contrast, expression of RILP-C33, the deletion mutant containing only the C-terminal half of RILP that does not interact with V1G1, decreased colocalization with Lamp1 (Fig. 4A,B). These data suggest that the localization of V1G1 on late endosomal and/or lysosomal membranes is regulated by RILP. Furthermore HA-tagged V1G1 partially colocalized with RILP on late endosomal and/or lysosomal structures labelled with Lamp1, whereas very little colocalization was detected with RILPC33 (Fig. 4A). Quantification of the immunofluorescence data (Fig. 4C) indicates that <10% of HA–V1G1 colocalized with RILPC33, thus confirming the interaction data.

In addition, we analyzed the effects of RILP silencing on the localization of HA-tagged V1G1 (Fig. 5). Interestingly, in RILP-depleted cells (RILPi) the colocalization of HA-tagged V1G1 with Lamp1 decreased, thus confirming the role of RILP in recruiting V1G1 to late endosomal and lysosomal membranes (Fig. 5). In addition, upon RILP silencing, the colocalization of HA-tagged V1G1 with giantin (also known as GOLGB1), a marker of the Golgi complex, strongly increased, indicating that, in the absence of RILP, V1G1 localizes mainly to other compartments (Fig. 5). Taken together, these data indicate that RILP recruits V1G1 to late endosomal and lysosomal membranes and suggest that RILP could be important for the assembly of the pump on these organelles.

RILP is a Rab7 effector protein and, thus, in order to investigate whether Rab7 is also involved in the regulation of V1G1, we transiently overexpressed or silenced Rab7 and looked at V1G1 abundance (Fig. 6A,B). Interestingly, Rab7 overexpression did not affect V1G1 abundance, whereas Rab7 silencing caused a strong decrease in the amount of V1G1 (Fig. 6A,B). Furthermore, the decrease in the amount of V1G1 caused by RILP overexpression was counteracted by Rab7 overexpression (Fig. 6C), thus suggesting that when RILP binds to Rab7 it does not induce V1G1 degradation.

In order to further investigate the role of Rab7 in the interaction between RILP and V1G1, we used HeLa cells to co-express HA–RILP, HA–RILPΔC2, HA–RILPC33, Myc–Rab7 and V1G1 in different combinations, and we immunoprecipitated V1G1 using a specific anti-V1G1 antibody (Fig. 6D). Immunoprecipitates were then analyzed for the presence of the different recombinant proteins. As expected, V1G1 was able to co-immunoprecipitate HA–RILP and HA–RILPΔC2 but not HA–RILPC33 (Fig. 6D, lanes 2–4). Interestingly, V1G1 was also able to co-immunoprecipitate Rab7 (Fig. 6D, lane 5), although in the two-hybrid system we did not detect interaction between these two proteins (supplementary material Fig. S1). However, co-immunoprecipitation of V1G1 and Rab7 was seen both upon exogenous HA–RILP expression (Fig. 6D, lane 7) and in the presence of endogenous RILP (Fig. 6D, lane 5). In RILP-silenced cells, Rab7 no longer co-immunoprecipitated with V1G1 (Fig. 6D, lane 6). These data indicate that V1G1 does not interact directly with Rab7 and that RILP is able to interact simultaneously with Rab7 and V1G1. Indeed, expression of the RILP truncated forms RILPΔC2 or RILPC33, which bind to and sequester V1G1 and Rab7, respectively, abolished the co-immunoprecipitation of Rab7 (Fig. 6D, lanes 12–13). Taken together, these data indicate that Rab7 recruits RILP to endolysosomal membranes, and RILP in turn recruits V1G1, whereas RILP induces V1G1 proteasomal degradation when it is not bound to Rab7.

RILP is required for the biogenesis of MVBs and, together with Rab7, it regulates late endocytic traffic (Progida et al., 2007). Acidification, triggered by V-ATPase, is required for the formation of MVBs (Aniento et al., 1996; Clague et al., 1994; Marshansky and Futai, 2008; Trajkovic et al., 2008). Thus, RILP might coordinate the biogenesis of MVBs through V-ATPase activity regulation, by controlling the amount and localization of V1G1. In order to investigate whether RILP influences V-ATPase activity, we used LysoTracker Red, a marker of acidic compartments that stains vesicles with an internal pH of <6.5, in HeLa cells depleted of RILP (RILPi) (Fig. 7A). The samples were analyzed by live microscopy, and LysoTracker Red intensity in RILPi samples relative to controls (scr) was quantified (Fig. 7B). Interestingly, the levels of LysoTracker Red intensity were increased more than twofold in RILPi cells, suggesting a role for RILP in controlling V-ATPase activity (Fig. 7A,B). Furthermore, the distribution of acidic compartments was more diffuse and peripheral in RILP-depleted cells compared with control cells.

Because we had demonstrated that RILP is important for acidification and controls V1G1 abundance and recruitment to late endosomal and lysosomal membranes, we decided to investigate the relevance of V1G1 to V-ATPase activity. In order to monitor how variations of V1G1 abundance affect V-ATPase activity, we monitored cathepsin D maturation in HeLa cells that overexpressed V1G1 or that were silenced for V1G1 either transiently or stably. Cathepsin D is synthesized as preprocathepsin D precursor, which is converted into procathepsin D (52 kDa) in the endoplasmic reticulum, and then undergoes further proteolytic processing in the acidic milieu of late endosomes and lysosomes, being converted in a 44-kDa form and finally into the 32-kDa mature form. Lysates of control HeLa cells and of V1G1-silenced or V1G1-overexpressing cells were subjected to western blot analysis with an anti-cathepsin-D antibody that recognized the 52-kDa, 44-kDa and 32-kDa forms (Fig. 7C). In control cells, the 52-kDa and the 44-kDa forms represented <30% of the total cathepsin D staining, whereas, as expected, treatment with bafilomycin A1, a high-affinity inhibitor of V-ATPase, caused the accumulation of the immature forms and reduced the amount of the mature 32-kDa form to <5% (Fig. 7C,D). Transient or stable overexpression (in two independent clones) of V1G1 caused a strong increase in the amount of immature cathepsin D forms, whereas the mature 32-kDa form represented <25% of the total cathepsin D staining (Fig. 7C,D). Similarly, transient or stable depletion of V1G1 (in two independent clones) caused an accumulation of the immature forms, whereas the mature 32-kDa form of cathepsin D accounted for <25% of the total cathepsin D staining (Fig. 7C,D). By contrast, no changes were detected in cells transfected with a construct expressing a control short-hairpin (sh)RNA (Fig. 7C,D).

As V1G1 seems to be important to ensure proper V-ATPase activity, we wanted to investigate the effect on late endosomes and lysosomes when cells were depleted of V1G1. We therefore used two different markers of acidic compartments – LysoTracker Red, which accumulates in acidic compartments, and LysoSensor DND-192, which labels less acidic compartments than LysoTracker Red – and we allowed HeLa cells that were stably depleted of V1G1 (in two independent clones) to internalize fluorescently labeled transferrin. Afterwards, we quantified the relative amount of transferrin that colocalized with LysoTracker-Red- or LysoSensor-positive endosomes, by using confocal microscopy analysis (supplementary material Fig. S2). Internalized transferrin mainly localizes to sorting and recycling early endosomes, and only a small percentage of it reaches late endosomes and lysosomes to be degraded (Dautry-Varsat et al., 1983; Hopkins and Trowbridge, 1983). Changes in the percentage of transferrin colocalizing with endocytic degradative compartments represent a clear alteration of the endocytic route (Roxrud et al., 2009).

The levels of transferrin in LysoTracker-positive structures increased in both clones depleted of V1G1, and the increase was stronger in LysoSensor-positive endosomes (supplementary material Fig. S2). These results confirm the importance of V1G1 for the V-ATPase activity, as the depletion of this subunit causes accumulation of transferrin in acidic compartments, suggesting impaired degradation, and thus malfunctioning of late endosomes and lysosomes. Taken together, these data demonstrate that changes in the amount of V1G1 impair V-ATPase activity.

V-ATPases are highly conserved enzymes that couple ATP hydrolysis to proton transport across membranes, and they consist of two domains and multiple subunits and are localized to the membrane of several organelles (Cipriano et al., 2008; Toei et al., 2010). The G subunit of the V-ATPase V1 domain is part, together with the E subunit, of the peripheral stator stalks responsible for structurally and functionally linking the peripheral V1 sector to the membrane V0 sector of the pump (Armbrüster et al., 2003; Ohira et al., 2006). There are two isoforms of the G subunit – G1 is ubiquitously expressed, whereas G2 is specifically expressed in central nervous system neurons (Murata et al., 2002). Here, we have identified the G1 subunit of V-ATPase as an RILP-interacting protein. We have proved that the interaction is physiological, as it was detected by co-immunoprecipitation of overexpressed (Fig. 1A) but also endogenous proteins (Fig. 1C). Furthermore, we have proved that the interaction is direct, as it was also detectable between bacterially expressed and purified proteins (Fig. 1D). In addition, our data indicate that V1G1 interacts with the N-terminal half of RILP (Figs 1,6D; supplementary material Fig. S1), in contrast to Rab7, which binds in the C-terminal half of RILP (Cantalupo et al., 2001). Interestingly, the RILP–V1G1 complex can interact with the dynactin projecting arm p150^Glued (Fig. 1E), suggesting that RILP could simultaneously control endosome acidification and translocation towards the microtubule organizing center.

Our data indicate that RILP regulates the recruitment of V1G1 to endosomes and lysosomes. Overexpression of RILP increases the colocalization of V1G1 with the late endosomal and lysosomal marker Lamp1, whereas expression of the RILPC33 mutant, which is unable to bind to V1G1, decreases it (Fig. 4). In addition, overexpression of RILP increases the amount of V1G1 present in membranes and drastically decreases V1G1 in the cytosol (Fig. 3A,B). In agreement, in RILP-silenced cells colocalization of V1G1 with Lamp1 decreases, whereas co-localization with giantin increases (Fig. 5), indicating that RILP is important for the recruitment of V1G1 to late endosomal and lysosomal membranes. Thus, GTP-bound Rab7 recruits RILP to late endosomes and lysosomes (Cantalupo et al., 2001), and RILP, in turn, recruits V1G1 to these organelles. In fact, we demonstrated by co-immunoprecipitation that RILP is able to bind simultaneously to V1G1 at its N-terminal half and Rab7 at its C-terminal half (Fig. 6D). Also, we established that V1G1 is not able to bind to Rab7 directly, because V1G1 is not able to interact with Rab7 in the two-hybrid system (supplementary material Fig. S1), and it is not able to co-immunoprecipitate Rab7 in the absence of RILP (Fig. 6D).

As V1G1 is important for V-ATPase assembly and function (Ohira et al., 2006; Tomashek et al., 1997), the recruitment of V1G1 to endosomal or lysosomal membranes is likely to contribute to the correct assembly and, consequently, functioning of the pump. In fact, our data show that V1G1, when binding to RILP, can also be in complex with other subunits of the cytosolic and membrane domain of V-ATPase (Fig. 3C), thus suggesting that RILP could control not only the recruitment of V1G1 to endosomal and lysosomal membranes but also the assembly of the entire pump. Furthermore, we have demonstrated that RILP modulates the activity of the pump on late endosomes and lysosomes, controlling V1G1 localization (Figs 4,5). Indeed, in RILP-depleted cells we detected mislocalization of V1G1 to other organelles (Fig. 5) and a different intracellular distribution and higher intensity of LysoTracker Red (Fig. 7A,B). These data, together with the previously reported inhibition of EGFR degradation and increase in Lamp levels upon RILP silencing (Progida et al., 2006; Progida et al., 2007), suggest that there is a block of endosomal maturation caused by the failed recruitment of the proton pump to Rab7-labeled endosomes.

We have shown previously that RILP is important for the formation of MVBs (Progida et al., 2007; Progida et al., 2006). As MVB formation is dependent on the luminal acidic pH of endosomes and thus on the acidification triggered by V-ATPase (Aniento et al., 1996; Clague et al., 1994; Marshansky and Futai, 2008; Trajkovic et al., 2008), our data support the idea that RILP might coordinate the biogenesis of MVBs through regulation of V1G1 activity.

Our data also demonstrate that RILP regulates V1G1 subunit abundance. Overexpression of RILP causes a reduction in the amount of V1G1, whereas, in RILP-silenced cells, the amount of V1G1 is strongly increased (Fig. 2A,B). Furthermore, we established that RILP induces V1G1 ubiquitylation-dependent proteasomal degradation (Fig. 2I). Thus, RILP promotes the recruitment of V1G1 to endosomal membranes but also induces V1G1 degradation. How are these two different and contrasting functions regulated? Our data indicate that overexpression of Rab7 has no effect on V1G1 abundance, whereas Rab7 silencing causes a strong decrease in the amount of V1G1 (Fig. 6A,B). Hence, these data suggest that when RILP is bound to Rab7 it recruits V1G1 to membranes, whereas an excess of RILP that is not bound to Rab7 (as in the case of RILP overexpression or Rab7 silencing) leads to V1G1 degradation. Therefore, the amount of V1G1 is strictly regulated by the availability of cytosolic RILP.

Previously, we have observed that expression of the Rab7T22N mutant, which is unable to recruit RILP to late endosomal or lysosomal membranes, causes a strong decrease in LysoTracker Red staining (Bucci et al., 2000), thus indicating an increase in pH of intracellular organelles, possibly caused by impaired V-ATPase assembly and function due to the lack of RILP-mediated recruitment of V1G1. These data seem to contrast with the fact that RILP silencing causes an increase in LysoTracker intensity (Fig. 7A,B). However, in cells expressing the Rab7T22N mutant, RILP is still present but it cannot be recruited to endosomes and could, instead, induce V1G1 degradation.

We also demonstrated that a proper amount of V1G1 seems to be fundamental for the correct functioning of the pump. Overexpression or silencing of the V1G1 subunit causes a strong inhibition of cathepsin D maturation, similar to what happens when treating the cells with bafilomycin A1, an inhibitor of the vacuolar pump (Fig. 7C,D). Moreover, V1G1 silencing causes the accumulation of transferrin in acidic compartments, suggesting that it is linked to malfunctioning of late endosomes and lysosomes (supplementary material Fig. S2). These data are in agreement with previously published data suggesting that changes in V1G1 expression destabilize the E subunit, influencing V-ATPase assembly (Ohira et al., 2006; Tomashek et al., 1997). The V1G1 subunit thus seems to be a crucial subunit not only for correct assembly of the vacuolar pump but also for V-ATPase function, as changes in the amount of V1G1 have strong negative effects on processes regulated by V-ATPase.

The discovery of the interaction between RILP and V1G1 is of importance, as it reveals a new mechanism of regulation of the V-ATPase, mediated by RILP. In particular, our data demonstrate that RILP regulates V1G1 abundance and is crucial for the regulation of V-ATPase at the level of late endosomes and lysosomes, promoting assembly and activation of the proton pump on these organelles (Fig. 8).

V-ATPase has been proposed as a drug target for several diseases and processes (Hinton et al., 2009; Kartner and Manolson, 2012). Plasma membrane V-ATPases are important for sperm maturation and viability (Pietrement et al., 2006), for acid secretion by certain kinds of renal cells (the malfunctioning of which causes renal tubular acidosis) (Karet et al., 1999; Smith et al., 2000) and for osteoclast bone degradation (Frattini et al., 2000; Toyomura et al., 2003), which, if altered, causes osteopetrosis and osteoporosis. Importantly, in many tumor cell lines, V-ATPases are abundant in plasma membranes, and their activity in plasma membranes has been directly correlated with invasiveness (Sennoune et al., 2004a; Sennoune et al., 2004b; Sennoune and Martinez-Zaguilan, 2012). The tumor acidic microenvironment plays a key role in tumor development and progression, as it influences resistance to chemotherapy, proliferation and formation of metastases (Hernandez et al., 2012). Thus, in order to fight cancer, the use of V-ATPase inhibitors and/or the regulation of V-ATPase assembly have been proposed to control V-ATPase activity (Kane, 2012; Pérez-Sayáns et al., 2009). Our data suggest that modulation of RILP expression could represent a mechanism to regulate V-ATPase activity. Furthermore, the discovery of a functional link between RILP and V-ATPase opens new possibilities for the cellular role of RILP that will have to be further investigated.

Restriction and modification enzymes were from New England Biolabs (Ipswich, MA), chemicals were from Sigma-Aldrich (St Louis, MO) and tissue culture reagents were from Sigma-Aldrich or Invitrogen (Carlsbad, CA). HeLa cells were grown in DMEM supplemented with 10% FBS, 2 mM glutamine, 100 U/ml penicillin and 10 mg/ml streptomycin, in an incubator at 37°C, under 5% CO₂. HeLa cells were stably transfected with pCDNA3_2xHA_V1G1 plasmid or stably silenced using the sh-RNA V1G1 plasmids (sc-36797-SH, Santa Cruz Biotechnology, Santa Cruz, CA). Stable cell lines were selected by adding G-418 (1 mg/ml, Calbiochem, Darmstadt, Germany) to cells transfected with pCDNA3_2xHA_V1G1 plasmid or by adding puromycin (3 µg/ml, Sigma-Aldrich) to V1G1-silenced cells. Where indicated, MG132 (10 µM, C2211 Sigma-Aldrich) was used for 3–5 h.

Most Rab7 and RILP constructs used in this study have been described previously (Cantalupo et al., 2001; Colucci et al., 2005a; Progida et al., 2006; Vitelli et al., 1997). The pCDNA3_2xHA_RILP construct was used to amplify a fragment using the following oligonucleotides; 5′-CTTAGAATTCTAATGGAGCCCAGGAGGGCGGCG-3′ and 5′-ATAAGAATGCGGCCGCTCACCATTCGGGCTCCTGTCCGTGCTG-3′. The fragment, coding for the N-terminal half of RILP, was then digested with EcoRI and NotI and cloned into the pCDNA3-2xHA vector to obtain a construct for the expression of an HA-tagged RILPΔC2. The V1G1 ORF was amplified from the two-hybrid pACT2-VIG1 construct using the following oligonucleotides; 5′-CTAAGATCTATGGCTAGTCAGTCTCAGGGG-3′ and 5′-CGCAAGCTTCTATCCATTTATGCGGTAGTT-3′. The amplified fragment was then digested with BglII and HindIII and cloned into the pEGFP-C1 vector to obtain a construct for the expression of GFP-tagged V1G1. The pCDNA3-V1G1 construct was obtained after cloning the BglII-HindIII fragment from the pEGFP-V1G1 vector into the pCDNA3 vector cut with BamHI and HindIII. To obtain the HA-tagged V1G1, an EcoRV-XhoI fragment containing the V1G1 ORF was cloned into the pCDNA3_2xHA plasmid. The pCMV6-AC-GFP-V1G1 (RG203317) vectors were obtained from Origene (Rockville, MD). The pEF-FLAG-p150^Glued construct has been described previously (Ohbayashi et al., 2012).

The construct pGBKT7-RILP was used to screen a human liver cDNA library in the pACT2 vector in AH109 yeast cells (Bartel et al., 1993; Suter et al., 2008). Transformants were plated onto synthetic medium lacking His, Leu and Trp. Colonies were picked 5 days later and then assayed for growth on medium lacking Ade, His, Leu and Trp and for β-galactosidase activity (Bartel et al., 1993). Specificity tests were performed by transforming AH109 yeast cells with the pGBKT7 vector or the pGBKT7-Rab7 construct (as a negative controls) or the pGBKT7-RILP wild-type construct, and with the PGADT7-V1G1 construct. Clones were then assayed for growth on selective medium and for β-galactosidase activity using o-nitrophenyl-β-d-galactoside as a substrate (Bartel et al., 1993).

Transfection was performed using Metafectene Pro or Metafectene Easy from Biontex (Martinsried, Germany) following the manufacturer's instructions. After 20 h of transfection, cells were processed for immunofluorescence or biochemical assays. For RNA interference, small-interfering (si)RNAs were purchased from MWG-Biotech (Ebersberg, Germany) or Sigma-Aldrich. We used the following oligonucleotides: siRNA-RILP2 sense, 5′-GAUCAAGGCCAAGAUGUUATT-3′ and antisense, 5′-UAACAUCUUGGCCUUGAUCTT-3′; control RNA sense, 5′-ACUUCGAGCGUGCAUGGCUTT-3′ and antisense, 5′-AGCCAUGCACGCUCGAAGUTT-3′; siRNA-V1G1 sense, 5′-AGAAGAAGCUCAGGCUGAATT-3′ and antisense, 5′-UUCTGCCTGAGCUUCUUCUTT-3′; siRNA-Rab7a sense, 5′-GGAUGACCUCUAGGAAGAATT-3′ and antisense, 5′-UUCUUCCUAGAGGUCAUCCTT-3′. RILP and Rab7 siRNAs were efficient in silencing as reported previously (De Luca et al., 2008; Progida et al., 2007; Spinosa et al., 2008). Briefly, HeLa cells were plated 1 day before transfection in tissue culture dishes (6-cm diameter). Cells were transfected with siRNAs using Oligofectamine (Invitrogen) for 72 h, re-plated and left for 48 h before performing further experiments.

Rabbit anti-RILP polyclonal antibodies have been described previously (Cantalupo et al., 2001). Goat polyclonal anti-RILP (1∶100, sc-82746), rabbit polyclonal anti-RILP (1∶100, sc-98331), mouse monoclonal 9E10 anti-Myc (1∶500, sc-40), rabbit polyclonal and mouse monoclonal anti-HA (1∶500, sc-805 and sc-7392, respectively), mouse monoclonal anti-V1G1 (1∶100, sc-25333), mouse monoclonal anti-V0D1 (1∶100, sc-81887), goat polyclonal anti-cathepsin-D (1∶500, sc-6486) and rabbit polyclonal anti-GAPDH (1∶10000, sc-25778) antibodies were from Santa Cruz Biotechnology. Rabbit polyclonal anti-V1C1 (1∶1000, 361–375), rabbit polyclonal anti-Rab7 (1∶1000, R4779), mouse monoclonal anti-Rab7 (1∶1000, R8779) and mouse monoclonal anti-tubulin (1∶3000, clone B512) were from Sigma-Aldrich. Chicken polyclonal anti-V1G1 (1∶1000, ab15853) and rabbit polyclonal anti-giantin (1∶1000, ab24586) were from Abcam (Cambridge, UK). Goat polyclonal anti-GST (1∶2000, 27-4577-01) was from GE Healthcare (Buckinghamshire, UK) and rabbit polyclonal anti-ubiquitin (1∶500, Z0458) was from DAKO (Glostrup, Denmark). Mouse monoclonal anti-p150 (1∶500, 610474) was from BD Bioscience. Secondary antibodies conjugated to fluorochromes or horseradish peroxidase (HRP) were from Invitrogen or SantaCruz Biotechnology.

For analysis of cytosolic and membrane distribution, pellets of HeLa cells were resuspended in homogenization buffer (8% sucrose in 3 mM imidazole). Cells were passed through a 26-gauge needle and centrifuged at 3800 g for 5 min at 4°C. The post-nuclear supernatant was centrifuged at 50,000 g for 2 h at 4°C and fractionated into the high-speed pellet (membrane) and supernantant (cytosol).

For immunoprecipitation, we used anti-HA affinity gel (Ezview Red Anti-HA E6779, Sigma) according to the manufacturer's instructions and as described previously (Cogli et al., 2013b). Co-immunoprecipitation of endogenous proteins in HeLa cells was performed using a crosslink immunprecipitation kit (Pierce, Rockford, IL), following the manufacturer's instructions and as described previously (Cogli et al., 2013a).

For direct interaction experiments, GST, GST-tagged and His-tagged proteins were expressed in bacteria and affinity purified as described previously (Chiariello et al., 1999). Glutathione resin alone or bound to purified GST, GST–RILPC33 or GST–RILP was incubated with purified His-tagged V1G1 in PBS with 2 mM MgCl₂ and 0.8 mM GTP for 1 h on a rotating wheel. Subsequently, samples were subjected to GST pull-down using the glutathione resin. Samples were then subjected to SDS-PAGE and western blotting. Pull-down experiments were as described previously (Cogli et al., 2013b). Briefly, purified His–V1G1 was bound to Ni–NTA resin and incubated with lysates of HeLa cells transfected with FLAG–p150^Glued in the presence or absence of purified GST–RILP for 2 h at 4°C. In addition, as a control, purified GST–RILP was bound to glutathione resin and incubated with lysates of HeLa cells transfected with FLAG–p150^Glued for 2 h at 4°C. After incubation, samples were processed for SDS-PAGE and western blot analysis.

HeLa cells were lysed with RIPA buffer (R0278, Sigma-Aldrich) plus protease inhibitor cocktail (Roche, Mannheim Germany). For ubiquitylation experiments, RIPA buffer was supplemented with 20 mM Na₄P₂O₇ pH 7.5, protease inhibitor cocktail (Roche), 50 mM NaF, 2 mM PMSF, 10 mM Na₃VO₄ in HEPES pH 7.5 and 5 µM N-ethylmaleimide. Lysates were loaded onto SDS-PAGE gels, and separated proteins were transferred onto polyvinylidene fluoride (PVDF) membrane from Millipore (Billerica, MA). The membrane was blocked in 5% milk in PBS for 30 min at room temperature, incubated with the appropriate antibody and then with a secondary antibody conjugated to HRP (diluted 1∶5000).

For incubation with the antibody against ubiquitin, after transferring, the membrane was subjected to a treatment in denaturing solution (6 M guanidium chloride, 20 mM Tris-HCl pH 7.4, 1 mM PMSF and 5 mM β-mercaptoethanol) for 30 min at 4°C, to make proteins more accessible to the antibody. After extensive washing in TBS-T buffer (25 mM Tris-HCl pH 8.0, 150 mM NaCl and 0.05% Tween) the membrane was blocked in 5% bovine serum albumin (BSA) in TBS (25 mM Tris-HCl pH 8, 150 mM NaCl) for 30 min at room temperature, incubated with the anti-ubiquitin antibody and then with a secondary antibody conjugated to HRP (diluted 1∶5000). Bands were visualized by using western blot Luminol Reagent (Santa Cruz) or SuperSignal West Pico (Pierce).

Cells grown on 11-mm round glass coverslips were permeabilized, fixed and incubated with antibodies as described previously (Bucci et al., 1992). Cells were viewed with a Zeiss LSM 510 confocal microscope. In some experiments, cells were incubated with LysoTracker Red or LysoSensor DND-192 (Invitrogen) and Alexa-Fluor-647- or Alexa-Fluor-555-labelled transferrin (Invitrogen) for 30 min. Cells were imaged by live microscopy, and the relative level of transferrin colocalizing with LysoTracker or LysoSensor was quantified.

Total RNA was extracted from HeLa cells by using the RNeasy mini kit according to the manufacturer's instructions (Qiagen, Hilden, Germany). The RNA retrotranscription protocol was performed as described previously (Progida et al., 2010). Quantitative real-time PCR was performed using SYBR Green JumpStart ReadyMix (Sigma) in the Smart Cycler II Real-Time PCR detection system (Cepheid, Sunnyvale, CA). The primers used were: GAPDH forward, 5′-GGTGGTCTCCTCTGACTTCAACA-3′; GAPDH reverse, 5′-GTTGCTGTAGCCAAATTCGTTGT-3′; V1G1 forward, 5′-GCCGAGAAGGTGTCCGAGGCCCG-3′; V1G1 reverse, 5′-GCGGTACTGTTCAATTTCAGCC-3′; RILP forward, 5′-CGGAAGCAGCGGAAGAAGATCAAG-3′; RILP reverse, 5′-GAGCAGGATCCATGGGCCAGC-3′. Primers were purchased from Eurofin MWG Operon (Ebersberg, Germany). The PCR program was as follows: 3 min at 94°C; 35 cycles of 30 s at 94°C, 30 s at 60°C and 30 s at 72°C; and 6 min at 75°C. The specificity of PCR products was checked by performing a melting-curve test.

We thank Anna Maria Rosaria Colucci (Novartis Vaccine Institute for Global Health, Siena, Italy) for help in the first two-hybrid screening and Mitsunori Fukuda (Tohoku University, Sendai, Japan) for kindly donating pEF-FLAG-p150^Glued plasmid.